1 2018-10-11 Doug Rupp <rupp@adacore.com>
2 Olivier Hainque <hainque@adacore.com>
4 * config/rs6000/vxworks.h (VXWORKS_RELAX_LINK_SPEC): New macro.
5 Pass --relax to the linker for RTPs.
6 (LINK_SPEC): Append VXWORKS_RELAX_LINK_SPEC.
8 2018-10-11 Andrew Stubbs <ams@codesourcery.com>
9 Jan Hubicka <jh@suse.cz>
10 Martin Jambor <mjambor@suse.cz>
12 * print-rtl.c (print_rtx_operand_codes_E_and_V): Print how many times
13 the same elements are repeated rather than printing all of them.
14 * read-rtl.c (rtx_reader::read_rtx_operand): Recognize and expand
16 * read-rtl-function.c (test_loading_repeat): New function.
17 (read_rtl_function_c_tests): Call test_loading_repeat.
18 * rtl-tests.c (test_dumping_repeat): New function.
19 (rtl_tests_c_tests): Call test_dumping_repeat.
21 2018-10-11 Richard Biener <rguenther@suse.de>
23 * config/i386/x86-tune-costs.h (bdver?_memcpy, bdver?_memset,
24 bdver?_cost): Unify to ...
25 (bdver_memcpy, bdver_memset, bdver_cost): ... this.
26 * config/i386/i386.c (processor_cost_table): Adjust.
28 2018-10-10 Eric Botcazou <ebotcazou@adacore.com>
31 * cgraphunit.c (cgraph_node::expand_thunk): Force DECL_IGNORED_P on
32 the thunk when expanding to GIMPLE.
34 2018-10-10 Bernd Edlinger <bernd.edlinger@hotmail.de>
36 * varasm.c (mergeable_string_section): Don't try to move zero-length
37 strings to the merge section.
39 2018-10-10 Uros Bizjak <ubizjak@gmail.com>
42 * config/i386/mmx.md (const_vector 0 -> mem splitter): New splitter.
44 2018-10-10 Jakub Jelinek <jakub@redhat.com>
47 * config/i386/i386-builtin.def (IX86_BUILTIN_RDPMC): Move from args set
50 2018-10-10 Richard Biener <rguenther@suse.de>
52 * config/i386/sse.md (reduc_plus_scal_v8df, reduc_plus_scal_v4df,
53 reduc_plus_scal_v2df, reduc_plus_scal_v16sf, reduc_plus_scal_v8sf,
54 reduc_plus_scal_v4sf): Merge into pattern reducing to half width
55 and recursing and pattern terminating the recursion on SSE
56 vector width using ix86_expand_reduc.
57 (reduc_sminmax_scal_<mode>): Split into part reducing to half
58 width and recursing and SSE2 vector variant doing the final
59 reduction with ix86_expand_reduc.
60 (reduc_uminmax_scal_<mode>): Likewise for the AVX512 variants
61 with terminating the recursion at AVX level, splitting that
64 2018-10-09 David Malcolm <dmalcolm@redhat.com>
66 * genmatch.c (error_cb): Rename to...
67 (diagnostic_cb): ...this, converting int params to enums.
68 (fatal_at): Update for renaming.
69 (warning_at): Likewise.
71 * input.c (selftest::ebcdic_execution_charset::apply):
72 Update for renaming of...
73 (selftest::ebcdic_execution_charset::on_error): ...this, renaming
75 (selftest::ebcdic_execution_charset::on_diagnostic): ...this,
76 converting level and reason to enums.
77 (class selftest::lexer_error_sink): Rename to...
78 (class selftest::lexer_test_options): ...this, renaming field
79 "m_errors" to "m_diagnostics".
80 (selftest::lexer_test_options::apply): Update for renaming of...
81 (selftest::lexer_test_options::on_error): ...this, renaming to...
82 (selftest::lexer_test_options::on_diagnostic): ...this
83 converting level and reason to enums.
84 (selftest::test_lexer_string_locations_raw_string_unterminated):
86 * opth-gen.awk (struct cpp_reason_option_codes_t): Use enum for
89 2018-10-09 Paul A. Clarke <pc@us.ibm.com>
91 * config.gcc (powerpc*-*-*): Add pmmintrin.h to extra_headers.
92 * config/rs6000/pmmintrin.h: New file.
94 2018-10-09 Eric Botcazou <ebotcazou@adacore.com>
96 PR tree-optimization/86659
97 * gimple-match.h (gimple_match_op constructors): Initialize reverse.
99 2018-10-09 Richard Biener <rguenther@suse.de>
101 PR tree-optimization/63155
102 * tree-ssa-structalias.c: Include tree-ssa.h.
103 (get_constraint_for_ssa_var): For undefs return nothing_id.
104 (find_func_aliases): Cleanup PHI handling.
106 2018-10-09 Richard Biener <rguenther@suse.de>
108 * tree-vectorizer.c (dump_stmt_cost): Fix cut&paste missing
111 2018-10-09 Martin Liska <mliska@suse.cz>
113 * asan.c (asan_emit_stack_protection): If a stack variable
114 is located in a same file as current function, then emit
115 line info into variable definition string.
117 2018-10-08 Eric Botcazou <ebotcazou@adacore.com>
119 * print-rtl.c (rtx_writer::print_rtx_operand_code_i): Print column
122 2018-10-08 Eric Botcazou <ebotcazou@adacore.com>
124 * cgraphunit.c (cgraph_node::expand_thunk): Do not force DECL_IGNORED_P
127 2018-10-08 Eric Botcazou <ebotcazou@adacore.com>
129 PR tree-optimization/86659
130 * gimple-match.h (struct gimple_match_op): Add reverse field.
131 (gimple_match_op::set_op): New overloaded method.
132 * gimple-match-head.c (maybe_build_generic_op) <BIT_FIELD_REF>: Set
133 the REF_REVERSE_STORAGE_ORDER flag on the value.
134 (gimple_simplify) <GIMPLE_ASSIGN>: For BIT_FIELD_REF, propagate the
135 REF_REVERSE_STORAGE_ORDER flag and avoid simplifying if it is set.
137 2018-10-08 Richard Sandiford <richard.sandiford@arm.com>
140 * gimple-ssa-backprop.c (backprop::intersect_uses): Use
141 FOR_EACH_IMM_USE_FAST instead of FOR_EACH_IMM_USE_STMT.
143 2018-10-08 H.J. Lu <hongjiu.lu@intel.com>
146 * config/i386/avx512fintrin.h (_mm512_mask_fmaddsub_round_pd):
147 Defined with __builtin_ia32_vfmaddsubpd512_mask.
149 2018-10-08 Richard Biener <rguenther@suse.de>
151 * config/i386/x86-tune-costs.h (znver1_cost): Make AVX256 vector loads
152 cost the same as AVX128 ones.
154 2018-10-08 Paul Koning <ni1d@arrl.net>
156 * config/pdp11/pdp11-protos.h (output_block_move): Remove.
157 (expand_block_move): New function.
158 * config/pdp11/pdp11.c (output_block_move): Remove.
159 (expand_block_move): New function.
160 * config/pdp11/pdp11.h (MOVE_RATIO): New definition.
161 * config/pdp11/pdp11.md (movmemhi): Use expand_block_move.
162 (*movmemhi1): Remove.
164 2018-10-08 Robin Dapp <rdapp@linux.ibm.com>
166 * config/s390/2827.md: Increase latencies for some FP instructions.
168 2018-10-08 Richard Biener <rguenther@suse.de>
170 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
172 * tree-vectorizer.c (dump_stmt_cost): Add cost param and dump it.
173 * tree-vectorizer.h (dump_stmt_cost): Adjust.
174 (add_stmt_cost): Dump return value of the hook.
176 2018-10-08 Richard Biener <rguenther@suse.de>
178 PR tree-optimization/63155
179 * tree-ssa-propagate.c (add_ssa_edge): Do cheap check first.
180 (ssa_propagation_engine::ssa_propagate): Remove redundant
183 2018-10-05 Peter Bergner <bergner@linux.ibm.com>
185 PR rtl-optimization/86939
186 PR rtl-optimization/87479
187 * ira.h (non_conflicting_reg_copy_p): New prototype.
188 * ira-lives.c (ignore_reg_for_conflicts): New static variable.
189 (make_hard_regno_dead): Don't add conflicts for register
190 ignore_reg_for_conflicts.
191 (make_object_dead): Likewise.
192 (non_conflicting_reg_copy_p): New function.
193 (process_bb_node_lives): Set ignore_reg_for_conflicts for copies.
194 Remove special conflict handling of REAL_PIC_OFFSET_TABLE_REGNUM.
195 * lra-lives.c (ignore_reg_for_conflicts): New static variable.
196 (make_hard_regno_dead): Don't add conflicts for register
197 ignore_reg_for_conflicts. Remove special conflict handling of
198 REAL_PIC_OFFSET_TABLE_REGNUM. Remove now unused argument
199 check_pic_pseudo_p and update callers.
200 (mark_pseudo_dead): Don't add conflicts for register
201 ignore_reg_for_conflicts.
202 (process_bb_lives): Set ignore_reg_for_conflicts for copies.
204 2018-10-05 Andrew Waterman <andrew@sifive.com>
205 Jim Wilson <jimw@sifive.com>
207 * config/riscv/riscv.md (f<quiet_pattern>_quiet<ANYF:mode><X:mode>4):
208 Add define_expand. Add ! HONOR_SNANS check to current pattern. Add
209 new pattern using HONOR_SNANS that emits one extra instruction.
211 2018-10-05 Segher Boessenkool <segher@kernel.crashing.org>
213 * config/rs6000/rs6000.md (unnamed mfcr scc_comparison_operator
214 patterns): Merge SI and DI patterns to a GPR pattern.
215 (unnamed define_insn and define_split for record form of that): Merge
216 to a single define_insn_and_split pattern.
218 2018-10-05 David Malcolm <dmalcolm@redhat.com>
221 * input.c (expand_location_to_spelling_point): Add param "aspect"
222 and use rather than hardcoding LOCATION_ASPECT_CARET.
223 (get_substring_ranges_for_loc): Handle the case of a single token
224 within a macro expansion.
225 * input.h (expand_location_to_spelling_point): Add "aspect" param,
226 defaulting to LOCATION_ASPECT_CARET.
228 2018-10-05 Paul Koning <ni1d@arrl.net>
230 * config/pdp11/pdp11.c (TARGET_CXX_GUARD_TYPE): Define.
231 (TARGET_CXX_CLASS_DATA_ALWAYS_COMDAT): Define.
232 (TARGET_CXX_LIBRARY_RTTI_COMDAT): Define.
233 (pdp11_guard_type): New function.
235 2018-10-05 Paul Koning <ni1d@arrl.net>
237 * config/pdp11/pdp11.h (FLOAT_TYPE_SIZE): Always 32.
238 * config/pdp11/pdp11.opt (mfloat32): Remove.
240 * doc/invoke.texi (pdp11 -mfloat32): Remove:
241 (pdp11 -mfloat64): Remove.
243 2018-10-05 Uros Bizjak <ubizjak@gmail.com>
245 * config/i386/i386.md (*cmpxf_cc_i387): Remove pattern.
246 (*cmp<mode>_cc_i387): Ditto.
247 (*cmpu<mode>_cc_i387): Ditto.
248 (*cmp<X87MODEF:mode>_<SWI24:mode>_cc_i387): Ditto.
249 * config/i386/i386.c (ix86_expand_fp_compare): Remove
251 <case IX86_FPCMP_SAHF>: Do not generate pattern with HImode clobber.
252 Emit x86_sahf_1 pattern.
253 (ix86_expand_compare): Update call to ix86_expand_fp_compare.
254 (ix86_expand_carry_flag_compare): Ditto.
256 2018-10-05 Uros Bizjak <ubizjak@gmail.com>
258 * config/i386/i386.md (*cmpxf_i387): Change operand 2 predicate
259 to reg_or_0_operand. Add "C" constraint.
260 (*cmpxf_cc_i387): Ditto.
261 (*cmp<mode>_i387): Change operand 2 predicate
262 to nonimm_or_0_operand. Add "C" constraint.
263 (*cmp<mode>_cc_i387): Ditto.
264 (*cmp<mode>_0_i387): Remove insn pattern.
265 (*cmp<mode>_0_cc_i387): Ditto.
267 2018-10-05 Uros Bizjak <ubizjak@gmail.com>
269 * config/i386/constraints.md ("C"): Do not depend on TARGET_SSE.
270 * config/i386/predicates.md (nonimm_or_0_operand): Rename
271 from vector_move_operand. Update all uses.
273 2018-10-05 Martin Sebor <msebor@redhat.com>
275 PR tree-optimization/87490
276 * builtins.c (expand_builtin_strnlen): Handle a null data.decl
279 2018-10-05 Richard Biener <rguenther@suse.de>
281 PR tree-optimization/63155
282 * tree-ssa-ccp.c (ccp_propagate::visit_phi): Avoid excess
283 vertical space in dumpfiles.
284 * tree-ssa-propagate.h
285 (ssa_propagation_engine::process_ssa_edge_worklist): Remove.
286 * tree-ssa-propagate.c (cfg_blocks_back): New global.
287 (ssa_edge_worklist_back): Likewise.
288 (curr_order): Likewise.
289 (cfg_blocks_get): Remove abstraction.
290 (cfg_blocks_add): Likewise.
291 (cfg_blocks_empty_p): Likewise.
292 (add_ssa_edge): Add to current or next worklist based on
294 (add_control_edge): Likewise.
295 (ssa_propagation_engine::process_ssa_edge_worklist): Fold
297 (ssa_propagation_engine::ssa_propagate): ... here. Unify
298 iteration from CFG and SSA edge worklist so we process
299 everything in RPO order, prioritizing forward progress
301 (ssa_prop_init): Allocate new worklists, do not dump
303 (ssa_prop_fini): Free new worklists.
305 2018-10-05 Richard Biener <rguenther@suse.de>
307 * tree-core.h (tree_block::abstract_flag): Remove.
308 (tree_block::block_num): Make full 32bits.
309 * tree.def (BLOCK): Remove docs about BLOCK_ABSTRACT.
310 * tree.h (BLOCK_ABSTRACT): Remove.
311 * dwarf2out.c (gen_lexical_block_die): Remove dead code
312 resulting from BLOCK_ABSTRACT being always false.
313 (gen_inlined_subroutine_die): Likewise.
314 (gen_block_die): Likewise.
315 * tree.c (block_ultimate_origin): Likewise.
316 * tree-pretty-print.c (dump_block_node): Remove code dealing
318 * tree-ssa-live.c (dump_scope_block): Likewise.
319 * tree-streamer-in.c (unpack_ts_block_value_fields): Likewise.
320 * tree-streamer-out.c (pack_ts_block_value_fields): Likewise.
322 2018-10-05 Richard Biener <rguenther@suse.de>
324 * config/i386/i386.c (ix86_add_stmt_cost): When scalar cost
325 is asked for initialize mode to the component mode of the
328 2018-10-05 H.J. Lu <hongjiu.lu@intel.com>
331 * config/i386/gnu-user.h (ASM_SPEC): Don't pass -msse2avx to
333 * config/i386/gnu-user64.h (ASM_SPEC): Likewise.
335 2018-10-05 Segher Boessenkool <segher@kernel.crashing.org>
338 * config/rs6000/rs6000-builtin.def (RS6000_BUILTIN_SET_FPSCR_DRN): Use
340 * config/rs6000/rs6000.md (rs6000_set_fpscr_rn): Require the operand
341 to be DImode. When using mffscrn, force the operand to a register.
343 2018-10-04 Uros Bizjak <ubizjak@gmail.com>
345 * config/i386/i386.md (*fop_<X87MODEF:mode>_2_i387): Macroize insn
346 from *fop_<MODEF:mode>_2_i387 and *fop_xf_2_i387 using
347 X87MODEF mode iterator.
348 (*fop_<X87MODEF:mode>_3_i387): Macroize insn from
349 *fop_<MODEF:mode>_3_i387 and *fop_xf_3_i387 using
350 X87MODEF mode iterator.
352 2018-10-04 Vinay Kumar <vinay.kumar@blackfigtech.com>
354 * doc/invoke.texi (-Wno-prio-ctor-dtor): Document new warning
357 2018-10-04 David Malcolm <dmalcolm@redhat.com>
359 * Makefile.in (OBJS): Add opt-problem.o.
360 * dump-context.h: Include "selftest.h.
361 (selftest::temp_dump_context): New forward decl.
362 (class dump_context): Make friend of class
363 selftest::temp_dump_context.
364 (dump_context::dump_loc_immediate): New decl.
365 (class dump_pretty_printer): Move here from dumpfile.c.
366 (class temp_dump_context): Move to namespace selftest.
367 (temp_dump_context::temp_dump_context): Add param
368 "forcibly_enable_dumping".
369 (selftest::verify_dumped_text):
370 (ASSERT_DUMPED_TEXT_EQ): Move here from dumpfile.c.
371 (selftest::verify_item):
372 (ASSERT_IS_TEXT): Move here from dumpfile.c.
373 (ASSERT_IS_TREE): Likewise.
374 (ASSERT_IS_GIMPLE): Likewise.
375 * dumpfile.c (dump_context::dump_loc): Move immediate dumping
377 (dump_context::dump_loc_immediate): ...this new function.
378 (class dump_pretty_printer): Move to dump-context.h.
379 (dump_switch_p_1): Don't enable MSG_PRIORITY_REEMITTED.
380 (opt_info_switch_p_1): Enable MSG_PRIORITY_REEMITTED.
381 (temp_dump_context::temp_dump_context): Move to "selftest"
382 namespace. Add param "forcibly_enable_dumping", and use it to
383 conditionalize the use of m_pp;
384 (selftest::verify_dumped_text): Make non-static.
385 (ASSERT_DUMPED_TEXT_EQ): Move to dump-context.h.
386 (selftest::verify_item): Make non-static.
387 (ASSERT_IS_TEXT): Move to dump-context.h.
388 (ASSERT_IS_TREE): Likewise.
389 (ASSERT_IS_GIMPLE): Likewise.
390 (selftest::test_capture_of_dump_calls): Pass "true" for new
391 param of temp_dump_context.
392 * dumpfile.h (enum dump_flag): Add MSG_PRIORITY_REEMITTED, adding
393 it to MSG_ALL_PRIORITIES. Update values of TDF_COMPARE_DEBUG and
395 * opt-problem.cc: New file.
396 * opt-problem.h: New file.
397 * optinfo-emit-json.cc
398 (selftest::test_building_json_from_dump_calls): Pass "true" for
399 new param of temp_dump_context.
400 * optinfo.cc (optinfo_kind_to_dump_flag): New function.
401 (optinfo::emit_for_opt_problem): New function.
402 (optinfo::emit): Clarity which emit_item is used.
403 * optinfo.h (optinfo::get_dump_location): New accessor.
404 (optinfo::emit_for_opt_problem): New decl.
405 (optinfo::emit): Make const.
406 * selftest-run-tests.c (selftest::run_tests): Call
407 selftest::opt_problem_cc_tests.
408 * selftest.h (selftest::opt_problem_cc_tests): New decl.
409 * tree-data-ref.c (dr_analyze_innermost): Convert return type from
410 bool to opt_result, converting fprintf messages to
411 opt_result::failure_at calls. Add "stmt" param for use by the
413 (create_data_ref): Pass "stmt" to the dr_analyze_innermost call.
414 (runtime_alias_check_p): Convert return type from bool to
415 opt_result, converting dump_printf calls to
416 opt_result::failure_at, using the statement DDR_A for their
418 (find_data_references_in_stmt): Convert return type from bool to
419 opt_result, converting "return false" to opt_result::failure_at
421 * tree-data-ref.h: Include "opt-problem.h".
422 (dr_analyze_innermost): Convert return type from bool to opt_result,
423 and add a const gimple * param.
424 (find_data_references_in_stmt): Convert return type from bool to
426 (runtime_alias_check_p): Likewise.
427 * tree-predcom.c (find_looparound_phi): Pass "init_stmt" to
428 dr_analyze_innermost.
429 * tree-vect-data-refs.c (vect_mark_for_runtime_alias_test):
430 Convert return type from bool to opt_result, adding a message for
431 the PARAM_VECT_MAX_VERSION_FOR_ALIAS_CHECKS zero case.
432 (vect_analyze_data_ref_dependence): Convert return type from bool
433 to opt_result. Change sense of return type from "false"
434 effectively meaning "no problems" to "false" meaning a problem,
435 so that "return false" becomes "return opt_result::success".
436 Convert "return true" calls to opt_result::failure_at, using
437 the location of statement A rather than vect_location.
438 (vect_analyze_data_ref_dependences): Convert return type from bool
440 (verify_data_ref_alignment): Likewise, converting dump_printf_loc
441 calls to opt_result::failure_at, using the stmt location rather
443 (vect_verify_datarefs_alignment): Convert return type from bool
445 (vect_enhance_data_refs_alignment): Likewise. Split local "stat"
446 into multiple more-tightly-scoped copies.
447 (vect_analyze_data_refs_alignment): Convert return type from bool
449 (vect_analyze_data_ref_accesses): Likewise, converting a
450 "return false" to a "return opt_result::failure_at", adding a
452 (vect_prune_runtime_alias_test_list): Convert return type from
453 bool to opt_result, converting dump_printf_loc to
454 opt_result::failure_at. Add a %G to show the pertinent statement,
455 and use the stmt's location rather than vect_location.
456 (vect_find_stmt_data_reference): Convert return type from
457 bool to opt_result, converting dump_printf_loc to
458 opt_result::failure_at, using stmt's location.
459 (vect_analyze_data_refs): Convert return type from bool to
460 opt_result. Convert "return false" to "return
461 opt_result::failure_at", adding messages as needed.
462 * tree-vect-loop.c (vect_determine_vf_for_stmt_1): Convert return
463 type from bool to opt_result.
464 (vect_determine_vf_for_stmt): Likewise.
465 (vect_determine_vectorization_factor): Likewise, converting
466 dump_printf_loc to opt_result::failure_at, using location of phi
467 rather than vect_location.
468 (vect_analyze_loop_form_1): Convert return type from bool to
469 opt_result, converting dump_printf_loc calls, retaining the use of
471 (vect_analyze_loop_form): Convert return type from loop_vec_info
472 to opt_loop_vec_info.
473 (vect_analyze_loop_operations): Convert return type from bool to
474 opt_result, converting dump_printf_loc calls, using the location
475 of phi/stmt rather than vect_location where available. Convert
476 various "return false" to "return opt_result::failure_at" with
477 "unsupported phi" messages.
478 (vect_get_datarefs_in_loop): Convert return type from bool to
479 opt_result. Add a message for the
480 PARAM_LOOP_MAX_DATAREFS_FOR_DATADEPS failure.
481 (vect_analyze_loop_2): Convert return type from bool to
482 opt_result. Ensure "ok" is set to a opt_result::failure_at before
483 each "goto again;", adding new messages where needed.
484 Add "unsupported grouped {store|load}" messages.
485 (vect_analyze_loop): Convert return type from loop_vec_info to
487 * tree-vect-slp.c (vect_analyze_slp): Convert return type from
489 * tree-vect-stmts.c (process_use): Likewise, converting
490 dump_printf_loc call and using stmt location, rather than
492 (vect_mark_stmts_to_be_vectorized): Likeise.
493 (vect_analyze_stmt): Likewise, adding a %G.
494 (vect_get_vector_types_for_stmt): Convert return type from bool to
495 opt_result, converting dump_printf_loc calls and using stmt
496 location, rather than vect_location.
497 (vect_get_mask_type_for_stmt): Convert return type from tree to
498 opt_tree, converting dump_printf_loc calls and using stmt location.
499 * tree-vectorizer.c: Include "opt-problem.h.
500 (try_vectorize_loop_1): Flag "Analyzing loop at" dump message as
501 MSG_PRIORITY_INTERNALS. Convert local "loop_vinfo" from
502 loop_vec_info to opt_loop_vec_info. If if fails, and dumping is
503 enabled, use it to report at the top level "couldn't vectorize
504 loop" followed by the problem.
505 * tree-vectorizer.h (opt_loop_vec_info): New typedef.
506 (vect_mark_stmts_to_be_vectorized): Convert return type from bool
508 (vect_analyze_stmt): Likewise.
509 (vect_get_vector_types_for_stmt): Likewise.
510 (tree vect_get_mask_type_for_stmt): Likewise.
511 (vect_analyze_data_ref_dependences): Likewise.
512 (vect_enhance_data_refs_alignment): Likewise.
513 (vect_analyze_data_refs_alignment): Likewise.
514 (vect_verify_datarefs_alignment): Likewise.
515 (vect_analyze_data_ref_accesses): Likewise.
516 (vect_prune_runtime_alias_test_list): Likewise.
517 (vect_find_stmt_data_reference): Likewise.
518 (vect_analyze_data_refs): Likewise.
519 (vect_analyze_loop): Convert return type from loop_vec_info to
521 (vect_analyze_loop_form): Likewise.
522 (vect_analyze_slp): Convert return type from bool to opt_result.
524 2018-10-04 David Malcolm <dmalcolm@redhat.com>
526 * doc/invoke.texi (-fopt-info): Document new "internals"
528 * dump-context.h (dump_context::apply_dump_filter_p): New decl.
529 * dumpfile.c (dump_options): Update for renaming of MSG_ALL to
531 (optinfo_verbosity_options): Add "internals".
532 (kind_as_string): Update for renaming of MSG_ALL to MSG_ALL_KINDS.
533 (dump_context::apply_dump_filter_p): New member function.
534 (dump_context::dump_loc): Use apply_dump_filter_p rather than
535 explicitly masking the dump_kind.
536 (dump_context::begin_scope): Increment the scope depth first. Use
537 apply_dump_filter_p rather than explicitly masking the dump_kind.
538 (dump_context::emit_item): Use apply_dump_filter_p rather than
539 explicitly masking the dump_kind.
540 (dump_dec): Likewise.
541 (dump_hex): Likewise.
542 (dump_switch_p_1): Default to MSG_ALL_PRIORITIES.
543 (opt_info_switch_p_1): Default to MSG_PRIORITY_USER_FACING.
544 (opt_info_switch_p): Update handling of default
545 MSG_OPTIMIZED_LOCATIONS to cope with default of
546 MSG_PRIORITY_USER_FACING.
547 (dump_basic_block): Use apply_dump_filter_p rather than explicitly
548 masking the dump_kind.
549 (selftest::test_capture_of_dump_calls): Update test_dump_context
550 instances to use MSG_ALL_KINDS | MSG_PRIORITY_USER_FACING rather
551 than MSG_ALL. Generalize scope test to be run at all four
552 combinations of with/without MSG_PRIORITY_USER_FACING and
553 MSG_PRIORITY_INTERNALS, adding examples of explicit priority
554 for each of the two values.
555 * dumpfile.h (enum dump_flag): Add comment about the MSG_* flags.
556 Rename MSG_ALL to MSG_ALL_KINDS. Add MSG_PRIORITY_USER_FACING,
557 MSG_PRIORITY_INTERNALS, and MSG_ALL_PRIORITIES, updating the
558 values for TDF_COMPARE_DEBUG and TDF_ALL_VALUES.
559 (AUTO_DUMP_SCOPE): Add a note to the comment about the interaction
561 * tree-vect-loop-manip.c (vect_loop_versioning): Mark versioning
562 dump messages as MSG_PRIORITY_USER_FACING.
563 * tree-vectorizer.h (DUMP_VECT_SCOPE): Add a note to the comment
564 about the interaction with MSG_PRIORITY_*.
566 2018-10-04 Bernd Edlinger <bernd.edlinger@hotmail.de>
568 * varasm.c (output_constant): Add new parameter merge_strings.
569 Make strings properly zero terminated in merge string sections.
570 (mergeable_string_section): Don't fail if the last char is non-zero.
571 (assemble_variable_contents): Handle merge string sections.
572 (assemble_variable): Likewise.
573 (assemble_constant_contents): Likewise.
574 (output_constant_def_contents): Likewise.
575 (output_constructor_array_range,
576 output_constructor_regular_field): Adjust call to output_constant.
577 (output_object_block): Adjust call to assemble_constant_contents
578 and assemble_variable_contents.
580 2018-10-04 Martin Liska <mliska@suse.cz>
583 * cgraphunit.c (process_function_and_variable_attributes):
584 Warn about a function with alias attribute and a body.
586 2018-10-04 Martin Liska <mliska@suse.cz>
589 * multiple_target.c (redirect_to_specific_clone): New function.
590 (ipa_target_clone): Use it.
591 * tree-inline.c: Fix comment.
593 2018-10-04 David Malcolm <dmalcolm@redhat.com>
595 * dumpfile.c (gcc::dump_manager::dump_manager): Initialize new
597 (gcc::dump_manager::~dump_manager): Free m_optinfo_filename.
598 (gcc::dump_manager::register_pass): New member function, adapted
599 from loop body in gcc::pass_manager::register_pass, adding a
600 call to update_dfi_for_opt_info.
601 (gcc::dump_manager::opt_info_enable_passes): Store the
602 -fopt-info options into the new fields. Move the loop
604 (gcc::dump_manager::update_dfi_for_opt_info): ...this new member
606 * dumpfile.h (struct opt_pass): New forward decl.
607 (gcc::dump_manager::register_pass): New decl.
608 (gcc::dump_manager::update_dfi_for_opt_info): New decl.
609 (class gcc::dump_manager): Add fields "m_optgroup_flags",
610 "m_optinfo_flags", and "m_optinfo_filename".
611 * passes.c (gcc::pass_manager::register_pass): Move all of the
612 dump-handling code to gcc::dump_manager::register_pass.
614 2018-10-04 Peter Bergner <bergner@linux.ibm.com>
616 PR rtl-optimization/87466
617 * target.def (setjmp_preserves_nonvolatile_regs_p): New target hook.
618 * doc/tm.texi.in (TARGET_SETJMP_PRESERVES_NONVOLATILE_REGS_P): New hook.
619 * doc/tm.texi: Regenerate.
620 * ira-lives.c (process_bb_node_lives): Use the new target hook.
621 * lra-lives.c (process_bb_lives): Likewise.
622 * config/rs6000/rs6000.c (TARGET_SETJMP_PRESERVES_NONVOLATILE_REGS_P):
625 2018-10-04 Tamar Christina <tamar.christina@arm.com>
627 * params.c (add_params): Fix initialization.
629 2018-10-04 Martin Liska <mliska@suse.cz>
631 PR gcov-profile/84107
632 * tree-profile.c (init_ic_make_global_vars):
633 Remove ic_void_ptr_var and ic_gcov_type_ptr_var.
634 Come up with new ic_tuple* variables. Emit
635 __gcov_indirect_call{,_topn} variables.
636 (gimple_gen_ic_profiler): Access the variable
638 (gimple_gen_ic_func_profiler): Access
639 __gcov_indirect_call.callee field.
640 (gimple_init_gcov_profiler): Use ptr_type_node.
641 * value-prof.c (gimple_ic): Use ptr_type_node.
643 2018-10-04 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
645 PR tree-optimization/85787
646 * ipa-pure-const.c (malloc_candidate_p_1): Move most of malloc_candidate_p
647 into this function and add support for detecting multiple phis.
648 (DUMP_AND_RETURN): Move from malloc_candidate_p into top-level macro.
650 2018-10-04 Martin Liska <mliska@suse.cz>
653 * ipa-inline.c (inline_to_all_callers_1):
654 Call ultimate_alias_target for node being inlined.
656 2018-10-03 Jeff Law <law@redhat.com>
658 * gimple-ssa-sprintf.c (format_string): Do not hardcode size of
660 * tree.c (get_typenode_from_name): Moved from fortran/trans-types.c.
661 * tree.h (get_typenode_from_name): Prototype.
663 2018-10-03 Uros Bizjak <ubizjak@gmail.com>
665 * config/i386/i386.md (*cmp<X87MODEF:mode>_<SWI24:mode>_i387):
666 Change operand 2 predicate to nonimmediate_operand.
667 (*cmp<X87MODEF:mode>_<SWI24:mode>_cc_i387): Ditto.
669 2018-10-03 Martin Sebor <msebor@redhat.com>
670 Jeff Law <law@redhat.com>
672 * gimple-ssa-sprintf.c (struct fmtresult): Add new member and
674 (get_string_length): Detect unterminated arrays.
675 (format_string): Same.
676 (format_directive): Warn about unterminated arrays.
677 (handle_gimple_call): Mark statements with no_warning as needed.
679 2018-10-03 Jim Wilson <jimw@sifive.com>
681 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): For ABI_ILP32E,
682 also define __riscv_abi_rve. Delete trailing white space.
684 2018-10-03 Paul Koning <ni1d@arrl.net>
686 Enable LRA register allocator for PDP11.
687 * config/pdp11/constraints.md (Q): Use define_memory_constraint.
690 * config/pdp11/pdp11.c (pdp11_lra_p): New function.
691 * config/pdp11/pdp11.opt (-mlra): New option.
692 * doc/invoke.texi (PDP-11 Options): Document -mlra.
694 2018-10-03 Uros Bizjak <ubizjak@gmail.com>
696 * config/i386/i386.md (*<absneg:code>extendsfdf2): Remove.
697 (*<absneg:code>extend<mode>xf2): Ditto.
699 2018-10-03 Aldy Hernandez <aldyh@redhat.com>
701 PR tree-optimization/87415
702 * tree-vrp.c (set_value_range_with_overflow): Special case one bit
705 2018-10-02 Jeff Law <law@redhat.com>
707 * gimple-fold.c (get_range_strlen): Only set *nonstr when
708 an unterminated string is discovered. Bubble up range
709 even for unterminated strings.
710 (gimple_fold_builtin_strlen): Do not fold if get_range_strlen
711 indicates the string was not terminated via NONSTR.
713 2018-10-03 Aldy Hernandez <aldyh@redhat.com>
715 * tree-vrp.c (extract_range_from_unary_expr): Special case all
717 Do not do anything special for anti-ranges.
719 2018-10-03 Jérôme Lambourg <lambourg@adacore.com>
721 * config/arm/vxworks.h (ARM_TARGET2_DWARF_FORMAT): Adjust to
722 DW_EH_PE_pcrel | DW_EH_PE_indirect for RTPs.
724 2018-10-03 Martin Liska <mliska@suse.cz>
726 PR gcov-profile/86109
727 * coverage.c (coverage_begin_function): Do not
728 mark lambdas as artificial.
729 * tree-core.h (struct GTY): Remove tm_clone_flag
730 and introduce new lambda_function.
731 * tree.h (DECL_LAMBDA_FUNCTION): New macro.
733 2018-10-02 Aaron Sawdey <acsawdey@linux.ibm.com>
736 * config/rs6000/rs6000-string.c (expand_strn_compare): Check that both
737 P8_VECTOR and VSX are enabled.
739 2018-10-02 Andreas Krebbel <krebbel@linux.ibm.com>
741 * config/s390/driver-native.c (s390_host_detect_local_cpu): Add
742 0x3907 as CPU model number.
744 2018-10-02 Andreas Krebbel <krebbel@linux.ibm.com>
746 * common/config/s390/s390-common.c: Rename PF_ARCH12 to PF_Z14.
747 * config/s390/s390.h (enum processor_flags): Rename PF_ARCH12 to
748 PF_Z14. Rename TARGET_CPU_ARCH12 to TARGET_CPU_Z14,
749 TARGET_CPU_ARCH12_P to TARGET_CPU_Z14_P, TARGET_ARCH12 to
750 TARGET_Z14, and TARGET_ARCH12_P to TARGET_Z14_P.
751 * config/s390/s390.md: Likewise. Rename also the cpu attribute
752 value from arch12 to z14.
754 2018-10-02 Uros Bizjak <ubizjak@gmail.com>
756 * config/i386/i386.md (fxam<mode>2_i387_with_temp): Remove.
758 (isinf<mode>2): Ditto.
760 2018-10-02 Uros Bizjak <ubizjak@gmail.com>
762 * config/i386/i386.c (ix86_emit_i387_round): Extend op1 to XFmode
763 before emitting fxam. Perform calculations in XFmode.
765 2018-10-02 Marc Glisse <marc.glisse@inria.fr>
767 * match.pd (((X /[ex] A) +- B) * A): New transformation.
769 2018-10-02 Marc Glisse <marc.glisse@inria.fr>
772 * fold-const.c (fold_plusminus_mult_expr): Handle complex and vectors.
773 * tree.c (signed_or_unsigned_type_for): Handle complex.
775 2018-10-02 Jeff Law <law@redhat.com>
777 * gimple-fold.c (get_range_strlen): Remove dead code.
779 2018-10-02 Martin Sebor <msebor@redhat.com>
780 Jeff Law <law@redhat.com>
782 * builtins.c (unterminated_array): Add new arguments.
783 If argument is not terminated, bubble up size and exact
785 (expand_builtin_strnlen): Detect, avoid expanding
786 and diagnose unterminated arrays.
787 (c_strlen): Fill in offset of start of unterminated strings.
788 * builtins.h (unterminated_array): Update prototype.
790 2018-10-02 Richard Biener <rguenther@suse.de>
792 * config/i386/sse.md (reduc_plus_scal_v4df): Avoid the use
793 of haddv4df, first reduce to SSE width and exploit the fact
794 that we only need element zero with the reduction result.
795 (reduc_plus_scal_v2df): Likewise.
797 2018-10-02 Eric Botcazou <ebotcazou@adacore.com>
799 * dojump.h (do_jump): Delete.
800 (do_jump_1): Likewise.
801 (split_comparison): Move around.
802 * dojump.c (do_jump): Make static.
803 (do_jump_1): Likewise.
804 (jumpifnot): Move around.
805 (jumpifnot_1): Likewise.
807 (jumpif_1): Likewise.
808 * expr.c (expand_expr_real_1): Call jumpif[not] instead of do_jump.
810 2018-10-02 Eric Botcazou <ebotcazou@adacore.com>
812 * reorg.c (make_return_insns): Use emit_copy_of_insn_after for the
813 insns in the delay slot and add_insn_after for the jump insn.
815 2018-10-02 Richard Biener <rguenther@suse.de>
817 * tree-inline.c (expand_call_inline): Use the location of
818 the callee declaration for the inline-entry marker.
819 * final.c (notice_source_line): Remove special-casing of
820 NOTE_INSN_INLINE_ENTRY.
822 2018-10-01 Carl Love <cel@us.ibm.com>
825 * config/rs6000/rs6000-builtin.def (__builtin_mffsl): New.
826 (__builtin_mtfsb0): New.
827 (__builtin_mtfsb1): New.
828 ( __builtin_set_fpscr_rn): New.
829 (__builtin_set_fpscr_drn): New.
830 * config/rs6000/rs6000.c (rs6000_expand_mtfsb_builtin): Add.
831 (rs6000_expand_set_fpscr_rn_builtin): Add.
832 (rs6000_expand_set_fpscr_drn_builtin): Add.
833 (rs6000_expand_builtin): Add case statement entries for
834 RS6000_BUILTIN_MTFSB0, RS6000_BUILTIN_MTFSB1,
835 RS6000_BUILTIN_SET_FPSCR_RN, RS6000_BUILTIN_SET_FPSCR_DRN,
836 RS6000_BUILTIN_MFFSL.
837 (rs6000_init_builtins): Add ftype initialization and def_builtin
838 calls for __builtin_mffsl, __builtin_mtfsb0, __builtin_mtfsb1,
839 __builtin_set_fpscr_rn, __builtin_set_fpscr_drn.
840 * config/rs6000.md (rs6000_mtfsb0, rs6000_mtfsb1, rs6000_mffscrn,
841 rs6000_mffscdrn): Add define_insn.
842 (rs6000_set_fpscr_rn, rs6000_set_fpscr_drn): Add define_expand.
843 * doc/extend.texi: Add documentation for the builtins.
845 2018-10-01 Richard Biener <rguenther@suse.de>
847 PR tree-optimization/87465
848 * tree-ssa-loop-ivcanon.c (tree_estimate_loop_size): Fix typo
849 causing branch miscounts.
851 2018-10-01 Tamar Christina <tamar.christina@arm.com>
853 * common/config/aarch64/aarch64-common.c (TARGET_OPTION_DEFAULT_PARAM,
854 aarch64_option_default_param): New.
856 (TARGET_OPTION_VALIDATE_PARAM, aarch64_option_validate_param): New.
857 * config/aarch64/aarch64.c (aarch64_override_options_internal): Simplify
858 stack-clash protection validation code.
860 2018-10-01 Tamar Christina <tamar.christina@arm.com>
862 * params.c (validate_param): New.
863 (add_params): Use it.
864 (set_param_value): Refactor param validation into validate_param.
865 (diagnostic.h): Include.
866 * diagnostic.h (diagnostic_ready_p): New.
868 2018-10-01 Tamar Christina <tamar.christina@arm.com>
870 * params.c (set_param_value):
871 Add index of parameter being validated.
872 * common/common-target.def (option_validate_param): New.
873 * common/common-targhooks.h (default_option_validate_param): New.
874 * common/common-targhooks.c (default_option_validate_param): New.
875 * doc/tm.texi.in (TARGET_OPTION_VALIDATE_PARAM): New.
876 * doc/tm.texi: Regenerate.
878 2018-10-01 Tamar Christina <tamar.christina@arm.com>
881 * config/aarch64/aarch64.c (aarch64_override_options_internal):
882 Add validation for stack-clash parameters and set defaults.
884 2018-10-01 Tamar Christina <tamar.christina@arm.com>
887 * configure.ac: Add stack-clash-protection-guard-size.
888 * doc/install.texi: Document it.
889 * config.in (DEFAULT_STK_CLASH_GUARD_SIZE): New.
890 * params.def: Update comment for guard-size.
891 (PARAM_STACK_CLASH_PROTECTION_GUARD_SIZE,
892 PARAM_STACK_CLASH_PROTECTION_PROBE_INTERVAL): Update description.
893 * configure: Regenerate.
895 2018-10-01 Tamar Christina <tamar.christina@arm.com>
898 * config/aarch64/aarch64.h (STACK_CLASH_MIN_BYTES_OUTGOING_ARGS,
899 STACK_DYNAMIC_OFFSET): New.
900 * config/aarch64/aarch64.c (aarch64_layout_frame):
901 Update outgoing args size.
902 (aarch64_stack_clash_protection_alloca_probe_range,
903 TARGET_STACK_CLASH_PROTECTION_ALLOCA_PROBE_RANGE): New.
905 2018-10-01 Tamar Christina <tamar.christina@arm.com>
908 * explow.c (anti_adjust_stack_and_probe_stack_clash): Support custom
910 * target.def (stack_clash_protection_alloca_probe_range): New.
911 (stack_clash_protection_final_dynamic_probe): Remove.
912 * targhooks.h (default_stack_clash_protection_alloca_probe_range) New.
913 (default_stack_clash_protection_final_dynamic_probe): Remove.
914 * targhooks.c: Likewise.
915 * doc/tm.texi.in (TARGET_STACK_CLASH_PROTECTION_ALLOCA_PROBE_RANGE): New.
916 (TARGET_STACK_CLASH_PROTECTION_FINAL_DYNAMIC_PROBE): Remove.
917 * doc/tm.texi: Regenerate.
919 2018-10-01 Tamar Christina <tamar.christina@arm.com>
922 * config/aarch64/aarch64-protos.h (aarch64_output_probe_sve_stack_clash): New.
923 * config/aarch64/aarch64.c (aarch64_output_probe_sve_stack_clash,
924 aarch64_clamp_to_uimm12_shift): New.
925 (aarch64_allocate_and_probe_stack_space): Add SVE specific section.
926 * config/aarch64/aarch64.md (probe_sve_stack_clash): New.
928 2018-10-01 Tamar Christina <tamar.christina@arm.com>
931 * config/aarch64/aarch64.c (aarch64_layout_frame): Add assert.
933 2018-10-01 Jeff Law <law@redhat.com>
934 Richard Sandiford <richard.sandiford@linaro.org>
935 Tamar Christina <tamar.christina@arm.com>
938 * config/aarch64/aarch64.md
939 (probe_stack_range): Add k (SP) constraint.
940 * config/aarch64/aarch64.h (STACK_CLASH_CALLER_GUARD,
941 STACK_CLASH_MAX_UNROLL_PAGES): New.
942 * config/aarch64/aarch64.c (aarch64_output_probe_stack_range): Emit
943 stack probes for stack clash.
944 (aarch64_allocate_and_probe_stack_space): New.
945 (aarch64_expand_prologue): Use it.
946 (aarch64_expand_epilogue): Likewise and update IP regs re-use criteria.
947 (aarch64_sub_sp): Add emit_move_imm optional param.
949 2018-10-01 MCC CS <deswurstes@users.noreply.github.com>
951 PR tree-optimization/87261
952 * match.pd: Remove trailing whitespace.
953 Add (x & y) | ~(x | y) -> ~(x ^ y),
954 (~x | y) ^ (x ^ y) -> x | ~y and (x ^ y) | ~(x | y) -> ~(x & y)
956 2018-10-01 Claudiu Zissulescu <claziss@synopsys.com>
958 * config/arc/arc.md (*add_n): Clean up pattern, update instruction
960 (ashlsi3_insn): Update instruction constraints.
961 (ashrsi3_insn): Likewise.
963 (add_shift): Likewise.
964 * config/arc/constraints.md (Csz): New 32 bit constraint. It
965 avoids placing in the limm field small constants which, otherwise,
966 could end into a small instruction.
968 2018-10-01 Claudiu Zissulescu <claziss@synopsys.com>
970 * config/arc/arc.md (maddsidi4_split): Don't use dmac if the
971 destination register is not odd-even.
972 (umaddsidi4_split): Likewise.
974 2018-10-01 Richard Biener <rguenther@suse.de>
976 * tree-inline.c (expand_call_inline): Store origin of fn
977 in BLOCK_ABSTRACT_ORIGIN for the inline BLOCK.
978 * tree.c (block_ultimate_origin): Simplify and do some
981 2018-09-30 Uros Bizjak <ubizjak@gmail.com>
983 * config/i386/mmx.md (EMMS): New int iterator.
984 (emms): New int attribute.
985 (mmx_<emms>): Macroize insn from *mmx_emms and *mmx_femms using
986 EMMS int iterator. Explicitly declare clobbers.
987 (mmx_emms): Remove expander.
989 * config/i386/predicates.md (emms_operation): Remove predicate.
990 (vzeroall_pattern): New predicate.
991 (vzeroupper_pattern): Rename from vzeroupper_operation.
992 * config/i386/i386.c (ix86_avx_u128_mode_after): Use
993 vzeroupper_pattern and vzeroall_pattern predicates.
995 2018-09-30 Peter Bergner <bergner@linux.ibm.com>
997 PR rtl-optimization/86939
998 * ira-lives.c (make_hard_regno_born): Rename from this...
999 (make_hard_regno_live): ... to this. Remove update to conflict
1000 information. Update function comment.
1001 (make_hard_regno_dead): Add conflict information update. Update
1003 (make_object_born): Rename from this...
1004 (make_object_live): ... to this. Remove update to conflict information.
1005 Update function comment.
1006 (make_object_dead): Add conflict information update. Update function
1008 (mark_pseudo_regno_live): Call make_object_live.
1009 (mark_pseudo_regno_subword_live): Likewise.
1010 (mark_hard_reg_dead): Update function comment.
1011 (mark_hard_reg_live): Call make_hard_regno_live.
1012 (process_bb_node_lives): Likewise.
1013 * lra-lives.c (make_hard_regno_born): Rename from this...
1014 (make_hard_regno_live): ... to this. Remove update to conflict
1015 information. Remove now uneeded check_pic_pseudo_p argument.
1016 Update function comment.
1017 (make_hard_regno_dead): Add check_pic_pseudo_p argument and add update
1018 to conflict information. Update function comment.
1019 (mark_pseudo_live): Remove update to conflict information. Update
1021 (mark_pseudo_dead): Add conflict information update.
1022 (mark_regno_live): Call make_hard_regno_live.
1023 (mark_regno_dead): Call make_hard_regno_dead with new arguement.
1024 (process_bb_lives): Call make_hard_regno_live and make_hard_regno_dead.
1026 2018-09-29 H.J. Lu <hongjiu.lu@intel.com>
1029 * config/i386/i386.c (construct_container): Use TImode for
1030 BLKmode values in 2 integer registers.
1032 2018-09-29 Jeff Law <law@redhat.com>
1034 * builtins.c (unterminated_array): Pass in c_strlen_data * to
1035 c_strlen rather than just a tree *.
1036 (c_strlen): Change NONSTR argument to a c_strlen_data pointer.
1037 Update recursive calls appropriately. If caller did not provide a
1038 suitable data pointer, create a local one. When a non-terminated
1039 string is discovered, bubble up information about the string via the
1040 c_strlen_data object.
1041 * builtins.h (c_strlen): Update prototype.
1042 (c_strlen_data): New structure.
1043 * gimple-fold.c (get_range_strlen): Update calls to c_strlen.
1044 For a type 2 call, if c_strlen indicates a non-terminated string
1045 use the length of the non-terminated string.
1046 (gimple_fold_builtin_stpcpy): Update calls to c_strlen.
1048 2018-09-29 Jakub Jelinek <jakub@redhat.com>
1051 * config/i386/avx512fintrin.h (_mm512_abs_pd, _mm512_mask_abs_pd): Use
1052 __m512d type for __A argument rather than __m512.
1054 2018-09-28 John David Anglin <danglin@gcc.gnu.org>
1056 * match.pd (simple_comparison): Don't optimize if either operand is
1057 a function pointer when target needs function pointer canonicalization.
1059 2018-09-28 Segher Boessenkool <segher@kernel.crashing.org>
1061 * config/rs6000/driver-rs6000.c (asm_names): Adjust the entries for
1062 power5 .. power9 to remove indirection.
1063 * config/rs6000/rs6000.h (ASM_CPU_POWER5_SPEC, ASM_CPU_POWER6_SPEC,
1064 ASM_CPU_POWER7_SPEC, ASM_CPU_POWER8_SPEC, ASM_CPU_POWER9_SPEC,
1065 ASM_CPU_476_SPEC): Delete.
1066 (ASM_CPU_SPEC): Adjust.
1067 (EXTRA_SPECS): Delete asm_cpu_power5, asm_cpu_power6, asm_cpu_power7,
1068 asm_cpu_power8, asm_cpu_power9, asm_cpu_476.
1070 2018-09-28 Segher Boessenkool <segher@kernel.crashing.org>
1072 * config.in: Delete HAVE_AS_DCI.
1073 * config/powerpcspe/powerpcspe.h: Treat HAVE_AS_DCI as always true.
1074 * config/rs6000/rs6000.h: Ditto.
1075 * configure.ac: Delete HAVE_AS_DCI.
1076 * configure: Regenerate.
1078 2018-09-28 Segher Boessenkool <segher@kernel.crashing.org>
1080 * config.in (HAVE_AS_LWSYNC): Delete.
1081 * config/powerpcspe/powerpcspe.h (TARGET_LWSYNC_INSTRUCTION): Delete.
1082 * config/powerpcspe/sync.md (*lwsync): Always generate lwsync, never
1084 * config/rs6000/rs6000.h (TARGET_LWSYNC_INSTRUCTION): Delete.
1085 * config/rs6000/sync.md (*lwsync): Always generate lwsync, never do it
1087 * configure.ac: Delete HAVE_AS_LWSYNC.
1088 * configure: Regenerate.
1090 2018-09-28 Eric Botcazou <ebotcazou@adacore.com>
1091 Pierre-Marie de Rodat <derodat@adacore.com>
1093 * calls.c (expand_call): Try to do a tail call for thunks at -O0 too.
1094 * cgraph.h (struct cgraph_thunk_info): Add indirect_offset.
1095 (cgraph_node::create_thunk): Add indirect_offset parameter.
1096 (thunk_adjust): Likewise.
1097 * cgraph.c (cgraph_node::create_thunk): Add indirect_offset parameter
1098 and initialize the corresponding field with it.
1099 (cgraph_node::dump): Dump indirect_offset field.
1100 * cgraphclones.c (duplicate_thunk_for_node): Deal with indirect_offset.
1101 * cgraphunit.c (cgraph_node::analyze): Be prepared for external thunks.
1102 (thunk_adjust): Add indirect_offset parameter and deal with it.
1103 (cgraph_node::expand_thunk): Deal with the indirect_offset field and
1104 pass it to thunk_adjust. Do not call the target hook if it's non-zero
1105 or if the thunk is external or local. Fix formatting. Do not chain
1106 the RESULT_DECL to BLOCK_VARS. Pass the static chain to the target,
1107 if any, in the GIMPLE representation.
1108 * ipa-icf.c (sem_function::equals_wpa): Deal with indirect_offset.
1109 * lto-cgraph.c (lto_output_node): Write indirect_offset field.
1110 (input_node): Read indirect_offset field.
1111 * tree-inline.c (expand_call_inline): Pass indirect_offset field in the
1112 call to thunk_adjust.
1113 * tree-nested.c (struct nesting_info): Add thunk_p field.
1114 (create_nesting_tree): Set it.
1115 (convert_all_function_calls): Copy static chain from targets to thunks.
1116 (finalize_nesting_tree_1): Return early for thunks.
1117 (unnest_nesting_tree_1): Do not finalize thunks.
1118 (gimplify_all_functions): Do not gimplify thunks.
1120 2018-09-28 David Malcolm <dmalcolm@redhat.com>
1122 * opt-suggestions.c (option_proposer::build_option_suggestions):
1123 Release "option_values".
1125 2018-09-28 David Malcolm <dmalcolm@redhat.com>
1127 * coverage.c (get_coverage_counts): Convert problem-reporting dump
1128 messages from MSG_OPTIMIZED_LOCATIONS to MSG_MISSED_OPTIMIZATION.
1129 * dumpfile.c (kind_as_string): New function.
1130 (dump_loc): Rather than a hardcoded prefix of "note: ", use
1131 kind_as_string to vary the prefix based on dump_kind.
1132 (selftest::test_capture_of_dump_calls): Update for above.
1134 2018-09-28 Uros Bizjak <ubizjak@gmail.com>
1136 * config/i386/i386.h (SSE_REGNO): Fix check for FIRST_REX_SSE_REG.
1137 (GET_SSE_REGNO): Rename from SSE_REGNO. Update all uses for rename.
1139 2018-09-28 Uros Bizjak <ubizjak@gmail.com>
1141 * config/i386/i386.h (CC_REGNO): Remove FPSR_REG.
1142 * config/i386/i386.c (ix86_fixed_condition_code_regs): Use
1143 INVALID_REGNUM instead of FPSR_REG.
1144 (ix86_md_asm_adjust): Do not clobber FPSR_REG.
1145 * config/i386/i386.md: Update comment of FP compares.
1146 (fldenv): Do not clobber FPSR_REG.
1148 2018-09-28 Richard Biener <rguenther@suse.de>
1150 * tree.h (BLOCK_ORIGIN): New.
1151 * omp-expand.c (grid_expand_target_grid_body): Assign
1152 BLOCK_ORIGIN to BLOCK_ABSTRACT_ORIGIN.
1153 * tree-inline.c (remap_block): Likewise.
1154 * auto-profile.c (get_function_decl_from_block): Simplify
1155 by eliding the BLOCK_ABSTRACT_ORIGIN chasing.
1156 * langhooks.c (lhd_print_error_function): Likewise.
1157 * optinfo-emit-json.cc (optrecord_json_writer::inlining_chain_to):
1159 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
1160 * tree.c (block_nonartificial_location): Likewise.
1161 (block_ultimate_origin): Likewise.
1162 * tree-pretty-print.c (percent_K_format): Likewise. Remove
1163 no longer needed LTO case.
1165 2018-09-28 Andrew Stubbs <ams@codesourcery.com>
1166 Jan Hubicka <jh@suse.cz>
1167 Martin Jambor <mjambor@suse.cz>
1169 * simplify-rtx.c (simplify_merge_mask): New function.
1170 (simplify_ternary_operation): Use it, also see if VEC_MERGEs with the
1171 same masks are used in op1 or op2.
1172 (test_vec_merge): New function.
1173 (test_vector_ops): Call test_vec_merge.
1175 2018-09-28 Eric Botcazou <ebotcazou@adacore.com>
1177 * config/sparc/sparc-protos.h (sparc_branch_cost): Declare.
1178 * config/sparc/sparc.h (BRANCH_COST): Call sparc_branch_cost.
1179 * config/sparc/sparc.c (struct processor_costs): Add branch_cost field.
1180 (cypress_costs): Set it.
1181 (supersparc_costs): Likewise.
1182 (hypersparc_costs): Likewise.
1183 (leon_cost): Likewise.
1184 (leon3_costs): Likewise.
1185 (sparclet_costs): Likewise.
1186 (ultrasparc_costs): Likewise.
1187 (ultrasparc_costs): Likewise.
1188 (niagara_costs): Likewise.
1189 (niagara2_costs): Likewise.
1190 (niagara3_costs): Likewise.
1191 (niagara4_costs): Likewise.
1192 (niagara7_costs): Likewise.
1193 (m8_costs): Likewise.
1194 (TARGET_CAN_FOLLOW_JUMP): Define.
1195 (pass_work_around_errata::gate): Minor tweak.
1196 (sparc_option_override): Remove MASK_FSMULD mask for V7 processors.
1197 Do not set both MASK_VIS4 and MASK_VIS4B for M8 processor.
1198 Automaitcally clear MASK_FSMULD mask for V7 processors.
1199 (sparc_can_follow_jump): New static function.
1200 (output_ubranch): Deal with CROSSING_JUMP_P.
1201 (sparc_use_sched_lookahead): Rewrite using switch statement.
1202 (sparc_issue_rate): Reorder.
1203 (sparc_branch_cost): New function.
1205 2018-09-27 Martin Sebor <msebor@redhat.com>
1207 * tree.h (tree_to_shwi): Add attribute nonnull and pure.
1208 (tree_to_poly_int64, tree_to_uhwi, tree_to_poly_uint64): Same.
1209 (int_fits_type_p): Same.
1211 2018-09-27 Uros Bizjak <ubizjak@gmail.com>
1213 * config/i386/i386.md (FPCR_REG): Remove.
1214 (UNSPEC_FLDCW): Remove.
1215 (x86_fnstcw_1): Use (const_int 0) instead of FPCR_REG.
1216 (x86_fldcw_1): Remove insn pattern.
1217 (fnstenv): Do not clobber FPCR_REG.
1219 * config/i386/i386.h (FIXED_REGISTERS) Remove fpsr register.
1220 (CALL_USED_REGISTERS): Ditto.
1221 (REG_ALLOC_ORDER): Ditto.
1222 (REG_CLASS_CONTENTS): Ditto.
1223 (HI_REGISTER_NAMES): Ditto.
1224 (ADDITIONAL_REGISTER_NAMES): Use defines instead
1225 of numerical constants.
1226 * config/i386/i386.c (regclass_map): Remove fpsr register.
1227 (dbx_register_map): Ditto.
1228 (dbx64_register_map): Ditto.
1229 (svr4_dbx_register_map): Ditto.
1230 (print_reg): Do not handle FPCR_REG.
1232 2018-09-27 Segher Boessenkool <segher@kernel.crashing.org>
1235 * config.in (HAVE_AS_CMPB, HAVE_AS_DFP, HAVE_AS_FPRND, HAVE_AS_MFPGPR,
1236 HAVE_AS_POPCNTB, HAVE_AS_POPCNTD, HAVE_AS_POWER8, HAVE_AS_POWER9):
1237 Delete, always treat as true.
1238 * config/powerpcspe/powerpcspe.c (rs6000_option_override_internal):
1239 Ditto. Simplify remaining code.
1240 * config/powerpcspe/powerpcspe.h: Ditto.
1241 * config/rs6000/rs6000.c (rs6000_option_override_internal): Ditto.
1242 Simplify remaining code.
1243 (rs6000_expand_builtin): Ditto.
1244 * config/rs6000/rs6000.h: Ditto.
1245 * configure.ac: Ditto.
1246 * configure: Regenerate.
1248 2018-09-27 Martin Liska <mliska@suse.cz>
1250 * coverage.c (get_coverage_counts): Revert the formatting
1251 of missing profile opt info.
1253 2018-09-27 Richard Biener <rguenther@suse.de>
1257 * dwarf2out.c (set_block_origin_self): Do not mark outermost
1258 block as we do not output that.
1259 (gen_inlined_subroutine_die): Elide the originally outermost
1260 block, matching what we do for concrete instances.
1261 (decls_for_scope): Add parameter specifying whether to recurse
1264 2018-09-27 Andrew Stubbs <ams@codesourcery.com>
1265 Tom de Vries <tom@codesourcery.com>
1269 * expmed.c (emit_cstore): Fix handling of result_mode == BImode and
1270 STORE_FLAG_VALUE == 1.
1272 2018-09-27 Andreas Krebbel <krebbel@linux.ibm.com>
1274 * config/s390/s390.md (PPA_TX_ABORT, PPA_OOO_BARRIER): New
1275 constant definitions.
1276 ("tx_assist"): Replace magic number with PPA_TX_ABORT.
1277 ("*ppa"): Enable pattern also for -march=zEC12 -mno-htm.
1278 ("speculation_barrier"): New expander definition.
1280 2018-09-26 Indu Bhagat <indu.bhagat@oracle.com>
1282 PR gcov-profile/86957
1283 * common.opt: New warning option -Wmissing-profile.
1284 * coverage.c (get_coverage_counts): Add warning for missing .gcda file.
1285 * doc/invoke.texi: Document -Wmissing-profile.
1287 2018-09-26 Jim Wilson <jimw@sifive.com>
1289 * config/riscv/riscv.md (subsi3_extended2): Add J constraint.
1290 (negdi2, negsi2, negsi2_extended, negsi2_extended2): New.
1292 2018-09-26 Martin Sebor <msebor@redhat.com>
1294 * tree.c (zerop): Change return type to bool.
1295 (integer_zerop, integer_onep, integer_each_onep): Same.
1296 (integer_all_onesp, integer_minus_onep, integer_pow2p): Same.
1297 (integer_nonzerop, integer_truep, tree_ctz, real_zerop): Same.
1298 (real_onep, real_minus_onep, chain_index): Same.
1299 (print_type_hash_statistics, type_list_equal): Same.
1300 * tree.h (zerop): Same.
1301 (zerop, integer_zerop, integer_onep, integer_each_onep): Same.
1302 (integer_all_onesp, integer_minus_onep, integer_pow2p): Same.
1303 (integer_nonzerop, integer_truep, tree_ctz, real_zerop): Same.
1304 (real_onep, real_minus_onep, chain_index): Same.
1305 (print_type_hash_statistics, type_list_equal): Same.
1307 2018-09-26 Jim Wilson <jimw@sifive.com>
1309 * config/riscv/riscv.h (FUNCTION_ARG_REGNO_P): Fix comment.
1311 2018-09-26 Jakub Jelinek <jakub@redhat.com>
1314 * config/i386/i386.c: Include debug.h and dwarf2out.h.
1315 (output_indirect_thunk): Emit DW_CFA_def_cfa_offset after the
1318 2018-09-25 Andrew Stubbs <ams@codesourcery.com>
1320 * builtins.c (get_builtin_sync_mem): Force address mode conversion.
1322 2018-09-26 Uros Bizjak <ubizjak@gmail.com>
1324 * config/i386/i386.h (enum reg_class): Remove FP_TOP_SSE_REGS
1325 and FP_SECOND_SSE_REGS.
1326 (REG_CLASS_NAMES): Ditto.
1327 (REG_CLASS_CONTENTS): Ditto.
1328 * config/i386/i386.c (ix86_preferred_reload_class) Do not handle
1329 FP_TOP_SSE_REGS and FP_SECOND_SSE_REGS classes.
1330 (ix86_preferred_output_reload_class): Ditto.
1331 * config/i386/i386.md (fix_trunc<mode>_i387_fisttp): Change "=&1f"
1332 clobber constraint to "=&f".
1333 (fix_truncdi_i387): Ditto.
1334 (lrintxfdi2): Ditto.
1335 (fistdi2_<rounding>): Ditto.
1336 (fpremxf4_i387): Change "=u" constraint to "=f".
1337 (fprem1xf4_i387): Ditto.
1339 (fptanxf4_i387): Ditto.
1340 (fxtractxf3_i387): Ditto.
1341 (fscalexf4_i387): Ditto.
1342 (atan2xf3): Change "u" constraint to "f".
1343 (fyl2xxf3_i387): Ditto.
1344 (fyl2xp1xf3_i387): Ditto.
1346 2018-09-26 Uros Bizjak <ubizjak@gmail.com>
1349 * config/i386/i386.h (NUM_MODES_FOR_MODE_SWITCHING): Update
1350 for removed I387_MASK_PM entity.
1353 2018-09-26 Jeff Law <law@redhat.com>
1355 2018-09-26 Alexey Neyman <stilor@att.net>
1357 * graphite.h: Include <isl/id.h> and <isl/space.h>; these
1358 headers are no longer pulled in by <isl/val.h>.
1360 2018-09-26 Richard Biener <rguenther@suse.de>
1363 * dwarf2out.c (gen_lexical_block_die): Do not equate inline
1364 or concrete instance DIE to the tree. Create abstract origin
1365 attributes also for concrete instances.
1367 2018-09-26 Alexey Neyman <stilor@att.net>
1369 * graphite.h: Include <isl/id.h> and <isl/space.h>; these
1370 headers are no longer pulled in by <isl/val.h>.
1372 2018-09-26 Matthew Malcomson <matthew.malcomson@arm.com>
1374 * config/arm/arm.c (arm_split_compare_and_swap, arm_split_atomic_op):
1375 Use new helper functions.
1376 * config/arm/sync.md (atomic_load<mode>, atomic_store<mode>):
1377 Use new helper functions.
1378 * config/arm/aarch-common-protos.h (aarch_mm_needs_acquire,
1379 aarch_mm_needs_release): New declarations.
1380 * config/arm/aarch-common.c (aarch_mm_needs_acquire,
1381 aarch_mm_needs_release): New.
1383 2018-09-26 Eric Botcazou <ebotcazou@adacore.com>
1385 * config/arm/arm.c (arm_reorg): Skip Thumb reorg pass for thunks.
1386 (arm32_output_mi_thunk): Deal with long calls.
1388 2018-09-26 Richard Biener <rguenther@suse.de>
1392 * tree-inline.c (expand_call_inline): When the location
1393 of the call is UNKNOWN_LOCATION use DECL_SOURCE_LOCATION
1394 or BUILTINS_LOCATION for the BLOCK_SOURCE_LOCATION of
1395 the inserted BLOCK to make inlined_function_outer_scope_p
1397 * dwarf2out.c (add_call_src_coords_attributes): Do not add
1398 coords for reserved locations.
1400 2018-09-25 Segher Boessenkool <segher@kernel.crashing.org>
1402 * config/rs6000/rs6000.md (*movcc_internal1): Use set_attr_alternative.
1403 (*call_indirect_nonlocal_sysv<mode>): Ditto.
1404 (*call_value_indirect_nonlocal_sysv<mode>): Ditto.
1405 (*sibcall_nonlocal_sysv<mode>): Ditto.
1406 (*sibcall_value_nonlocal_sysv<mode>): Ditto.
1407 (<bd>_<mode>): Ditto.
1408 (<bd>tf_<mode>): Ditto.
1410 2018-09-25 Segher Boessenkool <segher@kernel.crashing.org>
1412 * config/rs6000/altivec.md (*altivec_mov<mode>): Write the output
1413 control string as a list of templates instead of as C code.
1414 (*altivec_movti): Ditto.
1415 * config/rs6000/darwin.md (movdf_low_di): Ditto.
1417 2018-09-25 Jim Wilson <jimw@sifive.com>
1419 * config/riscv/riscv.c (riscv_split_symbol): Mark auipc label as weak
1420 when target symbol is weak.
1422 2018-09-25 Bernd Edlinger <bernd.edlinger@hotmail.de>
1425 * builtins.c (unterminated_array): Simplify.
1426 * expr.c (string_constant): Handle SSA_NAME. Add more exceptions
1427 where pointer arithmetic is safe.
1429 2018-09-25 Segher Boessenkool <segher@kernel.crashing.org>
1432 * config/rs6000/altivec.md (altivec_vspltb): Use
1433 const_0_to_15_operand instead of u5bit_cint_operand.
1434 (*altivec_vspltb_internal): Ditto.
1435 (altivec_vspltb_direct): Ditto.
1436 (altivec_vsplth): Use const_0_to_7_operand instead of
1438 (*altivec_vsplth_internal): Ditto.
1439 (altivec_vsplth_direct): Ditto.
1440 (altivec_vspltw): Use const_0_to_3_operand instead of
1442 (*altivec_vspltw_internal): Ditto.
1443 (altivec_vspltw_direct): Ditto.
1444 (altivec_vspltsf): Ditto.
1445 (*altivec_vspltsf_internal): Ditto.
1446 * config/rs6000/rs6000.c (rs6000_expand_binop_builtin): Handle the
1447 various splats with the proper size immediate. Reorder the various
1448 cases by ascending size of immediate, and put all such together.
1450 2018-09-25 Richard Biener <rguenther@suse.de>
1453 * dwarf2out.c (add_AT_external_die_ref): Remove now redundant
1455 (maybe_create_die_with_external_ref): Do not create
1456 DW_TAG_imported_unit here.
1457 (add_abstract_origin_attribute): Handle external BLOCK refs.
1458 (dwarf2out_abstract_function): Simplify LTO case.
1459 (dwarf2out_early_finish): Create DW_TAG_imported_unit explicitely
1460 rather than using maybe_create_die_with_external_ref.
1462 2018-09-25 Uros Bizjak <ubizjak@gmail.com>
1465 * config/i386/i386.md (frndintxf2_mask_pm): Remove.
1466 (frndintxf2_mask_pm_i387): Ditto.
1467 (nearbyintxf2): Rewrite expander pattern to match rintxf2.
1468 Enable for !flag_trapping_math.
1469 (nearbyint<mode>2): Enable x87 modes for !flag_trapping_math.
1470 Enable SSE modes for TARGET_SSE4_1 and expand them with round insn.
1471 Change operand 1 predicate to nonimmediate_operand.
1472 (attr "i387_cw"): Remove mask_pm.
1473 * config/i386/i386.h (enum ix86_stack_slot): Remove SLOT_CW_MASK_PM.
1474 (enum ix86_entity): Remove I387_MASK_PM.
1475 * config/i386/i386.c (ix86_i387_mode_needed): Do not
1476 handle I387_MASK_PM.
1477 (ix86_mode_needed): Ditto.
1478 (ix86_mode_after): Ditto.
1479 (ix86_mode_entry): Ditto.
1480 (ix86_mode_exit): Ditto.
1481 (emit_i387_cw_initialization): Do not handle I387_CW_MASK_PM.
1483 2018-09-25 Jakub Jelinek <jakub@redhat.com>
1485 * vr-values.c (vr_values::vr_values): Initialize to_remove_edges and
1486 to_update_switch_stmts to vNULL instead of calling create on them
1489 2018-09-25 Richard Biener <rguenther@suse.de>
1491 PR tree-optimization/87402
1492 * tree-ssa-sccvn.c (SSA_VISITED): Remove unused function.
1493 (visit_phi): Re-instantiate handling of supposed to be VARYING
1494 but non-VARYING backedge value.
1496 2018-09-25 Richard Biener <rguenther@suse.de>
1499 * dwarf2out.c (struct sym_off_pair): New.
1500 (external_die_map): New global.
1501 (lookup_decl_die): When in LTO create DIEs lazily from the
1503 (lookup_block_die): New function, create DIEs lazily in LTO.
1504 (equate_block_to_die): New function.
1505 (dwarf2out_die_ref_for_decl): During WPA get the association
1506 from the external DIE map.
1507 (dwarf2out_register_external_die): Record mapping into the
1509 (maybe_create_die_with_external_ref): New function split out from
1510 DIE generation part of old dwarf2out_register_external_die.
1511 (add_abstract_origin_attribute): Do not return the DIE. When
1512 in LTO reference externals directly.
1513 (dwarf2out_abstract_function): When in LTO ignore calls for
1514 decls with external DIEs (already present abstract instances).
1515 (gen_call_site_die): Adjust.
1516 (add_high_low_attributes): Likewise.
1517 (gen_lexical_block_die): Likewise.
1518 (gen_inlined_subroutine_die): Likewie.
1519 (gen_block_die): Likewise.
1520 (dwarf2out_inline_entry): Likewise.
1521 (dwarf2out_early_finish): In LTRANS phase create DW_TAG_imported_unit
1524 2018-09-25 Martin Liska <mliska@suse.cz>
1526 * ipa-fnsummary.c (estimate_node_size_and_time): Scale by two
1527 integers and not by a float value.
1529 2018-09-25 Martin Liska <mliska@suse.cz>
1532 * dbgcnt.c (dbg_cnt_process_single_pair): Return false
1534 * dumpfile.c (dump_enable_all): Remove extra parenthesis.
1535 * gcov-tool.c: Declare the function with ATTRIBUTE_NORETURN.
1536 * godump.c (go_format_type): Remove extra parenthesis.
1538 2018-09-25 Martin Liska <mliska@suse.cz>
1540 * alias.c (set_dest_equal_p): Remove unused function.
1541 * config/i386/i386.c (def_builtin_pure2): Likewise.
1542 * diagnostic-show-locus.c (class layout): Remove
1544 (layout::layout): Likewise here.
1545 * dump-context.h (class temp_dump_context): Likewise.
1546 * dwarf2out.c (add_AT_fde_ref): Remove unused function.
1547 (add_AT_loclistsptr): Likewise.
1548 (add_AT_offset): Likewise.
1549 (get_AT_hi_pc): Likewise.
1550 (is_comdat_die): Likewise.
1551 (type_is_enum): Likewise.
1552 (ceiling): Likewise.
1553 (add_AT_vms_delta): Likewise.
1554 (is_class_die): Likewise.
1555 * edit-context.c (class line_event): Remove unused field.
1556 * graphite-sese-to-poly.c (tree_int_to_gmp): Remove
1558 * ipa-cp.c (ipa_get_vr_lat): Likewise.
1559 * lra-constraints.c (ok_for_index_p_nonstrict): Likewise.
1560 (ok_for_base_p_nonstrict): Likewise.
1561 * tree-chrec.c (is_not_constant_evolution): Likewise.
1562 (chrec_fold_poly_cst): Likewise.
1563 * tree-if-conv.c (has_pred_critical_p): Likewise.
1564 * tree-ssa-coalesce.c (print_exprs): Likewise.
1565 * tree-ssa-pre.c (bitmap_set_contains_expr): Likewise.
1566 * tree-ssa-uninit.c (is_and_or_or_p): Likewise.
1567 * tree-vrp.c (value_ranges_intersect_p): Likewise.
1568 (value_range_nonnegative_p): Likewise.
1570 2018-09-25 Martin Liska <mliska@suse.cz>
1572 * config/powerpcspe/powerpcspe.c (rs6000_output_function_epilogue):
1573 Do not handle "GNU Pascal".
1574 * config/rs6000/rs6000.c (rs6000_output_function_epilogue):
1576 * config/sparc/sparc.c (sparc_pass_by_reference): Remove Pascal
1577 from documentation. Likewise.
1578 * dbxout.c (dbxout_range_type): Likewise.
1579 * doc/cpp.texi: Likewise.
1580 * doc/extend.texi: Likewise.
1581 * doc/frontends.texi: Likewise.
1582 * doc/invoke.texi: Remove Pascal entry.
1583 * tree.def (CLEANUP_POINT_EXPR): Likewise.
1584 * doc/rtl.texi (MODE_FUNCTION): Remove not used entry.
1586 2018-09-25 Martin Liska <mliska@suse.cz>
1589 * doc/invoke.texi: Document all parameters and remove default
1592 2018-09-25 Ilya Leoshkevich <iii@linux.ibm.com>
1595 * rtl.c (rtx_code_size): Take into account that EQ_ATTR_ALT
1596 contains HOST_WIDE_INTs when computing its size.
1598 2018-09-24 Jim Wilson <jimw@sifive.com>
1601 * config/riscv/riscv.h (STACK_BOUNDARY): Test riscv_abi == ABI_ILP32E
1603 (ABI_STACK_BOUNDARY, MAX_ARGS_IN_REGISTERS): Likewise.
1605 2018-09-24 Andrew Pinski <apinski@marvell.com>
1607 * config/aarch64/aarch64.c (aarch_macro_fusion_pair_p): Don't
1608 access prev before checking it for NULLness in the
1609 AARCH64_FUSE_CMP_BRANCH case.
1611 2018-09-24 H.J. Lu <hongjiu.lu@intel.com>
1614 * config/i386/i386.c (rest_of_insert_endbranch): Set
1615 endbr_queued_at_entrance to true and don't insert ENDBR if
1616 x86_function_profiler will be called.
1617 (x86_function_profiler): Insert ENDBR if endbr_queued_at_entrance
1619 * config/i386/i386.h (machine_function): Add
1620 endbr_queued_at_entrance.
1622 2018-09-24 Ilya Leoshkevich <iii@linux.ibm.com>
1624 * genattrtab.c (mk_attr_alt): Use alternative_mask.
1625 (attr_rtx_1): Adjust caching to match the new EQ_ATTR_ALT field
1627 (check_attr_test): Use alternative_mask.
1628 (get_attr_value): Likewise.
1629 (compute_alternative_mask): Use alternative_mask and XWINT.
1630 (make_alternative_compare): Use alternative_mask.
1631 (attr_alt_subset_p): Use XWINT.
1632 (attr_alt_subset_of_compl_p): Likewise.
1633 (attr_alt_intersection): Use alternative_mask and XWINT.
1634 (attr_alt_union): Likewise.
1635 (attr_alt_complement): Use HOST_WIDE_INT and XWINT.
1636 (mk_attr_alt): Use alternative_mask and HOST_WIDE_INT.
1637 (simplify_test_exp): Use alternative_mask and XWINT.
1638 (write_test_expr): Use alternative_mask and XWINT, adjust bit
1639 number calculation to support 64 bits. Generate code that
1640 checks 64-bit masks.
1641 (main): Use alternative_mask.
1642 * rtl.def (EQ_ATTR_ALT): Change field types from ii to ww.
1644 2018-09-24 Ilya Leoshkevich <iii@linux.ibm.com>
1647 * config/s390/s390.c (s390_emit_epilogue): Do not use PARALLEL
1648 RETURN+USE when returning via %r14.
1650 2018-09-24 Martin Liska <mliska@suse.cz>
1652 * gcov.c (output_lines): Print colorization legend
1653 for both flag_use_colors and flag_use_hotness_colors.
1656 2018-09-24 Martin Liska <mliska@suse.cz>
1658 * coverage.c (get_coverage_counts): Use warning_at
1659 with current_function_decl location. Use %qD in warning
1662 2018-09-24 Martin Liska <mliska@suse.cz>
1664 * memory-block.h (memory_block_pool::release): Annotate with
1665 valgrind that the memory is not accessible.
1667 2018-09-24 Martin Liska <mliska@suse.cz>
1670 * asan.c: Make asan_handled_variables extern.
1672 * cfgexpand.c (expand_stack_vars): Make sure
1673 a representative is unpoison if another
1674 variable in the partition is handled by
1675 use-after-scope sanitization.
1677 2018-09-24 Richard Biener <rguenther@suse.de>
1679 PR tree-optimization/63155
1680 * tree-ssa-propagate.c (add_ssa_edge): Avoid adding PHIs to
1681 the worklist when the edge of the respective argument isn't
1684 2018-09-23 Uros Bizjak <ubizjak@gmail.com>
1686 * config/i386/i386.h (enum reg_class): Rename MASK_REGS to
1687 ALL_MASK_REGS and MASK_EVEX_REGS to MASK_REGS.
1688 (MASK_CLASS_P): Update for rename.
1689 (MAYBE_MASK_CLASS_P): Ditto.
1690 (REG_CLASS_NAMES): Update.
1691 (REG_CLASS_CONTENT): Update.
1692 * config/i386/i386.c (regclass_map): Update for MASK_REG
1693 and ALL_MASK_REGS rename.
1694 * config/i386/constraints.md (Yk): Update for rename.
1697 2018-09-23 Uros Bizjak <ubizjak@gmail.com>
1699 * config/i386/i386.h (enum reg_class): Remove
1700 EVEX_SSE_REGS and MOD4_SSE_REGS.
1701 (REG_CLASS_NAMES): Update.
1702 (REG_CLASS_CONTENT): Update.
1703 * config/i386/i386.c (regclass_map): Declare AVX-512 SSE
1704 registers as ALL_SSE_REGS.
1705 (ix86_additional_allocno_class_p): Remove.
1706 (TARGET_ADDITIONAL_ALLOCNO_CLASS_P): Remove.
1707 (ix86_register_priority): Lower priority of EVEX SSE registers.
1708 Use IN_RANGE macro where appropriate.
1709 (ix86_hard_regno_mode_ok): Merge AVX-5124FMAPS and
1710 AVX-5124VNNIW checks.
1711 (ix86_modes_tieable_p): Tie 512-bit SSE modes.
1712 * config/i386/sse.md (avx5124fmaddps_4fmaddps)
1713 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddps_maskz)
1714 (avx5124fmaddps_4fmaddss, avx5124fmaddps_4fmaddss_mask)
1715 (avx5124fmaddps_4fmaddss_maskz, avx5124fmaddps_4fnmaddps)
1716 (avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddps_maskz)
1717 (avx5124fmaddps_4fnmaddss, avx5124fmaddps_4fnmaddss_mask)
1718 (avx5124fmaddps_4fnmaddss_maskz, avx5124vnniw_vp4dpwssd)
1719 (avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssd_maskz)
1720 (avx5124vnniw_vp4dpwssds, avx5124vnniw_vp4dpwssds_mask)
1721 (avx5124vnniw_vp4dpwssds_maskz): Use "v" instead of "Yh" constraint.
1722 * config/i386/constraints.md (Yh): Remove.
1724 2018-09-23 Uros Bizjak <ubizjak@gmail.com>
1726 * config/i386/i386.c (regclass_map): Declare integer REX registers
1729 2018-09-23 Gerald Pfeifer <gerald@pfeifer.com>
1731 * doc/service.texi (Service): Switch the fsf.org link to https.
1733 2018-09-22 Chung-Ju Wu <jasonwucj@gmail.com>
1736 * config/nds32/nds32.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
1737 Define to speculation_safe_value_not_needed.
1739 2018-09-21 Florian Weimer <fweimer@redhat.com>
1742 * doc/extend.texi (Common Function Attributes): Mention that
1743 noreturn suppresses tail call optimization.
1745 2018-09-21 Jeff Law <law@redhat.com>
1747 * gimple-ssa-evrp.c (evrp_dom_walker::cleanup): Call
1748 vr_values::cleanup_edges_and_switches.
1749 * tree-vrp.c (to_remove_edges, to_update_switch_stmts): Moved into
1751 (identify_jump_threads): Remove EDGE_IGNORE handling.
1752 (execute_vrp): Move handling of to_remove_edges and
1753 to_update_switch_stmts into vr_values class member functions.
1754 * tree-vrp.h (switch_update, to_remove_edges): Remove declarations.
1755 (to_update_switch_stmts): Likewise.
1756 * vr-values.c: Include cfghooks.h.
1757 (vr_values::vr_values): Initialize to_remove_edges and
1758 to_update_switch_stmts.
1759 (vr_values::~vr_values): Verify to_remove_edges and
1760 to_update_switch_stmts are empty.
1761 (vr_values::simplify_switch_using_ranges): Set EDGE_IGNORE as needed.
1762 (vr_values::cleanup_edges_and_switches): New member function.
1763 * vr-values.h (vr_values): Add cleanup_edges_and_switches member
1764 function. Add new data members.
1766 2018-09-21 David Malcolm <dmalcolm@redhat.com>
1768 PR tree-optimization/87309
1769 * dumpfile.c (dump_context::begin_scope): Filter the dump_loc
1770 calls with pflags and alt_flags.
1771 (selftest::test_capture_of_dump_calls): Add test of interaction of
1772 MSG_OPTIMIZED_LOCATIONS with AUTO_DUMP_SCOPE.
1774 2018-09-21 Olivier Hainque <hainque@adacore.com>
1776 * config.gcc: Factorize and comment inclusion of vxworks-dummy.h.
1778 2018-09-21 Olivier Hainque <hainque@adacore.com>
1780 * config/vxworks.h (CLEAR_INSN_CACHE): #define to 1.
1782 2018-09-21 Olivier Hainque <hainque@adacore.com>
1784 * config/vxworks.h (VXWORKS_LIBGCC_SPEC): Remove -lc_internal.
1785 Merge block comment with the one ahead of VXWORKS_LIBS_RTP. Then:
1786 (VXWORKS_LIBS_RTP): Minor reordering.
1788 2018-09-21 Olivier Hainque <hainque@adacore.com>
1790 * config/vxworks.h (STARTFILE_PREFIX_SPEC): Define.
1791 (VXWORKS_LIBS_DIR_RTP): Remove definition and use.
1793 2018-09-21 Olivier Hainque <hainque@adacore.com>
1795 * config/vxworks.h (SIZE_TYPE): Account for TARGET_VXWORKS64.
1796 (PTRDIFF_TYPE): Likewise.
1798 2018-09-21 Olivier Hainque <hainque@adacore.com>
1800 * config.gcc: Enforce def of TARGET_VXWORKS64 to 1 from
1801 triplet, similar to support for VxWorks7.
1802 * config/vxworks-dummy.h: Provide a default definition
1803 of TARGET_VXWORKS64 to 0.
1805 2018-09-21 Olivier Hainque <hainque@adacore.com>
1807 * config/vxworks.h (TARGET_VXWORKS7): Move default definition ...
1808 * config/vxworks-dummy.h: here.
1810 2018-09-21 Olivier Hainque <hainque@adacore.com>
1812 * config.gcc: Prepend vxworks-dummy.h to tm_file for powerpc*
1814 2018-09-21 Shaokun Zhang <zhangshaokun@hisilicon.com>
1815 Bo Zhou <zbo.zhou@hisilicon.com>
1817 * config/aarch64/aarch64-cores.def (tsv110): New CPU.
1818 * config/aarch64/aarch64-tune.md: Regenerated.
1819 * doc/invoke.texi (AArch64 Options/-mtune): Add "tsv110".
1820 * config/aarch64/aarch64.c (tsv110_tunings): New tuning table.
1821 * config/aarch64/aarch64-cost-tables.h: Add "tsv110" extra costs.
1823 2018-09-21 Andrew Stubbs <ams@codesourcery.com>
1824 Julian Brown <julian@codesourcery.com>
1826 * builtins.c (get_builtin_sync_mem): Handle address spaces.
1828 2018-09-21 Eric Botcazou <ebotcazou@adacore.com>
1830 * config/rs6000/rs6000.c (rs6000_function_ok_for_sibcall): Return false
1831 if the call takes a static chain.
1833 2018-09-21 Martin Liska <mliska@suse.cz>
1835 * auto-profile.c (autofdo_source_profile::read): Do not
1837 (read_profile): Do not add working sets.
1838 (read_autofdo_file): Remove sum_all.
1839 (afdo_callsite_hot_enough_for_early_inline): Remove const
1841 * coverage.c (struct counts_entry): Remove gcov_summary.
1842 (read_counts_file): Read new GCOV_TAG_OBJECT_SUMMARY,
1843 do not support GCOV_TAG_PROGRAM_SUMMARY.
1844 (get_coverage_counts): Remove summary and expected
1846 * coverage.h (get_coverage_counts): Likewise.
1847 * doc/gcov-dump.texi: Remove -w option.
1848 * gcov-dump.c (dump_working_sets): Remove.
1849 (main): Do not support '-w' option.
1850 (print_usage): Likewise.
1851 (tag_summary): Likewise.
1852 * gcov-io.c (gcov_write_summary): Do not dump
1854 (gcov_read_summary): Likewise.
1855 (gcov_histo_index): Remove.
1856 (gcov_histogram_merge): Likewise.
1857 (compute_working_sets): Likewise.
1858 * gcov-io.h (GCOV_TAG_OBJECT_SUMMARY): Mark
1860 (GCOV_TAG_PROGRAM_SUMMARY): Mark it obsolete.
1861 (GCOV_TAG_SUMMARY_LENGTH): Adjust.
1862 (GCOV_HISTOGRAM_SIZE): Remove.
1863 (GCOV_HISTOGRAM_BITVECTOR_SIZE): Likewise.
1864 (struct gcov_summary): Simplify rapidly just
1865 to runs and sum_max fields.
1866 (gcov_histo_index): Remove.
1867 (NUM_GCOV_WORKING_SETS): Likewise.
1868 (compute_working_sets): Likewise.
1869 * gcov-tool.c (print_overlap_usage_message): Remove
1870 trailing empty line.
1871 * gcov.c (read_count_file): Read GCOV_TAG_OBJECT_SUMMARY.
1872 (output_lines): Remove program related line.
1873 * ipa-profile.c (ipa_profile): Do not consider GCOV histogram.
1874 * lto-cgraph.c (output_profile_summary): Do not stream GCOV
1876 (input_profile_summary): Do not read it.
1877 (merge_profile_summaries): And do not merge it.
1878 (input_symtab): Do not call removed function.
1879 * modulo-sched.c (sms_schedule): Do not print sum_max.
1880 * params.def (HOT_BB_COUNT_FRACTION): Reincarnate param that was
1881 removed when histogram method was invented.
1882 (HOT_BB_COUNT_WS_PERMILLE): Mention that it's used only in LTO
1884 * postreload-gcse.c (eliminate_partially_redundant_load): Fix
1886 * predict.c (get_hot_bb_threshold): Use HOT_BB_COUNT_FRACTION
1887 and dump selected value.
1888 * profile.c (add_working_set): Remove.
1889 (get_working_sets): Likewise.
1890 (find_working_set): Likewise.
1891 (get_exec_counts): Do not work with working sets.
1892 (read_profile_edge_counts): Do not inform as sum_max is removed.
1893 (compute_branch_probabilities): Likewise.
1894 (compute_value_histograms): Remove argument for call of
1895 get_coverage_counts.
1896 * profile.h: Do not make gcov_summary const.
1898 2018-09-21 Monk Chiang <sh.chiang04@gmail.com>
1900 * config.gcc (nds32*-*-*): Set TARGET_DEFAULT_TLSDESC_TRAMPOLINE=0.
1902 2018-09-21 Eric Botcazou <ebotcazou@adacore.com>
1904 PR tree-optimization/86990
1905 * gimple-ssa-store-merging.c (imm_store_chain_info:coalesce_immediate):
1906 Check that the entire merged store group is made of constants only for
1909 2018-09-20 Allan Sandfeld Jensen <allan.jensen@qt.io>
1911 * gcc.c (LINK_COMMAND_SPEC): Handle -r like -nostdlib.
1912 (VTABLE_VERIFICATION_SPEC): Likewise.
1913 (SANITIZER_EARLY_SPEC): Likewise.
1914 (SANITIZER_SPEC): Likewise.
1915 * config/darwin.h (LINK_COMMAND_SPEC): Likewise.
1916 * doc/invoke.texi (Link Options): Document -r.
1918 2018-09-20 Richard Biener <rguenther@suse.de>
1921 * gimplify.c (gimplify_expr): Retain alignment of
1922 addressable lvalue in dereference.
1924 2018-09-20 Alexandre Oliva <aoliva@redhat.com>
1927 * configure.ac: Check for .loc is_stmt support.
1928 * configure, config.in: Rebuilt.
1929 * dwarf2out.c (dwarf2out_source_line): Skip is_stmt
1932 2018-09-20 Segher Boessenkool <segher@kernel.crashing.org>
1934 * config/rs6000/rs6000.opt (misel=no, misel=yes): Delete.
1935 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -misel=yes and
1938 2018-09-20 Segher Boessenkool <segher@kernel.crashing.org>
1940 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
1942 * config/rs6000/rs6000.c (rs6000_debug_vector_unit): Delete
1945 2018-09-20 Marek Polacek <polacek@redhat.com>
1947 * doc/invoke.texi: Add -Wno-init-list-lifetime to C++ Language Options.
1949 2018-09-20 Richard Sandiford <richard.sandiford@arm.com>
1951 PR tree-optimization/87288
1952 * tree-vect-loop.c (vect_analyze_loop_2): Take PEELING_FOR_GAPS
1953 into account when determining PEELING_FOR_NITERS.
1955 2018-09-20 Richard Sandiford <richard.sandiford@arm.com>
1957 PR tree-optimization/86877
1958 * tree-vect-loop.c (vect_analyze_loop_2): Call
1959 vect_verify_datarefs_alignment.
1961 2018-09-19 Marek Polacek <polacek@redhat.com>
1963 * doc/invoke.texi: Document -Wclass-conversion.
1965 2018-09-19 John David Anglin <danglin@gcc.gnu.org>
1967 * config/pa/pa.c (pa_adjust_priority): Delete.
1968 (TARGET_SCHED_ADJUST_PRIORITY): Delete define.
1970 * config/pa/pa.md (atomic_storeqi): Restore deleted expander.
1971 (atomic_storehi): Likewise.
1972 (atomic_storesi): Likewise.
1973 (atomic_loaddi): Restore compare and swap exchange loop code.
1975 2018-09-19 Segher Boessenkool <segher@kernel.crashing.org>
1977 PR rtl-optimization/86902
1978 * combine.c (try_combine): When changing the CC mode used, don't change
1979 an unrelated mode in other_insn to that new CC mode.
1981 2018-09-19 David Malcolm <dmalcolm@redhat.com>
1983 * tree-data-ref.c (runtime_alias_check_p): Use formatted printing
1984 with %T in place of calls to dump_generic_expr.
1985 (prune_runtime_alias_test_list): Likewise.
1986 (create_runtime_alias_checks): Likewise.
1987 * tree-vect-data-refs.c (vect_check_nonzero_value): Likewise.
1988 (vect_analyze_data_ref_dependence): Likewise.
1989 (vect_slp_analyze_data_ref_dependence): Likewise.
1990 (vect_record_base_alignment): Likewise. Use %G in place of call
1991 to dump_gimple_stmt.
1992 (vect_compute_data_ref_alignment): Likewise.
1993 (verify_data_ref_alignment): Likewise.
1994 (vect_find_same_alignment_drs): Likewise.
1995 (vect_analyze_group_access_1): Likewise.
1996 (vect_analyze_data_ref_accesses): Likewise.
1997 (dependence_distance_ge_vf): Likewise.
1998 (dump_lower_bound): Likewise.
1999 (vect_prune_runtime_alias_test_list): Likewise.
2000 (vect_find_stmt_data_reference): Likewise.
2001 (vect_analyze_data_refs): Likewise.
2002 (vect_create_addr_base_for_vector_ref): Likewise.
2003 (vect_create_data_ref_ptr): Likewise.
2004 * tree-vect-loop-manip.c (vect_set_loop_condition): Likewise.
2005 (vect_can_advance_ivs_p): Likewise.
2006 (vect_update_ivs_after_vectorizer): Likewise.
2007 (vect_gen_prolog_loop_niters): Likewise.
2008 (vect_prepare_for_masked_peels): Likewise.
2009 * tree-vect-loop.c (vect_determine_vf_for_stmt): Likewise.
2010 (vect_determine_vectorization_factor): Likewise.
2011 (vect_is_simple_iv_evolution): Likewise.
2012 (vect_analyze_scalar_cycles_1): Likewise.
2013 (vect_analyze_loop_operations): Likewise.
2014 (report_vect_op): Likewise.
2015 (vect_is_slp_reduction): Likewise.
2016 (check_reduction_path): Likewise.
2017 (vect_is_simple_reduction): Likewise.
2018 (vect_create_epilog_for_reduction): Likewise.
2019 (vect_finalize_reduction:): Likewise.
2020 (vectorizable_induction): Likewise.
2021 (vect_transform_loop_stmt): Likewise.
2022 (vect_transform_loop): Likewise.
2023 (optimize_mask_stores): Likewise.
2024 * tree-vect-patterns.c (vect_pattern_detected): Likewise.
2025 (vect_split_statement): Likewise.
2026 (vect_recog_over_widening_pattern): Likewise.
2027 (vect_recog_average_pattern): Likewise.
2028 (vect_determine_min_output_precision_1): Likewise.
2029 (vect_determine_precisions_from_range): Likewise.
2030 (vect_determine_precisions_from_users): Likewise.
2031 (vect_mark_pattern_stmts): Likewise.
2032 (vect_pattern_recog_1): Likewise.
2033 * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
2034 (vect_record_max_nunits): Likewise.
2035 (vect_build_slp_tree_1): Likewise.
2036 (vect_build_slp_tree_2): Likewise.
2037 (vect_print_slp_tree): Likewise.
2038 (vect_analyze_slp_instance): Likewise.
2039 (vect_detect_hybrid_slp_stmts): Likewise.
2040 (vect_detect_hybrid_slp_1): Likewise.
2041 (vect_slp_analyze_operations): Likewise.
2042 (vect_slp_analyze_bb_1): Likewise.
2043 (vect_transform_slp_perm_load): Likewise.
2044 (vect_schedule_slp_instance): Likewise.
2045 * tree-vect-stmts.c (vect_mark_relevant): Likewise.
2046 (vect_mark_stmts_to_be_vectorized): Likewise.
2047 (vect_init_vector_1): Likewise.
2048 (vect_get_vec_def_for_operand): Likewise.
2049 (vect_finish_stmt_generation_1): Likewise.
2050 (vect_check_load_store_mask): Likewise.
2051 (vectorizable_call): Likewise.
2052 (vectorizable_conversion): Likewise.
2053 (vectorizable_operation): Likewise.
2054 (vectorizable_load): Likewise.
2055 (vect_analyze_stmt): Likewise.
2056 (vect_is_simple_use): Likewise.
2057 (vect_get_vector_types_for_stmt): Likewise.
2058 (vect_get_mask_type_for_stmt): Likewise.
2059 * tree-vectorizer.c (increase_alignment): Likewise.
2061 2018-09-19 Andrew Stubbs <ams@codesourcery.com>
2063 * doc/rtl.texi: Adjust vec_select description.
2064 * simplify-rtx.c (simplify_binary_operation_1): Allow VEC_SELECT to use
2065 non-constant selectors.
2067 2018-09-19 Matthew Malcomson <matthew.malcomson@arm.com>
2069 * config/aarch64/aarch64-protos.h
2070 (aarch64_offset_9bit_signed_unscaled_p): New declaration.
2071 * config/aarch64/aarch64.md (arches): New "rcpc8_4" attribute value.
2072 (arch_enabled): Add check for "rcpc8_4" attribute value of "arch".
2073 * config/aarch64/aarch64.h (AARCH64_FL_RCPC8_4): New bitfield.
2074 (AARCH64_FL_FOR_ARCH8_4): Include AARCH64_FL_RCPC8_4.
2075 (AARCH64_FL_PROFILE): Move index so flags are ordered.
2076 (AARCH64_ISA_RCPC8_4): New flag.
2077 * config/aarch64/aarch64.c (offset_9bit_signed_unscaled_p): Renamed
2078 to aarch64_offset_9bit_signed_unscaled_p.
2079 * config/aarch64/atomics.md (atomic_store<mode>): Allow offset
2081 * config/aarch64/constraints.md (Ust): New constraint.
2082 * config/aarch64/predicates.md.
2083 (aarch64_9bit_offset_memory_operand): New predicate.
2084 (aarch64_rcpc_memory_operand): New predicate.
2086 2018-09-19 Eric Botcazou <ebotcazou@adacore.com>
2088 PR rtl-optimization/87361
2089 * rtlanal.c (nonzero_bits1): Revert accidental change.
2091 2018-09-19 Richard Biener <rguenther@suse.de>
2093 PR tree-optimization/87349
2094 PR tree-optimization/87342
2095 * tree-ssa-sccvn.c (do_rpo_vn): Iterate max_rpo computation.
2097 2018-09-18 Marek Polacek <polacek@redhat.com>
2099 P1064R0 - Allowing Virtual Function Calls in Constant Expressions
2100 * gimple-fold.c (gimple_get_virt_method_for_vtable): Adjust assert.
2102 2018-09-18 Segher Boessenkool <segher@kernel.crashing.org>
2104 * config/rs6000/rs6000.md: Remove old "Cygnus sibcall" comment.
2106 2018-09-18 Segher Boessenkool <segher@kernel.crashing.org>
2108 PR rtl-optimization/86882
2109 * rtlanal.c (reg_overlap_mentioned_p): Handle CLOBBER.
2111 2018-09-18 Uros Bizjak <ubizjak@gmail.com>
2113 * config/i386/i386.md (*<code>extend<mode>xf2): Macroize insn from
2114 *<code>extendsfxf2 and *<code>extenddfxf2 using MODEF mode iterator.
2116 2018-09-18 Jonathan Wakely <jwakely@redhat.com>
2119 * doc/invoke.texi (Link Options): Fix formatting and grammar.
2121 2018-09-18 Richard Biener <rguenther@suse.de>
2124 * tree-ssa-coalesce.c (tree_int_map_hasher): Remove.
2125 (compute_samebase_partition_bases): Likewise.
2126 (coalesce_ssa_name): Always use compute_optimized_partition_bases.
2127 (gimple_can_coalesce_p): Simplify.
2129 2018-09-18 Hans-Peter Nilsson <hp@bitrange.com>
2131 Handle a library implementation of ffs calling __builtin_ffs.
2132 * config/mmix/mmix.c (TARGET_INIT_LIBFUNCS): Override with...
2133 (mmix_init_libfuncs): New function: make __builtin_ffs expand
2136 2018-09-17 David Malcolm <dmalcolm@redhat.com>
2138 * diagnostic-show-locus.c (class layout_range): Add field
2140 (layout_range::layout_range): Add "original_idx" param and use it
2141 to initialize new field.
2142 (make_range): Use 0 for original_idx.
2143 (layout::layout): Pass in index to calls to
2144 maybe_add_location_range.
2145 (layout::maybe_add_location_range): Add param "original_idx" and
2146 pass it on to layout_range.
2147 (layout::print_any_labels): Pass on range->m_original_idx to
2149 (gcc_rich_location::add_location_if_nearby): Use 0 for
2151 * gcc-rich-location.h (text_range_label::get_text): Update for new
2153 (range_label_for_type_mismatch::get_text): Likewise.
2155 2018-09-17 Uros Bizjak <ubizjak@gmail.com>
2157 * config/i386/i386.c (ix86_emit_i387_log1p): Emit fldln2 earlier.
2159 2018-09-17 David Malcolm <dmalcolm@redhat.com>
2161 * gimple-ssa-sprintf.c (fmtwarn): Update for introduction of
2162 format_string_diagnostic_t.
2163 (fmtwarn_n): Likewise.
2164 * substring-locations.c
2165 (format_string_diagnostic_t::format_string_diagnostic_t) New ctor.
2166 (format_warning_n_va): Convert to...
2167 (format_string_diagnostic_t::emit_warning_n_va): ...this.
2168 (format_warning_va): Convert to...
2169 (format_string_diagnostic_t::emit_warning_va): ...this.
2170 (format_warning_at_substring): Convert to...
2171 (format_string_diagnostic_t::emit_warning): ...this.
2172 (format_warning_at_substring_n): Convert to...
2173 (format_string_diagnostic_t::emit_warning_n): ...this.
2174 * substring-locations.h (class format_string_diagnostic_t): New
2176 (format_warning_va): Convert to
2177 format_string_diagnostic_t::emit_warning_va.
2178 (format_warning_n_va): Convert to
2179 format_string_diagnostic_t::emit_warning_n_va.
2180 (format_warning_at_substring): Convert to
2181 format_string_diagnostic_t::emit_warning.
2182 (format_warning_at_substring_n): Convert to
2183 format_string_diagnostic_t::emit_warning_n.
2185 2018-09-17 Cesar Philippidis <cesar@codesourcery.com>
2186 Bernd Schmidt <bernds_cb1@t-online.de>
2188 * config/nvptx/nvptx.md (atomic_fetch_<logic><mode>): Enable with
2191 2018-09-17 Uros Bizjak <ubizjak@gmail.com>
2193 * config/i386/i386.md (truncxf<mode>2_i387_noop_unspec): Change
2194 operand 0 predicate to nonimmediate operand.
2195 (rint<mode>2_frndint): Remove insn pattern.
2196 (rint<mode>2): Change operand 1 predicate to general_operand.
2197 Extend operand 1 to XFmode and generate rintxf2 insn.
2198 (frndintxf2_<rounding>): Rename from frndint<mode>2_<rounding>.
2199 Do not use X87MODEF mode macro.
2200 (frndintxf2_<rounding>_i387): Rename from
2201 frndint<mode>2_<rounding>_i387. Do not use X87MODEF mode macro.
2202 (<rounding_insn><mode>2): For non-SSE modes, extend operand 1
2203 to XFmode and generate significandxf3 insn.
2205 2018-09-17 Richard Biener <rguenther@suse.de>
2207 PR tree-optimization/87328
2208 * tree-ssa-sccvn.c (process_bb): Remove assertion about not
2209 visiting unexecutable backedges when not iterating.
2210 (do_rpo_vn): Mark all edges not executable even when not
2213 2018-09-17 Martin Jambor <mjambor@suse.cz>
2216 * doc/invoke.texi (Warning Options): Likewise.
2218 2018-09-17 Richard Biener <rguenther@suse.de>
2220 PR tree-optimization/87301
2221 * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_cleanup): Properly
2222 clean EH info from leftover copy assignments.
2224 2018-09-17 Martin Liska <mliska@suse.cz>
2226 PR gcov-profile/85871
2227 * gcov.c (output_intermediate_file): Fix out of bounds
2230 2018-09-17 Vineet Gupta <vgupta@synopsys.com>
2232 * config/arc/arc.c: Object attributes for core4 not reflected
2234 * config/arc/arc.h: Don't restrict DBNZ to core3 (core4 includes
2237 2018-09-17 Alexey Brodkin <abrodkin@synopsys.com>
2239 * config/arc/linux.h (LINK_EH_SPEC): Add missing space.
2241 2018-09-17 Martin Liska <mliska@suse.cz>
2243 * doc/gcov.texi: Document new option --use-hotness-colors.
2244 * gcov.c (struct source_info): Declare new field.
2245 (source_info::source_info): Set default for maximum_count.
2246 (print_usage): Add new -q option.
2247 (process_args): Process it.
2248 (accumulate_line_info): Save src->maximum_count.
2249 (output_line_beginning): Make color line number if
2250 flag_use_hotness_colors is set.
2251 (output_line_details): Pass default argument value.
2252 (output_lines): Pass src->maximum_count.
2254 2018-09-17 Martin Liska <mliska@suse.cz>
2256 * common/config/i386/i386-common.c (ix86_get_valid_option_values):
2257 Use processor_names table.
2258 * config/i386/i386.c (ix86_default_align): Use
2259 processor_cost_table for alignment values.
2260 (ix86_option_override_internal): Use processor_names.
2261 (ix86_function_specific_print): Likewise.
2262 * config/i386/i386.h (struct processor_costs):
2263 Add alignment values.
2264 (struct ptt): Remove and replace with const char *.
2265 * config/i386/x86-tune-costs.h (struct processor_costs):
2266 Declare default alignments for all costs.
2268 2018-09-17 Aldy Hernandez <aldyh@redhat.com>
2270 * tree-vrp.c (extract_range_from_unary_expr): Do not special case
2271 symbolics or VR_VARYING ranges for ABS_EXPR.
2272 * wide-int-range.cc (wide_int_range_abs): Return positive numbers
2273 when range will wrap.
2275 2018-09-15 Eric Botcazou <ebotcazou@adacore.com>
2278 * cfgexpand.c (expand_gimple_basic_block): Be prepared for a BARRIER
2279 before and after a JUMP_TABLE_DATA.
2281 2018-09-14 John David Anglin <danglin@gcc.gnu.org>
2284 * dojump.c (do_compare_and_jump): Canonicalize function pointers
2285 when one operand is a function pointer. Use POINTER_TYPE_P and
2286 FUNC_OR_METHOD_TYPE_P.
2287 * expr.c (do_store_flag): Use POINTER_TYPE_P and FUNC_OR_METHOD_TYPE_P.
2288 * fold-const.c (build_range_check): Likewise.
2289 * match.pd (simple_comparison): Likewise.
2291 2018-09-14 David Malcolm <dmalcolm@redhat.com>
2294 * spellcheck.c (get_edit_distance_cutoff): New function.
2295 (selftest::test_edit_distance_unit_test_oneway): Rename to...
2296 (selftest::test_get_edit_distance_one_way): ...this.
2297 (selftest::test_get_edit_distance_unit): Rename to...
2298 (selftest::test_get_edit_distance_both_ways): ...this.
2299 (selftest::test_edit_distances): Move tests to this new function,
2300 and test some more pairs of strings. Update for above renaming.
2301 (selftest::get_old_cutoff): New function.
2302 (selftest::test_get_edit_distance_cutoff): New function.
2303 (selftest::assert_suggested_for): New function.
2304 (ASSERT_SUGGESTED_FOR): New macro.
2305 (selftest::assert_not_suggested_for): New function.
2306 (ASSERT_NOT_SUGGESTED_FOR): New macro.
2307 (selftest::test_suggestions): New function.
2308 (selftest::spellcheck_c_tests): Move test_get_edit_distance_unit
2309 tests to selftest::test_edit_distances and call it. Add calls to
2310 selftest::test_get_edit_distance_cutoff and
2311 selftest::test_suggestions.
2312 * spellcheck.h (get_edit_distance_cutoff): New function declaration.
2313 (best_match::consider): Replace hard-coded cutoff calculation with
2315 (best_match::get_cutoff): New declaration.
2316 (best_match::get_best_meaningful_candidate): Likewise.
2318 2018-09-14 Bernd Edlinger <bernd.edlinger@hotmail.de>
2320 * builtins.c (fold_builtin_strlen): Remove TODO comment.
2322 2018-09-14 Bernd Edlinger <bernd.edlinger@hotmail.de>
2325 2018-07-30 Bernd Edlinger <bernd.edlinger@hotmail.de>
2327 * tree-ssa-forwprop.c (simplify_builtin_call): Don't create a not NUL
2328 terminated string literal.
2330 2018-09-14 Martin Sebor <msebor@redhat.com>
2332 * builtins.c (unterminated_array): Handle ARRAY_REF.
2333 (expand_builtin_stpcpy_1): Detect unterminated char arrays.
2334 * builtins.h (unterminated_array): Declare extern.
2335 * gimple-fold.c (gimple_fold_builtin_stpcpy): Detect unterminated
2337 (gimple_fold_builtin_sprintf): Propagate NO_WARNING to transformed
2340 2018-09-14 Martin Sebor <msebor@redhat.com>
2341 Jeff Law <law@redhat.com>
2343 * builtins.c (unterminated_array): New.
2344 (expand_builtin_strcpy): Adjust.
2345 (expand_builtin_strcpy_args): Detect unterminated arrays.
2346 * gimple-fold.c (get_maxval_strlen): Add argument. Detect
2347 unterminated arrays.
2348 * gimple-fold.h (get_maxval_strlen): Add argument.
2349 (gimple_fold_builtin_strcpy): Detec unterminated arrays.
2351 * gimple-fold.c (get_range_strlen): Add argument.
2352 (get_maxval_strlen): Adjust.
2353 * gimple-fold.h (get_range_strlen): Add argument.
2355 2018-09-14 Wei Xiao <wei3.xiao@intel.com>
2357 * config/i386/movdirintrin.h: Fix copyright year.
2359 2018-09-14 Uros Bizjak <ubizjak@gmail.com>
2361 * reg-stack.c: Include regs.h.
2362 (replace_reg): Assert that mode is MODE_FLOAT or MODE_COMPLEX_FLOAT.
2363 (emit_pop_insn): Default pop insn mode to the reg_raw_mode of
2364 FIRST_STACK_REG, not DFmode.
2365 (emit_swap_insn): Default swap insn mode to the reg_raw_mode of
2366 FIRST_STACK_REG, not XFmode. Explicitly construct swap RTX.
2367 (change stack): Default register mode to the reg_raw_mode of
2368 FIRST_STACK_REG, not DFmode.
2369 * config/i386/i386.md (*swap<mode>): Remove insn pattern.
2370 (*swapxf): Rename from swapxf.
2372 2018-09-14 Carl Love <cel@us.ibm.com>
2374 * config/rs6000/emmintrin.h: Add _MM_SHUFFLE2.
2375 * config/rs6000/xmmintrin.h: Add _MM_SHUFFLE.
2377 2018-09-14 Segher Boessenkool <segher@kernel.crashing.org>
2380 * config/rs6000/rs6000.md (*mov<mode>_hardfloat64): Add Z to the Y
2383 2018-09-14 Sam Tebbs <sam.tebbs@arm.com>
2386 * config/aarch64/aarch64.md (*aarch64_bfxilsi_uxtw): Define.
2388 2018-09-14 Jason Merrill <jason@redhat.com>
2390 Fix --enable-gather-detailed-mem-stats.
2391 * hash-table.c (hash_table_usage): Change from variable to function.
2392 * hash-table.h: Adjust.
2393 * Makefile.in: Add missing dependencies on hash-table.h.
2395 2018-09-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2397 PR tree-optimization/87259
2399 (pass_cse_reciprocals::execute): Run optimize_recip_sqrt after
2400 execute_cse_reciprocals_1 has tried transforming.
2402 2018-09-14 Aldy Hernandez <aldyh@redhat.com>
2404 * tree-vrp.c (extract_range_from_binary_expr_1): Normalize
2405 VR_VARYING for PLUS/MINUS_EXPR.
2407 2018-09-14 Ilya Leoshkevich <iii@linux.ibm.com>
2409 * config/s390/s390-passes.def (INSERT_PASS_BEFORE): Improve
2412 2018-09-14 Richard Biener <rguenther@suse.de>
2415 * tree-ssa-coalesce.c (ssa_conflicts_merge): Remove conflict
2416 bits for the merged partition.
2418 2018-09-13 Martin Sebor <msebor@redhat.com>
2419 Bernd Edlinger <bernd.edlinger@hotmail.de>
2421 * builtins.h (c_srlen): Add argument.
2422 * builtins.c (warn_string_no_nul): New function.
2423 (c_strlen): Add argument and use it. Update recursive calls.
2424 Pass DECL argument to string_constant to get info on non
2425 terminated strings. Update *NONSTR as needed.
2426 (fold_builtin_strlen): Add argument to calls to c_strlen.
2427 Warn for unterminated arrays.
2428 (warn_string_no_null): Add prototype.
2429 * expr.c (string_constant): Update arguments. Update recursive
2430 calls appropriately. Detect missing NUL terminator and outermost
2431 declaration its missing in.
2432 Improve checks for arrays with nonzero lower bound or elements
2433 that are not a single byte. Simplify offset computation.
2434 Simplify checks for non-NUL terminated strings.
2435 * gimple-fold.c (get_range_strlen): Add argument to c_strlen call.
2436 * gimple-ssa-sprintf.c (get_string_length): Remove unnecessary code.
2438 2018-09-13 Bernd Edlinger <bernd.edlinger@hotmail.de>
2440 * builtins.c (c_strlen): Handle not zero terminated STRING_CSTs
2442 * fold-const.c (c_getstr): Fix function comment. Remove unused third
2443 argument. Fix range checks.
2444 * fold-const.h (c_getstr): Adjust protoype.
2445 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid folding when
2446 string is constant but contains no NUL byte.
2448 * expr.c (string_constant): Adjust function comment.
2449 Remove bogus check for zero termination.
2451 * fold-const.c (c_getstr): Clamp STRING_LENGTH to STRING_SIZE.
2453 * varasm.c (compare_constant): Compare type size of STRING_CSTs.
2454 (get_constant_size): Don't make STRING_CSTs larger than they are.
2455 (check_string_literal): New check function for STRING_CSTs.
2456 (output_constant): Use it.
2458 2018-09-13 Eric Botcazou <ebotcazou@adacore.com>
2461 * config/visium/visium.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define.
2463 2018-09-13 Richard Biener <rguenther@suse.de>
2465 PR tree-optimization/87263
2466 * tree-ssa-sccvn.c (visit_phi): Revert some earlier changes.
2467 (struct unwind_state): Add max_rpo field.
2468 (do_rpo_vn): Allow up-to-date loop state to be used when not iterating.
2469 Compute max_rpo, the max RPO number a block can be backwards reached
2470 from. Re-write non-iterating mode to a RPO ordered worklist approach,
2471 separating it from the iterating mode.
2473 2018-09-13 Vlad Lazar <vlad.lazar@arm.com>
2475 * haifa-sched.c (rank_for_schedule): Schedule by INSN_COST.
2476 (rfs_decision): New scheduling decision.
2478 2018-09-13 Richard Biener <rguenther@suse.de>
2481 * tree-ssa-sccvn.c (vn_nary_op_insert_into): Fix assert.
2482 (vn_nary_op_insert_pieces_predicated): Do not write useless
2483 valid_dominated_by_p entry outside of the allocated storage.
2485 2018-09-13 Omar Sandoval <osandov@osandov.com>
2486 Tom de Vries <tdevries@suse.de>
2489 * dwarf2out.c (is_c): New function.
2490 (add_subscript_info): Add DW_AT_count of 0 for C zero-length arrays.
2492 2018-09-13 Sam Tebbs <sam.tebbs@arm.com>
2495 * config/aarch64/aarch64.md (*aarch64_bfxil):
2497 * config/aarch64/constraints.md (Ulc): Define.
2498 * config/aarch64/aarch64-protos.h (aarch64_high_bits_all_ones_p):
2500 * config/aarch64/aarch64.c (aarch64_high_bits_all_ones_p):
2503 2018-09-13 Vlad Lazar <vlad.lazar@arm.com>
2505 * config/aarch64/aarch64.h (TARGET_COMPUTE_FRAME_LAYOUT): Define.
2506 * config/aarch64/aarch64.c (aarch64_expand_prologue): Remove
2507 aarch64_layout_frame call.
2508 (aarch64_expand_epilogue): Likewise.
2509 (aarch64_initial_elimination_offset): Likewise.
2510 (aarch64_get_separate_components): Likewise.
2511 (aarch64_use_return_insn_p): Likewise.
2512 (aarch64_layout_frame): Remove unneeded check.
2514 2018-09-13 Jakub Jelinek <jakub@redhat.com>
2516 * configure.ac: Only append
2517 " : (reconfigured) $TOPLEVEL_CONFIGURE_ARGUMENTS" to
2518 gcc_config_arguments if it was never reconfigured or last reconfigure
2519 was with different arguments.
2520 * configure: Regenerated.
2522 2018-09-13 Jakub Jelinek <jakub@redhat.com>
2523 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2526 * expr.c (maybe_optimize_pow2p_mod_cmp): New function.
2527 (maybe_optimize_mod_cmp): Use it if integer_pow2p treeop1.
2529 2018-09-13 Jakub Jelinek <jakub@redhat.com>
2531 PR tree-optimization/87287
2532 * fold-const.c (fold_binary_loc) <case EQ_EXPR>: Move signed modulo
2533 X % C == 0 to X % (unsigned) C == 0 optimization to ...
2534 * match.pd (X % C == 0): ... here. New optimization.
2536 2018-09-12 Jakub Jelinek <jakub@redhat.com>
2539 * expr.h (maybe_optimize_mod_cmp): Declare.
2540 * expr.c (mod_inv): New function.
2541 (maybe_optimize_mod_cmp): New function.
2542 (do_store_flag): Use it.
2543 * cfgexpand.c (expand_gimple_cond): Likewise.
2545 2018-09-09 Cesar Philippidis <cesar@codesourcery.com>
2546 Julian Brown <julian@codesourcery.com>
2549 * gimplify.c (gimplify_scan_omp_clauses): Set
2550 target_firstprivatize_array_bases in OpenACC parallel and kernels
2551 region contexts. Remove GOMP_MAP_FIRSTPRIVATE_REFERENCE clauses from
2552 OpenACC data regions.
2554 2018-09-12 Uros Bizjak <ubizjak@gmail.com>
2556 * config/i386/i386.md (sqrt_extend<mode>xf3_i387): Remove.
2557 (sqrt<mode>2): Extend operand 1 to XFmode and generate sqrtxf3 insn.
2559 2018-09-12 Richard Biener <rguenther@suse.de>
2561 PR tree-optimization/87280
2562 * tree-ssa-sccvn.c (process_bb): Handle the case of executable
2563 edge but unreachable target.
2564 (do_rpo_vn): For conservatively handling a PHI only mark
2565 the backedge executable but not the block reachable.
2567 2018-09-12 Richard Biener <rguenther@suse.de>
2569 PR tree-optimization/87266
2570 * tree-ssa-sccvn.c (do_rpo_vn): Always iterate to not yet
2573 2018-09-12 Andreas Krebbel <krebbel@linux.ibm.com>
2575 * config/s390/s390.md (PFPO_RND_MODE_DFP, PFPO_RND_MODE_BFP): New
2577 ("trunc<BFP:mode><DFP_ALL:mode>2")
2578 ("trunc<DFP_ALL:mode><BFP:mode>2")
2579 ("extend<BFP:mode><DFP_ALL:mode>2")
2580 ("extend<DFP_ALL:mode><BFP:mode>2"): Set proper rounding mode
2581 according to the target operand type.
2583 2018-09-12 Jakub Jelinek <jakub@redhat.com>
2584 Andreas Krebbel <krebbel@linux.ibm.com>
2586 PR tree-optimization/86844
2587 * gimple-ssa-store-merging.c
2588 (imm_store_chain_info::coalesce_immediate): For overlapping stores, if
2589 there are any overlapping stores in between them, make sure they are
2590 also coalesced or we give up completely.
2592 2018-09-12 Jakub Jelinek <jakub@redhat.com>
2595 * fold-const.c (fold_ternary_loc) <case COND_EXPR>: Verify also that
2596 BIT_AND_EXPR's second operand is a power of two. Formatting fix.
2598 2018-09-12 Tom de Vries <tdevries@suse.de>
2600 * common.opt (gdescribe-dies): Add option.
2601 * dwarf2out.c (add_name_and_src_coords_attributes): Add description
2602 attribute for artifical and nameless decls.
2603 (dwarf2out_register_external_die): Add description attribute to
2604 external reference die.
2605 (add_desc_attribute): New functions.
2606 (gen_subprogram_die): Add description attribute to
2607 DW_TAG_call_site_parameter.
2608 * tree-pretty-print.c (print_generic_expr_to_str): New function.
2609 * tree-pretty-print.h (print_generic_expr_to_str): Declare.
2610 * doc/invoke.texi (@item Debugging Options): Add -gdescribe-dies and
2612 (@item -gdescribe-dies): Add.
2614 2018-09-12 Aldy Hernandez <aldyh@redhat.com>
2616 * tree-vrp.c (vrp_shift_undefined_p): Remove.
2617 (extract_range_from_binary_expr_1: Call
2618 wide_int_range_shift_undefined_p instead of vrp_shift_undefined_p.
2619 * wide-int-range.h (wide_int_range_shift_undefined_p): Do not
2622 2018-09-12 Aldy Hernandez <aldyh@redhat.com>
2624 * gimple-ssa-warn-alloca.c
2625 (alloca_type_and_limit::alloca_type_and_limit): Initialize limit
2626 field for ALLOCA_BOUND_*_LARGE.
2628 2018-09-11 Nathan Sidwell <nathan@acm.org>
2630 * gcc.c (load_specs, execute, run_attempt): Use %qs not '%s'.
2632 2018-09-11 Uros Bizjak <ubizjak@gmail.com>
2634 * reg-stack.c (subst_asm_stack_regs): Call replace_reg also
2635 for clobbers. Remove obsolete comment.
2637 2018-09-11 Uros Bizjak <ubizjak@gmail.com>
2639 * config/i386/i386.md (define_attr "type"): Remove mpxmov, mpxmk,
2640 mpxchk, mpxld and mpxst types.
2641 (define_attr length_immediate): Remove all processing of mpx types.
2642 (define_attr prefix_0f): Ditto.
2643 (define_attr memory): Ditto.
2645 2018-09-11 Uros Bizjak <ubizjak@gmail.com>
2647 * config/i386/i386.md (fyl2x_extend<mode>xf3_i387): Remove.
2648 (log<mode>2): Change operand 1 predicate to general_operand.
2649 Extend operand 1 to XFmode and generate logxf3 insn.
2650 (log10<mode>2): Change operand 1 predicate to general_operand.
2651 Extend operand 1 to XFmode and generate log10xf3 insn.
2652 (log2<mode>2): Change operand 1 predicate to general_operand.
2653 Extend operand 1 to XFmode and generate log2xf3 insn.
2654 (fyl2xp1_extend<mode>xf3_i387): Remove.
2655 (log1p<mode>2): Change operand 1 predicate to general_operand.
2656 Extend operand 1 to XFmode and generate log1pxf3 insn.
2657 (fxtract_extend<mode>xf3_i387): Remove.
2658 (logb<mode>2): Change operand 1 predicate to general_operand.
2659 Extend operand 1 to XFmode and generate logbxf3 insn.
2660 (ilogb<mode>2): Change operand 1 predicate to general_operand.
2661 Extend operand 1 to XFmode and generate fxtractxf3_i387 insn.
2662 (significand<mode>2): Change operand 1 predicate to general_operand.
2663 Extend operand 1 to XFmode and generate significandxf3 insn.
2665 2018-09-11 Nathan Sidwell <nathan@acm.org>
2667 * gcc.c (perror_with_name, pfatal_with_name): Delete.
2668 (load_specs): Use fatal_error.
2669 (DELETE_IF_ORDINARY, process_command): Use error.
2670 (execute, run_attempt): Use fatal_error.
2672 2018-09-11 Andrew Stubbs <ams@codesourcery.com>
2674 * diagnostic-core.h (sorry_at): New prototype.
2675 * diagnostic.c (sorry_at): New function.
2677 2018-09-11 Aldy Hernandez <aldyh@redhat.com>
2679 * tree-vrp.c (extract_range_from_binary_expr_1): Treat all divisions
2680 by zero as VR_UNDEFINED.
2682 2018-09-10 Uros Bizjak <ubizjak@gmail.com>
2684 * config/i386/i386.md (<sincos>xf2): Rename from *<sincos>xf2_i387.
2685 (*<sincos>_extend<mode>xf2_i387): Remove insn pattern.
2686 (<sincos>mode2): New expander.
2687 (sincos_extend<mode>xf3_i387): Remove insn pattern.
2688 (sincos -> sin, cos splitters): Remove splitter patterns.
2689 (sincos<mode>3): Change operand 2 predicate to general_operand.
2690 Extend operand 2 to XFmode and generate sincosxf3 insn.
2691 (fptanxf4_i387): Change mode of operands 0 and 3 to SFmode.
2692 Change operand 3 predicate to const1_operand.
2693 (fptan_extend<mode>xf4_i387): Remove insn pattern.
2694 (tanxf2): Update operands in the call to fptanxf4_i387.
2695 (tan<mode>2): Change operand 1 predicate to general_operand.
2696 Extend operand 1 to XFmode and generate tanxf3 insn.
2697 (atan2xf3): Rename from *fpatanxf3_i387.
2698 (fpatan_extend<mode>xf3_i387): Remove insn pattern.
2699 (atan2xf3): Remove expander.
2700 (atan2<mode<3): Change operand 1 and 2 predicates to general_operand.
2701 Extend operands 1 and 2 to XFmode and generate atan2xf3 insn.
2702 (atan<mode>2): Change operand 1 predicate to general_operand.
2703 Extend operand 1 to XFmode and generate atanxf3 insn.
2705 2018-09-10 Uros Bizjak <ubizjak@gmail.com>
2707 * config/i386/i386.md (x87/SSE constant load splitter): Use
2708 memory_operand instead of nonimmediate_operand for input operand
2711 2018-09-09 Uros Bizjak <ubizjak@gmail.com>
2713 * config/i386/i386.md (float partial SSE register stall splitter): Move
2714 splitter near its instruction pattern.
2715 (float_extend partial SSE register stall splitter): Ditto.
2716 (float_truncate partial SSE register stall splitter): Ditto.
2718 2018-09-09 Hans-Peter Nilsson <hp@bitrange.com>
2721 * config/mmix/mmix.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
2722 to speculation_safe_value_not_needed.
2725 * config/mmix/mmix.c (mmix_assemble_integer): Handle byte-size
2726 non-CONST_INT rtx:es using assemble_integer_with_op ".byte".
2727 (MMIX_CFUN_NEEDS_SAVED_EH_RETURN_ADDRESS): Don't call
2728 leaf_function_p, instead use has_hard_reg_initial_val.
2730 2018-09-09 Nathan Sidwell <nathan@acm.org>
2732 * gcc.h (pfatal_with_name): Don't declare here.
2733 * gcc.c (pfatal_with_name): Make static.
2735 2018-09-09 Xianmiao Qu <xianmiao_qu@c-sky.com>
2737 * config/csky/csky.md (*cskyv2_adddi3, *cskyv2_subdi3): Add
2740 2018-09-08 John David Anglin <danglin@gcc.gnu.org>
2742 PR rtl-optimization/85458
2743 * sel-sched.c (sel_target_adjust_priority): Allow backend adjust
2744 priority hook to reduce the priority of EXPR.
2746 2018-09-07 Uros Bizjak <ubizjak@gmail.com>
2748 * config/i386/i386.md (float<SWI48x:mode><MODEF:mode>2) Enable
2749 DImode for x87 on 32bit targets. Conditionally disable x87 modes
2750 with X87_ENABLE_FLOAT. Remove preparation code.
2751 (*float<SWI48:mode><MODEF:mode>2): Rename from
2752 *float<SWI48:mode><MODEF:mode>2_mixed. Handle x87, SSE and mixed
2753 math using "enabled" attribute.
2754 (*floatdi<MODEF:mode>2_i387): Rename from
2755 *float<SWI48x:mode><MODEF:mode>2_i387. Handle only DImode and
2756 enable for 32bit targets only.
2757 (floatdi<X87MODEF:mode>2_i387_with_xmm pre-reload splitter): New
2759 (floatdi<X87MODEF:mode>2_i387_with_xmm): Use register_operand
2760 as operand 1 predicate. Rewrite as define_insn_and_split.
2761 (floatdi<X87MODEF:mode>2_i387_with_xmm memory input splitter): Remove.
2763 2018-09-06 Uros Bizjak <ubizjak@gmail.com>
2765 * reg-stack.c (get_true_reg) <case FLOAT_TRUNCATE>: Reorder
2766 to fallthru to FLOAT case.
2768 2018-09-06 Will Schmidt <will_schmidt@vnet.ibm.com>
2771 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Update logic
2772 around folding of vec_sl to handle out of range shift values.
2774 2018-09-06 Uros Bizjak <ubizjak@gmail.com>
2776 * config/i386/i386.md (fix_trunc<mode>_fisttp_i387_1): Remove.
2777 Update callers to gen_fix_trunc<mode>_i387_fisttp
2778 (fix_trunc<mode>_i387_fisttp): Change operand 0 predicate to
2779 nonimmediate_operand.
2780 (fix_trunc<mode>_i387_fisttp_with_temp): Remove insn pattern
2781 and corresponding splitters.
2782 (*fix_trunc<mode>_i387_1): Always expand with fix_trunc<mode>_i387.
2783 (fix_truncdi_i387): Change operand 0 predicate to nonimmediate_operand.
2784 (fix_truncdi_i387_with_temp): Remove insn pattern
2785 and corresponding splitters.
2786 (fix_trunc<mode>_i387): Change operand 0 predicate to
2787 nonimmediate_operand.
2788 (fix_trunc<mode>_i387_with_temp): Remove insn pattern
2789 and corresponding splitters.
2790 (*fistdi2_1): Remove.
2792 (fistdi2_with_temp): Remove insn pattern and corresponding splitters.
2793 (lrintxfdi2): Remove expander. Reimplement as define_insn.
2794 (*fist<mode>2_1): Remove.
2795 (fist<mode>2): Ditto.
2796 (fist<mode>2_with_temp): Remove insn pattern and corresponding
2798 (lrintxf<mode>2): Remove expander. Reimplement as define_insn.
2799 (*fist<mode>2_<rounding>_1): Always expand with fist<mode>2_<rounding>.
2800 (fistdi2_<rounding>): Change operand 0 predicate to
2801 nonimmediate_operand.
2802 (fistdi2_<rounding>_with_temp): Remove insn pattern
2803 and corresponding splitters.
2804 (fist<mode>2_<rounding>): Change operand 0 predicate to
2805 nonimmediate_operand.
2806 (fist<mode>2_<rounding>_with_temp): Remove insn pattern
2807 and corresponding splitters.
2809 (*fixuns_trunc<mode>si2_avx512f_zext): Depend on TARGET_SSE_MATH.
2811 2018-09-06 Bernd Edlinger <bernd.edlinger@hotmail.de>
2813 * varasm.c (output_constructor_regular_field): Check TYPE_SIZE_UNIT of
2816 2018-09-06 Will Schmidt <will_schmidt@vnet.ibm.com>
2818 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support for
2819 early gimple folding of vec_splat().
2820 * tree-vect-generic.c: Remove static from tree_vec_extract() definition.
2821 * gimple-fold.h: Add an extern define for tree_vec_extract().
2823 2018-09-06 Will Schmidt <will_schmidt@vnet.ibm.com>
2825 * config/rs6000/rs6000.c (fold_mergehl_helper): Add types_compatible_p
2826 wrappers around TREE_TYPE comparisons.
2828 2018-09-06 Ilya Leoshkevich <iii@linux.ibm.com>
2831 * config/s390/predicates.md: Add nonsym_memory_operand.
2832 * config/s390/s390.c (s390_legitimize_cs_operand): If operand
2833 contains a SYMBOL_REF, load it into an intermediate pseudo.
2834 (s390_emit_compare_and_swap): Legitimize operand.
2835 * config/s390/s390.md: Use the new nonsym_memory_operand
2836 with UNSPECV_CAS patterns.
2838 2018-09-06 Ilya Leoshkevich <iii@linux.ibm.com>
2841 * config/s390/s390-passes.def: New file.
2842 * config/s390/s390-protos.h (class rtl_opt_pass): Add forward
2844 (make_pass_s390_early_mach): Add declaration.
2845 * config/s390/s390.c (make_pass_s390_early_mach):
2846 (s390_option_override): Remove dynamic registration.
2847 * config/s390/t-s390: Add s390-passes.def.
2849 2018-09-06 Ilya Leoshkevich <iii@linux.ibm.com>
2851 * config/s390/s390.c (s390_decompose_constant_pool_ref):
2852 Remove UNSPEC_LTREL_BASE check.
2853 (annotate_constant_pool_refs): Likewise.
2854 (find_constant_pool_ref): Likewise.
2855 (find_ltrel_base): Removed.
2856 (replace_ltrel_base): Removed.
2857 (s390_mainpool_finish): Remove replace_ltrel_base call.
2858 (s390_chunkify_start): Remove pending LTREL_BASE logic.
2859 (s390_chunkify_finish): Remove replace_ltrel_base call.
2860 * config/s390/s390.md: Remove UNSPEC_LTREL_BASE.
2862 2018-09-06 Hans-Peter Nilsson <hp@axis.com>
2865 * config/cris/cris.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
2866 to speculation_safe_value_not_needed.
2868 2018-09-05 Cesar Philippidis <cesar@codesourcery.com>
2869 Bernd Schmidt <bernds_cb1@t-online.de>
2871 * config/nvptx/nvptx-opts.h: New file.
2872 * config/nvptx/nvptx.c (nvptx_file_start): Print the correct .target.
2873 * config/nvptx/nvptx.h: Include "nvptx-opts.h".
2875 (TARGET_SM35): New macro.
2876 * config/nvptx/nvptx.md (atomic_fetch_<logic><mode>): Enable with the
2878 * config/nvptx/nvptx.opt (ptx_isa, sm_30, sm_35): New enum and its
2880 (misa=): New option.
2881 * doc/invoke.texi (Nvidia PTX Options): Document -misa.
2883 2018-09-05 Uros Bizjak <ubizjak@gmail.com>
2885 * config/i386/i386.md (truncdfsf2): Remove expander.
2886 (truncdfsf2_with_temp): Ditto.
2887 (truncxf<mode>2): Ditto.
2888 (*truncdfsf_fast_mixed): Remove insn pattern.
2889 (*truncdfsf_fast_i387): Ditto.
2890 (*truncdfsf_mixed): Ditto.
2891 (*truncdfsf_i387): Ditto.
2892 (*truncdfsf2_i387_1): Ditto.
2893 (*truncxfsf2_mixed): Ditto.
2894 (*truncxfdf2_mixed): Ditto.
2895 (*truncxf<mode>2_i387_noop): Ditto. Update callers
2896 to call gen_truncxf<mode>2 instead.
2897 (*truncxf<mode>2_i387): Remove.
2898 (reg->reg splitters): Remove splitter pattern.
2899 (reg->mem splitters): Ditto.
2901 (truncdfsf2): New insn pattern.
2902 (truncxf<mode>2): Ditto.
2904 2018-09-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2906 * tree-ssa-math-opts.c (is_mult_by): New function.
2907 (is_square_of): Use the above.
2908 (optimize_recip_sqrt): New function.
2909 (pass_cse_reciprocals::execute): Use the above.
2911 2018-09-05 Richard Biener <rguenther@suse.de>
2914 * tree-ssa-sccvn.c (rpo_elim::eliminate_push_avail): Make sure
2915 to zero-init the emplaced vec.
2917 2018-09-05 Martin Liska <mliska@suse.cz>
2919 PR tree-optimization/87205
2920 * tree-switch-conversion.c (pass_lower_switch::execute):
2921 Group cases for switch statements.
2923 2018-09-05 Richard Biener <rguenther@suse.de>
2925 PR tree-optimization/87217
2926 * tree-ssa-sccvn.c (vuse_valueize): New.
2927 (vn_reference_lookup_pieces): Use it.
2928 (vn_reference_lookup): Likewise.
2930 2018-09-05 Nathan Sidwell <nathan@acm.org>
2933 * stor-layout.c (place_field): Scan forwards to check last
2934 bitfield when ms_bitfield_placement is in effect.
2936 2018-09-05 Richard Biener <rguenther@suse.de>
2939 * tree-vect-stmts.c (vectorizable_simd_clone_call): Fix bogus
2942 2018-09-05 Siddhesh Poyarekar <siddhesh@sourceware.org>
2943 Kugan Vivekanandarajah <kugan.vivekanandarajah@linaro.org>
2945 * config/aarch64/falkor-tag-collision-avoidance.c: New file.
2946 * config.gcc (extra_objs): Build it.
2947 * config/aarch64/t-aarch64 (falkor-tag-collision-avoidance.o):
2949 * config/aarch64/aarch64-passes.def
2950 (pass_tag_collision_avoidance): New pass.
2951 * config/aarch64/aarch64.c (qdf24xx_tunings): Add
2952 AARCH64_EXTRA_TUNE_RENAME_LOAD_REGS to tuning_flags.
2953 (aarch64_classify_address): Remove static qualifier.
2954 (aarch64_address_info, aarch64_address_type): Move to...
2955 * config/aarch64/aarch64-protos.h: ... here.
2956 (make_pass_tag_collision_avoidance): New function.
2957 * config/aarch64/aarch64-tuning-flags.def (rename_load_regs):
2960 2018-09-05 Martin Liska <mliska@suse.cz>
2962 * doc/gcov.texi: Update documentation of humar
2964 * gcov.c (format_count): Print one decimal place, it provides
2965 more fine number of situations like '1G' vs. '1.4G'.
2967 2018-09-05 Martin Liska <mliska@suse.cz>
2970 * config/rs6000/rs6000.opt: Mark the option as Deprecated.
2971 * optc-gen.awk: Allow 'Var' for Deprecated options in order
2972 to generate a MASK value.
2974 2018-09-04 H.J. Lu <hongjiu.lu@intel.com>
2977 * dwarf2out.c (based_loc_descr): Allow hard frame pointer even
2978 if frame pointer isn't used.
2979 (compute_frame_pointer_to_fb_displacement): Likewise.
2981 2018-09-04 Jakub Jelinek <jakub@redhat.com>
2984 * common/config/i386/i386-common.c (OPTION_MASK_ISA_XSAVEOPT_SET,
2985 OPTION_MASK_ISA_XSAVES_SET, OPTION_MASK_ISA_XSAVEC_SET): Use
2986 OPTION_MASK_ISA_XSAVE_SET instead of OPTION_MASK_ISA_XSAVE.
2987 (OPTION_MASK_ISA_XSAVE_UNSET): Add OPTION_MASK_ISA_XSAVES_UNSET
2988 and OPTION_MASK_ISA_XSAVEC_UNSET.
2990 2018-09-04 Max Filippov <jcmvbkbc@gmail.com>
2992 * config/xtensa/xtensa.c (xtensa_expand_atomic): Reorder AND and
2993 XOR operations in NAND case.
2995 2018-09-04 Aldy Hernandez <aldyh@redhat.com>
2997 * wide-int-range.cc (wide_int_range_convert): New.
2998 * wide-int-range.h (wide_int_range_convert): New.
2999 * tree-vrp.c (extract_range_from_unary_expr): Abstract wide int
3000 code into wide_int_range_convert.
3001 (extract_range_into_wide_ints): Do not munge anti range constants
3002 into the entire domain. Just return the range back.
3004 2018-09-04 Martin Liska <mliska@suse.cz>
3006 * genmatch.c (output_line_directive): Add new argument
3008 (dt_simplify::gen_1): Encapsulate dump within __builtin_expect.
3010 2018-09-04 Jonathan Wakely <jwakely@redhat.com>
3012 * doc/invoke.texi (Option Summary): Add whitespace.
3014 * doc/invoke.texi (Option Summary): Add -Waligned-new.
3016 2018-09-04 Richard Biener <rguenther@suse.de>
3018 PR tree-optimization/87211
3019 * tree-ssa-sccvn.c (visit_phi): When value-numbering to a
3020 backedge value we're supposed to treat as VARYING also number
3021 the PHI to VARYING in case it got a different value-number already.
3023 2018-09-04 Aldy Hernandez <aldyh@redhat.com>
3025 * tree-vrp.c (vrp_can_optimize_bit_op): Remove.
3026 (extract_range_from_binary_expr_1): Do not call
3027 vrp_can_optimize_bit_op.
3028 * wide-int-range.cc (wide_int_range_can_optimize_bit_op): Make
3030 (wide_int_range_get_mask_and_bounds): New.
3031 (wide_int_range_optimize_bit_op): New.
3032 (wide_int_range_bit_ior): Call wide_int_range_optimize_bit_op.
3033 (wide_int_range_bit_and): Same.
3034 * wide-int-range.h (wide_int_range_can_optimize_bit_op): Remove.
3035 (wide_int_range_optimize_bit_op): New.
3036 (wide_int_range_get_mask_and_bounds): New.
3038 2018-09-04 Richard Biener <rguenther@suse.de>
3040 PR tree-optimization/87176
3041 * tree-ssa-sccvn.c (visit_phi): Remove redundant allsame
3042 variable. When value-numbering a virtual PHI node make sure
3043 to not value-number to the backedge value.
3045 2018-09-04 Jonathan Wakely <jwakely@redhat.com>
3047 * doc/extend.texi (Long Long, Hex Floats): Document support for
3048 long long and hex floats in more recent versions of ISO C++.
3050 2018-09-03 Richard Biener <rguenther@suse.de>
3052 PR tree-optimization/87177
3053 * tree-ssa-sccvn.c (vuse_ssa_val): Revert previous change, keep
3056 2018-09-03 Alexander Monakov <amonakov@ispras.ru>
3058 * bb-reorder.c (edge_order): Convert to C-qsort-style
3059 tri-state comparator.
3060 (reorder_basic_blocks_simple): Change std::stable_sort to
3063 2018-09-03 Alexander Monakov <amonakov@ispras.ru>
3065 * tree-loop-distribution.c (offset_cmp): Convert to C-qsort-style
3066 tri-state comparator.
3067 (fuse_memset_builtins): Change std::stable_sort to gcc_stablesort.
3069 2018-09-03 Alexander Monakov <amonakov@ispras.ru>
3071 * sort.cc (struct sort_ctx): New field 'nlim'. Use it...
3072 (mergesort): ... here as maximum count for using netsort.
3073 (gcc_qsort): Set nlim to 3 if stable sort is requested.
3074 (gcc_stablesort): New.
3075 * system.h (gcc_stablesort): Declare.
3077 2018-09-03 Alexander Monakov <amonakov@ispras.ru>
3079 * sort.cc (gcc_qsort) [CHECKING_P]: Call qsort_chk.
3080 * system.h (qsort): Always redirect to gcc_qsort. Update comment.
3081 * vec.c (qsort_chk): Do not call gcc_qsort. Update comment.
3083 2018-09-03 Segher Boessenkool <segher@kernel.crashing.org>
3085 * config/rs6000/rs6000.md (*mov<mode>_hardfloat32): Remove %U from the
3086 lxsdx and stxsdx alternatives.
3087 (*mov<mode>_hardfloat64): Ditto.
3088 * config/rs6000/vsx.md (*vsx_extract_<mode>_store): Ditto.
3090 2018-09-03 Richard Biener <rguenther@suse.de>
3092 PR tree-optimization/87200
3093 * tree-ssa-sccvn.c (vn_nary_build_or_lookup_1): Valueize a
3096 2018-09-03 Martin Liska <mliska@suse.cz>
3098 PR tree-optimization/87201
3099 * tree-switch-conversion.c (switch_decision_tree::balance_case_nodes):
3100 Fix parenthesis in an expression.
3102 2018-09-03 Richard Biener <rguenther@suse.de>
3104 PR tree-optimization/87197
3105 * tree-ssa-sccvn.c (vn_nary_build_or_lookup_1): Mark the new def
3106 visited. CSE the VN_INFO hashtable lookup.
3108 PR tree-optimization/87169
3109 * tree-ssa-sccvn.c (do_rpo_vn): When marking loops for not
3110 iterating make sure there's no extra backedges from irreducible
3111 regions feeding the header. Mark the destination block
3114 2018-09-03 Martin Liska <mliska@suse.cz>
3117 * common/common-target.def: Add TARGET_GET_VALID_OPTION_VALUES.
3118 * common/common-targhooks.c (default_get_valid_option_values):
3120 * common/common-targhooks.h (default_get_valid_option_values):
3122 * common/config/i386/i386-common.c: Move processor_target_table
3124 (ix86_get_valid_option_values): New function.
3125 (TARGET_GET_VALID_OPTION_VALUES): New macro.
3126 * config/i386/i386.c (struct ptt): Move to i386-common.c.
3127 (PTA_*): Move all defined masks into i386-common.c.
3128 (ix86_function_specific_restore): Use new processor_cost_table.
3129 * config/i386/i386.h (struct ptt): Moved from i386.c.
3130 (struct pta): Likewise.
3131 * doc/tm.texi: Document new TARGET_GET_VALID_OPTION_VALUES.
3132 * doc/tm.texi.in: Likewise.
3133 * opt-suggestions.c (option_proposer::suggest_option):
3134 Pass prefix to build_option_suggestions.
3135 (option_proposer::get_completions): Likewise.
3136 (option_proposer::build_option_suggestions): Use the new target
3138 * opts.c (struct option_help_tuple): New struct.
3139 (print_filtered_help): Use the new target hook.
3141 2018-09-03 Martin Liska <mliska@suse.cz>
3144 * predict.c (set_even_probabilities): Add likely_edges
3145 argument and handle cases where we have precisely one
3147 (combine_predictions_for_bb): Catch also likely_edges.
3148 (tree_predict_by_opcode): Handle gswitch statements.
3149 * tree-cfg.h (find_case_label_for_value): New declaration.
3150 (find_taken_edge_switch_expr): Likewise.
3151 * tree-switch-conversion.c (switch_decision_tree::balance_case_nodes):
3152 Find pivot in decision tree based on probabily, not by number of
3155 2018-09-02 Gerald Pfeifer <gerald@pfeifer.com>
3157 * doc/standards.texi (Standards): Update Objective-C reference.
3159 2018-09-01 Gerald Pfeifer <gerald@pfeifer.com>
3161 * doc/install.texi (Prerequisites): Update link for MPC.
3163 2018-09-01 Michael Matz <matz@suse.de>
3165 PR tree-optimization/87074
3166 * gimple-loop-jam.c (unroll_jam_possible_p): Check loop exit
3167 PHIs for outer-loop uses.
3169 2018-09-01 Gerald Pfeifer <gerald@pfeifer.com>
3171 * doc/generic.texi (OpenMP): Adjust link to openmp.org.
3172 * doc/invoke.texi (C Dialect Options): Ditto.
3174 2018-09-01 Gerald Pfeifer <gerald@pfeifer.com>
3176 * doc/install.texi (Prerequisites): Adjust link mpfr.org.
3178 2018-08-31 Richard Biener <rguenther@suse.de>
3180 PR tree-optimization/87168
3181 * tree-ssa-sccvn.c (SSA_VAL): Add visited output parameter.
3182 (rpo_elim::eliminate_avail): When OP was not visited it must
3185 2018-08-31 David Malcolm <dmalcolm@redhat.com>
3187 * tree-vrp.c (copy_value_range): Convert param "from" from
3188 "value_range *" to "const value_range *".
3189 (range_is_null): Likewise for param "vr".
3190 (range_int_cst_p): Likewise.
3191 (range_int_cst_singleton_p): Likewise.
3192 (symbolic_range_p): Likewise.
3193 (value_ranges_intersect_p): Likewise for both params.
3194 (value_range_nonnegative_p): Likewise for param "vr".
3195 (value_range_constant_singleton): Likewise.
3196 (vrp_set_zero_nonzero_bits): Likewise for param "ar".
3197 (extract_range_into_wide_ints): Likewise for param "vr".
3198 (extract_range_from_multiplicative_op): Likewise for params "vr0"
3200 (vrp_can_optimize_bit_op): Likewise.
3201 (extract_range_from_binary_expr_1): Likewise for params "vr0_" and
3203 (extract_range_from_unary_expr): Likewise.
3204 (debug_value_range): Likewise for param "vr".
3205 (value_range::dump): Add "const" qualifier.
3206 (vrp_prop::check_array_ref): Convert local "vr" from
3207 "value_range *" to "const value_range *".
3208 (vrp_prop::check_mem_ref): Likewise.
3209 (vrp_prop::visit_stmt): Likewise for local "old_vr".
3210 (vrp_intersect_ranges_1): Likewise for param "vr_1".
3211 (vrp_intersect_ranges): Likewise.
3212 (simplify_stmt_for_jump_threading): Likewise for local "vr".
3213 (vrp_prop::vrp_finalize): Likewise.
3214 * tree-vrp.h (value_range::dump): Add "const" qualifier.
3215 (vrp_intersect_ranges): Add "const" qualifier to params as above.
3216 (extract_range_from_unary_expr): Likewise.
3217 (value_range_constant_singleton): Likewise.
3218 (symbolic_range_p): Likewise.
3219 (copy_value_range): Likewise.
3220 (extract_range_from_binary_expr_1): Likewise.
3221 (range_int_cst_p): Likewise.
3222 (vrp_set_zero_nonzero_bits): Likewise.
3223 (range_int_cst_singleton_p): Likewise.
3225 2018-08-31 Vlad Lazar <vlad.lazar@arm.com>
3227 * config/aarch64/arm_neon.h (vabsd_s64): New.
3228 (vnegd_s64): Likewise.
3230 2018-08-31 Martin Jambor <mjambor@suse.cz>
3232 * ipa-cp.c (estimate_local_effects): Replace wrong MAX with MIN.
3234 2018-08-31 Martin Liska <mliska@suse.cz>
3236 * ipa-icf.c (sem_item::add_type): Use
3237 sem_item::m_type_hash_cache.
3238 * ipa-icf.h: Move the cache from sem_item_optimizer
3241 2018-08-31 Nathan Sidwell <nathan@acm.org>
3243 * doc/extend.texi (Backwards Compatibility): Remove implicit
3244 extern C leeway of () being (...).
3246 2018-08-31 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3248 * ipa-inline.c (can_inline_edge_by_limits_p): Fix typos in comment.
3250 2018-08-31 Segher Boessenkool <segher@kernel.crashing.org>
3254 * config/rs6000/rs6000.md (lround<mode>di2): Gate on TARGET_FPRND.
3256 2018-08-31 Jakub Jelinek <jakub@redhat.com>
3259 * expmed.c (expand_mult_const): Use immed_wide_int_const instead of
3260 gen_int_mode. Formatting fixes.
3262 2018-08-30 Sandra Loosemore <sandra@codesourcery.com>
3264 * target.def (custom_function_descriptors): Improve documentation.
3265 * doc/tm.texi.in (Trampolines): Expand discussion of function
3266 descriptors and move TARGET_CUSTOM_FUNCTION_DESCRIPTORS to the
3267 beginning of the section.
3268 * doc/tm.texi: Regenerated.
3270 2018-08-30 Jose E. Marchesi <jose.marchesi@oracle.com>
3272 * cfg.h (class auto_edge_flag): Spell out the template-id of the
3273 base class in the initializer list. This is a workaround for
3274 building with older GCC.
3275 (class auto_bb_flag): Likewise.
3277 2018-08-30 Aaron Sawdey <acsawdey@linux.ibm.com>
3279 * config/rs6000/altivec.md (altivec_eq<mode>): Remove star.
3280 (altivec_vcmpequ<VI_char>_p): Remove star.
3281 * config/rs6000/rs6000-string.c (do_load_for_compare): Support
3283 (expand_strncmp_vec_sequence): New function.
3284 (emit_final_str_compare_vec): New function.
3285 (expand_strn_compare): Add support for vector strncmp.
3286 * config/rs6000/rs6000.opt (-mstring-compare-inline-limit): Change
3287 length specification to bytes.
3288 * config/rs6000/vsx.md (vsx_ld_elemrev_v16qi_internal): Remove star.
3289 (vcmpnezb_p): New pattern.
3290 * doc/invoke.texi (RS/6000 and PowerPC Options): Update documentation
3291 for option -mstring-compare-inline-limit.
3293 2018-08-30 Thiago Macieira <thiago.macieira@intel.com>
3295 * config/i386/i386.c (PTA_WESTMERE): Remove PTA_AES.
3296 (PTA_SKYLAKE): Add PTA_AES.
3297 (PTA_GOLDMONT): Likewise.
3299 2018-08-29 Jan Hubicka <jh@suse.cz>
3302 * lto-opts.c (lto_write_options): Always stream PIC/PIE mode.
3303 * lto-wrapper.c (merge_and_complain): Fix merging of PIC/PIE.
3305 2018-08-29 Jan Hubicka <jh@suse.cz>
3307 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not follow
3309 (hash_tree): Do not visit TYPE_STUB_DECL.
3310 * tree-streamer-out.c (write_ts_type_common_tree_pointers): Do not
3311 stream TYPE_STUB_DECL.
3312 * tree-streamer-in.c (lto_input_ts_type_common_tree_pointers): Likewise.
3313 * ipa-utils.h (type_with_linkage_p): Do not rely on TYPE_STUB_DECL
3314 after free_lang_data.
3315 (type_in_anonymous_namespace_p): Likewise.
3317 2018-08-29 Jan Hubicka <jh@suse.cz>
3319 * sreal.h (SREAL_PART_BITS): Change to 31; remove seemingly unnecessary
3320 comment that it has to be even number.
3321 (class sreal): Change m_sig type to int32_t.
3322 * sreal.c (sreal::dump, sreal::to_int, opreator+, operator-): Use
3323 int64_t for temporary calculations.
3324 (sreal_verify_basics): Drop one bit from minimum and maximum.
3326 2018-08-30 Richard Biener <rguenther@suse.de>
3328 PR tree-optimization/87147
3329 * tree-ssa-sccvn.c (SSA_VISITED): New function.
3330 (visit_phi): When the degenerate result is from the backedge and
3331 we didn't visit its definition yet drop to VARYING.
3332 (do_rpo_vn): Properly mark blocks with incoming backedges as executable.
3334 2018-08-29 Jan Hubicka <jh@suse.cz>
3336 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not walk
3338 (hash_tree): Likewise.
3340 2018-08-29 Jan Hubicka <jh@suse.cz>
3342 * tree.c (find_decls_types_r): Walk also TYPE_NEXT_PTR_TO
3343 and TYPE_NEXT_REF_TO.
3345 2018-08-29 Jan Hubicka <jh@suse.cz>
3347 * sreal.h (SREAL_PART_BITS): Change to 31; remove seemingly unnecessary
3348 comment that it has to be even number.
3349 (class sreal): Change m_sig type to int32_t.
3350 * sreal.c (sreal::dump, sreal::to_int, opreator+, operator-): Use
3351 int64_t for temporary calculations.
3352 (sreal_verify_basics): Drop one bit from minimum and maximum.
3354 2018-08-30 Tamar Christina <tamar.christina@arm.com>
3356 * config/aarch64/aarch64.c (aarch64_expand_movmem): Set TImode max.
3358 2018-08-30 Vlad Lazar <vlad.lazar@arm.com>
3361 * expmed.c (canonicalize_comparison): Use wi::sub instead of wi::add
3362 if to_add is negative.
3364 2018-08-29 Bernd Edlinger <bernd.edlinger@hotmail.de>
3367 * builtins.c (c_strlen): Improve range checks.
3369 2018-08-29 Martin Sebor <msebor@redhat.com>
3370 Jeff Law <law@redhat.com>
3372 PR tree-optimization/86714
3373 PR tree-optimization/86711
3374 * builtins.c (c_strlen): Add arguments to call to string_constant.
3375 * expr.c (string_constant): Add argument. Detect missing nul
3376 terminator and outermost declaration it's missing in.
3377 * expr.h (string_constant): Add argument.
3378 * fold-const.c (read_from_constant_string): Add arguments to call to
3380 (c_getstr): Likewise.
3381 * tree-ssa-forwprop.c (simplify_builtin_call): Likewise.
3383 * tree-ssa-strlen.c (get_stridx): Likewise.
3385 2018-08-29 Jan Hubicka <jh@suse.cz>
3387 * tree-streamer-in.c (lto_input_ts_function_decl_tree_pointers):
3388 Do not stream DECL_VINDEX.
3389 * tree-streamer-out.c (write_ts_function_decl_tree_pointers): Likewise.
3390 * tree.c (free_lang_data_in_decl): Clear DECL_VINDEX.
3391 (decl_function_context): Use DECL_VIRTUAL_P rather than DECL_VINDEX.
3393 2018-08-29 Richard Biener <rguenther@suse.de>
3395 * tree-ssa-sccvn.c (vuse_ssa_val): Return NULL for unvisited
3396 virtual operands that are not default defs to honor region
3398 (rpo_vn_valueize): Remove ineffective code here.
3400 2018-08-29 Richard Biener <rguenther@suse.de>
3402 PR tree-optimization/87132
3403 * tree-ssa-alias.c (get_continuation_for_phi): Do not translate
3404 when skipping defs reachable over backedges.
3406 2018-08-29 Richard Biener <rguenther@suse.de>
3408 * tree-core.h: Document use of deprecated_flag in SSA_NAME.
3409 * tree.h (SSA_NAME_POINTS_TO_READONLY_MEMORY): Define.
3410 * tree-into-ssa.c (pass_build_ssa::execute): Initialize
3411 function parameters SSA_NAME_POINTS_TO_READONLY_MEMORY from fnspec.
3412 * tree-ssa-sccvn.c (const_parms, init_const_parms): Remove.
3413 (vn_reference_lookup_3): Remove use of const_parms.
3414 (free_rpo_vn): Do not free const_parms.
3415 (do_rpo_vn): Do not call init_const_parms.
3416 * tree-ssa-alias.c (refs_may_alias_p_1): Honor
3417 SSA_NAME_POINTS_TO_READONLY_MEMORY.
3418 (call_may_clobber_ref_p_1): Likewise.
3420 2018-08-29 Alexander Monakov <amonakov@ispras.ru>
3423 * invoke.texi (Optimization Options): List -ftree-scev-cprop.
3425 (-ftree-scev-cprop): Document.
3427 2018-08-29 Jan Hubicka <jh@suse.cz>
3429 * sreal.h (normalize, normalize_up, normalize_down): Add new_sig/new_exp
3431 (sreal constructor): Update.
3432 * sreal.c (sreal:operator+, sreal:operator-, sreal:operator*,
3433 sreal:operator/): Update.
3435 2018-08-29 Martin Liska <mliska@suse.cz>
3437 * tree-switch-conversion.c (switch_conversion::expand):
3438 Strenghten assumption about gswitch statements.
3440 2018-08-29 Richard Biener <rguenther@suse.de>
3442 PR tree-optimization/87117
3443 * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_stmt): Only
3444 re-value-number released SSA VDEFs.
3446 2018-08-29 Richard Biener <rguenther@suse.de>
3448 PR tree-optimization/87126
3449 * tree-ssa-sccvn.c (vn_reference_insert): Remove assert.
3451 2018-08-28 Jim Wilson <jimw@sifive.com>
3453 * config/riscv/pic.md: Rewrite.
3454 * config/riscv/riscv.c (riscv_address_insns): Return cost of 3 for
3456 * config/riscv/riscv.md (ZERO_EXTEND_LOAD): Delete.
3457 (SOFTF, default_load, softload, softstore): New.
3459 2018-08-28 Jeff Law <law@redhat.com>
3461 * fold-const.c (fold_binary_loc): Remove recently added assert.
3463 2018-08-28 Joern Rennecke <joern.rennecke@riscy-ip.com>
3465 * genpreds.c (write_predicate_subfunction): Also add ATTRIBUTE_UNUSED
3466 to OP parmeter of generated function.
3468 2018-08-28 MCC CS <deswurstes@users.noreply.github.com>
3470 PR tree-optimization/87009
3471 * match.pd: Add boolean optimizations.
3473 2018-08-28 Martin Sebor <msebor@redhat.com>
3476 * calls.c (alloc_max_size): Treat HOST_WIDE_INT special.
3477 * gimple-ssa-warn-alloca.c (adjusted_warn_limit): New function.
3478 (pass_walloca::gate): Use it.
3479 (alloca_call_type): Same.
3480 (pass_walloca::execute): Same.
3481 * stor-layout.c (layout_decl): Treat HOST_WIDE_INT special.
3483 2018-08-28 David Malcolm <dmalcolm@redhat.com>
3485 * dumpfile.h (ATTRIBUTE_GCC_DUMP_PRINTF): Change version check on
3486 GCC_VERSION for usage of "__gcc_dump_printf__" format from
3489 2018-08-28 Richard Biener <rguenther@suse.de>
3491 PR tree-optimization/87124
3492 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Guard against
3493 constants before looking up avail.
3495 2018-08-28 Jakub Jelinek <jakub@redhat.com>
3498 * calls.c (maybe_warn_nonstring_arg): Punt early if
3499 warn_stringop_overflow is zero. Don't call get_range_strlen
3500 on 3rd argument, keep iterating until lenrng[1] is INTEGER_CST.
3501 Swap comparison operands to have constants on rhs. Only use
3502 lenrng[1] if non-NULL and INTEGER_CST. Don't uselessly
3503 increment lenrng[0].
3505 2018-08-28 Richard Sandiford <richard.sandiford@arm.com>
3507 * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Fix unguarded
3508 use of tree_to_shwi. Remove duplicated test for the size being
3509 a whole number of bytes.
3511 2018-08-28 Richard Biener <rguenther@suse.de>
3513 PR tree-optimization/87117
3514 * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_cleanup):
3515 Handle removed stmt without LHS (GIMPLE_NOP).
3517 2018-08-28 Richard Biener <rguenther@suse.de>
3519 PR tree-optimization/87117
3520 * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Exclude
3521 void which is is_gimple_reg_type by checking for COMPLETE_TYPE_P.
3523 2018-08-28 Richard Biener <rguenther@suse.de>
3525 PR tree-optimization/87117
3526 * tree-ssa-pre.c (compute_avail): Do not make expressions
3527 with predicated values available.
3528 (get_expr_value_id): Assert we do not run into predicated value
3531 2018-08-28 Richard Biener <rguenther@suse.de>
3533 PR tree-optimization/87117
3534 * tree-ssa-operands.c (add_stmt_operand): STRING_CST may
3535 get virtual operands.
3536 (get_expr_operands): Handle STRING_CST like other decls.
3538 2018-08-28 Martin Liska <mliska@suse.cz>
3540 * tree.h: Update documentation of fndecl_built_in_p
3544 2018-08-27 Jeff Law <law@redhat.com>
3545 PR tree-optimization/87110
3546 * tree-ssa-dse.c (compute_trims): Handle non-constant
3549 2018-08-27 Martin Sebor <msebor@redhat.com>
3551 PR tree-optimization/86914
3552 * tree-ssa-strlen.c (maybe_set_strlen_range): Avoid MEM_REF.
3554 2018-08-27 Martin Sebor <msebor@redhat.com>
3556 PR tree-optimization/87112
3557 * builtins.c (expand_builtin_strnlen): Convert c_strlen result to
3558 the type of the bound argument.
3560 2018-08-27 Jeff Law <law@redhat.com>
3562 * tree-ssa-dse.c (compute_trims): Handle case where the reference's
3563 type does not have a TYPE_SIZE_UNIT.
3565 2018-08-27 Steve Ellcey <sellcey@cavium.com>
3567 * config/aarch64/aarch64-speculation.cc: Replace include of cfg.h
3568 with include of backend.h.
3570 2018-08-27 Richard Biener <rguenther@suse.de>
3572 PR tree-optimization/86927
3573 * tree-vect-loop.c (vect_create_epilog_for_reduction): Properly
3574 use const cond reduction code.
3576 2018-08-27 Alexander Monakov <amonakov@ispras.ru>
3578 PR tree-optimization/85758
3579 * match.pd ((X & Y) ^ Y): Add :s qualifier to inner expression.
3581 2018-08-27 David Malcolm <dmalcolm@redhat.com>
3584 * diagnostic-show-locus.c (class layout_range): Update for
3585 conversion of show_caret_p to a tri-state.
3586 (layout_range::layout_range): Likewise.
3587 (make_range): Likewise.
3588 (layout::maybe_add_location_range): Likewise.
3589 (layout::should_print_annotation_line_p): Don't show annotation
3590 lines for ranges that are SHOW_LINES_WITHOUT_RANGE.
3591 (layout::get_state_at_point): Update for conversion of
3592 show_caret_p to a tri-state. Bail out early for
3593 SHOW_LINES_WITHOUT_RANGE, so that such ranges don't affect
3594 underlining or source colorization.
3595 (gcc_rich_location::add_location_if_nearby): Update for conversion
3596 of show_caret_p to a tri-state.
3597 (selftest::test_one_liner_multiple_carets_and_ranges): Likewise.
3598 (selftest::test_one_liner_fixit_replace_equal_secondary_range):
3600 (selftest::test_one_liner_labels): Likewise.
3601 * gcc-rich-location.c (gcc_rich_location::add_expr): Update for
3602 conversion of show_caret_p to a tri-state.
3603 * pretty-print.c (text_info::set_location): Likewise.
3604 * pretty-print.h (text_info::set_location): Likewise.
3605 * substring-locations.c (format_warning_n_va): Likewise.
3606 * tree-diagnostic.c (default_tree_printer): Likewise.
3607 * tree-pretty-print.c (newline_and_indent): Likewise.
3609 2018-08-27 David Malcolm <dmalcolm@redhat.com>
3612 * diagnostic-show-locus.c (get_line_span_for_fixit_hint): Show the
3613 line above for line-insertion fix-it hints.
3614 (selftest::test_fixit_insert_containing_newline): Update the
3615 expected results, and add a test with line-numbering enabled.
3617 2018-08-27 Martin Liska <mliska@suse.cz>
3620 * sanopt.c (sanitize_rewrite_addressable_params): Ignore
3621 params with DECL_HAS_VALUE_EXPR_P.
3623 2018-08-27 Martin Liska <mliska@suse.cz>
3625 * config/i386/i386.c (ix86_expand_set_or_movmem): Dump
3626 selected expansion strategy.
3628 2018-08-27 Martin Liska <mliska@suse.cz>
3630 * builtins.h (is_builtin_fn): Remove and fndecl_built_in_p.
3631 * builtins.c (is_builtin_fn): Likewise.
3632 * attribs.c (diag_attr_exclusions): Use new function
3633 fndecl_built_in_p and remove check for FUNCTION_DECL if
3635 (builtin_mathfn_code): Likewise.
3636 (fold_builtin_expect): Likewise.
3637 (fold_call_expr): Likewise.
3638 (fold_builtin_call_array): Likewise.
3639 (fold_call_stmt): Likewise.
3640 (set_builtin_user_assembler_name): Likewise.
3641 (is_simple_builtin): Likewise.
3642 * calls.c (gimple_alloca_call_p): Likewise.
3643 (maybe_warn_nonstring_arg): Likewise.
3644 * cfgexpand.c (expand_call_stmt): Likewise.
3645 * cgraph.c (cgraph_update_edges_for_call_stmt_node): Likewise.
3646 (cgraph_edge::verify_corresponds_to_fndecl): Likewise.
3647 (cgraph_node::verify_node): Likewise.
3648 * cgraphclones.c (build_function_decl_skip_args): Likewise.
3649 (cgraph_node::create_clone): Likewise.
3650 * config/arm/arm.c (arm_insert_attributes): Likewise.
3651 * config/i386/i386.c (ix86_gimple_fold_builtin): Likewise.
3652 * dse.c (scan_insn): Likewise.
3653 * expr.c (expand_expr_real_1): Likewise.
3654 * fold-const.c (operand_equal_p): Likewise.
3655 (fold_binary_loc): Likewise.
3656 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Likewise.
3657 * gimple-low.c (lower_stmt): Likewise.
3658 * gimple-pretty-print.c (dump_gimple_call): Likewise.
3659 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
3661 * gimple.c (gimple_build_call_from_tree): Likewise.
3662 (gimple_call_builtin_p): Likewise.
3663 (gimple_call_combined_fn): Likewise.
3664 * gimplify.c (gimplify_call_expr): Likewise.
3665 (gimple_boolify): Likewise.
3666 (gimplify_modify_expr): Likewise.
3667 (gimplify_addr_expr): Likewise.
3668 * hsa-gen.c (gen_hsa_insns_for_call): Likewise.
3669 * ipa-cp.c (determine_versionability): Likewise.
3670 * ipa-fnsummary.c (compute_fn_summary): Likewise.
3671 * ipa-param-manipulation.c (ipa_modify_formal_parameters): Likewise.
3672 * ipa-split.c (visit_bb): Likewise.
3673 (split_function): Likewise.
3674 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
3675 * lto-cgraph.c (input_node): Likewise.
3676 * lto-streamer-out.c (write_symbol): Likewise.
3677 * omp-low.c (setjmp_or_longjmp_p): Likewise.
3678 (lower_omp_1): Likewise.
3679 * predict.c (strip_predict_hints): Likewise.
3680 * print-tree.c (print_node): Likewise.
3681 * symtab.c (symtab_node::output_to_lto_symbol_table_p): Likewise.
3682 * trans-mem.c (is_tm_irrevocable): Likewise.
3683 (is_tm_load): Likewise.
3684 (is_tm_simple_load): Likewise.
3685 (is_tm_store): Likewise.
3686 (is_tm_simple_store): Likewise.
3687 (is_tm_abort): Likewise.
3688 (tm_region_init_1): Likewise.
3689 * tree-call-cdce.c (gen_shrink_wrap_conditions): Likewise.
3690 * tree-cfg.c (verify_gimple_call): Likewise.
3691 (move_stmt_r): Likewise.
3692 (stmt_can_terminate_bb_p): Likewise.
3693 * tree-eh.c (lower_eh_constructs_2): Likewise.
3694 * tree-if-conv.c (if_convertible_stmt_p): Likewise.
3695 * tree-inline.c (remap_gimple_stmt): Likewise.
3696 (copy_bb): Likewise.
3697 (estimate_num_insns): Likewise.
3698 (fold_marked_statements): Likewise.
3699 * tree-sra.c (scan_function): Likewise.
3700 * tree-ssa-ccp.c (surely_varying_stmt_p): Likewise.
3701 (optimize_stack_restore): Likewise.
3702 (pass_fold_builtins::execute): Likewise.
3703 * tree-ssa-dce.c (mark_stmt_if_obviously_necessary): Likewise.
3704 (mark_all_reaching_defs_necessary_1): Likewise.
3705 * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Likewise.
3706 * tree-ssa-forwprop.c (simplify_builtin_call): Likewise.
3707 (pass_forwprop::execute): Likewise.
3708 * tree-ssa-loop-im.c (stmt_cost): Likewise.
3709 * tree-ssa-math-opts.c (pass_cse_reciprocals::execute): Likewise.
3710 * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Likewise.
3711 * tree-ssa-strlen.c (get_string_length): Likewise.
3712 * tree-ssa-structalias.c (handle_lhs_call): Likewise.
3713 (find_func_aliases_for_call): Likewise.
3714 * tree-ssa-ter.c (find_replaceable_in_bb): Likewise.
3715 * tree-stdarg.c (optimize_va_list_gpr_fpr_size): Likewise.
3716 * tree-tailcall.c (find_tail_calls): Likewise.
3717 * tree.c (need_assembler_name_p): Likewise.
3718 (free_lang_data_in_decl): Likewise.
3719 (get_call_combined_fn): Likewise.
3720 * ubsan.c (is_ubsan_builtin_p): Likewise.
3721 * varasm.c (incorporeal_function_p): Likewise.
3722 * tree.h (DECL_BUILT_IN): Remove and replace with
3724 (DECL_BUILT_IN_P): Transfort to fndecl_built_in_p.
3725 (fndecl_built_in_p): New.
3727 2018-08-27 Martin Liska <mliska@suse.cz>
3729 PR tree-optimization/86847
3730 * tree-switch-conversion.c (switch_decision_tree::dump_case_nodes):
3731 Dump also subtree probability.
3732 (switch_decision_tree::do_jump_if_equal): New function.
3733 (switch_decision_tree::emit_case_nodes): Handle special
3734 situations in balanced tree that can be emitted much simpler.
3735 Fix calculation of probabilities that happen in tree expansion.
3736 * tree-switch-conversion.h (struct cluster): Add
3738 (struct simple_cluster): Likewise.
3739 (struct case_tree_node): Add new function has_child.
3740 (do_jump_if_equal): New.
3742 2018-08-27 Martin Liska <mliska@suse.cz>
3744 * tree-switch-conversion.c (bit_test_cluster::find_bit_tests):
3745 Add new argument to bit_test_cluster constructor.
3746 (bit_test_cluster::emit): Set bits really number of values
3748 (bit_test_cluster::hoist_edge_and_branch_if_true): Add
3749 probability argument.
3750 * tree-switch-conversion.h (struct bit_test_cluster):
3751 Add m_handles_entire_switch.
3753 2018-08-27 Martin Liska <mliska@suse.cz>
3755 PR tree-optimization/86702
3756 * tree-switch-conversion.c (jump_table_cluster::emit):
3757 Make probabilities even for values in jump table
3758 according to number of cases handled.
3759 (switch_decision_tree::compute_cases_per_edge): Pass
3760 argument to reset_out_edges_aux function.
3761 (switch_decision_tree::analyze_switch_statement): Likewise.
3762 * tree-switch-conversion.h (switch_decision_tree::reset_out_edges_aux):
3765 2018-08-27 Martin Liska <mliska@suse.cz>
3767 * cfgexpand.c (expand_asm_stmt): Use label_to_block and pass
3768 cfun argument explicitly.
3769 * gimple-pretty-print.c (dump_gimple_switch): Likewise.
3770 * hsa-gen.c (gen_hsa_insns_for_switch_stmt): Use new
3771 function gimple_switch_default_bb.
3772 (convert_switch_statements):
3774 * ipa-fnsummary.c (set_switch_stmt_execution_predicate):
3775 * stmt.c (label_to_block_fn): Use label_to_block and pass
3776 cfun argument explicitly and use gimple_switch_label_bb.
3777 (expand_case): Likewise.
3778 * tree-cfg.c (lower_phi_internal_fn): Use label_to_block and pass
3779 cfun argument explicitly. Likewise.
3780 (make_edges_bb): Likewise.
3781 (make_cond_expr_edges): Likewise.
3782 (get_cases_for_edge): Likewise.
3783 (make_gimple_switch_edges): Likewise.
3784 (label_to_block_fn): Likewise.
3785 (label_to_block): Likewise.
3786 (make_goto_expr_edges): Likewise.
3787 (make_gimple_asm_edges): Likewise.
3788 (main_block_label): Likewise.
3789 (group_case_labels_stmt): Likewise.
3790 (find_taken_edge_computed_goto): Likewise.
3791 (find_taken_edge_switch_expr): Likewise.
3792 (gimple_verify_flow_info): Likewise.
3793 (gimple_redirect_edge_and_branch): Likewise.
3794 (gimple_switch_label_bb): New function.
3795 (gimple_switch_default_bb): Likewise.
3796 (gimple_switch_edge): Likewise.
3797 (gimple_switch_default_edge): Likewise.
3798 * tree-cfg.h (label_to_block_fn): Remove and replace ...
3799 (label_to_block): ... with this.
3800 (gimple_switch_label_bb): New.
3801 (gimple_switch_default_bb): Likewise.
3802 (gimple_switch_edge): Likewise.
3803 (gimple_switch_default_edge): Likewise.
3804 * tree-cfgcleanup.c (convert_single_case_switch): Use
3805 new gimple functions and pass new argument to label_to_block.
3806 (cleanup_control_flow_bb):
3807 * tree-eh.c (make_eh_dispatch_edges): Use label_to_block and pass
3808 cfun argument explicitly.
3809 (make_eh_edges): Likewise.
3810 (redirect_eh_dispatch_edge): Likewise.
3811 (lower_resx): Likewise.
3812 (lower_eh_dispatch): Likewise.
3813 (maybe_remove_unreachable_handlers): Likewise.
3814 (unsplit_eh): Likewise.
3815 (cleanup_empty_eh): Likewise.
3816 (verify_eh_edges): Likewise.
3817 (verify_eh_dispatch_edge): Likewise.
3818 * tree-ssa-dom.c (record_edge_info): Likewise.
3819 * tree-ssa-forwprop.c (simplify_gimple_switch_label_vec): Likewise.
3820 * tree-ssa-threadedge.c (thread_around_empty_blocks): Likewise.
3821 (thread_through_normal_block): Likewise.
3822 * tree-ssa-uncprop.c (associate_equivalences_with_edges): Likewise.
3823 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds):
3824 * tree-switch-conversion.c (switch_conversion::collect): Use new
3826 (switch_conversion::check_final_bb): Likewise.
3827 (switch_conversion::gather_default_values): Pass new argument
3829 (switch_conversion::build_constructors): Likewise.
3830 (switch_decision_tree::compute_cases_per_edge): Use new
3831 gimple_switch_edge function.
3832 (switch_decision_tree::analyze_switch_statement): Pass new argument
3834 (switch_decision_tree::try_switch_expansion): Use
3835 gimple_switch_default_edge.
3836 * tree-vrp.c (find_switch_asserts): Pass new argument
3838 * vr-values.c (vr_values::vrp_visit_switch_stmt): Likewise.
3839 (vr_values::simplify_switch_using_ranges): Likewise.
3841 2018-08-27 Richard Biener <rguenther@suse.de>
3843 * cfganal.h (rev_post_order_and_mark_dfs_back_seme): Declare.
3844 * cfganal.c (rev_post_order_and_mark_dfs_back_seme): New function.
3846 * tree-ssa-sccvn.h (struct vn_pval): New structure.
3847 (struct vn_nary_op_s): Add unwind_to member. Add
3848 predicated_values flag and put result into a union together
3849 with a linked list of vn_pval.
3850 (struct vn_ssa_aux): Add name member to make maintaining
3851 a map of SSA name to vn_ssa_aux possible. Remove no longer
3852 needed info, dfsnum, low, visited, on_sccstack, use_processed
3853 and range_info_anti_range_p members.
3854 (run_scc_vn, vn_eliminate, free_scc_vn, vn_valueize): Remove.
3855 (do_rpo_vn, run_rpo_vn, eliminate_with_rpo_vn, free_rpo_vn):
3857 (vn_valueize): New global.
3858 (vn_context_bb): Likewise.
3859 (VN_INFO_RANGE_INFO, VN_INFO_ANTI_RANGE_P, VN_INFO_RANGE_TYPE,
3860 VN_INFO_PTR_INFO): Remove.
3861 * tree-ssa-sccvn.c: ... (rewrite)
3862 (pass_fre::execute): For -O2+ initialize loops and run
3863 RPO VN in optimistic mode (iterating). For -O1 and -Og
3864 run RPO VN in non-optimistic mode.
3865 * params.def (PARAM_SCCVN_MAX_SCC_SIZE): Remove.
3866 (PARAM_RPO_VN_MAX_LOOP_DEPTH): Add.
3867 * doc/invoke.texi (sccvn-max-scc-size): Remove.
3868 (rpo-vn-max-loop-depth): Document.
3869 * tree-ssa-alias.c (walk_non_aliased_vuses): Stop walking
3870 when valuezing the VUSE signals we walked out of the region.
3871 * tree-ssa-pre.c (phi_translate_1): Ignore predicated values.
3872 (phi_translate): Set VN context block to use for availability
3874 (compute_avail): Likewise.
3875 (pre_valueize): New function.
3876 (pass_pre::execute): Adjust to the RPO VN API.
3878 * tree-ssa-loop-ivcanon.c: Include tree-ssa-sccvn.h.
3879 (propagate_constants_for_unrolling): Remove.
3880 (tree_unroll_loops_completely): Perform value-numbering
3881 on the unrolled bodies loop parent.
3883 2018-08-27 Richard Biener <rguenther@suse.de>
3885 * tree-ssa-pre.c (compute_antic): Re-use inverted postorder
3886 for partial antic compute.
3888 2018-08-27 Jakub Jelinek <jakub@redhat.com>
3890 PR rtl-optimization/87065
3891 * combine.c (simplify_if_then_else): Formatting fix.
3892 (if_then_else_cond): Guard MULT optimization with SCALAR_INT_MODE_P
3894 (known_cond): Don't return const_true_rtx for vector modes. Use
3895 CONST0_RTX instead of const0_rtx. Formatting fixes.
3897 2018-08-27 Martin Liska <mliska@suse.cz>
3899 PR gcov-profile/87069
3900 * gcov.c (process_file): Record files already processed
3901 and warn about a file being processed multiple times.
3903 2018-08-27 Martin Liska <mliska@suse.cz>
3906 * config/aarch64/aarch64.c (aarch64_override_options_internal):
3907 Set default values for x_aarch64_*_string strings.
3908 * config/aarch64/aarch64.opt: Remove --{march,mcpu,mtune}==
3909 prefix. For -mabi do not print '=ABI' in help and use
3910 <option_value> format for -msve-vector-bits and -moverride
3913 2018-08-26 Jeff Law <law@redhat.com>
3915 * config/mips/frame-header-opt.c: Include "backend.h" rather than
3918 2018-08-26 Marek Polacek <polacek@redhat.com>
3920 PR c++/87029, Implement -Wredundant-move.
3921 * doc/invoke.texi: Document -Wredundant-move.
3923 2018-08-25 Martin Sebor <msebor@redhat.com>
3925 PR tree-optimization/87059
3926 * builtins.c (expand_builtin_strncmp): Convert MIN_EXPR operand
3927 to the same type as the other.
3928 * fold-const.c (fold_binary_loc): Assert expectation.
3930 2018-08-25 Iain Sandoe <iain@sandoe.co.uk>
3932 * config/darwin.c (machopic_legitimize_pic_address): Clean up
3933 extraneous parentheses, dead code section and formatting.
3935 2018-08-24 David Malcolm <dmalcolm@redhat.com>
3938 * diagnostic-show-locus.c (layout::layout): Ensure the margin is
3939 wide enough for jumps in the line-numbering to be visible.
3940 (layout::print_gap_in_line_numbering): New member function.
3941 (layout::calculate_line_spans): When using line numbering, merge
3942 line spans that are only 1 line apart.
3943 (diagnostic_show_locus): When printing line numbers, show gaps in
3944 line numbering directly, rather than printing headers.
3945 (selftest::test_diagnostic_show_locus_fixit_lines): Add test of
3946 line-numbering with multiple line spans.
3947 (selftest::test_fixit_insert_containing_newline_2): Add test of
3948 line-numbering, in which the spans are close enough to be merged.
3950 2018-08-24 Aldy Hernandez <aldyh@redhat.com>
3952 * gimple-ssa-evrp-analyze.c (set_ssa_range_info): Pass value_range
3953 to range_includes_zero_p. Do not special case VR_ANTI_RANGE.
3954 * tree-vrp.c (range_is_nonnull): Remove.
3955 (range_includes_zero_p): Accept value_range instead of min/max.
3956 (extract_range_from_binary_expr_1): Do not early bail on
3958 Use range_includes_zero_p instead of range_is_nonnull.
3959 (extract_range_from_unary_expr): Use range_includes_zero_p instead
3960 of range_is_nonnull.
3961 (vrp_meet_1): Pass value_range to range_includes_zero_p. Do not
3962 special case VR_ANTI_RANGE.
3963 (vrp_finalize): Same.
3964 * tree-vrp.h (range_includes_zero_p): Pass value_range as argument
3966 (range_is_nonnull): Remove.
3967 * vr-values.c (vrp_stmt_computes_nonzero): Use
3968 range_includes_zero_p instead of range_is_nonnull.
3969 (extract_range_basic): Pass value_range to range_includes_zero_p
3970 instead of range_is_nonnull.
3972 2018-08-24 Uros Bizjak <ubizjak@gmail.com>
3974 * emit-rtl.c (init_emit_once): Do not emit MODE_POINTER_BOUNDS RTXes.
3975 * emit-rtl.h (rtl_data): Remove return_bnd.
3976 * explow.c (trunc_int_for_mode): Do not handle POINTER_BOUNDS_MODE_P.
3977 * function.c (diddle_return_value): Do not handle crtl->return_bnd.
3978 * genmodes.c (complete_mode): Do not handle MODE_POINTER_BOUNDS.
3979 (POINTER_BOUNDS_MODE): Remove definition.
3980 (make_pointer_bounds_mode): Remove.
3981 (get_mode_class): Do not handle MODE_POINTER_BOUNDS.
3982 * machmode.h (POINTER_BOUNDS_MODE_P): Remove definition.
3983 (scalare_mode::includes_p): Do not handle MODE_POINTER_BOUNDS.
3984 * mode-classes.def: Do not define MODE_POINTER_BOUNDS.
3985 * stor-layout.c (int_mode_for_mode): Do not handle MODE_POINTER_BOUNDS.
3986 * tree-core.h (enum tree_index): Remove TI_POINTER_BOUNDS_TYPE.
3987 * varasm.c (output_constant_pool_2): Do not handle MODE_POINTER_BOUNDS.
3989 * config/i386/i386-modes.def (BND32, BND64): Remove.
3990 * config/i386/i386.c (dbx_register_map): Remove bound registers.
3991 (dbx64_register_map): Ditto.
3992 (svr4_dbx_register_map): Ditto.
3993 (indirect_thunk_bnd_needed): Remove.
3994 (indirect_thunks_bnd_used): Ditto.
3995 (indirect_return_bnd_needed): Ditto.
3996 (indirect_return_via_cx_bnd): Ditto.
3997 (enum indirect_thunk_prefix): Remove indirect_thunk_prefix_bnd.
3998 (indirect_thunk_name): Remove handling of indirect_thunk_prefix_bnd.
3999 (output_indirect_thunk): Ditto. Remove need_prefix argument.
4000 (output_indirect_thunk_function): Remove handling of
4001 indirect_return_bnd_needed, indirect_return_via_cx_bnd,
4002 indirect_thunk_bnd_needed and indirect_thunks_bnd_used variables.
4003 (ix86_save_reg): Remove handling of crtl->return_bnd.
4004 (ix86_legitimate_constant_p): Remove handling of POINTER_BOUNDS_MODE_P.
4005 (ix86_print_operand_address_as): Remove handling of UNSPEC_BNDMK_ADDR
4006 and UNSPEC_BNDLX_ADDR.
4007 (ix86_output_indirect_branch_via_reg): Remove handling of
4008 indirect_thunk_prefix_bnd.
4009 (ix86_output_indirect_branch_via_push): Ditto.
4010 (ix86_output_function_return): Ditto.
4011 (ix86_output_indirect_function_return): Ditto.
4012 (avoid_func_arg_motion): Do not handle UNSPEC_BNDSTX.
4013 * config/i386/i386.h (FIXED_REGISTERS): Remove bound registers.
4014 (CALL_USED_REGISTERS): Ditto.
4015 (REG_ALLOC_ORDER): Update for removal of bound registers.
4016 (HI_REGISTER_NAMES): Ditto.
4017 * config/i386/i386.md (UNSPEC_BNDMK, UNSPEC_BNDMK_ADDR, UNSPEC_BNDSTX)
4018 (UNSPEC_BNDLDX, UNSPEC_BNDLDX_ADDR, UNSPEC_BNDCL, UNSPEC_BNDCU)
4019 (UNSPEC_BNDCN, UNSPEC_MPX_FENCE): Remove.
4020 (BND0_REG, BND1_REG, BND2_REG, BND3_REG): Remove
4021 (FIRST_PSEUDO_REG): Update.
4022 (BND): Remove mode iterator.
4023 * config/i386/predicates.md (bnd_mem_operator): Remove.
4025 2018-08-24 Richard Sandiford <richard.sandiford@arm.com>
4027 * tree-vect-stmts.c (vectorizable_bswap): Handle variable-length
4030 2018-08-24 Richard Sandiford <richard.sandiford@arm.com>
4032 * tree-vect-slp.c (vect_transform_slp_perm_load): Separate out
4033 the case in which the permute needs only a single element and
4034 repeats for every vector of the result. Extend that case to
4035 handle variable-length vectors.
4036 * tree-vect-stmts.c (vectorizable_load): Update accordingly.
4038 2018-08-24 H.J. Lu <hongjiu.lu@intel.com>
4041 * dwarf2out.c (save_macinfo_strings): Call set_indirect_string
4042 on DW_MACINFO_start_file for -gsplit-dwarf -g3.
4044 2018-08-24 Richard Biener <rguenther@suse.de>
4046 * cfg.h (struct control_flow_graph): Add edge_flags_allocated and
4047 bb_flags_allocated members.
4048 (auto_flag): New RAII class for allocating flags.
4049 (auto_edge_flag): New RAII class for allocating edge flags.
4050 (auto_bb_flag): New RAII class for allocating bb flags.
4051 * cfgloop.c (verify_loop_structure): Allocate temporary edge
4053 * cfganal.c (dfs_enumerate_from): Remove use of visited sbitmap
4054 in favor of temporarily allocated BB flag.
4055 * hsa-brig.c: Re-order includes.
4056 * hsa-dump.c: Likewise.
4057 * hsa-regalloc.c: Likewise.
4058 * print-rtl.c: Likewise.
4059 * profile-count.c: Likewise.
4061 2018-08-24 Segher Boessenkool <segher@kernel.crashing.org>
4064 * config/rs6000/rs6000.c (toc_relative_expr_p): Check that the base is
4067 2018-08-24 Aldy Hernandez <aldyh@redhat.com>
4070 * wide-int-range.cc (wide_int_range_div): Do not ignore result
4071 from wide_int_range_multiplicative_op.
4073 2018-08-23 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
4075 * tree-vect-data-refs.c (vect_grouped_store_supported): Fix typo
4078 2018-08-23 Giuliano Belinassi <giuliano.belinassi@usp.br>
4080 * genmatch.c (parser::parse_operation): Fix typo 'exapnded'
4083 2018-08-23 Alexander Monakov <amonakov@ispras.ru>
4085 * tree-scalar-evolution.c (final_value_replacement_loop): Dump
4086 full GENERIC expression used for replacement.
4088 2018-08-23 Aldy Hernandez <aldyh@redhat.com>
4090 * tree-vrp.c (abs_extent_range): Remove.
4091 (extract_range_into_wide_ints): Pass wide ints by reference.
4092 (extract_range_from_binary_expr_1): Rewrite the *DIV_EXPR code.
4093 Pass wide ints by reference in all calls to
4094 extract_range_into_wide_ints.
4095 * wide-int-range.cc (wide_int_range_div): New.
4096 * wide-int-range.h (wide_int_range_div): New.
4097 (wide_int_range_includes_zero_p): New.
4098 (wide_int_range_zero_p): New.
4100 2018-08-23 Matthew Malcomson <matthew.malcomson@arm.com>
4102 * config/aarch64/aarch64.md (arches): New enum.
4103 (arch): New enum attr.
4104 (arch_enabled): New attr.
4105 (enabled): Now uses arch_enabled only.
4106 (simd, sve, fp16): Removed attribute.
4107 (fp): Attr now defined in terms of 'arch'.
4108 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64, *movti_aarch64,
4109 *movhf_aarch64, <optab><fcvt_target><GPF:mode>2,
4110 <FCVT_F2FIXED:fcvt_fixed_insn><GPF:mode>3,
4111 <FCVT_FIXED2F:fcvt_fixed_insn><GPI:mode>3): Merge 'fp' and 'simd'
4112 attributes into 'arch'.
4113 (*movsf_aarch64, *movdf_aarch64, *movtf_aarch64, *add<mode>3_aarch64,
4114 subdi3, neg<mode>2, <optab><mode>3, one_cmpl<mode>2,
4115 *<NLOGICAL:optab>_one_cmpl<mode>3, *xor_one_cmpl<mode>3,
4116 *aarch64_ashl_sisd_or_int_<mode>3, *aarch64_lshr_sisd_or_int_<mode>3,
4117 *aarch64_ashr_sisd_or_int_<mode>3, *aarch64_sisd_ushl): Convert use of
4118 'simd' attribute into 'arch'.
4119 (load_pair_sw_<SX:mode><SX2:mode>, load_pair_dw_<DX:mode><DX2:mode>,
4120 store_pair_sw_<SX:mode><SX2:mode>, store_pair_dw_<DX:mode><DX2:mode>):
4121 Convert use of 'fp' attribute to 'arch'.
4122 * config/aarch64/aarch64-simd.md (move_lo_quad_internal_<mode>,
4123 move_lo_quad_internal_<mode>): (different modes) Merge 'fp' and 'simd'
4125 (move_lo_quad_internal_be_<mode>, move_lo_quad_internal_be_<mode>):
4126 (different modes) Merge 'fp' and 'simd' into 'arch'.
4127 (*aarch64_combinez<mode>, *aarch64_combinez_be<mode>): Merge 'fp' and
4130 2018-08-23 Segher Boessenkool <segher@kernel.crashing.org>
4132 PR rtl-optimization/87026
4133 * expmed.c (canonicalize_comparison): If we can no longer create
4134 pseudoregisters, don't.
4136 2018-08-23 Richard Earnshaw <rearnsha@arm.com>
4139 * config/arm/arm-protos.h (arm_emit_speculation_barrier): New
4141 * config/arm/arm.c (speculation_barrier_libfunc): New static
4143 (arm_init_libfuncs): Initialize it.
4144 (arm_emit_speculation_barrier): New function.
4145 * config/arm/arm.md (speculation_barrier): Call
4146 arm_emit_speculation_barrier for architectures that do not have
4148 (speculation_barrier_insn): Only match on Armv7 or later.
4150 2018-08-23 Richard Biener <rguenther@suse.de>
4153 * tree-inline.c (copy_bb): Drop unused __builtin_va_arg_pack_len
4156 2018-08-23 Richard Sandiford <richard.sandiford@arm.com>
4158 * config/aarch64/aarch64.c (aarch64_evpc_sve_tbl): Fix handling
4159 of single-vector TBLs.
4160 (aarch64_vectorize_vec_perm_const): Set one_vector_p when only
4163 2018-08-23 Richard Sandiford <richard.sandiford@arm.com>
4166 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const_1): Fix
4167 aarch64_evpc_tbl guard.
4169 2018-08-22 Bernd Edlinger <bernd.edlinger@hotmail.de>
4171 * tree-ssa-dse.c (compute_trims): Avoid folding away undefined
4174 2018-08-22 Martin Sebor <msebor@redhat.com>
4177 * tree-pretty-print.c (pretty_print_string): Add argument.
4178 (dump_generic_node): Call to pretty_print_string with string size.
4180 2018-08-22 Segher Boessenkool <segher@kernel.crashing.org>
4182 PR rtl-optimization/86771
4183 * combine.c (try_combine): Do not allow splitting a resulting PARALLEL
4184 of two SETs into those two SETs, one to be placed at i2, if that SETs
4185 destination is modified between i2 and i3.
4187 2018-08-22 Richard Sandiford <richard.sandiford@arm.com>
4189 PR tree-optimization/86725
4190 * tree-vect-loop.c (vect_inner_phi_in_double_reduction_p): New
4192 (vect_analyze_scalar_cycles_1): Check it.
4194 2018-08-22 Richard Sandiford <richard.sandiford@arm.com>
4196 PR tree-optimization/86725
4197 * tree-vect-loop.c (vect_is_simple_reduction): When treating
4198 an outer loop phi as a double reduction, make sure that the
4199 single user of the phi result is an inner loop phi.
4201 2018-08-22 Richard Sandiford <richard.sandiford@arm.com>
4203 * tree-vect-data-refs.c (vect_analyze_group_access_1): Convert
4204 grouped stores with gaps to a strided group.
4206 2018-08-22 Richard Sandiford <richard.sandiford@arm.com>
4208 * tree-vect-stmts.c (get_group_load_store_type)
4209 (get_load_store_type): Only test STMT_VINFO_STRIDED_P for the
4210 first statement in a group.
4212 2018-08-22 Iain Sandoe <iain@sandoe.co.uk>
4214 * config/darwin.h (LINK_COMMAND_SPEC_A): Sync LTO options with
4215 the sequence used in gcc/gcc.c.
4217 2018-08-22 Iain Sandoe <iain@sandoe.co.uk>
4220 * gcc-ar.c (main): Don’t try to invoke the plug-in if we’re not
4223 2018-08-22 Iain Sandoe <iain@sandoe.co.uk>
4225 * config/darwin10.h (LINK_GCC_C_SEQUENCE_SPEC): Adjust to use the
4226 Darwin10-specific unwinder-shim.
4227 * config/darwin12.h (LINK_GCC_C_SEQUENCE_SPEC): Remove.
4228 * config/rs6000/darwin.h (DARWIN_CRT1_SPEC, DARWIN_DYLIB1_SPEC):
4229 New to cater for Darwin10 Rosetta.
4231 2018-08-22 Iain Sandoe <iain@sandoe.co.uk>
4233 * config/i386/i386.c (ix86_output_addr_diff_elt): Move the MACH-O
4234 specific test before the one for HAVE_AS_GOTOFF_IN_DATA.
4236 2018-08-22 Iain Sandoe <iain@sandoe.co.uk>
4241 * gcc/dwarf2out.c (FUNC_SECOND_SECT_LABEL): New.
4242 (dwarf2out_switch_text_section): Generate a local label for the
4243 second function sub-section and apply it as the second FDE start
4245 * gcc/final.c (final_scan_insn_1): Emit second FDE label after the
4246 second sub-section start.
4248 2018-08-22 Richard Biener <rguenther@suse.de>
4250 PR tree-optimization/86988
4251 * tree-vrp.c (vrp_prop::check_mem_ref): Bail out on VLAs.
4253 2018-08-22 Richard Biener <rguenther@suse.de>
4255 PR tree-optimization/86945
4256 * tree-cfg.c (generate_range_test): Use unsigned arithmetic.
4258 2018-08-22 Alexandre Oliva <oliva@adacore.com>
4260 * config/rs6000/rs6000.c (SMALL_DATA_RELOC, SMALL_DATA_REG): Add
4261 a comment about how uses of r2 for .sdata2 come about.
4263 2018-08-22 Alexandre Oliva <aoliva@redhat.com>
4265 * tree-ssa-reassoc.c (is_reassociable_op): Fix cut&pasto.
4267 2018-08-21 Marek Polacek <polacek@redhat.com>
4269 PR c++/86981, Implement -Wpessimizing-move.
4270 * doc/invoke.texi: Document -Wpessimizing-move.
4272 2018-08-21 Jan Hubicka <jh@suse.cz>
4274 * tree.c (find_decls_types_r): Do not check for redundant typedefs.
4275 * tree.h (is_redundant_typedef): Remove.
4276 * dwarf2out.c (is_redundant_typedef): Turn into static function.
4278 2018-08-21 Jan Hubicka <jh@suse.cz>
4280 * tree.c (free_lang_data_in_decl): Remove types from DECL_CONTEXT
4283 2018-08-21 Tamar Christina <tamar.christina@arm.com>
4285 * expmed.c (extract_low_bits): Reject invalid subregs early.
4287 2018-08-21 Bernd Edlinger <bernd.edlinger@hotmail.de>
4290 * tree-ssa-strlen.c (adjust_last_stmt): Avoid folding away undefined
4293 2018-08-21 Rasmus Villemoes <rv@rasmusvillemoes.dk>
4295 * config/vxworks.h: Guard vxworks_asm_out_constructor and
4296 vxworks_asm_out_destructor by !HAVE_INITFINI_ARRAY_SUPPORT
4297 * config/vxworks.c: Likewise.
4299 2018-08-21 Rasmus Villemoes <rv@rasmusvillemoes.dk>
4301 * config/vxworks.c: Set targetm.have_ctors_dtors
4302 if HAVE_INITFINI_ARRAY_SUPPORT.
4303 * config/vxworks.h: Set SUPPORTS_INIT_PRIORITY
4304 if HAVE_INITFINI_ARRAY_SUPPORT.
4306 2018-08-21 Rasmus Villemoes <rv@rasmusvillemoes.dk>
4308 * config/vxworks.h: Add $(WIND_BASE)/target/h/wrn/coreip to
4309 default search path for VxWorks < 7.
4311 2018-08-21 Bernd Edlinger <bernd.edlinger@hotmail.de>
4313 * gimple-ssa-sprintf.c (decl_constant_value): Remove.
4314 (get_format_string): Refer to c_getstr.
4316 2018-08-21 Tom de Vries <tdevries@suse.de>
4318 * cgraph.h (debuginfo_early_init, debuginfo_init, debuginfo_fini)
4319 (debuginfo_start, debuginfo_stop, debuginfo_early_start)
4320 (debuginfo_early_stop): Declare.
4321 * cgraphunit.c (debuginfo_early_init, debuginfo_init, debuginfo_fini)
4322 (debuginfo_start, debuginfo_stop, debuginfo_early_start)
4323 (debuginfo_early_stop): New function.
4324 (symbol_table::finalize_compilation_unit): Call debuginfo_early_start
4325 and debuginfo_early_stop.
4326 * dwarf2out.c (dwarf2out_finish, dwarf2out_early_finish): Dump dwarf.
4327 * toplev.c (compile_file): Call debuginfo_start and debuginfo_stop.
4328 (general_init): Call debuginfo_early_init.
4329 (finalize): Call debuginfo_fini.
4330 (do_compile): Call debuginfo_init.
4331 * doc/invoke.texi (@gccoptlist): Add -fdump-debug and
4333 (@item -fdump-debug, @item -fdump-earlydebug): Add.
4335 2018-08-21 Tom de Vries <tdevries@suse.de>
4337 * dwarf2out.c (print_dw_val, print_loc_descr, print_die): Handle
4338 flag_dump_noaddr and flag_dump_unnumbered.
4340 2018-08-21 Aldy Hernandez <aldyh@redhat.com>
4342 * wide-int-range.cc (wide_int_range_abs): New.
4343 (wide_int_range_order_set): Rename from wide_int_range_min_max.
4344 * wide-int-range.h (wide_int_range_abs): New.
4345 (wide_int_range_min_max): New.
4346 * tree-vrp.c (extract_range_from_unary_expr): Rewrite ABS_EXPR
4347 case to call wide_int_range_abs.
4348 Rewrite MIN/MAX_EXPR to call wide_int_range_min_max.
4349 (extract_range_from_abs_expr): Delete.
4351 2018-08-20 Michael Meissner <meissner@linux.ibm.com>
4354 * config/rs6000/rs6000.md (extendsi<mode>2): Change constraints
4355 from 'Y' to 'YZ' to enable the LWAX instruction to be generated
4358 2018-08-20 Nathan Sidwell <nathan@acm.org>
4359 Jeff Law <law@redhat.com>
4361 * config/s390/s390-c (s390_macro_to_expand): Use cpp_macro_p.
4362 * config/spu/spu-c.c (spu_macro_to_expand): Likewise.
4364 2018-08-20 David Malcolm <dmalcolm@redhat.com>
4367 * attribs.c (diag_attr_exclusions): Add auto_diagnostic_group instance.
4368 (decl_attributes): Likewise.
4369 * calls.c (maybe_warn_nonstring_arg): Add auto_diagnostic_group
4371 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
4372 * diagnostic-core.h (class auto_diagnostic_group): New class.
4373 * diagnostic.c (diagnostic_initialize): Initialize the new fields.
4374 (diagnostic_report_diagnostic): Handle the first diagnostics within
4376 (emit_diagnostic): Add auto_diagnostic_group instance.
4378 (inform_n): Likewise.
4379 (warning): Likewise.
4380 (warning_at): Likewise.
4381 (warning_n): Likewise.
4382 (pedwarn): Likewise.
4383 (permerror): Likewise.
4385 (error_n): Likewise.
4386 (error_at): Likewise.
4388 (fatal_error): Likewise.
4389 (internal_error): Likewise.
4390 (internal_error_no_backtrace): Likewise.
4391 (auto_diagnostic_group::auto_diagnostic_group): New ctor.
4392 (auto_diagnostic_group::~auto_diagnostic_group): New dtor.
4393 * diagnostic.h (struct diagnostic_context): Add fields
4394 "diagnostic_group_nesting_depth",
4395 "diagnostic_group_emission_count", "begin_group_cb",
4397 * gimple-ssa-isolate-paths.c (find_implicit_erroneous_behavior):
4398 Add auto_diagnostic_group instance(s).
4399 (find_explicit_erroneous_behavior): Likewise.
4400 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Likewise.
4401 * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Likewise.
4402 * gimplify.c (warn_implicit_fallthrough_r): Likewise.
4403 (gimplify_va_arg_expr): Likewise.
4404 * hsa-gen.c (HSA_SORRY_ATV): Likewise.
4405 (HSA_SORRY_AT): Likewise.
4406 * ipa-devirt.c (compare_virtual_tables): Likewise.
4407 (warn_odr): Likewise.
4408 * multiple_target.c (expand_target_clones): Likewise.
4409 * opts-common.c (cmdline_handle_error): Likewise.
4410 * reginfo.c (globalize_reg): Likewise.
4411 * substring-locations.c (format_warning_n_va): Likewise.
4412 * tree-inline.c (expand_call_inline): Likewise.
4413 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Likewise.
4414 * tree-ssa-loop-niter.c
4415 (do_warn_aggressive_loop_optimizations): Likewise.
4416 * tree-ssa-uninit.c (warn_uninit): Likewise.
4417 * tree.c (warn_deprecated_use): Likewise.
4419 2018-08-20 H.J. Lu <hongjiu.lu@intel.com>
4422 * config/i386/i386.md (eh_return): Always update EH return
4423 address in word_mode.
4425 2018-08-20 Chung-Lin Tang <cltang@codesourcery.com>
4427 * targhooks.c (std_gimplify_va_arg_expr): Properly handle case of when
4428 TARGET_SPLIT_COMPLEX_ARG is defined.
4430 2018-08-20 Bernd Edlinger <bernd.edlinger@hotmail.de>
4432 * expr.c (store_field): Change gcc_assert to gcc_checking_assert.
4434 2018-08-20 Bernd Edlinger <bernd.edlinger@hotmail.de>
4437 * expr.c (expand_assignment): Assert that bitpos is positive.
4438 (store_field): Likewise
4439 (expand_expr_real_1): Make sure that bitpos is positive.
4440 * config/alpha/alpha.h (CONSTANT_ADDRESS_P): Avoid signed
4443 2018-08-20 Nathan Sidwell <nathan@acm.org>
4445 * Makefile.in (CPP_ID_DATA_H): Delete.
4446 (CPP_INTERNAL_H): Don't add it.
4447 (GTFILES): Replace CPP_ID_DATA_H with CPPLIB_H.
4448 * gengtype.c (open_base_files): Replace cpp-id-data.h with cpplib.h
4450 2018-08-20 Richard Biener <rguenther@suse.de>
4452 PR tree-optimization/78655
4453 * tree-vrp.c (extract_range_from_binary_expr_1): Make
4454 pointer + offset nonnull if either operand is nonnull work.
4456 2018-08-20 Tom de Vries <tdevries@suse.de>
4458 * dwarf2out.c (add_scalar_info): Don't add reference to existing die
4459 unless the referenced die describes the added property using
4460 DW_AT_location or DW_AT_const_value. Fall back to exprloc case.
4461 Otherwise, add a DW_AT_location to the referenced die.
4463 2018-08-19 Uros Bizjak <ubizjak@gmail.com>
4466 * config/i386/i386.c (ix86_rtx_costs) [case SET]: Check source for
4467 register_operand when calling ix86_set_reg_reg_cost.
4468 [case CONST_INT, case CONST, case LABEL_REF, case SYMBOL_REF]:
4469 Set *total to 0 for operands that satisfy x86_64_immediate_operand
4470 predicate and to 1 otherwise.
4472 2018-08-18 Iain Sandoe <iain@sandoe.co.uk>
4474 * config/darwin.c (darwin_override_options): If -gsplit-dwarf is set,
4475 emit a diagnostic that it is not supported and reset the option.
4476 * config/darwin.h (DRIVER_SELF_SPECS): Note that gsplit-dwarf is not
4477 supported and consume the option. (ASM_FINAL_SPEC): New.
4479 2018-08-17 Segher Boessenkool <segher@kernel.crashing.org>
4481 * doc/md.texi (Patterns): Use @ref instead of @xref in the middle of
4484 2018-08-17 Sandra Loosemore <sandra@codesourcery.com>
4486 C-SKY port: Documentation
4488 * doc/extend.texi (C-SKY Function Attributes): New section.
4489 * doc/invoke.texi (Option Summary): Add C-SKY options.
4490 (C-SKY Options): New section.
4491 * doc/md.texi (Machine Constraints): Document C-SKY constraints.
4493 2018-08-17 Jojo <jijie_rong@c-sky.com>
4494 Huibin Wang <huibin_wang@c-sky.com>
4495 Sandra Loosemore <sandra@codesourcery.com>
4496 Chung-Lin Tang <cltang@codesourcery.com>
4498 C-SKY port: Backend implementation
4500 * config/csky/*: New.
4501 * common/config/csky/*: New.
4503 2018-08-17 Jojo <jijie_rong@c-sky.com>
4504 Huibin Wang <huibin_wang@c-sky.com>
4505 Sandra Loosemore <sandra@codesourcery.com>
4506 Chung-Lin Tang <cltang@codesourcery.com>
4507 Andrew Jenner <andrew@codesourcery.com>
4509 C-SKY port: Configury
4511 * config.gcc (csky-*-*): New.
4512 * configure.ac: Add csky to targets for dwarf2 debug_line support.
4513 * configure: Regenerated.
4515 2018-08-17 David Malcolm <dmalcolm@redhat.com>
4517 * dump-context.h: Include "dumpfile.h".
4518 (dump_context::dump_printf_va): Convert final param from va_list
4519 to va_list *. Convert from ATTRIBUTE_PRINTF to
4520 ATTRIBUTE_GCC_DUMP_PRINTF.
4521 (dump_context::dump_printf_loc_va): Likewise.
4522 * dumpfile.c: Include "stringpool.h".
4523 (make_item_for_dump_printf_va): Delete.
4524 (make_item_for_dump_printf): Delete.
4525 (class dump_pretty_printer): New class.
4526 (dump_pretty_printer::dump_pretty_printer): New ctor.
4527 (dump_pretty_printer::emit_items): New member function.
4528 (dump_pretty_printer::emit_any_pending_textual_chunks): New member
4530 (dump_pretty_printer::emit_item): New member function.
4531 (dump_pretty_printer::stash_item): New member function.
4532 (dump_pretty_printer::format_decoder_cb): New member function.
4533 (dump_pretty_printer::decode_format): New member function.
4534 (dump_context::dump_printf_va): Reimplement in terms of
4535 dump_pretty_printer.
4536 (dump_context::dump_printf_loc_va): Convert final param from va_list
4538 (dump_context::begin_scope): Reimplement call to
4539 make_item_for_dump_printf.
4540 (dump_printf): Update for change to dump_printf_va.
4541 (dump_printf_loc): Likewise.
4542 (selftest::test_capture_of_dump_calls): Convert "stmt" from
4543 greturn * to gimple *. Add a test_decl. Add tests of dump_printf
4544 with %T, %E, and %G.
4545 * dumpfile.h (ATTRIBUTE_GCC_DUMP_PRINTF): New macro.
4546 (dump_printf): Replace ATTRIBUTE_PRINTF_2 with
4547 ATTRIBUTE_GCC_DUMP_PRINTF (2, 3).
4548 (dump_printf_loc): Replace ATTRIBUTE_PRINTF_3 with
4549 ATTRIBUTE_GCC_DUMP_PRINTF (3, 0).
4550 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Convert
4551 use of HOST_WIDE_INT_PRINT_DEC on unsigned HOST_WIDE_INT "count"
4552 within a dump_printf_loc call to "%wu".
4553 (vector_alignment_reachable_p): Merge two dump_printf[_loc] calls,
4554 converting a use of HOST_WIDE_INT_PRINT_DEC to "%wd". Add a
4555 missing space after "=".
4556 * tree-vect-loop.c (vect_analyze_loop_2) Within a dump_printf
4557 call, convert use of HOST_WIDE_INT_PRINT_DEC to "%wd".
4558 * tree-vect-slp.c (vect_slp_bb): Within a dump_printf_loc call,
4559 convert use of HOST_WIDE_INT_PRINT_UNSIGNED to "%wu".
4560 * tree-vectorizer.c (try_vectorize_loop_1): Likewise. Remove
4561 duplicate "vectorized" from message.
4563 2018-08-17 Szabolcs Nagy <szabolcs.nagy@arm.com>
4565 * config/arm/arm-builtins.c (arm_init_simd_builtin_types): Clear
4566 polyNxK_t element's TYPE_STRING_FLAG.
4568 2018-08-17 Segher Boessenkool <segher@kernel.crashing.org>
4570 * config/rs6000/rs6000.md (*cbranch, *creturn): Name these patterns
4571 (they were unnamed before). Fix comments.
4573 2018-08-17 Nathan Sidwell <nathan@acm.org>
4575 * cppbuiltin.c: Include "cpplib.h", not "cpp-id-data.h".
4577 2018-08-17 Richard Biener <rguenther@suse.de>
4579 PR tree-optimization/86841
4580 * wide-int-range.cc (wide_int_range_lshift): Use to_uhwi.
4582 2018-08-17 Martin Liska <mliska@suse.cz>
4584 * common.opt: Remove Warn, Init and Report for options with
4585 Ignore/Deprecated flag. Warning is done automatically for
4587 * config/i386/i386.opt: Likewise.
4588 * config/ia64/ia64.opt: Likewise.
4589 * config/rs6000/rs6000.opt: Likewise.
4590 * cppbuiltin.c (define_builtin_macros_for_compilation_flags):
4591 Remove usage of flag_check_pointer_bounds.
4592 * lto-wrapper.c (merge_and_complain): Do not handle
4593 OPT_fcheck_pointer_bounds.
4594 (append_compiler_options): Likewise.
4595 * opt-functions.awk: Do not handle Deprecated.
4596 * optc-gen.awk: Check that Var, Report and Init are not
4597 used for an option with Ignore/Deprecated flag.
4598 * opts-common.c (decode_cmdline_option): Do not report
4600 (read_cmdline_option): Report warning for OPT_SPECIAL_deprecated
4602 * opts.h (struct cl_option): Remove cl_deprecated flag.
4603 (CL_ERR_DEPRECATED): Remove error enum value.
4605 2018-08-17 Richard Biener <rguenther@suse.de>
4608 * tree-inline.c (copy_bb): When inlining __builtin_va_arg_pack_len ()
4609 across a va-arg-pack using call adjust its return value accordingly.
4611 2018-08-16 Martin Sebor <msebor@redhat.com>
4613 PR tree-optimization/86853
4614 * gimple-ssa-sprintf.c (struct format_result): Rename member.
4615 (struct fmtresult): Add member and initialize it in ctors.
4616 (format_character): Handle %C. Extend range to NUL. Set MAYFAIL.
4617 (format_string): Handle %S the same as %ls. Set MAYFAIL.
4618 (format_directive): Set POSUNDER4K when MAYFAIL is set.
4619 (parse_directive): Handle %C same as %c.
4620 (sprintf_dom_walker::compute_format_length): Adjust.
4621 (is_call_safe): Adjust.
4623 2018-08-16 Bernd Edlinger <bernd.edlinger@hotmail.de>
4625 * builtins.c (c_strlen): Add new parameter eltsize. Use it
4626 for determining how to count the elements.
4627 * builtins.h (c_strlen): Adjust prototype.
4628 * expr.c (string_constant): Add new parameter mem_size.
4629 Set *mem_size appropriately.
4630 * expr.h (string_constant): Adjust protoype.
4631 * gimple-fold.c (get_range_strlen): Add new parameter eltsize.
4632 * gimple-fold.h (get_range_strlen): Adjust prototype.
4633 * gimple-ssa-sprintf.c (get_string_length): Add new parameter eltsize.
4634 (format_string): Call get_string_length with eltsize.
4636 2018-08-16 David Malcolm <dmalcolm@redhat.com>
4638 * diagnostic.c (default_diagnostic_start_span_fn): Call pp_string
4639 to emit the span, rather than setting it as the prefix.
4641 2018-08-16 David Malcolm <dmalcolm@redhat.com>
4643 * diagnostic-show-locus.c (layout::start_annotation_line): Add
4644 "margin_char" parameter, defaulting to space. Use it in place
4645 of pp_space for the initial part of the margin.
4646 (layout::print_leading_fixits): Use '+' when filling the margin
4647 of line-insertion fix-it hints.
4649 2018-08-16 Segher Boessenkool <segher@kernel.crashing.org>
4651 * config/rs6000/rs6000.md (two unnamed define_insn and define_split):
4654 2018-08-16 Segher Boessenkool <segher@kernel.crashing.org>
4656 * config/rs6000/altivec.md: Don't set length attribute to the default
4658 * config/rs6000/darwin.md: Ditto.
4659 * config/rs6000/dfp.md: Ditto.
4660 * config/rs6000/htm.md: Ditto.
4661 * config/rs6000/rs6000.md: Ditto.
4662 * config/rs6000/sync.md: Ditto.
4663 * config/rs6000/vsx.md: Ditto.
4665 2018-08-16 Segher Boessenkool <segher@kernel.crashing.org>
4667 * config/rs6000/altivec.md: Don't set length attribute to the default
4668 value, for branch instructions.
4669 * config/rs6000/darwin.md: Ditto.
4670 * config/rs6000/rs6000.md: Ditto.
4672 2018-08-16 Segher Boessenkool <segher@kernel.crashing.org>
4674 * config/rs6000/rs6000.md (length): Always define as const_int 4.
4675 (unnamed conditional branch define_insn): Set length to 4 or 8
4676 depending on offset.
4677 (<bd>_<mode>): Similar, for alternative 0.
4678 (<bd>tf_<mode>): Ditto.
4680 2018-08-16 Tamar Christina <tamar.christina@arm.com>
4682 * expr.c (copy_blkmode_to_reg): Perform larger copies when safe.
4684 2018-08-16 Matthew Malcomson <matthew.malcomson@arm.com>
4686 * doc/rtl.texi: Replace old RTX class names with new names.
4689 2018-08-16 Vlad Lazar <vlad.lazar@arm.com>
4691 * expmed.h (canonicalize_comparison): New declaration.
4692 * expmed.c (canonicalize_comparison, equivalent_cmp_code): New function.
4693 * expmed.c (emit_store_flag_1): Add call to canonicalize_comparison.
4694 * optabs.c (prepare_cmp_insn): Likewise.
4695 * rtl.h (unsigned_condition_p): New function which checks if a
4696 comparison operator is unsigned.
4698 2018-08-16 Nathan Sidwell <nathan@acm.org>
4700 * config/rs6000/rs6000-c.c (rs6000_macro_to_expend): Use cpp_macro_p.
4701 * config/powerpcspc/powerpcspe-c.c (rs6000_macro_to_expend): Likewise.
4703 2018-08-16 Tamar Christina <tamar.christina@arm.com>
4706 * config/arm/arm.c (arm_can_change_mode_class): Disallow subreg.
4707 * config/arm/neon.md (movv4hf, movv8hf): Refactored to..
4708 (mov<mov>): ..this and enable unconditionally.
4710 2018-08-16 Tamar Christina <tamar.christina@arm.com>
4712 * config/arm/neon.md (*neon_mov<mode>): Remove reg-to-reg alternative.
4714 2018-08-16 Sam Tebbs <sam.tebbs@arm.com>
4716 * config/aarch64/aarch64.opt (mlow-precision-recip-sqrt)
4717 (mlow-precision-sqrt, mlow-precision-div, mverbose-cost-dump): Replace
4718 "Common" with "Target".
4720 2018-08-15 Uros Bizjak <ubizjak@gmail.com>
4722 * config/i386/i386.opt (mmitigate-rop): Mark as deprecated.
4723 * doc/invoke.texi (mmitigate-rop): Remove.
4724 * config/i386/i386.c: Do not include "regrename.h".
4725 (ix86_rop_should_change_byte_p, reg_encoded_number)
4726 (ix86_get_modrm_for_rop, set_rop_modrm_reg_bits, ix86_mitigate_rop):
4728 (ix86_reorg): Remove call to ix86_mitigate_rop.
4729 * config/i386/i386.md (attr "modrm_class"): Remove.
4730 (cmp<mode>_ccno_1, mov<mode>_xor, movstrict<mode>_xor)
4731 (x86_mov<mode>cc_0_m1. x86_mov<mode>cc_0_m1_se)
4732 (x86_mov<mode>cc_0_m1_neg): Remove modrm_class attribute override.
4734 2018-08-15 Will Schmidt <will_schmidt@vnet.ibm.com>
4736 * config/rs6000/rs600.c (rs6000_gimple_fold_builtin): Add entries to
4737 allow folding of mergeh() and mergel() for the float and double types.
4738 (fold_mergehl_helper): Rework to handle building a permute tree
4741 2018-08-15 Uros Bizjak <ubizjak@gmail.com>
4743 * config/i386/i386.c (expand_vec_perm_movs): Enable V4SFmode
4746 2018-08-15 David Malcolm <dmalcolm@redhat.com>
4748 * common.opt (fdiagnostics-show-labels): New option.
4749 * diagnostic-show-locus.c (class layout_range): Add field
4751 (class layout): Add field "m_show_labels_p".
4752 (layout_range::layout_range): Add param "label" and use it to
4754 (make_range): Pass in NULL for new "label" param of layout_range's
4756 (layout::layout): Initialize m_show_labels_p.
4757 (layout::maybe_add_location_range): Pass in loc_range->m_label
4758 when constructing layout_range instances.
4759 (struct line_label): New struct.
4760 (layout::print_any_labels): New member function.
4761 (layout::print_line): Call it if label-printing is enabled.
4762 (selftest::test_one_liner_labels): New test.
4763 (selftest::test_diagnostic_show_locus_one_liner): Call it.
4764 * diagnostic.c (diagnostic_initialize): Initialize
4765 context->show_labels_p.
4766 * diagnostic.h (struct diagnostic_context): Add field
4768 * doc/invoke.texi (Diagnostic Message Formatting Options): Add
4769 -fno-diagnostics-show-labels.
4770 * dwarf2out.c (gen_producer_string): Add
4771 OPT_fdiagnostics_show_labels to the ignored options.
4772 * gcc-rich-location.c (gcc_rich_location::add_expr): Add "label"
4774 (gcc_rich_location::maybe_add_expr): Likewise.
4775 * gcc-rich-location.h (gcc_rich_location::gcc_rich_location): Add
4776 label" param, defaulting to NULL.
4777 (gcc_rich_location::add_expr): Add "label" param.
4778 (gcc_rich_location::maybe_add_expr): Likewise.
4779 (class text_range_label): New class.
4780 (class range_label_for_type_mismatch): New class.
4781 * gimple-ssa-sprintf.c (fmtwarn): Pass NULL for new label params
4782 of format_warning_va.
4783 (fmtwarn_n): Likewise for new params of format_warning_n_va.
4784 * lto-wrapper.c (merge_and_complain): Add
4785 OPT_fdiagnostics_show_labels to the "pick one setting" options.
4786 (append_compiler_options): Likewise to the dropped options.
4787 (append_diag_options): Likewise to the passed-on options.
4788 * opts.c (common_handle_option): Handle the new option.
4789 * selftest-diagnostic.c
4790 (test_diagnostic_context::test_diagnostic_context): Enable
4792 * substring-locations.c: Include "gcc-rich-location.h".
4793 (format_warning_n_va): Add "fmt_label" and "param_label" params
4794 and use them as appropriate.
4795 (format_warning_va): Add "fmt_label" and "param_label" params,
4796 passing them on to format_warning_n_va.
4797 (format_warning_at_substring): Likewise.
4798 (format_warning_at_substring_n): Likewise.
4799 * substring-locations.h (format_warning_va): Add "fmt_label" and
4800 "param_label" params.
4801 (format_warning_n_va): Likewise.
4802 (format_warning_at_substring): Likewise.
4803 (format_warning_at_substring_n): Likewise.
4804 * toplev.c (general_init): Initialize global_dc->show_labels_p.
4806 2018-08-15 Qing Zhao <qing.zhao@oracle.com>
4809 * builtins.c (expand_builtin_memcmp): Do not expand the call
4810 when overflow is detected.
4812 2018-08-15 Martin Sebor <msebor@redhat.com>
4814 PR tree-optimization/71625
4815 * config/aarch64/aarch64-builtins.c
4816 (aarch64_init_simd_builtin_types): Clear Poly8_t's TYPE_STRING_FLAG.
4818 2018-08-15 Ilya Leoshkevich <iii@linux.ibm.com>
4820 * config/s390/s390.c (s390_reorg): Remove loop.
4822 2018-08-15 Iain Sandoe <iain@sandoe.co.uk>
4825 (darwin_function_switched_text_sections): Delete.
4826 * gcc/config/darwin.h
4827 (TARGET_ASM_FUNCTION_SWITCHED_TEXT_SECTIONS): Likewise.
4829 2018-08-15 Iain Sandoe <iain@sandoe.co.uk>
4832 * config/darwin.h: (DEBUG_STR_OFFSETS_SECTION, DEBUG_LOCLISTS_SECTION,
4833 DEBUG_RNGLISTS_SECTION) new macros. (DEBUG_PUBNAMES_SECTION,
4834 DEBUG_PUBTYPES_SECTION) update to include GNU variant.
4836 2018-08-15 Martin Liska <mliska@suse.cz>
4838 PR tree-optimization/86925
4839 * predict.c (expr_expected_value_1): When taking
4840 later predictor, assign also probability.
4841 Use fold_build2_initializer_loc in order to fold
4842 the expression in -frounding-math.
4844 2018-08-14 Allan Sandfeld Jensen <allan.jensen@qt.io>
4846 * config/i386/i386.c (expand_vec_perm_movs): New method matching movs
4848 (expand_vec_perm_1): Try the new method.
4850 2018-08-14 Ilya Leoshkevich <iii@linux.ibm.com>
4853 * lra-lives.c (remove_some_program_points_and_update_live_ranges):
4854 Check whether lra_live_max_point is 0 before dividing.
4856 2018-08-14 Martin Sebor <msebor@redhat.com>
4858 PR tree-optimization/86650
4859 * tree-vrp.c (vrp_prop::check_array_ref): Print an inform message.
4860 (vrp_prop::check_mem_ref): Same.
4862 2018-08-13 Liu Hao <lh_mouse@126.com>
4864 * pretty-print.c (eat_esc_sequence): Swap the foreground and
4865 background colors if the COMMON_LVB_REVERSE_VIDEO flag is set,
4866 and clear it thereafter, as it only works for DBCS.
4868 2018-08-13 Liu Hao <lh_mouse@126.com>
4870 * pretty-print.c (mingw_ansi_fputs): Do not call _close() on the
4871 handle returned by _get_osf_handle().
4873 2018-08-13 Will Schmidt <will_schmidt@vnet.ibm.com>
4875 * gcc/config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support
4876 for folding vec_perm.
4878 2018-08-13 Will Schmidt <will_schmidt@vnet.ibm.com>
4880 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin):
4881 Add support for gimple-folding of vec_pack() and vec_unpack()
4884 2018-08-13 Will Schmidt <will_schmidt@vnet.ibm.com>
4886 * config/rs6000/rs6000.c (rs6000_builtin_valid_without_lhs): Add
4887 vec_xst variants to the list.
4888 (rs6000_gimple_fold_builtin): Add support for folding unaligned
4889 vector loads and stores.
4891 2018-08-13 David Edelsohn <dje.gcc@gmail.com>
4893 * config.gcc (rs6000-ibm-aix4.x): Delete.
4894 (rs6000-ibm-aix5.1): Delete.
4895 (rs6000-ibm-aix5.2): Delete.
4896 (rs6000-ibm-aix5.3): Delete.
4897 * config/rs6000/aix43.h: Delete.
4898 * config/rs6000/aix51.h: Delete.
4899 * config/rs6000/aix52.h: Delete.
4900 * config/rs6000/t-aix43: Delete.
4902 2018-08-13 Ilya Leoshkevich <iii@linux.ibm.com>
4904 * config/s390/s390.c (s390_decompose_constant_pool_ref):
4906 (s390_decompose_address): Factor out constant pool ref
4909 2018-08-12 Chung-Ju Wu <jasonwucj@gmail.com>
4911 * config/nds32/nds32-predicates.c
4912 (nds32_can_use_bclr_p): Change return type as bool.
4913 (nds32_can_use_bset_p): Ditto.
4914 (nds32_can_use_btgl_p): Ditto.
4915 (nds32_can_use_bitci_p): Ditto.
4916 * config/nds32/nds32-protos.h
4917 (nds32_can_use_bclr_p): Change declaration.
4918 (nds32_can_use_bset_p): Ditto.
4919 (nds32_can_use_btgl_p): Ditto.
4920 (nds32_can_use_bitci_p): Ditto.
4922 2018-08-12 Chung-Ju Wu <jasonwucj@gmail.com>
4924 * config/nds32/nds32.c (nds32_expand_prologue, nds32_expand_epilogue):
4925 Support -msched-prolog-epilog option.
4926 * config/nds32/nds32.opt (msched-prolog-epilog): New option.
4928 2018-08-12 Chung-Ju Wu <jasonwucj@gmail.com>
4930 * common/config/nds32/nds32-common.c
4931 (nds32_option_optimization_table): Enalbe -malways-align.
4933 2018-08-12 Chung-Ju Wu <jasonwucj@gmail.com>
4935 * config.gcc (nds32*): Add nds32_isr.h and nds32_init.inc in
4937 * common/config/nds32/nds32-common.c (nds32_handle_option): Handle
4938 OPT_misr_secure_ case.
4939 * config/nds32/nds32-isr.c: Implementation of backward compatibility.
4940 * config/nds32/nds32-protos.h (nds32_isr_function_critical_p): New.
4941 * config/nds32/nds32.c (nds32_attribute_table): Add critical and
4943 * config/nds32/nds32.h (nds32_isr_nested_type): Add NDS32_CRITICAL.
4944 (nds32_isr_info): New field security_level.
4945 (TARGET_ISR_VECTOR_SIZE_4_BYTE): New macro.
4946 * config/nds32/nds32.md (return_internal): Consider critical attribute.
4947 * config/nds32/nds32.opt (misr-secure): New option.
4948 * config/nds32/nds32_init.inc: New file.
4949 * config/nds32/nds32_isr.h: New file.
4951 2018-08-11 John David Anglin <danglin@gcc.gnu.org>
4953 * config/pa/pa.md (UNSPEC_MEMORY_BARRIER): New unspec enum.
4954 Update comment for atomic instructions.
4955 (atomic_storeqi, atomic_storehi, atomic_storesi, atomic_storesf,
4956 atomic_loaddf, atomic_loaddf_1, atomic_storedf, atomic_storedf_1):
4958 (atomic_loaddi): Revise fence expansion to only emit fence prior to
4959 load for __ATOMIC_SEQ_CST model.
4960 (atomic_loaddi_1): Remove float register target.
4961 (atomic_storedi): Handle CONST_INT values.
4962 (atomic_storedi_1): Remove float register source. Add special case
4964 (memory_barrier): New expander and insn.
4966 2018-08-11 Jakub Jelinek <jakub@redhat.com>
4968 PR tree-optimization/86835
4969 * tree-ssa-math-opts.c (insert_reciprocals): Even when inserting
4970 new_stmt after def_gsi, make sure to insert new_square_stmt after
4971 that stmt, not 2 stmts before it.
4973 2018-08-10 Alexander Monakov <amonakov@ispras.ru>
4976 * config/i386/i386.md (<s>mul<mode>3_highpart): Use DWIH mode iterator
4979 2018-08-10 Martin Liska <mliska@suse.cz>
4982 * builtin-types.def (BT_FN_LONG_LONG_LONG_DOUBLE): Add new
4984 * builtins.c (expand_builtin_expect_with_probability):
4986 (expand_builtin_expect_with_probability): New function.
4987 (build_builtin_expect_predicate): Add new argumnet probability
4988 for BUILT_IN_EXPECT_WITH_PROBABILITY.
4989 (fold_builtin_expect):
4992 * builtins.def (BUILT_IN_EXPECT_WITH_PROBABILITY):
4993 * builtins.h (fold_builtin_expect): Set new argument.
4994 * doc/extend.texi: Document __builtin_expect_with_probability.
4995 * doc/invoke.texi: Likewise.
4996 * gimple-fold.c (gimple_fold_call): Pass new argument.
4997 * ipa-fnsummary.c (find_foldable_builtin_expect): Handle
4998 also BUILT_IN_EXPECT_WITH_PROBABILITY.
4999 * predict.c (get_predictor_value): New function.
5000 (expr_expected_value): Add new argument probability. Assume
5001 that predictor and probability are always non-null.
5002 (expr_expected_value_1): Likewise. For __builtin_expect and
5003 __builtin_expect_with_probability set probability. Handle
5004 combination in binary expressions.
5005 (tree_predict_by_opcode): Simplify code by simply calling
5006 get_predictor_value.
5007 (pass_strip_predict_hints::execute): Add handling of
5008 BUILT_IN_EXPECT_WITH_PROBABILITY.
5009 * predict.def (PRED_BUILTIN_EXPECT_WITH_PROBABILITY): Add
5011 * tree.h (DECL_BUILT_IN_P): New function.
5013 2018-08-10 Martin Liska <mliska@suse.cz>
5015 PR tree-optimization/85799
5016 * passes.def: Add argument for pass_strip_predict_hints.
5017 * predict.c (class pass_strip_predict_hints): Add new argument
5019 (strip_predictor_early): New function.
5020 (pass_strip_predict_hints::execute): Call the function to
5022 (strip_predict_hints): New function.
5023 * predict.def: Fix comment.
5025 2018-08-10 Thomas Preud'homme <thomas.preudhomme@linaro.org>
5027 * Makefile.in: Clarify which tm.texi to copy over to assert the
5028 right to grant a GFDL license for all.
5030 2018-08-09 Jeff Law <law@redhat.com>
5032 * config/m68k/m68k.c (m68k_adjust_decorated_operand): Remove
5035 2018-08-09 Andreas Schwab <schwab@linux-m68k.org>
5037 * config/m68k/m68k-protos.h (m68k_final_prescan_insn): Remove
5040 2018-08-09 Richard Sandiford <richard.sandiford@arm.com>
5042 * tree-vect-loop.c (vectorizable_reduction): Allow inner-loop
5043 reductions for variable-length vectors.
5045 2018-08-09 David Malcolm <dmalcolm@redhat.com>
5048 * common.opt (fdiagnostics-show-line-numbers): New option.
5049 * diagnostic-show-locus.c (class layout): Add fields
5050 "m_show_line_numbers_p" and "m_linenum_width";
5051 (num_digits): New function.
5052 (test_num_digits): New function.
5053 (layout::layout): Initialize new fields. Update m_x_offset
5054 logic to handle any left margin.
5055 (layout::print_source_line): Print line number when requested.
5056 (layout::start_annotation_line): New member function.
5057 (layout::print_annotation_line): Call it.
5058 (layout::print_leading_fixits): Likewise.
5059 (layout::print_trailing_fixits): Likewise. Update calls to
5060 move_to_column for new parameter.
5061 (layout::get_x_bound_for_row): Add "add_left_margin" param and use
5062 it to potentially call start_annotation_line.
5063 (layout::show_ruler): Call start_annotation_line.
5064 (selftest::test_line_numbers_multiline_range): New selftest.
5065 (selftest::diagnostic_show_locus_c_tests): Call test_num_digits
5066 and selftest::test_line_numbers_multiline_range.
5067 * diagnostic.c (diagnostic_initialize): Initialize
5068 show_line_numbers_p.
5069 * diagnostic.h (struct diagnostic_context): Add field
5070 "show_line_numbers_p".
5071 * doc/invoke.texi (Diagnostic Message Formatting Options): Add
5072 -fno-diagnostics-show-line-numbers.
5073 * dwarf2out.c (gen_producer_string): Add
5074 OPT_fdiagnostics_show_line_numbers to the ignored options.
5075 * lto-wrapper.c (merge_and_complain): Likewise to the "pick
5076 one setting" options.
5077 (append_compiler_options): Likewise to the dropped options.
5078 (append_diag_options): Likewise to the passed-on options.
5079 * opts.c (common_handle_option): Handle the new option.
5080 * toplev.c (general_init): Set up global_dc->show_line_numbers_p.
5082 2018-08-09 Kelvin Nilsen <kelvin@gcc.gnu.org>
5084 * doc/extend.texi (PowerPC AltiVec Built-in Functions Available on
5085 ISA 2.07): Correct spelling of bcdsub to be __builtin_bcdsub. Add
5086 third argument of type "const signed char" to descriptions of
5087 __builtin_bcdadd, __builtin_bcdadd_lt, __builtin_bcdadd_eq,
5088 __builtin_bcdadd_gt, __builtin_bcdadd_ov, __builtin_bcdsub,
5089 __builtin_bcdsub_lt, __builtin_bcdsub_eq, __builtin_bcdsub_gt,
5090 __builtin_bcdsub_ov functions.
5092 2018-08-09 Richard Sandiford <richard.sandiford@arm.com>
5094 PR tree-optimization/86858
5095 * tree-vect-loop.c (vect_is_simple_reduction): Restore
5096 flow_bb_inside_loop_p calls.
5098 2018-08-09 Richard Sandiford <richard.sandiford@arm.com>
5100 PR tree-optimization/86871
5101 * tree-vect-stmts.c (vect_transform_stmt): Use gimple_get_lhs
5102 instead of gimple_assign_lhs.
5104 2018-08-09 Richard Earnshaw <rearnsha@arm.com>
5107 * config/aarch64/aarch64.md (add<mode>3_carryinC_zero): Add missing
5108 register constraint to operand 0.
5109 (add<mode>3_carryinC): Likewise.
5110 (add<mode>3_carryinV_zero, add<mode>3_carryinV): Likewise.
5112 2018-08-09 Martin Liska <mliska@suse.cz>
5115 * common.opt: Remove extra line.
5117 2018-08-09 Martin Liska <mliska@suse.cz>
5119 * params.def (PARAM_ALIGN_LOOP_ITERATIONS): Remove double dots
5120 at the end of a line, make first letter capital and end up
5121 a sentence with a dot.
5122 (PARAM_LOOP_INTERCHANGE_STRIDE_RATIO): Likewise.
5123 (PARAM_LOOP_BLOCK_TILE_SIZE): Likewise.
5124 (PARAM_GRAPHITE_MAX_NB_SCOP_PARAMS): Likewise.
5125 (PARAM_GRAPHITE_MAX_ARRAYS_PER_SCOP): Likewise.
5126 (PARAM_MAX_ISL_OPERATIONS): Likewise.
5127 (PARAM_GRAPHITE_ALLOW_CODEGEN_ERRORS): Likewise.
5128 (PARAM_PROFILE_FUNC_INTERNAL_ID): Likewise.
5129 (PARAM_INDIR_CALL_TOPN_PROFILE): Likewise.
5130 (PARAM_SLP_MAX_INSNS_IN_BB): Likewise.
5131 (PARAM_IPA_CP_EVAL_THRESHOLD): Likewise.
5132 (PARAM_IPA_CP_RECURSION_PENALTY): Likewise.
5133 (PARAM_IPA_CP_SINGLE_CALL_PENALTY): Likewise.
5134 (PARAM_IPA_CP_LOOP_HINT_BONUS): Likewise.
5135 (PARAM_IPA_CP_ARRAY_INDEX_HINT_BONUS): Likewise.
5136 (PARAM_TREE_REASSOC_WIDTH): Likewise.
5137 (PARAM_HSA_GEN_DEBUG_STORES): Likewise.
5138 (PARAM_MAX_SPECULATIVE_DEVIRT_MAYDEFS): Likewise.
5139 (PARAM_MAX_VRP_SWITCH_ASSERTIONS): Likewise.
5141 2018-08-09 Andreas Krebbel <krebbel@linux.ibm.com>
5144 * config/s390/s390.c (s390_option_override_internal): Reduce the
5145 stack-clash-protection-probe-interval param if it would be too big
5148 2018-08-08 Andreas Schwab <schwab@linux-m68k.org>
5151 * config/m68k/m68k.h (FINAL_PRESCAN_INSN): Don't define.
5152 * config/m68k/m68k.c (handle_move_double): Don't call
5153 m68k_final_prescan_insn.
5154 (m68k_adjust_decorated_operand): Renamed from
5155 m68k_final_prescan_insn, remove first and third operand and
5157 (print_operand): Call it.
5158 (print_operand_address): Call it.
5160 2018-08-08 Nathan Sidwell <nathan@acm.org>
5162 * diagnostic.c (diagnostic_report_current_module): Use
5163 linemap_included_from & linemap_included_from_linemap.
5165 2018-08-08 Hongbo Zhang <hongbo.zhang@linaro.org>
5167 * config/aarch64/aarch64-cores.def: Add phecda core.
5168 * config/aarch64/aarch64-tune.md: Regenerate.
5169 * doc/invoke.texi: Add phecda core.
5171 2018-08-08 Andreas Krebbel <krebbel@linux.ibm.com>
5174 * config/s390/constraints.md ("NxHD0", "NxSD0"): New constraint
5176 * config/s390/s390.md ("movti"): Add more alternatives for
5177 constant to GPR copies.
5179 2018-08-08 Andreas Krebbel <krebbel@linux.ibm.com>
5181 * config/s390/s390.c: Fix whitespace damage throughout the file.
5182 * config/s390/s390.h: Likewise.
5183 * config/s390/tpf.h: Likewise.
5185 2018-08-08 Ilya Leoshkevich <iii@linux.ibm.com>
5187 * config/s390/s390.c (s390_loadrelative_operand_p):
5188 Remove TARGET_CPU_ZARCH usages.
5189 (s390_rtx_costs): Likewise.
5190 (s390_legitimate_constant_p): Likewise.
5191 (s390_cannot_force_const_mem): Likewise.
5192 (legitimate_reload_constant_p): Likewise.
5193 (s390_preferred_reload_class): Likewise.
5194 (legitimize_pic_address): Likewise.
5195 (legitimize_tls_address): Likewise.
5196 (s390_split_branches): Removed.
5197 (s390_add_execute): Removed.
5198 (s390_dump_pool): Remove TARGET_CPU_ZARCH usages.
5199 (s390_mainpool_start): Likewise.
5200 (s390_mainpool_finish): Likewise.
5201 (s390_mainpool_cancel): Removed.
5202 (s390_chunkify_start): Remove TARGET_CPU_ZARCH usages.
5203 (s390_chunkify_cancel): Likewise.
5204 (s390_return_addr_rtx): Likewise.
5205 (s390_register_info): Remove split_branches_pending_p uages.
5206 (s390_optimize_register_info): Likewise.
5207 (s390_init_frame_layout): Remove TARGET_CPU_ZARCH and
5208 split_branches_pending_p usages.
5209 (s390_can_eliminate): Remove TARGET_CPU_ZARCH usages.
5210 (s390_load_got): Likewise.
5211 (s390_expand_split_stack_prologue): Likewise.
5212 (output_asm_nops): Likewise.
5213 (s390_function_profiler): Likewise.
5214 (s390_emit_call): Likewise.
5215 (s390_conditional_register_usage): Likewise.
5216 (s390_optimize_prologue): Likewise.
5217 (s390_reorg): Remove TARGET_CPU_ZARCH and
5218 split_branches_pending_p usages.
5219 (s390_option_override_internal): Remove TARGET_CPU_ZARCH
5221 (s390_output_indirect_thunk_function): Likewise.
5222 * config/s390/s390.h (TARGET_CPU_ZARCH): Removed.
5223 (TARGET_CPU_ZARCH_P): Removed.
5224 (struct machine_function): Remove split_branches_pending_p.
5225 * config/s390/s390.md: Remove TARGET_CPU_ZARCH usages.
5227 2018-08-08 Ilya Leoshkevich <iii@linux.ibm.com>
5229 * common/config/s390/s390-common.c (processor_flags_table):
5231 * config.gcc: Remove with_arch/with_tune support.
5232 * config/s390/2064.md: Remove cpu attribute comparisons.
5233 * config/s390/driver-native.c (s390_host_detect_local_cpu):
5235 * config/s390/linux.h (ASM_SPEC):
5236 Remove -march support.
5237 * config/s390/s390-c.c (s390_cpu_cpp_builtins_internal):
5238 Use a table to get an arch level.
5239 * config/s390/s390-opts.h (enum processor_type):
5241 * config/s390/s390.c
5242 (processor_table): Remove entries, add arch_level values.
5243 (s390_issue_rate): Remove cases.
5244 (s390_option_override): Adjust
5245 s390_option_override_internal() call.
5246 (s390_option_override_internal): Remove deprecation warning.
5247 (s390_valid_target_attribute_tree): Adjust
5248 s390_option_override_internal() call.
5249 * config/s390/s390.h (struct s390_processor):
5250 Share with s390-c.c, add arch_level field.
5251 * config/s390/s390.md:
5252 Remove occurrences in cpu attribute.
5253 * config/s390/s390.opt: Remove -march/-mtune support.
5254 * config/s390/tpf.h (ASM_SPEC): Remove -march support.
5255 * doc/invoke.texi: Remove deprecation warning.
5257 2018-08-08 Luis Machado <luis.machado@linaro.org>
5259 * config/aarch64/aarch64.c (qdf24xx_vector_cost): New static
5261 (qdf24xx_tunings): Set vector cost structure to
5262 qdf24xx_vector_cost.
5264 * config/aarch64/aarch64.c (qdf24xx_addrcost_table)
5265 <register_sextend>: Set to 3.
5267 2018-08-07 Richard Sandiford <richard.sandiford@arm.com>
5270 * config/aarch64/iterators.md (FRECP, frecp_suffix): Delete.
5271 * config/aarch64/aarch64-simd.md
5272 (aarch64_frecp<FRECP:frecp_suffix><mode>): Fold FRECPE into...
5273 (@aarch64_frecpe<mode>): ...here and the move FRECPX to...
5274 (aarch64_frecpx<mode>): ...this new pattern.
5275 * config/aarch64/aarch64-simd-builtins.def: Remove comment
5276 about aarch64_frecp<FRECP:frecp_suffix><mode>.
5278 2018-08-07 Martin Liska <mliska@suse.cz>
5281 * predict.c (expr_expected_value_1): Handle DECL_IS_MALLOC,
5282 BUILT_IN_REALLOC and DECL_IS_OPERATOR_NEW.
5283 * predict.def (PRED_MALLOC_NONNULL): New predictor.
5284 * doc/extend.texi: Document that malloc attribute adds
5287 2018-08-06 John David Anglin <danglin@gcc.gnu.org>
5290 * config/pa/pa.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
5291 Define to speculation_safe_value_not_needed.
5293 2018-08-06 Jeff Law <law@redhat.com>
5295 * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Pass down
5296 the vr_values instance to cprop_into_stmt.
5297 (cprop_into_stmt): Pass vr_values instance down to cprop_operand.
5298 (cprop_operand): Also query EVRP to determine if OP is a constant.
5300 2018-08-06 Nathan Sidwell <nathan@acm.org>
5302 * diagnostic.c (diagnostic_report_current_module): Reroll
5303 included-at loop. Translate text.
5305 2018-08-06 David Malcolm <dmalcolm@redhat.com>
5307 * function-tests.c (selftest::test_expansion_to_rtl): Call
5308 free_after_compilation.
5310 2018-08-06 Alan Hayward <alan.hayward@arm.com>
5312 * config/aarch64/aarch64.md: Add clobber highs to tls_desc.
5314 2018-08-06 Andreas Krebbel <krebbel@linux.ibm.com>
5316 * config/s390/s390.c (s390_loop_unroll_adjust): Prevent small
5317 loops with memory block operations from getting unrolled.
5319 2018-08-06 Ulrich Weigand <uweigand@de.ibm.com>
5322 * config/spu/spu.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
5323 Define to speculation_safe_value_not_needed.
5325 2018-08-06 Jeff Law <law@redhat.com>
5327 * reload1.c (forget_old_reloads_1): Adjust CLOBBER_HIGH
5330 2018-08-06 Jozef Lawrynowicz <jozef.l@mittosystems.com>
5333 * gcc/tree.c (build_common_tree_nodes): Initialize integer_types array
5334 with all enabled __intN types.
5336 * gcc/testsuite/gcc.target/msp430/pr86662.c: New test.
5338 2018-08-06 Alan Hayward <alan.hayward@arm.com>
5340 * alias.c (record_set): Check for clobber high.
5341 * cfgexpand.c (expand_gimple_stmt): Likewise.
5342 * combine-stack-adj.c (single_set_for_csa): Likewise.
5343 * combine.c (find_single_use_1): Likewise.
5344 (set_nonzero_bits_and_sign_copies): Likewise.
5345 (get_combine_src_dest): Likewise.
5346 (is_parallel_of_n_reg_sets): Likewise.
5347 (try_combine): Likewise.
5348 (record_dead_and_set_regs_1): Likewise.
5349 (reg_dead_at_p_1): Likewise.
5350 (reg_dead_at_p): Likewise.
5351 * dce.c (deletable_insn_p): Likewise.
5352 (mark_nonreg_stores_1): Likewise.
5353 (mark_nonreg_stores_2): Likewise.
5354 * df-scan.c (df_find_hard_reg_defs): Likewise.
5355 (df_uses_record): Likewise.
5356 (df_get_call_refs): Likewise.
5357 * dwarf2out.c (mem_loc_descriptor): Likewise.
5358 * haifa-sched.c (haifa_classify_rtx): Likewise.
5359 * ira-build.c (create_insn_allocnos): Likewise.
5360 * ira-costs.c (scan_one_insn): Likewise.
5361 * ira.c (equiv_init_movable_p): Likewise.
5362 (rtx_moveable_p): Likewise.
5363 (interesting_dest_for_shprep): Likewise.
5364 * jump.c (mark_jump_label_1): Likewise.
5365 * postreload-gcse.c (record_opr_changes): Likewise.
5366 * postreload.c (reload_cse_simplify): Likewise.
5367 (struct reg_use): Add source expr.
5368 (reload_combine): Check for clobber high.
5369 (reload_combine_note_use): Likewise.
5370 (reload_cse_move2add): Likewise.
5371 (move2add_note_store): Likewise.
5372 * print-rtl.c (print_pattern): Likewise.
5373 * recog.c (decode_asm_operands): Likewise.
5374 (store_data_bypass_p): Likewise.
5375 (if_test_bypass_p): Likewise.
5376 * regcprop.c (kill_clobbered_value): Likewise.
5377 (kill_set_value): Likewise.
5378 * reginfo.c (reg_scan_mark_refs): Likewise.
5379 * reload1.c (maybe_fix_stack_asms): Likewise.
5380 (eliminate_regs_1): Likewise.
5381 (elimination_effects): Likewise.
5382 (mark_not_eliminable): Likewise.
5383 (scan_paradoxical_subregs): Likewise.
5384 (forget_old_reloads_1): Likewise.
5385 * reorg.c (find_end_label): Likewise.
5386 (try_merge_delay_insns): Likewise.
5387 (redundant_insn): Likewise.
5388 (own_thread_p): Likewise.
5389 (fill_simple_delay_slots): Likewise.
5390 (fill_slots_from_thread): Likewise.
5391 (dbr_schedule): Likewise.
5392 * resource.c (update_live_status): Likewise.
5393 (mark_referenced_resources): Likewise.
5394 (mark_set_resources): Likewise.
5395 * rtl.c (copy_rtx): Likewise.
5396 * rtlanal.c (reg_referenced_p): Likewise.
5397 (single_set_2): Likewise.
5398 (noop_move_p): Likewise.
5399 (note_stores): Likewise.
5400 * sched-deps.c (sched_analyze_reg): Likewise.
5401 (sched_analyze_insn): Likewise.
5403 2018-08-06 Alan Hayward <alan.hayward@arm.com>
5405 * cse.c (invalidate_reg): New function extracted from...
5406 (invalidate): ...here.
5407 (canonicalize_insn): Check for clobber high.
5408 (invalidate_from_clobbers): invalidate clobber highs.
5409 (invalidate_from_sets_and_clobbers): Likewise.
5410 (count_reg_usage): Check for clobber high.
5411 (insn_live_p): Likewise.
5412 * cselib.c (cselib_expand_value_rtx_1):Likewise.
5413 (cselib_invalidate_regno): Check for clobber in setter.
5414 (cselib_invalidate_rtx): Pass through setter.
5415 (cselib_invalidate_rtx_note_stores):
5416 (cselib_process_insn): Check for clobber high.
5417 * cselib.h (cselib_invalidate_rtx): Add operand.
5419 2018-08-06 Alan Hayward <alan.hayward@arm.com>
5421 * lra-eliminations.c (lra_eliminate_regs_1): Check for clobber high.
5422 (mark_not_eliminable): Likewise.
5423 * lra-int.h (struct lra_insn_reg): Add clobber high marker.
5424 * lra-lives.c (process_bb_lives): Check for clobber high.
5425 * lra.c (new_insn_reg): Remember clobber highs.
5426 (collect_non_operand_hard_regs): Check for clobber high.
5427 (lra_set_insn_recog_data): Likewise.
5428 (add_regs_to_insn_regno_info): Likewise.
5429 (lra_update_insn_regno_info): Likewise.
5431 2018-08-06 Alan Hayward <alan.hayward@arm.com>
5433 * rtl.h (reg_is_clobbered_by_clobber_high): Add declarations.
5434 * rtlanal.c (reg_is_clobbered_by_clobber_high): Add function.
5436 2018-08-06 Alan Hayward <alan.hayward@arm.com>
5438 * emit-rtl.c (verify_rtx_sharing): Check for CLOBBER_HIGH.
5439 (copy_insn_1): Likewise.
5440 (gen_hard_reg_clobber_high): New gen function.
5441 * genconfig.c (walk_insn_part): Check for CLOBBER_HIGH.
5442 * genemit.c (gen_exp): Likewise.
5443 (gen_emit_seq): Pass through info.
5444 (gen_insn): Check for CLOBBER_HIGH.
5445 (gen_expand): Pass through info.
5446 (gen_split): Likewise.
5447 (output_add_clobbers): Likewise.
5448 * genrecog.c (validate_pattern): Check for CLOBBER_HIGH.
5449 (remove_clobbers): Likewise.
5450 * rtl.h (gen_hard_reg_clobber_high): New declaration.
5452 2018-08-06 Alan Hayward <alan.hayward@arm.com>
5454 * doc/rtl.texi (clobber_high): Add.
5455 (parallel): Add in clobber high
5456 * rtl.c (rtl_check_failed_code3): Add function.
5457 * rtl.def (CLOBBER_HIGH): Add expression.
5458 * rtl.h (RTL_CHECKC3): Add macro.
5459 (rtl_check_failed_code3): Add declaration.
5460 (XC3EXP): Add macro.
5462 2018-08-05 H.J. Lu <hongjiu.lu@intel.com>
5465 * config/i386/i386.c (ix86_finalize_stack_frame_flags): Set
5466 cfun->machine->max_used_stack_alignment if needed.
5468 2018-08-04 Martin Sebor <msebor@redhat.com>
5470 PR tree-optimization/86571
5471 * gimple-ssa-sprintf.c (format_floating): Extend upper bound of
5474 2018-08-03 Sandra Loosemore <sandra@codesourcery.com>
5477 * config/nios2/nios2.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
5480 2018-08-03 Jeff Law <law@redhat.com>
5483 * config/mn10300/mn10300.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
5484 Define to speculation_safe_value_not_needed.
5486 2018-08-03 David Malcolm <dmalcolm@redhat.com>
5488 * doc/gcov.texi (-x): Remove duplicate "to".
5489 * doc/invoke.texi (-Wnoexcept-type): Remove duplicate "calls".
5490 (-Wif-not-aligned): Remove duplicate "is".
5491 (-flto): Remove duplicate "the".
5492 (MicroBlaze Options): In examples of "-mcpu=cpu-type", remove
5493 duplicate "v5.00.b".
5494 (MSP430 Options): Remove duplicate "and" from the description
5495 of "-mgprel-sec=regexp".
5496 (x86 Options): Remove duplicate copies of "vmldLog102" and
5497 vmlsLog104 from description of "-mveclibabi=type".
5499 2018-08-03 Richard Sandiford <richard.sandiford@arm.com>
5501 * internal-fn.h (first_commutative_argument): Declare.
5502 * internal-fn.c (first_commutative_argument): New function.
5503 * tree-vect-slp.c (vect_get_and_check_slp_defs): Remove extra
5504 restrictions for pattern statements. Use first_commutative_argument
5505 to look for commutative operands in calls to internal functions.
5507 2018-08-03 Aldy Hernandez <aldyh@redhat.com>
5509 * Makefile.in (wide-int-range.o): New.
5510 * tree-vrp.c: Move all the wide_int_* functions to...
5511 * wide-int-range.cc: ...here.
5512 * tree-vrp.h: Move all the wide_int_* prototypes to...
5513 * wide-int-range.h: ...here.
5515 2018-08-03 Tom de Vries <tdevries@suse.de>
5517 * common/config/nvptx/nvptx-common.c (nvptx_except_unwind_info): Return
5519 * config/nvptx/nvptx.c (TARGET_ASM_BYTE_OP): Remove define.
5520 * except.c (output_function_exception_table): Do early exit if
5521 targetm_common.except_unwind_info (&global_options) == UI_NONE.
5523 2018-08-03 Martin Liska <mliska@suse.cz>
5525 * predict.c (dump_prediction): Change to 2 digits
5528 2018-08-03 Siddhesh Poyarekar <siddhesh@sourceware.org>
5530 * config/aarch64/falkor.md (falkor_am_1_vxvy_vxvy): Move
5532 (falkor_am_1_gtov_gtov): ... a new insn reservation.
5534 2018-07-19 Ilya Leoshkevich <iii@linux.ibm.com>
5536 * config/nds32/nds32.c (nds32_hard_regno_mode_ok): Replace > with >=.
5537 * df-problems.c (df_remove_dead_eq_notes): Replace > with >=.
5538 * dwarf2out.c (mem_loc_descriptor): Replace > with >=.
5539 * lra-constraints.c (spill_hard_reg_in_range): Replace <= with <.
5540 * lra-remat.c (call_used_input_regno_present_p): Replace <= with <.
5542 2018-08-02 David Malcolm <dmalcolm@redhat.com>
5544 * diagnostic-show-locus.c (diagnostic_show_locus): Use
5545 pp_take_prefix when saving the existing prefix.
5546 * diagnostic.c (diagnostic_append_note): Likewise.
5547 * langhooks.c (lhd_print_error_function): Likewise.
5548 * pretty-print.c (pp_set_prefix): Drop the "const" from "prefix"
5549 param's type. Free the existing prefix.
5550 (pp_take_prefix): New function.
5551 (pretty_printer::pretty_printer): Drop the prefix parameter.
5552 Rename the length parameter to match the comment.
5553 (pretty_printer::~pretty_printer): Free the prefix.
5554 * pretty-print.h (pretty_printer::pretty_printer): Drop the prefix
5556 (struct pretty_printer): Drop the "const" from "prefix" field's
5557 type and clarify memory management.
5558 (pp_set_prefix): Drop the "const" from the 2nd param.
5559 (pp_take_prefix): New decl.
5561 2018-08-02 Aaron Sawdey <acsawdey@linux.ibm.com>
5563 * config/rs6000/rs6000-string.c (select_block_compare_mode): Move test
5564 for word_mode_ok here instead of passing as argument.
5565 (expand_block_compare): Change select_block_compare_mode() call.
5566 (expand_strncmp_gpr_sequence): New function.
5567 (expand_strn_compare): Make use of expand_strncmp_gpr_sequence.
5569 2018-08-02 Jeff Law <law@redhat.com>
5572 * config/m68k/m68k.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
5573 Define to speculation_safe_value_not_needed.
5576 * config/h8300/h8300.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
5577 Define to speculation_safe_value_not_needed.
5579 2018-08-02 Tom de Vries <tdevries@suse.de>
5582 * common/config/nvptx/nvptx-common.c (nvptx_except_unwind_info): New
5583 function. Return UI_TARGET unconditionally.
5584 (TARGET_EXCEPT_UNWIND_INFO): Redefine to nvptx_except_unwind_info.
5585 * config/nvptx/nvptx.c (TARGET_ASM_BYTE_OP): Emit commented out '.byte'.
5587 2018-08-02 Richard Sandiford <richard.sandiford@arm.com>
5589 * genemit.c (print_overload_test): Fix typo.
5591 2018-08-02 Richard Biener <rguenther@suse.de>
5593 PR tree-optimization/86816
5594 * tree-ssa-tail-merge.c (tail_merge_valueize): New function
5595 which checks for value availability before querying it.
5596 (gvn_uses_equal): Use it.
5597 (same_succ_hash): Likewise.
5598 (gimple_equal_p): Likewise.
5600 2018-08-02 Nick Clifton <nickc@redhat.com>
5603 * config/stormy16/stormy16.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
5604 Define to speculation_safe_value_not_needed.
5607 * config/v850/v850.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
5608 Define to speculation_safe_value_not_needed.
5611 * config/v850/v850.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
5612 Define to speculation_safe_value_not_needed.
5615 * config/rx/rx.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
5616 Define to speculation_safe_value_not_needed.
5619 * config/msp430/msp430.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
5620 Define to speculation_safe_value_not_needed.
5623 * config/mcore/mcore.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
5624 Define to speculation_safe_value_not_needed.
5627 * config/m32r/m32r.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
5628 Define to speculation_safe_value_not_needed.
5631 * config/iq2000/iq2000.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
5632 Define to speculation_safe_value_not_needed.
5635 * config/frv/frv.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define to
5636 speculation_safe_value_not_needed.
5639 * config/fr30/fr30.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define
5640 to speculation_safe_value_not_needed.
5642 2018-08-02 Richard Sandiford <richard.sandiford@arm.com>
5644 * doc/md.texi: Expand the documentation of instruction names
5645 to mention port-local uses. Document '@' in pattern names.
5646 * read-md.h (overloaded_instance, overloaded_name): New structs.
5648 (md_reader::handle_overloaded_name): New member function.
5649 (md_reader::get_overloads): Likewise.
5650 (md_reader::m_first_overload): New member variable.
5651 (md_reader::m_next_overload_ptr): Likewise.
5652 (md_reader::m_overloads_htab): Likewise.
5653 * read-md.c (md_reader::md_reader): Initialize m_first_overload,
5654 m_next_overload_ptr and m_overloads_htab.
5655 * read-rtl.c (iterator_group): Add "type" and "get_c_token" fields.
5656 (get_mode_token, get_code_token, get_int_token): New functions.
5657 (map_attr_string): Add an optional argument that passes back
5658 the associated iterator.
5659 (overloaded_name_hash, overloaded_name_eq_p, named_rtx_p):
5660 (md_reader::handle_overloaded_name, add_overload_instance): New
5662 (apply_iterators): Handle '@' names. Report an error if '@'
5663 is used without iterators.
5664 (initialize_iterators): Initialize the new iterator_group fields.
5665 * genopinit.c (handle_overloaded_code_for)
5666 (handle_overloaded_gen): New functions.
5667 (main): Use them to print declarations of maybe_code_for_* and
5668 maybe_gen_* functions, and inline definitions of code_for_* and gen_*.
5669 * genemit.c (print_overload_arguments, print_overload_test)
5670 (handle_overloaded_code_for, handle_overloaded_gen): New functions.
5671 (main): Use it to print definitions of maybe_code_for_* and
5672 maybe_gen_* functions.
5673 * config/aarch64/aarch64.c (aarch64_split_128bit_move): Use
5674 gen_aarch64_mov{low,high}_di and gen_aarch64_movdi_{low,high}
5675 instead of explicit mode checks.
5676 (aarch64_split_simd_combine): Likewise gen_aarch64_simd_combine.
5677 (aarch64_split_simd_move): Likewise gen_aarch64_split_simd_mov.
5678 (aarch64_emit_load_exclusive): Likewise gen_aarch64_load_exclusive.
5679 (aarch64_emit_store_exclusive): Likewise gen_aarch64_store_exclusive.
5680 (aarch64_expand_compare_and_swap): Likewise
5681 gen_aarch64_compare_and_swap and gen_aarch64_compare_and_swap_lse
5682 (aarch64_gen_atomic_cas): Likewise gen_aarch64_atomic_cas.
5683 (aarch64_emit_atomic_swap): Likewise gen_aarch64_atomic_swp.
5684 (aarch64_constant_pool_reload_icode): Delete.
5685 (aarch64_secondary_reload): Use code_for_aarch64_reload_movcp
5686 instead of aarch64_constant_pool_reload_icode. Use
5687 code_for_aarch64_reload_mov instead of explicit mode checks.
5688 (rsqrte_type, get_rsqrte_type, rsqrts_type, get_rsqrts_type): Delete.
5689 (aarch64_emit_approx_sqrt): Use gen_aarch64_rsqrte instead of
5690 get_rsqrte_type and gen_aarch64_rsqrts instead of gen_rqrts_type.
5691 (recpe_type, get_recpe_type, recps_type, get_recps_type): Delete.
5692 (aarch64_emit_approx_div): Use gen_aarch64_frecpe instead of
5693 get_recpe_type and gen_aarch64_frecps instead of get_recps_type.
5694 (aarch64_atomic_load_op_code): Delete.
5695 (aarch64_emit_atomic_load_op): Likewise.
5696 (aarch64_gen_atomic_ldop): Use UNSPECV_ATOMIC_* instead of
5697 aarch64_atomic_load_op_code. Use gen_aarch64_atomic_load
5698 instead of aarch64_emit_atomic_load_op.
5699 * config/aarch64/aarch64.md (aarch64_reload_movcp<GPF_TF:mode><P:mode>)
5700 (aarch64_reload_movcp<VALL:mode><P:mode>, aarch64_reload_mov<mode>)
5701 (aarch64_movdi_<mode>low, aarch64_movdi_<mode>high)
5702 (aarch64_mov<mode>high_di, aarch64_mov<mode>low_di): Add a '@'
5703 character before the pattern name.
5704 * config/aarch64/aarch64-simd.md (aarch64_split_simd_mov<mode>)
5705 (aarch64_rsqrte<mode>, aarch64_rsqrts<mode>)
5706 (aarch64_simd_combine<mode>, aarch64_frecpe<mode>)
5707 (aarch64_frecps<mode>): Likewise.
5708 * config/aarch64/atomics.md (atomic_compare_and_swap<mode>)
5709 (aarch64_compare_and_swap<mode>, aarch64_compare_and_swap<mode>_lse)
5710 (aarch64_load_exclusive<mode>, aarch64_store_exclusive<mode>)
5711 (aarch64_atomic_swp<mode>, aarch64_atomic_cas<mode>)
5712 (aarch64_atomic_load<atomic_ldop><mode>): Likewise.
5714 2018-08-02 Richard Sandiford <richard.sandiford@arm.com>
5716 * config/aarch64/aarch64.c (aarch64_float_const_representable_p):
5717 Allow HFmode constants if TARGET_FP_F16INST.
5719 2018-08-02 Jackson Woodruff <jackson.woodruff@arm.com>
5722 * config/aarch64/aarch64.c (aarch64_operands_adjust_ok_for_ldpstp):
5723 No longer check last store for clobber of address register.
5725 2018-08-02 Martin Liska <mliska@suse.cz>
5727 PR gcov-profile/86817
5728 * gcov.c (process_all_functions): New function.
5730 (process_file): Move functions processing to
5731 process_all_functions.
5733 2018-08-02 David Malcolm <dmalcolm@redhat.com>
5735 * dumpfile.c (dump_user_location_t::dump_user_location_t): Add
5736 "const" to the "gimple *" and "rtx_insn *" parameters.
5737 * dumpfile.h (dump_user_location_t::dump_user_location_t):
5739 (dump_location_t::dump_location_t): Likewise.
5741 2018-08-01 Martin Sebor <msebor@redhat.com>
5743 PR tree-optimization/86650
5744 * gimple-pretty-print.c (percent_G_format): Accept a "gimple *"
5745 rather than a "gcall *". Directly pass the data of interest
5746 to percent_K_format, rather than building a temporary CALL_EXPR
5748 * gimple-fold.c (gimple_fold_builtin_strncpy): Adjust.
5749 (gimple_fold_builtin_strncat): Adjust.
5750 * gimple-ssa-warn-restrict.h (check_bounds_or_overlap): Replace
5751 gcall* argument with gimple*.
5752 * gimple-ssa-warn-restrict.c (check_call): Same.
5753 (wrestrict_dom_walker::before_dom_children): Same.
5754 (builtin_access::builtin_access): Same.
5755 (check_bounds_or_overlap): Same
5756 (maybe_diag_overlap): Same.
5757 (maybe_diag_offset_bounds): Same.
5758 * tree-diagnostic.c (default_tree_printer): Move usage of
5759 EXPR_LOCATION (t) and TREE_BLOCK (t) from within percent_K_format
5761 * tree-pretty-print.c (percent_K_format): Add argument.
5762 * tree-pretty-print.h: Add argument.
5763 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Adjust.
5764 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Adjust.
5765 (maybe_diag_stxncpy_trunc): Same.
5766 (handle_builtin_stxncpy): Same.
5767 (handle_builtin_strcat): Same.
5769 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
5771 * match.pd: Optimise pointer range checks.
5773 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
5775 PR tree-optimization/86758
5776 * tree-vect-stmts.c (vectorizable_simd_clone_call): Don't try
5777 to remove pattern statements.
5779 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
5781 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Use the
5782 result of dfs_enumerate_from when constructing stmt_vec_infos,
5783 instead of additionally calling get_loop_body.
5785 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
5787 * tree-vectorizer.h (vect_create_data_ref_ptr): Remove inv_p
5789 * tree-vect-data-refs.c (vect_create_data_ref_ptr): Likewise.
5790 When creating an iv, assert that the step is not known to be zero.
5791 (vect_setup_realignment): Update call accordingly.
5792 * tree-vect-stmts.c (vectorizable_store): Likewise.
5793 (vectorizable_load): Likewise. Handle VMAT_INVARIANT separately.
5795 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
5797 * tree-vectorizer.h (vect_stmt_to_vectorize): New function.
5798 * tree-vect-loop.c (vect_update_vf_for_slp): Use it.
5799 (vectorizable_reduction): Likewise.
5800 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
5801 (vect_detect_hybrid_slp_stmts): Likewise.
5802 * tree-vect-stmts.c (vect_is_simple_use): Likewise.
5804 2018-08-01 Aldy Hernandez <aldyh@redhat.com>
5806 * tree-vrp (zero_nonzero_bits_from_bounds): Rename to...
5807 (wide_int_set_zero_nonzero_bits): ...this.
5808 (zero_nonzero_bits_from_vr): Rename to...
5809 (vrp_set_zero_nonzero_bits): ...this.
5810 (extract_range_from_multiplicative_op_1): Abstract wide int
5812 (wide_int_range_multiplicative_op): ...here.
5813 (extract_range_from_binary_expr_1): Extract wide int binary
5814 operations into their own functions.
5815 (wide_int_range_lshift): New.
5816 (wide_int_range_can_optimize_bit_op): New.
5817 (wide_int_range_shift_undefined_p): New.
5818 (wide_int_range_bit_xor): New.
5819 (wide_int_range_bit_ior): New.
5820 (wide_int_range_bit_and): New.
5821 (wide_int_range_trunc_mod): New.
5822 (extract_range_into_wide_ints): New.
5823 (vrp_shift_undefined_p): New.
5824 (extract_range_from_multiplicative_op): New.
5825 (vrp_can_optimize_bit_op): New.
5826 * tree-vrp.h (value_range::dump): New.
5827 (wide_int_range_multiplicative_op): New.
5828 (wide_int_range_lshift):New.
5829 (wide_int_range_shift_undefined_p): New.
5830 (wide_int_range_bit_xor): New.
5831 (wide_int_range_bit_ior): New.
5832 (wide_int_range_bit_and): New.
5833 (wide_int_range_trunc_mod): New.
5834 (zero_nonzero_bits_from_bounds): Rename to...
5835 (wide_int_set_zero_nonzero_bits): ...this.
5836 (zero_nonzero_bits_from_vr): Rename to...
5837 (vrp_set_zero_nonzero_bits): ...this.
5838 (range_easy_mask_min_max): Rename to...
5839 (wide_int_range_can_optimize_bit_op): this.
5840 * vr-values.c (simplify_bit_ops_using_ranges): Rename
5841 zero_nonzero_bits_from_vr into vrp_set_zero_nonzero_bits.
5843 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
5845 * tree-vectorizer.h (vect_orig_stmt): New function.
5846 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Use it.
5847 * tree-vect-loop.c (vect_model_reduction_cost): Likewise.
5848 (vect_create_epilog_for_reduction): Likewise.
5849 (vectorizable_live_operation): Likewise.
5850 * tree-vect-slp.c (vect_find_last_scalar_stmt_in_slp): Likewise.
5851 (vect_detect_hybrid_slp_stmts, vect_schedule_slp): Likewise.
5852 * tree-vect-stmts.c (vectorizable_call): Likewise.
5853 (vectorizable_simd_clone_call, vect_remove_stores): Likewise.
5855 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
5857 * tree-vectorizer.h (vect_transform_stmt): Remove grouped_store
5859 * tree-vect-stmts.c (vect_transform_stmt): Likewise.
5860 * tree-vect-loop.c (vect_transform_loop_stmt): Update call accordingly.
5861 (vect_transform_loop): Likewise.
5862 * tree-vect-slp.c (vect_schedule_slp_instance): Likewise.
5864 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
5866 * tree-vectorizer.h (vect_schedule_slp): Return void.
5867 * tree-vect-slp.c (vect_schedule_slp_instance): Likewise.
5868 (vect_schedule_slp): Likewise.
5870 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
5872 * tree-vect-loop.c (vect_transform_loop_stmt): Remove slp_scheduled
5874 (vect_transform_loop): Update calls accordingly. Schedule SLP
5875 instances before the main loop, if any exist.
5877 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
5879 PR tree-optimization/86749
5880 * tree-vect-patterns.c (vect_determine_min_output_precision_1):
5881 If the lhs is used in a COND_EXPR, check that it is being used
5882 as the "then" or "else" value.
5884 2018-08-01 Tom de Vries <tdevries@suse.de>
5887 * config/nvptx/nvptx.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define to
5888 speculation_safe_value_not_needed.
5890 2018-08-01 Richard Biener <rguenther@suse.de>
5892 * tree-ssa-sccvn.c (visit_phi): Compare invariant addresses
5895 2018-08-01 Martin Liska <mliska@suse.cz>
5897 * value-prof.c (gimple_divmod_fixed_value_transform): Unify
5898 format how successful transformation is dumped.
5899 (gimple_mod_pow2_value_transform): Likewise.
5900 (gimple_mod_subtract_transform): Likewise.
5901 (gimple_stringops_transform): Likewise.
5903 2018-08-01 Martin Liska <mliska@suse.cz>
5906 * value-prof.c (interesting_stringop_to_profile_p):
5907 Simplify the code and add BUILT_IN_MEMMOVE.
5908 (gimple_stringops_transform): Likewise.
5910 2018-08-01 Sam Tebbs <sam.tebbs@arm.com>
5912 * config/aarch64/aarch64-simd.md
5913 (*aarch64_get_lane_zero_extendsi<mode>): Rename to...
5914 (*aarch64_get_lane_zero_extend<GPI:mode><VDQQH:mode>): ... This and
5915 use GPI iterator instead of SI mode.
5917 2018-08-01 Richard Earnshaw <rearnsha@arm.com>
5919 * config/rs6000/rs6000.md (speculation_barrier): Renamed from
5920 rs6000_speculation_barrier.
5921 * config/rs6000/rs6000.c (rs6000_expand_builtin): Adjust for
5922 new barrier pattern name.
5924 2018-08-01 Richard Earnshaw <rearnsha@arm.com>
5926 * config/i386/i386.md (unspecv): Add UNSPECV_SPECULATION_BARRIER.
5927 (speculation_barrier): New insn.
5929 2018-08-01 Richard Biener <rguenther@suse.de>
5932 * graphite.h: Include isl/id.h and isl/space.h to allow build
5935 2018-08-01 Jan Willem Jagersma <jwjagersma@gmail.com>
5938 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
5939 mode for COFF targets.
5940 * defaults.h (TARGET_COFF): Define.
5941 * config/i386/djgpp.h (TARGET_ASM_LTO_START, TARGET_ASM_LTO_END,
5942 TARGET_COFF): Define.
5943 (i386_djgpp_asm_lto_start, i386_djgpp_asm_lto_end): Declare.
5944 * config/i386/djgpp.c (saved_debug_info_level): New static variable.
5945 (i386_djgpp_asm_lto_start, i386_djgpp_asm_lto_end): New functions.
5947 2018-07-31 Alexandre Oliva <oliva@adacore.com>
5949 * gimple-streamer-in.c (input_bb): Restore BB discriminator.
5950 * gimple-streamer-out.c (output_bb): Save it.
5951 * lto-streamer-in.c (input_struct_function_base): Restore
5952 instance discriminator if available. Create map on demand.
5953 * lto-streamer-out.c (output_struct_function_base): Save it if
5955 * final.c (decl_to_instance_map): Document LTO strategy.
5957 2018-07-31 Alexandre Oliva <oliva@adacore.com>
5958 Olivier Hainque <hainque@adacore.com>
5960 * debug.h (decl_to_instance_map_t): New type.
5961 (decl_to_instance_map): Declare.
5962 (maybe_create_decl_to_instance_map): New inline function.
5963 * final.c (bb_discriminator, last_bb_discriminator): New statics,
5964 to track basic block discriminators.
5965 (final_start_function_1): Initialize them.
5966 (final_scan_insn_1): On NOTE_INSN_BASIC_BLOCK, track
5968 (decl_to_instance_map): New variable.
5969 (map_decl_to_instance, maybe_set_discriminator): New functions.
5970 (notice_source_line): Set discriminator.
5972 2018-07-31 Ian Lance Taylor <iant@golang.org>
5974 * targhooks.c (default_have_speculation_safe_value): Add
5977 2018-07-31 David Malcolm <dmalcolm@redhat.com>
5979 * dump-context.h: Include "pretty-print.h".
5980 (dump_context::refresh_dumps_are_enabled): New decl.
5981 (dump_context::emit_item): New decl.
5982 (class dump_context): Add fields "m_test_pp" and
5984 (temp_dump_context::temp_dump_context): Add param "test_pp_flags".
5985 (temp_dump_context::get_dumped_text): New decl.
5986 (class temp_dump_context): Add field "m_pp".
5987 * dumpfile.c (refresh_dumps_are_enabled): Convert to...
5988 (dump_context::refresh_dumps_are_enabled): ...and add a test for
5990 (set_dump_file): Update for above change.
5991 (set_alt_dump_file): Likewise.
5992 (dump_loc): New overload, taking a pretty_printer *.
5993 (dump_context::dump_loc): Call end_any_optinfo. Dump the location
5994 to any test pretty-printer.
5995 (make_item_for_dump_gimple_stmt): New function, adapted from
5996 optinfo::add_gimple_stmt.
5997 (dump_context::dump_gimple_stmt): Call it, and use the result,
5998 eliminating the direct usage of dump_file and alt_dump_file in
5999 favor of indirectly using them via emit_item.
6000 (make_item_for_dump_gimple_expr): New function, adapted from
6001 optinfo::add_gimple_expr.
6002 (dump_context::dump_gimple_expr): Call it, and use the result,
6003 eliminating the direct usage of dump_file and alt_dump_file in
6004 favor of indirectly using them via emit_item.
6005 (make_item_for_dump_generic_expr): New function, adapted from
6007 (dump_context::dump_generic_expr): Call it, and use the result,
6008 eliminating the direct usage of dump_file and alt_dump_file in
6009 favor of indirectly using them via emit_item.
6010 (make_item_for_dump_printf_va): New function, adapted from
6011 optinfo::add_printf_va.
6012 (make_item_for_dump_printf): New function.
6013 (dump_context::dump_printf_va): Call make_item_for_dump_printf_va,
6014 and use the result, eliminating the direct usage of dump_file and
6015 alt_dump_file in favor of indirectly using them via emit_item.
6016 (make_item_for_dump_dec): New function.
6017 (dump_context::dump_dec): Call it, and use the result,
6018 eliminating the direct usage of dump_file and alt_dump_file in
6019 favor of indirectly using them via emit_item.
6020 (make_item_for_dump_symtab_node): New function, adapted from
6021 optinfo::add_symtab_node.
6022 (dump_context::dump_symtab_node): Call it, and use the result,
6023 eliminating the direct usage of dump_file and alt_dump_file in
6024 favor of indirectly using them via emit_item.
6025 (dump_context::begin_scope): Reimplement, avoiding direct usage
6026 of dump_file and alt_dump_file in favor of indirectly using them
6028 (dump_context::emit_item): New member function.
6029 (temp_dump_context::temp_dump_context): Add param "test_pp_flags".
6030 Set up test pretty-printer on the underlying context. Call
6031 refresh_dumps_are_enabled.
6032 (temp_dump_context::~temp_dump_context): Call
6033 refresh_dumps_are_enabled.
6034 (temp_dump_context::get_dumped_text): New member function.
6035 (selftest::verify_dumped_text): New function.
6036 (ASSERT_DUMPED_TEXT_EQ): New macro.
6037 (selftest::test_capture_of_dump_calls): Run all tests twice, with
6038 and then without optinfo enabled. Add uses of
6039 ASSERT_DUMPED_TEXT_EQ to all tests. Add test of nested scopes.
6040 * dumpfile.h: Update comment for the dump_* API.
6041 * optinfo-emit-json.cc
6042 (selftest::test_building_json_from_dump_calls): Update for new
6043 param for temp_dump_context ctor.
6044 * optinfo.cc (optinfo_item::optinfo_item): Remove "owned" param
6045 and "m_owned" field.
6046 (optinfo_item::~optinfo_item): Likewise.
6047 (optinfo::add_item): New member function.
6048 (optinfo::emit): Update comment.
6049 (optinfo::add_string): Delete.
6050 (optinfo::add_printf): Delete.
6051 (optinfo::add_printf_va): Delete.
6052 (optinfo::add_gimple_stmt): Delete.
6053 (optinfo::add_gimple_expr): Delete.
6054 (optinfo::add_tree): Delete.
6055 (optinfo::add_symtab_node): Delete.
6056 (optinfo::add_dec): Delete.
6057 * optinfo.h (class dump_context): New forward decl.
6058 (optinfo::add_item): New decl.
6059 (optinfo::add_string): Delete.
6060 (optinfo::add_printf): Delete.
6061 (optinfo::add_printf_va): Delete.
6062 (optinfo::add_gimple_stmt): Delete.
6063 (optinfo::add_gimple_expr): Delete.
6064 (optinfo::add_tree): Delete.
6065 (optinfo::add_symtab_node): Delete.
6066 (optinfo::add_dec): Delete.
6067 (optinfo::add_poly_int): Delete.
6068 (optinfo_item::optinfo_item): Remove "owned" param.
6069 (class optinfo_item): Remove field "m_owned".
6071 2018-07-31 Jozef Lawrynowicz <jozef.l@mittosystems.com>
6074 * gcc/cfgexpand.c (set_parm_rtl): Use the alignment of Pmode when
6075 MAX_SUPPORTED_STACK_ALIGNMENT would otherwise be exceeded by the
6076 requested variable alignment.
6077 (expand_one_ssa_partition): Likewise.
6078 (expand_one_var): Likewise.
6080 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
6082 * config/pdp11/pdp11.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
6083 to speculation_safe_value_not_needed.
6085 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
6087 * targhooks.h (speculation_safe_value_not_needed): New prototype.
6088 * targhooks.c (speculation_safe_value_not_needed): New function.
6089 * target.def (have_speculation_safe_value): Update documentation.
6090 * doc/tm.texi: Regenerated.
6092 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
6094 * config/aarch64/iterators.md (ALLI_TI): New iterator.
6095 * config/aarch64/aarch64.md (despeculate_copy<ALLI_TI:mode>): New
6097 (despeculate_copy<ALLI:mode>_insn): New insn.
6098 (despeculate_copyti_insn): New insn.
6099 (despeculate_simple<ALLI:mode>): New insn
6100 (despeculate_simpleti): New insn.
6101 * config/aarch64/aarch64.c (aarch64_speculation_safe_value): New
6103 (TARGET_SPECULATION_SAFE_VALUE): Redefine to
6104 aarch64_speculation_safe_value.
6105 (aarch64_print_operand): Handle const0_rtx in modifier 'H'.
6107 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
6109 * config/aarch64/aarch64-speculation.cc: New file.
6110 * config/aarch64/aarch64-passes.def (pass_track_speculation): Add
6111 before pass_reorder_blocks.
6112 * config/aarch64/aarch64-protos.h (make_pass_track_speculation): Add
6114 * config/aarch64/aarch64.c (aarch64_conditional_register_usage): Fix
6115 X14 and X15 when tracking speculation.
6116 * config/aarch64/aarch64.md (register name constants): Add
6117 SPECULATION_TRACKER_REGNUM and SPECULATION_SCRATCH_REGNUM.
6118 (unspec): Add UNSPEC_SPECULATION_TRACKER.
6119 (speculation_barrier): New insn attribute.
6120 (cmp<mode>): Allow SP in comparisons.
6121 (speculation_tracker): New insn.
6122 (speculation_barrier): Add speculation_barrier attribute.
6123 * config/aarch64/t-aarch64: Add make rule for aarch64-speculation.o.
6124 * config.gcc (aarch64*-*-*): Add aarch64-speculation.o to extra_objs.
6125 * doc/invoke.texi (AArch64 Options): Document -mtrack-speculation.
6127 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
6129 * config/aarch64/aarch64.md (cb<optab><mode>1): Disable when
6130 aarch64_track_speculation is true.
6131 (tb<optab><mode>1): Likewise.
6132 * config/aarch64/aarch64.c (aarch64_split_compare_regs): Do not
6133 generate CB[N]Z when tracking speculation.
6134 (aarch64_split_compare_and_swap): Likewise.
6135 (aarch64_split_atomic_op): Likewise.
6137 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
6139 * config/aarch64/aarch64.opt (mtrack-speculation): New target option.
6141 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
6143 * config/aarch64.md (unspecv): Add UNSPECV_SPECULAION_BARRIER.
6144 (speculation_barrier): New insn.
6146 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
6148 * config/arm/unspecs.md (unspecv): Add VUNSPEC_SPECULATION_BARRIER.
6149 * config/arm/arm.md (speculation_barrier): New expand.
6150 (speculation_barrier_insn): New pattern.
6152 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
6154 * builtin-types.def (BT_FN_PTR_PTR_VAR): New function type.
6155 (BT_FN_I1_I1_VAR, BT_FN_I2_I2_VAR, BT_FN_I4_I4_VAR): Likewise.
6156 (BT_FN_I8_I8_VAR, BT_FN_I16_I16_VAR): Likewise.
6157 * builtin-attrs.def (ATTR_NOVOPS_NOTHROW_LEAF_LIST): New attribute
6159 * builtins.def (BUILT_IN_SPECULATION_SAFE_VALUE_N): New builtin.
6160 (BUILT_IN_SPECULATION_SAFE_VALUE_PTR): New internal builtin.
6161 (BUILT_IN_SPECULATION_SAFE_VALUE_1): Likewise.
6162 (BUILT_IN_SPECULATION_SAFE_VALUE_2): Likewise.
6163 (BUILT_IN_SPECULATION_SAFE_VALUE_4): Likewise.
6164 (BUILT_IN_SPECULATION_SAFE_VALUE_8): Likewise.
6165 (BUILT_IN_SPECULATION_SAFE_VALUE_16): Likewise.
6166 * builtins.c (expand_speculation_safe_value): New function.
6167 (expand_builtin): Call it.
6168 * doc/cpp.texi: Document predefine __HAVE_SPECULATION_SAFE_VALUE.
6169 * doc/extend.texi: Document __builtin_speculation_safe_value.
6170 * doc/md.texi: Document "speculation_barrier" pattern.
6171 * doc/tm.texi.in: Pull in TARGET_SPECULATION_SAFE_VALUE and
6172 TARGET_HAVE_SPECULATION_SAFE_VALUE.
6173 * doc/tm.texi: Regenerated.
6174 * target.def (have_speculation_safe_value, speculation_safe_value): New
6176 * targhooks.c (default_have_speculation_safe_value): New function.
6177 (default_speculation_safe_value): New function.
6178 * targhooks.h (default_have_speculation_safe_value): Add prototype.
6179 (default_speculation_safe_value): Add prototype.
6181 2018-07-31 David Malcolm <dmalcolm@redhat.com>
6183 * dump-context.h (dump_context::dump_loc): New decl.
6184 * dumpfile.c (dump_context::dump_loc): New member function.
6185 (dump_context::dump_gimple_stmt_loc): Reimplement using dump_loc
6186 and dump_gimple_stmt.
6187 (dump_context::dump_gimple_expr_loc): Likewise, using
6189 (dump_context::dump_generic_expr_loc): Likewise, using
6191 (dump_context::dump_printf_loc_va): Likewise, using
6193 (dump_context::begin_scope): Explicitly using the global function
6194 "dump_loc", rather than the member function.
6196 2018-07-31 Martin Sebor <msebor@redhat.com>
6198 PR tree-optimization/86741
6199 * tree-vrp.c (vrp_prop::check_mem_ref): Avoid incomplete types.
6201 2018-07-31 Andreas Krebbel <krebbel@linux.ibm.com>
6203 * config/s390/s390.c (s390_expand_setmem): Make the unrolling to
6204 depend on whether prefetch instructions will be emitted or not.
6205 Use TARGET_SETMEM_PFD for checking whether prefetch instructions
6206 will be emitted or not.
6207 * config/s390/s390.h (TARGET_SETMEM_PREFETCH_DISTANCE)
6208 (TARGET_SETMEM_PFD): New macros.
6210 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6212 * tree-vectorizer.h (stmt_vec_info): Turn back into a typedef.
6213 (NULL_STMT_VEC_INFO): Delete.
6214 (stmt_vec_info::operator*): Likewise.
6215 (stmt_vec_info::operator gimple *): Likewise.
6216 * tree-vect-loop.c (vectorizable_reduction): Use NULL instead
6217 of NULL_STMT_VEC_INFO.
6218 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
6219 (vect_reassociating_reduction_p): Likewise.
6220 * tree-vect-stmts.c (vect_build_gather_load_calls): Likewise.
6221 (vectorizable_store): Likewise.
6222 * tree-vectorizer.c (vec_info::set_vinfo_for_stmt): Likewise.
6223 (vec_info::free_stmt_vec_infos): Likewise.
6225 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6227 * tree-vectorizer.h (vect_stmt_in_region_p): Delete.
6228 * tree-vectorizer.c (vect_stmt_in_region_p): Likewise.
6230 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6232 * tree-vectorizer.h (vec_info::new_vinfo_for_stmt)
6233 (vec_info::set_vinfo_for_stmt, vec_info::free_stmt_vec_infos)
6234 (vec_info::free_stmt_vec_info): New private member functions.
6235 (set_stmt_vec_info_vec, free_stmt_vec_infos, vinfo_for_stmt)
6236 (set_vinfo_for_stmt, new_stmt_vec_info, free_stmt_vec_info): Delete.
6237 * tree-parloops.c (gather_scalar_reductions): Remove calls to
6238 set_stmt_vec_info_vec and free_stmt_vec_infos.
6239 * tree-vect-loop.c (_loop_vec_info): Remove call to
6240 set_stmt_vec_info_vec.
6241 * tree-vect-stmts.c (new_stmt_vec_info, set_stmt_vec_info_vec)
6242 (free_stmt_vec_infos, free_stmt_vec_info): Delete in favor of...
6243 * tree-vectorizer.c (vec_info::new_stmt_vec_info)
6244 (vec_info::set_vinfo_for_stmt, vec_info::free_stmt_vec_infos)
6245 (vec_info::free_stmt_vec_info): ...these new functions. Remove
6246 assignments in {vec_info::,}new_stmt_vec_info that are redundant
6247 with the clearing in the xcalloc.
6248 (stmt_vec_info_vec): Delete.
6249 (vec_info::vec_info): Don't call set_stmt_vec_info_vec.
6250 (vectorize_loops): Likewise.
6251 (vec_info::~vec_info): Remove argument from call to
6252 free_stmt_vec_infos.
6253 (vec_info::add_stmt): Remove vinfo argument from call to
6256 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6258 * tree-vectorizer.h (free_stmt_vec_info): Take a stmt_vec_info
6259 rather than a gimple stmt.
6260 * tree-vect-stmts.c (free_stmt_vec_info): Likewise. Don't free
6261 information for pattern statements when passed the original
6262 statement; instead wait to be passed the pattern statement itself.
6263 Don't call set_vinfo_for_stmt here.
6264 (free_stmt_vec_infos): Update call to free_stmt_vec_info.
6265 * tree-vect-loop.c (_loop_vec_info::~loop_vec_info): Don't free
6266 stmt_vec_infos here.
6267 * tree-vect-slp.c (_bb_vec_info::~bb_vec_info): Likewise.
6268 * tree-vectorizer.c (vec_info::remove_stmt): Nullify the statement's
6269 stmt_vec_infos entry.
6271 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6273 * tree-vectorizer.h (vec_info::replace_stmt): Declare.
6274 * tree-vectorizer.c (vec_info::replace_stmt): New function.
6275 * tree-vect-slp.c (vect_remove_slp_scalar_calls): Use it.
6276 * tree-vect-stmts.c (vectorizable_call): Likewise.
6277 (vectorizable_simd_clone_call): Likewise.
6279 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6281 * tree-vectorizer.h (vec_info::remove_stmt): Declare.
6282 * tree-vectorizer.c (vec_info::remove_stmt): New function.
6283 * tree-vect-loop-manip.c (vect_set_loop_condition): Use it.
6284 * tree-vect-loop.c (vect_transform_loop): Likewise.
6285 * tree-vect-slp.c (vect_schedule_slp): Likewise.
6286 * tree-vect-stmts.c (vect_remove_stores): Likewise.
6288 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6290 * tree-vectorizer.h (vec_info::lookup_dr): New member function.
6291 (vect_dr_stmt): Delete.
6292 * tree-vectorizer.c (vec_info::lookup_dr): New function.
6293 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Use it instead
6295 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
6296 (vect_analyze_data_ref_dependence, vect_record_base_alignments)
6297 (vect_verify_datarefs_alignment, vect_peeling_supportable)
6298 (vect_analyze_data_ref_accesses, vect_prune_runtime_alias_test_list)
6299 (vect_analyze_data_refs): Likewise.
6300 (vect_slp_analyze_data_ref_dependence): Likewise. Take a vec_info
6302 (vect_find_same_alignment_drs): Likewise.
6303 (vect_slp_analyze_node_dependences): Update calls accordingly.
6304 (vect_analyze_data_refs_alignment): Likewise. Use vec_info::lookup_dr
6305 instead of DR_VECT_AUX.
6306 (vect_get_peeling_costs_all_drs): Take a loop_vec_info instead
6307 of a vector data references. Use vec_info::lookup_dr instead of
6309 (vect_peeling_hash_get_lowest_cost): Update calls accordingly.
6310 (vect_enhance_data_refs_alignment): Likewise. Use vec_info::lookup_dr
6311 instead of DR_VECT_AUX.
6313 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6315 * tree-vectorizer.h (_loop_vec_info::unaligned_dr): Change to
6317 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Update
6319 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
6320 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
6321 (vect_gen_prolog_loop_niters): Likewise.
6323 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6325 * tree-vectorizer.h (set_dr_misalignment, dr_misalignment)
6326 (DR_TARGET_ALIGNMENT, aligned_access_p, known_alignment_for_access_p)
6327 (vect_known_alignment_in_bytes, vect_dr_behavior)
6328 (vect_get_scalar_dr_size): Take references as dr_vec_infos
6329 instead of data_references. Update calls to other routines for
6330 which the same change has been made.
6331 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Take
6332 dr_vec_infos instead of stmt_vec_infos.
6333 (vect_analyze_data_ref_dependence): Update call accordingly.
6334 (vect_slp_analyze_data_ref_dependence)
6335 (vect_record_base_alignments): Use DR_VECT_AUX.
6336 (vect_calculate_target_alignment, vect_compute_data_ref_alignment)
6337 (vect_update_misalignment_for_peel, verify_data_ref_alignment)
6338 (vector_alignment_reachable_p, vect_get_data_access_cost)
6339 (vect_peeling_supportable, vect_analyze_group_access_1)
6340 (vect_analyze_group_access, vect_analyze_data_ref_access)
6341 (vect_vfa_segment_size, vect_vfa_access_size, vect_vfa_align)
6342 (vect_compile_time_alias, vect_small_gap_p)
6343 (vectorizable_with_step_bound_p, vect_duplicate_ssa_name_ptr_info):
6344 (vect_supportable_dr_alignment): Take references as dr_vec_infos
6345 instead of data_references. Update calls to other routines for
6346 which the same change has been made.
6347 (vect_verify_datarefs_alignment, vect_get_peeling_costs_all_drs)
6348 (vect_find_same_alignment_drs, vect_analyze_data_refs_alignment)
6349 (vect_slp_analyze_and_verify_node_alignment)
6350 (vect_analyze_data_ref_accesses, vect_prune_runtime_alias_test_list)
6351 (vect_create_addr_base_for_vector_ref, vect_create_data_ref_ptr)
6352 (vect_setup_realignment): Use dr_vec_infos. Update calls after
6354 (_vect_peel_info::dr): Replace with...
6355 (_vect_peel_info::dr_info): ...this new field.
6356 (vect_peeling_hash_get_most_frequent)
6357 (vect_peeling_hash_choose_best_peeling): Update accordingly.
6358 (vect_peeling_hash_get_lowest_cost):
6359 (vect_enhance_data_refs_alignment): Likewise. Update calls to other
6360 routines for which the same change has been made.
6361 (vect_peeling_hash_insert): Likewise. Take a dr_vec_info instead of a
6363 * tree-vect-loop-manip.c (get_misalign_in_elems)
6364 (vect_gen_prolog_loop_niters): Use dr_vec_infos. Update calls after
6366 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
6367 * tree-vect-stmts.c (vect_get_store_cost, vect_get_load_cost)
6368 (vect_truncate_gather_scatter_offset, compare_step_with_zero)
6369 (get_group_load_store_type, get_negative_load_store_type)
6370 (vect_get_data_ptr_increment, vectorizable_store)
6371 (vectorizable_load): Likewise.
6372 (ensure_base_align): Take a dr_vec_info instead of a data_reference.
6373 Update calls to other routines for which the same change has been made.
6375 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6377 * tree-vectorizer.h (vec_info::move_dr): New member function.
6378 (dataref_aux): Rename to...
6379 (dr_vec_info): ...this and add "dr" and "stmt" fields.
6380 (_stmt_vec_info::dr_aux): Update accordingly.
6381 (_stmt_vec_info::data_ref_info): Delete.
6382 (STMT_VINFO_GROUPED_ACCESS, DR_GROUP_FIRST_ELEMENT)
6383 (DR_GROUP_NEXT_ELEMENT, DR_GROUP_SIZE, DR_GROUP_STORE_COUNT)
6384 (DR_GROUP_GAP, DR_GROUP_SAME_DR_STMT, REDUC_GROUP_FIRST_ELEMENT):
6385 (REDUC_GROUP_NEXT_ELEMENT, REDUC_GROUP_SIZE): Use dr_aux.dr instead
6387 (STMT_VINFO_DATA_REF): Likewise. Turn into an lvalue.
6388 (STMT_VINFO_DR_INFO): New macro.
6389 (DR_VECT_AUX): Use STMT_VINFO_DR_INKFO and vect_dr_stmt.
6390 (set_dr_misalignment): Update after rename of dataref_aux.
6391 (vect_dr_stmt): Move earlier in file. Return dr_aux.stmt.
6392 * tree-vect-stmts.c (new_stmt_vec_info): Remove redundant
6393 initialization of STMT_VINFO_DATA_REF.
6394 * tree-vectorizer.c (vec_info::move_dr): New function.
6395 * tree-vect-patterns.c (vect_recog_bool_pattern)
6396 (vect_recog_mask_conversion_pattern)
6397 (vect_recog_gather_scatter_pattern): Use it.
6398 * tree-vect-data-refs.c (vect_analyze_data_refs): Initialize
6399 the "dr" and "stmt" fields of dr_vec_info instead of
6400 STMT_VINFO_DATA_REF.
6402 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6404 * tree-vectorizer.h (_stmt_vec_info::pattern_stmt_p): New field.
6405 (is_pattern_stmt_p): Use it.
6406 * tree-vect-patterns.c (vect_init_pattern_stmt): Set pattern_stmt_p
6407 on pattern statements.
6409 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6411 * tree-vect-patterns.c (vect_mark_pattern_stmts): Take the
6412 original stmt as a stmt_vec_info rather than a gimple stmt.
6413 (vect_pattern_recog_1): Take the statement directly as a
6414 stmt_vec_info, rather than via a gimple_stmt_iterator.
6415 Update call to vect_mark_pattern_stmts.
6416 (vect_pattern_recog): Update calls accordingly.
6418 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6420 * tree-vectorizer.h (vect_get_vec_defs_for_stmt_copy)
6421 (vect_get_vec_def_for_stmt_copy): Take a vec_info rather than
6422 a vect_def_type for the first argument.
6423 * tree-vect-stmts.c (vect_get_vec_defs_for_stmt_copy): Likewise.
6424 (vect_get_vec_def_for_stmt_copy): Likewise. Return the original
6425 operand if it isn't defined by a vectorized statement.
6426 (vect_build_gather_load_calls): Remove the mask_dt argument and
6427 update calls to vect_get_vec_def_for_stmt_copy.
6428 (vectorizable_bswap): Likewise the dt argument.
6429 (vectorizable_call): Update calls to vectorizable_bswap and
6430 vect_get_vec_def_for_stmt_copy.
6431 (vectorizable_simd_clone_call, vectorizable_assignment)
6432 (vectorizable_shift, vectorizable_operation, vectorizable_condition)
6433 (vectorizable_comparison): Update calls to
6434 vect_get_vec_def_for_stmt_copy.
6435 (vectorizable_store): Likewise. Remove now-unnecessary calls to
6437 (vect_get_loop_based_defs): Remove dt argument and update call
6438 to vect_get_vec_def_for_stmt_copy.
6439 (vectorizable_conversion): Update calls to vect_get_loop_based_defs
6440 and vect_get_vec_def_for_stmt_copy.
6441 (vectorizable_load): Update calls to vect_build_gather_load_calls
6442 and vect_get_vec_def_for_stmt_copy.
6443 * tree-vect-loop.c (vect_create_epilog_for_reduction)
6444 (vectorizable_reduction, vectorizable_live_operation): Update calls
6445 to vect_get_vec_def_for_stmt_copy.
6447 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6449 * tree-vect-data-refs.c (vect_record_base_alignment): Replace vec_info
6450 and gimple stmt arguments with a stmt_vec_info.
6451 (vect_record_base_alignments): Update calls accordingly.
6452 * tree-vect-slp.c (vect_record_max_nunits): Replace vec_info
6453 and gimple stmt arguments with a stmt_vec_info.
6454 (vect_build_slp_tree_1): Remove vinfo argument and update call
6455 to vect_record_max_nunits.
6456 (vect_build_slp_tree_2): Update calls to vect_build_slp_tree_1
6457 and vect_record_max_nunits.
6459 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6461 * tree-vectorizer.h (nested_in_vect_loop_p): Move further down
6462 file and take a stmt_vec_info instead of a gimple stmt.
6463 (supportable_widening_operation, vect_finish_replace_stmt)
6464 (vect_finish_stmt_generation, vect_get_store_rhs)
6465 (vect_get_vec_def_for_operand_1, vect_get_vec_def_for_operand)
6466 (vect_get_vec_defs, vect_init_vector, vect_transform_stmt)
6467 (vect_remove_stores, vect_analyze_stmt, vectorizable_condition)
6468 (vect_get_smallest_scalar_type, vect_check_gather_scatter)
6469 (vect_create_data_ref_ptr, bump_vector_ptr)
6470 (vect_permute_store_chain, vect_setup_realignment)
6471 (vect_transform_grouped_load, vect_record_grouped_load_vectors)
6472 (vect_create_addr_base_for_vector_ref, vectorizable_live_operation)
6473 (vectorizable_reduction, vectorizable_induction)
6474 (get_initial_def_for_reduction, is_simple_and_all_uses_invariant)
6475 (vect_get_place_in_interleaving_chain): Take stmt_vec_infos rather
6476 than gimple stmts as arguments.
6477 * tree-vect-data-refs.c (vect_get_smallest_scalar_type)
6478 (vect_preserves_scalar_order_p, vect_slp_analyze_node_dependences)
6479 (can_group_stmts_p, vect_check_gather_scatter)
6480 (vect_create_addr_base_for_vector_ref, vect_create_data_ref_ptr)
6481 (bump_vector_ptr, vect_permute_store_chain, vect_setup_realignment)
6482 (vect_permute_load_chain, vect_shift_permute_load_chain)
6483 (vect_transform_grouped_load)
6484 (vect_record_grouped_load_vectors): Likewise.
6485 * tree-vect-loop.c (vect_fixup_reduc_chain)
6486 (get_initial_def_for_reduction, vect_create_epilog_for_reduction)
6487 (vectorize_fold_left_reduction, is_nonwrapping_integer_induction)
6488 (vectorizable_reduction, vectorizable_induction)
6489 (vectorizable_live_operation, vect_loop_kill_debug_uses): Likewise.
6490 * tree-vect-patterns.c (type_conversion_p, adjust_bool_stmts)
6491 (vect_get_load_store_mask): Likewise.
6492 * tree-vect-slp.c (vect_get_place_in_interleaving_chain)
6493 (vect_analyze_slp_instance, vect_mask_constant_operand_p): Likewise.
6494 * tree-vect-stmts.c (vect_mark_relevant)
6495 (is_simple_and_all_uses_invariant)
6496 (exist_non_indexing_operands_for_use_p, process_use)
6497 (vect_init_vector_1, vect_init_vector, vect_get_vec_def_for_operand_1)
6498 (vect_get_vec_def_for_operand, vect_get_vec_defs)
6499 (vect_finish_stmt_generation_1, vect_finish_replace_stmt)
6500 (vect_finish_stmt_generation, vect_truncate_gather_scatter_offset)
6501 (compare_step_with_zero, vect_get_store_rhs, get_group_load_store_type)
6502 (get_negative_load_store_type, get_load_store_type)
6503 (vect_check_load_store_mask, vect_check_store_rhs)
6504 (vect_build_gather_load_calls, vect_get_strided_load_store_ops)
6505 (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
6506 (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
6507 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
6508 (get_group_alias_ptr_type, vectorizable_store, hoist_defs_of_uses)
6509 (vectorizable_load, vectorizable_condition, vectorizable_comparison)
6510 (vect_analyze_stmt, vect_transform_stmt, vect_remove_stores)
6511 (supportable_widening_operation): Likewise.
6513 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6515 * tree-vect-data-refs.c (vect_describe_gather_scatter_call): Take
6516 a stmt_vec_info instead of a gcall.
6517 (vect_check_gather_scatter): Update call accordingly.
6518 * tree-vect-loop-manip.c (iv_phi_p): Take a stmt_vec_info instead
6520 (vect_can_advance_ivs_p, vect_update_ivs_after_vectorizer)
6521 (slpeel_update_phi_nodes_for_loops):): Update calls accordingly.
6522 * tree-vect-loop.c (vect_transform_loop_stmt): Take a stmt_vec_info
6523 instead of a gimple stmt.
6524 (vect_transform_loop): Update calls accordingly.
6525 * tree-vect-slp.c (vect_split_slp_store_group): Take and return
6526 stmt_vec_infos instead of gimple stmts.
6527 (vect_analyze_slp_instance): Update use accordingly.
6528 * tree-vect-stmts.c (read_vector_array, write_vector_array)
6529 (vect_clobber_variable, vect_stmt_relevant_p, permute_vec_elements)
6530 (vect_use_strided_gather_scatters_p, vect_build_all_ones_mask)
6531 (vect_build_zero_merge_argument, vect_get_gather_scatter_ops)
6532 (vect_gen_widened_results_half, vect_get_loop_based_defs)
6533 (vect_create_vectorized_promotion_stmts, can_vectorize_live_stmts):
6534 Take a stmt_vec_info instead of a gimple stmt and pass stmt_vec_infos
6535 down to subroutines.
6537 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6539 * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Change the type
6540 of the worklist from a vector of gimple stmts to a vector of
6542 * tree-vect-stmts.c (vect_mark_relevant, process_use)
6543 (vect_mark_stmts_to_be_vectorized): Likewise
6545 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6547 * tree-vect-loop.c (vect_analyze_loop_operations): Look up the
6548 statement before passing it to vect_analyze_stmt.
6549 (vect_create_epilog_for_reduction): Use a stmt_vec_info to walk
6550 the chain of phi vector definitions. Track the exit phi via its
6552 (vectorizable_reduction): Set cond_stmt_vinfo directly from the
6553 STMT_VINFO_REDUC_DEF.
6554 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Use
6555 stmt_vec_infos to handle the statement chains.
6556 (vect_get_slp_defs): Record the first statement in the node
6557 using a stmt_vec_info.
6558 * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Look up
6559 statements here and pass their stmt_vec_info down to subroutines.
6560 (vect_init_vector_1): Hoist call to vinfo_for_stmt and pass it
6561 down to vect_finish_stmt_generation.
6562 (vect_init_vector, vect_get_vec_defs, vect_finish_replace_stmt)
6563 (vect_finish_stmt_generation): Call vinfo_for_stmt and pass
6564 stmt_vec_infos to subroutines.
6565 (vect_remove_stores): Use stmt_vec_infos to handle the statement
6568 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6570 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences):
6571 (vect_check_gather_scatter, vect_create_data_ref_ptr, bump_vector_ptr)
6572 (vect_permute_store_chain, vect_setup_realignment)
6573 (vect_permute_load_chain, vect_shift_permute_load_chain)
6574 (vect_transform_grouped_load): Use stmt_vec_info rather than gimple
6575 stmts internally, and when passing values to other vectorizer routines.
6576 * tree-vect-loop-manip.c (vect_can_advance_ivs_p): Likewise.
6577 * tree-vect-loop.c (vect_analyze_scalar_cycles_1)
6578 (vect_analyze_loop_operations, get_initial_def_for_reduction)
6579 (vect_create_epilog_for_reduction, vectorize_fold_left_reduction)
6580 (vectorizable_reduction, vectorizable_induction)
6581 (vectorizable_live_operation, vect_transform_loop_stmt)
6582 (vect_transform_loop): Likewise.
6583 * tree-vect-patterns.c (vect_reassociating_reduction_p)
6584 (vect_recog_widen_op_pattern, vect_recog_mixed_size_cond_pattern)
6585 (vect_recog_bool_pattern, vect_recog_gather_scatter_pattern): Likewise.
6586 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
6587 (vect_slp_analyze_node_operations_1): Likewise.
6588 * tree-vect-stmts.c (vect_mark_relevant, process_use)
6589 (exist_non_indexing_operands_for_use_p, vect_init_vector_1)
6590 (vect_mark_stmts_to_be_vectorized, vect_get_vec_def_for_operand)
6591 (vect_finish_stmt_generation_1, get_group_load_store_type)
6592 (get_load_store_type, vect_build_gather_load_calls)
6593 (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
6594 (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
6595 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
6596 (vectorizable_store, vectorizable_load, vectorizable_condition)
6597 (vectorizable_comparison, vect_analyze_stmt, vect_transform_stmt)
6598 (supportable_widening_operation): Likewise.
6599 (vect_get_vector_types_for_stmt): Likewise.
6600 * tree-vectorizer.h (vect_dr_behavior): Likewise.
6602 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6604 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
6605 (vect_slp_analyze_node_dependences, vect_analyze_data_ref_accesses)
6606 (vect_permute_store_chain, vect_permute_load_chain)
6607 (vect_shift_permute_load_chain, vect_transform_grouped_load): Avoid
6608 repeated stmt_vec_info lookups.
6609 * tree-vect-loop-manip.c (vect_can_advance_ivs_p): Likewise.
6610 (vect_update_ivs_after_vectorizer): Likewise.
6611 * tree-vect-loop.c (vect_is_simple_reduction): Likewise.
6612 (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
6613 * tree-vect-patterns.c (adjust_bool_stmts): Likewise.
6614 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
6615 (vect_bb_slp_scalar_cost): Likewise.
6616 * tree-vect-stmts.c (get_group_alias_ptr_type): Likewise.
6618 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6620 * tree-vect-data-refs.c (vect_check_gather_scatter): Pass the
6621 gcall rather than the generic gimple stmt to gimple_call_internal_fn.
6622 (vect_get_smallest_scalar_type, can_group_stmts_p): Use dyn_cast
6623 to get gassigns and gcalls, rather than operating on generc gimple
6625 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p)
6626 (vect_mark_stmts_to_be_vectorized, vectorizable_store)
6627 (vectorizable_load, vect_analyze_stmt): Likewise.
6628 * tree-vect-loop.c (vectorizable_reduction): Likewise gphi.
6630 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6632 * tree-vectorizer.h (get_earlier_stmt, get_later_stmt): Take and
6633 return stmt_vec_infos rather than gimple stmts. Do not accept
6635 (vect_find_last_scalar_stmt_in_slp): Return a stmt_vec_info instead
6637 * tree-vect-slp.c (vect_find_last_scalar_stmt_in_slp): Likewise.
6638 Update use of get_later_stmt.
6639 (vect_get_constant_vectors): Update call accordingly.
6640 (vect_schedule_slp_instance): Likewise
6641 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Likewise.
6642 (vect_slp_analyze_instance_dependence): Likewise.
6643 (vect_preserves_scalar_order_p): Update use of get_earlier_stmt.
6645 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6647 * tree-vectorizer.h (stmt_info_for_cost::stmt): Replace with...
6648 (stmt_info_for_cost::stmt_info): ...this new field.
6649 (add_stmt_costs): Update accordingly.
6650 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost)
6651 (vect_get_known_peeling_cost): Likewise.
6652 (vect_estimate_min_profitable_iters): Likewise.
6653 * tree-vect-stmts.c (record_stmt_cost): Likewise.
6655 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6657 * tree-vectorizer.h (_loop_vec_info::may_misalign_stmts): Change
6658 from an auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
6659 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Update
6661 * tree-vect-loop-manip.c (vect_create_cond_for_align_checks): Likewise.
6663 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6665 * tree-vectorizer.h (_stmt_vec_info::same_dr_stmt): Change from
6666 a gimple stmt to a stmt_vec_info.
6667 * tree-vect-stmts.c (vectorizable_load): Update accordingly.
6669 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6671 * tree-vectorizer.h (vec_info::grouped_stores): Change from
6672 an auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
6673 (_loop_vec_info::reduction_chains): Likewise.
6674 * tree-vect-loop.c (vect_fixup_scalar_cycles_with_patterns): Update
6676 * tree-vect-slp.c (vect_analyze_slp): Likewise.
6678 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6680 * tree-vectorizer.h (_stmt_vec_info::first_element): Change from
6681 a gimple stmt to a stmt_vec_info.
6682 (_stmt_vec_info::next_element): Likewise.
6683 * tree-vect-data-refs.c (vect_update_misalignment_for_peel)
6684 (vect_slp_analyze_and_verify_node_alignment)
6685 (vect_analyze_group_access_1, vect_analyze_group_access)
6686 (vect_small_gap_p, vect_prune_runtime_alias_test_list)
6687 (vect_create_data_ref_ptr, vect_record_grouped_load_vectors)
6688 (vect_supportable_dr_alignment): Update accordingly.
6689 * tree-vect-loop.c (vect_fixup_reduc_chain): Likewise.
6690 (vect_fixup_scalar_cycles_with_patterns, vect_is_slp_reduction)
6691 (vect_is_simple_reduction, vectorizable_reduction): Likewise.
6692 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
6693 * tree-vect-slp.c (vect_build_slp_tree_1)
6694 (vect_attempt_slp_rearrange_stmts, vect_supported_load_permutation_p)
6695 (vect_split_slp_store_group, vect_analyze_slp_instance)
6696 (vect_analyze_slp, vect_transform_slp_perm_load): Likewise.
6697 * tree-vect-stmts.c (vect_model_store_cost, vect_model_load_cost)
6698 (get_group_load_store_type, get_load_store_type)
6699 (get_group_alias_ptr_type, vectorizable_store, vectorizable_load)
6700 (vect_transform_stmt, vect_remove_stores): Likewise.
6702 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6704 * tree-vectorizer.h (vect_dr_stmt): Return a stmt_vec_info rather
6706 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
6707 (vect_slp_analyze_data_ref_dependence, vect_record_base_alignments)
6708 (vect_calculate_target_alignmentm, vect_compute_data_ref_alignment)
6709 (vect_update_misalignment_for_peel, vect_verify_datarefs_alignment)
6710 (vector_alignment_reachable_p, vect_get_data_access_cost)
6711 (vect_get_peeling_costs_all_drs, vect_peeling_hash_get_lowest_cost)
6712 (vect_peeling_supportable, vect_enhance_data_refs_alignment)
6713 (vect_find_same_alignment_drs, vect_analyze_data_refs_alignment)
6714 (vect_analyze_group_access_1, vect_analyze_group_access)
6715 (vect_analyze_data_ref_access, vect_analyze_data_ref_accesses)
6716 (vect_vfa_access_size, vect_small_gap_p, vect_analyze_data_refs)
6717 (vect_supportable_dr_alignment): Remove vinfo_for_stmt from the
6718 result of vect_dr_stmt and use the stmt_vec_info instead of
6719 the associated gimple stmt.
6720 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
6721 (vect_gen_prolog_loop_niters): Likewise.
6722 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
6724 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6726 * tree-vectorizer.h (_slp_tree::stmts): Change from a vec<gimple *>
6727 to a vec<stmt_vec_info>.
6728 * tree-vect-slp.c (vect_free_slp_tree): Update accordingly.
6729 (vect_create_new_slp_node): Take a vec<gimple *> instead of a
6731 (_slp_oprnd_info::def_stmts): Change from a vec<gimple *>
6732 to a vec<stmt_vec_info>.
6733 (bst_traits::value_type, bst_traits::value_type): Likewise.
6734 (bst_traits::hash): Update accordingly.
6735 (vect_get_and_check_slp_defs): Change the stmts parameter from
6736 a vec<gimple *> to a vec<stmt_vec_info>.
6737 (vect_two_operations_perm_ok_p, vect_build_slp_tree_1): Likewise.
6738 (vect_build_slp_tree): Likewise.
6739 (vect_build_slp_tree_2): Likewise. Update uses of
6740 SLP_TREE_SCALAR_STMTS.
6741 (vect_print_slp_tree): Update uses of SLP_TREE_SCALAR_STMTS.
6742 (vect_mark_slp_stmts, vect_mark_slp_stmts_relevant)
6743 (vect_slp_rearrange_stmts, vect_attempt_slp_rearrange_stmts)
6744 (vect_supported_load_permutation_p, vect_find_last_scalar_stmt_in_slp)
6745 (vect_detect_hybrid_slp_stmts, vect_slp_analyze_node_operations_1)
6746 (vect_slp_analyze_node_operations, vect_slp_analyze_operations)
6747 (vect_bb_slp_scalar_cost, vect_slp_analyze_bb_1)
6748 (vect_get_constant_vectors, vect_get_slp_defs)
6749 (vect_transform_slp_perm_load, vect_schedule_slp_instance)
6750 (vect_remove_slp_scalar_calls, vect_schedule_slp): Likewise.
6751 (vect_analyze_slp_instance): Build up a vec of stmt_vec_infos
6752 instead of gimple stmts.
6753 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Change
6754 the stores parameter for a vec<gimple *> to a vec<stmt_vec_info>.
6755 (vect_slp_analyze_instance_dependence): Update uses of
6756 SLP_TREE_SCALAR_STMTS.
6757 (vect_slp_analyze_and_verify_node_alignment): Likewise.
6758 (vect_slp_analyze_and_verify_instance_alignment): Likewise.
6759 * tree-vect-loop.c (neutral_op_for_slp_reduction): Likewise.
6760 (get_initial_defs_for_reduction): Likewise.
6761 (vect_create_epilog_for_reduction): Likewise.
6762 (vectorize_fold_left_reduction): Likewise.
6763 * tree-vect-stmts.c (vect_prologue_cost_for_slp_op): Likewise.
6764 (vect_model_simple_cost, vectorizable_shift, vectorizable_load)
6765 (can_vectorize_live_stmts): Likewise.
6767 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6769 * tree-vectorizer.h (_loop_vec_info::reductions): Change from an
6770 auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
6771 (vect_force_simple_reduction): Take and return stmt_vec_infos rather
6773 * tree-parloops.c (valid_reduction_p): Take a stmt_vec_info instead
6775 (gather_scalar_reductions): Update after above interface changes.
6776 * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Likewise.
6777 (vect_is_simple_reduction): Take and return stmt_vec_infos rather
6779 (vect_force_simple_reduction): Likewise.
6780 * tree-vect-patterns.c (vect_pattern_recog_1): Update use of
6781 LOOP_VINFO_REDUCTIONS.
6782 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
6784 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6786 * tree-vectorizer.h (_stmt_vec_info::reduc_def): Change from
6787 a gimple stmt to a stmt_vec_info.
6788 * tree-vect-loop.c (vect_active_double_reduction_p)
6789 (vect_force_simple_reduction, vectorizable_reduction): Update
6792 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6794 * tree-vectorizer.h (_slp_tree::vec_stmts): Change from a
6795 vec<gimple *> to a vec<stmt_vec_info>.
6796 * tree-vect-loop.c (vect_create_epilog_for_reduction): Change
6797 the reduction_phis argument from a vec<gimple *> to a
6799 (vectorizable_reduction): Likewise the phis local variable that
6800 is passed to vect_create_epilog_for_reduction. Update for new type
6801 of SLP_TREE_VEC_STMTS.
6802 (vectorizable_induction): Update for new type of SLP_TREE_VEC_STMTS.
6803 (vectorizable_live_operation): Likewise.
6804 * tree-vect-slp.c (vect_get_slp_vect_defs): Likewise.
6805 (vect_transform_slp_perm_load, vect_schedule_slp_instance): Likewise.
6807 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6809 * tree-vectorizer.h (_stmt_vec_info::vectorized_stmt): Change from
6810 a gimple stmt to a stmt_vec_info.
6811 (vectorizable_condition, vectorizable_live_operation)
6812 (vectorizable_reduction, vectorizable_induction): Pass back the
6813 vectorized statement as a stmt_vec_info.
6814 * tree-vect-data-refs.c (vect_record_grouped_load_vectors): Update
6815 use of STMT_VINFO_VEC_STMT.
6816 * tree-vect-loop.c (vect_create_epilog_for_reduction): Likewise,
6817 accumulating the inner phis that feed the STMT_VINFO_VEC_STMT
6818 as stmt_vec_infos rather than gimple stmts.
6819 (vectorize_fold_left_reduction): Change vec_stmt from a gimple stmt
6821 (vectorizable_live_operation): Likewise.
6822 (vectorizable_reduction, vectorizable_induction): Likewise,
6823 updating use of STMT_VINFO_VEC_STMT.
6824 * tree-vect-stmts.c (vect_get_vec_def_for_operand_1): Update use
6825 of STMT_VINFO_VEC_STMT.
6826 (vect_build_gather_load_calls, vectorizable_bswap, vectorizable_call)
6827 (vectorizable_simd_clone_call, vectorizable_conversion)
6828 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
6829 (vectorizable_store, vectorizable_load, vectorizable_condition)
6830 (vectorizable_comparison, can_vectorize_live_stmts): Change vec_stmt
6831 from a gimple stmt to a stmt_vec_info.
6832 (vect_transform_stmt): Update use of STMT_VINFO_VEC_STMT. Pass a
6833 pointer to a stmt_vec_info to the vectorizable_* routines.
6835 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6837 * tree-vectorizer.h (_stmt_vec_info::related_stmt): Change from
6838 a gimple stmt to a stmt_vec_info.
6839 (is_pattern_stmt_p): Update accordingly.
6840 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Likewise.
6841 (vect_record_grouped_load_vectors): Likewise.
6842 * tree-vect-loop.c (vect_determine_vf_for_stmt): Likewise.
6843 (vect_fixup_reduc_chain, vect_update_vf_for_slp): Likewise.
6844 (vect_model_reduction_cost): Likewise.
6845 (vect_create_epilog_for_reduction): Likewise.
6846 (vectorizable_reduction, vectorizable_induction): Likewise.
6847 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
6848 Return the stmt_vec_info for the pattern statement.
6849 (vect_set_pattern_stmt): Update use of STMT_VINFO_RELATED_STMT.
6850 (vect_split_statement, vect_mark_pattern_stmts): Likewise.
6851 * tree-vect-slp.c (vect_detect_hybrid_slp_stmts): Likewise.
6852 (vect_detect_hybrid_slp, vect_get_slp_defs): Likewise.
6853 * tree-vect-stmts.c (vect_mark_relevant): Likewise.
6854 (vect_get_vec_def_for_operand_1, vectorizable_call): Likewise.
6855 (vectorizable_simd_clone_call, vect_analyze_stmt, new_stmt_vec_info)
6856 (free_stmt_vec_info, vect_is_simple_use): Likewise.
6858 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6860 * tree-vectorizer.h (vect_finish_replace_stmt): Return a stmt_vec_info
6861 (vect_finish_stmt_generation): Likewise.
6862 * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
6863 (vect_finish_replace_stmt, vect_finish_stmt_generation): Likewise.
6864 (vect_build_gather_load_calls): Use the return value of the above
6865 functions instead of a separate call to vinfo_for_stmt. Use narrow
6866 scopes for the input gimple stmt and wider scopes for the associated
6867 stmt_vec_info. Use vec_info::lookup_def when setting these
6868 stmt_vec_infos from an SSA_NAME definition.
6869 (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
6870 (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
6871 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
6872 (vectorizable_store, vectorizable_load, vectorizable_condition)
6873 (vectorizable_comparison): Likewise.
6874 * tree-vect-loop.c (vectorize_fold_left_reduction): Likewise.
6875 (vectorizable_reduction): Likewise.
6877 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6879 * tree-vectorizer.h (vect_is_simple_use): Add an optional
6880 stmt_vec_info * parameter before the optional gimple **.
6881 * tree-vect-stmts.c (vect_is_simple_use): Likewise.
6882 (process_use, vect_get_vec_def_for_operand_1): Update callers.
6883 (vect_get_vec_def_for_operand, vectorizable_shift): Likewise.
6884 * tree-vect-loop.c (vectorizable_reduction): Likewise.
6885 (vectorizable_live_operation): Likewise.
6886 * tree-vect-patterns.c (type_conversion_p): Likewise.
6887 (vect_look_through_possible_promotion): Likewise.
6888 (vect_recog_rotate_pattern): Likewise.
6889 * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
6891 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6893 * tree-vectorizer.h (stmt_vec_info): Temporarily change from
6894 a typedef to a wrapper class.
6895 (NULL_STMT_VEC_INFO): New macro.
6896 (vec_info::stmt_infos): Change to vec<stmt_vec_info>.
6897 (stmt_vec_info::operator*): New function.
6898 (stmt_vec_info::operator gimple *): Likewise.
6899 (set_vinfo_for_stmt): Use NULL_STMT_VEC_INFO.
6900 (add_stmt_costs): Likewise.
6901 * tree-vect-loop-manip.c (iv_phi_p): Likewise.
6902 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost)
6903 (vect_get_known_peeling_cost): Likewise.
6904 (vect_estimate_min_profitable_iters): Likewise.
6905 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
6906 * tree-vect-slp.c (vect_remove_slp_scalar_calls): Likewise.
6907 * tree-vect-stmts.c (vect_build_gather_load_calls): Likewise.
6908 (vectorizable_store, free_stmt_vec_infos): Likewise.
6909 (new_stmt_vec_info): Change return type of xcalloc to
6912 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6914 * tree-vectorizer.h (vec_info::lookup_single_use): Declare.
6915 * tree-vectorizer.c (vec_info::lookup_single_use): New function.
6916 * tree-vect-loop.c (vectorizable_reduction): Use it instead of
6917 a single_imm_use-based sequence.
6918 * tree-vect-stmts.c (supportable_widening_operation): Likewise.
6920 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6922 * tree-vectorizer.h (vec_info::lookup_def): Declare.
6923 * tree-vectorizer.c (vec_info::lookup_def): New function.
6924 * tree-vect-patterns.c (vect_get_internal_def): Use it.
6925 (vect_widened_op_tree): Likewise.
6926 * tree-vect-stmts.c (vect_is_simple_use): Likewise.
6927 * tree-vect-loop.c (vect_analyze_loop_operations): Likewise.
6928 (vectorizable_reduction): Likewise.
6929 (vect_valid_reduction_input_p): Take a stmt_vec_info instead
6931 (vect_is_slp_reduction): Update calls accordingly. Use
6932 vec_info::lookup_def.
6933 (vect_is_simple_reduction): Likewise
6934 * tree-vect-slp.c (vect_detect_hybrid_slp_1): Use vec_info::lookup_def.
6936 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6938 * tree-vectorizer.h (vec_info::lookup_stmt): Declare.
6939 * tree-vectorizer.c (vec_info::lookup_stmt): New function.
6940 * tree-vect-loop.c (vect_determine_vf_for_stmt): Use it instead
6942 (vect_determine_vectorization_factor, vect_analyze_scalar_cycles_1)
6943 (vect_compute_single_scalar_iteration_cost, vect_analyze_loop_form)
6944 (vect_update_vf_for_slp, vect_analyze_loop_operations)
6945 (vect_is_slp_reduction, vectorizable_induction)
6946 (vect_transform_loop_stmt, vect_transform_loop): Likewise.
6947 * tree-vect-patterns.c (vect_init_pattern_stmt):
6948 (vect_determine_min_output_precision_1, vect_determine_precisions)
6949 (vect_pattern_recog): Likewise.
6950 * tree-vect-stmts.c (vect_analyze_stmt, vect_transform_stmt): Likewise.
6951 * config/powerpcspe/powerpcspe.c (rs6000_density_test): Likewise.
6952 * config/rs6000/rs6000.c (rs6000_density_test): Likewise.
6953 * tree-vect-slp.c (vect_detect_hybrid_slp_stmts): Likewise.
6954 (vect_detect_hybrid_slp_1, vect_detect_hybrid_slp_2)
6955 (vect_detect_hybrid_slp): Likewise. Change the walk_stmt_info
6956 info field from a loop to a loop_vec_info.
6958 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6960 * tree-vectorizer.h (stmt_vec_info): Move typedef earlier in file.
6961 (vec_info::add_stmt): Declare.
6962 * tree-vectorizer.c (vec_info::add_stmt): New function.
6963 * tree-vect-data-refs.c (vect_create_data_ref_ptr): Use it.
6964 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Likewise.
6965 (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
6966 (vectorizable_induction): Likewise.
6967 * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Likewise.
6968 * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
6969 (vectorizable_simd_clone_call, vectorizable_store): Likewise.
6970 (vectorizable_load): Likewise.
6971 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
6972 (vect_recog_bool_pattern, vect_recog_mask_conversion_pattern)
6973 (vect_recog_gather_scatter_pattern): Likewise.
6974 (append_pattern_def_seq): Likewise. Remove a check that is
6975 performed by add_stmt itself.
6977 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6979 * tree-vect-loop.c (vectorizable_reduction): Fix an instance in
6980 which make_ssa_name was called with new_stmt before new_stmt
6983 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6985 * tree-vect-loop.c (vect_valid_reduction_input_p): New function,
6987 (vect_is_slp_reduction): ...here...
6988 (vect_is_simple_reduction): ...and here. Remove repetition of tests
6989 that are already known to be false.
6991 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6993 * tree-vectorizer.h (vect_free_slp_instance): Add a final_p parameter.
6994 * tree-vect-slp.c (vect_free_slp_tree): Likewise. Don't update
6995 STMT_VINFO_NUM_SLP_USES when it's true.
6996 (vect_free_slp_instance): Add a final_p parameter and pass it to
6998 (vect_build_slp_tree_2): Update call to vect_free_slp_instance.
6999 (vect_analyze_slp_instance): Likewise.
7000 (vect_slp_analyze_operations): Likewise.
7001 (vect_slp_analyze_bb_1): Likewise.
7002 * tree-vectorizer.c (vec_info): Likewise.
7003 * tree-vect-loop.c (vect_transform_loop): Likewise.
7005 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
7007 * tree-vect-loop.c (vectorizable_reduction): Assert that the
7008 function is not called for second and subsequent members of
7011 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
7013 * tree-vect-loop.c (get_initial_def_for_reduction): Move special
7014 cases for nested loops from here to ...
7015 (vect_create_epilog_for_reduction): ...here. Only call
7016 vect_is_simple_use for inner-loop reductions.
7018 2018-07-31 Martin Liska <mliska@suse.cz>
7020 PR gcov-profile/85338
7021 PR gcov-profile/85350
7022 PR gcov-profile/85372
7023 * profile.c (struct location_triplet): New.
7024 (struct location_triplet_hash): Likewise.
7025 (output_location): Do not output a BB that
7026 is already recorded for a line.
7027 (branch_prob): Use streamed_locations.
7029 2018-07-31 Martin Liska <mliska@suse.cz>
7031 PR gcov-profile/85370
7032 * coverage.c (coverage_begin_function): Do not mark target
7033 clones as artificial functions.
7035 2018-07-31 Martin Liska <mliska@suse.cz>
7037 PR gcov-profile/83813
7038 PR gcov-profile/84758
7039 PR gcov-profile/85217
7040 PR gcov-profile/85332
7041 * profile.c (branch_prob): Do not record GOTO expressions
7042 for GIMPLE statements which locations are already streamed.
7044 2018-07-31 Olivier Hainque <hainque@adacore.com>
7046 * gcc.c (handle_spec_function): Accept a soft_matched_part
7047 argument, as do_spec_1. Pass it down to ...
7048 (eval_spec_function): Accept a soft_matched_part argument,
7049 and pass it down to ...
7050 (do_spec_2): Accept a soft_matched_part argument, and pass
7051 it down to do_spec_1.
7052 (do_spec_1): Pass soft_matched_part to handle_spec_function.
7053 (handle_braces): Update call to handle_spec_function.
7054 (driver::set_up_specs): Update calls to do_spec_2.
7055 (compare_debug_dump_opt_spec_function): Likewise.
7056 (compare_debug_self_opt_spec_function): Likewise.
7058 2018-07-31 Olivier Hainque <hainque@adacore.com>
7060 * common.opt (nolibc): New option.
7061 * doc/invoke.texi (Link Options): Document it.
7062 * gcc.c (LINK_GCC_C_SEQUENCE_SPEC): Honor nolibc.
7063 * config/alpha/linux.h: Likewise.
7064 * config/arc/elf.h: Likewise.
7065 * config/arm/uclinux-elf.h: Likewise.
7066 * config/arm/unknown-elf.h: Likewise.
7067 * config/avr/avrlibc.h: Likewise.
7068 * config/bfin/bfin.h: Likewise.
7069 * config/bfin/linux.h: Likewise.
7070 * config/bfin/uclinux.h: Likewise.
7071 * config/darwin.h: Likewise.
7072 * config/darwin10.h: Likewise.
7073 * config/darwin12.h: Likewise.
7074 * config/gnu-user.h: Likewise.
7075 * config/lm32/uclinux-elf.h: Likewise.
7076 * config/pa/pa-hpux11.h: Likewise.
7077 * config/pa/pa64-hpux.h: Likewise.
7078 * config/sparc/sparc.h: Likewise.
7080 2018-07-31 Olivier Hainque <hainque@adacore.com>
7082 * gcc.c (getenv_spec_function): Prepend '/' to value for allowed
7083 undefined variables.
7085 2018-07-30 Segher Boessenkool <segher@kernel.crashing.org>
7088 * config/arm/arm.c (arm_block_set_aligned_vect): Use gen_int_mode
7091 2018-07-30 Bernd Edlinger <bernd.edlinger@hotmail.de>
7093 * tree-ssa-forwprop.c (simplify_builtin_call): Don't create a not NUL
7094 terminated string literal.
7096 2018-07-30 Segher Boessenkool <segher@kernel.crashing.org>
7098 PR rtl-optimization/85160
7099 * combine.c (is_just_move): New function.
7100 (try_combine): Allow combining two instructions into two if neither of
7101 the original instructions was a move.
7103 2018-07-30 Alexander Monakov <amonakov@ispras.ru>
7106 * doc/extend.texi (Global Register Variables): Discourage use of type
7108 (Local Register Variables): Likewise.
7110 2018-07-30 Richard Sandiford <richard.sandiford@arm.com>
7112 PR tree-optimization/86506
7113 * hwint.h (ceil_log2): Resync with hwint.c implementation.
7115 2018-07-30 Ilya Leoshkevich <iii@linux.ibm.com>
7118 * lra-constraints.c (spill_hard_reg_in_range): When selecting the
7119 hard_regno, make sure no insn between `from` and `to` clobbers it.
7121 2018-07-30 Cesar Philippidis <cesar@codesourcery.com>
7122 Tom de Vries <tdevries@suse.de>
7124 * config/nvptx/nvptx.c (PTX_GANG_DEFAULT): Rename to ...
7125 (PTX_DEFAULT_RUNTIME_DIM): ... this.
7126 (nvptx_goacc_validate_dims): Set default worker and gang dims to
7127 PTX_DEFAULT_RUNTIME_DIM.
7128 (nvptx_dim_limit): Ignore GOMP_DIM_WORKER.
7130 2018-07-29 John David Anglin <danglin@gcc.gnu.org>
7132 * config/pa/pa.c (pa_output_addr_vec): Align address table.
7133 * config/pa/pa.h (JUMP_TABLES_IN_TEXT_SECTION): Revise comment.
7134 * config/pa/pa32-linux.h (JUMP_TABLES_IN_TEXT_SECTION): Define.
7136 2018-07-27 Michael Meissner <meissner@linux.ibm.com>
7138 * config/rs6000/constraints.md (wG constraint): Delete, no longer
7140 * config/rs6000/predicates.md (p9_fusion_reg_operand): Rename
7141 predicate to reflect toc fusion has been deleted.
7142 (toc_fusion_mem_raw): Delete, no longer used.
7143 (toc_fusion_mem_wrapped): Likewise.
7144 * config/rs6000/rs6000-cpus.def (POWERPC_MASKS): Delete toc
7146 * config/rs6000/rs6000-protos.h (fusion_wrap_memory_address):
7147 Delete, no longer used.
7148 * config/rs6000/rs6000.c (struct rs6000_reg_addr): Delete fields
7149 meant to be used for toc fusion.
7150 (rs6000_debug_print_mode): Delete toc fusion debugging.
7151 (rs6000_debug_reg_global): Likewise.
7152 (rs6000_init_hard_regno_mode_ok): Delete setting up fields for toc
7153 fusion and secondary reload support that were never used.
7154 (rs6000_option_override_internal): Delete TOC fusion, that was only
7155 partially defined, and it did not work unless you also used the
7157 (rs6000_legitimate_address_p): Delete TOC fusion support.
7158 (rs6000_opt_masks): Likewise.
7159 (fusion_wrap_memory_address): Delete function, no longer used.
7160 (fusion_split_address); Delete TOC fusion support.
7161 * config/rs6000/rs6000.h (TARGET_TOC_FUSION_INT): Delete, no
7162 longer used with toc fusion being deleted.
7163 (TARGET_TOC_FUSION_FP): Likewise.
7164 * config/rs6000/rs6000.md (UNSPEC_FUSION_ADDIS): Delete TOC fusion
7166 (toc fusion spliter): Delete TOC fusion support.
7167 (toc_fusionload_<mode>): Likewise.
7168 (toc_fusionload_di): Likewise.
7169 (fusion_gpr_load_<mode>): Delete generator function, this insn no
7170 longer needs to be named. Rename predicate to delete TOC fusion.
7171 (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_load): Likewise.
7172 (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_store): Likewise.
7173 (fusion_vsx_<P:mode>_<GPR_FUSION:mode>_load): Likewise.
7174 (fusion_vsx_<P:mode>_<GPR_FUSION:mode>_store): Likewise.
7175 (p9 fusion peephole2s): Rename predicate to delete TOC fusion.
7177 2018-07-27 Kelvin Nilsen <kelvin@gcc.gnu.org>
7179 * doc/extend.texi (Basic PowerPC Built-in Functions Available on
7180 ISA 2.05): Replace __uint128_t with __uint128 and __int128_t with
7181 __int128 in built-in function prototypes.
7182 (PowerPC AltiVec Built-in Functions on ISA 2.07): Likewise.
7183 (PowerPC AltiVec Built-in Functions on ISA 3.0): Likewise.
7185 2018-07-27 Martin Sebor <msebor@redhat.com>
7187 PR tree-optimization/86696
7188 * tree-ssa-strlen.c (get_min_string_length): Handle all integer
7189 types, including enums.
7190 (handle_char_store): Be prepared for the above function to fail.
7192 2018-07-26 Qing Zhao <qing.zhao@oracle.com>
7194 * builtins.c (inline_expand_builtin_string_cmp): Disable inlining
7195 when optimization level is lower than 2 or optimize for size.
7197 2018-07-26 Martin Sebor <msebor@redhat.com>
7199 PR tree-optimization/86043
7200 PR tree-optimization/86042
7201 * tree-ssa-strlen.c (handle_builtin_memcpy): Handle strict overlaps.
7202 (get_string_cst_length): Rename...
7203 (get_min_string_length): ...to this. Add argument.
7204 (handle_char_store): Extend to handle multi-character stores by
7206 * tree.c (initializer_zerop): Use new argument. Handle MEM_REF.
7207 * tree.h (initializer_zerop): Add argument.
7209 2018-07-26 Jakub Jelinek <jakub@redhat.com>
7212 * omp-low.c (scan_sharing_clauses): Don't ignore map clauses for
7213 declare target to variables if they have always,{to,from,tofrom} map
7216 2018-07-26 Martin Liska <mliska@suse.cz>
7219 * lto-wrapper.c: Add linker_output as prefix
7220 for ltrans_output_file.
7222 2018-07-26 Segher Boessenkool <segher@kernel.crashing.org>
7224 PR rtl-optimization/85805
7225 * combine.c (reg_nonzero_bits_for_combine): Only use the last set
7226 value for hard registers if that was written in the same mode.
7228 2018-07-26 Martin Liska <mliska@suse.cz>
7230 PR gcov-profile/86536
7231 * gcov.c (format_gcov): Use printf format %.*f directly
7232 and do not handle special values.
7234 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
7236 * common/config/arc/arc-common.c (arc_option_optimization_table):
7237 Update default optimizations for size.
7239 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
7241 * config/arc/arc.md (movsf_insn): Add short instruction selection.
7242 * config/arc/constraints.md (CfZ): New constraint.
7243 * config/arc/fpu.md (addssf3_fpu): Use CfZ constraint.
7244 (subsf3_fpu): Likewise.
7245 (cmpsf_fpu): Likewise.
7246 (cmpsf_fpu_uneq): Likewise.
7248 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
7250 * config/arc/arc.c (compact_memory_operand_p): Check for uncached
7252 (arc_is_uncached_mem_p): uncached applies to both the variable and
7255 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
7257 * config/arc/arc.h (ADDITIONAL_REGISTER_NAMES): Add additional
7260 2018-07-25 David Malcolm <dmalcolm@redhat.com>
7262 * optinfo-emit-json.cc (class optrecord_json_writer): Convert
7263 field "m_scopes" from vec to auto_vec.
7265 2018-07-25 Martin Liska <mliska@suse.cz>
7267 * config/powerpcspe/powerpcspe-protos.h (rs6000_loop_align): Fix
7270 2018-07-25 Richard Biener <rguenther@suse.de>
7273 * dwarf2out.c (dwarf2out_decl): Do not handle nested functions
7274 special wrt context_die late.
7275 (gen_subprogram_die): Re-use DIEs in local scope.
7277 2018-07-25 Richard Sandiford <richard.sandiford@arm.com>
7279 PR tree-optimization/86644
7280 * hwint.c (ceil_log2): Fix comment. Return 0 for 0.
7282 2018-07-25 Martin Liska <mliska@suse.cz>
7285 * dumpfile.c: And excluded values with TDF_ALL_VALUES.
7286 * dumpfile.h (enum dump_flag): Defince TDF_ALL_VALUES.
7288 2018-07-25 Martin Liska <mliska@suse.cz>
7291 * params.def: Explain ASan abbreviation and provide
7292 a documentation link.
7294 2018-07-24 Martin Sebor <msebor@redhat.com>
7296 PR tree-optimization/86622
7297 PR tree-optimization/86532
7298 * builtins.h (string_length): Declare.
7299 * builtins.c (c_strlen): Correct handling of non-constant offsets.
7300 (check_access): Be prepared for non-constant length ranges.
7301 (string_length): Make extern.
7302 * expr.c (string_constant): Only handle the minor non-constant
7303 array index. Use string_constant to compute the length of
7304 a generic string constant.
7306 2018-07-24 Richard Sandiford <richard.sandiford@arm.com>
7308 PR tree-optimization/86618
7309 * tree-vect-stmts.c (vectorizable_call): Don't take the address
7310 of LOOP_VINFO_MASKS (loop_vinfo) when loop_vinfo is null.
7312 2018-07-24 David Malcolm <dmalcolm@redhat.com>
7314 PR tree-optimization/86636
7315 * json.cc (json::object::set): Fix comment. Add assertions.
7316 (json::array::append): Move here from json.h. Add comment and an
7318 (json::string::string): Likewise.
7319 * json.h (json::array::append): Move to json.cc.
7320 (json::string::string): Likewise.
7321 * optinfo-emit-json.cc
7322 (optrecord_json_writer::impl_location_to_json): Assert that we
7323 aren't attempting to write out UNKNOWN_LOCATION, or an ad-hoc
7324 wrapper around it. Expand the location once, rather than three
7326 (optrecord_json_writer::inlining_chain_to_json): Fix the check for
7327 UNKNOWN_LOCATION, to use LOCATION_LOCUS to look through ad-hoc
7329 (optrecord_json_writer::optinfo_to_json): Likewise, in four
7330 places. Fix some overlong lines.
7332 2018-07-24 Matthew Malcomson <matthew.malcomson@arm.com>
7334 * config/aarch64/aarch64-simd.md
7335 (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w<mode>): Split into...
7336 (aarch64_<ANY_EXTEND:su>subw<mode>): ... This...
7337 (aarch64_<ANY_EXTEND:su>addw<mode>): ... And this.
7338 (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w<mode>_internal): Split into...
7339 (aarch64_<ANY_EXTEND:su>subw<mode>_internal): ... This...
7340 (aarch64_<ANY_EXTEND:su>addw<mode>_internal): ... And this.
7341 (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w2<mode>_internal): Split into...
7342 (aarch64_<ANY_EXTEND:su>subw2<mode>_internal): ... This...
7343 (aarch64_<ANY_EXTEND:su>addw2<mode>_internal): ... And this.
7345 2018-07-24 Jakub Jelinek <jakub@redhat.com>
7348 * expmed.c (expand_divmod): Punt if d == HOST_WIDE_INT_MIN
7349 and size > HOST_BITS_PER_WIDE_INT. For size > HOST_BITS_PER_WIDE_INT
7350 and abs_d == d, do the power of two handling if profitable.
7352 2018-07-24 Richard Biener <rguenther@suse.de>
7354 * match.pd: Add BIT_FIELD_REF canonicalizations.
7356 2018-07-23 Bernd Edlinger <bernd.edlinger@hotmail.de>
7359 * genmatch.c (dt_operand::gen_match_op): Avoid folding volatile values.
7361 2018-07-23 Bernd Edlinger <bernd.edlinger@hotmail.de>
7363 * gimple-fold.c (gimple_fold_builtin_printf): Don't create a not NUL
7364 terminated STRING_CST object.
7366 2018-07-23 Bernd Edlinger <bernd.edlinger@hotmail.de>
7368 hsa-dump.c (dump_hsa_symbol): Avoid out of scope access to buf.
7370 2018-07-23 Segher Boessenkool <segher@kernel.crashing.org>
7372 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Adjust.
7373 * config/rs6000/rs6000-protos.h (rs6000_split_v4si_init): Delete.
7374 * config/rs6000/rs6000.c (rs6000_expand_vector_init): Always force
7375 the elements into a register.
7376 (rs6000_split_v4si_init_di_reg): Delete.
7377 (rs6000_split_v4si_init): Delete.
7378 * config/rs6000/vsx.md (unspec): Delete UNSPEC_VSX_VEC_INIT.
7379 (vsx_init_v4si): Rewrite as a define_expand.
7381 2018-07-23 Segher Boessenkool <segher@kernel.crashing.org>
7383 * config/rs6000/rs6000.md (splitters for rldimi and rlwimi with the
7384 zero_extend argument from memory): New.
7386 2018-07-22 Martin Sebor <msebor@redhat.com>
7389 * gimple-ssa-warn-alloca.c (alloca_call_type_by_arg): Avoid
7390 diagnosing calls with unknown arguments unless -Walloca-larger-than
7391 is restricted to less than PTRDIFF_MAX bytes.
7393 2018-07-22 Gerald Pfeifer <gerald@pfeifer.com>
7395 * doc/gcov.texi (Invoking Gcov): Editorial changes.
7397 2018-07-20 David Malcolm <dmalcolm@redhat.com>
7399 * pretty-print.c (text_info::set_location): Remove redundant
7400 "line_table" parameter from call to rich_location::set_range.
7402 2018-07-20 Martin Sebor <msebor@redhat.com>
7405 * builtins.c (expand_builtin_alloca): Adjust.
7406 * calls.c (alloc_max_size): Simplify.
7407 * cgraphunit.c (cgraph_node::expand): Adjust.
7408 * common.opt (larger_than_size, warn_frame_larger_than): Remove
7410 (frame_larger_than_size): Same.
7411 (-Wframe-larger-than, -Wlarger-than, -Wstack-usage): Change options
7412 to take a HOST_WIDE_INT argument and accept a byte-size suffix.
7414 * doc/invoke.texi (GCC Command Options): Document option arguments.
7415 Explain byte-size arguments and suffixes.
7416 (-Wvla-larger-than, -Wno-alloc-size-larger-than): Update.
7417 (-Wno-alloca-larger-than, -Wno-vla-larger-than): Same.
7418 (-Wframe-larger-than, -Wlarger-than, -Wstack-usage): Same.
7419 * doc/options.texi (UInteger): Expand.
7420 (Host_Wide_Int, ByteSize): Document new properties.
7421 * final.c (final_start_function_1): Include sizes in an error message.
7422 * function.c (frame_offset_overflow): Same.
7423 * gimple-ssa-warn-alloca.c (pass_walloca::gate): Adjust.
7424 (alloca_call_type_by_arg): Change function argument to HOST_WIDE_INT.
7425 Diagnose unbounded alloca calls only for limits of less than
7427 (alloca_call_type): Adjust. Diagnose possibly out-of-bounds alloca
7428 calls and VLA size only for limits of less than PTRDIFF_MAX. Same
7430 (pass_walloca::execute): Adjust. Diagnose alloca calls in loops
7431 only for limits of less than PTRDIFF_MAX.
7432 * langhooks-def.h (lhd_handle_option): Change function argument
7434 * langhooks.c (lhd_handle_option): Same.
7435 * langhooks.h (handle_option): Same.
7436 * opt-functions.awk (switch_bit_fields): Handle Host_Wide_Int and
7438 (var_type, var_type_struct): Same.
7439 (var_set): Handle ByteSize flag.
7440 * optc-gen.awk: Add comments to output to ease debugging. Make
7441 use of HOST_WIDE_INT where appropriate.
7442 * opts-gen-save.awk: Use %lx to format unsigned long.
7443 * opth-gen.awk: Change function argument to HOST_WIDE_INT.
7444 * opts-common.c (integral_argument): Return HOST_WIDE_INT and add
7445 arguments. Parse bytes-size suffixes.
7446 (enum_arg_to_value): Change function argument to HOST_WIDE_INT.
7447 (enum_value_to_arg): Same.
7448 (decode_cmdline_option): Handle cl_host_wide_int. Adjust.
7449 (handle_option): Adjust.
7450 (generate_option): Change function argument to HOST_WIDE_INT.
7451 (cmdline_handle_error): Adjust.
7452 (read_cmdline_option): Change function argument to HOST_WIDE_INT.
7453 (set_option): Change function argument to HOST_WIDE_INT.
7454 (option_enabled): Handle cl_host_wide_int.
7455 (get_option_state): Handle CLVC_SIZE.
7456 (control_warning_option): Same.
7457 * opts.c (common_handle_option): Change function argument to
7458 HOST_WIDE_INT. Remove handling of OPT_Walloca_larger_than_ and
7459 OPT_Wvla_larger_than_.
7460 * opts.h (enum cl_var_type): Add an enumerator.
7461 * stor-layout.c (layout_decl): Print a more meaningful warning.
7462 * toplev.c (output_stack_usage): Adjust.
7464 2018-07-20 Qing Zhao <qing.zhao@oracle.com>
7466 * builtins.c (expand_builtin_memcmp): Delete the last parameter for
7467 call to inline_expand_builtin_string_cmp.
7468 (expand_builtin_strcmp): Likewise.
7469 (expand_builtin_strncmp): Likewise.
7470 (inline_string_cmp): Delete the last parameter, change char_type_node
7471 to unsigned_char_type_node for strcmp/strncmp, add conversions to the
7473 (inline_expand_builtin_string_cmp): Delete the last parameter, give up
7474 the inlining expansion on target where the type of the call has same or
7475 narrower precision than unsigned char.
7477 2018-07-20 David Malcolm <dmalcolm@redhat.com>
7479 * Makefile.in (OBJS): Add json.o and optinfo-emit-json.o.
7480 (CFLAGS-optinfo-emit-json.o): Define TARGET_NAME.
7481 * common.opt (fsave-optimization-record): New option.
7482 * coretypes.h (struct kv_pair): Move here from dumpfile.c.
7483 * doc/invoke.texi (-fsave-optimization-record): New option.
7484 * dumpfile.c: Include "optinfo-emit-json.h".
7485 (struct kv_pair): Move to coretypes.h.
7486 (optgroup_options): Make non-static.
7487 (dump_context::end_scope): Call
7488 optimization_records_maybe_pop_dump_scope.
7489 * dumpfile.h (optgroup_options): New decl.
7490 * json.cc: New file.
7492 * optinfo-emit-json.cc: New file.
7493 * optinfo-emit-json.h: New file.
7494 * optinfo.cc: Include "optinfo-emit-json.h".
7495 (optinfo::emit): Call optimization_records_maybe_record_optinfo.
7496 (optinfo_enabled_p): Check optimization_records_enabled_p.
7497 (optinfo_wants_inlining_info_p): Likewise.
7498 * optinfo.h: Update comment.
7499 * profile-count.c (profile_quality_as_string): New function.
7500 * profile-count.h (profile_quality_as_string): New decl.
7501 (profile_count::quality): New accessor.
7502 * selftest-run-tests.c (selftest::run_tests): Call json_cc_tests
7503 and optinfo_emit_json_cc_tests.
7504 * selftest.h (selftest::json_cc_tests): New decl.
7505 (selftest::optinfo_emit_json_cc_tests): New decl.
7506 * toplev.c: Include "optinfo-emit-json.h".
7507 (compile_file): Call optimization_records_finish.
7508 (do_compile): Call optimization_records_start.
7509 * tree-ssa-live.c: Include optinfo.h.
7510 (remove_unused_scope_block_p): Retain inlining information if
7511 optinfo_wants_inlining_info_p returns true.
7513 2018-07-20 Richard Biener <rguenther@suse.de>
7516 * dwarf2out.c (dwarf2out_die_ref_for_decl): Test in_lto_p
7517 to cover -flto-partition=none.
7519 2018-07-20 Martin Liska <mliska@suse.cz>
7521 * tree.h (DECL_LOCATION_RANGE): Remove unused macro.
7522 (get_decl_source_range): Remove unused function.
7524 2018-07-20 Richard Biener <rguenther@suse.de>
7526 * tree-ssa-sccvn.h (struct vn_nary_op_s): Add next member.
7527 (struct vn_phi_s): Likewise.
7528 (struct vn_reference_s): Likewise.
7529 * tree-ssa-sccvn.c (vn_nary_op_hasher::equal): Add shortcut
7530 for searching the slot of an entry known to be in the hash itself.
7531 (vn_phi_hasher::equal): Likewise.
7532 (vn_reference_hasher::equal): Likewise.
7533 (last_inserted_ref, last_inserted_phi, last_inserted_nary): New
7535 (optimistic_info, current_info): Remove, keeping only valid_info.
7536 (vn_reference_lookup_1): Remove fallback lookup.
7537 (vn_reference_lookup_2): Likewise.
7538 (vn_nary_op_lookup_1): Likewise.
7539 (vn_phi_lookup): Likewise.
7540 (vn_nary_build_or_lookup_1): Make sure to not chain the built
7542 (vn_reference_insert): Adjust, chain the inserted hash element
7543 at last_inserted_ref.
7544 (vn_reference_insert_pieces): Likewise.
7545 (visit_reference_op_call): Likewise.
7546 (vn_nary_op_insert_into): Chain the inserted hash element at
7548 (vn_nary_op_insert_pieces): Adjust.
7549 (vn_nary_op_insert): Likewise.
7550 (vn_nary_op_insert_stmt): Likewise.
7551 (vn_phi_insert): Adjust, chain the inserted hash element at
7553 (process_scc): Remove clearing and copying the optimistic
7554 table. Instead remove elements inserted during an optimistic
7555 iteration from the single table we maintain.
7556 (init_scc_vn): Adjust.
7557 (free_scc_vn): Likewise.
7558 (sccvn_dom_walker::record_cond): Likewise.
7559 (sccvn_dom_walker::after_dom_children): Likewise.
7561 2018-07-19 Martin Sebor <msebor@redhat.com>
7563 PR tree-optimization/84047
7564 PR tree-optimization/83776
7565 * tree-vrp.c (vrp_prop::check_mem_ref): New function.
7566 (check_array_bounds): Call it.
7568 2018-07-19 Martin Sebor <msebor@redhat.com>
7570 * align.h (align_flags): Use member initialization.
7572 2018-07-19 David Malcolm <dmalcolm@redhat.com>
7574 * Makefile.in (OBJS): Add optinfo.o.
7575 * coretypes.h (class symtab_node): New forward decl.
7576 (struct cgraph_node): New forward decl.
7577 (class varpool_node): New forward decl.
7578 * dump-context.h: New file.
7579 * dumpfile.c: Include "optinfo.h", "dump-context.h", "cgraph.h",
7581 (refresh_dumps_are_enabled): Use optinfo_enabled_p.
7582 (set_dump_file): Call dumpfile_ensure_any_optinfo_are_flushed.
7583 (set_alt_dump_file): Likewise.
7584 (dump_context::~dump_context): New dtor.
7585 (dump_gimple_stmt): Move implementation to...
7586 (dump_context::dump_gimple_stmt): ...this new member function.
7587 Add the stmt to any pending optinfo, creating one if need be.
7588 (dump_gimple_stmt_loc): Move implementation to...
7589 (dump_context::dump_gimple_stmt_loc): ...this new member function.
7590 Start a new optinfo and add the stmt to it.
7591 (dump_gimple_expr): Move implementation to...
7592 (dump_context::dump_gimple_expr): ...this new member function.
7593 Add the stmt to any pending optinfo, creating one if need be.
7594 (dump_gimple_expr_loc): Move implementation to...
7595 (dump_context::dump_gimple_expr_loc): ...this new member function.
7596 Start a new optinfo and add the stmt to it.
7597 (dump_generic_expr): Move implementation to...
7598 (dump_context::dump_generic_expr): ...this new member function.
7599 Add the tree to any pending optinfo, creating one if need be.
7600 (dump_generic_expr_loc): Move implementation to...
7601 (dump_context::dump_generic_expr_loc): ...this new member
7602 function. Add the tree to any pending optinfo, creating one if
7604 (dump_printf): Move implementation to...
7605 (dump_context::dump_printf_va): ...this new member function. Add
7606 the text to any pending optinfo, creating one if need be.
7607 (dump_printf_loc): Move implementation to...
7608 (dump_context::dump_printf_loc_va): ...this new member function.
7609 Start a new optinfo and add the stmt to it.
7610 (dump_dec): Move implementation to...
7611 (dump_context::dump_dec): ...this new member function. Add the
7612 value to any pending optinfo, creating one if need be.
7613 (dump_context::dump_symtab_node): New member function.
7614 (dump_context::get_scope_depth): New member function.
7615 (dump_context::begin_scope): New member function.
7616 (dump_context::end_scope): New member function.
7617 (dump_context::ensure_pending_optinfo): New member function.
7618 (dump_context::begin_next_optinfo): New member function.
7619 (dump_context::end_any_optinfo): New member function.
7620 (dump_context::s_current): New global.
7621 (dump_context::s_default): New global.
7622 (dump_scope_depth): Delete global.
7623 (dumpfile_ensure_any_optinfo_are_flushed): New function.
7624 (dump_symtab_node): New function.
7625 (get_dump_scope_depth): Reimplement in terms of dump_context.
7626 (dump_begin_scope): Likewise.
7627 (dump_end_scope): Likewise.
7628 (selftest::temp_dump_context::temp_dump_context): New ctor.
7629 (selftest::temp_dump_context::~temp_dump_context): New dtor.
7630 (selftest::verify_item): New function.
7631 (ASSERT_IS_TEXT): New macro.
7632 (ASSERT_IS_TREE): New macro.
7633 (ASSERT_IS_GIMPLE): New macro.
7634 (selftest::test_capture_of_dump_calls): New test.
7635 (selftest::dumpfile_c_tests): Call it.
7636 * dumpfile.h (dump_printf, dump_printf_loc, dump_basic_block)
7637 (dump_generic_expr_loc, dump_generic_expr, dump_gimple_stmt_loc)
7638 (dump_gimple_stmt, dump_dec): Gather these related decls and add a
7639 descriptive comment.
7640 (dump_function, print_combine_total_stats, enable_rtl_dump_file)
7641 (dump_node, dump_bb): Move these unrelated decls.
7642 (class dump_manager): Add leading comment.
7643 * optinfo.cc: New file.
7644 * optinfo.h: New file.
7646 2018-07-19 Michael Collison <michael.collison@arm.com>
7647 Richard Henderson <rth@redhat.com>
7649 * config/aarch64/aarch64.md (subv<GPI>4, usubv<GPI>4): New patterns.
7650 (subti): Handle op1 zero.
7651 (subvti4, usub4ti4): New.
7652 (*sub<GPI>3_compare1_imm): New.
7653 (sub<GPI>3_carryinCV): New.
7654 (*sub<GPI>3_carryinCV_z1_z2, *sub<GPI>3_carryinCV_z1): New.
7655 (*sub<GPI>3_carryinCV_z2, *sub<GPI>3_carryinCV): New.
7657 2018-07-19 Michael Collison <michael.collison@arm.com>
7658 Richard Henderson <rth@redhat.com>
7660 * config/aarch64/aarch64.md: (addv<GPI>4, uaddv<GPI>4): New.
7661 (addti3): Create simpler code if low part is already known to be 0.
7662 (addvti4, uaddvti4): New.
7663 (*add<GPI>3_compareC_cconly_imm): New.
7664 (*add<GPI>3_compareC_cconly): New.
7665 (*add<GPI>3_compareC_imm): New.
7666 (*add<GPI>3_compareC): Rename from add<GPI>3_compare1; do not
7667 handle constants within this pattern..
7668 (*add<GPI>3_compareV_cconly_imm): New.
7669 (*add<GPI>3_compareV_cconly): New.
7670 (*add<GPI>3_compareV_imm): New.
7671 (add<GPI>3_compareV): New.
7672 (add<GPI>3_carryinC, add<GPI>3_carryinV): New.
7673 (*add<GPI>3_carryinC_zero, *add<GPI>3_carryinV_zero): New.
7674 (*add<GPI>3_carryinC, *add<GPI>3_carryinV): New.
7675 ((*add<GPI>3_compareC_cconly_imm): Replace 'ne' operator
7676 with 'comparison' operator.
7677 (*add<GPI>3_compareV_cconly_imm): Ditto.
7678 (*add<GPI>3_compareV_cconly): Ditto.
7679 (*add<GPI>3_compareV_imm): Ditto.
7680 (add<GPI>3_compareV): Ditto.
7681 (add<mode>3_carryinC): Ditto.
7682 (*add<mode>3_carryinC_zero): Ditto.
7683 (*add<mode>3_carryinC): Ditto.
7684 (add<mode>3_carryinV): Ditto.
7685 (*add<mode>3_carryinV_zero): Ditto.
7686 (*add<mode>3_carryinV): Ditto.
7688 2018-07-19 Michael Collison <michael.collison@arm.com>
7689 Richard Henderson <rth@redhat.com>
7691 * config/aarch64/aarch64-modes.def (CC_V): New.
7692 * config/aarch64/aarch64-protos.h
7693 (aarch64_addti_scratch_regs): Declare
7694 (aarch64_subvti_scratch_regs): Declare.
7695 (aarch64_expand_subvti): Declare.
7696 (aarch64_gen_unlikely_cbranch): Declare
7697 * config/aarch64/aarch64.c (aarch64_select_cc_mode): Test
7698 for signed overflow using CC_Vmode.
7699 (aarch64_get_condition_code_1): Handle CC_Vmode.
7700 (aarch64_gen_unlikely_cbranch): New function.
7701 (aarch64_addti_scratch_regs): New function.
7702 (aarch64_subvti_scratch_regs): New function.
7703 (aarch64_expand_subvti): New function.
7705 2018-07-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
7707 * config/aarch64/aarch64-option-extensions.def: New entry for profile
7709 * config/aarch64/aarch64.h (AARCH64_FL_PROFILE): New.
7710 * doc/invoke.texi (aarch64-feature-modifiers): New entry for profile
7713 2018-07-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
7716 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
7717 address check not strict.
7719 2018-07-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
7721 * config/aarch64/aarch64-simd.md (aarch64_simd_mov<VQ:mode>): Replace
7723 (store_pair_lanes<mode>): Likewise.
7724 * config/aarch64/aarch64-protos.h (aarch64_addr_query_type): Add new
7725 enum value 'ADDR_QUERY_LDP_STP_N'.
7726 * config/aarch64/aarch64.c (aarch64_addr_query_type): Likewise.
7727 (aarch64_print_address_internal): Add declaration.
7728 (aarch64_print_ldpstp_address): Remove.
7729 (aarch64_classify_address): Adapt mode for 'ADDR_QUERY_LDP_STP_N'.
7730 (aarch64_print_operand): Change printing of 'y'.
7731 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Use
7732 new enum value 'ADDR_QUERY_LDP_STP_N', don't hardcode mode and use
7733 'true' rather than '1'.
7734 * config/aarch64/constraints.md (Uml): Likewise.
7735 (Uml): Rename to Umn.
7738 2018-07-19 Richard Biener <rguenther@suse.de>
7740 * tree-ssa-sccvn.h (struct vn_phi_s): Make phiargs member
7742 * tree-ssa-sccvn.c: Remove alloc-pool.h use.
7743 (vn_phi_hasher): Derive from nofree_ptr_hash and remove remove method.
7744 (vn_reference_hasher): Likewise.
7745 (struct vn_tables_s): Remove obstack and alloc-pool members.
7746 (vn_tables_obstack, vn_tables_insert_obstack): New global obstacks.
7747 (vn_nary_build_or_lookup_1): Manually build in vn_tables_insert_obstack.
7748 (vn_reference_insert): Allocate from obstack instead of from alloc-pool.
7749 (vn_reference_insert_pieces): Likewise.
7750 (alloc_vn_nary_op_noinit): Adjust.
7751 (vn_nary_op_insert_stmt): Allocate phiargs in-place.
7752 (vn_phi_eq): Adjust.
7753 (shared_lookup_phiargs): Remove.
7754 (vn_phi_lookup): Allocate temporary vn_phi_s on the stack.
7755 (vn_phi_insert): Allocate from obstack instead of from alloc-pool.
7756 (visit_reference_op_call): Likewise.
7757 (copy_nary, copy_phi, copy_reference): Remove.
7758 (process_scc): Rewind the obstack when iterating. Do not
7759 copy the elements to valid_info but just move them from one
7760 hashtable to the other.
7761 (allocate_vn_table): Adjust.
7762 (free_vn_table): Likewise.
7763 (init_scc_vn): Likewise.
7764 (free_scc_vn): Likewise.
7766 2018-07-19 H.J. Lu <hongjiu.lu@intel.com>
7769 * config/i386/i386.c (rest_of_insert_endbranch): Lookup
7770 indirect_return as function type attribute.
7771 (ix86_attribute_table): Change indirect_return to function
7773 * doc/extend.texi: Update indirect_return attribute.
7775 2018-07-19 Aldy Hernandez <aldyh@redhat.com>
7777 * wide-int.h (widest2_int): New.
7778 * gimple-fold.c (arith_overflowed_p): Use it.
7779 * tree.h (widest2_int_cst): New.
7780 * tree-vrp.c (wide_int_binop_overflow): Rename from
7781 vrp_int_const_binop.
7782 Rewrite to work on trees.
7783 (extract_range_from_multiplicative_op_1): Abstract code to...
7784 (wide_int_range_min_max): ...here.
7785 (wide_int_range_cross_product): ...and here.
7786 (extract_range_from_binary_expr_1): Abstract overflow code to...
7787 (wide_int_range_mult_wrapping): ...here.
7788 * tree-vrp.h (wide_int_range_cross_product): New.
7789 (wide_int_range_mult_wrapping): New.
7791 2018-07-19 Andrew Senkevich <andrew.senkevich@intel.com>
7792 Julia Koval <julia.koval@intel.com>
7794 * config/i386/x86-tune-costs.h (skylake_memcpy,
7795 skylake_memset): Replace rep_prefix with unrolling for size 512.
7797 2018-07-18 Kugan Vivekanandarajah <kuganv@linaro.org>
7800 * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): Handle
7801 comparision with EQ_EXPR in last stmt.
7803 2018-07-18 Kelvin Nilsen <kelvin@gcc.gnu.org>
7805 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Rename
7806 this subsection to "PowerPC AltiVec/VSX Built-in Functions".
7807 (PowerPC AltiVec/VSX Built-in Functions): New name for subsection
7808 previously known as "PowerPC AltiVec Built-in Functions". Move
7809 some material to new subsubsections "PowerPC AltiVec Built-in
7810 Functions on ISA 2.06" and "PowerPC AltiVec Built-in Functions on
7812 (PowerPC Altivec Built-in Functions on ISA 2.05): New subsubsection.
7813 (PowerPC Altivec Built-in Functions on ISA 2.06): Likewise.
7814 (PowerPC Altivec Built-in Functions on ISA 2.07): Likewise.
7815 (PowerPC Altivec Built-in Functions on ISA 3.0): Likewise.
7817 2018-07-18 Richard Biener <rguenther@suse.de>
7819 PR tree-optimization/86557
7820 * tree-vect-patterns.c (vect_recog_divmod_pattern): Also handle
7823 2018-07-18 Ilya Leoshkevich <iii@linux.ibm.com>
7825 * config/s390/s390.c (s390_function_profiler): Generate CFI.
7827 2018-07-17 Jeff Law <law@redhat.com>
7829 * config/arm/arm.c (get_label_padding): Update for recent
7830 changes to label_to_alignment.
7832 PR tree-optimization/86010
7833 * tree-ssa-dse.c (compute_trims): Fix typo/thinko.
7835 * config/mips/mips.c (vr4130_align_insns): Update for recent
7836 changes to label_to_alignment.
7838 * config/frv/frv.c (frv_label_align): Update for recent changes
7839 to label_to_alignment.
7841 * config/nios2/nios2.c (nios2_label_align): Update for recent
7842 changes which dropped ALIGN_LABELS_LOG.
7844 2018-07-17 Andreas Schwab <schwab@linux-m68k.org>
7846 * config/m68k/m68k.md (umulsi3_highpart+1, const_umulsi3_highpart)
7847 (smulsi3_highpart+1, const_smulsi3_highpart): Add CC_STATUS_INIT.
7849 2018-07-17 Claudiu Zissulescu <claziss@synopsys.com>
7851 * config/arc/arc.c (arc_label_align): Use align_labels instead of
7852 deprecated align_labels_log.
7854 2018-07-17 Richard Biener <rguenther@suse.de>
7857 * dwarf2out.c (init_sections_and_labels): Always generate
7858 a debug_line_str_section for early LTO debug.
7859 (dwarf2out_finish): Reset debug_line_str_hash output early.
7860 Bump counter for extra dwarf5 .debug_loc labels to not conflict
7862 (dwarf2out_early_finish): Output debug_line_str.
7864 2018-07-17 Robin Dapp <rdapp@linux.ibm.com>
7866 * config/s390/s390.c (preferred_la_operand_p): Do not use LA with
7867 index register on z196 or later.
7869 2018-07-17 Robin Dapp <rdapp@linux.ibm.com>
7871 * config/s390/s390.c (s390_default_align): Set default function
7873 (s390_override_options_after_change): Call s390_default align.
7874 (s390_option_override_internal): Call s390_default_align.
7875 (TARGET_OVERRIDE_OPTIONS_AFTER_CHANGE): Define.
7877 2018-07-17 Jakub Jelinek <jakub@redhat.com>
7880 * omp-low.c (create_task_copyfn): Copy over also fields corresponding
7881 to _looptemp_ clauses, other than the first two.
7883 2018-07-17 Martin Liska <mliska@suse.cz>
7885 * opts.c: Do not enable OPT_falign_* for -Os.
7887 2018-07-17 Martin Liska <mliska@suse.cz>
7889 * align.h (MAX_CODE_ALIGN): New.
7890 (MAX_CODE_ALIGN_VALUE): New.
7891 * common/config/i386/i386-common.c (ix86_handle_option):
7892 (MAX_CODE_ALIGN): Moved to align.h.
7893 * final.c (MAX_CODE_ALIGN): Likewise.
7894 * opts.c (parse_and_check_align_values):
7895 (MAX_CODE_ALIGN): Likewise.
7896 (MAX_CODE_ALIGN_VALUE): Likewise.
7898 2018-07-17 Martin Liska <mliska@suse.cz>
7900 * config/i386/att.h (ASM_OUTPUT_ALIGN): Fix spacing
7901 in order to fulfil coding style.
7902 * config/i386/cygming.h (ASM_OUTPUT_ALIGN): Likewise.
7903 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
7904 * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
7905 * config/iq2000/iq2000.h (ASM_OUTPUT_ALIGN): Likewise.
7906 * config/pa/pa.h (ASM_OUTPUT_ALIGN): Likewise.
7907 * config/sparc/sol2.h (ASM_OUTPUT_ALIGN_WITH_NOP): Likewise.
7908 * config/sparc/sparc.h (ASM_OUTPUT_ALIGN): Likewise.
7909 * config/visium/visium.h (ASM_OUTPUT_ALIGN): Likewise.
7910 (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
7912 2018-07-17 Martin Liska <mliska@suse.cz>
7914 * align.h: New file.
7915 * config/alpha/alpha.c (alpha_align_insns_1): Use align_functions
7917 * config/i386/i386.c (ix86_avoid_jump_mispredicts): Use new return type
7918 align_flags of label_to_alignment.
7919 * config/m32r/m32r.h (LOOP_ALIGN): Wrap returned values into
7921 * config/m68k/m68k.c: Do not use removed align_labels_value and
7923 * config/nds32/nds32.h (JUMP_ALIGN): Wrap result into align_flags class.
7924 (LOOP_ALIGN): Likewise.
7925 (LABEL_ALIGN): Likewise.
7926 * config/powerpcspe/powerpcspe.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
7927 Remove not used macro.
7928 (rs6000_loop_align): Change return type to align_flags.
7929 (rs6000_loop_align_max_skip): Remove.
7930 * config/rs6000/rs6000-protos.h (rs6000_loop_align):
7931 Change return type to align_flags.
7932 * config/rs6000/rs6000.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
7933 Remove not used macro.
7934 (rs6000_loop_align): Change return type to align_flags.
7935 (rs6000_loop_align_max_skip): Remove.
7936 * config/rx/rx.h (JUMP_ALIGN): Wrap integer values
7937 * config/rx/rx-protos.h (rx_align_for_label): Make it
7939 * config/rx/rx.c (rx_align_for_label): Change return type
7941 (rx_max_skip_for_label): Remove TARGET_ASM_*_ALIGN_MAX_SKIP
7943 into align_flags class.
7944 (LABEL_ALIGN): Likewise.
7945 (LOOP_ALIGN): Likewise.
7946 * config/s390/s390.c (s390_label_align): Use align_flags
7948 (s390_asm_output_function_label): Likewise.
7949 * config/sh/sh.c (sh_override_options_after_change):
7950 Use align_flags class directly without macros.
7951 (find_barrier): Likewise.
7952 (barrier_align): Likewise.
7953 (sh_loop_align): Likewise.
7954 * config/spu/spu.c (spu_option_override):
7955 Use align_flags_tuple::get_value instead of removed macros.
7956 (spu_sched_init): Likewise.
7957 * config/spu/spu.h (GTY): Likewise.
7958 * config/visium/visium.c (visium_option_override):
7959 Set "8" as default secondary alignment.
7960 * config/visium/visium.h (SUBALIGN_LOG): Define to 3
7961 in order to guarantee secondary alignment of 8.
7962 * coretypes.h: Include align.h header file.
7963 * doc/tm.texi: Remove TARGET_ASM_JUMP_ALIGN_MAX_SKIP,
7964 TARGET_ASM_LOOP_ALIGN_MAX_SKIP, TARGET_ASM_LABEL_ALIGN_MAX_SKIP
7965 and TARGET_ASM_LABEL_ALIGN_AFTER_BARRIER_MAX_SKIP macros.
7966 * doc/tm.texi.in: Likewise.
7967 * final.c (struct label_alignment): Remove not used structure.
7968 (LABEL_ALIGN): Change type to align_flags.
7969 (LOOP_ALIGN): Likewise.
7970 (JUMP_ALIGN): Likewise.
7971 (default_loop_align_max_skip): Remove.
7972 (default_label_align_max_skip): Likewise.
7973 (default_jump_align_max_skip): Likewise.
7974 (default_label_align_after_barrier_max_skip):
7975 (LABEL_TO_ALIGNMENT): Change to access label_align vector.
7976 (LABEL_TO_MAX_SKIP): Remove.
7977 (label_to_alignment): Return align_flags type instead of integer.
7978 (label_to_max_skip): Remove.
7979 (align_fuzz): Use align_flags type.
7980 (compute_alignments): Use align_flags type and use align_flags::max
7981 to combine multiple alignments.
7982 (grow_label_align): Grow vec instead of C array.
7983 (update_alignments): Assign just LABEL_TO_ALIGNMENT.
7984 (shorten_branches): Use align_flags type and use align_flags::max
7985 to combine multiple alignments.
7986 (final_scan_insn_1): Remove usage of secondary alignment that comes
7987 from label alignment, but instead use proper secondary alignment
7988 which is computed in grow_label_align.
7989 * flags.h (struct align_flags_tuple): Move to align.h.
7990 (struct align_flags): Likewise.
7991 (state_align_loops): Rename to align_loops.
7992 (state_align_jumps): Rename to align_jumps.
7993 (state_align_labels): Rename to align_labels.
7994 (state_align_functions): Rename to align_functions.
7995 (align_loops_log): Remove.
7996 (align_jumps_log): Remove.
7997 (align_labels_log): Remove.
7998 (align_functions_log): Remove.
7999 (align_loops_max_skip): Remove.
8000 (align_jumps_max_skip): Remove.
8001 (align_labels_max_skip): Remove.
8002 (align_functions_max_skip): Remove.
8003 (align_loops_value): Remove.
8004 (align_jumps_value): Remove.
8005 (align_labels_value): Remove.
8006 (align_functions_value): Remove.
8007 * output.h (label_to_alignment): Change return type to align_flags.
8008 (label_to_max_skip): Remove.
8009 * target.def: Remove loop_align_max_skip, label_align_max_skip,
8010 jump_align_max_skip macros.
8011 * targhooks.h (default_loop_align_max_skip): Remove.
8012 (default_label_align_max_skip): Likewise.
8013 (default_jump_align_max_skip): Likewise.
8014 (default_label_align_after_barrier_max_skip): Remove.
8015 * toplev.c (read_log_maxskip): Use ::normalize function.
8016 (parse_N_M): Remove not used argument and also call ::normalize.
8017 (parse_alignment_opts): Do not pass unused arguments.
8018 * varasm.c (assemble_start_function): Use directly align_functions
8019 instead of removed macros.
8020 * system.h: Do not poison removed macros.
8022 2018-07-17 Jakub Jelinek <jakub@redhat.com>
8025 * gimplify.c (gimplify_omp_for): Ensure taskloop firstprivatized init
8026 and cond temporaries don't have reference type if iterator has
8027 pointer type. For init use &for_pre_body instead of pre_p if
8028 for_pre_body is non-empty.
8030 2018-07-16 Segher Boessenkool <segher@kernel.crashing.org>
8032 * config/rs6000/rs6000.md (trunc<mode>sf2): Expand truncates of
8033 double-double modes to SFmode directly directly.
8034 (trunc<mode>sf2_fprs): Delete.
8036 2018-07-16 Segher Boessenkool <segher@kernel.crashing.org>
8038 * config/rs6000/rs6000.c (init_float128_ibm): Use the correct names
8039 for conversions between IFmode and the decimal floating point modes.
8040 (init_float128_ieee): Use the correct names for conversions between
8041 KFmode and the decimal floating point modes.
8043 2018-07-16 Segher Boessenkool <segher@kernel.crashing.org>
8045 * config/rs6000/rs6000.c (init_float128_ibm): Use more correct names
8046 for the conversions between TDmode and IFmode.
8047 (init_float128_ieee): Use more correct names for the conversions
8048 between TDmode and KFmode.
8050 2018-07-16 Jakub Jelinek <jakub@redhat.com>
8052 PR tree-optimization/86526
8053 * builtins.c (expand_builtin_memcmp): Formatting fixes.
8054 (inline_expand_builtin_string_cmp): Likewise.
8055 (inline_string_cmp): Likewise. Use c_readstr instead of
8056 builtin_memcpy_read_str. Add unit_mode temporary.
8058 2018-07-16 Bernd Edlinger <bernd.edlinger@hotmail.de>
8061 * builtins.c (check_access): Bail out if range[0] is no INTEGER_CST.
8062 * expr.c (string_constant): Fix the element size of ARRAY_TYPE.
8064 2018-07-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
8066 * doc/extend.texi (PowerPC AltiVec Built-in Functions):
8067 Alphabetize prototypes of built-in functions, separating out
8068 built-in functions that are listed in this section but should be
8069 described elsewhere.
8071 2018-07-16 Uros Bizjak <ubizjak@gmail.com>
8074 * expmed.c (emit_store_flag): Do not emit setcc followed by a
8075 conditional move when trapping comparison was split to a
8076 non-trapping one (and vice versa).
8078 2018-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
8080 * config/s390/s390.c (s390_function_profiler): Generate nops
8081 instead of profiler call sequences.
8082 * config/s390/s390.opt: Add the new option.
8084 2018-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
8086 * config/s390/s390.c (s390_function_profiler): Generate
8087 __mcount_loc section.
8088 * config/s390/s390.opt: Add the new option.
8090 2018-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
8092 * common.opt: Add the new warning.
8093 * config/s390/s390.c (s390_function_profiler): Emit "brasl
8094 %r0,__fentry__" when -mfentry is specified.
8095 (s390_option_override_internal): Disallow -mfentry for 31-bit
8097 * config/s390/s390.opt: Add the new option.
8099 2018-07-16 Richard Biener <rguenther@suse.de>
8102 * dwarf2out.c (dwarf2out_register_external_die): Assign DIE parents
8103 for function-local FUNCTION_DECL and RESULT_DECL immediately.
8105 2018-07-16 Martin Liska <mliska@suse.cz>
8108 * ipa-pure-const.c (malloc_candidate_p): Revert ::get
8111 2017-07-16 Claudiu Zissulescu <claziss@synopsys.com>
8113 * config/arc/arcHS.md: Update ARCHS scheduling rules.
8115 2017-07-16 Claudiu Zissulescu <claziss@synopsys.com>
8117 * config/arc/arc-arch.h (arc_tune_attr): Add new tune parameters
8119 * config/arc/arc-cpus.def (hs4x): New cpu.
8121 * config/arc/arc-tables.opt: Regenerate.
8122 * config/arc/arc.c (arc_sched_issue_rate): New function.
8123 (TARGET_SCHED_ISSUE_RATE): Define.
8124 (TARGET_SCHED_EXPOSED_PIPELINE): Likewise.
8125 * config/arc/arc.md (attr type): Add fpu_fuse, fpu_sdiv, fpu_ddiv,
8127 (attr tune): Add ARCHS4x tune values.
8128 (attr tune_dspmpy): Define.
8129 (*tst): Correct instruction type.
8130 * config/arc/arcHS.md: Don't use this automaton for ARCHS4x cpus.
8131 * config/arc/arcHS4x.md: New file.
8132 * config/arc/fpu.md: Update instruction type attributes.
8133 * config/arc/t-multilib: Regenerate.
8135 2018-07-16 Tom de Vries <tdevries@suse.de>
8138 * var-tracking.c (vt_initialize): Fix pre_dec handling.
8140 2018-07-16 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
8142 * config/aarch64/atomics.md (aarch64_store_execlusive<mode>): Add
8145 2018-07-16 Eric Botcazou <ebotcazou@adacore.com>
8147 PR tree-optimization/86514
8148 * tree-ssa-reassoc.c (init_range_entry) <CASE_CONVERT>: Return for a
8149 conversion to a boolean type from a type with greater precision.
8151 2018-07-16 Tom de Vries <tdevries@suse.de>
8153 * var-tracking.c (vt_initialize): Print adjusted insn slim if
8154 dump_flags request TDF_SLIM.
8156 2018-07-16 Aldy Hernandez <aldyh@redhat.com>
8158 * fold-const.c (int_const_binop_1): Abstract...
8159 (wide_int_binop): ...wide int code here.
8160 (poly_int_binop): ...poly int code here.
8161 Abstract the rest of int_const_binop_1 into int_const_binop.
8162 * fold-const.h (wide_int_binop): New.
8163 * tree-vrp.c (vrp_int_const_binop): Call wide_int_binop.
8164 Remove useless PLUS/MINUS_EXPR case.
8165 (zero_nonzero_bits_from_vr): Move wide int code...
8166 (zero_nonzero_bits_from_bounds): ...here.
8167 (extract_range_from_binary_expr_1): Move mask optimization code...
8168 (range_easy_mask_min_max): ...here.
8169 * tree-vrp.h (zero_nonzero_bits_from_bounds): New.
8170 (range_easy_mask_min_max): New.
8172 2018-07-15 Jeff Law <law@redhat.com>
8175 * config/sh/sh.c (output_mi_thunk): Remove dead conditional
8178 2018-07-14 Jim Wilson <jimw@sifive.com>
8180 * config/riscv/linux.h (TARGET_ASM_FILE_END): New.
8182 2018-07-14 Paul Koning <ni1d@arrl.net>
8184 * config/pdp11/pdp11.c (pdp11_rtx_costs): Bugfixes.
8186 2018-07-13 Jan Hubicka <hubicka@ucw.cz>
8188 * lto-streamer-out.c (copy_function_or_variable): Dump info about
8191 2018-07-13 Bill Schmidt <wschmidt@linux.ibm.com>
8192 Steve Munroe <munroesj52@gmail.com>
8194 * config/rs6000/emmintrin.h (_mm_and_si128): New function.
8195 (_mm_andnot_si128): Likewise.
8196 (_mm_or_si128): Likewise.
8197 (_mm_xor_si128): Likewise.
8199 2018-07-13 Qing Zhao <qing.zhao@oracle.com>
8202 * builtins.c (expand_builtin_memcmp): Inline the calls first
8203 when result_eq is false.
8204 (expand_builtin_strcmp): Inline the calls first.
8205 (expand_builtin_strncmp): Likewise.
8206 (inline_string_cmp): New routine. Expand a string compare
8207 call by using a sequence of char comparison.
8208 (inline_expand_builtin_string_cmp): New routine. Inline expansion
8209 a call to str(n)cmp/memcmp.
8210 * doc/invoke.texi (--param builtin-string-cmp-inline-length):
8212 * params.def (BUILTIN_STRING_CMP_INLINE_LENGTH): New.
8214 2018-07-13 Richard Earnshaw <rearnsha@arm.com>
8216 * config/arm/driver-arm.c: Include arm-native.h.
8217 (host_detect_local_cpu): Use auto-generated data tables.
8218 (vendors, arm_cpu_table): Delete. Move part information to ...
8219 * config/arm/arm-cpus.in: ... here.
8220 * config/arm/parsecpu.awk (gen_native): New function.
8221 (vendor, part): New CPU fields.
8222 (END): Add support for building the native CPU detection tables.
8223 * config/arm/t-arm (arm-native.h): Add build rule.
8224 (driver-arm.o): Add dependency on arm-native.h.
8226 2018-07-13 Richard Biener <rguenther@suse.de>
8229 * match.pd (addr1 - addr2): Allow either of the operand to
8232 2018-07-13 Tom de Vries <tdevries@suse.de>
8234 * tree-inline.c (remap_ssa_name): Save and reuse debug exprs generated
8237 2018-07-13 Jackson Woodruff <jackson.woodruff@arm.com>
8239 * config/aarch64/aarch64.c (aarch64_operands_adjust_ok_for_ldpstp): Use
8240 arrays instead of numbered variables.
8242 2018-07-13 Eric Botcazou <ebotcazou@adacore.com>
8244 * config/sparc/sparc-protos.h (sparc_compute_frame_size): Delete.
8245 * config/sparc/sparc.c (sparc_compute_frame_size): Make static.
8247 2018-07-13 Richard Biener <rguenther@suse.de>
8250 * dwarf2out.c (gen_type_die_with_usage): Use scope_die_for
8251 instead of get_context_die.
8253 2018-07-13 Kugan Vivekanandarajah <kuganv@linaro.org>
8254 Richard Biener <rguenther@suse.de>
8257 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Check
8258 that the loop latch destination where phi is defined.
8260 2018-07-12 Kito Cheng <kito.cheng@gmail.com>
8262 * config/riscv/riscv.c (enum riscv_privilege_levels): Add UNKNOWN_MODE.
8263 (riscv_expand_epilogue): Add assertion to check interrupt mode.
8264 (riscv_set_current_function): Extract getting interrupt type to new
8266 (riscv_get_interrupt_type): New function.
8267 (riscv_merge_decl_attributes): New function, checking interrupt type is
8269 (TARGET_MERGE_DECL_ATTRIBUTES): Define.
8271 2018-07-12 Paul Koning <ni1d@arrl.net>
8273 * config/pdp11/pdp11.c (pdp11_output_def): Fix typo in .set
8276 2018-07-12 Paul Koning <ni1d@arrl.net>
8278 * doc/rtl.texi (REG_NONNEG): Remove decrement and branch until
8279 zero reference, add doloop_end instead.
8280 * doc/md.texi (decrement_and_branch_until_zero): Remove.
8281 (Looping patterns): Remove decrement_and_branch_until_zero. Add
8282 detail for doloop_end.
8284 2018-07-12 Martin Sebor <msebor@redhat.com>
8287 * attribs.c (decl_attributes): Reject conflicting attributes before
8288 calling attribute handlers.
8290 2018-07-12 Jan Hubicka <hubicka@ucw.cz>
8292 * dumpfile.c (gcc::dump_manager::get_dump_file_name): Add PART
8294 (gcc::dump_manager::get_dump_file_name): likewise.
8295 (dump_begin): Likewise.
8296 * dumpfile.h (dump_begin): Update prototype.
8297 (gcc::dump_manager::get_dump_file_name,
8298 gcc::dump_manager::get_dump_file_name): Update prototype.
8300 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
8302 * internal-fn.h (vectorizable_internal_fn_p): New function.
8303 * tree-vect-slp.c (compatible_calls_p): Likewise.
8304 (vect_build_slp_tree_1): Remove nops argument. Handle calls
8305 to internal functions.
8306 (vect_build_slp_tree_2): Update call to vect_build_slp_tree_1.
8308 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
8310 * fold-const.h (inverse_conditions_p): Declare.
8311 * fold-const.c (inverse_conditions_p): New function.
8312 * match.pd: Use inverse_conditions_p. Add folds of view_converts
8313 that test the inverse condition of a conditional internal function.
8314 * internal-fn.h (vectorized_internal_fn_supported_p): Declare.
8315 * internal-fn.c (internal_fn_mask_index): Handle conditional
8317 (vectorized_internal_fn_supported_p): New function.
8318 * tree-if-conv.c: Include internal-fn.h and fold-const.h.
8319 (any_pred_load_store): Replace with...
8320 (need_to_predicate): ...this new variable.
8321 (redundant_ssa_names): New variable.
8322 (ifcvt_can_use_mask_load_store): Move initial checks to...
8323 (ifcvt_can_predicate): ...this new function. Handle tree codes
8324 for which a conditional internal function exists.
8325 (if_convertible_gimple_assign_stmt_p): Use ifcvt_can_predicate
8326 instead of ifcvt_can_use_mask_load_store. Update after variable
8328 (predicate_load_or_store): New function, split out from
8329 predicate_mem_writes.
8330 (check_redundant_cond_expr): New function.
8331 (value_available_p): Likewise.
8332 (predicate_rhs_code): Likewise.
8333 (predicate_mem_writes): Rename to...
8334 (predicate_statements): ...this. Use predicate_load_or_store
8335 and predicate_rhs_code.
8336 (combine_blocks, tree_if_conversion): Update after above name changes.
8337 (ifcvt_local_dce): Handle redundant_ssa_names.
8338 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): Handle
8339 general conditional functions.
8340 * tree-vect-stmts.c (vectorizable_call): Likewise.
8342 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
8343 Alan Hayward <alan.hayward@arm.com>
8344 David Sherwood <david.sherwood@arm.com>
8346 * internal-fn.h (can_interpret_as_conditional_op_p): Declare.
8347 * internal-fn.c (can_interpret_as_conditional_op_p): New function.
8348 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Handle conditional
8349 plus and minus and convert them into IFN_COND_FMA-based sequences.
8350 (convert_mult_to_fma): Handle conditional plus and minus.
8352 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
8354 * doc/md.texi (cond_fma, cond_fms, cond_fnma, cond_fnms): Document.
8355 * optabs.def (cond_fma_optab, cond_fms_optab, cond_fnma_optab)
8356 (cond_fnms_optab): New optabs.
8357 * internal-fn.def (COND_FMA, COND_FMS, COND_FNMA, COND_FNMS): New
8359 (FMA): Use DEF_INTERNAL_FLT_FN rather than DEF_INTERNAL_FLT_FLOATN_FN.
8360 * internal-fn.h (get_conditional_internal_fn): Declare.
8361 (get_unconditional_internal_fn): Likewise.
8362 * internal-fn.c (cond_ternary_direct): New macro.
8363 (expand_cond_ternary_optab_fn): Likewise.
8364 (direct_cond_ternary_optab_supported_p): Likewise.
8365 (FOR_EACH_COND_FN_PAIR): Likewise.
8366 (get_conditional_internal_fn): New function.
8367 (get_unconditional_internal_fn): Likewise.
8368 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 5.
8369 (gimple_match_op::gimple_match_op): Add a new overload for 5
8371 (gimple_match_op::set_op): Likewise.
8372 (gimple_resimplify5): Declare.
8373 * genmatch.c (decision_tree::gen): Generate simplifications for
8375 * gimple-match-head.c (gimple_simplify): Define an overload for
8376 5 operands. Handle calls with 5 arguments in the top-level overload.
8377 (convert_conditional_op): Handle conversions from unconditional
8378 internal functions to conditional ones.
8379 (gimple_resimplify5): New function.
8380 (build_call_internal): Pass a fifth operand.
8381 (maybe_push_res_to_seq): Likewise.
8382 (try_conditional_simplification): Try converting conditional
8383 internal functions to unconditional internal functions.
8384 Handle 3-operand unconditional forms.
8385 * match.pd (UNCOND_TERNARY, COND_TERNARY): Operator lists.
8386 Define ternary equivalents of the current rules for binary conditional
8388 * config/aarch64/aarch64.c (aarch64_preferred_else_value): Handle
8390 * config/aarch64/iterators.md (UNSPEC_COND_FMLA, UNSPEC_COND_FMLS)
8391 (UNSPEC_COND_FNMLA, UNSPEC_COND_FNMLS): New unspecs.
8392 (optab): Handle them.
8393 (SVE_COND_FP_TERNARY): New int iterator.
8394 (sve_fmla_op, sve_fmad_op): New int attributes.
8395 * config/aarch64/aarch64-sve.md (cond_<optab><mode>)
8396 (*cond_<optab><mode>_2, *cond_<optab><mode_4)
8397 (*cond_<optab><mode>_any): New SVE_COND_FP_TERNARY patterns.
8399 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
8401 * target.def (preferred_else_value): New target hook.
8402 * doc/tm.texi.in (TARGET_PREFERRED_ELSE_VALUE): New hook.
8403 * doc/tm.texi: Regenerate.
8404 * targhooks.h (default_preferred_else_value): Declare.
8405 * targhooks.c (default_preferred_else_value): New function.
8406 * internal-fn.h (conditional_internal_fn_code): Declare.
8407 * internal-fn.c (FOR_EACH_CODE_MAPPING): New macro.
8408 (get_conditional_internal_fn): Use it.
8409 (conditional_internal_fn_code): New function.
8410 * gimple-match.h (gimple_match_cond): New struct.
8411 (gimple_match_op): Add a cond member function.
8412 (gimple_match_op::gimple_match_op): Update all forms to take a
8414 * genmatch.c (expr::gen_transform): Use the same condition as res_op
8415 for the suboperation, but don't specify a particular else_value.
8416 * tree-ssa-sccvn.c (vn_nary_simplify, vn_reference_lookup_3)
8417 (visit_nary_op, visit_reference_op_load): Pass
8418 gimple_match_cond::UNCOND to the gimple_match_op constructor.
8419 * gimple-match-head.c: Include tree-eh.h
8420 (convert_conditional_op): New function.
8421 (maybe_resimplify_conditional_op): Likewise.
8422 (gimple_resimplify1): Call maybe_resimplify_conditional_op.
8423 (gimple_resimplify2): Likewise.
8424 (gimple_resimplify3): Likewise.
8425 (gimple_resimplify4): Likewise.
8426 (maybe_push_res_to_seq): Return null for conditional operations.
8427 (try_conditional_simplification): New function.
8428 (gimple_simplify): Call it. Pass conditions to the gimple_match_op
8430 * match.pd: Fold VEC_COND_EXPRs of an IFN_COND_* call to a new
8432 * config/aarch64/aarch64.c (aarch64_preferred_else_value): New
8434 (TARGET_PREFERRED_ELSE_VALUE): Redefine.
8436 2018-07-12 Jan Hubicka <hubicka@ucw.cz>
8438 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
8440 (hash_tree): Do not hash DECL_FCONTEXT
8441 * tree-streamer-in.c (lto_input_ts_field_decl_tree_pointers):
8442 Do not stream DECL_FCONTEXT.
8443 * tree-streamer-out.c (write_ts_field_decl_tree_pointers): Likewise.
8444 * tree.c (free_lang_data_in_decl): Free DECL_FCONTEXT.
8446 2018-07-12 Richard Biener <rguenther@suse.de>
8449 * dwarf2out.c (gen_block_die): Only output blocks when they have
8450 at least one !DECL_IGNORED_P variable.
8452 2018-07-12 Richard Biener <rguenther@suse.de>
8455 * config/gnu-user.h (GNU_USER_TARGET_NO_PTHREADS_LIB_SPEC):
8456 Remove -mieee-fp handling.
8458 2018-07-12 Richard Biener <rguenther@suse.de>
8460 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Remove bogus
8461 left-over from last patch.
8463 2018-07-12 Jakub Jelinek <jakub@redhat.com>
8465 PR tree-optimization/86492
8466 * gimple-ssa-store-merging.c
8467 (imm_store_chain_info::coalesce_immediate_stores): Call
8468 check_no_overlap even for the merge_overlapping case. Formatting fix.
8470 2018-07-12 Richard Biener <rguenther@suse.de>
8473 * fold-const.c (fold_binary_op_with_conditional_arg): Do not
8474 move possibly trapping operations into the conditional.
8476 2018-07-12 Richard Biener <rguenther@suse.de>
8478 * tree-ssa-sccvn.c (mprts_hook_cnt): Remove.
8479 (vn_lookup_simplify_result): Remove recursion limit applied
8481 (vn_nary_build_or_lookup_1): Adjust.
8482 (try_to_simplify): Likewise.
8483 * gimple-match-head.c (gimple_resimplify1): Instead apply one
8485 (gimple_resimplify2): Likewise.
8486 (gimple_resimplify3): Likewise.
8487 (gimple_resimplify4): Likewise.
8489 2018-07-11 Jakub Jelinek <jakub@redhat.com>
8491 * config/i386/avx512bitalgintrin.h (_mm512_mask_bitshuffle_epi64_mask):
8492 Use __mmask64 type instead of __mmask8 for __M argument.
8493 * config/i386/avx512fintrin.h (_mm512_mask_xor_epi64,
8494 _mm512_maskz_xor_epi64): Use __mmask8 type instead of __mmask16 for
8496 (_mm512_mask_cmpneq_epi64_mask): Use __mmask8 type instead of
8497 __mmask16 for __M argument.
8498 (_mm512_maskz_insertf32x4, _mm512_maskz_inserti32x4,
8499 _mm512_mask_insertf32x4, _mm512_mask_inserti32x4): Cast last argument
8500 to __mmask16 instead of __mmask8.
8501 * config/i386/avx512vlintrin.h (_mm_mask_add_ps, _mm_maskz_add_ps,
8502 _mm256_mask_add_ps, _mm256_maskz_add_ps, _mm_mask_sub_ps,
8503 _mm_maskz_sub_ps, _mm256_mask_sub_ps, _mm256_maskz_sub_ps,
8504 _mm256_maskz_cvtepi32_ps, _mm_maskz_cvtepi32_ps): Use __mmask8 type
8505 instead of __mmask16 for __U argument.
8506 * config/i386/avx512vlbwintrin.h (_mm_mask_cmp_epi8_mask): Use
8507 __mmask16 instead of __mmask8 for __U argument.
8508 (_mm256_mask_cmp_epi8_mask): Use __mmask32 instead of __mmask16 for
8510 (_mm256_cmp_epi8_mask): Use __mmask32 return type instead of
8512 (_mm_mask_cmp_epu8_mask): Use __mmask16 instead of __mmask8 for __U
8514 (_mm256_mask_cmp_epu8_mask): Use __mmask32 instead of __mmask16 for
8516 (_mm256_cmp_epu8_mask): Use __mmask32 return type instead of
8518 (_mm_mask_cmp_epi16_mask): Cast last argument to __mmask8 instead
8520 (_mm256_mask_cvtepi8_epi16): Use __mmask16 instead of __mmask32 for
8522 (_mm_mask_cvtepi8_epi16): Use __mmask8 instead of __mmask32 for
8524 (_mm256_mask_cvtepu8_epi16): Use __mmask16 instead of __mmask32 for
8526 (_mm_mask_cvtepu8_epi16): Use __mmask8 instead of __mmask32 for
8528 (_mm256_mask_cmpneq_epu8_mask, _mm256_mask_cmplt_epu8_mask,
8529 _mm256_mask_cmpge_epu8_mask, _mm256_mask_cmple_epu8_mask): Change
8530 return type as well as __M argument type and all casts from __mmask8
8532 (_mm256_mask_cmpneq_epu16_mask, _mm256_mask_cmplt_epu16_mask,
8533 _mm256_mask_cmpge_epu16_mask, _mm256_mask_cmple_epu16_mask): Change
8534 return type as well as __M argument type and all casts from __mmask8
8536 (_mm256_mask_cmpneq_epi8_mask, _mm256_mask_cmplt_epi8_mask,
8537 _mm256_mask_cmpge_epi8_mask, _mm256_mask_cmple_epi8_mask): Change
8538 return type as well as __M argument type and all casts from __mmask8
8540 (_mm256_mask_cmpneq_epi16_mask, _mm256_mask_cmplt_epi16_mask,
8541 _mm256_mask_cmpge_epi16_mask, _mm256_mask_cmple_epi16_mask): Change
8542 return type as well as __M argument type and all casts from __mmask8
8544 * config/i386/avx512vbmi2vlintrin.h (_mm_mask_shrdi_epi32,
8545 _mm_mask_shldi_epi32): Cast last argument to __mmask8 instead of
8548 2018-07-11 Grazvydas Ignotas <notasas@gmail.com>
8550 * config/i386/avx512bwintrin.h: (_mm512_mask_cmp_epi8_mask,
8551 _mm512_mask_cmp_epu8_mask): Use __mmask64 type instead of __mmask32
8554 2018-07-11 Paul Koning <ni1d@arrl.net>
8556 * doc/md.texi (define_subst): Document how multiple occurrences of
8557 the same argument in the replacement pattern are handled.
8559 2018-07-11 Paul Koning <ni1d@arrl.net>
8561 * doc/extend.texi (Common Variable Attributes): Move "mode" into
8563 (Common Type Attributes): Add "mode" attribute.
8565 2018-07-11 Jan Hubicka <hubicka@ucw.cz>
8567 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not
8568 stream DECL_ORIGINAL_TYPE.
8569 (DFS::DFS_write_tree_body): Drop hack handling local external decls.
8570 (hash_tree): Do not walk DECL_ORIGINAL_TYPE.
8571 * tree-streamer-in.c (lto_input_ts_decl_non_common_tree_pointers):
8572 Do not walk original type.
8573 * tree-streamer-out.c (streamer_write_chain): Drop hack handling
8575 (write_ts_decl_non_common_tree_pointers): Do not stream
8577 * tree.c (free_lang_data_in_decl): Clear DECL_ORIGINAL_TYPE.
8578 (find_decls_types_r): Do not walk DEC_ORIGINAL_TYPE.
8580 2018-07-11 Aldy Hernandez <aldyh@redhat.com>
8582 * tree-ssa-threadupdate.c (thread_through_all_blocks): Do not jump
8583 thread twice from the same starting edge.
8585 2018-07-11 Aldy Hernandez <aldyh@redhat.com>
8587 * vr-values.c (gimple_stmt_nonzero_p): Abstract common code to...
8588 * gimple.c (gimple_call_nonnull_result_p): ...here...
8589 (gimple_call_nonnull_arg): ...and here.
8590 * gimple.h (gimple_call_nonnull_result_p): New.
8591 (gimple_call_nonnull_arg): New.
8593 2018-07-11 Richard Earnshaw <rearnsha@arm.com>
8595 * config/arm/arm-cpus.in: Move information from fpu field of each
8596 cpu definition to the isa field.
8597 * config/arm/parsecpu.awk (fpu): Delete match rule.
8598 (gen_comm_data): Don't add bits from the CPU's FPU entry.
8600 2018-07-11 Richard Biener <rguenther@suse.de>
8603 * dwarf2out.c (init_sections_and_labels): Use
8604 output_asm_line_debug_info consistently.
8605 (dwarf2out_early_finish): Likewise.
8606 (dwarf2out_finish): Remove DW_AT_stmt_list from early generated
8609 2018-07-11 Richard Biener <rguenther@suse.de>
8611 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1):
8612 Rework father_bb setting in a way to avoid propagating constants
8613 multiple times on a loop body.
8615 2018-07-10 Mark Wielaard <mark@klomp.org>
8618 * dwarf2out.c (output_macinfo_op): Fix dwarf_FORM typo in gcc_assert.
8620 2018-07-10 Richard Biener <rguenther@suse.de>
8622 * hash-map.h (hash_map::iterator::operator*): Return
8623 references to key and value.
8625 2018-07-10 Jakub Jelinek <jakub@redhat.com>
8628 * gimplify.c (find_combined_omp_for): Add DATA argument, in addition
8629 to finding the inner OMP_FOR/OMP_SIMD stmt find non-trivial wrappers,
8630 BLOCKs with BLOCK_VARs, OMP_PARALLEL in between, OMP_FOR in between.
8631 (gimplify_omp_for): For composite loops, move outer
8632 OMP_{DISTRIBUTE,TASKLOOP,FOR,PARALLEL} right around innermost
8633 OMP_FOR/OMP_SIMD if there are any non-trivial wrappers. For class
8634 iterators add any needed clauses. Allow OMP_FOR_ORIG_DECLS to contain
8635 TREE_LIST for both the original class iterator and the "last" helper
8636 var. Gimplify OMP_FOR_PRE_BODY before the outermost composite
8637 loop, remember has_decl_expr from outer composite loops for the
8638 innermost OMP_SIMD in TREE_PRIVATE bit on OMP_FOR_INIT.
8640 2018-07-09 Martin Sebor <msebor@redhat.com>
8644 * builtins.c (c_strlen): Avoid out-of-bounds warnings when
8645 accessing implicitly initialized array elements.
8646 * expr.c (string_constant): Handle string initializers of
8647 character arrays within aggregates.
8648 * gimple-fold.c (fold_array_ctor_reference): Add argument.
8649 Store element offset. As a special case, handle zero size.
8650 (fold_nonarray_ctor_reference): Same.
8651 (fold_ctor_reference): Add argument. Store subobject offset.
8652 * gimple-fold.h (fold_ctor_reference): Add argument.
8654 2018-07-09 Paul Koning <ni1d@arrl.net>
8656 * config/pdp11/pdp11.c (pdp11_addr_cost): New function.
8657 (pdp11_insn_cost): New function.
8658 (pdp11_md_asm_adjust): New function.
8659 (TARGET_INVALID_WITHIN_DOLOOP): Define.
8660 (pdp11_rtx_costs): Update to match machine better.
8661 (output_addr_const_pdp11): Correct format mismatch warnings.
8662 * config/pdp11/pdp11.h (SLOW_BYTE_ACCESS): Correct definition.
8663 * config/pdp11/pdp11.md: General change to add base_cost and/or
8664 length attributes for use by new pdp11_insn_cost function.
8665 (MIN_BRANCH): Correct definition.
8667 (doloop_end): Use standard pattern name for looping pattern.
8668 (doloop_end_nocc): New.
8669 (movsf): Add another constraint alternative.
8670 (zero_extendqihi2): Add constraint alternatives for not in place
8672 (zero_extendhisi2): Remove.
8673 (shift patterns): Add CC handling variants.
8677 (define_peephole2): New peephole to recognize mov that sets CC for
8680 2018-07-09 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8682 * config/sparc/sparc.c (sparc_fold_builtin) <SPARC_BUILTIN_PDIST,
8683 SPARC_BUILTIN_PDISTN>: Adapt for signature change of wi::neg,
8686 2018-07-09 Jakub Jelinek <jakub@redhat.com>
8689 * real.c (real_nextafter): Return true if result is denormal.
8691 2018-07-09 Martin Liska <mliska@suse.cz>
8693 * common.opt: Add back wrongly removed attribute.
8695 2018-07-09 Richard Biener <rguenther@suse.de>
8698 * dwarf2out.c (gen_block_die): For an early generated DIE
8699 always output high/low PC attributes.
8701 2018-07-09 Tom de Vries <tdevries@suse.de>
8703 * cfgexpand.c (expand_debug_source_expr): Handle VAR_DECL.
8704 * tree-inline.c (remap_ssa_name): Handle default def ssa_name mapping
8705 onto VAR_DECL with abstract origin.
8707 2018-07-07 Jim Wilson <jimw@sifive.com>
8709 * config/riscv/riscv.c (TARGET_CUSTOM_FUNCTION_DESCRIPTORS): New.
8711 2018-07-07 Tom de Vries <tdevries@suse.de>
8713 * tree-dfa.c (dump_enumerated_decls): Handle cfun->cfg == NULL.
8715 2018-07-07 Aldy Hernandez <aldyh@redhat.com>
8717 * tree-vrp.c (vrp_int_const_binop): Change overflow type to
8719 (combine_bound): Use wide-int overflow calculation instead of
8721 * calls.c (maybe_warn_alloc_args_overflow): Change overflow type to
8723 * fold-const.c (int_const_binop_2): Same.
8724 (extract_muldiv_1): Same.
8725 (fold_div_compare): Same.
8726 (fold_abs_const): Same.
8728 * poly-int.h (add): Same.
8732 * predict.c (predict_iv_comparison): Same.
8733 * profile-count.c (slow_safe_scale_64bit): Same.
8734 * simplify-rtx.c (simplify_const_binary_operation): Same.
8735 * tree-chrec.c (tree_fold_binomial): Same.
8736 * tree-data-ref.c (split_constant_offset_1): Same.
8737 * tree-if-conv.c (idx_within_array_bound): Same.
8738 * tree-scalar-evolution.c (iv_can_overflow_p): Same.
8739 * tree-ssa-phiopt.c (minmax_replacement): Same.
8740 * tree-vect-loop.c (is_nonwrapping_integer_induction): Same.
8741 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Same.
8742 * vr-values.c (vr_values::adjust_range_with_scev): Same.
8743 * wide-int.cc (wi::add_large): Same.
8744 (wi::mul_internal): Same.
8745 (wi::sub_large): Same.
8746 (wi::divmod_internal): Same.
8747 * wide-int.h: Change overflow type to overflow_type for neg, add,
8748 mul, smul, umul, div_trunc, div_floor, div_ceil, div_round,
8749 mod_trunc, mod_ceil, mod_round, add_large, sub_large,
8750 mul_internal, divmod_internal.
8751 (overflow_type): New enum.
8752 (accumulate_overflow): New.
8754 2018-07-06 Kugan Vivekanandarajah <kugan.vivekanandarajah@linaro.org>
8756 * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): New.
8757 (tree_ssa_phiopt_worker): Call cond_removal_in_popcount_pattern.
8759 2018-07-06 Kugan Vivekanandarajah <kuganv@linaro.org>
8761 * tree-ssa-loop-niter.c (number_of_iterations_popcount): If popcount
8762 argument is checked for zero before entering loop, avoid checking again.
8764 2018-07-06 Kugan Vivekanandarajah <kuganv@linaro.org>
8766 * gimplify.h (generic_expr_could_trap_p): Set as global function.
8767 * gimplify.h (generic_expr_could_trap_p): Likwise.
8768 * tree-scalar-evolution.c (expression_expensive_p): Handle COND_EXPR.
8770 2018-07-06 Jakub Jelinek <jakub@redhat.com>
8772 PR tree-optimization/86401
8773 * fold-const.c (fold_binary_loc) <case BIT_AND_EXPR>: Move the
8774 ((A & N) + B) & M -> (A + B) & M etc. optimization into ...
8775 (fold_bit_and_mask): ... here. New helper function for match.pd.
8776 * fold-const.h (fold_bit_and_mask): Declare.
8777 * match.pd (((A & N) + B) & M -> (A + B) & M): New optimization.
8779 2018-07-06 Peter Bergner <bergner@linux.ibm.com>
8782 * target.def (translate_mode_attribute): New hook.
8783 * targhooks.h (default_translate_mode_attribute): Declare.
8784 * targhooks.c (default_translate_mode_attribute): New function.
8785 * doc/tm.texi.in (TARGET_TRANSLATE_MODE_ATTRIBUTE): New hook.
8786 * doc/tm.texi: Regenerate.
8787 * config/rs6000/rs6000.c (TARGET_TRANSLATE_MODE_ATTRIBUTE): Define.
8788 (rs6000_translate_mode_attribute): New function.
8790 2018-07-06 Paul Koning <ni1d@arrl.net>
8792 * doc/md.texi (define_split): Document DONE and FAIL.
8793 (define_peephole2): Ditto.
8795 2018-07-05 Jeff Law <law@redhat.com>
8797 PR tree-optimization/86010
8798 * tree-ssa-dse.c (compute_trims): More aggressively trim at
8799 both the head and tail of mem* and str* calls.
8801 2018-07-05 Jim Wilson <jimw@sifive.com>
8803 * config.gcc (riscv*-*-*): When setting xlen, handle riscv-*.
8805 2018-07-05 Indu Bhagat <indu.bhagat@oracle.com>
8807 * config/aarch64/aarch64-simd.md: correct flags text for
8808 MIN_EXPR replacement.
8810 2018-07-05 James Clarke <jrtc27@jrtc27.com>
8812 * configure: Regenerated.
8814 2018-07-05 Carl Love <cel@us.ibm.com>
8816 * config/rs6000/rs6000-c.c: Map ALTIVEC_BUILTIN_VEC_UNPACKH for
8817 float argument to VSX_BUILTIN_DOUBLEH_V4SF.
8818 Map ALTIVEC_BUILTIN_VEC_UNPACKL for float argument to
8819 VSX_BUILTIN_DOUBLEL_V4SF.
8821 2018-07-05 Martin Sebor <msebor@redhat.com>
8824 * tree-ssa-strlen.c (maybe_set_strlen_range): Use type size rather
8825 than its domain to compute its the upper bound of a char array.
8827 2018-07-05 Nathan Sidwell <nathan@acm.org>
8829 Replace NO_IMPLICIT_EXTERN_C with SYSTEM_IMPLICIT_EXTERN_C.
8830 * doc/cpp.texi: Update comment.
8831 * doc/tm.texi: Rebuilt.
8832 * doc/tm.texi.in (NO_IMPLICIT_EXTERN_C): Replace with ...
8833 (SYSTEM_IMPLICIT_EXTERN_C): ... this, opposite sense.
8834 * doc/extend.texi (Backwards Compatibility): Clarify it is system
8835 headers affected by extern "C".
8836 * system.h: Poison NO_IMPLICIT_EXTERN_C.
8837 * config/alpha/alpha.h, config/arm/uclinux-elf.h,
8838 config/bfin/elf.h, config/cris/cris.h, config/darwin.h,
8839 config/dragonfly.h, config/freebsd.h, config/gnu-user.h,
8840 config/i386/cygming.h, config/i386/djgpp.h, config/i386/nto.h,
8841 config/ia64/hpux.h, config/lm32/lm32.h, config/lm32/uclinux-elf.h,
8842 config/lynx.h, config/mips/elf.h, config/mmix/mmix.h,
8843 config/netbsd.h, config/pa/pa-hpux.h, config/powerpcspe/sysv4.h,
8844 config/riscv/elf.h, config/rs6000/sysv4.h, config/rtems.h,
8845 config/s390/tpf.h, config/sh/newlib.h, config/sol2.h,
8846 config/sparc/openbsd64.h, config/sparc/sp-elf.h,
8847 config/sparc/sp64-elf.h, config/spu/spu.h,
8848 config/stormy16/stormy16.h, config/v850/v850.h,
8849 config/visium/visium.h, config/vx-common.h, config/xtensa/elf.h: Don't
8850 define NO_IMPLICIT_EXTERN_C.
8851 * config/rs6000/aix.h: Set SYSTEM_IMPLICIT_EXTERN_C.
8853 2018-07-05 Tamar Christina <tamar.christina@arm.com>
8856 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
8857 instead of GET_MODE_SIZE when comparing Units.
8859 2018-07-05 Tamar Christina <tamar.christina@arm.com>
8862 * rtlanal.c (set_noop_p): Constrain on mode change,
8863 include hard-reg-set.h
8865 2018-07-05 Tamar Christina <tamar.christina@arm.com>
8867 * config/aarch64/aarch64.c (aarch64_expand_movmem): Fix mode size.
8869 2018-07-05 Jakub Jelinek <jakub@redhat.com>
8872 2018-07-04 Maxim Ostapenko <m.ostapenko@samsung.com>
8875 * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
8877 * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
8879 2018-07-04 Maxim Ostapenko <m.ostapenko@samsung.com>
8882 * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
8884 * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
8886 2018-07-04 Eric Botcazou <ebotcazou@adacore.com>
8889 * expmed.c (choose_multiplier): Fix incorrect comparison with mask.
8891 2018-07-04 Aldy Hernandez <aldyh@redhat.com>
8893 * tree-vrp.c (extract_range_from_binary_expr_1): Initialize
8894 neg_*_op* variables.
8896 2018-07-04 Martin Liska <mliska@suse.cz>
8898 * tree-switch-conversion.c: Define
8899 max_ratio_for_speed and max_ratio_for_size constants.
8901 2018-07-04 Denys Vlasenko <dvlasenk@redhat.com>
8902 Martin Liska <mliska@suse.cz>
8907 * common.opt: Rename align options with 'str_' prefix.
8908 * common/config/i386/i386-common.c (set_malign_value): New
8910 (ix86_handle_option): Use it to set -falign-* options/
8911 * config/aarch64/aarch64-protos.h (struct tune_params): Change
8912 type from int to string.
8913 * config/aarch64/aarch64.c: Update default values from int
8915 * config/alpha/alpha.c (alpha_override_options_after_change):
8917 * config/arm/arm.c (arm_override_options_after_change_1): Likewise.
8918 * config/i386/dragonfly.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
8919 max skip conditionally.
8920 * config/i386/freebsd.h (SUBALIGN_LOG): New.
8921 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
8922 max skip conditionally.
8923 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
8924 max skip conditionally.
8925 * config/i386/gnu-user.h (SUBALIGN_LOG): New.
8926 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
8927 max skip conditionally.
8928 * config/i386/i386.c (struct ptt): Change type from int to
8930 (ix86_default_align): Set default values.
8931 * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Print
8932 max skip conditionally.
8933 * config/i386/iamcu.h (SUBALIGN_LOG): New.
8934 (ASM_OUTPUT_MAX_SKIP_ALIGN):
8935 * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN):
8936 * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
8937 max skip conditionally.
8938 * config/i386/openbsdelf.h (SUBALIGN_LOG): New.
8939 (ASM_OUTPUT_MAX_SKIP_ALIGN) Print max skip conditionally.:
8940 * config/i386/x86-64.h (SUBALIGN_LOG): New.
8941 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
8942 max skip conditionally.
8943 (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
8944 * config/ia64/ia64.c (ia64_option_override): Set default values
8945 for alignment options.
8946 * config/m68k/m68k.c: Handle new str_align_* options.
8947 * config/mips/mips.c (mips_set_compression_mode): Change
8949 (mips_option_override): Set default values for options.
8950 * config/powerpcspe/powerpcspe.c (rs6000_option_override_internal):
8952 * config/rs6000/rs6000.c (rs6000_option_override_internal):
8954 * config/rx/rx.c (rx_option_override): Likewise.
8955 * config/rx/rx.h (JUMP_ALIGN): Use align_jumps_log.
8956 (LABEL_ALIGN): Use align_labels_log.
8957 (LOOP_ALIGN): Use align_loops_align.
8958 * config/s390/s390.c (s390_asm_output_function_label): Use new
8960 * config/sh/sh.c (sh_override_options_after_change):
8961 Change type of constants.
8962 * config/spu/spu.c (spu_sched_init): Likewise.
8963 * config/sparc/sparc.c (sparc_option_override): Set default
8965 * config/visium/visium.c (visium_option_override): Likewise.
8966 * config/visium/visium.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Do not
8967 emit p2align format with last argument if it's not needed.
8968 * doc/invoke.texi: Document extended format of -falign-*.
8969 * final.c: Use align_labels alignment.
8970 * flags.h (struct target_flag_state): Change type to use
8972 (struct align_flags_tuple): New.
8973 (struct align_flags): Likewise.
8974 (align_loops_log): Redefine macro to use new types.
8975 (align_loops_max_skip): Redefine macro to use new types.
8976 (align_jumps_log): Redefine macro to use new types.
8977 (align_jumps_max_skip): Redefine macro to use new types.
8978 (align_labels_log): Redefine macro to use new types.
8979 (align_labels_max_skip): Redefine macro to use new types.
8980 (align_functions_log): Redefine macro to use new types.
8981 (align_loops): Redefine macro to use new types.
8982 (align_jumps): Redefine macro to use new types.
8983 (align_labels): Redefine macro to use new types.
8984 (align_functions): Redefine macro to use new types.
8985 (align_functions_max_skip): Redefine macro to use new types.
8986 (align_loops_value): New macro.
8987 (align_jumps_value): New macro.
8988 (align_labels_value): New macro.
8989 (align_functions_value): New macro.
8990 * function.c (invoke_set_current_function_hook): Propagate
8991 alignment values from flags to global variables default in
8993 * ipa-icf.c (sem_function::equals_wpa): Use
8994 cl_optimization_option_eq instead of memcmp.
8995 * lto-streamer.h (cl_optimization_stream_out): Support streaming
8997 (cl_optimization_stream_in): Likewise.
8998 * optc-save-gen.awk: Support strings in cl_optimization.
8999 * opth-gen.awk: Likewise.
9000 * opts.c (finish_options): Remove error checking of invalid
9002 (MAX_CODE_ALIGN): Remove.
9003 (MAX_CODE_ALIGN_VALUE): Likewise.
9004 (parse_and_check_align_values): New function.
9005 (check_alignment_argument): Likewise.
9006 (common_handle_option): Use check_alignment_argument.
9007 * opts.h (parse_and_check_align_values): Declare.
9008 * toplev.c (init_alignments): Remove.
9009 (read_log_maxskip): New.
9010 (parse_N_M): Likewise.
9011 (parse_alignment_opts): Likewise.
9012 (backend_init_target): Remove usage of init_alignments.
9013 * toplev.h (parse_alignment_opts): Declare.
9014 * tree-streamer-in.c (streamer_read_tree_bitfields): Add new
9016 * tree-streamer-out.c (streamer_write_tree_bitfields): Likewise.
9017 * tree.c (cl_option_hasher::equal): New.
9018 * varasm.c: Use new global macros.
9020 2018-07-04 Denys Vlasenko <dvlasenk@redhat.com>
9022 * config/i386/dragonfly.h: (ASM_OUTPUT_MAX_SKIP_ALIGN):
9023 Use a simpler align directive also if MAXSKIP = ALIGN-1.
9024 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
9025 * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
9026 * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
9027 * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
9028 * config/i386/freebsd.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Remove "If N
9029 is large, do at least 8 byte alignment" code. Add SUBALIGN_LOG
9030 define. Use a simpler align directive also if MAXSKIP = ALIGN-1.
9031 * config/i386/gnu-user.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
9032 * config/i386/iamcu.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
9033 * config/i386/openbsdelf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
9034 * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
9036 2018-07-04 Martin Liska <mliska@suse.cz>
9037 Jonathan Wakely <jwakely@redhat.com>
9039 * coverage.c: Use correct type.
9040 * doc/invoke.texi: Language correction.
9042 2018-07-03 H.J. Lu <hongjiu.lu@intel.com>
9045 * config/i386/i386.c (rest_of_insert_endbranch): Also generate
9046 ENDBRANCH for non-tail call which may return via indirect branch.
9047 * doc/extend.texi: Document indirect_return attribute.
9049 2018-07-03 Martin Sebor <msebor@redhat.com>
9051 PR tree-optimization/86274
9052 * gimple-ssa-sprintf.c (fmtresult::type_max_digits): Verify
9054 (format_floating): Correct handling of infinities and NaNs.
9056 2018-07-03 Martin Sebor <msebor@redhat.com>
9058 * print-tree.c (print_real_cst): New function.
9059 (print_node_brief): Call it.
9060 (print_node): Ditto.
9062 2018-07-03 Jeff Law <law@redhat.com>
9064 * config/h8300/h8300.md (logical<mode>3_sn, logical<mode>3): Merge
9065 into a single pattern.
9067 * config/h8300/h8300.md (ors code_iterator): New.
9068 (bsetqi_msx, bnotqi_msx patterns and splitters): Consolidate into
9069 a single pattern and single splitter.
9070 (bsethi_msx, bnothi_msx patterns): Consolidate into a single pattern.
9071 (iorqi3_1, xorqi3_1): Likewise.
9072 (iorqi3, xorqi3 expanders): Similarly.
9074 * config/h8300/h8300.md (movmd_internal_normal): Consolidated with
9075 (movmd_internal) into a single pattern using the P mode iterator.
9076 (movmd splitters): Similarly.
9077 (stpcpy_internal_normal, stpcpy_internal): Similarly for thes patterns.
9078 (movsd splitters): Similarly.
9080 * config/h8300/h8300.c (h8300_insn_length_from_table): Consolidate
9081 ADDB, ADDW and ADDL into a single ADD attribute which selects the
9082 right table based on the size of the operand.
9083 * config/h8300/h8300.md (length_table): Corresponding changes. All
9084 references to "addb", "addw" and "addl" changed to "add".
9085 (btst patterns): Merge two variants into a single pattern.
9086 (tstqi, tsthi): Likewise.
9087 (addhi3_incdec, addsi3_incdec): Likewise.
9088 (subhi3_h8300hs, subsi3_h8300hs): Likewise.
9089 (mulhi3, mulsi3): Likewise.
9090 (udivhi3, udivsi3): Likewise.
9091 (divhi3, divsi3): Likewise.
9092 (andorqi3, andorhi3, andorsi3): Likewise.
9094 2018-07-03 Uros Bizjak <ubizjak@gmail.com>
9097 * config/i386/sse.md (uavg<mode>3_ceil): New expander.
9098 (<sse2_avx2>_uavg<mode>3<mask_name>): Simplify expander.
9100 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
9102 PR tree-optimization/85694
9103 * config/aarch64/iterators.md (HADD, RHADD): New int iterators.
9104 (u): Handle UNSPEC_SHADD, UNSPEC_UHADD, UNSPEC_SRHADD and
9106 * config/aarch64/aarch64-simd.md (<u>avg<mode>3_floor)
9107 (<u>avg<mode>3_ceil): New patterns.
9109 2018-07-03 David Malcolm <dmalcolm@redhat.com>
9111 * gcc.dg/vect/slp-perm-1.c: Remove "note: " prefix from
9112 scan-tree-dump directive.
9113 * gcc.dg/vect/slp-perm-2.c: Likewise.
9114 * gcc.dg/vect/slp-perm-3.c: Likewise.
9115 * gcc.dg/vect/slp-perm-5.c: Likewise.
9116 * gcc.dg/vect/slp-perm-6.c: Likewise.
9117 * gcc.dg/vect/slp-perm-7.c: Likewise.
9118 * gcc.dg/vect/slp-perm-8.c: Likewise.
9120 2018-07-03 Marek Polacek <polacek@redhat.com>
9123 * gimple-fold.c (size_must_be_zero_p): Check the type of the size.
9125 2018-07-03 Richard Biener <rguenther@suse.de>
9128 * tree-ssa-structalias.c (find_func_clobbers): Properly
9129 handle indirect calls.
9131 2018-07-03 Jeff Law <law@redhat.com>
9133 * config/h8300/h8300.md (HSI, QHSI, QHSIF): New mode iterators.
9134 (shifts): New code iterator.
9135 (movqi, movhi, movsi, movsf expanders): Consolidate into a single
9136 expander. Fix HImode handling on H8/SX.
9137 (addqi3, addhi3, addsi3 expanders): Consolidate into a single expander.
9138 (subqi3, subhi3, subsi3 expanders): Likewise.
9139 (andqi3, andhi3, andsi3 expanders): Likewise.
9140 (iorqi3, iorhi3, iorsi3 expanders): Likewise.
9141 (xorqi3, xorhi3, xorsi3 expanders): Likewise.
9142 (negqi2, neghi2, negsi2, negsf2 expanders): Likewise.
9143 (one_cmplqi2, one_cmplhi2, one_cmplsi2): Likewise.
9144 (zero_extendqihi2, zero_extendqisi2): Likewise.
9145 (extendqihi2, extendqisi2): Likewise.
9146 (rotlqi3, rotlhi3, rotlsi3): Likewise.
9147 (neghi2_h8300, negsi2_h8300): Likewise for these patterns.
9148 (rotlqi3_1, rotlhi3_1): Likewise.
9149 (logicalhi3_sn, logicalsi3_sn): Likewise.
9150 (logicalhi3, logicalsi3): Likewise.
9152 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
9154 * tree-vect-patterns.c (vect_recog_rotate_pattern)
9155 (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
9156 (vect_recog_mixed_size_cond_pattern, adjust_bool_pattern_cast)
9157 (adjust_bool_pattern, vect_recog_bool_pattern): Pass the vector
9158 type to append_pattern_def_seq instead of creating a stmt_vec_info
9160 (build_mask_conversion): Likewise. Remove vinfo argument.
9161 (vect_add_conversion_to_patterm): Likewise, renaming to...
9162 (vect_add_conversion_to_pattern): ...this.
9163 (vect_recog_mask_conversion_pattern): Update call to
9164 build_mask_conversion. Pass the vector type to
9165 append_pattern_def_seq here too.
9166 (vect_recog_gather_scatter_pattern): Update call to
9167 vect_add_conversion_to_pattern.
9169 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
9171 * tree-vect-patterns.c (new_pattern_def_seq): Delete.
9172 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
9173 (vect_recog_widen_op_pattern, vect_recog_over_widening_pattern)
9174 (vect_recog_rotate_pattern, vect_synth_mult_by_constant): Don't set
9175 STMT_VINFO_PATTERN_DEF_SEQ to null here.
9176 (vect_recog_pow_pattern, vect_recog_vector_vector_shift_pattern)
9177 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern): Use
9178 append_pattern_def_seq instead of new_pattern_def_seq.
9179 (vect_recog_divmod_pattern): Do both of the above.
9180 (vect_pattern_recog_1): Assert that STMT_VINO_PATTERN_DEF_SEQ
9183 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
9185 * tree-vect-patterns.c (vect_recog_dot_prod_pattern):
9186 (vect_recog_sad_pattern, vect_recog_widen_op_pattern)
9187 (vect_recog_widen_mult_pattern, vect_recog_pow_pattern):
9188 (vect_recog_widen_sum_pattern, vect_recog_over_widening_pattern)
9189 (vect_recog_average_pattern, vect_recog_cast_forwprop_pattern)
9190 (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
9191 (vect_recog_vector_vector_shift_pattern, vect_synth_mult_by_constant)
9192 (vect_recog_mult_pattern, vect_recog_divmod_pattern)
9193 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
9194 (vect_recog_mask_conversion_pattern): Replace vec<gimple *>
9195 parameter with a single stmt_vec_info.
9196 (vect_recog_func_ptr): Likewise.
9197 (vect_recog_gather_scatter_pattern): Likewise, folding in...
9198 (vect_try_gather_scatter_pattern): ...this.
9199 (vect_pattern_recog_1): Remove stmts_to_replace and just pass
9200 the stmt_vec_info of the statement to be matched. Don't clear
9201 STMT_VINFO_RELATED_STMT.
9202 (vect_pattern_recog): Update call accordingly.
9204 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
9206 PR tree-optimization/85694
9207 * doc/md.texi (avgM3_floor, uavgM3_floor, avgM3_ceil)
9208 (uavgM3_ceil): Document new optabs.
9209 * doc/sourcebuild.texi (vect_avg_qi): Document new target selector.
9210 * internal-fn.def (IFN_AVG_FLOOR, IFN_AVG_CEIL): New internal
9212 * optabs.def (savg_floor_optab, uavg_floor_optab, savg_ceil_optab)
9213 (savg_ceil_optab): New optabs.
9214 * tree-vect-patterns.c (vect_recog_average_pattern): New function.
9215 (vect_vect_recog_func_ptrs): Add it.
9216 * tree-vect-stmts.c (vectorizable_call): Get the type of the zero
9217 constant directly from the associated lhs.
9219 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
9221 * tree-vect-patterns.c (vect_split_statement): New function.
9222 (vect_convert_input): Use it to try to split an existing cast.
9224 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
9226 * poly-int.h (print_hex): New function.
9227 * dumpfile.h (dump_dec, dump_hex): Declare.
9228 * dumpfile.c (dump_dec, dump_hex): New poly_wide_int functions.
9229 * tree-vectorizer.h (_stmt_vec_info): Add min_output_precision,
9230 min_input_precision, operation_precision and operation_sign.
9231 * tree-vect-patterns.c (vect_get_range_info): New function.
9232 (vect_same_loop_or_bb_p, vect_single_imm_use)
9233 (vect_operation_fits_smaller_type): Delete.
9234 (vect_look_through_possible_promotion): Add an optional
9235 single_use_p parameter.
9236 (vect_recog_over_widening_pattern): Rewrite to use new
9237 stmt_vec_info infomration. Handle one operation at a time.
9238 (vect_recog_cast_forwprop_pattern, vect_narrowable_type_p)
9239 (vect_truncatable_operation_p, vect_set_operation_type)
9240 (vect_set_min_input_precision): New functions.
9241 (vect_determine_min_output_precision_1): Likewise.
9242 (vect_determine_min_output_precision): Likewise.
9243 (vect_determine_precisions_from_range): Likewise.
9244 (vect_determine_precisions_from_users): Likewise.
9245 (vect_determine_stmt_precisions, vect_determine_precisions): Likewise.
9246 (vect_vect_recog_func_ptrs): Put over_widening first.
9248 (vect_pattern_recog): Call vect_determine_precisions.
9250 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
9252 * tree-vect-patterns.c (vect_mark_pattern_stmts): Remove pattern
9253 statements that have been replaced by further pattern statements.
9254 (vect_pattern_recog_1): Clear STMT_VINFO_PATTERN_DEF_SEQ on failure.
9256 2018-07-03 Richard Biener <rguenther@suse.de>
9258 * tree-vect-stmts.c (vect_is_simple_use): Consolidate dumping,
9259 always set *dt. Dump vectype in vectype overload.
9260 * dumpfile.h (dump_gimple_expr): New function.
9261 (dump_gimple_expr_loc): Likewise.
9262 * dumpfile.c (dump_gimple_expr): New function.
9263 (dump_gimple_expr_loc): Likewise.
9265 2018-07-02 Jeff Law <law@redhat.com>
9267 * config/h8300/h8300.md (movqi_h8300, movqi_h8300hs): Consolidate
9268 the H8/300, H8/300H and H8/S variants into a single pattern.
9269 (movhi_h8300, movqi_h8300hs): Similarly.
9270 (pushqi_h8300hs, pushhi_h8300hs): Consolidate into a single pattern.
9271 (QHI mode iterator): New.
9273 * config/h8300/h8300.md: Remove trailing whitespace.
9275 2018-07-02 Jim Wilson <jimw@sifive.com>
9277 * config/riscv/riscv.c (riscv_expand_epilogue): Use emit_jump_insn
9278 instead of emit_insn for interrupt returns.
9279 * config/riscv/riscv.md (riscv_met): Add (return) to rtl.
9280 (riscv_sret, riscv_uret): Likewise.
9282 2018-07-02 David Malcolm <dmalcolm@redhat.com>
9284 * pretty-print.c (selftest::test_pp_format): Move save and restore
9285 of quotes to class auto_fix_quotes, and add an instance.
9286 * selftest.c: Include "intl.h".
9287 (selftest::auto_fix_quotes::auto_fix_quotes): New ctor.
9288 (selftest::auto_fix_quotes::~auto_fix_quotes): New dtor.
9289 * selftest.h (selftest::auto_fix_quotes): New class.
9291 2018-07-02 Richard Henderson <richard.henderson@linaro.org>
9293 * config/aarch64/aarch64-protos.h, config/aarch64/aarch64.c
9294 (aarch64_sve_prepare_conditional_op): Remove.
9295 * config/aarch64/aarch64-sve.md (cond_<SVE_INT_BINARY><SVE_I>):
9296 Allow aarch64_simd_reg_or_zero as select operand; remove
9297 the aarch64_sve_prepare_conditional_op call.
9298 (cond_<SVE_INT_BINARY_SD><SVE_SDI>): Likewise.
9299 (cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
9300 (*cond_<SVE_INT_BINARY><SVE_I>_z): New pattern.
9301 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_z): New pattern.
9302 (*cond_<SVE_COND_FP_BINARY><SVE_F>_z): New pattern.
9303 (*cond_<SVE_INT_BINARY><SVE_I>_any): New pattern.
9304 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_any): New pattern.
9305 (*cond_<SVE_COND_FP_BINARY><SVE_F>_any): New pattern
9306 and a splitters to match all of the *_any patterns.
9307 * config/aarch64/predicates.md (aarch64_sve_any_binary_operator): New.
9309 * config/aarch64/iterators.md (SVE_INT_BINARY_REV): Remove.
9310 (SVE_COND_FP_BINARY_REV): Remove.
9311 (sve_int_op_rev, sve_fp_op_rev): New.
9312 * config/aarch64/aarch64-sve.md (*cond_<SVE_INT_BINARY><SVE_I>_0): New.
9313 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_0): New.
9314 (*cond_<SVE_COND_FP_BINARY><SVE_F>_0): New.
9315 (*cond_<SVE_INT_BINARY><SVE_I>_2): Rename, add movprfx alternative.
9316 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_2): Similarly.
9317 (*cond_<SVE_COND_FP_BINARY><SVE_F>_2): Similarly.
9318 (*cond_<SVE_INT_BINARY><SVE_I>_3): Similarly; use sve_int_op_rev.
9319 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_3): Similarly.
9320 (*cond_<SVE_COND_FP_BINARY><SVE_F>_3): Similarly; use sve_fp_op_rev.
9322 * config/aarch64/aarch64-sve.md (cond_<SVE_COND_FP_BINARY><SVE_F>):
9323 Remove match_dup 1 from the inner unspec.
9324 (*cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
9326 * config/aarch64/aarch64.md (movprfx): New attr.
9327 (length): Default movprfx to 8.
9328 * config/aarch64/aarch64-sve.md (*mul<SVE_I>3): Add movprfx alt.
9329 (*madd<SVE_I>, *msub<SVE_I): Likewise.
9330 (*<su>mul<SVE_I>3_highpart): Likewise.
9331 (*<SVE_INT_BINARY_SD><SVE_SDI>3): Likewise.
9332 (*v<ASHIFT><SVE_I>3): Likewise.
9333 (*<su><MAXMIN><SVE_I>3): Likewise.
9334 (*<su><MAXMIN><SVE_F>3): Likewise.
9335 (*fma<SVE_F>4, *fnma<SVE_F>4): Likewise.
9336 (*fms<SVE_F>4, *fnms<SVE_F>4): Likewise.
9337 (*div<SVE_F>4): Likewise.
9339 2018-07-02 Richard Sandiford <richard.sandiford@arm.com>
9341 * tree-vect-patterns.c (vect_recog_widen_shift_pattern): Fix typo
9344 2018-07-02 Richard Biener <rguenther@suse.de>
9346 PR tree-optimization/86363
9347 * tree-ssa-sccvn.c (vn_reference_lookup_3): Check the
9348 memset argument refers to a non-variable address.
9350 2018-07-02 Aldy Hernandez <aldyh@redhat.com>
9352 * tree-vrp.c (extract_range_from_binary_expr_1): Abstract a lot of the
9353 {PLUS,MINUS}_EXPR code to...
9354 (adjust_symbolic_bound): ...here,
9355 (combine_bound): ...here,
9356 (set_value_range_with_overflow): ...and here.
9358 2018-07-02 Aldy Hernandez <aldyh@redhat.com>
9360 * tree-vrp.c (extract_range_from_unary_expr): Abstract ABS_EXPR
9362 (extract_range_from_abs_expr): ...here.
9364 2018-07-02 Eric Botcazou <ebotcazou@adacore.com>
9366 * config/i386/i386.c (ix86_finalize_stack_frame_flags): Do not overrule
9367 -fno-omit-frame-pointer when not optimizing.
9369 2018-07-02 Martin Liska <mliska@suse.cz>
9372 * ipa-pure-const.c (malloc_candidate_p): Revert usage of ::get.
9373 (propagate_nothrow): Likewise.
9375 2018-07-02 Martin Liska <mliska@suse.cz>
9378 * ipa-inline.c (early_inliner): Revert wrongly added ::get call.
9380 2018-07-02 David Malcolm <dmalcolm@redhat.com>
9382 * dumpfile.c (dump_generic_expr_loc): Undo removal of this
9383 function in r262149, changing "loc" param from source_location to
9384 const dump_location_t &.
9385 * dumpfile.h (dump_generic_expr_loc): Undo removal of this
9386 declaration, as above.
9388 2018-07-01 Paul Koning <ni1d@arrl.net>
9390 * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
9391 -munit-asm, -mgnu-asm, -mdec-asm.
9392 * config/pdp11/pdp11-protos.h (pdp11_gen_int_label): New.
9393 (pdp11_output_labelref): New.
9394 (pdp11_output_def): New.
9395 (pdp11_output_addr_vec_elt): New.
9396 * config/pdp11/pdp11.c: Use tab between opcode and operands. Use
9397 %# and %@ format codes.
9398 (pdp11_option_override): New.
9399 (TARGET_ASM_FILE_START_FILE_DIRECTIVE): Define.
9400 (pdp11_output_ident): New.
9401 (pdp11_asm_named_section): New.
9402 (pdp11_asm_init_sections): New.
9403 (pdp11_file_start): New.
9404 (pdp11_file_end): New.
9405 (output_ascii): Use .ascii/.asciz for -mdec-asm.
9406 (pdp11_asm_print_operand): Update %# and %$ for -mdec-asm. Add
9407 %o, like %c but octal.
9408 (pdp11_option_override): New.
9409 * config/pdp11/pdp11.h (TEXT_SECTION_ASM_OP): Update for
9411 (DATA_SECTION_ASM_OP): Ditto.
9412 (READONLY_DATA_SECTION_ASM_OP): New.
9413 (IS_ASM_LOGICAL_LINE_SEPARATOR): New.
9414 (ASM_GENERATE_INTERNAL_LABEL): Use new function.
9415 (ASM_OUTPUT_LABELREF): Ditto.
9416 (ASM_OUTPUT_DEF): Ditto.
9417 (ASM_OUTPUT_EXTERNAL): New.
9418 (ASM_OUTPUT_SOURCE_FILENAME): New.
9419 (ASM_OUTPUT_ADDR_VEC_ELT): Use new function.
9420 (ASM_OUTPUT_SKIP): Update for -mdec-asm.
9421 * config/pdp11/pdp11.md: Use tab between opcode and operands. Use
9422 %# and %@ format codes.
9423 * config/pdp11/pdp11.opt (mgnu-asm): New.
9424 (mdec-asm): Conflicts with -mgnu-asm and -munix-asm.
9425 (munix-asm): Conflicts with -mdec-asm and -mgnu-asm.
9426 * doc/invoke.txt (PDP-11 Options): Add -mgnu-asm.
9428 2018-07-01 Aldy Hernandez <aldyh@redhat.com>
9430 * tree-ssa-threadupdate.c (mark_threaded_blocks): Avoid
9431 dereferencing path[] beyond its length.
9433 (debug_all_paths): New.
9434 (rewire_first_differing_edge): New.
9435 (adjust_paths_after_duplication): New.
9436 (duplicate_thread_path): Call adjust_paths_after_duplication.
9438 (thread_through_all_blocks): Add new argument to
9439 duplicate_thread_path.
9441 2018-06-30 Jim Wilson <jimw@sifive.com>
9443 * config/riscv/predicates.md (p2m1_shift_operand): New.
9444 (high_mask_shift_operand): New.
9445 * config/riscv/riscv.md (lshrsi3_zero_extend_3+1): New combiner
9446 pattern using p2m1_shift_operand.
9447 (lshsi3_zero_extend_3+2): New combiner pattern using
9448 high_mask_shift_operand.
9450 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
9452 * tree-vect-patterns.c (vect_get_external_def_edge): New function,
9454 (vect_recog_rotate_pattern): ...here.
9455 (vect_convert_input): Try to insert casts of invariants in the
9457 * tree-vect-loop-manip.c (vect_loop_versioning): Don't require the
9458 preheader to be empty.
9460 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
9462 * tree-vect-patterns.c (append_pattern_def_seq): Take an optional
9463 vector type. If given, install it in the new statement's
9465 (vect_element_precision): New function.
9466 (vect_unpromoted_value): New struct.
9467 (vect_unpromoted_value::vect_unpromoted_value): New function.
9468 (vect_unpromoted_value::set_op): Likewise.
9469 (vect_look_through_possible_promotion): Likewise.
9470 (vect_joust_widened_integer, vect_joust_widened_type): Likewise.
9471 (vect_widened_op_tree, vect_convert_input): Likewise.
9472 (vect_convert_inputs, vect_convert_output): Likewise.
9473 (vect_recog_dot_prod_pattern): Use vect_look_through_possible_promotion
9474 to handle the optional cast of the multiplication result and
9475 vect_widened_op_tree to detect the widened multiplication itself.
9476 Do not require the input and output of promotion casts to have
9477 the same sign, but base the signedness of the operation on the
9478 input rather than the result. If the pattern includes two
9479 promotions, check that those promotions have the same sign.
9480 Do not restrict the MULT_EXPR handling to a double-width result;
9481 handle quadruple-width results and wider. Use vect_convert_inputs
9482 to convert the inputs to the common type.
9483 (vect_recog_sad_pattern): Use vect_look_through_possible_promotion
9484 to handle the optional cast of the ABS result. Also allow a sign
9485 change or a sign extension between the ABS and MINUS.
9486 Use vect_widened_op_tree to detect the widened subtraction and use
9487 vect_convert_inputs to convert the inputs to the common type.
9488 (vect_handle_widen_op_by_const): Delete.
9489 (vect_recog_widen_op_pattern): New function.
9490 (vect_recog_widen_mult_pattern): Use it.
9491 (vect_recog_widen_shift_pattern): Likewise.
9492 (vect_recog_widen_sum_pattern): Use
9493 vect_look_through_possible_promotion to handle the promoted
9496 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
9498 * gimple-iterator.c (gsi_for_stmt): Add a new overload that takes
9499 the containing gimple_seq *.
9500 * gimple-iterator.h (gsi_for_stmt): Declare it.
9501 * tree-vect-patterns.c (vect_recog_dot_prod_pattern)
9502 (vect_recog_sad_pattern, vect_recog_widen_sum_pattern)
9503 (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
9504 (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
9505 (vect_recog_mask_conversion_pattern): Remove STMT_VINFO_IN_PATTERN_P
9507 (vect_init_pattern_stmt, vect_set_pattern_stmt): New functions,
9509 (vect_mark_pattern_stmts): ...here. Handle cases in which the
9510 statement being replaced is part of an existing pattern
9511 definition sequence, inserting the new pattern statements before
9513 (vect_pattern_recog_1): Don't return a bool. If the statement
9514 is already part of a pattern, instead apply pattern matching
9515 to the pattern definition statements. Don't clear the
9516 STMT_VINFO_RELATED_STMT if is_pattern_stmt_p.
9517 (vect_pattern_recog): Don't break after the first match;
9518 continue processing the pattern definition statements instead.
9519 Don't bail out for STMT_VINFO_IN_PATTERN_P here.
9521 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
9523 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
9524 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
9525 (vect_recog_widen_sum_pattern): Use it.
9527 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
9529 * tree-vect-loop.c (vectorizable_reduction): Assert that the
9530 phi is not a pattern statement and has not been replaced by
9531 a pattern statement.
9532 * tree-vect-patterns.c (type_conversion_p): Don't check
9533 STMT_VINFO_IN_PATTERN_P.
9534 (vect_recog_vector_vector_shift_pattern): Likewise.
9535 (vect_recog_dot_prod_pattern): Expect vect_is_simple_use to return
9536 the pattern statement rather than the original statement; check
9537 directly for a WIDEN_MULT_EXPR here.
9538 * tree-vect-slp.c (vect_get_and_check_slp_defs): Expect
9539 vect_is_simple_use to return the pattern statement rather
9540 than the original statement; use is_pattern_stmt_p to check
9541 for such a pattern statement.
9542 * tree-vect-stmts.c (process_use): Expect vect_is_simple_use
9543 to return the pattern statement rather than the original statement;
9544 don't do the same transformation here.
9545 (vect_is_simple_use): If the defining statement has been replaced
9546 by a pattern statement, return the pattern statement instead.
9547 Remove the corresponding (local) transformation from the vectype
9550 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
9552 * tree-vectorizer.h (vect_is_simple_use): Move the gimple ** to the
9553 end and default to null.
9554 * tree-vect-loop.c (vect_create_epilog_for_reduction)
9555 (vectorizable_reduction): Update calls accordingly, dropping the
9556 gimple ** argument if the passed-back statement isn't needed.
9557 * tree-vect-patterns.c (vect_get_internal_def, type_conversion_p)
9558 (vect_recog_rotate_pattern): Likewise.
9559 (vect_recog_mask_conversion_pattern): Likewise.
9560 * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
9561 (vect_mask_constant_operand_p): Likewise.
9562 * tree-vect-stmts.c (is_simple_and_all_uses_invariant, process_use):
9563 (vect_model_simple_cost, vect_get_vec_def_for_operand): Likewise.
9564 (get_group_load_store_type, get_load_store_type): Likewise.
9565 (vect_check_load_store_mask, vect_check_store_rhs): Likewise.
9566 (vectorizable_call, vectorizable_simd_clone_call): Likewise.
9567 (vectorizable_conversion, vectorizable_assignment): Likewise.
9568 (vectorizable_shift, vectorizable_operation): Likewise.
9569 (vectorizable_store, vect_is_simple_cond): Likewise.
9570 (vectorizable_condition, vectorizable_comparison): Likewise.
9571 (get_same_sized_vectype, vect_get_mask_type_for_stmt): Likewise.
9572 (vect_is_simple_use): Rename the def_stmt argument to def_stmt_out
9573 and move it to the end. Cope with null def_stmt_outs.
9575 2018-06-30 Bernd Edlinger <bernd.edlinger@hotmail.de>
9577 * Makefile.in (FLAGS_TO_PASS): Add RANLIB_FOR_TARGET.
9579 2018-06-29 Jeff Law <law@redhat.com>
9581 * config/v850/v850.c (v850_legitimate_address_p): Handle large
9582 displacements for TARGET_V850E2V3 and newer.
9583 (TARGET_LRA_P): Remove. Defaults to LRA now.
9584 * config/v850/v850.md (sign23byte_load): Remove.
9585 (unsign23byte_load, sign23hword_load, unsign23hword_load): Likewise.
9586 (23word_load, 23byte_store, 23hword_store, 23word_store): Likewise.
9588 2018-06-29 Martin Liska <mliska@suse.cz>
9591 * coverage.c (coverage_init): Mangle full path name.
9592 * doc/invoke.texi: Document the change.
9593 * gcov-io.c (mangle_path): New.
9594 * gcov-io.h (mangle_path): Likewise.
9595 * gcov.c (mangle_name): Use mangle_path for path mangling.
9597 2018-06-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9599 * config/arm/arm.c (output_move_double): Don't allow STRD instructions
9600 if starting source register is not even.
9602 2018-06-29 Martin Liska <mliska@suse.cz>
9604 PR tree-optimization/86263
9605 * tree-switch-conversion.c (switch_decision_tree::try_switch_expansion):
9606 Make edge redirection.
9608 2018-06-29 David Malcolm <dmalcolm@redhat.com>
9610 * dumpfile.c (dump_loc): Add indentation based on scope depth.
9611 (dump_scope_depth): New variable.
9612 (get_dump_scope_depth): New function.
9613 (dump_begin_scope): New function.
9614 (dump_end_scope): New function.
9615 * dumpfile.h (get_dump_scope_depth): New declaration.
9616 (dump_begin_scope): New declaration.
9617 (dump_end_scope): New declaration.
9618 (class auto_dump_scope): New class.
9619 (AUTO_DUMP_SCOPE): New macro.
9620 * tree-vectorizer.h (DUMP_VECT_SCOPE): Reimplement in terms of
9623 2018-06-29 Richard Biener <rguenther@suse.de>
9625 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences): Assert
9626 compute_all_dependences succeeds.
9627 * tree-vect-loop.c (vect_get_datarefs_in_loop): Fail early if we
9628 exceed --param loop-max-datarefs-for-datadeps.
9630 2018-06-29 Jakub Jelinek <jakub@redhat.com>
9632 * config/rs6000/t-rs6000: Append rs6000-modes.h to TM_H.
9634 2018-06-28 Uros Bizjak <ubizjak@gmail.com>
9637 * config/i386/sse.md (*vec_extractv4si_0_zext_sse4): Use
9638 alternative 0 in preferred_for_speed attribute.
9640 2018-06-28 Paul Koning <ni1d@arrl.net>
9642 * config/pdp11/pdp11-protos.h (pdp11_shift_length): New function.
9643 * config/pdp11/pdp11.c (pdp11_shift_length): New function.
9644 * config/pdp11/pdp11.h (ADJUST_INSN_LENGTH): Remove.
9645 * config/pdp11/pdp11.md: Correct "length" attribute calculation
9646 for shift insn patterns.
9648 2018-06-28 David Malcolm <dmalcolm@redhat.com>
9650 * cgraph.c (cgraph_node::get_body): Replace assignments to
9651 "dump_file" with calls to set_dump_file.
9652 * dumpfile.c (alt_dump_file): Make static, and group with...
9653 (alt_flags): ...this definition.
9654 (dumps_are_enabled): New variable.
9655 (refresh_dumps_are_enabled): New function.
9656 (set_dump_file): New function.
9657 (set_alt_dump_file): New function.
9658 (gcc::dump_manager::dump_start): Replace assignments to
9659 "dump_file" and "alt_dump_file" with calls to set_dump_file and
9661 (gcc::dump_manager::dump_finish): Likewise.
9662 * dumpfile.h (alt_dump_file): Delete decl.
9663 (dumps_are_enabled): New variable decl.
9664 (set_dump_file): New function decl.
9665 (dump_enabled_p): Rewrite in terms of new "dumps_are_enabled"
9667 * tree-nested.c (lower_nested_functions): Replace assignments to
9668 "dump_file" with calls to set_dump_file.
9670 2018-06-28 Eric Botcazou <ebotcazou@adacore.com>
9672 * tree-cfg.c (verify_gimple_in_cfg): Call verify_location on the
9673 goto_locus of each outgoing edge of each basic block.
9675 2018-06-28 Richard Biener <rguenther@suse.de>
9677 * dwarf2out.c (decl_scope_table): Remove.
9678 (push_decl_scope): Likewise.
9679 (pop_decl_scope): Likewise.
9680 (gen_type_die_for_member): Do not call push/pop_decl_scope.
9681 (gen_struct_or_union_type_die): Likewise.
9682 (gen_tagged_type_die): Likewise.
9683 (dwarf2out_init): Do not initialize decl_scope_table.
9684 (dwarf2out_c_finalize): Do not free it.
9686 2018-06-28 Richard Biener <rguenther@suse.de>
9688 * dwarf2out.c (gen_subprogram_die): Use is_unit_die when
9689 deciding whether to not re-use a DIE.
9691 2018-06-28 Richard Biener <rguenther@suse.de>
9693 * dwarf2out.c (gen_subprogram_die): Always re-use DIEs with an
9694 DW_AT_abstract_origin attribute.
9696 2018-06-28 Martin Liska <mliska@suse.cz>
9698 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
9699 Use newly introduced constants.
9700 * tree-switch-conversion.h (struct jump_table_cluster):
9701 Define max_ratio_for_size and max_ratio_for_speed.
9703 2018-06-28 Martin Liska <mliska@suse.cz>
9705 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
9706 Add new checking assert to catch invalid state.
9707 (jump_table_cluster::can_be_handled): Handle single case
9709 (jump_table_cluster::is_beneficial): Bail out for such case.
9710 (bit_test_cluster::find_bit_tests):
9711 Add new checking assert to catch invalid state.
9712 (bit_test_cluster::can_be_handled): Handle single case
9714 (bit_test_cluster::is_beneficial): Bail out for such case.
9715 (switch_decision_tree::analyze_switch_statement):
9718 2018-06-28 Martin Liska <mliska@suse.cz>
9720 * common.opt: Introduce -completion option.
9721 * gcc.c (driver_handle_option): Handle it.
9722 (driver::main): Print completions if completion
9724 * opt-suggestions.c (option_proposer::get_completions):
9726 (option_proposer::suggest_completion): Likewise.
9727 (option_proposer::find_param_completions): Likewise.
9728 (verify_autocompletions): Likewise.
9729 (test_completion_valid_options): Likewise.
9730 (test_completion_valid_params): Likewise.
9731 (in_completion_p): Likewise.
9732 (empty_completion_p): Likewise.
9733 (test_completion_partial_match): Likewise.
9734 (test_completion_garbage): Likewise.
9735 (opt_proposer_c_tests): Likewise.
9736 * opt-suggestions.h: Declare new functions.
9737 * opts.c (common_handle_option): Handle OPT__completion_.
9738 * selftest-run-tests.c (selftest::run_tests): Add
9739 opt_proposer_c_tests.
9740 * selftest.c (assert_str_startswith): New.
9741 * selftest.h (assert_str_startswith): Likewise.
9742 (opt_proposer_c_tests): New.
9743 (ASSERT_STR_STARTSWITH): Likewise.
9745 2018-06-28 Martin Liska <mliska@suse.cz>
9747 * Makefile.in: Add opt-suggestions.o.
9748 * gcc-main.c: Include opt-suggestions.h.
9749 * gcc.c (driver::driver): Likewise.
9750 (driver::~driver): Remove m_option_suggestions.
9751 (driver::build_option_suggestions): Moved to option_proposer.
9752 (driver::suggest_option): Likewise.
9753 (driver::handle_unrecognized_options): Use option_proposer.
9754 * gcc.h (class driver): Add new memver m_option_proposer.
9755 * opt-suggestions.c: New file.
9756 * opt-suggestions.h: New file.
9758 2018-06-28 Martin Liska <mliska@suse.cz>
9760 * vec.h (class auto_string_vec): New (moved from auto_argvec).
9761 (auto_string_vec::~auto_string_vec): Likewise.
9763 2018-06-28 Eric Botcazou <ebotcazou@adacore.com>
9765 * tree-inline.h (struct copy_body_data): Move remapping_type_depth and
9766 prevent_decl_creation_for_types fields up and add reset_location field.
9767 * tree-inline.c (remap_gimple_stmt): Force input_location on the new
9768 statement if id->reset_location is true.
9769 (copy_edges_for_bb): Do not set goto_locus on the new edges if
9770 id->reset_location is true.
9771 (copy_phis_for_bb): Force input_location on the arguments if
9772 id->reset_location is true.
9773 (expand_call_inline): Set id->reset_location if DECL_IGNORED_P
9774 is set on the function to be inlined.
9776 2018-06-27 Stephan Bergmann <sbergman@redhat.com>
9778 * doc/invoke.texi (Debugging Options): Add -gsplit-dwarf.
9780 2018-06-27 Dimitar Dimitrov <dimitar@dinux.eu>
9782 * lra-eliminations.c (update_reg_eliminate): Mark all spanning hard
9783 registers for Pmode.
9784 * lra-lives.c (check_pseudos_live_through_calls): Mark all spanning
9785 hard registers for the clobbered pseudo.
9787 2018-06-27 Paul Koning <ni1d@arrl.net>
9789 * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
9790 mutually exclusive options.
9791 * config/pdp11/constraints.md (h): New constraint.
9792 (O): Update definition to match shift code generation.
9793 (D): New constraint.
9794 * config/pdp11/pdp11-modes.def (CCNZ): Define mode.
9796 * config/pdp11/pdp11-protos.h (int_no_side_effect_operand): New
9798 (output_jump): Change arguments.
9799 (pdp11_fixed_cc_regs): New function.
9800 (pdp11_cc_mode): Ditto.
9801 (pdp11_expand_shift): Ditto.
9802 (pdp11_assemble_shift): Ditto.
9803 (pdp11_small_shift): Ditto.
9804 (pdp11_branch_cost): Remove.
9805 * config/pdp11/pdp11.c (pdp11_assemble_integer): Remove comments
9807 (pdp11_register_move_cost): Update for CC registers.
9808 (pdp11_rtx_costs): Add case for LSHIFTRT.
9809 (pdp11_output_jump): Add CCNZ mode conditional branches.
9810 (notice_update_cc_on_set): Remove.
9811 (pdp11_cc_mode): New function.
9812 (simple_memory_operand): Correct pre/post decrement case.
9813 (no_side_effect_operand): New function.
9814 (pdp11_regno_reg_class): Add CC_REGS class.
9815 (pdp11_fixed_cc_regs): New function.
9816 (pdp11_small_shift): New function.
9817 (pdp11_expand_shift): New function to expand shift insns.
9818 (pdp11_assemble_shift): New function to output shifts.
9819 (pdp11_branch_cost): Remove.
9820 (pdp11_modes_tieable_p): Make QI/HI modes tieable.
9821 * config/pdp11/pdp11.h (SIZE_TYPE): Ensure 16-bit type.
9822 (WCHAR_TYPE): Ditto.
9823 (PTRDIFF_TYPE): Ditto.
9824 (ADJUST_INSN_LENGTH): New macro.
9825 (FIXED_REGISTERS): Add CC registers.
9826 (CALL_USED_REGISTERS): Ditto.
9828 (REG_CLASS_NAMES): Ditto.
9829 (REG_CLASS_CONTENTS): Ditto.
9830 (SELECT_CC_MODE): Use new function.
9831 (TARGET_FLAGS_REGNUM): New macro.
9832 (TARGET_FIXED_CONDITION_CODE_REGS): Ditto.
9833 (cc0_reg_rtx): Remove.
9834 (CC_STATUS_MDEP): Remove.
9835 (CC_STATUS_MDEFP_INIT): Remove.
9836 (CC_IN_FPU): Remove.
9837 (NOTICE_UPDATE_CC): Remove.
9838 (REGISTER_NAMES): Add CC registers.
9839 (BRANCH_COST): Change to constant 1.
9840 * config/pdp11/pdp11.md: Rewrite for CCmode condition code
9842 * config/pdp11/pdp11.opt (mbcopy): Remove.
9843 (mbcopy-builtin): Remove.
9844 (mbranch-cheap): Remove.
9845 (mbranch-expensive): Remove.
9846 * config/pdp11/predicates.md (expand_shift_operand): Update to
9847 match shift code generation.
9848 (ccnz_operator): New predicate.
9849 * doc/invoke.texi (PDP-11 Options): Remove deleted options
9850 -mbcopy, -mbcopy-builtin, -mbranch-cheap, -mbranch-expensive.
9851 Remove non-existent option -mabshi, -mno-abshi. Document mutually
9853 * doc/md.texi (PDP-11): Document new D and h constraints. Update
9854 description of O constraint.
9856 2018-06-27 Jeff Law <law@redhat.com>
9857 Austin Law <austinklaw@gmail.com>
9859 * config/v850/v850.md (addsi3_set_flags): New pattern.
9860 (subsi3_set_flags, negsi2_set_flags, andsi3_set_flags): Likewise.
9861 (iorsi3_set_flags, xorsi3_set_flags, one_cmplsi2_set_flags): Likewise.
9862 (zero_extendhisi2_v850_set_flags): Likewise.
9863 (zero_extendqisi2_v850_set_flags): Likewise.
9864 (ashlsi3_set_flags, ashlsi3_v850e2_set_flags): Likewise.
9865 (lshrsi3_set_flags, lshrsi3_v850e2_set_flags): Likewise.
9866 (ashrsi3_set_flags, ashrsi3_v850e2_set_flags): Likewise.
9868 * config/v850/v850-protos.h (notice_update_cc): Remove.
9869 * config/v850/v850.c (v850_compare_op0, v850_compare_op1): Remove.
9870 (v850_print_operand): Handle 'D' and "d".
9871 (v850_select_cc_mode): Remove ATTRIBUTE_UNUSED for last argument.
9872 Add handling of arithmetic/logical operations compared against zero.
9873 (v850_gen_float_compare): Remove ATTRIBUTE_UNUSED for last argument.
9874 Do not look at v850_compare_op, instead get mode from last argument.
9875 (v850_gen_compare): Remove
9876 (increment_stack): Use addsi3_clobber_flags to avoid splitting failure
9877 after reload for prologue insns.
9878 (expand_prologue): Account for CLOBBER of CC_REGNUM in various
9880 (construct_save_jarl): Likewise.
9881 (TARGET_FLAGS_REGNUM): Define.
9882 * config/v850/v850.h (v850_compare_op0, v850_compare_op1): Remove.
9883 (NOTICE_UPDATE_CC): Remove.
9884 * config/v850/v850.md (v850_tst1): Use (reg:CCZ CC_REGNUM) rather
9885 than cc0. Conditionalize on reload_completed.
9886 (cmpsi_insn, setfcc_insn): Likewise.
9887 (tst1 splitter): Turn into define_and_split which sets the flags
9889 (cstoresi4, cbranchsf4, cbranchdf4, cbranchsi4_insn): Likewise.
9890 (cbranchsi4, branch_normal, branch_invert): Do not expose cc0 here.
9891 (cstoresf4, cstoredf4): Clobber the flags.
9892 (cmpsi, cmpsf, cmpdf): Remove expanders.
9893 (setf_insn): Remove pattern.
9894 (addsi3): Turn into define_and_split which clobbers the flags after
9895 reload and a suitable pattern (addsi3_clobber_flags) for use after
9897 (subsi3, negsi2, andsi3, iorsi3, xorsi3, one_cmplsi2) Likewise.
9898 (ashlsi3, ashlsi3_v850e2, lshrsi3, lsh4si3_v850e2): Likewise.
9899 (ashrsi3, ashrsi3_v850e2): Likewise.
9900 (bins): Clobber the flags.
9901 (movsicc_normal_cc, movsicc_normal, movsicc_tst1): Likewise.
9902 (movsicc_tst1_revesed, sasf, swap and rotate patterns): Likewise.
9903 (fix_loop_counter, call_internal_short, call_internal_long): Likewise.
9904 (call_value_internal_short, call_value_internal_long): Likewise.
9905 (callt_save_interrupt, callt_return_interrupt): Likewise.
9906 (save_interrupt, return_interrupt): Likewise.
9907 (callt_save_all_interrupt, save_all_interrupt): Likewise.
9908 (_save_all_interrupt, callt_restore_all_interrupt): Likewise.
9909 (restore_all_interrupt, _restore_all_interrupt): Likewise.
9910 (All FP comparisons): Only allow after reload has completed.
9912 (divh, divhu): Tweak output template.
9913 (branch_z_normal, branch_z_invert): Remove
9914 (branch_nz_normal, branch_nz_invert): Likewise.
9915 (extendhisi_insn, extendqisi_insn): Do not clobber flags.
9917 * config/v850/v850-modes.def (CCZ, CCNZ): Add new modes.
9918 * config/v850/v850.c (notice_update_cc): Remove.
9919 * config/v850/v850.h (CC_OVERFLOW_UNUSABLE): Remove
9920 (CC_NO_CARRY): Likewise.
9921 (NOTICE_UPDATE_CC): Define to nothing.
9922 * config/v850/v850.md: Remove block comment on cc0 handling
9923 Remove "cc" attribute from all patterns. Remove cc_status handling
9924 from all patterns. Minor formatting fixes.
9926 2018-06-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9928 * config/aarch64/aarch64-cores.def (cortex-a76): New entry.
9929 (cortex-a76.cortex-a55): Likewise.
9930 * config/aarch64/aarch64-tune.md: Regenerate.
9931 * doc/invoke.texi (AArch64 Options): Document cortex-a76 and
9932 cortex-a76.cortex-a55.
9934 2018-06-27 Jeff Law <law@redhat.com>
9936 * config/v850/t-v850 (MULTILIB_OPTIONS): Remove 8byte-align.
9937 (MULTILIB_DIRNAMES): Similarly.
9939 2018-06-27 Eric Botcazou <ebotcazou@adacore.com>
9941 * gimple.h (gimple_return_retbnd): Delete.
9942 (gimple_return_set_retbnd): Likewise.
9943 * cgraphunit.c (cgraph_node::expand_thunk): Remove call to
9944 gimple_return_set_retbnd.
9945 * gimple-pretty-print.c (dump_gimple_return): Remove call to
9946 gimple_return_retbnd and adjust.
9947 * tree-inline.h (struct copy_body_data): Remove retbnd field.
9948 * tree-inline.c (remap_gimple_stmt): Remove handling of retbnd.
9949 Explicitly return NULL in a couple more cases. Move assertion
9950 on debug statements and remove unreachable code.
9951 (reset_debug_binding): Do not test id->retbnd.
9952 (expand_call_inline): Do not set it.
9954 2018-06-27 Rasmus Villemoes <rv@rasmusvillemoes.dk>
9956 * configure.ac: Add --disable-gcov option.
9957 * configure: Regenerate.
9958 * Makefile.in: Honour @enable_gcov@.
9959 * doc/install.texi: Document --disable-gcov.
9961 2018-06-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9963 * config/arm/arm-cpus.in (cortex-a76): New entry.
9964 (cortex-a76.cortex-a55): Likewise.
9965 * config/arm/arm-tables.opt: Regenerate.
9966 * config/arm/arm-tune.md: Likewise.
9967 * config/arm/driver-arm.c (arm_cpu_table): Add Cortex-A76 entry.
9968 * doc/invoke.texi (ARM Options): Document cortex-a76 and
9969 cortex-a76.cortex-a55.
9971 2018-06-27 Tamar Christina <tamar.christina@arm.com>
9974 * config/aarch64/aarch64.md (*movhf_aarch64): Add dup v0.4h pattern.
9976 2018-06-27 Siddhesh Poyarekar <siddhesh@sourceware.org>
9978 * config/aarch64/aarch64.h (CALL_USE_REGISTERS): Fix obsolete
9980 (EPILOGUE_USES): Likewise.
9982 2018-06-26 Eric Botcazou <ebotcazou@adacore.com>
9984 * tree-inline.c (remap_location): New function extracted from...
9985 (copy_edges_for_bb): Add ID parameter. Remap goto_locus.
9986 (copy_phis_for_bb): ...here. Call remap_location.
9987 (copy_cfg_body): Adjust call to copy_edges_for_bb.
9989 2018-06-26 Aaron Sawdey <acsawdey@linux.ibm.com>
9991 * config/rs6000/rs6000-string.c (expand_block_clear): Don't use
9992 unaligned vsx for 16B memset.
9994 2018-06-26 Segher Boessenkool <segher@kernel.crashing.org>
9997 * config/rs6000/rs6000.c (rs6000_init_builtins): Do not set
9998 ieee128_float_type_node to long_double_type_node unless
9999 TARGET_LONG_DOUBLE_128 is set.
10001 2018-06-26 David Malcolm <dmalcolm@redhat.com>
10003 * cfgloop.c (get_loop_location): Convert return type from
10004 location_t to dump_user_location_t, replacing INSN_LOCATION lookups
10005 by implicit construction from rtx_insn *, and using
10006 dump_user_location_t::from_function_decl for the fallback case.
10007 * cfgloop.h (get_loop_location): Convert return type from
10008 location_t to dump_user_location_t.
10009 * cgraphunit.c (walk_polymorphic_call_targets): Update call to
10010 dump_printf_loc to pass in a dump_location_t rather than a
10011 location_t, via the gimple stmt.
10012 * coverage.c (get_coverage_counts): Update calls to
10013 dump_printf_loc to pass in dump_location_t rather than a
10015 * doc/optinfo.texi (Dump types): Convert example of
10016 dump_printf_loc from taking "locus" to taking "insn". Update
10017 description of the "_loc" calls to cover dump_location_t.
10018 * dumpfile.c: Include "backend.h", "gimple.h", "rtl.h", and
10020 (dump_user_location_t::dump_user_location_t): New constructors,
10021 from gimple *stmt and rtx_insn *.
10022 (dump_user_location_t::from_function_decl): New function.
10023 (dump_loc): Make static.
10024 (dump_gimple_stmt_loc): Convert param "loc" from location_t to
10025 const dump_location_t &.
10026 (dump_generic_expr_loc): Delete.
10027 (dump_printf_loc): Convert param "loc" from location_t to
10028 const dump_location_t &.
10029 (selftest::test_impl_location): New function.
10030 (selftest::dumpfile_c_tests): New function.
10031 * dumpfile.h: Include "profile-count.h".
10032 (class dump_user_location_t): New class.
10033 (struct dump_impl_location_t): New struct.
10034 (class dump_location_t): New class.
10035 (dump_printf_loc): Convert 2nd param from source_location to
10036 const dump_location_t &.
10037 (dump_generic_expr_loc): Delete.
10038 (dump_gimple_stmt_loc): Convert 2nd param from source_location to
10039 const dump_location_t &.
10040 * gimple-fold.c (fold_gimple_assign): Update call to
10041 dump_printf_loc to pass in a dump_location_t rather than a
10042 location_t, via the gimple stmt.
10043 (gimple_fold_call): Likewise.
10044 * gimple-loop-interchange.cc
10045 (loop_cand::analyze_iloop_reduction_var): Update for change to
10046 check_reduction_path.
10047 (tree_loop_interchange::interchange): Update for change to
10048 find_loop_location.
10049 * graphite-isl-ast-to-gimple.c (scop_to_isl_ast): Update for
10050 change in return-type of find_loop_location.
10051 (graphite_regenerate_ast_isl): Likewise.
10052 * graphite-optimize-isl.c (optimize_isl): Likewise.
10053 * graphite.c (graphite_transform_loops): Likewise.
10054 * ipa-devirt.c (ipa_devirt): Update call to dump_printf_loc to
10055 pass in a dump_location_t rather than a location_t, via the
10057 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
10058 * ipa.c (walk_polymorphic_call_targets): Likewise.
10059 * loop-unroll.c (report_unroll): Convert "locus" param from
10060 location_t to dump_location_t.
10061 (decide_unrolling): Update for change to get_loop_location's
10063 * omp-grid.c (struct grid_prop): Convert field "target_loc" from
10064 location_t to dump_user_location_t.
10065 (grid_find_single_omp_among_assignments_1): Updates calls to
10066 dump_printf_loc to pass in a dump_location_t rather than a
10067 location_t, via the gimple stmt.
10068 (grid_parallel_clauses_gridifiable): Convert "tloc" from
10069 location_t to dump_location_t. Updates calls to dump_printf_loc
10070 to pass in a dump_location_t rather than a location_t, via the
10072 (grid_inner_loop_gridifiable_p): Likewise.
10073 (grid_dist_follows_simple_pattern): Likewise.
10074 (grid_gfor_follows_tiling_pattern): Likewise.
10075 (grid_target_follows_gridifiable_pattern): Likewise.
10076 (grid_attempt_target_gridification): Convert initialization
10077 of local "grid" from memset to zero-initialization; FIXME: does
10078 this require C++11? Update call to dump_printf_loc to pass in a
10079 optinfo_location rather than a location_t, via the gimple stmt.
10080 * profile.c (read_profile_edge_counts): Updates call to
10081 dump_printf_loc to pass in a dump_location_t rather than a
10083 (compute_branch_probabilities): Likewise.
10084 * selftest-run-tests.c (selftest::run_tests): Call
10086 * selftest.h (dumpfile_c_tests): New decl.
10087 * tree-loop-distribution.c (pass_loop_distribution::execute):
10088 Update for change in return type of find_loop_location.
10089 * tree-parloops.c (parallelize_loops): Likewise.
10090 * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Convert
10091 "locus" from location_t to dump_user_location_t.
10092 (canonicalize_loop_induction_variables): Likewise.
10093 * tree-ssa-loop-ivopts.c (tree_ssa_iv_optimize_loop): Update
10094 for change in return type of find_loop_location.
10095 * tree-ssa-loop-niter.c (number_of_iterations_exit): Update call
10096 to dump_printf_loc to pass in a dump_location_t rather than a
10097 location_t, via the stmt.
10098 * tree-ssa-sccvn.c (eliminate_dom_walker::before_dom_children):
10100 * tree-vect-loop-manip.c (find_loop_location): Convert return
10101 type from source_location to dump_user_location_t.
10102 (vect_do_peeling): Update for above change.
10103 (vect_loop_versioning): Update for change in type of
10105 * tree-vect-loop.c (check_reduction_path): Convert "loc" param
10106 from location_t to dump_user_location_t.
10107 (vect_estimate_min_profitable_iters): Update for change in type
10109 * tree-vect-slp.c (vect_print_slp_tree): Convert param "loc" from
10110 location_t to dump_location_t.
10111 (vect_slp_bb): Update for change in type of vect_location.
10112 * tree-vectorizer.c (vect_location): Convert from source_location
10113 to dump_user_location_t.
10114 (try_vectorize_loop_1): Update for change in vect_location's type.
10115 (vectorize_loops): Likewise.
10116 (increase_alignment): Likewise.
10117 * tree-vectorizer.h (vect_location): Convert from source_location
10118 to dump_user_location_t.
10119 (find_loop_location): Convert return type from source_location to
10120 dump_user_location_t.
10121 (check_reduction_path): Convert 1st param from location_t to
10122 dump_user_location_t.
10123 * value-prof.c (check_counter): Update call to dump_printf_loc to
10124 pass in a dump_user_location_t rather than a location_t; update
10125 call to error_at for change in type of "locus".
10126 (check_ic_target): Update call to dump_printf_loc to
10127 pass in a dump_user_location_t rather than a location_t, via the
10130 2018-06-26 Robin Dapp <rdapp@linux.vnet.ibm.com>
10132 * config/s390/s390.h (enum processor_flags): Do not use
10133 default tune parameter when -march was specified.
10135 2018-06-26 Jakub Jelinek <jakub@redhat.com>
10138 * config/i386/i386.md (setcc + movzbl to xor + setcc peephole2s):
10139 Check reg_overlap_mentioned_p in addition to reg_set_p with the same
10142 2018-06-26 Richard Biener <rguenther@suse.de>
10144 PR tree-optimization/86287
10146 * tree-vect-loop.c (vect_transform_loop_stmt): Fix read-after-free.
10147 (vect_analyze_loop): Initialize n_stmts.
10149 2018-06-26 Richard Biener <rguenther@suse.de>
10151 PR middle-end/86271
10152 * fold-const.c (fold_convertible_p): Pointer extension
10155 2018-06-26 Alexandre Oliva <aoliva@redhat.com>
10158 * dwarf2out.c (loc_list_has_views): Adjust comments.
10159 (dw_loc_list): Split single cross-partition range with
10162 2018-06-25 Jeff Law <law@redhat.com>
10164 * common/config/v850/v850-common.c (TARGET_DEFAULT_TARGET_FLAGS): Turn
10165 on -mbig-switch by default.
10167 * config/v850/predicates.md (const_float_1_operand): Fix match_code
10169 (const_float_0_operand): Remove unused predicate.
10170 * config/v850/v850.md (define_constants): Remove UNSPEC_LOOP.
10171 (define_c_enum unspec): Add LOOP, RCP and RSQRT constants.
10172 (recipsf2): New expander. Original pattern now called
10174 (recipdf2, recipdf2_insn): Similarly.
10175 (rsqrtsf2, rsqrtsf2_insn): Similarly
10176 (rsqrtdf2, rsqrtdf2_insn): Similarly
10178 2018-06-26 Gerald Pfeifer <gerald@pfeifer.com>
10180 * ginclude/stddef.h: Remove an obsolete comment on FreeBSD 5.
10181 Simplify logic for FreeBSD (twice).
10183 2018-06-25 Martin Sebor <msebor@redhat.com>
10185 PR tree-optimization/86204
10186 * tree-ssa-strlen.c (handle_builtin_strlen): Avoid storing
10187 a strnlen result if it's less than the length of the string.
10189 2018-06-25 Martin Sebor <msebor@redhat.com>
10191 PR tree-optimization/85700
10192 * gimple-fold.c (gimple_fold_builtin_strncat): Adjust comment.
10193 * tree-ssa-strlen.c (is_strlen_related_p): Handle integer subtraction.
10194 (maybe_diag_stxncpy_trunc): Distinguish strncat from strncpy.
10196 2018-06-25 Martin Sebor <msebor@redhat.com>
10198 * doc/extend.texi (Zero-length arrays): Update and clarify.
10200 2018-06-25 Michael Meissner <meissner@linux.ibm.com>
10202 * config.gcc (powerpc64le*): Revert January 16th, 2018 patch that
10203 added IEEE/IBM long double multilib support on PowerPC little
10204 endian Linux systems.
10205 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Likewise.
10206 (MULTILIB_DEFAULTS): Likewise.
10207 * config/rs6000/rs6000.c (rs6000_option_override_internal):
10209 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Likewise.
10210 * config/rs6000/t-ldouble-linux64le-ibm: Delete, no longer used.
10211 * config/rs6000/t-ldouble-linux64le-ieee: Delete, no longer used.
10213 2018-06-25 Alexander Monakov <amonakov@ispras.ru>
10215 PR middle-end/86311
10216 * sort.cc (REORDER_23): Avoid memcpy with same destination and source.
10217 (REORDER_45): Likewise.
10219 2018-06-25 Jeff Law <law@redhat.com>
10221 * config/v850/v850.md (divmodhi4): Make sure to sign extend the
10222 dividend to 32 bits. Adjust length.
10223 (udivmodhi4): Cleanup output template. Fix length.
10225 2018-06-25 Carl Love <cel@us.ibm.com>
10227 * config/rs6000/vsx.md: Change word selector to prefered location.
10229 2018-06-25 Richard Biener <rguenther@suse.de>
10231 PR tree-optimization/86304
10232 * tree-vectorizer.c (vectorize_loops): Walk over new possibly
10233 epilogue-if-converted loops as well.
10235 2018-06-25 Jan Hubicka <hubicka@ucw.cz>
10237 * lto-section-out.c (lto_begin_section): Do not print section
10238 name for noaddr and unnumbered dumps.
10240 2018-06-25 Richard Biener <rguenther@suse.de>
10242 * tree-vectorizer.h (struct vec_info_shared): New structure
10243 with parts split out from struct vec_info and loop_nest from
10244 struct _loop_vec_info.
10245 (struct vec_info): Adjust accordingly.
10246 (struct _loop_vec_info): Likewise.
10247 (LOOP_VINFO_LOOP_NEST): Adjust.
10248 (LOOP_VINFO_DATAREFS): Likewise.
10249 (LOOP_VINFO_DDRS): Likewise.
10250 (struct _bb_vec_info): Likewise.
10251 (BB_VINFO_DATAREFS): Likewise.
10252 (BB_VINFO_DDRS): Likewise.
10253 (struct _stmt_vec_info): Add dr_aux member.
10254 (DR_VECT_AUX): Adjust to refer to member of DR_STMTs vinfo.
10255 (DR_MISALIGNMENT_UNINITIALIZED): New.
10256 (set_dr_misalignment): Adjust.
10257 (dr_misalignment): Assert misalign isn't DR_MISALIGNMENT_UNINITIALIZED.
10258 (vect_analyze_loop): Adjust prototype.
10259 (vect_analyze_loop_form): Likewise.
10260 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
10261 Compute dependences lazily.
10262 (vect_record_base_alignments): Use shared datarefs/ddrs.
10263 (vect_verify_datarefs_alignment): Likewise.
10264 (vect_analyze_data_refs_alignment): Likewise.
10265 (vect_analyze_data_ref_accesses): Likewise.
10266 (vect_analyze_data_refs): Likewise.
10267 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Add
10268 constructor parameter for shared part.
10269 (vect_analyze_loop_form): Pass in shared part and adjust.
10270 (vect_analyze_loop_2): Pass in storage for the number of
10271 stmts. Move loop nest finding to the caller. Compute
10273 (vect_analyze_loop): Pass in shared part.
10274 (vect_transform_loop): Verify shared datarefs are unchanged.
10275 * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Add
10276 constructor parameter for shared part.
10277 (vect_slp_analyze_bb_1): Pass in shared part and adjust.
10278 (vect_slp_bb): Verify shared datarefs are unchanged before
10280 * tree-vect-stmts.c (ensure_base_align): Adjust for DR_AUX
10282 (new_stmt_vec_info): Initialize DR_AUX misalignment to
10283 DR_MISALIGNMENT_UNINITIALIZED.
10284 * tree-vectorizer.c (vec_info::vec_info): Add constructor
10285 parameter for shared part.
10286 (vec_info::~vec_info): Adjust.
10287 (vec_info_shared::vec_info_shared): New.
10288 (vec_info_shared::~vec_info_shared): Likewise.
10289 (vec_info_shared::save_datarefs): Likewise.
10290 (vec_info_shared::check_datarefs): Likewise.
10291 (try_vectorize_loop_1): Construct shared part live for analyses
10292 of a single loop for multiple vector sizes.
10293 * tree-parloops.c (gather_scalar_reductions): Adjust.
10295 2018-06-25 Richard Biener <rguenther@suse.de>
10297 * tree-vect-data-refs.c (vect_find_stmt_data_reference): Modify
10298 DR for SIMD lane accesses here and mark DR with (void *)-1 aux.
10299 (vect_analyze_data_refs): Remove similar code from here and
10300 simplify accordingly.
10302 2018-06-25 Richard Biener <rguenther@suse.de>
10304 * tree-vect-data-refs.c (vect_check_gather_scatter): Fail
10305 for reverse storage order accesses rather than asserting
10306 they cannot happen here.
10308 2018-06-25 Tom de Vries <tdevries@suse.de>
10311 * config/i386/i386.md (define_insn "*tls_global_dynamic_64_<mode>"):
10312 Use data16 instead of .byte for insn prefix.
10314 2018-06-25 Andreas Krebbel <krebbel@linux.ibm.com>
10317 * parser.c (make_char_string_pack): Pass this literal chars
10318 through cpp_interpret_string.
10319 (cp_parser_userdef_numeric_literal): Check the result of
10320 make_char_string_pack.
10322 2018-06-24 Maya Rashish <coypu@sdf.org>
10324 * ginclude/stddef.h: Simplify conditions around avoiding
10325 re-definition of __size_t.
10327 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
10329 * lto-streamer-out.c (tree_is_indexable): Make LABEL_DECL nonindexable
10330 unless it is forced or nonlocal; assert that we stream no IMPORTED_DECL.
10332 2018-06-22 Maya Rashish <coypu@sdf.org>
10334 * doc/invoke.texi (mno-fancy-math-387): Update for changes
10335 made to OpenBSD and NetBSD through the years.
10337 2018-06-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
10339 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
10340 behavior of vec_pack (vector double, vector double) to match
10341 behavior of vec_float2 (vector double, vector double).
10343 2018-06-22 Olivier Hainque <hainque@adacore.com>
10345 * gimplify.c (gimplify_function_tree): Prevent creation
10346 of a trampoline for the address of the current function
10347 passed to entry/exit instrumentation hooks.
10349 2018-06-22 Aaron Sawdey <acsawdey@linux.ibm.com>
10352 * config/rs6000/rs6000-string.c (expand_strn_compare): Handle -m32
10355 2018-06-22 Martin Liska <mliska@suse.cz>
10357 PR tree-optimization/86263
10358 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
10359 Bail out if is_enabled is false.
10360 * tree-switch-conversion.h (jump_table_cluster::is_enabled):
10362 (jump_table_cluster::is_enabled): New function.
10364 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
10366 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
10367 BINFO_BASE_ACCESSES and BINFO_VPTR_FIELD.
10368 * tree-streamer-in.c (streamer_read_tree_bitfields): Likewise.
10369 (lto_input_ts_binfo_tree_pointers): Likewise.
10370 * tree-streamer-out.c (streamer_write_tree_bitfields,
10371 write_ts_binfo_tree_pointers): Likewise.
10372 * tree.c (free_lang_data_in_binfo): Clear BINFO_VPTR_FIELD.
10374 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
10376 * tree.c (free_lang_data_in_type): Free all TYPE_VFIELDs.
10378 2018-06-22 Martin Liska <mliska@suse.cz>
10380 * symbol-summary.h (get): Make it pure and inline move
10381 functionality from ::get function.
10382 (get): Remove and inline into ::get and ::get_create.
10383 (get_create): Move code from ::get function.
10385 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
10388 * config/i386/sol2.h (CPP_SPEC): Don't pass -P for
10389 -x assembler-with-cpp.
10391 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
10393 * config/sol2.h (TARGET_OS_CPP_BUILTINS): Define
10394 _FILE_OFFSET_BITS=64 for C++.
10396 2018-06-21 Michael Meissner <meissner@linux.ibm.com>
10398 * config/rs6000/rs6000.md (extendtfif2): Add missing 128-bit
10399 conversion insn that shows up when pr85657-3.c is compiled using
10400 IEEE 128-bit long double.
10401 (neg<mode>2_internal): Use the correct mode to check whether the
10402 mode is IBM extended.
10403 * config/rs6000/rs6000.c (init_float128_ieee): Prevent complex
10404 multiply and divide external functions from being created more
10407 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
10409 * cfgrtl.c (fixup_reorder_chain): Do not emit NOPs in DECL_IGNORED_P
10411 (rtl_merge_blocks): Likewise. Do not emit a NOP if the location of
10412 the edge can be forwarded.
10413 (cfg_layout_merge_blocks): Likewise.
10415 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
10417 * except.c (finish_eh_generation): Commit edge insertions only after
10418 the EH edges have been redirected from post-landing to landing pads.
10420 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
10422 * tree-nested.c (get_frame_type): Use create_tmp_var_raw instead of
10423 create_tmp_var_for to create the FRAME decl.
10424 (finalize_nesting_tree_1): Do not unchain the FRAME decl.
10426 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
10428 * tree-inline.c (copy_edges_for_bb): Minor tweak.
10429 (maybe_move_debug_stmts_to_successors): Also reset the locus of the
10430 debug statement when resetting its value.
10431 (expand_call_inline): Copy the locus of the call onto the assignment
10432 of the return value, if any. Use local variable in more cases.
10434 2018-06-21 Martin Liska <mliska@suse.cz>
10436 * ipa-pure-const.c (propagate_nothrow): Use
10437 funct_state_summaries->get.
10438 (dump_malloc_lattice): Likewise.
10439 (propagate_malloc): Likewise.
10441 2018-06-21 Richard Biener <rguenther@suse.de>
10443 * lto-streamer-out.c (DFS::DFS_write_tree_body): Update outdated
10444 comment. Follow BLOCK_ABSTRACT_ORIGIN unconditionally.
10445 * tree-streamer-in.c (lto_input_ts_block_tree_pointers): Update
10447 * tree-streamer-out.c (write_ts_block_tree_pointers): Stream
10448 BLOCK_ABSTRACT_ORIGIN unconditionally.
10450 2018-06-21 David Malcolm <dmalcolm@redhat.com>
10452 * ipa-cp.c (ipcp_driver): Set edge_clone_summaries to NULL after
10454 * ipa-reference.c (ipa_reference_c_finalize): Delete
10455 ipa_ref_opt_sum_summaries and set it to NULL.
10457 2018-06-21 Tom de Vries <tdevries@suse.de>
10459 PR tree-optimization/85859
10460 * tree-ssa-tail-merge.c (stmt_local_def): Copy gimple_is_call
10461 test with comment from bb_no_side_effects_p.
10463 2018-06-21 Richard Biener <rguenther@suse.de>
10465 PR tree-optimization/86232
10466 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Adjust
10467 max for constant niter.
10469 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
10471 * config/aarch64/aarch64-simd.md
10472 (*aarch64_crypto_aes<aes_op>v16qi_xor_combine): New.
10474 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
10476 * config/aarch64/aarch64-simd.md (aarch64_crypto_aes<aes_op>v16qi):
10477 Make opernads of the unspec commutative.
10479 2018-06-21 Richard Biener <rguenther@suse.de>
10481 * tree-data-ref.c (dr_step_indicator): Handle NULL DR_STEP.
10482 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr):
10483 Avoid calling vect_mark_for_runtime_alias_test with gathers or scatters.
10484 (vect_analyze_data_ref_dependence): Re-order checks to deal with
10486 (vect_record_base_alignments): Do not record base alignment
10487 for gathers or scatters.
10488 (vect_compute_data_ref_alignment): Drop return value that is always
10489 true. Bail out early for gathers or scatters.
10490 (vect_enhance_data_refs_alignment): Bail out early for gathers
10492 (vect_find_same_alignment_drs): Likewise.
10493 (vect_analyze_data_refs_alignment): Remove dead code.
10494 (vect_slp_analyze_and_verify_node_alignment): Likewise.
10495 (vect_analyze_data_refs): For possible gathers or scatters do
10496 not create an alternate DR, just check their possible validity
10497 and mark them. Adjust DECL_NONALIASED handling to not rely
10498 on DR_BASE_ADDRESS.
10499 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Do not
10500 update inits of gathers or scatters.
10501 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern):
10502 Also copy gather/scatter flag to pattern vinfo.
10504 2018-06-20 Kelvin Nilsen <kelvin@gcc.gnu.org>
10506 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
10507 behavior of vec_packsu (vector unsigned long long, vector unsigned
10508 long long) to match behavior of vec_packs with same signature.
10510 2018-06-20 Chung-Lin Tang <cltang@codesourcery.com>
10511 Thomas Schwinge <thomas@codesourcery.com>
10512 Cesar Philippidis <cesar@codesourcery.com>
10514 * gimplify.c (gimplify_scan_omp_clauses): Add support for
10515 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
10516 (gimplify_adjust_omp_clauses): Likewise.
10517 (gimplify_oacc_declare_1): Add support for GOMP_MAP_RELEASE, remove
10518 support for GOMP_MAP_FORCE_{ALLOC,TO,FROM,TOFROM}.
10519 (gimplify_omp_target_update): Update handling of acc update and
10521 * omp-low.c (install_var_field): Remove unused parameter
10522 base_pointers_restrict.
10523 (scan_sharing_clauses): Remove base_pointers_restrict parameter.
10524 Update call to install_var_field. Handle OMP_CLAUSE_{IF_PRESENT,
10526 (omp_target_base_pointers_restrict_p): Delete.
10527 (scan_omp_target): Update call to scan_sharing_clauses.
10528 * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_{IF_PRESENT,
10530 * tree-nested.c (convert_nonlocal_omp_clauses): Handle
10531 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
10532 (convert_local_omp_clauses): Likewise.
10533 * tree-pretty-print.c (dump_omp_clause): Likewise.
10534 * tree.c (omp_clause_num_ops): Add entries for OMP_CLAUSE_{IF_PRESENT,
10536 (omp_clause_code_name): Likewise.
10538 2018-06-20 Jakub Jelinek <jakub@redhat.com>
10541 * var-tracking.c (use_narrower_mode_test): Check if shift amount can
10544 PR tree-optimization/86231
10545 * tree-vrp.c (union_ranges): For ( [ ) ] or ( )[ ] range and
10546 anti-range don't overwrite *vr0min before using it to compute *vr0max.
10548 2018-06-20 Tom de Vries <tdevries@suse.de>
10550 PR tree-optimization/86097
10551 * tree-ssa-loop-manip.c (canonicalize_loop_ivs): Also convert *nit to
10552 iv type if signedness of iv type is not the same as that of *nit.
10554 2018-06-20 Jakub Jelinek <jakub@redhat.com>
10556 * cfgrtl.c (rtl_verify_edges): Formatting fix. If bb->preds has any
10557 EDGE_EH edges, verify they are all EDGE_EH.
10559 2018-06-20 Maya Rashish <coypu@sdf.org>
10561 * ginclude/stddef.h: Limit #include <machine/ansi.h> to NetBSD.
10563 2018-06-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10565 * config/aarch64/aarch64-tuning-flags.def (no_ldp_stp_qregs): New.
10566 * config/aarch64/aarch64.c (xgene1_tunings): Add
10567 AARCH64_EXTRA_TUNE_NO_LDP_STP_QREGS to tune_flags.
10568 (aarch64_mode_valid_for_sched_fusion_p):
10569 Allow 16-byte modes.
10570 (aarch64_classify_address): Allow 16-byte modes for load_store_pair_p.
10571 * config/aarch64/aarch64-ldpstp.md: Add peepholes for LDP STP of
10573 * config/aarch64/aarch64-simd.md (load_pair<VQ:mode><VQ2:mode>):
10575 (vec_store_pair<VQ:mode><VQ2:mode>): Likewise.
10576 * config/aarch64/iterators.md (VQ2): New mode iterator.
10578 2018-06-20 Martin Liska <mliska@suse.cz>
10580 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
10581 Change default ratio from 10 to 8.
10583 2018-06-20 Martin Liska <mliska@suse.cz>
10585 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
10587 (bit_test_cluster::find_bit_tests): Likewise.
10588 (switch_decision_tree::analyze_switch_statement): Find clusters.
10589 * tree-switch-conversion.h (struct jump_table_cluster): Document
10592 2018-06-20 Martin Liska <mliska@suse.cz>
10594 * tree-switch-conversion.c (switch_conversion::collect):
10595 Record m_uniq property.
10596 (switch_conversion::expand): Bail out for special conditions.
10597 (group_cluster::~group_cluster): New.
10598 (group_cluster::group_cluster): Likewise.
10599 (group_cluster::dump): Likewise.
10600 (jump_table_cluster::emit): New.
10601 (switch_decision_tree::fix_phi_operands_for_edges): New.
10602 (struct case_node): Remove struct.
10603 (jump_table_cluster::can_be_handled): New.
10604 (case_values_threshold): Moved to header.
10605 (reset_out_edges_aux): Likewise.
10606 (jump_table_cluster::is_beneficial): New.
10607 (bit_test_cluster::can_be_handled): Likewise.
10608 (add_case_node): Remove.
10609 (bit_test_cluster::is_beneficial): New.
10610 (case_bit_test::cmp): New.
10611 (bit_test_cluster::emit): New.
10612 (expand_switch_as_decision_tree_p): Remove.
10613 (bit_test_cluster::hoist_edge_and_branch_if_true): New.
10614 (fix_phi_operands_for_edge): Likewise.
10615 (switch_decision_tree::analyze_switch_statement): New.
10616 (compute_cases_per_edge): Move ...
10617 (switch_decision_tree::compute_cases_per_edge): ... here.
10618 (try_switch_expansion): Likewise.
10619 (switch_decision_tree::try_switch_expansion): Likewise.
10620 (record_phi_operand_mapping): Likewise.
10621 (switch_decision_tree::record_phi_operand_mapping): Likewise.
10622 (emit_case_decision_tree): Likewise.
10623 (switch_decision_tree::emit): Likewise.
10624 (balance_case_nodes): Likewise.
10625 (switch_decision_tree::balance_case_nodes): Likewise.
10626 (dump_case_nodes): Likewise.
10627 (switch_decision_tree::dump_case_nodes): Likewise.
10628 (emit_jump): Likewise.
10629 (switch_decision_tree::emit_jump): Likewise.
10630 (emit_cmp_and_jump_insns): Likewise.
10631 (switch_decision_tree::emit_cmp_and_jump_insns): Likewise.
10632 (emit_case_nodes): Likewise.
10633 (switch_decision_tree::emit_case_nodes): Likewise.
10634 (conditional_probability): Remove.
10635 * tree-switch-conversion.h (enum cluster_type): New.
10637 (struct cluster): Likewise.
10638 (cluster::cluster): Likewise.
10639 (struct simple_cluster): Likewise.
10640 (simple_cluster::simple_cluster): Likewise.
10641 (struct group_cluster): Likewise.
10642 (struct jump_table_cluster): Likewise.
10643 (struct bit_test_cluster): Likewise.
10644 (struct min_cluster_item): Likewise.
10645 (struct case_tree_node): Likewise.
10646 (case_tree_node::case_tree_node): Likewise.
10647 (jump_table_cluster::case_values_threshold): Likewise.
10648 (struct case_bit_test): Likewise.
10649 (struct switch_decision_tree): Likewise.
10650 (struct switch_conversion): Likewise.
10651 (switch_decision_tree::reset_out_edges_aux): Likewise.
10653 2018-06-20 Martin Liska <mliska@suse.cz>
10655 * tree-switch-conversion.c (MAX_CASE_BIT_TESTS): Remove.
10656 (hoist_edge_and_branch_if_true): Likewise.
10657 (expand_switch_using_bit_tests_p): Likewise.
10658 (struct case_bit_test): Likewise.
10659 (case_bit_test_cmp): Likewise.
10660 (emit_case_bit_tests): Likewise.
10661 (switch_conversion::switch_conversion): New class.
10662 (struct switch_conv_info): Remove old struct.
10663 (collect_switch_conv_info): More to ...
10664 (switch_conversion::collect): ... this.
10665 (check_range): Likewise.
10666 (switch_conversion::check_range): Likewise.
10667 (check_all_empty_except_final): Likewise.
10668 (switch_conversion::check_all_empty_except_final): Likewise.
10669 (check_final_bb): Likewise.
10670 (switch_conversion::check_final_bb): Likewise.
10671 (create_temp_arrays): Likewise.
10672 (switch_conversion::create_temp_arrays): Likewise.
10673 (free_temp_arrays): Likewise.
10674 (gather_default_values): Likewise.
10675 (switch_conversion::gather_default_values): Likewise.
10676 (build_constructors): Likewise.
10677 (switch_conversion::build_constructors): Likewise.
10678 (constructor_contains_same_values_p): Likewise.
10679 (switch_conversion::contains_same_values_p): Likewise.
10680 (array_value_type): Likewise.
10681 (switch_conversion::array_value_type): Likewise.
10682 (build_one_array): Likewise.
10683 (switch_conversion::build_one_array): Likewise.
10684 (build_arrays): Likewise.
10685 (switch_conversion::build_arrays): Likewise.
10686 (gen_def_assigns): Likewise.
10687 (switch_conversion::gen_def_assigns): Likewise.
10688 (prune_bbs): Likewise.
10689 (switch_conversion::prune_bbs): Likewise.
10690 (fix_phi_nodes): Likewise.
10691 (switch_conversion::fix_phi_nodes): Likewise.
10692 (gen_inbound_check): Likewise.
10693 (switch_conversion::gen_inbound_check): Likewise.
10694 (process_switch): Use the newly created class.
10695 (switch_conversion::expand): New.
10696 (switch_conversion::~switch_conversion): New.
10697 * tree-switch-conversion.h: New file.
10699 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
10701 * tree-vectorizer.h (NUM_PATTERNS, vect_recog_func_ptr): Move to
10702 tree-vect-patterns.c.
10703 * tree-vect-patterns.c (vect_supportable_direct_optab_p): New function.
10704 (vect_recog_dot_prod_pattern): Use it. Remove the type_in argument.
10705 (vect_recog_sad_pattern): Likewise.
10706 (vect_recog_widen_sum_pattern): Likewise.
10707 (vect_recog_pow_pattern): Likewise. Check for a null vectype.
10708 (vect_recog_widen_shift_pattern): Remove the type_in argument.
10709 (vect_recog_rotate_pattern): Likewise.
10710 (vect_recog_mult_pattern): Likewise.
10711 (vect_recog_vector_vector_shift_pattern): Likewise.
10712 (vect_recog_divmod_pattern): Likewise.
10713 (vect_recog_mixed_size_cond_pattern): Likewise.
10714 (vect_recog_bool_pattern): Likewise.
10715 (vect_recog_mask_conversion_pattern): Likewise.
10716 (vect_try_gather_scatter_pattern): Likewise.
10717 (vect_recog_widen_mult_pattern): Likewise. Check for a null vectype.
10718 (vect_recog_over_widening_pattern): Likewise.
10719 (vect_recog_gather_scatter_pattern): Likewise.
10720 (vect_recog_func_ptr): Move from tree-vectorizer.h
10721 (vect_vect_recog_func_ptrs): Move further down the file.
10722 (vect_recog_func): Likewise. Remove the third argument.
10723 (NUM_PATTERNS): Define based on vect_vect_recog_func_ptrs.
10724 (vect_pattern_recog_1): Expect the pattern function to do any
10725 necessary target tests. Also expect it to provide a vector type.
10726 Remove the type_in handling.
10728 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
10730 * tree-vect-patterns.c (vect_pattern_detected): New function.
10731 (vect_recog_dot_prod_patternm, vect_recog_sad_pattern)
10732 (vect_recog_widen_mult_pattern, vect_recog_widen_sum_pattern)
10733 (vect_recog_over_widening_pattern, vect_recog_widen_shift_pattern
10734 (vect_recog_rotate_pattern, vect_recog_vector_vector_shift_pattern)
10735 (vect_recog_mult_pattern, vect_recog_divmod_pattern)
10736 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
10737 (vect_recog_mask_conversion_pattern)
10738 (vect_try_gather_scatter_pattern): Likewise.
10740 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
10742 * tree-vect-patterns.c (vect_get_internal_def): New function.
10743 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
10744 (vect_recog_vector_vector_shift_pattern, check_bool_pattern)
10745 (search_type_for_mask_1): Use it.
10747 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
10749 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
10750 redundant WIDEN_SUM_EXPR handling.
10751 (vect_recog_sad_pattern): Likewise.
10753 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
10755 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
10756 redundant check that the types of a PLUS_EXPR or MULT_EXPR agree.
10757 (vect_recog_sad_pattern): Likewise PLUS_EXPR, ABS_EXPR and MINUS_EXPR.
10758 (vect_recog_widen_mult_pattern): Likewise MULT_EXPR.
10759 (vect_recog_widen_sum_pattern): Likewise PLUS_EXPR.
10761 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
10763 * tree-vect-stmts.c (vectorizable_call): Make sure that we
10764 use the stmt_vec_info of the original bb statement for the
10765 new zero assignment, even if the call is part of a pattern.
10767 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
10769 * tree-vectorizer.h (_stmt_vec_info): Note above pattern_def_seq
10770 that the sequence is attached to the original statement rather
10771 than the pattern statement.
10772 * tree-vect-loop.c (vect_determine_vf_for_stmt): Take the
10773 PATTERN_DEF_SEQ from the original statement rather than
10774 the main pattern statement.
10775 * tree-vect-stmts.c (free_stmt_vec_info): Likewise.
10776 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Likewise.
10777 (vect_mark_pattern_stmts): Don't copy the PATTERN_DEF_SEQ.
10779 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
10781 * tree-vect-stmts.c (vect_analyze_stmt): Move the handling of pattern
10782 definition statements before the early exit for statements that aren't
10784 * tree-vect-loop.c (vect_transform_loop_stmt): New function,
10786 (vect_transform_loop): ...here. Process pattern definition
10787 statements without first checking whether the main pattern
10788 statement is live or relevant.
10790 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
10792 * tree-cfgcleanup.c (tree_forwarder_block_p): Do not return false at
10793 -O0 if the locus represent UNKNOWN_LOCATION but have different values.
10795 2018-06-19 Aaron Sawdey <acsawdey@linux.ibm.com>
10797 * config/rs6000/rs6000-string.c (select_block_compare_mode): Check
10798 TARGET_EFFICIENT_OVERLAPPING_UNALIGNED here instead of in caller.
10799 (do_and3, do_and3_mask, do_compb3, do_rotl3): New functions.
10800 (expand_block_compare): Change select_block_compare_mode call.
10801 (expand_strncmp_align_check): Use new functions, fix comment.
10802 (emit_final_str_compare_gpr): New function.
10803 (expand_strn_compare): Refactor and clean up code.
10804 * config/rs6000/vsx.md (vsx_mov<mode>_64bit): Remove *.
10806 2018-06-19 Tony Reix <tony.reix@atos.com>
10807 Damien Bergamini <damien.bergamini@atos.com>
10808 David Edelsohn <dje.gcc@gmail.com>
10810 * collect2.c (static_obj): New variable.
10811 (static_libs): New variable.
10812 (is_in_list): Uncomment declaration.
10813 (main): Track AIX libraries linked statically.
10814 (is_in_list): Uncomment definition.
10815 (scan_prog_file): Don't add AIX shared libraries initializer
10816 to constructor list if linking statically.
10818 2018-06-19 Max Filippov <jcmvbkbc@gmail.com>
10820 * config/xtensa/xtensa.md (UNSPEC_FRAME_BLOCKAGE): New unspec
10822 (allocate_stack, frame_blockage, *frame_blockage): New patterns.
10824 2018-06-19 Jan Hubicka <hubicka@ucw.cz>
10826 * tree.c (find_decls_types_r): Remove all non-VAR_DECLs from
10829 2018-06-19 Martin Liska <mliska@suse.cz>
10831 * config/i386/i386.c (ix86_can_inline_p): Do not use
10832 ipa_fn_summaries::get_create.
10833 * ipa-cp.c (ipcp_cloning_candidate_p): Replace get_create with
10835 (devirtualization_time_bonus): Likewise.
10836 (ipcp_propagate_stage): Likewise.
10837 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
10838 (edge_set_predicate): Likewise.
10839 (evaluate_conditions_for_known_args): Likewise.
10840 (evaluate_properties_for_edge): Likewise.
10841 (ipa_call_summary::reset): Tranform to ...
10842 (ipa_call_summary::~ipa_call_summary): ... this.
10843 (ipa_fn_summary::reset): Transform to ...
10844 (ipa_fn_summary::~ipa_fn_summary): ... this.
10845 (ipa_fn_summary_t::remove): Rename to ...
10846 (ipa_fn_summary_t::remove_callees): ... this.
10847 (ipa_fn_summary_t::duplicate): Use placement new
10848 instead of memory copy.
10849 (ipa_call_summary_t::duplicate): Likewise.
10850 (ipa_call_summary_t::remove): Remove.
10851 (dump_ipa_call_summary): Change get_create to get.
10852 (ipa_dump_fn_summary): Dump only when summary exists.
10853 (analyze_function_body): Use symbol_summary::get instead
10855 (compute_fn_summary): Likewise.
10856 (estimate_edge_devirt_benefit): Likewise.
10857 (estimate_edge_size_and_time): Likewise.
10858 (inline_update_callee_summaries): Likewise.
10859 (remap_edge_change_prob): Likewise.
10860 (remap_edge_summaries): Likewise.
10861 (ipa_merge_fn_summary_after_inlining): Likewise.
10862 (write_ipa_call_summary): Likewise.
10863 (ipa_fn_summary_write): Likewise.
10864 (ipa_free_fn_summary): Likewise.
10865 * ipa-fnsummary.h (struct GTY): Add new ctor and copy ctor.
10866 (struct ipa_call_summary): Likewise.
10867 * ipa-icf.c (sem_function::merge): Use symbol_summary::get instead
10869 * ipa-inline-analysis.c (do_estimate_edge_time): Likewise.
10870 (estimate_size_after_inlining): Likewise.
10871 (estimate_growth): Likewise.
10872 (growth_likely_positive): Likewise.
10873 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
10874 (inline_call): Likewise.
10875 * ipa-inline.c (caller_growth_limits): Likewise.
10876 (can_inline_edge_p): Likewise.
10877 (can_inline_edge_by_limits_p): Likewise.
10878 (compute_uninlined_call_time): Likewise.
10879 (compute_inlined_call_time): Likewise.
10880 (want_inline_small_function_p): Likewise.
10881 (edge_badness): Likewise.
10882 (update_caller_keys): Likewise.
10883 (update_callee_keys): Likewise.
10884 (inline_small_functions): Likewise.
10885 (inline_to_all_callers_1): Likewise.
10886 (dump_overall_stats): Likewise.
10887 (early_inline_small_functions): Likewise.
10888 (early_inliner): Likewise.
10889 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
10890 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
10891 * ipa-pure-const.c (malloc_candidate_p): Likewise.
10892 * ipa-split.c (execute_split_functions): Likewise.
10893 * symbol-summary.h: Likewise.
10894 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
10896 2018-06-19 Richard Biener <rguenther@suse.de>
10898 * tree-vectorizer.c (try_vectorize_loop_1): Split out of ...
10899 (vectorize_loops): ... here. Fix dbgcnt handling.
10900 (try_vectorize_loop): Wrap try_vectorize_loop_1.
10902 2018-06-19 Segher Boessenkool <segher@kernel.crashing.org>
10905 * config/rs6000/rs6000.md (rs6000_discover_homogeneous_aggregate): An
10906 ieee128 argument takes up only one (vector) register, not two (floating
10909 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
10911 * gimplify.c (gimplify_init_constructor): Really never clear for an
10912 incomplete constructor if CONSTRUCTOR_NO_CLEARING is set.
10914 2018-06-19 Richard Biener <rguenther@suse.de>
10916 PR tree-optimization/86179
10917 * tree-vect-patterns.c (vect_pattern_recog_1): Clean up
10918 after failed recognition.
10920 2018-06-18 Martin Sebor <msebor@redhat.com>
10922 PR middle-end/85602
10923 * calls.c (maybe_warn_nonstring_arg): Handle strncat.
10924 * tree-ssa-strlen.c (is_strlen_related_p): Make extern.
10925 Handle integer subtraction.
10926 (maybe_diag_stxncpy_trunc): Handle nonstring source arguments.
10927 * tree-ssa-strlen.h (is_strlen_related_p): Declare.
10929 2018-06-18 David Malcolm <dmalcolm@redhat.com>
10931 * config/frv/frv-protos.h (frv_ifcvt_modify_insn): Strengthen 3rd
10932 param from rtx to rtx_insn *.
10933 * config/frv/frv.c (frv_ifcvt_add_insn): Likewise for "insn"
10935 (frv_ifcvt_modify_insn): Likwise.
10936 (frv_ifcvt_modify_final): Likwise for local "existing_insn",
10937 adding an as_a <rtx_insn *> cast. Likewise for local "insn".
10938 * config/mips/mips.c (r10k_insert_cache_barriers): Add an
10939 as_a <rtx_insn *> cast to local "unprotected_region" once
10940 it's been established that it's not NULL or pc_rtx.
10941 * config/nds32/nds32-relax-opt.c (nds32_group_insns): Strengthen
10942 param "sethi" from rtx to rtx_insn *.
10943 (nds32_group_float_insns): Likewise for param "insn".
10944 * config/vax/vax-protos.h (vax_output_int_add): Likewise for 1st
10946 (vax_output_int_subtract): Likewise.
10947 * config/vax/vax.c (vax_output_int_add): Likewise for param
10949 (vax_output_int_subtract): Likewise.
10950 * emit-rtl.c (set_insn_deleted): Likewise, removing cast.
10951 (emit_pattern_after): Likewise for param "after".
10952 (emit_insn_after): Likewise.
10953 (emit_jump_insn_after): Likewise.
10954 (emit_call_insn_after): Likewise.
10955 (emit_debug_insn_after): Likewise.
10956 (emit_pattern_before): Likewise for param "before".
10957 (emit_insn_before): Likewise.
10958 (emit_jump_insn_before): Likewise.
10959 * final.c (get_insn_template): Likewise for param "insn", removing
10961 * output.h (get_insn_template): Likewise for 2nd param.
10962 * rtl.h (emit_insn_before): Likewise.
10963 (emit_jump_insn_before): Likewise.
10964 (emit_debug_insn_before_noloc): Likewise.
10965 (emit_insn_after): Likewise.
10966 (emit_jump_insn_after): Likewise.
10967 (emit_call_insn_after): Likewise.
10968 (emit_debug_insn_after): Likewise.
10969 (set_insn_deleted): Likewise for param.
10971 2018-06-18 Michael Meissner <meissner@linux.ibm.com>
10974 * config/rs6000/rs6000-modes.def (toplevel): Rework the 128-bit
10975 floating point modes, so that IFmode is numerically greater than
10976 TFmode, which is greater than KFmode using FRACTIONAL_FLOAT_MODE
10977 to declare the ordering. This prevents IFmode from being
10978 converted to TFmode when long double is IEEE 128-bit on an ISA 3.0
10979 machine. Include rs6000-modes.h to share the fractional values
10980 between genmodes* and the rest of the compiler.
10981 (IFmode): Likewise.
10982 (KFmode): Likewise.
10983 (TFmode): Likewise.
10984 * config/rs6000/rs6000-modes.h: New file.
10985 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Change the
10986 meaning of rs6000_long_double_size so that 126..128 selects an
10987 appropriate 128-bit floating point type.
10988 (rs6000_option_override_internal): Likewise.
10989 * config/rs6000/rs6000.h (toplevel): Include rs6000-modes.h.
10990 (TARGET_LONG_DOUBLE_128): Change the meaning of
10991 rs6000_long_double_size so that 126..128 selects an appropriate
10992 128-bit floating point type.
10993 (LONG_DOUBLE_TYPE_SIZE): Update comment.
10994 * config/rs6000/rs6000.md (trunciftf2): Correct the modes of the
10995 source and destination to match the standard usage.
10996 (truncifkf2): Likewise.
10997 (copysign<mode>3, IEEE iterator): Rework copysign of float128 on
10998 ISA 2.07 to use an explicit clobber, instead of passing in a
11000 (copysign<mode>3_soft): Likewise.
11002 2018-06-18 David Malcolm <dmalcolm@redhat.com>
11004 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
11005 Replace dump_printf_loc call with DUMP_VECT_SCOPE.
11006 (vect_slp_analyze_instance_dependence): Likewise.
11007 (vect_enhance_data_refs_alignment): Likewise.
11008 (vect_analyze_data_refs_alignment): Likewise.
11009 (vect_slp_analyze_and_verify_instance_alignment
11010 (vect_analyze_data_ref_accesses): Likewise.
11011 (vect_prune_runtime_alias_test_list): Likewise.
11012 (vect_analyze_data_refs): Likewise.
11013 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Likewise.
11014 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
11015 (vect_analyze_scalar_cycles_1): Likewise.
11016 (vect_get_loop_niters): Likewise.
11017 (vect_analyze_loop_form_1): Likewise.
11018 (vect_update_vf_for_slp): Likewise.
11019 (vect_analyze_loop_operations): Likewise.
11020 (vect_analyze_loop): Likewise.
11021 (vectorizable_induction): Likewise.
11022 (vect_transform_loop): Likewise.
11023 * tree-vect-patterns.c (vect_pattern_recog): Likewise.
11024 * tree-vect-slp.c (vect_analyze_slp): Likewise.
11025 (vect_make_slp_decision): Likewise.
11026 (vect_detect_hybrid_slp): Likewise.
11027 (vect_slp_analyze_operations): Likewise.
11028 (vect_slp_bb): Likewise.
11029 * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Likewise.
11030 (vectorizable_bswap): Likewise.
11031 (vectorizable_call): Likewise.
11032 (vectorizable_simd_clone_call): Likewise.
11033 (vectorizable_conversion): Likewise.
11034 (vectorizable_assignment): Likewise.
11035 (vectorizable_shift): Likewise.
11036 (vectorizable_operation): Likewise.
11037 * tree-vectorizer.h (DUMP_VECT_SCOPE): New macro.
11039 2018-06-18 Martin Sebor <msebor@redhat.com>
11041 PR tree-optimization/81384
11042 * builtin-types.def (BT_FN_SIZE_CONST_STRING_SIZE): New.
11043 * builtins.c (expand_builtin_strnlen): New function.
11044 (expand_builtin): Call it.
11045 (fold_builtin_n): Avoid setting TREE_NO_WARNING.
11046 * builtins.def (BUILT_IN_STRNLEN): New.
11047 * calls.c (maybe_warn_nonstring_arg): Handle BUILT_IN_STRNLEN.
11048 Warn for bounds in excess of maximum object size.
11049 * tree-ssa-strlen.c (maybe_set_strlen_range): Return tree representing
11050 single-value ranges. Handle strnlen.
11051 (handle_builtin_strlen): Handle strnlen.
11052 (strlen_check_and_optimize_stmt): Same.
11053 * doc/extend.texi (Other Builtins): Document strnlen.
11055 2018-06-18 Maya Rashish <coypu@sdf.org>
11057 * config/alpha/openbsd.h (TARGET_DEFAULT): Define.
11058 (LINK_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): Likewise.
11059 (INTMAX_TYPE, UINTMAX_TYPE, WINT_TYPE): Likewise.
11061 * config/alpha/elf.h (STARTFILE_SPEC, ENDFILE_SPEC): Move from
11063 * config/alpha/linux.h (STARTFILE_SPEC, ENDFILE_SPEC): Here.
11065 2018-06-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
11067 * tree.c (escaped_string::escape): Replace cast to char * by
11068 const_cast<char *> (unescaped).
11070 2018-06-18 Nick Clifton <nickc@redhat.com>
11073 * tree.c (escaped_string): New class. Converts an unescaped
11074 string into its escaped equivalent.
11075 (warn_deprecated_use): Use the new class to convert the
11076 deprecation message, if present.
11077 (test_escaped_strings): New self test.
11078 (test_c_tests): Add test_escaped_strings.
11079 * doc/extend.texi (deprecated): Add a note that the
11080 deprecation message is affected by the -fmessage-length
11081 option, and that control characters will be escaped.
11082 (#pragma GCC error): Document this pragma.
11083 (#pragma GCC warning): Likewise.
11084 * doc/invoke.texi (-fmessage-length): Document this option's
11085 effect on the #warning and #error preprocessor directives and
11086 the deprecated attribute.
11088 2018-06-18 Eric Botcazou <ebotcazou@adacore.com>
11090 * tree.c (decl_value_expr_lookup): Revert latest change.
11091 (decl_value_expr_insert): Likewise.
11093 2018-06-17 Eric Botcazou <ebotcazou@adacore.com>
11095 * gimplify.c (nonlocal_vlas): Delete.
11096 (nonlocal_vla_vars): Likewise.
11097 (gimplify_var_or_parm_decl): Do not add debug VAR_DECLs for non-local
11099 (gimplify_body): Do not create and destroy nonlocal_vlas.
11100 * tree-nested.c: Include diagnostic.h.
11101 (use_pointer_in_frame): Tweak.
11102 (lookup_field_for_decl): Add assertion and declare the transformation.
11103 (convert_nonlocal_reference_op) <PARM_DECL>: Rework and issue an
11104 internal error when the reference is in a wrong context. Do not
11105 create a debug decl by default.
11106 (note_nonlocal_block_vlas): Delete.
11107 (convert_nonlocal_reference_stmt) <GIMPLE_BIND>: Do not call it.
11108 (convert_local_reference_op) <PARM_DECL>: Skip the frame decl. Do not
11109 create a debug decl by default.
11110 (convert_gimple_call) <GIMPLE_CALL>: Issue an internal error when the
11111 call is in a wrong context.
11112 (fixup_vla_decls): New function.
11113 (finalize_nesting_tree_1): Adjust comment. Call fixup_vla_decls if no
11114 debug variables were created.
11115 * tree.c (decl_value_expr_lookup): Add checking assertion.
11116 (decl_value_expr_insert): Likewise.
11118 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
11120 PR middle-end/82479
11121 * ipa-fnsummary.c (will_be_nonconstant_expr_predicate): Handle CALL_EXPR.
11122 * tree-scalar-evolution.c (interpret_expr): Likewise.
11123 (expression_expensive_p): Likewise.
11124 * tree-ssa-loop-ivopts.c (contains_abnormal_ssa_name_p): Likewise.
11125 * tree-ssa-loop-niter.c (number_of_iterations_popcount): New.
11126 (number_of_iterations_exit_assumptions): Use number_of_iterations_popcount.
11127 (ssa_defined_by_minus_one_stmt_p): New.
11129 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
11131 PR middle-end/64946
11132 * cfgexpand.c (expand_debug_expr): Hande ABSU_EXPR.
11133 * config/i386/i386.c (ix86_add_stmt_cost): Likewise.
11134 * dojump.c (do_jump): Likewise.
11135 * expr.c (expand_expr_real_2): Check operand type's sign.
11136 * fold-const.c (const_unop): Handle ABSU_EXPR.
11137 (fold_abs_const): Likewise.
11138 * gimple-pretty-print.c (dump_unary_rhs): Likewise.
11139 * gimple-ssa-backprop.c (backprop::process_assign_use): Likesie.
11140 (strip_sign_op_1): Likesise.
11141 * match.pd: Add new pattern to generate ABSU_EXPR.
11142 * optabs-tree.c (optab_for_tree_code): Handle ABSU_EXPR.
11143 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
11144 * tree-eh.c (operation_could_trap_helper_p): Likewise.
11145 * tree-inline.c (estimate_operator_cost): Likewise.
11146 * tree-pretty-print.c (dump_generic_node): Likewise.
11147 * tree-vect-patterns.c (vect_recog_sad_pattern): Likewise.
11148 * tree.def (ABSU_EXPR): New.
11150 2018-06-16 Jakub Jelinek <jakub@redhat.com>
11152 PR middle-end/86095
11153 * common.opt (Wunsafe-loop-optimizations): Add Ignore, remove Var,
11154 documented as preserved for backward compatibility only.
11155 * doc/invoke.texi: Remove -Wunsafe-loop-optimizations documentation.
11157 PR rtl-optimization/86108
11158 * bb-reorder.c (create_forwarder_block): Renamed to ...
11159 (create_eh_forwarder_block): ... this. Split OLD_BB after labels and
11160 jump from new landing pad to the second part.
11161 (sjlj_fix_up_crossing_landing_pad, dw2_fix_up_crossing_landing_pad):
11164 2018-06-15 Jakub Jelinek <jakub@redhat.com>
11166 PR middle-end/85878
11167 * expr.c (expand_assignment): Remove now redundant COMPLEX_MODE_P
11168 check from first store_expr, use to_mode instead of GET_MODE (to_rtx).
11169 Only call store_expr for halves if the mode is the same.
11171 PR middle-end/86123
11172 * match.pd ((X / Y) == 0 -> X < Y): Don't transform complex divisions.
11173 Fix up comment formatting.
11175 2018-06-15 Bernd Edlinger <bernd.edlinger@hotmail.de>
11177 * typed-splay-tree.h (typed_splay_tree::remove): New function.
11178 (typed_splay_tree::closure,
11179 typed_splay_tree::inner_foreach_fn, typed_splay_tree::m_inner): Deleted.
11180 (typed_splay_tree::typed_splay_tree,
11181 typed_splay_tree::operator =): Declared private.
11182 (typed_splay_tree::splay_tree_key, typed_splay_tree::splay_tree_value,
11183 typed_splay_tree::splay_tree_node_s, typed_splay_tree::KDEL,
11184 typed_splay_tree::VDEL, typed_splay_tree::splay_tree_delete_helper,
11185 typed_splay_tree::rotate_left, typed_splay_tree::rotate_right,
11186 typed_splay_tree::splay_tree_splay,
11187 typed_splay_tree::splay_tree_foreach_helper,
11188 typed_splay_tree::splay_tree_insert,
11189 typed_splay_tree::splay_tree_remove,
11190 typed_splay_tree::splay_tree_lookup,
11191 typed_splay_tree::splay_tree_predecessor,
11192 typed_splay_tree::splay_tree_successor,
11193 typed_splay_tree::splay_tree_min,
11194 typed_splay_tree::splay_tree_max): Took over from splay-tree.c/.h.
11195 (typed_splay_tree::root, typed_splay_tree::comp,
11196 typed_splay_tree::delete_key,
11197 typed_splay_tree::delete_value): New data members.
11198 * typed-splay-tree.c (selftest::test_str_to_int): Add a test for
11199 typed_splay_tree::remove.
11201 2018-06-15 Matthew Fortune <matthew.fortune@mips.com>
11203 * config/mips/mips.h (ASM_SPEC): Pass through -mcrc, -mno-crc,
11204 -mginv and -mno-ginv to the assembler.
11205 * config/mips/mips.opt (-mcrc): New option.
11206 (-mginv): Likewise.
11207 * doc/invoke.text (-mcrc): Document.
11208 (-mginv): Likewise.
11210 2018-06-15 Nick Clifton <nickc@redhat.com>
11213 * tree.c (escaped_string): New class. Converts an unescaped
11214 string into its escaped equivalent.
11215 (warn_deprecated_use): Use the new class to convert the
11216 deprecation message, if present.
11217 (test_escaped_strings): New self test.
11218 (test_c_tests): Add test_escaped_strings.
11219 * doc/extend.texi (deprecated): Add a note that the
11220 deprecation message is affected by the -fmessage-length
11221 option, and that control characters will be escaped.
11222 (#pragma GCC error): Document this pragma.
11223 (#pragma GCC warning): Likewise.
11224 * doc/invoke.texi (-fmessage-length): Document this option's
11225 effect on the #warning and #error preprocessor directives and
11226 the deprecated attribute.
11228 2018-06-15 Richard Biener <rguenther@suse.de>
11230 * tree-vect-slp.c (vect_slp_bb): Dump MSG_OPTIMIZED_LOCATIONS
11231 here, also noting vector size used.
11232 * tree-vectorizer.c (vectorize_loops): Adjust. Note vector
11233 size used in MSG_OPTIMIZED_LOCATIONS dump.
11234 (pass_slp_vectorize::execute): Adjust.
11236 2018-06-15 Claudiu Zissulescu <claziss@synopsys.com>
11239 * config/arc/arc.c (arc_return_address_register): Fix
11242 2018-06-15 Richard Biener <rguenther@suse.de>
11244 PR middle-end/86159
11245 * tree-cfg.c (gimplify_build3): Do not strip sign conversions,
11246 leave useless conversion stripping to force_gimple_operand_gsi.
11247 (gimplify_build2): Likewise.
11248 (gimplify_build1): Likewise.
11250 2018-06-15 Richard Biener <rguenther@suse.de>
11252 PR middle-end/86076
11253 * tree-cfg.c (move_stmt_op): unshare invariant addresses
11254 before adjusting their block.
11256 2018-06-15 Sebastian Huber <sebastian.huber@embedded-brains.de>
11258 * config.gcc (riscv*-*-elf* | riscv*-*-rtems*): Use custom
11259 multilibs for *-*-rtems*.
11260 * config/riscv/t-rtems: New file.
11262 2018-06-14 Jakub Jelinek <jakub@redhat.com>
11264 PR middle-end/86122
11265 * match.pd ((A +- CST1) +- CST2): Punt if last resort
11266 unsigned_type_for returns NULL.
11269 * lower-subreg.c (find_decomposable_subregs): Don't decompose float
11270 subregs of multi-word pseudos unless the float mode has word size.
11272 2018-06-14 Richard Biener <rguenther@suse.de>
11274 PR middle-end/86139
11275 * tree-vect-generic.c (build_word_mode_vector_type): Remove
11276 duplicate and harmful type_hash_canon.
11277 * tree.c (type_hash_canon): Assert we didn't find ourselves.
11279 2018-06-14 Richard Biener <rguenther@suse.de>
11282 * tree-ssa-struct-alias.c (create_variable_info_for): Handle
11285 2018-06-14 Sebastian Huber <sebastian.huber@embedded-brains.de>
11287 * config/rtems.h (STDINT_LONG32): Define.
11289 2018-06-13 Matthew Fortune <matthew.fortune@mips.com>
11290 Prachi Godbole <prachi.godbole@imgtec.com>
11292 * config/mips/mips-cpus.def: Define P6600.
11293 * config/mips/mips-tables.opt: Regenerate.
11294 * config/mips/mips.c (mips_ucbranch_type): New enum.
11295 (mips_rtx_cost_data): Add support for P6600.
11296 (mips_issue_rate): Likewise.
11297 (mips_multipass_dfa_lookahead): Likewise.
11298 (mips_avoid_hazard): Likewise.
11299 (mips_reorg_process_insns): Likewise.
11300 (mips_classify_branch_p6600): New function.
11301 * config/mips/mips.h (TUNE_P6600): New define.
11302 (MIPS_ISA_LEVEL_SPEC): Infer mips64r6 from p6600.
11303 (ENABLE_LD_ST_PAIRS): Enable load/store bonding for p6600.
11304 * config/mips/mips.md: Include p6600.md.
11305 (processor): Add p6600.
11306 * config/mips/p6600.md: New file.
11307 * doc/invoke.texi: Add p6600 to supported architectures.
11309 2018-06-13 Martin Sebor <msebor@redhat.com>
11311 PR tree-optimization/86114
11312 * gimple-fold.c (gimple_fold_builtin_strlen): Only handle LHS
11314 * tree-ssa-strlen.c (maybe_set_strlen_range): Same.
11316 2018-06-13 Richard Biener <rguenther@suse.de>
11318 * tree-vect-patterns.c (vect_recog_vector_vector_shift_pattern):
11319 Properly set vector type of the intermediate stmt.
11320 * tree-vect-stmts.c (vectorizable_operation): The destination
11321 var always has vectype_out type.
11323 2018-06-13 Jeff Law <law@redhat.com>
11325 * config/rl78/rl78.c (move_elim_pass): Use TDF_NONE rather than
11326 integer 0 for argument to print_rtl_with_bb.
11327 (rl78_reorg): Likewise.
11329 2018-06-13 David Malcolm <dmalcolm@redhat.com>
11331 * config/arc/arc.c (hwloop_optimize): Strengthen local "end_label"
11332 from rtx to rtx_insn *.
11333 * config/bfin/bfin.c (hwloop_optimize): Likewise for local
11335 (add_sched_insns_for_speculation): Likewise for local "target",
11336 converting usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
11337 * config/c6x/c6x.c (reorg_split_calls): Strengthen param "call_labels"
11338 from rtx_insn ** to rtx_code_label **.
11339 (reorg_emit_nops): Likewise.
11340 (c6x_reorg): Likewise for local "call_labels".
11341 * config/sh/sh-protos.h (get_dest_uid): Strengthen 1st param from
11343 * config/sh/sh.c (dump_table): Strengthen local "lab" from rtx to
11344 rtx_code_label *, adding safe_as_a <rtx_code_label *> casts to
11345 the loops over LABEL_REFS.
11346 (fixup_addr_diff_vecs): Add as_a <rtx_insn *> to usage of
11348 (barrier_align): Convert usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
11349 (get_dest_uid): Strengthen param "label" from rtx to rtx_insn *.
11350 (split_branches): Strengthen local "olabel" from rtx to
11351 rtx_insn *, adding a safe_as_a cast.
11352 * emit-rtl.c (next_real_insn): Strengthen param from "rtx"
11354 (add_insn_after): Likewise for first two params.
11355 (add_insn_before): Likewise.
11356 (remove_insn): Likewise for param.
11357 (emit_pattern_before_noloc): Likewise for second and third params.
11358 (emit_jump_insn_before_noloc): Convert NULL_RTX to NULL.
11359 (emit_call_insn_before_noloc): Likewise.
11360 (emit_debug_insn_before_noloc): Strengthen "before" param from "rtx"
11362 (emit_barrier_before): Likewise.
11363 (emit_label_before): Strengthen "label" param from "rtx" to
11364 "rtx_code_label *". Strengthen "before" param from "rtx" to
11366 (emit_insn_after_1): Strengthen "after" param from "rtx" to
11368 (emit_pattern_after_noloc): Likewise.
11369 (emit_insn_after_noloc): Likewise.
11370 (emit_jump_insn_after_noloc): Likewise.
11371 (emit_call_insn_after_noloc): Likewise.
11372 (emit_debug_insn_after_noloc): Likewise.
11373 (emit_barrier_after): Likewise.
11374 (emit_label_after): Likewise for both params.
11375 (emit_pattern_after_setloc): Likewise for "after" param. Convert
11376 "loc" param from "int" to "location_t".
11377 (emit_insn_after_setloc): Likewise.
11378 (emit_jump_insn_after_setloc): Likewise.
11379 (emit_call_insn_after_setloc): Likewise.
11380 (emit_debug_insn_after_setloc): Likewise.
11381 (emit_pattern_before_setloc): Likewise for "before" param. Convert
11382 "loc" param from "int" to "location_t".
11383 (emit_pattern_before): Convert NULL_RTX to NULL.
11384 (emit_insn_before_setloc): Convert "loc" param from "int" to
11386 (emit_jump_insn_before_setloc): Likewise.
11387 (emit_call_insn_before_setloc): Likewise.
11388 (emit_debug_insn_before_setloc): Strengthen "before" param from rtx to
11389 rtx_insn *. Convert "loc" param from "int" to "location_t".
11390 * rtl.h (emit_insn_before_setloc, emit_jump_insn_before_setloc,
11391 emit_call_insn_before_setloc, emit_debug_insn_before_setloc):
11392 Convert 3rd param from "int" to "location_t".
11393 (emit_barrier_before, emit_barrier_after, next_real_insn):
11394 Strengthen param from rtx to rtx_insn *.
11395 (emit_label_before): Strengthen 1st param from "rtx" to
11396 "rtx_code_label *". Strengthen 2nd param from "rtx" to
11398 (emit_insn_after_noloc, emit_jump_insn_after_noloc,
11399 emit_call_insn_after_noloc, emit_debug_insn_after_noloc):
11400 Strengthen 2nd param from "rtx" to "rtx_insn *".
11401 (emit_insn_after_setloc, emit_jump_insn_after_setloc)
11402 emit_call_insn_after_setloc, emit_debug_insn_after_setloc):
11403 Likewise. Convert 3rd param from "int" to "location_t".
11404 (emit_label_after): Strengthen 1st param from "rtx" to
11405 "rtx_code_label *".
11406 (next_real_insn, remove_insn): Strengthen param from "rtx" to
11408 (add_insn_before, add_insn_after): Strengthen 1st and 2nd params
11409 from "rtx" to "rtx_insn *".
11411 2018-06-13 Jan Hubicka <hubicka@gcc.gnu.org>
11413 * cgraph.c (cgraph_node::get_untransformed_body): Dump function
11414 bodies streamed in with -Q.
11415 * dumpfile.c (dump_files): Add lto-stream-out dump file.
11416 * dumpfile.h (tree_dump_index): Add lto_stream_out.
11417 * gimple-streamer-out.c: Include gimple-pretty-print.h
11418 (output_bb): Dump stmts streamed.
11419 * lto-section-out.c: Include print-tree.h
11420 (lto_begin_section): Dump sections created.
11421 (lto_output_decl_index): Dump decl encoded.
11422 * lto-streamer-out.c: Include print-tree.h
11423 (create_output_block): Dump output block created.
11424 (DFS::DFS_write_tree_body): Dump DFS SCCs streamed.
11425 (output_function): Dump function output.
11426 (output_constructor): Dump constructor streamed.
11427 (write_global_stream): Output indexes encoded.
11428 (produce_asm_for_decls): Dump streams encoded.
11429 * lto-streamer.c (streamer_dump_file): New global var.
11430 * lto-streamer.h (streamer_dump_file): Declare.
11431 * passes.c (ipa_write_summaries): Initialize streamer dump.
11432 * varpool.c (varpool_node::get_constructor): Dump constructors streamed
11435 2018-06-13 Eric Botcazou <ebotcazou@adacore.com>
11438 * config/i386/winnt.c (i386_pe_seh_cold_init): Do not emit negative
11439 offsets for register save directives. Emit a second batch of save
11440 directives, if need be, when the function accesses prior frames.
11442 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
11444 * config/arc/fpu.md (fmasf4): Force operand to register.
11445 (fnmasf4): Likewise.
11447 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
11449 * config/arc/arc-protos.h (arc_pad_return): Remove.
11450 * config/arc/arc.c (machine_function): Remove force_short_suffix
11452 (arc_print_operand): Adjust printing of '&'.
11453 (arc_verify_short): Remove conditional printing of short suffix.
11454 (arc_final_prescan_insn): Remove reference to size_reason.
11455 (pad_return): New function.
11456 (arc_reorg): Call pad_return.
11457 (arc_pad_return): Remove.
11458 (arc_init_machine_status): Remove reference to force_short_suffix.
11459 * config/arc/arc.md (vunspec): Add VUNSPEC_ARC_BLOCKAGE.
11460 (attr length): When attribute iscompact is true force to 2
11461 regardless; in the case of maybe check if we want to force the
11462 instruction to have 4 bytes length.
11463 (nopv): Change it to generate 4 byte long nop as well.
11464 (blockage): New pattern.
11465 (simple_return): Remove call to arc_pad_return.
11466 (p_return_i): Likewise.
11468 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
11470 * config/arc/elf.h (LINK_GCC_C_SEQUENCE_SPEC): Define.
11472 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
11474 * config/arc/builtins.def (SYNC): SYNC instruction is valid on all
11477 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
11479 * config/arc/arc.c (atomic_exchangesi): EX instruction is default
11480 for ARC700 and ARCv2.
11482 2018-06-13 Chenghua Xu <paul.hua.gm@gmail.com>
11485 * config/mips/loongson.md (vec_setv4hi): Gen_lowpart for
11486 operands[2] instead of operands[1].
11489 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
11491 * lra-constraints.c (simplify_operand_subreg): In the paradoxical
11492 case, check whether the outer register overlaps an unallocatable
11493 register, not just whether it fits the required class.
11495 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
11497 * poly-int.h (can_div_trunc_p): Add new overload in which all values
11499 * alias.c (get_addr): Extend CONST_INT handling to poly_int_rtx_p.
11500 (memrefs_conflict_p): Likewise.
11501 (init_alias_analysis): Likewise.
11502 * cfgexpand.c (expand_debug_expr): Likewise.
11503 * combine.c (combine_simplify_rtx, force_int_to_mode): Likewise.
11504 * cse.c (fold_rtx): Likewise.
11505 * explow.c (adjust_stack, anti_adjust_stack): Likewise.
11506 * expr.c (emit_block_move_hints): Likewise.
11507 (clear_storage_hints, push_block, emit_push_insn): Likewise.
11508 (store_expr_with_bounds, reduce_to_bit_field_precision): Likewise.
11509 (emit_group_load_1): Use rtx_to_poly_int64 for group offsets.
11510 (emit_group_store): Likewise.
11511 (find_args_size_adjust): Use strip_offset. Use rtx_to_poly_int64
11512 to read the PRE/POST_MODIFY increment.
11513 * calls.c (store_one_arg): Use strip_offset.
11514 * rtlanal.c (rtx_addr_can_trap_p_1): Extend CONST_INT handling to
11516 (set_noop_p): Use rtx_to_poly_int64 for the elements selected
11518 * simplify-rtx.c (avoid_constant_pool_reference): Use strip_offset.
11519 (simplify_binary_operation_1): Extend CONST_INT handling to
11521 * var-tracking.c (compute_cfa_pointer): Take a poly_int64 rather
11522 than a HOST_WIDE_INT.
11523 (hard_frame_pointer_adjustment): Change from HOST_WIDE_INT to
11525 (adjust_mems, add_stores): Update accodingly.
11526 (vt_canonicalize_addr): Track polynomial offsets.
11527 (emit_note_insn_var_location): Likewise.
11528 (vt_add_function_parameter): Likewise.
11529 (vt_initialize): Likewise.
11531 2018-06-12 Jeff Law <law@redhat.com>
11533 * config.gcc (alpha*-*-freebsd*): Remove.
11534 * config/alpha/freebsd.h: Remove.
11536 2018-06-12 David Malcolm <dmalcolm@redhat.com>
11539 * spellcheck-tree.c (levenshtein_distance): Rename to...
11540 (get_edit_distance): ...this, and update for underlying renaming.
11541 * spellcheck-tree.h (levenshtein_distance): Rename to...
11542 (get_edit_distance): ...this.
11543 * spellcheck.c (levenshtein_distance): Rename to...
11544 (get_edit_distance): ...this. Convert from Levenshtein distance
11545 to Damerau-Levenshtein distance by supporting transpositions of
11546 adjacent characters. Rename "v1" to "v_next" and "v0" to
11548 (selftest::levenshtein_distance_unit_test_oneway): Rename to...
11549 (selftest::test_edit_distance_unit_test_oneway): ...this, and
11550 update for underlying renaming.
11551 (selftest::levenshtein_distance_unit_test): Rename to...
11552 (selftest::test_get_edit_distance_unit): ...this, and update for
11553 underlying renaming.
11554 (selftest::test_find_closest_string): Add example from PR 69968
11555 where transposition helps
11556 (selftest::test_metric_conditions): Update for renaming.
11557 (selftest::test_metric_conditions): Likewise.
11558 (selftest::spellcheck_c_tests): Likewise.
11559 * spellcheck.h (levenshtein_distance): Rename both overloads to...
11560 (get_edit_distance): ...this.
11561 (best_match::consider): Update for renaming.
11563 2018-06-12 Martin Sebor <msebor@redhat.com>
11565 PR tree-optimization/85259
11566 * builtins.c (compute_objsize): Handle constant offsets.
11567 * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Return
11568 true iff a warning has been issued.
11569 * gimple.h (gimple_nonartificial_location): New function.
11570 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Call
11571 gimple_nonartificial_location and handle -Wno-system-headers.
11572 (handle_builtin_stxncpy): Same.
11574 2018-06-12 Martin Sebor <msebor@redhat.com>
11577 * fold-const.c (operand_equal_p): Handle SAVE_EXPR.
11579 2018-06-12 Will Schmidt <will_schmidt@vnet.ibm.com>
11581 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
11582 BUILTIN_VEC_XST entries for pointer to double and long long.
11584 2018-06-12 H.J. Lu <hongjiu.lu@intel.com>
11587 * config/i386/gnu-user.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
11589 * config/i386/gnu-user64.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
11592 2018-06-12 Martin Liska <mliska@suse.cz>
11594 * doc/options.texi: Document IntegerRange.
11596 2018-06-12 Martin Liska <mliska@suse.cz>
11598 * config/i386/i386.opt: Make MPX-related options as Deprecated.
11599 * opt-functions.awk: Handle Deprecated flag.
11600 * opts-common.c (decode_cmdline_option): Handle cl_deprecated
11602 (read_cmdline_option): Report warning for a deprecated option.
11603 * opts.h (struct cl_option): Add new field cl_deprecated.
11604 (CL_ERR_DEPRECATED): New.
11606 2018-06-12 Martin Liska <mliska@suse.cz>
11608 * doc/options.texi: Document Deprecated option flag.
11610 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
11612 * config/arc/arc-arch.h (arc_extras): New enum.
11613 (arc_cpu_t):Add field extra.
11614 (arc_cpu_types): Consider the extras.
11615 * config/arc/arc-cpus.def: Add extras info.
11616 * config/arc/arc-opts.h (processor_type): Consider extra field.
11617 * config/arc/arc.c (arc_override_options): Handle extra field.
11619 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
11621 * config/arc/arc-arch.h: Update ARC_OPTX macro.
11622 * config/arc/arc-options.def (ARC_OPTX): Introduce a new doc
11624 * config/arc/arc.c (arc_init): Update pic warning.
11625 (irq_range): Update irq range parsing warnings.
11626 (arc_override_options): Update various warning messages.
11627 (arc_handle_aux_attribute): Likewise.
11629 2018-06-12 Robert Suchanek <robert.suchanek@mips.com>
11631 * config/mips/i6400.md (i6400_fpu_fadd): Remove frint.
11633 2018-06-12 Jozef Lawrynowicz <jozef.l@mittosystems.com>
11635 * doc/sourcebuild.texi: Document usage of line number 0 in verify
11636 compiler messages directives.
11638 2018-06-12 Matthew Fortune <mfortune@gmail.com>
11640 * config/mips/mips-cpus.def: New MIPS_CPU for i6500.
11641 * config/mips/mips-tables.opt: Regenerate.
11642 * config/mips/mips.h (MIPS_ISA_LEVEL_SPEC): Mark i6500 as
11644 * doc/invoke.texi: Document -march=i6500.
11646 2018-06-12 Prachi Godbole <prachi.godbole@imgtec.com>
11648 * config/mips/i6400.md (i6400_gpmuldiv): Remove cpu_unit.
11649 (i6400_gpmul): Add cpu_unit.
11650 (i6400_gpdiv): Likewise.
11651 (i6400_msa_add_d): Update reservations.
11652 (i6400_msa_int_add) Likewise.
11653 (i6400_msa_short_logic3) Likewise.
11654 (i6400_msa_short_logic2) Likewise.
11655 (i6400_msa_short_logic) Likewise.
11656 (i6400_msa_move) Likewise.
11657 (i6400_msa_cmp) Likewise.
11658 (i6400_msa_short_float2) Likewise.
11659 (i6400_msa_div_d) Likewise.
11660 (i6400_msa_long_logic1) Likewise.
11661 (i6400_msa_long_logic2) Likewise.
11662 (i6400_msa_mult) Likewise.
11663 (i6400_msa_long_float2) Likewise.
11664 (i6400_msa_long_float4) Likewise.
11665 (i6400_msa_long_float5) Likewise.
11666 (i6400_msa_long_float8) Likewise.
11667 (i6400_fpu_fadd): Include frint type.
11668 (i6400_fpu_store): New define_insn_reservation.
11669 (i6400_fpu_load): Likewise.
11670 (i6400_fpu_move): Likewise.
11671 (i6400_fpu_fcmp): Likewise.
11672 (i6400_fpu_fmadd): Likewise.
11673 (i6400_int_mult): Include imul3nc type and update reservation.
11674 (i6400_int_div): Include idiv3 type and update reservation.
11675 (i6400_int_load): Update to check type not move_type.
11676 (i6400_int_store): Likewise.
11677 (i6400_int_prefetch): Set zero latency.
11679 2018-06-12 Eric Botcazou <ebotcazou@adacore.com>
11681 * gcc.c: Document new %@{...} sequence.
11682 (LINK_COMMAND_SPEC): Use it for the -L switches.
11683 (cpp_unique_options): Use it for the -I switches.
11684 (at_file_argbuf): New global variable.
11685 (in_at_file): Likewise.
11686 (alloc_args): Create at_file_argbuf.
11687 (clear_args): Truncate at_file_argbuf.
11688 (store_arg): If in_at_file, push the argument onto at_file_argbuf.
11689 (open_at_file): New function.
11690 (close_at_file): Likewise.
11691 (create_at_file): Delete.
11692 (do_spec_1) <'i'>: Use open_at_file/close_at_file.
11695 (validate_switches_from_spec): Deal with %@{...} sequence.
11696 (validate_switches): Likewise.
11697 (driver::finalize): Call clear_args.
11699 2018-06-11 Rasmus Villemoes <rasmus.villemoes@prevas.dk>
11701 * config/vx-common.h (USE_TM_CLONE_REGISTRY): #define to 0.
11703 2018-06-11 Martin Sebor <msebor@redhat.com>
11705 * doc/invoke.texi (-Wall): List -Wc++17-compat.
11706 (Wno-class-memaccess): Add @opindex.
11707 (Wno-templates, Wno-multiple-inheritance): Same.
11708 (Wno-virtual-inheritance, Wno-namespaces): Same.
11709 (Wno-pedantic, Wno-chkp, Wcoverage-mismatch): Same.
11710 (Wno-format-overflow, Wno-format-truncation): Same.
11711 (Wno-unused-local-typedefs, Walloc-size-larger-than): Same.
11712 (Wno-alloc-size-larger-than, Wframe-larger-than): Same
11713 (Wno-stack-usage, Wno-c++-compat, Wno-c++11-compat): Same.
11714 (Wno-c++14-compat, Wno-c++17-compat, Wno-openmp-simd): Same.
11715 (Wno-unsuffixed-float-constants, Wno-addr-space-convert): Same.
11716 (Wno-misspelled-isr): Same.
11718 2018-06-11 Martin Sebor <msebor@redhat.com>
11720 * PR tree-optimization/86083
11721 * tree-ssa-strlen.c (handle_char_store): Use tree_expr_nonzero_p.
11723 2018-06-11 Zhouyi Zhou <zhouzhouyi@gmail.com>
11725 * tree-eh.c (lower_eh_constructs_2): Add a comma to comment.
11727 2018-06-11 Segher Boessenkool <segher@kernel.crashing.org>
11730 * config/rs6000/rs6000.md (*movdi_internal32): Put constraint modifiers
11731 on the correct operand.
11732 (*movdi_internal64): Ditto.
11734 2018-06-11 Martin Liska <mliska@suse.cz>
11736 PR tree-optimization/86089
11737 * tree-ssa-strlen.c (get_string_length): Move back removed hunk.
11739 2018-06-11 Julia Koval <julia.koval@intel.com>
11741 * config/i386/vaesintrin.h (_mm_aesdec_epi128, _mm_aesdeclast_epi128)
11742 _mm_aesenc_epi128, _mm_aesenclast_epi128: Remove.
11743 * config/i386/vpclmulqdqintrin.h (_mm_clmulepi64_epi128): Remove.
11745 2018-06-11 Olivier Hainque <hainque@adacore.com>
11747 * dwarf2out.c (gen_compile_unit_die): Fallout to DW_LANG_Ada83
11748 for Ada with strict dwarf2.
11750 2018-06-08 Peter Bergner <bergner@vnet.ibm.com>
11753 * config/rs6000/rs6000.c (mem_operand_gpr): Enable PRE_INC and PRE_DEC
11756 2018-06-08 Jan Hubicka <hubicka@ucw.cz>
11758 * dumpfile.c (FIRST_ME_AUTO_NUMBERED_DUMP): Bump to 4.
11760 2018-06-08 David Edelsohn <dje.gcc@gmail.com>
11762 * config/rs6000/rs6000.c (rs6000_passes_ieee128): Protect with #if
11765 2018-06-08 Martin Liska <mliska@suse.cz>
11767 * tree-cfg.h (debug_function): Fix argument type to match
11770 2018-06-08 Martin Liska <mliska@suse.cz>
11772 * config/powerpcspe/powerpcspe.c (rs6000_xcoff_visibility):
11773 Remove usage of MPX-related (and removed) fields.
11774 * config/rs6000/rs6000.c (rs6000_xcoff_visibility): Likewise.
11776 2018-06-08 David Malcolm <dmalcolm@redhat.com>
11778 * cfg.c (debug): Use TDF_NONE rather than 0.
11779 * cfghooks.c (debug): Likewise.
11780 * dumpfile.c (DUMP_FILE_INFO): Likewise; also for OPTGROUP.
11781 (struct dump_option_value_info): Convert to...
11782 (struct kv_pair): ...this template type.
11783 (dump_options): Convert to kv_pair<dump_flags_t>; use TDF_NONE
11785 (optinfo_verbosity_options): Likewise.
11786 (optgroup_options): Convert to kv_pair<optgroup_flags_t>; use
11788 (gcc::dump_manager::dump_register): Use optgroup_flags_t rather
11789 than int for "optgroup_flags" param.
11790 (dump_generic_expr_loc): Use dump_flags_t rather than int for
11792 (dump_dec): Likewise.
11793 (dump_finish): Use TDF_NONE rather than 0.
11794 (gcc::dump_manager::opt_info_enable_passes): Use optgroup_flags_t
11795 rather than int for "optgroup_flags" param. Use TDF_NONE rather
11796 than 0. Update for change to option_ptr.
11797 (opt_info_switch_p_1): Convert "optgroup_flags" param from int *
11798 to optgroup_flags_t *. Use TDF_NONE and OPTGROUP_NONE rather than
11799 0. Update for changes to optinfo_verbosity_options and
11801 (opt_info_switch_p): Convert optgroup_flags from int to
11803 (dump_basic_block): Use dump_flags_t rather than int
11804 for "dump_kind" param.
11805 * dumpfile.h (TDF_ADDRESS, TDF_SLIM, TDF_RAW, TDF_DETAILS,
11806 TDF_STATS, TDF_BLOCKS, TDF_VOPS, TDF_LINENO, TDF_UID)
11807 TDF_STMTADDR, TDF_GRAPH, TDF_MEMSYMS, TDF_RHS_ONLY, TDF_ASMNAME,
11808 TDF_EH, TDF_NOUID, TDF_ALIAS, TDF_ENUMERATE_LOCALS, TDF_CSELIB,
11809 TDF_SCEV, TDF_GIMPLE, TDF_FOLDING, MSG_OPTIMIZED_LOCATIONS,
11810 MSG_MISSED_OPTIMIZATION, MSG_NOTE, MSG_ALL, TDF_COMPARE_DEBUG,
11811 TDF_NONE): Convert from macros to...
11812 (enum dump_flag): ...this new enum.
11813 (dump_flags_t): Update to use enum.
11814 (operator|, operator&, operator~, operator|=, operator&=):
11815 Implement for dump_flags_t.
11816 (OPTGROUP_NONE, OPTGROUP_IPA, OPTGROUP_LOOP, OPTGROUP_INLINE,
11817 OPTGROUP_OMP, OPTGROUP_VEC, OPTGROUP_OTHER, OPTGROUP_ALL):
11818 Convert from macros to...
11819 (enum optgroup_flag): ...this new enum.
11820 (optgroup_flags_t): New typedef.
11821 (operator|, operator|=): Implement for optgroup_flags_t.
11822 (struct dump_file_info): Convert field "alt_flags" to
11823 dump_flags_t. Convert field "optgroup_flags" to
11825 (dump_basic_block): Use dump_flags_t rather than int for param.
11826 (dump_generic_expr_loc): Likewise.
11827 (dump_dec): Likewise.
11828 (dump_register): Convert param "optgroup_flags" to
11830 (opt_info_enable_passes): Likewise.
11831 * early-remat.c (early_remat::dump_edge_list): Use TDF_NONE rather
11833 * gimple-pretty-print.c (debug): Likewise.
11834 * gimple-ssa-store-merging.c (bswap_replace): Likewise.
11835 (merged_store_group::apply_stores): Likewise.
11836 * gimple-ssa-strength-reduction.c (insert_initializers): Likewise.
11837 * gimple.c (verify_gimple_pp): Likewise.
11838 * graphite-poly.c (print_pbb_body): Likewise.
11839 * passes.c (pass_manager::register_one_dump_file): Convert
11840 local "optgroup_flags" to optgroup_flags_t.
11841 * print-tree.c (print_node): Use TDF_NONE rather than 0.
11843 (debug_body): Likewise.
11844 * tree-pass.h (struct pass_data): Convert field "optgroup_flags"
11845 to optgroup_flags_t.
11846 * tree-pretty-print.c (print_struct_decl): Use TDF_NONE rather
11848 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Likewise.
11849 (convert_mult_to_fma): Likewise.
11850 * tree-ssa-reassoc.c (undistribute_ops_list): Likewise.
11851 * tree-ssa-sccvn.c (vn_eliminate): Likewise.
11852 * tree-vect-data-refs.c (dump_lower_bound): Convert param
11853 "dump_kind" to dump_flags_t.
11855 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
11857 * config/rs6000/rs6000.c (min, max): Delete.
11859 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
11861 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete mention of
11862 -mabi=spe and -mabi=no-spe.
11864 2018-06-08 Martin Liska <mliska@suse.cz>
11866 * ipa-pure-const.c (propagate_pure_const): Use ::get at places
11867 where we expect an existing summary.
11869 2018-06-08 Martin Liska <mliska@suse.cz>
11871 * ipa-inline-analysis.c (simple_edge_hints): Use ::get method.
11872 * ipa-inline.h (estimate_edge_growth): Likewise.
11874 2018-06-08 Martin Liska <mliska@suse.cz>
11876 * cgraph.c (function_version_hasher::hash): Use
11877 cgraph_node::get_uid ().
11878 (function_version_hasher::equal):
11879 * cgraph.h (cgraph_node::get_uid): New method.
11880 * ipa-inline.c (update_caller_keys): Use
11881 cgraph_node::get_uid ().
11882 (update_callee_keys): Likewise.
11883 * ipa-utils.c (searchc): Likewise.
11884 (ipa_reduced_postorder): Likewise.
11885 * lto-cgraph.c (input_node): Likewise.
11886 * passes.c (is_pass_explicitly_enabled_or_disabled): Likewise.
11887 * symbol-summary.h (symtab_insertion): Likewise.
11888 (symtab_removal): Likewise.
11889 (symtab_duplication): Likewise.
11890 * tree-pretty-print.c (dump_function_header): Likewise.
11891 * tree-sra.c (convert_callers_for_node): Likewise.
11893 2018-06-08 Martin Liska <mliska@suse.cz>
11895 * cgraph.c (symbol_table::create_edge): Always assign a new
11897 (symbol_table::free_edge): Do not recycle numbers.
11898 * cgraph.h (cgraph_edge::get): New method.
11899 * symbol-summary.h (symtab_removal): Use it.
11900 (symtab_duplication): Likewise.
11901 (call_summary::hashable_uid): Remove.
11903 2018-06-08 Martin Liska <mliska@suse.cz>
11905 * ipa-inline-analysis.c (inline_edge_removal_hook): Remove.
11906 (initialize_growth_caches): Remove.
11907 (free_growth_caches): Likewise.
11908 (do_estimate_edge_time): Use edge_growth_cache.
11909 (do_estimate_edge_size): Likewise.
11910 (do_estimate_edge_hints): Likewise.
11911 * ipa-inline.c (reset_edge_caches): Likewise.
11912 (recursive_inlining): Likewise.
11913 (inline_small_functions): Likewise.
11914 * ipa-inline.h (initialize_growth_caches): Remove.
11915 (estimate_edge_size): Likewise.
11916 (estimate_edge_time): Likewise.
11917 (estimate_edge_hints): Likewise.
11918 (reset_edge_growth_cache): Likewise.
11919 * symbol-summary.h (call_summary::remove): New method.
11921 2018-06-08 Martin Liska <mliska@suse.cz>
11923 * ipa-cp.c (class edge_clone_summary): New summary.
11924 (grow_edge_clone_vectors): Remove.
11925 (ipcp_edge_duplication_hook): Remove.
11926 (class edge_clone_summary_t): New call_summary class.
11927 (ipcp_edge_removal_hook): Remove.
11928 (edge_clone_summary_t::duplicate): New function.
11929 (get_next_cgraph_edge_clone): Use edge_clone_summaries.
11930 (create_specialized_node): Likewise.
11931 (ipcp_driver): Initialize edge_clone_summaries and do not
11934 2018-06-08 Martin Liska <mliska@suse.cz>
11936 * symbol-summary.h (get): New function.
11937 (call_summary::m_initialize_when_cloning): New class member.
11939 2018-06-08 Martin Liska <mliska@suse.cz>
11941 * cgraph.c (cgraph_node::remove): Do not recycle uid.
11942 * cgraph.h (symbol_table::release_symbol): Do not pass uid.
11943 (symbol_table::allocate_cgraph_symbol): Do not set uid.
11944 * passes.c (uid_hash_t): Record removed_nodes by their uids.
11945 (remove_cgraph_node_from_order): Use the removed_nodes set.
11946 (do_per_function_toporder): Likwise.
11947 * symbol-summary.h (symtab_insertion): Use cgraph_node::uid
11948 instead of summary_uid.
11949 (symtab_removal): Likewise.
11950 (symtab_duplication): Likewise.
11952 2018-06-08 Martin Liska <mliska@suse.cz>
11954 * ipa-cp.c (ipcp_store_bits_results): Use
11955 ipcp_transformation_sum.
11956 (ipcp_store_vr_results): Likewise.
11957 * ipa-prop.c (ipcp_grow_transformations_if_necessary): Renamed
11959 (ipcp_transformation_initialize): ... this.
11960 (ipa_set_node_agg_value_chain):
11961 (ipa_node_params_t::duplicate): Use ipcp_transformation_sum.
11962 (write_ipcp_transformation_info): Likewise.
11963 (read_ipcp_transformation_info): Likewise.
11964 (ipcp_update_bits): Likewise.
11965 (ipcp_update_vr): Likewise.
11966 (ipcp_transform_function): Likewise.
11967 * ipa-prop.h: Rename ipcp_transformation_summary to
11968 ipcp_transformation.
11969 (class ipcp_transformation_t): New function summary.
11970 (ipcp_get_transformation_summary): Use ipcp_transformation_sum.
11971 (ipa_get_agg_replacements_for_node): Likewise.
11973 2018-06-08 Martin Liska <mliska@suse.cz>
11975 * ipa-pure-const.c (struct funct_state_d): Do it class instead
11977 (class funct_state_summary_t): New function_summary class.
11978 (has_function_state): Remove.
11979 (get_function_state): Likewise.
11980 (set_function_state): Likewise.
11981 (add_new_function): Likewise.
11982 (funct_state_summary_t::insert): New function.
11983 (duplicate_node_data): Remove.
11984 (remove_node_data): Remove.
11985 (funct_state_summary_t::duplicate): New function.
11986 (register_hooks): Create new funct_state_summaries.
11987 (pure_const_generate_summary): Use it.
11988 (pure_const_write_summary): Likewise.
11989 (pure_const_read_summary): Likewise.
11990 (propagate_pure_const): Likewise.
11991 (propagate_nothrow): Likewise.
11992 (dump_malloc_lattice): Likewise.
11993 (propagate_malloc): Likewise.
11994 (execute): Do not register hooks, just remove summary
11996 (pass_ipa_pure_const::pass_ipa_pure_const): Simplify
11999 2018-06-08 Martin Liska <mliska@suse.cz>
12001 * ipa-reference.c (remove_node_data): Remove.
12002 (duplicate_node_data): Likewise.
12003 (class ipa_ref_var_info_summary_t): New class.
12004 (class ipa_ref_opt_summary_t): Likewise.
12005 (get_reference_vars_info): Use ipa_ref_var_info_summaries.
12006 (get_reference_optimization_summary): Use
12007 ipa_ref_opt_sum_summaries.
12008 (set_reference_vars_info): Remove.
12009 (set_reference_optimization_summary): Likewise.
12010 (ipa_init): Create summaries.
12011 (init_function_info): Use function summary.
12012 (ipa_ref_opt_summary_t::duplicate): New function.
12013 (ipa_ref_opt_summary_t::remove): New function.
12014 (get_read_write_all_from_node): Fix GNU coding style.
12015 (propagate): Use function summary.
12016 (write_node_summary_p): Fix GNU coding style.
12017 (stream_out_bitmap): Likewise.
12018 (ipa_reference_read_optimization_summary): Use function summary.
12019 (ipa_reference_c_finalize): Do not release hooks.
12021 2018-06-08 Martin Liska <mliska@suse.cz>
12023 * ipa-fnsummary.c (dump_ipa_call_summary): Use ::get method.
12024 (analyze_function_body): Extract multiple calls of get_create.
12025 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
12026 * ipa-inline.c (recursive_inlining): Use ::get method.
12027 * ipa-inline.h (estimate_edge_growth): Likewise.
12029 2018-06-08 Martin Liska <mliska@suse.cz>
12031 * hsa-common.h (enum hsa_function_kind): Rename HSA_NONE to
12033 (hsa_function_summary::hsa_function_summary): Use the new enum
12035 (hsa_gpu_implementation_p): Use hsa_summaries::get.
12036 * hsa-gen.c (hsa_get_host_function): Likewise.
12037 (get_brig_function_name): Likewise.
12038 * ipa-hsa.c (process_hsa_functions): Likewise.
12039 (ipa_hsa_write_summary): Likewise.
12040 * symbol-summary.h (symtab_duplication): Use ::get function/
12041 (get): New function.
12043 2018-06-08 Martin Liska <mliska@suse.cz>
12045 * config/i386/i386.c (ix86_can_inline_p): Use get_create instead
12047 * hsa-common.c (hsa_summary_t::link_functions): Likewise.
12048 (hsa_register_kernel): Likewise.
12049 * hsa-common.h (hsa_gpu_implementation_p): Likewise.
12050 * hsa-gen.c (hsa_get_host_function): Likewise.
12051 (get_brig_function_name): Likewise.
12052 (generate_hsa): Likewise.
12053 (pass_gen_hsail::execute): Likewise.
12054 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
12055 (devirtualization_time_bonus): Likewise.
12056 (ipcp_propagate_stage): Likewise.
12057 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
12058 (edge_set_predicate): Likewise.
12059 (evaluate_conditions_for_known_args): Likewise.
12060 (evaluate_properties_for_edge): Likewise.
12061 (ipa_fn_summary::reset): Likewise.
12062 (ipa_fn_summary_t::duplicate): Likewise.
12063 (dump_ipa_call_summary): Likewise.
12064 (ipa_dump_fn_summary): Likewise.
12065 (analyze_function_body): Likewise.
12066 (compute_fn_summary): Likewise.
12067 (estimate_edge_devirt_benefit): Likewise.
12068 (estimate_edge_size_and_time): Likewise.
12069 (estimate_calls_size_and_time): Likewise.
12070 (estimate_node_size_and_time): Likewise.
12071 (inline_update_callee_summaries): Likewise.
12072 (remap_edge_change_prob): Likewise.
12073 (remap_edge_summaries): Likewise.
12074 (ipa_merge_fn_summary_after_inlining): Likewise.
12075 (ipa_update_overall_fn_summary): Likewise.
12076 (read_ipa_call_summary): Likewise.
12077 (inline_read_section): Likewise.
12078 (write_ipa_call_summary): Likewise.
12079 (ipa_fn_summary_write): Likewise.
12080 (ipa_free_fn_summary): Likewise.
12081 * ipa-hsa.c (process_hsa_functions): Likewise.
12082 (ipa_hsa_write_summary): Likewise.
12083 (ipa_hsa_read_section): Likewise.
12084 * ipa-icf.c (sem_function::merge): Likewise.
12085 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
12086 (do_estimate_edge_time): Likewise.
12087 (estimate_size_after_inlining): Likewise.
12088 (estimate_growth): Likewise.
12089 (growth_likely_positive): Likewise.
12090 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
12091 (inline_call): Likewise.
12092 * ipa-inline.c (caller_growth_limits): Likewise.
12093 (can_inline_edge_p): Likewise.
12094 (can_inline_edge_by_limits_p): Likewise.
12095 (compute_uninlined_call_time): Likewise.
12096 (compute_inlined_call_time): Likewise.
12097 (want_inline_small_function_p): Likewise.
12098 (edge_badness): Likewise.
12099 (update_caller_keys): Likewise.
12100 (update_callee_keys): Likewise.
12101 (recursive_inlining): Likewise.
12102 (inline_small_functions): Likewise.
12103 (inline_to_all_callers_1): Likewise.
12104 (dump_overall_stats): Likewise.
12105 (early_inline_small_functions): Likewise.
12106 (early_inliner): Likewise.
12107 * ipa-inline.h (estimate_edge_growth): Likewise.
12108 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
12109 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
12110 * ipa-prop.h (IPA_NODE_REF): Likewise.
12111 (IPA_EDGE_REF): Likewise.
12112 * ipa-pure-const.c (malloc_candidate_p): Likewise.
12113 (propagate_malloc): Likewise.
12114 * ipa-split.c (execute_split_functions): Likewise.
12115 * symbol-summary.h: Rename get to get_create.
12117 (get_create): Likewise.
12118 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
12120 2018-06-08 Martin Liska <mliska@suse.cz>
12122 * symbol-summary.h (release): Move definition out of class
12124 (symtab_removal): Likewise.
12125 (symtab_duplication): Likewise.
12127 2018-06-08 Martin Liska <mliska@suse.cz>
12129 * symbol-summary.h (function_summary): Move constructor
12130 implementation out of class declaration.
12131 (release): Likewise.
12132 (symtab_insertion): Likewise.
12133 (symtab_removal): Likewise.
12134 (symtab_duplication): Likewise.
12137 2018-06-08 Martin Liska <mliska@suse.cz>
12139 * Makefile.in: Remove support for MPX (macros, related functions,
12140 fields in cgraph_node, ...).
12141 * builtin-types.def (BT_BND): Likewise.
12142 (BT_FN_BND_CONST_PTR): Likewise.
12143 (BT_FN_CONST_PTR_BND): Likewise.
12144 (BT_FN_VOID_PTR_BND): Likewise.
12145 (BT_FN_BND_CONST_PTR_SIZE): Likewise.
12146 (BT_FN_VOID_CONST_PTR_BND_CONST_PTR): Likewise.
12147 * builtins.c (expand_builtin_memcpy_with_bounds): Likewise.
12148 (expand_builtin_mempcpy_with_bounds): Likewise.
12149 (expand_builtin_memset_with_bounds): Likewise.
12150 (expand_builtin_memset_args): Likewise.
12151 (std_expand_builtin_va_start): Likewise.
12152 (expand_builtin): Likewise.
12153 (expand_builtin_with_bounds): Likewise.
12154 * builtins.def (DEF_BUILTIN_CHKP): Likewise.
12155 (DEF_LIB_BUILTIN_CHKP): Likewise.
12156 (DEF_EXT_LIB_BUILTIN_CHKP): Likewise.
12157 (DEF_CHKP_BUILTIN): Likewise.
12158 (BUILT_IN_MEMCPY): Likewise.
12159 (BUILT_IN_MEMMOVE): Likewise.
12160 (BUILT_IN_MEMPCPY): Likewise.
12161 (BUILT_IN_MEMSET): Likewise.
12162 (BUILT_IN_STPCPY): Likewise.
12163 (BUILT_IN_STRCAT): Likewise.
12164 (BUILT_IN_STRCHR): Likewise.
12165 (BUILT_IN_STRCPY): Likewise.
12166 (BUILT_IN_STRLEN): Likewise.
12167 (BUILT_IN_MEMCPY_CHK): Likewise.
12168 (BUILT_IN_MEMMOVE_CHK): Likewise.
12169 (BUILT_IN_MEMPCPY_CHK): Likewise.
12170 (BUILT_IN_MEMSET_CHK): Likewise.
12171 (BUILT_IN_STPCPY_CHK): Likewise.
12172 (BUILT_IN_STRCAT_CHK): Likewise.
12173 (BUILT_IN_STRCPY_CHK): Likewise.
12174 * calls.c (store_bounds): Likewise.
12175 (emit_call_1): Likewise.
12176 (special_function_p): Likewise.
12177 (maybe_warn_nonstring_arg): Likewise.
12178 (initialize_argument_information): Likewise.
12179 (finalize_must_preallocate): Likewise.
12180 (compute_argument_addresses): Likewise.
12181 (expand_call): Likewise.
12182 * cfgexpand.c (expand_call_stmt): Likewise.
12183 (expand_return): Likewise.
12184 (expand_gimple_stmt_1): Likewise.
12185 (pass_expand::execute): Likewise.
12186 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
12187 (cgraph_node::remove): Likewise.
12188 (cgraph_node::dump): Likewise.
12189 (cgraph_node::verify_node): Likewise.
12190 * cgraph.h (chkp_function_instrumented_p): Likewise.
12191 (symtab_node::get_alias_target): Likewise.
12192 (cgraph_node::can_remove_if_no_direct_calls_and_refs_p): Likewise.
12193 (cgraph_local_p): Likewise.
12194 * cgraphbuild.c (cgraph_edge::rebuild_edges): Likewise.
12195 (cgraph_edge::rebuild_references): Likewise.
12196 * cgraphunit.c (varpool_node::finalize_decl): Likewise.
12197 (walk_polymorphic_call_targets): Likewise.
12198 (cgraph_node::expand_thunk): Likewise.
12199 (symbol_table::output_weakrefs): Likewise.
12200 * common/config/i386/i386-common.c (OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET): Likewise.
12201 (ix86_handle_option): Likewise.
12202 * config/i386/constraints.md: Likewise.
12203 * config/i386/i386-builtin-types.def (BND): Likewise.
12207 * config/i386/i386-builtin.def (BDESC_END): Likewise.
12208 (BDESC_FIRST): Likewise.
12210 * config/i386/i386-c.c (ix86_target_macros_internal): Likewise.
12211 * config/i386/i386-protos.h (ix86_bnd_prefixed_insn_p): Likewise.
12212 * config/i386/i386.c (enum reg_class): Likewise.
12213 (ix86_target_string): Likewise.
12214 (ix86_option_override_internal): Likewise.
12215 (ix86_conditional_register_usage): Likewise.
12216 (ix86_valid_target_attribute_inner_p): Likewise.
12217 (ix86_set_indirect_branch_type): Likewise.
12218 (ix86_set_current_function): Likewise.
12219 (ix86_function_arg_regno_p): Likewise.
12220 (init_cumulative_args): Likewise.
12221 (ix86_function_arg_advance): Likewise.
12222 (ix86_function_arg): Likewise.
12223 (ix86_pass_by_reference): Likewise.
12224 (ix86_function_value_regno_p): Likewise.
12225 (ix86_function_value_1): Likewise.
12226 (ix86_function_value_bounds): Likewise.
12227 (ix86_return_in_memory): Likewise.
12228 (ix86_setup_incoming_vararg_bounds): Likewise.
12229 (ix86_va_start): Likewise.
12230 (indirect_thunk_need_prefix): Likewise.
12231 (print_reg): Likewise.
12232 (ix86_print_operand): Likewise.
12233 (ix86_expand_call): Likewise.
12234 (ix86_output_function_return): Likewise.
12235 (reg_encoded_number): Likewise.
12236 (BDESC_VERIFYS): Likewise.
12237 (ix86_init_mpx_builtins): Likewise.
12238 (ix86_init_builtins): Likewise.
12239 (ix86_emit_cmove): Likewise.
12240 (ix86_emit_move_max): Likewise.
12241 (ix86_expand_builtin): Likewise.
12242 (ix86_builtin_mpx_function): Likewise.
12243 (ix86_get_arg_address_for_bt): Likewise.
12244 (ix86_load_bounds): Likewise.
12245 (ix86_store_bounds): Likewise.
12246 (ix86_load_returned_bounds): Likewise.
12247 (ix86_store_returned_bounds): Likewise.
12248 (ix86_class_likely_spilled_p): Likewise.
12249 (ix86_hard_regno_mode_ok): Likewise.
12250 (x86_order_regs_for_local_alloc): Likewise.
12251 (ix86_mitigate_rop): Likewise.
12252 (ix86_bnd_prefixed_insn_p): Likewise.
12253 (ix86_mpx_bound_mode): Likewise.
12254 (ix86_make_bounds_constant): Likewise.
12255 (ix86_initialize_bounds): Likewise.
12256 (TARGET_LOAD_BOUNDS_FOR_ARG): Likewise.
12257 (TARGET_STORE_BOUNDS_FOR_ARG): Likewise.
12258 (TARGET_LOAD_RETURNED_BOUNDS): Likewise.
12259 (TARGET_STORE_RETURNED_BOUNDS): Likewise.
12260 (TARGET_CHKP_BOUND_MODE): Likewise.
12261 (TARGET_BUILTIN_CHKP_FUNCTION): Likewise.
12262 (TARGET_CHKP_FUNCTION_VALUE_BOUNDS): Likewise.
12263 (TARGET_CHKP_MAKE_BOUNDS_CONSTANT): Likewise.
12264 (TARGET_CHKP_INITIALIZE_BOUNDS): Likewise.
12265 * config/i386/i386.h (TARGET_MPX): Likewise.
12266 (TARGET_MPX_P): Likewise.
12267 (VALID_BND_REG_MODE): Likewise.
12268 (FIRST_BND_REG): Likewise.
12269 (LAST_BND_REG): Likewise.
12270 (enum reg_class): Likewise.
12271 (BND_REG_P): Likewise.
12272 (BND_REGNO_P): Likewise.
12273 (BNDmode): Likewise.
12274 (ADJUST_INSN_LENGTH): Likewise.
12275 * config/i386/i386.md: Likewise.
12276 * config/i386/i386.opt: Likewise.
12277 * config/i386/linux-common.h (LIBMPX_LIBS): Likewise.
12278 (defined): Likewise.
12279 (LINK_MPX): Likewise.
12280 (MPX_SPEC): Likewise.
12281 (LIBMPX_SPEC): Likewise.
12282 (LIBMPXWRAPPERS_SPEC): Likewise.
12283 (CHKP_SPEC): Likewise.
12284 * config/i386/predicates.md: Likewise.
12285 * dbxout.c (dbxout_type): Likewise.
12286 * doc/extend.texi: Likewise.
12287 * doc/invoke.texi: Likewise.
12288 * doc/md.texi: Likewise.
12289 * doc/tm.texi: Likewise.
12290 * doc/tm.texi.in: Likewise.
12291 * dwarf2out.c (is_base_type): Likewise.
12292 (gen_formal_types_die): Likewise.
12293 (gen_subprogram_die): Likewise.
12294 (gen_type_die_with_usage): Likewise.
12295 (gen_decl_die): Likewise.
12296 (dwarf2out_late_global_decl): Likewise.
12297 * expr.c (expand_assignment): Likewise.
12298 (emit_storent_insn): Likewise.
12299 (store_expr_with_bounds): Likewise.
12300 (store_expr): Likewise.
12301 (expand_expr_real_1): Likewise.
12302 * expr.h (store_expr_with_bounds): Likewise.
12303 * function.c (use_register_for_decl): Likewise.
12304 (struct bounds_parm_data): Likewise.
12305 (assign_parms_augmented_arg_list): Likewise.
12306 (assign_parm_find_entry_rtl): Likewise.
12307 (assign_parm_is_stack_parm): Likewise.
12308 (assign_parm_load_bounds): Likewise.
12309 (assign_bounds): Likewise.
12310 (assign_parms): Likewise.
12311 (expand_function_start): Likewise.
12312 * gcc.c (CHKP_SPEC): Likewise.
12313 * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
12314 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Likewise.
12315 (wrestrict_dom_walker::check_call): Likewise.
12316 * gimple.c (gimple_build_call_from_tree): Likewise.
12317 * gimple.h (enum gf_mask): Likewise.
12318 (gimple_call_with_bounds_p): Likewise.
12319 (gimple_call_set_with_bounds): Likewise.
12320 * gimplify.c (gimplify_init_constructor): Likewise.
12321 * ipa-cp.c (initialize_node_lattices): Likewise.
12322 (propagate_constants_across_call): Likewise.
12323 (find_more_scalar_values_for_callers_subset): Likewise.
12324 * ipa-hsa.c (process_hsa_functions): Likewise.
12325 * ipa-icf-gimple.c (func_checker::compare_gimple_call): Likewise.
12326 * ipa-icf.c (sem_function::merge): Likewise.
12327 * ipa-inline.c (early_inliner): Likewise.
12328 * ipa-pure-const.c (warn_function_noreturn): Likewise.
12329 (warn_function_cold): Likewise.
12330 (propagate_pure_const): Likewise.
12331 * ipa-ref.h (enum GTY): Likewise.
12332 * ipa-split.c (find_retbnd): Likewise.
12333 (consider_split): Likewise.
12334 (split_function): Likewise.
12335 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
12336 * ipa.c (walk_polymorphic_call_targets): Likewise.
12337 (symbol_table::remove_unreachable_nodes): Likewise.
12338 (process_references): Likewise.
12339 (cgraph_build_static_cdtor_1): Likewise.
12340 * lto-cgraph.c (lto_output_node): Likewise.
12341 (output_refs): Likewise.
12342 (compute_ltrans_boundary): Likewise.
12343 (input_overwrite_node): Likewise.
12344 (input_node): Likewise.
12345 (input_cgraph_1): Likewise.
12346 * params.def (PARAM_CHKP_MAX_CTOR_SIZE): Likewise.
12347 * passes.c (pass_manager::execute_early_local_passes): Likewise.
12348 (class pass_chkp_instrumentation_passes): Likewise.
12349 (make_pass_chkp_instrumentation_passes): Likewise.
12350 * passes.def: Likewise.
12351 * rtl.h (struct GTY): Likewise.
12352 (CALL_EXPR_WITH_BOUNDS_P): Likewise.
12353 * stor-layout.c (layout_type): Likewise.
12354 * symtab.c: Likewise.
12355 * target.def: Likewise.
12356 * targhooks.c (default_chkp_bound_type): Likewise.
12357 (default_chkp_bound_mode): Likewise.
12358 (default_builtin_chkp_function): Likewise.
12359 (default_chkp_function_value_bounds): Likewise.
12360 (default_chkp_make_bounds_constant): Likewise.
12361 (default_chkp_initialize_bounds): Likewise.
12362 * targhooks.h (default_chkp_bound_type): Likewise.
12363 (default_chkp_bound_mode): Likewise.
12364 (default_builtin_chkp_function): Likewise.
12365 (default_chkp_function_value_bounds): Likewise.
12366 (default_chkp_make_bounds_constant): Likewise.
12367 (default_chkp_initialize_bounds): Likewise.
12368 * toplev.c (compile_file): Likewise.
12369 (process_options): Likewise.
12370 * tree-core.h (DEF_BUILTIN): Likewise.
12371 (DEF_BUILTIN_CHKP): Likewise.
12372 * tree-inline.c (declare_return_variable): Likewise.
12373 (remap_gimple_stmt): Likewise.
12374 (copy_bb): Likewise.
12375 (initialize_inlined_parameters): Likewise.
12376 (expand_call_inline): Likewise.
12377 * tree-pass.h (make_pass_ipa_chkp_versioning): Likewise.
12378 (make_pass_ipa_chkp_early_produce_thunks): Likewise.
12379 (make_pass_ipa_chkp_produce_thunks): Likewise.
12380 (make_pass_chkp): Likewise.
12381 (make_pass_chkp_opt): Likewise.
12382 (make_pass_chkp_instrumentation_passes): Likewise.
12383 * tree-pretty-print.c (dump_generic_node): Likewise.
12384 * tree-ssa-ccp.c (insert_clobber_before_stack_restore): Likewise.
12385 * tree-ssa-dce.c (propagate_necessity): Likewise.
12386 (eliminate_unnecessary_stmts): Likewise.
12387 * tree-ssa-pre.c (create_expression_by_pieces): Likewise.
12388 * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
12389 * tree-ssa-sccvn.h: Likewise.
12390 * tree-ssa-strlen.c (get_string_length): Likewise.
12391 (valid_builtin_call): Likewise.
12392 (adjust_last_stmt): Likewise.
12393 (handle_builtin_strchr): Likewise.
12394 (handle_builtin_strcpy): Likewise.
12395 (handle_builtin_stxncpy): Likewise.
12396 (handle_builtin_memcpy): Likewise.
12397 (handle_builtin_strcat): Likewise.
12398 (strlen_check_and_optimize_stmt): Likewise.
12399 * tree-stdarg.c (expand_ifn_va_arg_1): Likewise.
12400 * tree-streamer-in.c: Likewise.
12401 * tree-streamer.c (record_common_node): Likewise.
12402 * tree.c (tree_code_size): Likewise.
12403 (wide_int_to_tree_1): Likewise.
12404 (type_contains_placeholder_1): Likewise.
12405 (build_common_tree_nodes): Likewise.
12406 * tree.def (POINTER_BOUNDS_TYPE): Likewise.
12407 * tree.h (POINTER_BOUNDS_TYPE_P): Likewise.
12408 (POINTER_BOUNDS_P): Likewise.
12409 (BOUNDED_TYPE_P): Likewise.
12410 (BOUNDED_P): Likewise.
12411 (CALL_WITH_BOUNDS_P): Likewise.
12412 (pointer_bounds_type_node): Likewise.
12413 * value-prof.c (gimple_ic): Likewise.
12414 * var-tracking.c (vt_add_function_parameters): Likewise.
12415 * varasm.c (make_decl_rtl): Likewise.
12416 (assemble_start_function): Likewise.
12417 (output_constant): Likewise.
12418 (maybe_assemble_visibility): Likewise.
12419 * varpool.c (ctor_for_folding): Likewise.
12420 * chkp-builtins.def: Remove.
12421 * ipa-chkp.c: Remove.
12422 * ipa-chkp.h: Remove.
12423 * rtl-chkp.c: Remove.
12424 * rtl-chkp.h: Remove.
12425 * tree-chkp-opt.c: Remove.
12426 * tree-chkp.c: Remove.
12427 * tree-chkp.h: Remove.
12429 2018-06-07 Carl Love <cel@us.ibm.com>
12431 * config/rs6000/vsx.md (vextract_fp_from_shorth,
12432 vextract_fp_from_shortl): Add BE support.
12434 2018-06-07 Paul Koning <ni1d@arrl.net>
12436 * compare-elim.c (try_merge_compare): Don't merge compare if
12437 address contains a side effect.
12438 (try_eliminate_compare): Likewise.
12440 2018-06-07 Olga Makhotina <olga.makhotina@intel.com>
12442 * config.gcc: Support "tremont".
12443 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "tremont".
12444 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
12446 * config/i386/i386.c (m_TREMONT): Define.
12447 (processor_target_table): Add "tremont".
12448 (PTA_TREMONT): Define.
12449 (ix86_lea_outperforms): Add TARGET_TREMONT.
12450 (get_builtin_code_for_version): Handle PROCESSOR_TREMONT.
12451 (fold_builtin_cpu): Add M_INTEL_TREMONT, replace M_INTEL_GOLDMONT
12452 and M_INTEL_GOLDMONT_PLUS.
12453 (fold_builtin_cpu): Add "tremont".
12454 (ix86_add_stmt_cost): Add TARGET_TREMONT.
12455 (ix86_option_override_internal): Add "tremont".
12456 * config/i386/i386.h (processor_costs): Define TARGET_TREMONT.
12457 (processor_type): Add PROCESSOR_TREMONT.
12458 * config/i386/x86-tune.def: Add m_TREMONT.
12459 * doc/invoke.texi: Add tremont as x86 -march=/-mtune= CPU type.
12461 2018-06-07 Jozef Lawrynowicz <jozef.l@mittosystems.com>
12463 * config/msp430/msp430.c (msp430_mcu_name): Set the "i" in the
12464 symbol defined for msp430i* devices to be lower case.
12466 2018-06-07 Richard Biener <rguenther@suse.de>
12468 * graphite-sese-to-poly.c (extract_affine): Avoid unneded
12469 wrapping. Properly wrap the result of a BIT_NOT_EXPR.
12470 Properly wrap signed arithmetic if overflow wraps.
12472 2018-06-07 Jakub Jelinek <jakub@redhat.com>
12474 PR tree-optimization/69615
12475 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): If rhs2 is lhs
12476 of a cast from a same precision integral SSA_NAME in a bb dominated
12477 by first_bb, retry with rhs2 set to the rhs1 of the cast. Don't emit
12478 cast to utype if rhs2 has already a compatible type.
12480 2018-06-07 Richard Biener <rguenther@suse.de>
12482 PR tree-optimization/85935
12483 * graphite-scop-detection.c (find_params_in_bb): Analyze
12484 condition operands with respect to the correct loop. Assert
12485 the analysis doesn't fail.
12487 2018-06-04 Carl Love <cel@us.ibm.com>
12489 * config/rs6000/vsx.md (first_match_index_<mode>): Calculate index
12490 using natural element order. Use gen_lshrsi3 instead of gen_ashrsi3
12491 as it is slightly cheaper.
12492 (first_match_or_eos_index_<mode>):
12493 Calculate index using natural element order.
12494 (first_match_index_<mode>):
12495 Calculate index using natural element order.
12496 (first_match_or_eos_index_<mode>):
12497 Calculate index using natural order.
12498 (define_insn vclzlsbb): Change to define_insn vclzlsbb_<mode>.
12499 for BE and LE modes.
12500 * config/rs6000/rs6000-c.c: Rename P9V_BUILTIN_VCLZLSBB,
12501 P9V_BUILTIN_VCLZLSBB_V16QI.
12502 * config/rs6000/rs6000-builtin.def: Make VCLZLSBB mode
12505 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
12507 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Adjust
12508 indentation and line wrap for many prototypes. Add missing
12509 @smallexample directives around block of prototypes for vec_xl and
12512 2018-06-05 Michael Meissner <meissner@linux.ibm.com>
12514 * config/rs6000/rs6000.c (rs6000_passes_ieee128): New boolean to
12515 track if we pass or return IEEE 128-bit floating point.
12516 (ieee128_mangling_gcc_8_1): New boolean to say whether to generate
12517 C++ mangling that is compatible with GCC 8.1.
12518 (TARGET_ASM_GLOBALIZE_DECL_NAME): Override target hook.
12519 (init_cumulative_args): Note if we pass or return IEEE 128-bit
12520 floating point types.
12521 (rs6000_function_arg_advance_1): Likewise.
12522 (rs6000_mangle_type): Optionally generate mangled names that match
12523 what GCC 8.1 generated for IEEE 128-bit floating point types.
12524 (rs6000_globalize_decl_name): If we have an external function that
12525 passes or returns IEEE 128-bit types, generate a weak reference
12526 from the mangled name used in GCC 8.1 to the current mangled
12528 (rs6000_init_builtins): Make __ibm128 use the long double type if
12529 long double is IBM extended double. Make __float128 use the long
12530 double type if long double is IEEE 128-bit.
12533 * config/rs6000/rs6000-builtin.def (BU_IBM128_2): New helper
12534 macro for __ibm128 built-in functions.
12535 (PACK_IF): Add __ibm128 pack/unpack functions.
12536 (UNPACK_IF): Likewise.
12537 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Do not
12538 enable long double built-in functions if long double is IEEE
12539 128-bit floating point.
12540 (rs6000_invalid_builtin): Update long double built-in function
12542 (rs6000_expand_builtin): For PACK_IF and UNPACK_IF built-in
12543 functions, adjust the built-in function to use the long double
12544 built-in function if __ibm128 and long double are the same type.
12545 * doc/extend.texi (PowerPC builtins): Update documention for
12546 __builtin_{,un}pack_longdouble. Add documentation for
12547 __builtin_{,un}pack_ibm128.
12549 2018-06-06 Jim Wilson <jimw@sifive.com>
12551 * config/riscv/riscv.c (enum riscv_privilege_levels): New.
12552 (struct machine_function): New field interrupt_mode.
12553 (riscv_handle_type_attribute): New function. Add forward declaration.
12554 (riscv_attribute_table) <interrupt>: Use riscv_handle_type_attribute.
12555 (riscv_expand_epilogue): Check interrupt_mode field.
12556 (riscv_set_current_function): Check interrupt attribute args and
12557 set interrupt_mode field.
12558 * config/riscv/riscv.md (UNSPECV_SRET, UNSPECV_URET): New.
12559 (riscv_sret, riscv_uret): New.
12560 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: Document
12561 new arguments to interrupt attribute.
12563 2018-06-06 Peter Bergner <bergner@vnet.ibm.com>
12566 * /config/rs6000/rs6000.h (ASM_CPU_SPEC): Add support for -mpower9.
12567 Don't handle -mcpu=power8 if -mpower9-vector is also used.
12569 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
12571 * config/rs6000/rs6000-builtin.def (VSX_BUILTIN_VEC_LD,
12572 VSX_BUILTIN_VEC_ST): Add comment to explain non-traditional uses.
12573 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
12574 several redundant entries.
12576 2018-06-06 David Malcolm <dmalcolm@redhat.com>
12578 * config/i386/i386-protos.h (ix86_expand_call): Strengthen return
12579 type from "rtx" to "rtx_insn *".
12580 * config/i386/i386.c (ix86_expand_split_stack_prologue): Likewise
12581 for local "call_insn", removing cast.
12582 (ix86_expand_call): Likewise, introducing a "call_insn" local.
12584 2018-06-06 Eric Botcazou <ebotcazou@adacore.com>
12586 PR tree-optimization/86066
12587 * gimple-ssa-store-merging.c (process_store): Do not bypass BIT_NOT_EXPR
12588 for BIT_INSERT_EXPR stores.
12590 2018-06-06 Richard Biener <rguenther@suse.de>
12592 PR tree-optimization/86062
12593 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
12594 component refs ontop
12595 of to be offsetted base.
12597 2018-06-06 Jozef Lawrynowicz <jozef.l@mittosystems.com>
12599 * config/msp430/msp430.c (msp430_attr): Allow interrupt handlers
12600 to be static and remove check on interrupt attribute name.
12602 2018-06-05 Kelvin Nilsen <kelvin@gcc.gnu.org>
12604 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
12605 volatile qualifier from vec_lvsl and vec_lvsr argument prototypes.
12607 2018-06-05 Steve Ellcey <sellcey@cavium.com>
12610 * config/aarch64/aarch64-protos.h (aarch64_err_no_fpadvsimd): Remove
12612 * config/aarch64/aarch64-protos..c (aarch64_err_no_fpadvsimd):
12613 Remove second argument, change how error is called.
12614 (aarch64_layout_arg): Remove second argument from
12615 aarch64_err_no_fpadvsimd call.
12616 (aarch64_init_cumulative_args): Ditto.
12617 (aarch64_gimplify_va_arg_expr): Ditto.
12618 * config/aarch64/aarch64.md (mov<mode>): Ditto.
12620 2018-06-05 Uros Bizjak <ubizjak@gmail.com>
12622 * config/i386/i386.md (simple_return_indirect_internal): New expander.
12623 (*simple_return_indirect_internal<mode>): Rename from
12624 simple_return_indirect_internal. Use W mode iterator.
12625 (rstorssp): New expander.
12626 (*rstorssp<mode>): Rename from rstorssp. Use P mode iterator.
12627 (clrssbsy): New expander.
12628 (*clrssbsy<mode>): Rename from clrssbsy. Use P mode iterator.
12630 2018-06-05 Andre Vieira <andre.simoesdiasvieira@arm.com>
12632 * config/arm/arm_cmse.h (cmse_nsfptr_create): Change typeof to
12634 (cmse_check_pointed_object): Likewise.
12636 2018-06-05 Martin Liska <mliska@suse.cz>
12638 PR gcov-profile/47618
12639 * doc/invoke.texi: Document how -fprofile-dir format
12642 2018-06-05 Richard Biener <rguenther@suse.de>
12644 * tree-cfgcleanup.c (cleanup_control_flow_pre): For edge
12645 removal pretend DOM info isn't available so we do not update
12646 it and only remove edges, not dominated blocks. Actually free
12647 DOM info in case we removed something. Remove unreachable blocks.
12648 (mfb_keep_latches): Work with either DOM info or marked backedges.
12649 (cleanup_tree_cfg_noloop): Do not remove unreachable blocks
12650 first. Mark backedges if DOM info isn't available.
12651 (Re-)compute DOM info after cleanup_control_flow_pre.
12653 2018-06-05 Richard Biener <rguenther@suse.de>
12655 * tree-cfg.c (struct locus_discrim_map): Store line, not location.
12656 (locus_discrim_hasher::hash): Adjust.
12657 (locus_discrim_hasher::equal): Likewise.
12658 (next_discriminator_for_locus): Work on line directly.
12659 (same_line_p): Pass in expanded locus1 as well.
12660 (assign_discriminators): Avoid redundant location expansions.
12662 2018-06-05 Richard Biener <rguenther@suse.de>
12664 PR tree-optimization/86046
12665 * tree-ssa.c (maybe_optimize_var): Clear DECL_GIMPLE_REG_P
12666 if required after clearing TREE_ADDRESSABLE.
12668 2018-06-05 Richard Biener <rguenther@suse.de>
12670 PR tree-optimization/86047
12671 * tree-ssa-loop.c (for_each_index): Glob handling of all
12672 decls and constants and really handle all of them.
12674 2018-06-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12677 * config/arm/arm-builtins.c (arm_type_qualifiers): Add
12678 qualifier_void_pointer and qualifier_const_void_pointer.
12679 (arm_ldc_qualifiers, arm_stc_qualifiers): Use the above.
12680 (arm_init_builtins): Handle the above.
12681 * config/arm/arm_acle.h (__arm_cdp, __arm_ldc, __arm_ldcl, __arm_stc,
12682 __arm_stcl, __arm_mcr, __arm_cdp2, __arm_ldc2, __arm_ldcl2, __arm_stc2,
12683 __arm_stcl2,__arm_mcr2, __arm_mcrr, __arm_mcrr2): Remove return for
12686 2018-06-05 Martin Liska <mliska@suse.cz>
12688 * auto-profile.c (read_autofdo_file): Do not use
12689 gcov_ctr_summary struct.
12690 (afdo_callsite_hot_enough_for_early_inline): Likewise.
12691 * coverage.c (struct counts_entry): Likewise.
12692 (read_counts_file): Read just single summary entry.
12693 (get_coverage_counts): Use gcov_summary struct.
12694 * coverage.h (get_coverage_counts): Likewise.
12695 * gcov-dump.c (dump_working_sets): Likewise.
12696 (tag_summary): Dump just single summary.
12697 * gcov-io.c (gcov_write_summary): Write just histogram
12699 (gcov_read_summary): Read just single summary.
12700 (compute_working_sets): Use gcov_summary struct.
12701 * gcov-io.h (GCOV_TAG_SUMMARY_LENGTH): Remove usage
12702 of GCOV_COUNTERS_SUMMABLE.
12703 (GCOV_COUNTERS_SUMMABLE): Remove.
12704 (GCOV_FIRST_VALUE_COUNTER): Replace with
12705 GCOV_COUNTER_V_INTERVAL.
12706 (struct gcov_ctr_summary): Remove.
12707 (struct gcov_summary): Directly use fields of former
12709 (compute_working_sets): Use gcov_summary struct.
12710 * gcov.c (read_count_file): Do not use ctrs fields.
12711 * lto-cgraph.c (merge_profile_summaries): Use gcov_summary
12713 * lto-streamer.h (struct GTY): Make profile_info gcov_summary
12715 * profile.c: Likewise.
12716 * profile.h: Likewise.
12718 2018-06-05 Martin Liska <mliska@suse.cz>
12720 PR gcov-profile/84846
12721 * gcov.c (output_lines): Print working directory only
12722 in intermediate format.
12724 2018-06-05 Andreas Krebbel <krebbel@linux.ibm.com>
12726 * config/s390/s390-builtin-types.def: Add void function type.
12727 * config/s390/s390-builtins.def: Use the function type for the
12730 2018-06-04 Jim Wilson <jimw@sifive.com>
12732 * config/riscv/riscv-protos.h (riscv_expand_epilogue): Change bool arg
12734 * config/riscv/riscv.c (riscv_for_each_saved_reg): New args epilogue
12735 and maybe_eh_return. Change regno to unsigned int. Use new args to
12736 handle EH_RETURN_DATA_REGNO registers properly.
12737 (riscv_expand_prologue): Pass new args to riscv_for_each_saved_reg.
12738 (riscv_expand_epilogue): Update comment. Change argument name and
12739 type. Update code to use new name and type. Pass new args to
12740 riscv_for_each_saved_reg. Only use EH_RETURN_STACKADJ_RTX when
12742 * config/riscv/riscv.md (NORMAL_RETURN): New.
12743 (SIBCALL_RETURN, EXCEPTION_RETURN): New.
12744 (epilogue, sibcall_epilogue): Update riscv_expand_epilogue arg.
12745 (eh_return): Call gen_eh_return_internal and emit barrier.
12746 (eh_return_internal): Call riscv_expand_epilogue.
12748 2018-06-04 Eric Botcazou <ebotcazou@adacore.com>
12750 * gimple-ssa-store-merging.c (struct merged_store_group): Move up
12751 bit_insertion field and declare can_be_merged_into method.
12752 (merged_store_group::can_be_merged_into): New method.
12753 (imm_store_chain_info::coalesce_immediate): Call it to decide whether
12754 consecutive non-overlapping stores can be merged. Turn MEM_REF stores
12755 into BIT_INSERT_EXPR stores if the group contains a non-MEM_REF store.
12757 2018-06-04 Richard Biener <rguenther@suse.de>
12759 PR tree-optimization/85955
12760 * builtins.c (fold_builtin_sincos): Convert pointers to
12761 destination to appropriate type before dereferencing.
12763 2018-06-04 Segher Boessenkool <segher@kernel.crashing.org>
12765 * config/rs6000/rs6000.md (abs<mode>2 for FLOAT128): Handle IFmode.
12767 2018-06-04 Richard Sandiford <richard.sandiford@linaro.org>
12769 * expr.c (expand_expr_real_1): Force the operand into memory if
12770 its TYPE_MODE is BLKmode and if there is no integer mode for
12771 the number of bits being extracted.
12773 2018-06-04 Jakub Jelinek <jakub@redhat.com>
12777 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
12778 Use vptestnm rather than vptestm in (=Yc,v,C) variant.
12780 2018-06-04 Richard Biener <rguenther@suse.de>
12782 * tree-cfgcleanup.c (cleanup_tree_cfg_1): Fold into...
12783 (cleanup_tree_cfg_noloop): ... single caller. Do
12784 start_recording_case_labels later.
12786 2018-06-04 Sebastian Peryt <sebastian.peryt@intel.com>
12788 * config/i386/cldemoteintrin.h: Change define from _X86INTRIN_H_INCLUDED
12789 to _IMMINTRIN_H_INCLUDED.
12790 * config/i386/pconfigintrin.h: Ditto.
12791 * config/i386/waitpkgintrin.h: Ditto.
12792 * config/i386/immintrin.h: Add includes for sgxintrin.h,
12793 pconfigintrin.h, waitpkgintrin.h and cldemoteintrin.h.
12794 * config/i386/x86intrin.h: Remove includes for mintrin.h, xmmintrin.h,
12795 emmintrin.h, pmmintrin.h, tmmintrin.h, smmintrin.h, wmmintrin.h,
12796 bmiintrin.h, bmi2intrin.h, lzcntintrin.h, sgxintrin.h, pconfigintrin.h,
12797 waitpkgintrin.h and cldemoteintrin.h.
12799 2018-06-04 Richard Biener <rguenther@suse.de>
12801 PR tree-optimization/86038
12802 * tracer.c (find_best_successor): Check probability for
12803 being initialized, bail out if not.
12805 2018-06-04 Richard Earnshaw <rearnsha@arm.com>
12808 * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
12809 of bits to ignore when comparing architectures.
12811 2018-06-04 Jakub Jelinek <jakub@redhat.com>
12813 PR tree-optimization/69615
12814 * fold-const.c (merge_ranges): If range1 is - [x, x] and x is the
12815 maximum or minimum of the type, try to merge it also as if
12816 range1 is + [-, x - 1] or + [x + 1, -].
12819 * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
12821 2018-06-03 Eric Botcazou <ebotcazou@adacore.com>
12823 PR tree-optimization/86034
12824 * gimple-ssa-store-merging.c (output_merged_store): Convert the RHS to
12825 the unsigned bitfield type in a bit insertion sequence if it does not
12826 have a larger precision than the bitfield size.
12827 (process_store): Also bypass widening conversions for BIT_INSERT_EXPR.
12829 2018-06-03 Kito Cheng <kito.cheng@gmail.com>
12831 * config/nds32/nds32-peephole2.md: Add new patterns for code size.
12833 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
12835 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3J.
12836 * config/nds32/nds32.c (nds32_option_override): Consider ARCH_V3J.
12837 * config/nds32/nds32.h (TARGET_ISA_V3): Add ARCH_V3J checking.
12838 * config/nds32/nds32.opt (march): Add enum value Value(ARCH_V3J).
12840 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
12842 * common/config/nds32/nds32-common.c (nds32_option_optimization_table):
12843 Disable -fdelete-null-pointer-checks for ELF toolchain.
12845 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
12846 Kito Cheng <kito.cheng@gmail.com>
12848 * config.gcc (nds32*): Use nds32-linux.opt and nds32-elf.opt.
12849 (nds32le-*-*, nds32be-*-*): Integrate checking process.
12850 (nds32*-*-*): Add glibc and uclibc conditions.
12851 * common/config/nds32/nds32-common.c (nds32_except_unwind_info): New.
12852 (TARGET_EXCEPT_UNWIND_INFO): Define.
12853 * config/nds32/elf.h: New file.
12854 * config/nds32/linux.h: New file.
12855 * config/nds32/nds32-elf.opt: New file.
12856 * config/nds32/nds32-linux.opt: New file.
12857 * config/nds32/nds32-fp-as-gp.c
12858 (pass_nds32_fp_as_gp::gate): Consider TARGET_LINUX_ABI.
12859 * config/nds32/nds32.c (nds32_conditional_register_usage): Consider
12861 (nds32_asm_file_end): Ditto.
12862 (nds32_print_operand): Ditto.
12863 (nds32_insert_attributes): Ditto.
12864 (nds32_init_libfuncs): New function.
12865 (TARGET_HAVE_TLS): Define.
12866 (TARGET_INIT_LIBFUNCS): Define.
12867 * config/nds32/nds32.h (TARGET_DEFAULT_RELAX): Apply different relax
12869 (TARGET_ELF): Apply different mcmodel setting.
12870 (LINK_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): The content has
12871 been migrated into elf.h and linux.h files.
12872 * config/nds32/nds32.md (add_pc): Consider TARGET_LINUX_ABI.
12873 * config/nds32/nds32.opt (mvh): Consider TARGET_LINUX_ABI.
12874 (mcmodel): The content has been migrated into nds32-elf.opt and
12875 nds32-linux.opt files.
12876 * config/nds32/t-elf: New file.
12877 * config/nds32/t-linux: New file.
12879 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
12880 Shiva Chen <shiva0217@gmail.com>
12882 * config/nds32/constants.md (unspec_volatile_element): Add
12883 UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END.
12884 * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp
12886 * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare.
12887 (make_pass_nds32_fp_as_gp): Declare.
12888 * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one
12890 (nds32_asm_function_end_prologue): Remove unused asm output.
12891 (nds32_asm_function_begin_epilogue): Remove unused asm output.
12892 (nds32_asm_file_start): Output necessary fp_as_gp information.
12893 (nds32_option_override): Adjust register usage.
12894 (nds32_expand_prologue): Consider fp_as_gp situation.
12895 (nds32_expand_prologue_v3push): Consider fp_as_gp situation.
12896 * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p.
12899 (simple_return): Ditto.
12900 (omit_fp_begin): Output special directive for fp_as_gp.
12901 (omit_fp_end): Output special directive for fp_as_gp.
12902 * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp,
12903 mforbid-fp-as-gp): New options.
12905 2018-06-01 Mark Wielaard <mark@klomp.org>
12907 * dwarf2out.c (dwarf2out_finish): Remove generation of
12908 DW_AT_loclists_base.
12910 2018-06-01 Eric Botcazou <ebotcazou@adacore.com>
12912 * gimple-ssa-store-merging.c: Include gimple-fold.h.
12913 (struct store_immediate_info): Document BIT_INSERT_EXPR stores.
12914 (struct merged_store_group): Add bit_insertion field.
12915 (dump_char_array): Use standard hexadecimal format.
12916 (merged_store_group::merged_store_group): Set bit_insertion to false.
12917 (merged_store_group::apply_stores): Use optimal buffer size. Deal
12918 with BIT_INSERT_EXPR stores. Move up code updating the mask and
12919 also print the mask in the dump file.
12920 (pass_store_merging::gate): Minor tweak.
12921 (imm_store_chain_info::coalesce_immediate): Fix wrong association
12922 of stores with groups in dump. Allow coalescing of BIT_INSERT_EXPR
12923 stores with INTEGER_CST stores.
12924 (count_multiple_uses) <BIT_INSERT_EXPR>: New case.
12925 (imm_store_chain_info::output_merged_store): Add try_bitpos variable
12926 and use it throughout. Generate bit insertion sequences if need be.
12927 (pass_store_merging::process_store): Remove redundant condition.
12928 Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR.
12930 2018-06-01 Segher Boessenkool <segher@kernel.crashing.org>
12932 * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
12933 the 128-bit floating point types. Fix function comment.
12935 2018-06-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12937 * config/aarch64/aarch64-simd.md
12938 (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler
12940 (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
12943 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
12945 PR tree-optimization/85989
12946 * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
12948 (backprop::intersect_uses): Check it when deciding whether this
12949 is a backedge reference.
12950 (backprop::process_block): Add each phi to m_visited_phis
12951 after visiting it, then clear it at the end.
12953 2018-06-01 Richard Biener <rguenther@suse.de>
12955 * tree-vectorizer.h (vect_dr_stmt): New function.
12956 (vect_get_load_cost): Adjust.
12957 (vect_get_store_cost): Likewise.
12958 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
12959 Use vect_dr_stmt instead of DR_SMTT.
12960 (vect_record_base_alignments): Likewise.
12961 (vect_calculate_target_alignment): Likewise.
12962 (vect_compute_data_ref_alignment): Likewise and make static.
12963 (vect_update_misalignment_for_peel): Likewise.
12964 (vect_verify_datarefs_alignment): Likewise.
12965 (vector_alignment_reachable_p): Likewise.
12966 (vect_get_data_access_cost): Likewise. Pass down
12967 vinfo to vect_get_load_cost/vect_get_store_cost instead of DR.
12968 (vect_get_peeling_costs_all_drs): Likewise.
12969 (vect_peeling_hash_get_lowest_cost): Likewise.
12970 (vect_enhance_data_refs_alignment): Likewise.
12971 (vect_find_same_alignment_drs): Likewise.
12972 (vect_analyze_data_refs_alignment): Likewise.
12973 (vect_analyze_group_access_1): Likewise.
12974 (vect_analyze_group_access): Likewise.
12975 (vect_analyze_data_ref_access): Likewise.
12976 (vect_analyze_data_ref_accesses): Likewise.
12977 (vect_vfa_segment_size): Likewise.
12978 (vect_small_gap_p): Likewise.
12979 (vectorizable_with_step_bound_p): Likewise.
12980 (vect_prune_runtime_alias_test_list): Likewise.
12981 (vect_analyze_data_refs): Likewise.
12982 (vect_supportable_dr_alignment): Likewise.
12983 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
12984 (vect_gen_prolog_loop_niters): Likewise.
12985 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
12986 * tree-vect-patterns.c (vect_recog_bool_pattern): Do not
12988 (vect_recog_mask_conversion_pattern): Likewise.
12989 (vect_try_gather_scatter_pattern): Likewise.
12990 * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info
12991 to vect_get_store_cost.
12992 (vect_get_store_cost): Get stmt_info instead of DR.
12993 (vect_model_load_cost): Pass stmt_info to vect_get_load_cost.
12994 (vect_get_load_cost): Get stmt_info instead of DR.
12996 2018-06-01 Richard Biener <rguenther@suse.de>
12998 PR middle-end/86017
12999 * gimple-fold.c (var_decl_component_p): Also allow offsetted
13000 vars wrapped in MEM_REFs.
13002 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
13004 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
13005 Fix subreg tests so that we only return a choice between
13006 GENERAL_REGS and FP_REGS if the original classes included both.
13008 2018-06-01 Richard Biener <rguenther@suse.de>
13011 * tree-ssa-structalias.c (get_function_part_constraint):
13012 Handle NULL fi->decl.
13013 (find_func_aliases_for_call): Properly handle indirect
13014 fi from direct call.
13015 (find_func_clobbers): Likewise.
13016 (ipa_pta_execute): Likewise.
13017 (create_variable_info_for): For functions that are ifunc_resolver
13018 resolve to a varinfo that contains the result of the resolver call.
13019 (associate_varinfo_to_alias): Do not treat ifunc resolvers as
13022 2018-05-31 Michael Collison <michael.collison@arm.com>
13024 * config/aarch64/aarch64.md:
13025 (*fix_to_zero_extenddfdi2): New pattern.
13026 * gcc.target/aarch64/fix_extend1.c: New testcase.
13028 2018-05-31 Qing Zhao <qing.zhao@oracle.com>
13030 PR middle-end/78809
13031 PR middle-end/83026
13032 * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
13033 and BUILT_IN_STRNCMP_EQ.
13034 * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
13035 BUILT_IN_STRNCMP_EQ.
13036 * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
13037 handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
13038 (gimple_fold_builtin): Likewise.
13039 * tree-ssa-strlen.c (compute_string_length): New function.
13040 (determine_min_obsize): New function.
13041 (handle_builtin_string_cmp): New function to handle calls to
13042 string compare functions.
13043 (strlen_optimize_stmt): Add handling to builtin string compare
13045 * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
13046 Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
13047 * tree.c (build_common_builtin_nodes): Add new defines of
13048 BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
13050 2018-05-31 Jakub Jelinek <jakub@redhat.com>
13053 * bb-reorder.c (pass_partition_blocks::gate): Return false for
13054 functions with naked attribute.
13056 2018-05-31 Uros Bizjak <ubizjak@gmail.com>
13058 * config/i386/sse.md (avx_vec_concat<mode>):
13059 Substitute concat_tg_mode mode attribute with xtg_mode.
13060 (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
13061 (concat_tg_mode): Remove mode attribute.
13063 2018-05-31 Martin Sebor <msebor@redhat.com>
13066 * calls.c (alloc_max_size): Correct a logic error/typo.
13067 Treat excessive arguments as infinite. Warn for invalid arguments.
13068 * doc/invoke.texi (-Walloc-size-larger-than): Update.
13070 2018-05-31 H.J. Lu <hongjiu.lu@intel.com>
13073 * config/i386/x86-tune.def: Re-enable partial_reg_dependency
13074 and movx for Haswell.
13076 2018-05-31 Chung-Lin Tang <cltang@codesourcery.com>
13077 Cesar Philippidis <cesar@codesourcery.com>
13079 PR middle-end/85879
13080 * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
13081 when emitting error on private/firstprivate reductions.
13082 * omp-low.c (lower_omp_target): Avoid reference-type processing
13083 on pointers for firstprivate clause.
13085 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
13087 * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
13090 * config/aarch64/aarch64-simd.md
13091 (aarch64_ld1x3<VALLDIF:mode>): New pattern.
13092 (aarch64_ld1_x3_<mode>): Likewise
13093 (aarch64_st1x2<VALLDIF:mode>): Likewise
13094 (aarch64_st1_x2_<mode>): Likewise
13095 (aarch64_st1x3<VALLDIF:mode>): Likewise
13096 (aarch64_st1_x3_<mode>): Likewise
13097 * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
13098 (vld1_s8_x3): Likewise.
13099 (vld1_u16_x3): Likewise.
13100 (vld1_s16_x3): Likewise.
13101 (vld1_u32_x3): Likewise.
13102 (vld1_s32_x3): Likewise.
13103 (vld1_u64_x3): Likewise.
13104 (vld1_s64_x3): Likewise.
13105 (vld1_f16_x3): Likewise.
13106 (vld1_f32_x3): Likewise.
13107 (vld1_f64_x3): Likewise.
13108 (vld1_p8_x3): Likewise.
13109 (vld1_p16_x3): Likewise.
13110 (vld1_p64_x3): Likewise.
13111 (vld1q_u8_x3): Likewise.
13112 (vld1q_s8_x3): Likewise.
13113 (vld1q_u16_x3): Likewise.
13114 (vld1q_s16_x3): Likewise.
13115 (vld1q_u32_x3): Likewise.
13116 (vld1q_s32_x3): Likewise.
13117 (vld1q_u64_x3): Likewise.
13118 (vld1q_s64_x3): Likewise.
13119 (vld1q_f16_x3): Likewise.
13120 (vld1q_f32_x3): Likewise.
13121 (vld1q_f64_x3): Likewise.
13122 (vld1q_p8_x3): Likewise.
13123 (vld1q_p16_x3): Likewise.
13124 (vld1q_p64_x3): Likewise.
13125 (vst1_s64_x2): Likewise.
13126 (vst1_u64_x2): Likewise.
13127 (vst1_f64_x2): Likewise.
13128 (vst1_s8_x2): Likewise.
13129 (vst1_p8_x2): Likewise.
13130 (vst1_s16_x2): Likewise.
13131 (vst1_p16_x2): Likewise.
13132 (vst1_s32_x2): Likewise.
13133 (vst1_u8_x2): Likewise.
13134 (vst1_u16_x2): Likewise.
13135 (vst1_u32_x2): Likewise.
13136 (vst1_f16_x2): Likewise.
13137 (vst1_f32_x2): Likewise.
13138 (vst1_p64_x2): Likewise.
13139 (vst1q_s8_x2): Likewise.
13140 (vst1q_p8_x2): Likewise.
13141 (vst1q_s16_x2): Likewise.
13142 (vst1q_p16_x2): Likewise.
13143 (vst1q_s32_x2): Likewise.
13144 (vst1q_s64_x2): Likewise.
13145 (vst1q_u8_x2): Likewise.
13146 (vst1q_u16_x2): Likewise.
13147 (vst1q_u32_x2): Likewise.
13148 (vst1q_u64_x2): Likewise.
13149 (vst1q_f16_x2): Likewise.
13150 (vst1q_f32_x2): Likewise.
13151 (vst1q_f64_x2): Likewise.
13152 (vst1q_p64_x2): Likewise.
13153 (vst1_s64_x3): Likewise.
13154 (vst1_u64_x3): Likewise.
13155 (vst1_f64_x3): Likewise.
13156 (vst1_s8_x3): Likewise.
13157 (vst1_p8_x3): Likewise.
13158 (vst1_s16_x3): Likewise.
13159 (vst1_p16_x3): Likewise.
13160 (vst1_s32_x3): Likewise.
13161 (vst1_u8_x3): Likewise.
13162 (vst1_u16_x3): Likewise.
13163 (vst1_u32_x3): Likewise.
13164 (vst1_f16_x3): Likewise.
13165 (vst1_f32_x3): Likewise.
13166 (vst1_p64_x3): Likewise.
13167 (vst1q_s8_x3): Likewise.
13168 (vst1q_p8_x3): Likewise.
13169 (vst1q_s16_x3): Likewise.
13170 (vst1q_p16_x3): Likewise.
13171 (vst1q_s32_x3): Likewise.
13172 (vst1q_s64_x3): Likewise.
13173 (vst1q_u8_x3): Likewise.
13174 (vst1q_u16_x3): Likewise.
13175 (vst1q_u32_x3): Likewise.
13176 (vst1q_u64_x3): Likewise.
13177 (vst1q_f16_x3): Likewise.
13178 (vst1q_f32_x3): Likewise.
13179 (vst1q_f64_x3): Likewise.
13180 (vst1q_p64_x3): Likewise.
13182 2018-05-30 Jozef Lawrynowicz <jozef.l@mittosystems.com>
13184 * config/msp430/msp430.c (msp430_output_labelref): Prepend
13185 user_label_prefix to name.
13187 * tree-core.h: Update comment about the format of NAME string
13188 passed to handler in attribute_spec.
13190 * config/msp430/msp430.md: Remove erroneous subreg expression from
13191 zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
13192 zero_extend{q,h}isi2.
13194 2018-05-30 Borislav Petkov <bp@suse.de>
13196 * doc/extend.texi: Document some architecture specific
13197 constraints and sort entries.
13199 2018-05-30 Martin Sebor <msebor@redhat.com>
13201 PR middle-end/85369
13202 * builtins.c (expand_builtin_stpcpy_1): New function.
13203 (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
13204 only if the former succeeds.
13206 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
13208 * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
13211 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
13213 * doc/invoke.texi (-flinker-output): Document
13215 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
13217 * passes.c (ipa_write_summaries): Only modify statements if body
13219 * cgraphunit.c (ipa_passes): Also produce intermeidate code when
13220 incrementally linking.
13221 (ipa_passes): Likewise.
13222 * lto-cgraph.c (lto_output_node): When incrementally linking do not
13223 pass down resolution info.
13224 * common.opt (flag_incremental_link): Update info.
13225 * gcc.c (plugin specs): Turn flinker-output=* to
13226 -plugin-opt=-linker-output-known
13227 * toplev.c (compile_file): Also cut compilation when doing incremental
13229 * flag-types. (enum lto_partition_model): Add
13230 LTO_LINKER_OUTPUT_NOLTOREL.
13231 (invoke.texi): Add -flinker-output docs.
13232 * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
13233 link same way as WPA; do not stream in dead initializers.
13235 * dwarf2out.c (dwarf2out_die_ref_for_decl,
13236 darf2out_register_external_decl): Support incremental link.
13238 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
13240 * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
13242 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
13244 * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
13245 it down to simple_object_copy_lto_debug_sections.
13246 (run_gcc): Determine incremental LTO link time and configure
13247 lto1 into non-wpa mode, disable renaming of debug sections.
13249 2018-05-30 Kelvin Nilsen <kelvin@gcc.gnu.org>
13251 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
13252 descriptions of various incorrectly documented functions.
13254 2018-05-30 Andre Vieira <andre.simoesdiasvieira@arm.com>
13257 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
13258 address check not strict.
13260 2018-05-30 Richard Biener <rguenther@suse.de>
13262 PR tree-optimization/85964
13263 * tracer.c (better_p): Drop initialized count check, we only
13264 call the function with initialized counts now.
13265 (find_best_successor): Do find a best edge if one
13266 has uninitialized count.
13267 (find_best_predecessor): Likewise. Do BB frequency check only
13268 if count is initialized.
13270 2017-05-30 Jackson Woodruff <jackson.woodruff@arm.com>
13272 * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
13273 (aarch64_ldrstr_offset_compare): New.
13274 (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
13275 load/store orderings.
13276 (aarch64_gen_adjusted_ldpstp): Likewise.
13278 2018-05-30 Wilco Dijkstra <wdijkstr@arm.com>
13280 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
13281 Check for subset of GENERAL_REGS and FP_REGS.
13282 * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
13285 2018-05-30 Richard Sandiford <richard.sandiford@linaro.org>
13287 * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
13288 and wi::to_poly_offset. Add the current offset and then check
13289 whether the sum fits, rather than using an unchecked addition of
13290 a checked term. Check for a shwi rather than a uhwi.
13291 * expr.c (get_bit_range): Use tree_to_poly_uint64.
13292 (store_constructor): Use poly_int_tree_p.
13293 (expand_expr_real_1): Likewise.
13294 * function.c (assign_temp): Likewise.
13295 * fold-const.c (const_binop): Use poly_int_tree_p and
13296 wi::to_poly_offset.
13297 (fold_indirect_ref_1): Likewise. Use multiple_p to attempt an exact
13299 * ipa-icf-gimple.c (func_checker::compare_operand): Use
13300 to_poly_offset for MEM offsets.
13301 * ipa-icf.c (sem_variable::equals): Likewise.
13302 * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
13303 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
13304 wi::to_poly_offset for BIT_FIELD_REF offsets.
13305 (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
13306 wi::to_poly_offset.
13307 * var-tracking.c (emit_note_insn_var_location): Use
13308 tree_to_poly_uint64.
13310 2018-05-29 Jim Wilson <jimw@sifive.com>
13312 * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
13314 2018-05-29 Uros Bizjak <ubizjak@gmail.com>
13317 * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
13318 Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
13320 (sse4_1_round<mode>2): Use nonimmediate_operand
13321 for operand 1 predicate.
13323 2018-05-29 Martin Sebor <msebor@redhat.com>
13324 Richard Biener <rguenther@suse.de>
13327 * calls.c (get_size_range): Call determine_value_range instead
13328 of get_value_range..
13329 * tree-vrp.h (determine_value_range): Declared new function.
13330 * tree-vrp.c (determine_value_range_1, determine_value_range): New.
13332 2018-05-29 Richard Biener <rguenther@suse.de>
13334 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
13335 sure to use non-pattern stmts for get_earlier_stmt arguments.
13336 * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
13337 called on pattern stmts.
13338 (get_later_stmt): Likewise.
13340 2018-05-29 Martin Liska <mliska@suse.cz>
13342 PR gcov-profile/85759
13343 * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
13346 2018-05-29 Jakub Jelinek <jakub@redhat.com>
13348 * tree-cfg.c (verify_gimple_assign_unary): Add checking for
13350 (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
13354 * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
13355 VEC_PACK_FLOAT_EXPR): New tree codes.
13356 * tree-pretty-print.c (op_code_prio): Handle
13357 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
13358 (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
13359 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
13360 * tree-inline.c (estimate_operator_cost): Likewise.
13361 * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
13362 * fold-const.c (const_binop): Likewise.
13363 (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
13364 VEC_UNPACK_FIX_TRUNC_LO_EXPR.
13365 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
13366 (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
13367 * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
13368 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
13369 * expr.c (expand_expr_real_2): Likewise.
13370 * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
13371 vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
13372 vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
13374 * optabs.c (expand_widen_pattern_expr): For
13375 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
13376 sign from result type rather than operand's type.
13377 (expand_binop_directly): For vec_packu_float_optab and
13378 vec_packs_float_optab allow result type to be different from operand's
13380 * optabs-tree.c (optab_for_tree_code): Handle
13381 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
13382 VEC_PACK_FLOAT_EXPR. Formatting fixes.
13383 * tree-vect-generic.c (expand_vector_operations_1): Handle
13384 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
13385 VEC_PACK_FLOAT_EXPR.
13386 * tree-vect-stmts.c (supportable_widening_operation): Handle
13388 (supportable_narrowing_operation): Handle FLOAT_EXPR.
13389 * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
13390 * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
13391 (float<floatunssuffix>v2div2sf2): ... this. Formatting fix.
13392 (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
13394 (vec_pack<floatprefix>_float_<mode>): New expander.
13395 (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
13397 (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
13398 vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
13399 * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
13400 vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
13401 vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
13403 * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
13404 VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
13405 (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
13406 VEC_PACK_FLOAT_EXPR): Document.
13408 2018-05-29 Richard Biener <rguenther@suse.de>
13410 * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
13412 (stmt_vec_info_vec): Make pointer.
13413 (init_stmt_vec_info_vec): Remove.
13414 (free_stmt_vec_info_vec): Likewise.
13415 (set_stmt_vec_info_vec): New function.
13416 (free_stmt_vec_infos): Likewise.
13417 (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
13418 (set_vinfo_for_stmt): Likewise.
13419 (get_earlier_stmt): Likewise.
13420 (get_later_stmt): Likewise.
13421 * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
13422 (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
13423 (vec_info::~vec_info): Free stmt_vec_infos.
13424 (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
13425 Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
13426 (pass_slp_vectorize::execute): Likewise.
13427 * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
13428 (free_stmt_vec_info_vec): Likewise.
13429 (set_stmt_vec_info_vec): New function.
13430 (free_stmt_vec_infos): Likewise.
13431 * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
13432 the global stmt_vec_info_vec.
13433 * tree-parloops.c (gather_scalar_reductions): Use
13434 set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
13437 2018-05-29 Richard Biener <rguenther@suse.de>
13439 * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
13441 2018-05-29 Martin Liska <mliska@suse.cz>
13442 David Malcolm <dmalcolm@redhat.com>
13444 * vec.c (test_reverse): New.
13445 (vec_c_tests): Add new test.
13446 * vec.h (vl_ptr>::reverse): New function.
13448 2018-05-29 Gerald Pfeifer <gerald@pfeifer.com>
13450 * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
13452 * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
13455 2018-05-28 Bernd Edlinger <bernd.edlinger@hotmail.de>
13457 * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
13459 2018-05-28 Richard Biener <rguenther@suse.de>
13461 PR tree-optimization/85933
13462 * tree-vect-data-refs.c (vect_record_base_alignments): Only
13463 look at stmts marked as vectorizable.
13465 2018-05-28 Richard Biener <rguenther@suse.de>
13467 PR tree-optimization/85934
13468 * tree-vect-generic.c (expand_vector_operations_1): Hoist
13469 vector boolean check before scalar optimization.
13471 2018-05-28 Jakub Jelinek <jakub@redhat.com>
13473 * doc/invoke.texi (ARM Options): Use @item instead of @itemx
13476 2018-05-28 Mark Wielaard <mark@klomp.org>
13478 * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
13479 if it is an expression containing a minus sign.
13481 2018-05-27 John David Anglin <danglin@gcc.gnu.org>
13483 * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
13485 2018-05-27 Paul Koning <ni1d@arrl.net>
13487 * config/pdp11/pdp11.md (truncsihi2): Remove.
13489 2018-05-27 Monk Chiang <sh.chiang04@gmail.com>
13490 Chung-Ju Wu <jasonwucj@gmail.com>
13492 * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
13494 (unaligned_store_dw): Ditto.
13495 * config/nds32/nds32-memory-manipulation.c
13496 (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
13497 (nds32_gen_dup_4_byte_to_word_value): Rename to ...
13498 (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
13499 (emit_setmem_word_loop): Rename to ...
13500 (emit_setmem_doubleword_loop): ... this.
13501 (nds32_gen_dup_4_byte_to_word_value): New function.
13502 (nds32_gen_dup_8_byte_to_double_word_value): New function.
13503 (nds32_expand_setmem_loop): Refine implementation.
13504 (nds32_expand_setmem_loop_v3m): Ditto.
13505 * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
13508 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
13510 * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
13512 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
13514 * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
13515 (nds32_init_machine_status): Initialize machine->attr_naked_p and
13516 machine->attr_no_prologue_p.
13517 (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
13518 (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
13519 (nds32_expand_epilogue): Consider attr_naked_p.
13520 (nds32_expand_epilogue_v3pop): Likewise.
13521 (nds32_can_use_return_insn): Likewise.
13522 * config/nds32/nds32.h (machine_function): Add attr_naked_p and
13523 attr_no_prologue_p fields.
13524 * config/nds32/nds32.opt (mret-in-naked-func): New option.
13526 2018-05-27 Jakub Jelinek <jakub@redhat.com>
13529 * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
13531 * config/i386/sse.md
13532 (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
13534 (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
13536 (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
13538 (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
13540 (*<floatsuffix>floatv2div2sf2): Rename to ...
13541 (*float<floatunssuffix>v2div2sf2): ... this.
13542 (<floatsuffix>floatv2div2sf2_mask): Rename to ...
13543 (float<floatunssuffix>v2div2sf2_mask): ... this.
13544 (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
13545 (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
13546 (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
13548 (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
13550 (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
13552 (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
13554 (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
13556 (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
13558 (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
13559 (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
13560 (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
13561 gen_ufix_truncv8dfv8si2.
13562 * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
13563 __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
13564 __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
13565 __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
13566 __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
13567 __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
13568 __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
13569 Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
13571 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
13575 * varasm.c (assemble_alias): Lookup ifunc attribute on error.
13577 2018-05-25 Jim Wilson <jimw@sifive.com>
13579 * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
13580 * config/riscv/riscv.c (struct machine_function): Add
13581 interrupt_handler_p and attribute_checked_p fields.
13582 (riscv_attribute_table): Add interrupt.
13583 (riscv_interrupt_type_p): New.
13584 (riscv_save_reg_p): Save extra regs for interrupt handler.
13585 (riscv_use_save_libcall): Return false for interrupt handler.
13586 (riscv_first_stack_step): Add forward declaration.
13587 (riscv_compute_frame_info): New local interrupt_save_t1. Set it
13588 for interrupt handler with large frame. Use it for saved reg list.
13589 (riscv_expand_prologue): Move flag_stack_usage_info support to
13590 eliminate duplication.
13591 (riscv_expand_epilogue): Generate mret for interrupt handler.
13592 (riscv_epilogue_uses): New.
13593 (riscv_can_use_return_insn): Return false for interrupt handler.
13594 (riscv_function_ok_for_sibcall): Likewise.
13595 (riscv_set_current_function): Add interrupt handler support.
13596 * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
13597 * config/riscv/riscv.md (UNSPECV_MRET): New.
13599 (riscv_frflags, riscv_fsflags): Use tab after opcode.
13601 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
13603 2018-05-25 Bill Schmidt <wschmidt@linux.ibm.com>
13605 PR tree-optimization/85712
13606 * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
13607 this candidate has already been replaced in-situ by a copy.
13609 2018-05-25 Jason Merrill <jason@redhat.com>
13611 PR c++/80485 - inline function non-zero address.
13612 * symtab.c (nonzero_address): Check DECL_COMDAT.
13614 2018-05-25 Uros Bizjak <ubizjak@gmail.com>
13617 * config/alpha/alpha.md (ashlsi3): New insn pattern.
13618 (*ashlsi_se): Rename from *ashldi_se. Define as sign
13619 extension of SImode operation. Use const123_operand predicate.
13620 (*saddsi_1): Remove.
13621 (*saddl_se_1): Ditto.
13622 (*ssubsi_1): Ditto.
13623 (*ssubl_se_1): Ditto.
13624 * config/alpha/predicates.md (const123_operand): New predicate.
13625 * config/alpha/constraints.md (P): Use IN_RANGE.
13627 2018-05-25 Richard Biener <rguenther@suse.de>
13629 * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
13631 (ref_maybe_used_by_stmt_p): Likewise.
13632 (stmt_may_clobber_ref_p): Likewise.
13633 (stmt_may_clobber_ref_p_1): Likewise.
13634 * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
13636 (ref_maybe_used_by_stmt_p): Likewise.
13637 (stmt_may_clobber_ref_p): Likewise.
13638 (stmt_may_clobber_ref_p_1): Likewise.
13639 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
13640 the alias oracle to disambiguate DRs with stmts DR analysis
13642 (vect_analyze_data_refs): Do not give up on not analyzable
13643 DRs for BB vectorization. Remove code truncating the dataref
13646 2018-05-25 Jakub Jelinek <jakub@redhat.com>
13649 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
13650 Add (=Yk,v,C) variant using vptestm insn. Use TARGET_AVX512BW
13651 in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
13653 2018-05-25 Richard Biener <rguenther@suse.de>
13655 * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
13656 function, combining stmt data ref gathering and fatal analysis
13658 (vect_analyze_data_refs): Remove now redudnant code and simplify.
13659 * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
13660 vect_analyze_loop_2 and use vect_find_stmt_data_reference.
13661 * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
13662 * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
13664 2018-05-25 Bin Cheng <bin.cheng@arm.com>
13666 PR tree-optimization/85720
13667 * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
13668 SCC if all partitions are builtins.
13669 (version_loop_by_alias_check): New parameter. Generate cancelable
13670 runtime alias check if all partitions are builtins.
13671 (distribute_loop): Update call to above function.
13673 2018-05-25 Bin Cheng <bin.cheng@arm.com>
13675 * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
13676 (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
13677 (parm_default_def_partition_arg): Ditto.
13678 (set_parm_default_def_partition): Ditto.
13679 (get_parm_default_def_partitions): Ditto and make it static.
13680 (get_undefined_value_partitions): Ditto and make it static.
13681 (remove_ssa_form): Refactor call to init_var_map here.
13682 * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
13683 computation for loop region.
13684 (coalesce_partitions, compute_optimized_partition_bases): Ditto.
13685 (register_default_def): Delete.
13686 (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
13687 (parm_default_def_partition_arg): Ditto.
13688 (set_parm_default_def_partition): Ditto.
13689 (get_parm_default_def_partitions): Ditto and make it static.
13690 (get_undefined_value_partitions): Ditto and make it static.
13691 (coalesce_with_default, coalesce_with_default): Update comment.
13692 (create_coalesce_list_for_region): New func factored out from
13693 create_outofssa_var_map.
13694 (populate_coalesce_list_for_outofssa): New func factored out from
13695 create_outofssa_var_map and coalesce_ssa_name.
13696 (create_outofssa_var_map): Delete.
13697 (coalesce_ssa_name): Refactor to support live range computation.
13698 * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
13699 (get_parm_default_def_partitions): Delete.
13700 (get_undefined_value_partitions): Ditto.
13701 * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
13702 computation for loop region.
13703 (new_tree_live_info, loe_visit_block): Ditto.
13704 (live_worklist, set_var_live_on_entry): Ditto.
13705 (calculate_live_on_exit, verify_live_on_entry): Ditto.
13706 * tree-ssa-live.h (struct _var_map): New fields.
13707 (init_var_map): Change decl.
13708 (region_contains_p): New.
13710 2018-05-25 Bin Cheng <bin.cheng@arm.com>
13712 * tree-ssa-live.h (live_merge_and_clear): Delete.
13714 2018-05-25 Richard Biener <rguenther@suse.de>
13717 * tree-dump.c (dequeue_and_dump): Remove access to removed
13718 operand 2 of a SWITCH_EXPR.
13720 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
13722 * doc/sourcebuild.texi (vect_double_cond_arith): Include
13723 multiplication and division.
13724 * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
13725 (cond_udiv@var{m}, cond_umod@var{m}): Document.
13726 * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
13727 (cond_udiv_optab, cond_umod_optab): New optabs.
13728 * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
13729 (IFN_COND_RDIV): New internal functions.
13730 * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
13731 TRUNC_MOD_EXPR and RDIV_EXPR.
13732 * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
13733 * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
13735 (SVE_INT_BINARY): Include mult.
13736 (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
13737 (optab, sve_int_op): Handle mult.
13738 (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
13740 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
13741 for SVE_INT_BINARY_SD.
13743 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
13745 * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
13746 (optab, sve_int_op): Handle div and udiv.
13747 * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
13748 for SVE_INT_BINARY_SD.
13749 (*<optab><mode>3): New insn for the same.
13751 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
13753 * tree-vect-patterns.c: Include predict.h.
13754 (vect_recog_divmod_pattern): Restrict check for division support
13755 to when optimizing for size.
13757 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
13759 * doc/sourcebuild.texi (vect_double_cond_arith: Document.
13760 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
13761 (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
13762 (gimple_match_op::set_op): Likewise.
13763 (gimple_resimplify4): Declare.
13764 * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
13765 (expr::gen_transform): Likewise.
13766 (decision_tree::gen): Generate a simplification routine for 4 operands.
13767 * gimple-match-head.c (gimple_simplify): Add an overload for
13768 4 operands. In the top-level function, handle up to 4 call
13769 arguments and call gimple_resimplify4.
13770 (gimple_resimplify4): New function.
13771 (build_call_internal): Pass a fourth operand.
13772 (maybe_push_to_seq): Likewise.
13773 * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
13774 Fold VEC_COND_EXPRs of an operation and a default value into
13775 an IFN_COND_* function if possible.
13776 * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
13778 (SVE_COND_FP_BINARY): Include them.
13779 (optab, sve_fp_op): Handle them.
13780 (SVE_INT_BINARY_REV): New code iterator.
13781 (SVE_COND_FP_BINARY_REV): New int iterator.
13782 (commutative): New int attribute.
13783 * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
13785 * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
13787 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
13788 (*cond_<optab><mode>): New patterns for reversed operands.
13790 2018-05-25 Richard Biener <rguenther@suse.de>
13792 * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
13793 (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
13794 (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
13795 (STMT_VINFO_GROUPED_ACCESS): Adjust.
13796 * tree-vect-data-refs.c (everywhere): Adjust users.
13797 * tree-vect-loop.c (everywhere): Likewise.
13798 * tree-vect-slp.c (everywhere): Likewise.
13799 * tree-vect-stmts.c (everywhere): Likewise.
13800 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
13802 2018-05-25 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
13804 * configure.ac (gcc_cv_as_section_has_e): Move to common section.
13806 (gcc_cv_as_section_exclude): ... this.
13807 Try Solaris as #exclude syntax.
13808 * configure: Regenerate.
13809 * config.in: Regenerate.
13810 * config/i386/i386.c (i386_solaris_elf_named_section): Handle
13812 * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
13813 [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
13815 * varasm.c (default_elf_asm_named_section): Don't check if
13816 HAVE_GAS_SECTION_EXCLUDE is defined.
13818 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
13820 * doc/md.texi: Update the documentation of the cond_* optabs
13821 to mention the new final operand. Fix GET_MODE_NUNITS call.
13822 Describe the scalar case too.
13823 * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
13824 * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
13826 (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
13827 (get_conditional_internal_fn): Update comment.
13828 * tree-vect-loop.c (vectorizable_reduction): Pass the original
13829 accumulator value as a final argument to conditional functions.
13830 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
13831 a define_expand and add an "else" operand. Assert for now that
13832 the else operand is equal to operand 2. Use SVE_INT_BINARY and
13833 SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
13834 (*cond_<optab><mode>): New patterns.
13835 * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
13836 (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
13837 (UNSPEC_COND_EOR): Delete.
13838 (optab): Remove associated mappings.
13839 (SVE_INT_BINARY): New code iterator.
13840 (sve_int_op): Remove int attribute and add "minus" to the code
13842 (SVE_COND_INT_OP): Delete.
13843 (SVE_COND_FP_OP): Rename to...
13844 (SVE_COND_FP_BINARY): ...this.
13846 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
13848 * optabs.c (can_reuse_operands_p): New function.
13849 (maybe_legitimize_operands): Try to reuse the results for
13852 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
13854 * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
13855 Add {q} suffix to insn mnemonic.
13857 2018-05-23 Jozef Lawrynowicz <jozef.l@mittosystems.com>
13859 * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
13860 (msp430_warn_func_return): New.
13862 2018-05-24 Roger Sayle <roger@nextmovesoftware.com>
13864 * fold-const.c (tree_nonzero_bits): New function.
13865 * fold-const.h (tree_nonzero_bits): Likewise.
13866 * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
13867 friends. POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
13869 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
13873 * varasm.c (assemble_alias): Check ifunc_resolver only on
13876 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
13879 * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
13880 when memory input operand is handled.
13882 2018-05-24 Luis Machado <luis.machado@linaro.org>
13884 * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
13886 (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
13888 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
13890 * match.pd: Delay FMA folds until after vectorization.
13892 2018-05-24 Andre Vieira <andre.simoesdiasvieira@arm.com>
13895 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
13896 address check not strict.
13898 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
13900 * gimple-match.h (gimple_match_op): New class.
13901 (mprts_hook): Replace parameters with a gimple_match_op *.
13902 (maybe_build_generic_op): Likewise.
13903 (gimple_simplified_result_is_gimple_val): Replace parameters with
13904 a const gimple_match_op *.
13905 (gimple_simplify): Replace code_helper * and tree * parameters with
13906 a gimple_match_op * parameter.
13907 (gimple_resimplify1): Replace code_helper *, tree and tree *
13908 parameters with a gimple_match_op * parameter.
13909 (gimple_resimplify2): Likewise.
13910 (gimple_resimplify3): Likewise.
13911 (maybe_push_res_to_seq): Replace code_helper, tree and tree *
13912 parameters with a gimple_match_op * parameter.
13913 * gimple-match-head.c (gimple_simplify): Change prototypes of
13914 auto-generated functions to take a gimple_match_op * instead of
13915 separate code_helper * and tree * parameters. Make the same
13916 change in the top-level overload and update calls to the
13917 gimple_resimplify routines. Update calls to the auto-generated
13918 functions and to maybe_push_res_to_seq in the publicly-facing
13919 operation-specific gimple_simplify overloads.
13920 (gimple_match_op::MAX_NUM_OPS): Define.
13921 (gimple_resimplify1): Replace rcode and ops with a single res_op
13922 parameter. Update call to gimple_simplify.
13923 (gimple_resimplify2): Likewise.
13924 (gimple_resimplify3): Likewise.
13925 (mprts_hook): Replace parameters with a gimple_match_op *.
13926 (maybe_build_generic_op): Likewise.
13927 (build_call_internal): Replace type, nargs and ops with
13928 a gimple_match_op *.
13929 (maybe_push_res_to_seq): Replace res_code, type and ops parameters
13930 with a single gimple_match_op *. Update calls to mprts_hook,
13931 build_call_internal and gimple_simplified_result_is_gimple_val.
13932 Factor out code that is common to the tree_code and combined_fn cases.
13933 * genmatch.c (expr::gen_transform): Replace tem_code and
13934 tem_ops with a gimple_match_op called tem_op. Update calls
13935 to the gimple_resimplify functions and maybe_push_res_to_seq.
13936 (dt_simplify::gen_1): Manipulate res_op instead of res_code and
13937 res_ops. Update call to the gimple_resimplify functions.
13938 (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
13939 (decision_tree::gen): Make the functions take a gimple_match_op *
13940 called res_op instead of separate res_code and res_ops parameters.
13941 Update call accordingly.
13942 * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
13943 and ops with a single res_op parameter. Update calls to
13944 maybe_build_generic_op and maybe_push_res_to_seq.
13945 (fold_stmt_1): Update calls to gimple_simplify and
13946 replace_stmt_with_simplification.
13947 (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
13948 and gimple_simplified_result_is_gimple_val.
13949 * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
13951 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
13952 with a gimple_match_op *.
13953 (vn_nary_build_or_lookup): Likewise. Update call to
13954 vn_nary_build_or_lookup_1.
13955 (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
13956 gimple_match_op *. Update calls to the gimple_resimplify routines
13957 and to gimple_simplified_result_is_gimple_val.
13958 (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
13959 Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
13960 (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
13961 (visit_nary_op): Likewise.
13962 (visit_reference_op_load): Likewise.
13964 2018-05-23 Luis Machado <luis.machado@linaro.org>
13966 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
13967 modifier for printing the step amount.
13969 2018-05-23 Jozef Lawrynowicz <jozef.l@somniumtech.com>
13972 * tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
13975 2018-05-23 Segher Boessenkool <segher@kernel.crashing.org>
13977 * doc/sourcebuild.texi (Endianness): New subsubsection.
13979 2018-05-23 Luis Machado <luis.machado@linaro.org>
13981 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
13982 <prefetch_dynamic_strides>: New const bool field.
13983 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
13984 prefetch_dynamic_strides.
13985 (exynosm1_prefetch_tune): Likewise.
13986 (thunderxt88_prefetch_tune): Likewise.
13987 (thunderx_prefetch_tune): Likewise.
13988 (thunderx2t99_prefetch_tune): Likewise.
13989 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
13991 (aarch64_override_options_internal): Update to set
13992 PARAM_PREFETCH_DYNAMIC_STRIDES.
13993 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
13994 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
13995 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
13996 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
13997 prefetch-dynamic-strides setting.
13999 2018-05-23 Luis Machado <luis.machado@linaro.org>
14001 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
14002 <minimum_stride>: New const int field.
14003 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
14004 minimum_stride field defaulting to -1.
14005 (exynosm1_prefetch_tune): Likewise.
14006 (thunderxt88_prefetch_tune): Likewise.
14007 (thunderx_prefetch_tune): Likewise.
14008 (thunderx2t99_prefetch_tune): Likewise.
14009 (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
14010 <default_opt_level>: Set to 3.
14011 (aarch64_override_options_internal): Update to set
14012 PARAM_PREFETCH_MINIMUM_STRIDE.
14013 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
14014 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
14015 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
14016 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
14017 stride is constant and is below the minimum stride threshold.
14019 2018-05-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14021 * config/arm/arm-cpus.in (mode26): Delete.
14022 (armv4): Delete mode26 reference.
14023 * config/arm/arm.c (arm_configure_build_target): Delete use of
14026 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
14028 * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
14030 (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
14031 Rewrite expander pattern. Emit gen_floatunssi<mode>2_i387_with_xmm
14033 (floatunsdisf2): Rewrite expander pattern. Hanlde TARGET_AVX512F.
14034 (floatunsdidf2): Ditto.
14036 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
14038 * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
14039 (fixuns_trunc<mode>si2_avx512f): Ditto.
14040 (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
14041 (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
14042 Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
14044 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
14046 PR rtl-optimization/79985
14047 * df-scan.c (df_insn_refs_collect): Remove special case for
14048 global registers and asm statements.
14050 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
14052 * extend.texi (Global Register Variables): Rewrite the bullet list.
14053 Note that the register is available for allocation. Note that access
14054 via inline asm must use constraints. Add note about async-signal
14055 handlers. Remove paragraph about automagic register selection.
14057 2018-05-23 Richard Biener <rguenther@suse.de>
14059 * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
14060 of fixed offset from memset VN.
14062 2018-05-23 Bill Schmidt <wschmidt@linux.ibm.com>
14064 * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
14065 first_interp field.
14066 (alloc_cand_and_find_basis): Initialize first_interp field.
14067 (slsr_process_mul): Modify first_interp field.
14068 (slsr_process_add): Likewise.
14069 (slsr_process_cast): Modify first_interp field for each new
14071 (slsr_process_copy): Likewise.
14072 (dump_candidate): Dump first_interp field.
14073 (replace_mult_candidate): Process all interpretations, not just
14075 (replace_rhs_if_not_dup): Likewise.
14076 (replace_one_candidate): Likewise.
14078 2018-05-23 Wilco Dijkstra <wdijkstr@arm.com>
14080 * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
14082 (aarch64_needs_frame_chain): New function.
14083 (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
14085 2018-05-23 Sudakshina Das <sudi.das@arm.com>
14088 * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
14089 Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
14090 * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
14091 * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
14092 as true for strict-align.
14093 (aarch64_can_inline_p): Perform checks even when callee has no
14094 attributes to check for strict alignment.
14095 * doc/extend.texi (AArch64 Function Attributes): Document
14097 * doc/invoke.texi: (AArch64 Options): Likewise.
14099 2018-05-23 Richard Sandiford <richard.sandiford@linaro.org>
14101 PR tree-optimization/85853
14102 * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
14103 the handling of the root of the node to...
14104 (vect_slp_analyze_node_operations_1): ...this new function,
14105 and run the whole thing with the child nodes' def types
14106 set according to their SLP node's def type.
14108 2018-05-23 Richard Biener <rguenther@suse.de>
14110 PR middle-end/85874
14111 * tree-data-ref.c (create_runtime_alias_checks): Defer
14112 and ignore overflow warnings.
14114 2018-05-23 Yury Gribov <tetra2005@gmail.com>
14116 PR tree-optimization/85822
14117 * tree-vrp.c (is_masked_range_test): Fix handling of negative
14120 2018-05-23 Richard Biener <rguenther@suse.de>
14122 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
14123 memset constants via native_interpret_expr.
14125 2018-05-22 H.J. Lu <hongjiu.lu@intel.com>
14128 * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
14130 (cgraph_node::create_alias): Likewise.
14131 (cgraph_node::get_availability): Check ifunc_resolver instead
14132 of looking up ifunc attribute.
14133 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
14134 * varasm.c (do_assemble_alias): Likewise.
14135 (assemble_alias): Likewise.
14136 (default_binds_local_p_3): Likewise.
14137 * cgraph.h (cgraph_node): Add ifunc_resolver.
14138 (cgraph_node::only_called_directly_or_aliased_p): Return false
14139 for IFUNC resolver.
14140 * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
14142 * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
14143 is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
14144 (symtab_node::binds_to_current_def_p): Check ifunc_resolver
14145 instead of looking up ifunc attribute.
14147 2018-05-22 Luis Machado <luis.machado@linaro.org>
14149 * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
14151 2018-05-22 Martin Sebor <msebor@redhat.com>
14153 PR middle-end/85359
14154 * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
14155 only when expasion succeeds.
14156 (expand_builtin_strcmp): Same.
14157 (expand_builtin_strncmp): Same.
14159 2018-05-22 Martin Sebor <msebor@redhat.com>
14161 * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
14163 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
14164 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14166 * config/aarch64/aarch64-ldpstp.md: Replace uses of
14167 aarch64_mem_pair_operand with memory_operand and delete operand swapping
14169 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
14170 Add check for legitimate_address.
14171 (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
14172 (aarch64_swap_ldrstr_operands): New.
14173 * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
14176 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
14177 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14179 * config/aarch64/aarch64.md: New patterns to generate stp
14181 (store_pair_sw, store_pair_dw): New patterns to generate stp for
14182 single words and double words.
14183 (load_pair_sw, load_pair_dw): Likewise.
14184 (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
14186 (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
14188 * config/aarch64/aarch64-ldpstp.md: Modify peephole
14189 for different mode ldpstp and add peephole for merged zero stores.
14190 Likewise for loads.
14191 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
14193 (aarch64_gen_store_pair): Rename calls to match new patterns.
14194 (aarch64_gen_load_pair): Rename calls to match new patterns.
14195 * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
14196 (load_pair<DREG:mode><DREG2:mode>): ... This.
14197 (store_pair<mode>): Rename to...
14198 (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
14199 * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
14200 New mode iterators.
14201 (V_INT_EQUIV): Handle SImode.
14202 * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
14205 2018-05-22 Martin Sebor <msebor@redhat.com>
14208 * calls.c (maybe_warn_nonstring_arg): Use string length to set
14209 or ajust the presumed bound on an operation to avoid unnecessary
14212 2018-05-22 Martin Sebor <msebor@redhat.com>
14214 PR tree-optimization/85826
14215 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
14216 assuming that a DECL necesarily has a constant size.
14218 2018-05-22 Richard Sandiford <richard.sandiford@linaro.org>
14220 PR middle-end/85862
14221 * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
14223 2018-05-22 Richard Biener <rguenther@suse.de>
14225 PR tree-optimization/85834
14226 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
14227 non-constant and non-zero memset arguments.
14229 2018-05-22 Martin Liska <mliska@suse.cz>
14232 * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
14234 2018-05-22 Richard Biener <rguenther@suse.de>
14236 PR tree-optimization/85863
14237 * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
14238 comparisons when vectype is specified.
14239 (vectorizable_condition): Do not specify vectype for
14240 vect_is_simple_cond when SLP vectorizing.
14242 2018-05-21 Michael Meissner <meissner@linux.ibm.com>
14245 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
14246 define __ibm128 as long double.
14247 * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
14248 as a distinct type when IEEE 128-bit support is enabled.
14249 (init_float128_ieee): Fix up conversions between IFmode and IEEE
14250 128-bit types to use the correct functions.
14251 (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
14252 convert between 128-bit floating point types that have different
14253 modes but the same representation, instead of using gen_lowpart to
14255 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
14257 (IFKF_reg): New attributes to give the register constraints for
14259 (extend<mode>tf2_internal): New insns to mark an explicit
14260 conversion between 128-bit floating point types that have a
14261 different mode but share the same representation.
14263 2018-05-21 Richard Sandiford <richard.sandiford@linaro.org>
14265 PR tree-optimization/85814
14266 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
14267 a null return from get_strinfo when unsharing the next
14268 strinfo in the chain.
14270 2018-05-21 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
14273 * varasm.c (weak_finish): Clean up weak_decls.
14275 2018-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14277 * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
14278 UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
14279 UNSPEC_UADALP values.
14280 * config/aarch64/iterators.md (ABAL): New int iterator.
14283 (sur): Add mappings for the above.
14284 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
14286 (aarch64_<sur>abal<mode>_4): Likewise.
14287 (aarch64_<sur>adalp<mode>_3): Likewise.
14288 (<sur>sadv16qi): New define_expand.
14290 2018-05-21 Alexander Nesterovskiy <alexander.nesterovskiy@intel.com>
14292 * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
14293 (*movdf_internal): Ditto.
14294 (*rcpsf2_sse): Ditto.
14295 (*rsqrtsf2_sse): Ditto.
14296 (*sqrt<mode>2_sse): Ditto.
14298 2018-05-21 Tamar Christina <tamar.christina@arm.com>
14300 * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
14302 (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
14303 * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
14304 veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
14305 vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
14307 * config/aarch64/arm_neon.h: Likewise.
14308 * config/aarch64/iterators.md (VQ_I): New.
14310 2018-05-21 Alexey Brodkin <abrodkin@synopsys.com>
14312 * config.gcc: Add arc/t-multilib-linux to tmake_file for
14314 * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
14317 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
14319 * config/nds32/constraints.md (S): New constraint.
14320 * config/nds32/nds32.md (call_internal): Use constraint S.
14321 (call_value_internal): Likewise.
14322 (sibcall_internal): Likewise.
14323 (sibcall_value_internal): Likewise.
14325 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
14326 Chung-Ju Wu <jasonwucj@gmail.com>
14328 * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
14329 into consideration.
14331 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
14332 Chung-Ju Wu <jasonwucj@gmail.com>
14334 * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
14335 (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
14336 (nds32_rtx_costs_impl): Simplify.
14337 (nds32_address_cost_impl): Simplify.
14338 (nds32_init_rtx_costs): New function.
14339 (nds32_rtx_costs_speed_prefer): Likewise.
14340 (nds32_rtx_costs_size_prefer): Likewise.
14341 (nds32_address_cost_speed_prefer): Likewise.
14342 (nds32_address_cost_speed_fwprop): Likewise.
14343 (nds32_address_cost_size_prefer): Likewise.
14344 * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
14345 * config/nds32/nds32.c (nds32_option_override): Use
14346 nds32_init_rtx_costs function.
14348 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
14350 * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
14351 * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
14352 (TARGET_PIPELINE_N8): Likewise.
14353 (TARGET_PIPELINE_N10): Likewise.
14354 (TARGET_PIPELINE_N13): Likewise.
14355 (TARGET_PIPELINE_GRAYWOLF): Likewise.
14357 2018-05-19 Monk Chiang <sh.chiang04@gmail.com>
14359 * config/nds32/nds32-fpu.md: Update copyright year.
14361 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
14363 * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
14365 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
14367 * config/nds32/nds32.c
14368 (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
14369 * config/nds32/nds32.opt (minline-asm-r15): New option.
14371 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
14373 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
14375 * config/nds32/nds32.md (abssi2): New pattern.
14377 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
14379 * config/i386/i386.md (rex64namesuffix): New mode attribute.
14380 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
14381 Merge insn pattern from sse_cvtsi2ss<round_name> and
14382 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
14383 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
14384 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
14385 using SWI48 mode iterator.
14386 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
14387 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
14388 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
14389 pattern from sse_cvttss2si<round_saeonly_name>
14390 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
14391 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
14392 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
14393 using SWI48 mode iterator.
14394 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
14395 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
14396 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
14397 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
14398 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
14399 using SWI48 mode iterator.
14400 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
14401 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
14402 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
14403 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
14404 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
14405 SWI48 mode iterator.
14406 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
14407 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
14408 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
14409 pattern from sse_cvttsd2si<round_saeonly_name>
14410 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
14412 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
14414 * config/nds32/nds32-md-auxiliary.c
14415 (nds32_valid_smw_lwm_base_p): Refine.
14416 (nds32_output_smw_single_word): Refine.
14417 (nds32_output_smw_double_word): New.
14418 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
14420 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
14422 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
14423 (nds32_output_stack_pop): Refine.
14424 (nds32_expand_unaligned_load): Refine.
14425 (nds32_expand_unaligned_store): Refine.
14427 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
14428 Chung-Ju Wu <jasonwucj@gmail.com>
14430 * config/nds32/constants.md: Add TP_REGNUM constant.
14431 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
14432 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
14434 * config/nds32/nds32-doubleword.md: Consider flag_pic.
14435 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
14436 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
14437 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
14438 and PIC code generation.
14439 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
14441 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
14443 * config/nds32/nds32.md: Support TLS and PIC.
14444 * config/nds32/nds32.c: Support TLS and PIC.
14445 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
14446 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
14449 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
14451 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
14452 mode with E_ prefix.
14454 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
14455 Chung-Ju Wu <jasonwucj@gmail.com>
14457 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
14458 * config/nds32/nds32-md-auxiliary.c
14459 (symbolic_reference_mentioned_p): New.
14460 (nds32_legitimize_ict_address): New.
14461 (nds32_expand_ict_move): New.
14462 (nds32_indirect_call_referenced_p): New.
14463 (nds32_symbol_binds_local_p): Delete.
14464 (nds32_long_call_p): Modify.
14465 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
14466 * config/nds32/nds32-protos.h
14467 (symbolic_reference_mentioned_p): Declare.
14468 (nds32_legitimize_ict_address): Declare.
14469 (nds32_expand_ict_move): Declare.
14470 (nds32_indirect_call_referenced_p): Declare.
14471 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
14472 (nds32_relax_group): Use nds32_ict_const_p as condition.
14473 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
14474 (nds32_asm_file_start): Output ict_model directive in asm code.
14475 (nds32_legitimate_address_p): Consider indirect call.
14476 (nds32_print_operand): Consider indirect call.
14477 (nds32_print_operand_address): Consider indirect call.
14478 (nds32_insert_attributes): Handle "indirect_call" attribute.
14479 (TARGET_LEGITIMATE_ADDRESS_P): Define.
14480 (TARGET_LEGITIMATE_CONSTANT_P): Define.
14481 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
14482 (TARGET_DELEGITIMIZE_ADDRESS): Define.
14483 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
14484 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
14485 (TARGET_ICT_MODEL_SMALL): Define.
14486 (TARGET_ICT_MODEL_LARGE): Define.
14487 * config/nds32/nds32.md (movsi): Consider ict model.
14488 (call, call_value): Consider ict model.
14489 (sibcall, sibcall_value): Consider ict model.
14490 * config/nds32/nds32.opt (mict-model): New option.
14491 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
14494 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
14495 Monk Chiang <sh.chiang04@gmail.com>
14496 Jim Wilson <jimw@sifive.com>
14498 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
14499 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
14500 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
14501 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
14502 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
14503 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
14504 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
14505 compute save_libcall_adjustment properly.
14506 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
14507 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
14508 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
14509 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
14510 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
14511 (ABI_SPEC): Handle mabi=ilp32e.
14512 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
14513 (RVE): Add RVE mask.
14514 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
14515 <-march>: Add rv32e as an example.
14517 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
14520 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
14521 (intra_create_variable_infos): Handle C++ constructors.
14523 2018-05-18 Martin Liska <mliska@suse.cz>
14525 * passes.def: Remove a redundant pass.
14527 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
14530 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
14532 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14534 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
14536 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
14538 (armv2, armv2a, armv3, armv3m): Delete architectures.
14539 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
14540 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
14541 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
14543 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
14544 (*mulsidi3adddi): Likewise.
14545 (mulsidi3): Likewise.
14546 (*mulsidi3_nov6): Likewise.
14547 (umulsidi3): Likewise.
14548 (umulsidi3_nov6): Likewise.
14549 (umaddsidi4): Likewise.
14550 (*umulsidi3adddi): Likewise.
14551 (smulsi3_highpart): Likewise.
14552 (*smulsi3_highpart_nov6): Likewise.
14553 (umulsi3_highpart): Likewise.
14554 (*umulsi3_highpart_nov6): Likewise.
14555 * config/arm/arm.h (arm_arch3m): Delete.
14556 * config/arm/arm.c (arm_arch3m): Delete.
14557 (arm_option_override_internal): Update armv3-related comment.
14558 (arm_configure_build_target): Delete use of isa_bit_mode32.
14559 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
14560 (arm_rtx_costs_internal): Delete check of arm_arch3m.
14561 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
14562 (mulsa3): Likewise.
14563 (mulusa3): Likewise.
14564 * config/arm/arm-protos.h (arm_arch3m): Delete.
14565 * config/arm/arm-tables.opt: Regenerate.
14566 * config/arm/arm-tune.md: Likewise.
14567 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
14568 deleted architectures.
14570 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14572 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
14573 (armv5t, armv5te): New features.
14574 (ARMv5, ARMv5e): Delete fgroups.
14575 (ARMv5t, ARMv5te): Adjust for above changes.
14576 (ARMv6m): Likewise.
14577 (armv5, armv5e): Delete arches.
14578 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
14580 (*call_reg_arm): Likewise.
14581 (*call_value_reg_armv5): Likewise.
14582 (*call_value_reg_arm): Likewise.
14583 (*call_symbol): Likewise.
14584 (*call_value_symbol): Likewise.
14585 (*sibcall_insn): Likewise.
14586 (*sibcall_value_insn): Likewise.
14587 (clzsi2): Likewise.
14588 (prefetch): Likewise.
14589 (define_split and define_peephole2 dependent on arm_arch5):
14591 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
14593 (TARGET_ARM_QBIT): Likewise.
14594 (TARGET_DSP_MULTIPLY): Likewise.
14595 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
14596 (arm_arch5, arm_arch5e): Delete.
14597 (arm_arch5t, arm_arch5te): Declare.
14598 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
14599 (arm_arch5t): Declare.
14600 (arm_option_reconfigure_globals): Update for the above.
14601 (arm_options_perform_arch_sanity_checks): Update comment, replace
14602 use of arm_arch5 with arm_arch5t.
14603 (use_return_insn): Likewise.
14604 (arm_emit_call_insn): Likewise.
14605 (output_return_instruction): Likewise.
14606 (arm_final_prescan_insn): Likewise.
14607 (arm_coproc_builtin_available): Likewise.
14608 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
14609 arm_arch5e with arm_arch5t and arm_arch5te.
14610 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
14611 (arm_arch5t, arm_arch5te): Declare.
14612 * config/arm/arm-tables.opt: Regenerate.
14613 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
14614 * config/arm/t-multilib: Likewise.
14615 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
14616 instead of arm_arch5.
14617 (*call_reg_thumb1): Likewise.
14618 (*call_value_reg_thumb1_v5): Likewise.
14619 (*call_value_reg_thumb1): Likewise.
14620 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
14622 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
14624 2018-05-18 Martin Liska <mliska@suse.cz>
14626 PR gcov-profile/84846
14627 * doc/gcov.texi: Document -t option of gcov tool.
14629 2018-05-18 Martin Liska <mliska@suse.cz>
14631 PR gcov-profile/84846
14632 * gcov.c (print_usage): Add new -t option.
14633 (process_args): Handle the option.
14634 (generate_results): Use stdout as output when requested by
14637 2018-05-18 Martin Liska <mliska@suse.cz>
14639 PR gcov-profile/84846
14640 * coverage.c (coverage_init): Write PWD to .gcno file.
14641 * doc/gcov.texi: Document how working directory is printed.
14642 * gcov-dump.c (dump_gcov_file): Print PWD.
14643 * gcov.c (output_intermediate_file): Likewise.
14644 (read_graph_file): Read PWD string.
14645 (output_lines): Print PWD.
14647 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
14649 PR middle-end/85817
14650 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
14651 for retval and return false if all args to phi are zero.
14653 2018-05-18 Richard Biener <rguenther@suse.de>
14655 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
14657 (evrp_dom_walker::before_dom_children): Call it.
14659 2018-05-18 Richard Biener <rguenther@suse.de>
14661 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
14662 results when processing array refs with variable index.
14664 2018-05-18 Toon Moene <toon@moene.org>
14666 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
14667 directly after that of -floop-interchange. Indicate that both
14668 options are enabled by default when specifying -O3.
14670 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14672 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
14673 iterator. Delete separate integer-mode vec_set<mode> expander.
14674 (aarch64_simd_vec_setv2di): Delete.
14675 (vec_setv2di): Delete.
14676 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
14677 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
14678 the "w, r" alternative.
14680 2018-05-18 Martin Liska <mliska@suse.cz>
14682 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
14683 * tree-pass.h (make_pass_lower_switch_O0): New function.
14684 * tree-switch-conversion.c (node_has_low_bound): Remove.
14685 (node_has_high_bound): Likewise.
14686 (node_is_bounded): Likewise.
14687 (class pass_lower_switch): Make it a template type and create
14689 (pass_lower_switch::execute): Add template argument.
14690 (make_pass_lower_switch): New function.
14691 (make_pass_lower_switch_O0): New function.
14692 (do_jump_if_equal): Remove.
14693 (emit_case_nodes): Simplify to just handle all 3 cases and leave
14694 all the hard work to tree optimization passes.
14696 2018-05-18 Martin Liska <mliska@suse.cz>
14698 * dbgcnt.c (limit_low): Renamed from limit.
14699 (limit_high): New variable.
14700 (dbg_cnt_is_enabled): Check for upper limit.
14701 (dbg_cnt): Adjust dumping.
14702 (dbg_cnt_set_limit_by_index): Add new argument for high
14704 (dbg_cnt_set_limit_by_name): Likewise.
14705 (dbg_cnt_process_single_pair): Parse new format.
14706 (dbg_cnt_process_opt): Use strtok.
14707 (dbg_cnt_list_all_counters): Remove 'value' and add
14709 * doc/invoke.texi: Document changes.
14711 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
14713 * doc/sourcebuild.texi (scalar_all_fma): Document.
14714 * tree.def (FMA_EXPR): Delete.
14715 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
14716 * internal-fn.c (ternary_direct): New macro.
14717 (expand_ternary_optab_fn): Likewise.
14718 (direct_ternary_optab_supported_p): Likewise.
14719 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
14720 * builtins.c (fold_builtin_fma): Delete.
14721 (fold_builtin_3): Don't call it.
14722 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
14723 * expr.c (expand_expr_real_2): Likewise.
14724 * fold-const.c (operand_equal_p): Likewise.
14725 (fold_ternary_loc): Likewise.
14726 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
14727 * gimple.c (DEFTREECODE): Likewise.
14728 * gimplify.c (gimplify_expr): Likewise.
14729 * optabs-tree.c (optab_for_tree_code): Likewise.
14730 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
14731 * tree-eh.c (operation_could_trap_p): Likewise.
14732 (stmt_could_throw_1_p): Likewise.
14733 * tree-inline.c (estimate_operator_cost): Likewise.
14734 * tree-pretty-print.c (dump_generic_node): Likewise.
14735 (op_code_prio): Likewise.
14736 * tree-ssa-loop-im.c (stmt_cost): Likewise.
14737 * tree-ssa-operands.c (get_expr_operands): Likewise.
14738 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
14739 * fold-const-call.h (fold_fma): Delete.
14740 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
14741 CFN_FNMA and CFN_FNMS.
14742 (fold_fma): Delete.
14743 * genmatch.c (combined_fn): New enum.
14744 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
14745 (commutative_op): New function.
14746 (commutate): Use it. Handle more than 2 operands.
14747 (dt_operand::gen_gimple_expr): Use commutative_op.
14748 (parser::parse_expr): Allow :c to be used with non-binary
14749 operators if the commutative operand is known.
14750 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
14751 CFN_FMS, CFN_FNMA and CFN_FNMS.
14752 (backprop::process_assign_use): Remove FMA_EXPR handling.
14753 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
14754 (gen_hsa_fma): New function.
14755 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
14756 IFN_FNMA and IFN_FNMS.
14757 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
14758 * gimple-fold.h (follow_all_ssa_edges): Declare.
14759 * gimple-fold.c (follow_all_ssa_edges): New function.
14760 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
14761 gimple_build interface and use follow_all_ssa_edges to fold the result.
14762 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
14763 instead of checking for optabs directly.
14764 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
14765 rather than FMA_EXPRs.
14766 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
14767 call to IFN_FMA instead of an FMA_EXPR.
14769 2018-05-17 Jim Wilson <jimw@sifive.com>
14771 * expr.c (do_tablejump): When converting index to Pmode, if we have a
14772 sign extended promoted subreg, and the range does not have the sign bit
14773 set, then do a sign extend.
14775 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
14776 test, check for sign extended subreg and/or constant operands, and
14777 do a sign extend in that case.
14779 2018-05-17 Steve Ellcey <sellcey@cavium.com>
14781 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
14782 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
14784 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
14785 Change logics_shift_reg to logics_shift_imm.
14786 (thunderx2t99_fp_loadpair_basic): Delete.
14787 (thunderx2t99_fp_storepair_basic): Delete.
14788 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
14789 (thunderx2t99_asimd_polynomial): Delete.
14790 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
14791 and neon_fp_mul_d_scalar_q.
14792 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
14793 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
14794 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
14795 (thunderx2t99_asimd_lut): Add missing tbl types.
14796 (thunderx2t99_asimd_ext): Delete.
14797 (thunderx2t99_asimd_load1_1_mult): Delete.
14798 (thunderx2t99_asimd_load1_2_mult): Delete.
14799 (thunderx2t99_asimd_load1_ldp): New.
14800 (thunderx2t99_asimd_load1): New.
14801 (thunderx2t99_asimd_load2): Add missing *load2* types.
14802 (thunderx2t99_asimd_load3): New.
14803 (thunderx2t99_asimd_load4): New.
14804 (thunderx2t99_asimd_store1_1_mult): Delete.
14805 (thunderx2t99_asimd_store1_2_mult): Delete.
14806 (thunderx2t99_asimd_store2_mult): Delete.
14807 (thunderx2t99_asimd_store2_onelane): Delete.
14808 (thunderx2t99_asimd_store_stp): New.
14809 (thunderx2t99_asimd_store1): New.
14810 (thunderx2t99_asimd_store2): New.
14811 (thunderx2t99_asimd_store3): New.
14812 (thunderx2t99_asimd_store4): New.
14814 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
14816 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
14817 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
14819 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
14820 Segher Boessenkool <segher@kernel.crashing.org>
14823 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
14826 2018-05-17 Richard Biener <rguenther@suse.de>
14828 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
14829 for pruning loop and prune defs feeding only already visited PHIs.
14831 2018-05-17 Richard Biener <rguenther@suse.de>
14833 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
14835 2018-05-17 Bin Cheng <bin.cheng@arm.com>
14836 Richard Biener <rguenther@suse.de>
14838 PR tree-optimization/85793
14839 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
14840 for VMAT_ELEMENTWISE.
14842 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
14844 * internal-fn.h (lookup_internal_fn): Declare
14845 * internal-fn.c (lookup_internal_fn): New function.
14846 * gimple.c (gimple_build_call_from_tree): Handle calls to
14847 internal functions.
14848 * gimple-pretty-print.c (dump_gimple_call): Print "." before
14849 internal function names.
14850 * tree-pretty-print.c (dump_generic_node): Likewise.
14851 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
14853 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
14855 * gimple-fold.h (gimple_build): Make the function forms take
14856 combined_fn rather than built_in_function.
14857 (gimple_simplify): Likewise.
14858 * gimple-match-head.c (gimple_simplify): Likewise.
14859 * gimple-fold.c (gimple_build): Likewise.
14860 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
14861 rather than gimple_build_call_internal.
14862 (get_initial_defs_for_reduction): Likewise.
14863 (vect_create_epilog_for_reduction): Likewise.
14864 (vectorizable_live_operation): Likewise.
14866 2018-05-17 Martin Liska <mliska@suse.cz>
14868 * gimple-ssa-sprintf.c (format_directive): Do not use
14869 space in between 'G_' and '('.
14871 2018-05-17 Jakub Jelinek <jakub@redhat.com>
14874 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
14875 even if the mask is not all ones.
14878 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
14880 (ix86_gimple_fold_builtin): Likewise.
14883 * config/i386/i386.c: Include tree-vector-builder.h.
14884 (ix86_vector_shift_count): New function.
14885 (ix86_fold_builtin): Fold shift builtins by scalar count.
14886 (ix86_gimple_fold_builtin): Likewise.
14888 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
14889 _mm512_setzero): New intrinsics.
14891 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
14892 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14894 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
14895 code generation for cases where splatting a value is not useful.
14896 * simplify-rtx.c (simplify_ternary_operation): Simplify
14897 vec_merge across a vec_duplicate and a paradoxical subreg forming
14898 a vector mode to a vec_concat.
14900 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
14902 * config.gcc: Support "goldmont-plus".
14903 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
14905 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
14906 PROCESSOR_GOLDMONT_PLUS.
14907 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
14908 (processor_target_table): Add "goldmont-plus".
14909 (PTA_GOLDMONT_PLUS): Define.
14910 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
14911 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
14912 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
14913 (fold_builtin_cpu): Add "goldmont-plus".
14914 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
14915 (ix86_option_override_internal): Add "goldmont-plus".
14916 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
14917 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
14918 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
14919 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
14921 2018-05-17 Richard Biener <rguenther@suse.de>
14923 PR tree-optimization/85757
14924 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
14925 remove defs that only feed that PHI from further processing.
14927 2018-05-16 Jim Wilson <jimw@sifive.com>
14929 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
14931 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
14933 2018-05-16 Mark Wielaard <mark@klomp.org>
14935 * dwarf2out.c (count_index_strings): New function.
14936 (output_indirect_strings): Call count_index_strings and generate
14937 header for dwarf_version >= 5.
14939 2018-05-16 Mark Wielaard <mark@klomp.org>
14941 * dwarf2out.c (dwarf_FORM): New function.
14942 (set_indirect_string): Use dwarf_FORM.
14943 (reset_indirect_string): Likewise.
14944 (size_of_die): Likewise.
14945 (value_format): Likewise.
14946 (output_die): Likewise.
14947 (add_skeleton_AT_string): Likewise.
14948 (output_macinfo_op): Likewise.
14949 (index_string): Likewise.
14950 (output_index_string_offset): Likewise.
14951 (output_index_string): Likewise.
14952 (count_index_strings): Likewise.
14954 2018-05-16 Carl Love <cel@us.ibm.com>
14956 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
14957 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
14959 2018-05-16 Martin Jambor <mjambor@suse.cz>
14961 * ipa-prop.c (ipa_free_all_edge_args): Remove.
14962 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
14964 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
14966 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
14967 (fnma<mode>4): Likewise.
14968 (fms<mode>4): Likewise.
14969 (fnms<mode>4): Likewise.
14970 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
14971 (aarch64_fnma<mode>4): Likewise.
14972 (aarch64_fms<mode>4): Likewise.
14973 (aarch64_fnms<mode>4): Likewise.
14974 (aarch64_fnmadd<mode>4): Likewise.
14976 2018-05-16 Jason Merrill <jason@redhat.com>
14978 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
14980 2018-05-16 Richard Biener <rguenther@suse.de>
14982 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
14983 (dump_stmt_cost): Declare.
14984 (add_stmt_cost): Dump cost we add.
14985 (add_stmt_costs): New function.
14986 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
14987 No longer exported.
14988 (vect_analyze_stmt): Adjust prototype.
14989 (vectorizable_condition): Likewise.
14990 (vectorizable_live_operation): Likewise.
14991 (vectorizable_reduction): Likewise.
14992 (vectorizable_induction): Likewise.
14993 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
14994 cost vector to pass to vectorizable_ and record afterwards.
14995 (vect_model_reduction_cost): Take cost vector argument and adjust.
14996 (vect_model_induction_cost): Likewise.
14997 (vectorizable_reduction): Likewise.
14998 (vectorizable_induction): Likewise.
14999 (vectorizable_live_operation): Likewise.
15000 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
15001 SLP_TREE_NUMBER_OF_VEC_STMTS.
15002 (vect_analyze_slp_cost_1): Remove.
15003 (vect_analyze_slp_cost): Likewise.
15004 (vect_slp_analyze_node_operations): Take visited args and
15005 a target cost vector. Avoid processing already visited stmt sets.
15006 (vect_slp_analyze_operations): Use a local cost vector to gather
15007 costs and register those of non-discarded instances.
15008 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
15009 (vect_schedule_slp_instance): Remove copying of
15010 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
15012 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
15013 adding cost. Record cost entry location.
15014 (vect_prologue_cost_for_slp_op): Function to compute cost of
15015 a constant or invariant generated for SLP vect in the prologue,
15016 split out from vect_analyze_slp_cost_1.
15017 (vect_model_simple_cost): Make static. Adjust for SLP costing.
15018 (vect_model_promotion_demotion_cost): Likewise.
15019 (vect_model_store_cost): Likewise, make static.
15020 (vect_model_load_cost): Likewise.
15021 (vectorizable_bswap): Add cost vector arg and adjust.
15022 (vectorizable_call): Likewise.
15023 (vectorizable_simd_clone_call): Likewise.
15024 (vectorizable_conversion): Likewise.
15025 (vectorizable_assignment): Likewise.
15026 (vectorizable_shift): Likewise.
15027 (vectorizable_operation): Likewise.
15028 (vectorizable_store): Likewise.
15029 (vectorizable_load): Likewise.
15030 (vectorizable_condition): Likewise.
15031 (vectorizable_comparison): Likewise.
15032 (can_vectorize_live_stmts): Likewise.
15033 (vect_analyze_stmt): Likewise.
15034 (vect_transform_stmt): Adjust calls to vectorizable_*.
15035 * tree-vectorizer.c: Include gimple-pretty-print.h.
15036 (dump_stmt_cost): New function.
15038 2018-05-16 Richard Biener <rguenther@suse.de>
15040 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
15041 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
15042 * tree-ssa-dse.c: Include tree-ssa-loop.h.
15043 (check_name): New callback.
15044 (dse_classify_store): Track cycles via a visited bitmap of PHI
15045 defs and simplify handling of in-loop and across loop dead stores
15046 and properly fail for loop-variant refs. Handle byte-tracking with
15047 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
15050 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
15052 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
15053 (vect_get_mask_type_for_stmt): Likewise.
15054 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
15056 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
15057 to determine the statement's vector type and the vector type that
15058 should be used for calculating nunits. Deal with cases in which
15059 the type has to be deferred.
15060 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
15061 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
15062 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
15063 (vect_determine_vf_for_stmt): New functions, split out from...
15064 (vect_determine_vectorization_factor): ...here.
15065 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
15066 (vect_get_mask_type_for_stmt): New functions, split out from
15067 vect_determine_vectorization_factor.
15069 2018-05-16 Richard Biener <rguenther@suse.de>
15071 * tree-cfg.c (verify_gimple_assign_ternary): Properly
15072 verify the [VEC_]COND_EXPR embedded comparison.
15074 2018-05-15 Martin Sebor <msebor@redhat.com>
15076 PR tree-optimization/85753
15077 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
15078 RECORD_TYPE in addition to ARRAY_TYPE.
15080 2018-05-15 Martin Sebor <msebor@redhat.com>
15082 PR middle-end/85643
15083 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
15085 2018-05-15 Richard Biener <rguenther@suse.de>
15087 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
15088 add by_clobber_p one. Change algorithm to collect all defs
15089 representing uses we need to walk and try reducing them to
15090 a single one before failing.
15091 (dse_dom_walker::dse_optimize_stmt): Adjust.
15093 2018-05-13 Mark Wielaard <mark@klomp.org>
15095 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
15096 (size_of_loc_descr): Likewise.
15097 (output_loc_operands): Likewise.
15098 (output_loc_operands_raw): Likewise.
15099 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
15100 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
15101 (hash_loc_operands): Likewise.
15102 (compare_loc_operands): Likewise.
15104 2018-05-14 Mark Wielaard <mark@klomp.org>
15106 * dwarf2out.c (count_index_addrs): New function.
15107 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
15109 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
15111 PR tree-optimization/83648
15112 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
15113 return value as malloc candidate.
15115 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
15118 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite
15119 param as true in call to suggest_attribute.
15121 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
15123 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
15124 -mreadonly-in-sdata.
15126 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15128 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
15130 (aarch64_crypto_aesd_fused): Likewise.
15132 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
15134 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
15135 (movsi_aarch64): Likewise.
15136 (load_pairsi): Likewise.
15137 (load_pairdi): Likewise.
15138 (store_pairsi): Likewise.
15139 (store_pairdi): Likewise.
15140 (load_pairsf): Likewise.
15141 (load_pairdf): Likewise.
15142 (store_pairsf): Likewise.
15143 (store_pairdf): Likewise.
15144 (zero_extend): Likewise.
15145 (trunc): Swap alternatives.
15146 (fcvt_target): Add '?' to prefer w over r.
15148 2018-05-14 Jakub Jelinek <jakub@redhat.com>
15151 * config/i386/i386.md: Disallow non-commutative arithmetics in
15152 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
15153 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
15154 in the peephole2 before it.
15156 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
15158 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
15159 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
15160 (ix86_handle_option): Handle -mcldemote.
15161 * config.gcc: New header.
15162 * config/i386/cldemoteintrin.h: New file.
15163 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
15164 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
15166 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
15167 OPTION_MASK_ISA_CLDEMOTE.
15168 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
15169 (ix86_valid_target_attribute_inner_p): Ditto.
15170 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
15171 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
15172 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
15173 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
15174 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
15176 * config/i386/i386.opt: Add -mcldemote.
15177 * config/i386/x86intrin.h: New header.
15178 * doc/invoke.texi: Add -mcldemote.
15180 2018-05-14 Richard Biener <rguenther@suse.de>
15182 * doc/match-and-simplify.texi: Adjust :s documentation.
15184 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
15186 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
15187 intended memcpy size.
15188 (REORDER_45): Likewise.
15190 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
15192 * sort.cc: New file.
15193 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
15194 * vec.c (qsort_chk): Use gcc_qsort.
15195 * Makefile.in (OBJS-libcommon): Add sort.o.
15196 (build/sort.o): New target. Use it...
15197 (BUILD_RTL): ... here, and...
15198 (build/gencfn-macros): ... here, and...
15199 (build/genmatch): ... here.
15201 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
15202 Chung-Ju Wu <jasonwucj@gmail.com>
15204 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
15205 * config/nds32/nds32-graywolf.md: New file.
15206 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
15207 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
15209 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
15210 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
15211 * config/nds32/nds32.md (pipeline_model): Add graywolf.
15212 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
15213 * config/nds32/pipelines.md: Include n15 settings.
15215 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
15216 Chung-Ju Wu <jasonwucj@gmail.com>
15218 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
15219 * config/nds32/nds32-n13.md: New file.
15220 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
15221 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
15223 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
15224 * config/nds32/nds32.md (pipeline_model): Add n13.
15225 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
15226 * config/nds32/pipelines.md: Include n13 settings.
15228 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
15229 Chung-Ju Wu <jasonwucj@gmail.com>
15231 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
15232 * config/nds32/nds32-n10.md: New file.
15233 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
15234 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
15236 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
15237 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
15238 * config/nds32/nds32.md (pipeline_model): Add n10.
15239 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
15240 * config/nds32/pipelines.md: Include n10 settings.
15242 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
15243 Kito Cheng <kito.cheng@gmail.com>
15244 Chung-Ju Wu <jasonwucj@gmail.com>
15246 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
15247 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
15248 Add enum values for DSP extension instructions.
15249 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
15251 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
15252 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
15253 New code iterators.
15254 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
15255 * config/nds32/nds32-dspext.md: New file for DSP implementation.
15256 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
15257 * config/nds32/nds32-intrinsic.md: Likewise.
15258 * config/nds32/nds32_intrinsic.h: Likewise.
15259 * config/nds32/nds32-md-auxiliary.c: Likewise.
15260 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
15261 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
15262 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
15263 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
15264 * config/nds32/nds32-protos.h: New declarations for DSP extension.
15265 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
15266 TYPE_DMAC in switch statement.
15267 * config/nds32/nds32.c: New checking and implementation for DSP
15268 extension instructions.
15269 * config/nds32/nds32.h: Likewise.
15270 * config/nds32/nds32.md: Likewise.
15271 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
15272 * config/nds32/predicates.md: Implement new predicates for DSP
15275 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
15277 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
15278 Reformat alternatives and attributes so it is easier to identify
15279 which constraints/attributes go with which instruction.
15280 (mov<mode>_hardfloat32, FMOVE64): Likewise.
15281 (mov<mode>_softfloat32, FMOVE64): Likewise.
15282 (mov<mode>_hardfloat64, FMOVE64): Likewise.
15283 (mov<mode>_softfloat64, FMOVE64): Likewise.
15285 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
15287 * doc/extend.texi (PowerPC Built-in Functions): Rename this
15289 (Basic PowerPC Built-in Functions): The new name of the
15290 subsection previously known as "PowerPC Built-in Functions".
15291 (Basic PowerPC Built-in Functions Available on all Configurations):
15293 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
15294 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
15295 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
15296 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
15298 2018-05-11 Martin Jambor <mjambor@suse.cz>
15301 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
15304 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
15307 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
15309 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
15311 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
15312 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
15313 (ix86_handle_option): Handle -mwaitpkg.
15314 * config.gcc: New header.
15315 * config/i386/cpuid.h (bit_WAITPKG): New bit.
15316 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
15317 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
15319 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
15320 OPTION_MASK_ISA_WAITPKG.
15321 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
15322 (ix86_option_override_internal): Add PTA_WAITPKG.
15323 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
15324 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
15325 IX86_BUILTIN_TPAUSE.
15326 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
15327 __builtin_ia32_umwait and __builtin_ia32_tpause.
15328 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
15329 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
15330 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
15331 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
15332 UNSPECV_TPAUSE): New.
15333 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
15334 * config/i386/i386.opt: Add -mwaitpkg.
15335 * config/i386/waitpkgintrin.h: New file.
15336 * config/i386/x86intrin.h: New header.
15337 * doc/invoke.texi: Add -mwaitpkg.
15339 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
15342 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
15344 (cortex-m0): Use armv6s-m isa.
15345 (cortex-m0plus): Likewise.
15346 (cortex-m1): Likewise.
15347 (cortex-m0.small-multiply): Likewise.
15348 (cortex-m0plus.small-multiply): Likewise.
15349 (cortex-m1.small-multiply): Likewise.
15351 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
15352 Jakub Jelinek <jakub@redhat.com>
15354 PR tree-optimization/85692
15355 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
15356 source permute as well.
15358 2018-05-11 Martin Liska <mliska@suse.cz>
15361 * doc/extend.texi: Document LLVM style format for no_sanitize
15364 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
15366 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
15367 mode_supports_vsx_dform_quad to mode_supports_dq_form.
15368 (mode_supports_vsx_dform_quad): Likewise.
15369 (mode_supports_vmx_dform): Move these functions to be next to the
15370 other mode_supports functions.
15371 (mode_supports_dq_form): Likewise.
15372 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
15373 mode_supports_dq_form.
15374 (reg_offset_addressing_ok_p): Likewise.
15375 (offsettable_ok_by_alignment): Likewise.
15376 (rs6000_legitimate_offset_address_p): Likewise.
15377 (legitimate_lo_sum_address_p): Likewise.
15378 (rs6000_legitimize_address): Likewise.
15379 (rs6000_legitimize_reload_address): Likewise.
15380 (rs6000_secondary_reload_inner): Likewise.
15381 (rs6000_preferred_reload_class): Likewise.
15382 (rs6000_output_move_128bit): Likewise.
15384 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
15386 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
15387 Generate SImode target register for null target.
15388 <case IX86_BUILTIN_XGETBV>: Ditto.
15389 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
15390 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
15392 2018-05-10 Carl Love <cel@us.ibm.com>
15394 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
15395 dcbtt and dcbtstt if operands[2] is 0.
15397 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
15400 * config/i386/sse.md (usadv64qi): New expander.
15402 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
15404 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
15405 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
15406 -maltivec=be support.
15407 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
15408 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
15409 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
15410 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
15411 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
15412 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
15413 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
15414 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
15415 altivec_vsumsws): Adjust.
15416 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
15417 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
15418 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
15420 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
15421 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
15422 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
15423 (altivec_lve<VI_char>x): Delete expand.
15424 (*altivec_lve<VI_char>x_internal): Rename to...
15425 (altivec_lve<VI_char>x): ... this.
15426 (altivec_lvxl_<mode>): Delete expand.
15427 (*altivec_lvxl_<mode>_internal): Rename to ...
15428 (altivec_lvxl_<mode>): ... this.
15429 (altivec_stvxl_<mode>): Delete expand.
15430 (*altivec_stvxl_<mode>_internal): Rename to ...
15431 (altivec_stvxl_<mode>): ... this.
15432 (altivec_stve<VI_char>x): Delete expand.
15433 (*altivec_stve<VI_char>x_internal): Rename to ...
15434 (altivec_stve<VI_char>x): ... this.
15435 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
15436 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
15437 reduc_plus_scal_<mode>): Adjust.
15438 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
15440 (rs6000_cpu_cpp_builtins): Adjust.
15441 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
15442 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
15443 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
15444 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
15445 -maltivec=be support.
15446 (rs6000_split_vec_extract_var): Adjust.
15447 (rs6000_split_v4si_init): Adjust.
15448 (swap_selector_for_mode): Delete.
15449 (altivec_expand_lvx_be, altivec_expand_stvx_be,
15450 altivec_expand_stvex_be): Delete.
15451 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
15452 -maltivec=be support.
15453 (rs6000_gimple_fold_builtin): Ditto.
15454 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
15456 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
15457 (TARGET_DIRECT_MOVE_64BIT): Adjust.
15458 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
15459 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
15460 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
15461 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
15462 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
15463 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
15464 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
15465 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
15466 anonymous split): Adjust.
15467 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
15468 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
15470 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
15472 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
15473 when --with-gxx-include-dir is also specified.
15474 * configure: Regenerate.
15476 2018-05-09 Jim Wilson <jimw@sifive.com>
15479 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
15480 * config/riscv/t-withmultilib: New.
15481 * config/riscv/withmultilib.h: New.
15482 * doc/install.texi: Document RISC-V --with-multilib-list support.
15484 2018-05-09 Richard Biener <rguenther@suse.de>
15486 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
15488 (vect_bb_vectorization_profitable_p): Adjust. Compute
15489 actual scalar cost using the cost vector and the add_stmt_cost
15492 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
15494 PR rtl-optimization/85645
15495 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
15496 in the REG_CFA_REGISTER note for LR, don't leave it empty.
15498 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
15500 PR rtl-optimization/85645
15501 * shrink-wrap.c (spread_components): Return a boolean saying if
15502 anything was changed.
15503 (try_shrink_wrapping_separate): Iterate spread_components until
15504 nothing changes anymore.
15506 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
15508 PR rtl-optimization/85645
15509 * regrename.c (build_def_use): Also kill the chains that include the
15510 destination of a REG_CFA_REGISTER note.
15512 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
15514 PR rtl-optimization/85645
15515 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
15516 insn that has a REG_CFA_REGISTER note.
15518 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
15520 * cfgexpand.c (expand_clobber): New function.
15521 (expand_gimple_stmt_1): Use it.
15522 * tree-vect-stmts.c (vect_clobber_variable): New function,
15524 (vectorizable_simd_clone_call): ...here.
15525 (vectorizable_store): Emit a clobber either side of an
15526 IFN_STORE_LANES sequence.
15527 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
15529 2018-05-09 Tom de Vries <tom@codesourcery.com>
15532 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
15533 (define_insn "trap_if_false"): Add exit after trap.
15535 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
15537 PR rtl-optimization/85638
15538 * bb-reorder.c: Include common/common-target.h.
15539 (create_forwarder_block): New function extracted from...
15540 (fix_up_crossing_landing_pad): ...here. Rename into...
15541 (dw2_fix_up_crossing_landing_pad): ...this.
15542 (sjlj_fix_up_crossing_landing_pad): New function.
15543 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
15544 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
15545 from both partitions and exit the loop after one iteration.
15547 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
15550 * doc/extend.texi (PowerPC Built-in Functions): Rename this
15552 (Basic PowerPC Built-in Functions): The new name of the
15553 subsection previously known as "PowerPC Built-in Functions".
15554 (Basic PowerPC Built-in Functions Available on all Configurations):
15556 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
15558 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
15560 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
15562 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
15565 2018-05-08 Jim Wilson <jimw@sifive.com>
15567 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
15568 (LD_EMUL_SUFFIX): New.
15569 (LINK_SPEC): Use it.
15571 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
15573 * doc/extend.texi (PowerPC Built-in Functions): Rename this
15575 (Basic PowerPC Built-in Functions): The new name of the
15576 subsection previously known as "PowerPC Built-in Functions".
15577 (Basic PowerPC Built-in Functions Available on all Configurations):
15579 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
15581 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
15583 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
15585 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
15588 2018-05-08 Jakub Jelinek <jakub@redhat.com>
15591 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
15592 after cmpelim optimization.
15594 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
15596 * config.gcc: Support "goldmont".
15597 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
15598 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
15599 PROCESSOR_GOLDMONT.
15600 * config/i386/i386.c (m_GOLDMONT): Define.
15601 (processor_target_table): Add "goldmont".
15602 (PTA_GOLDMONT): Define.
15603 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
15604 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
15605 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
15606 (fold_builtin_cpu): Add "goldmont".
15607 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
15608 (ix86_option_override_internal): Add "goldmont".
15609 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
15610 (processor_type): Add PROCESSOR_GOLDMONT.
15611 * config/i386/i386.md: Add CPU "glm".
15612 * config/i386/glm.md: New file.
15613 * config/i386/x86-tune.def: Add m_GOLDMONT.
15614 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
15616 2018-05-08 Jakub Jelinek <jakub@redhat.com>
15619 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
15621 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
15622 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
15623 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
15626 * config/i386/i386.c (ix86_fold_builtin): Handle
15627 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
15630 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
15631 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
15633 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
15636 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
15637 (check_arch): Likewise.
15638 (check_fpu): Return the result rather than printing it.
15639 (end arch): Fix operator precedence.
15640 (end cpu): Likewise.
15641 (END): Print the result from check_fpu.
15643 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
15644 Alan Hayward <alan.hayward@arm.com>
15645 David Sherwood <david.sherwood@arm.com>
15647 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
15648 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
15649 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
15650 (*fcmuo<mode>_and): New patterns.
15652 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
15654 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
15655 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
15656 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
15657 (cmp_op, sve_imm_con): New code attributes.
15658 (SVE_COND_INT_CMP, imm_con): Delete.
15659 (cmp_op): Remove above unspecs from int attribute.
15660 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
15662 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
15663 comparison-specific unspecs.
15664 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
15665 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
15666 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
15667 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
15668 (*vec_fcm<cmp_op><mode>): Rename to...
15669 (*fcm<cmp_op><mode>): ...this and adjust likewise.
15670 (*vec_fcmuo<mode>): Rename to...
15671 (*fcmuo<mode>): ...this and adjust likewise.
15672 (*pred_fcm<cmp_op><mode>): New pattern.
15673 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
15674 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
15676 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
15678 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
15679 (aarch64_emit_sve_predicated_cond): New function.
15680 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
15681 (aarch64_emit_unspec_cond_or): Replace with...
15682 (aarch64_emit_sve_or_conds): ...this new function. Use
15683 aarch64_emit_sve_ptrue_op for the individual comparisons and
15684 aarch64_emit_binop to OR them together.
15685 (aarch64_emit_inverted_unspec_cond): Replace with...
15686 (aarch64_emit_sve_inverted_cond): ...this new function. Use
15687 aarch64_emit_sve_ptrue_op for the comparison and
15688 aarch64_emit_unop to invert the result.
15689 (aarch64_expand_sve_vec_cmp_float): Update after the above
15690 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
15692 2018-05-07 Nathan Sidwell <nathan@acm.org>
15694 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
15695 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
15696 (Backwards Compatibility): Likewise.
15698 2018-05-07 Luis Machado <luis.machado@linaro.org>
15702 2018-05-07 Luis Machado <luis.machado@linaro.org>
15704 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
15705 <prefetch_dynamic_strides>: New const bool field.
15706 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
15707 prefetch_dynamic_strides.
15708 (exynosm1_prefetch_tune): Likewise.
15709 (thunderxt88_prefetch_tune): Likewise.
15710 (thunderx_prefetch_tune): Likewise.
15711 (thunderx2t99_prefetch_tune): Likewise.
15712 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
15714 (aarch64_override_options_internal): Update to set
15715 PARAM_PREFETCH_DYNAMIC_STRIDES.
15716 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
15717 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
15718 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
15719 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
15720 prefetch-dynamic-strides setting.
15722 2018-05-07 Luis Machado <luis.machado@linaro.org>
15724 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
15725 <minimum_stride>: New const int field.
15726 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
15727 minimum_stride field.
15728 (exynosm1_prefetch_tune): Likewise.
15729 (thunderxt88_prefetch_tune): Likewise.
15730 (thunderx_prefetch_tune): Likewise.
15731 (thunderx2t99_prefetch_tune): Likewise.
15732 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
15733 (aarch64_override_options_internal): Update to set
15734 PARAM_PREFETCH_MINIMUM_STRIDE.
15735 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
15736 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
15737 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
15738 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
15739 stride is constant and is below the minimum stride threshold.
15741 2018-05-07 Luis Machado <luis.machado@linaro.org>
15743 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
15746 2018-05-07 Luis Machado <luis.machado@linaro.org>
15748 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
15749 <prefetch_dynamic_strides>: New const bool field.
15750 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
15751 prefetch_dynamic_strides.
15752 (exynosm1_prefetch_tune): Likewise.
15753 (thunderxt88_prefetch_tune): Likewise.
15754 (thunderx_prefetch_tune): Likewise.
15755 (thunderx2t99_prefetch_tune): Likewise.
15756 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
15758 (aarch64_override_options_internal): Update to set
15759 PARAM_PREFETCH_DYNAMIC_STRIDES.
15760 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
15761 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
15762 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
15763 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
15764 prefetch-dynamic-strides setting.
15766 2018-05-07 Luis Machado <luis.machado@linaro.org>
15768 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
15769 <minimum_stride>: New const int field.
15770 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
15771 minimum_stride field.
15772 (exynosm1_prefetch_tune): Likewise.
15773 (thunderxt88_prefetch_tune): Likewise.
15774 (thunderx_prefetch_tune): Likewise.
15775 (thunderx2t99_prefetch_tune): Likewise.
15776 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
15777 (aarch64_override_options_internal): Update to set
15778 PARAM_PREFETCH_MINIMUM_STRIDE.
15779 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
15780 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
15781 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
15782 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
15783 stride is constant and is below the minimum stride threshold.
15785 2018-05-06 Jakub Jelinek <jakub@redhat.com>
15788 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
15789 the type is addressable. Don't force op into register if it has
15792 2018-05-05 Roland McGrath <mcgrathr@google.com>
15795 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
15796 any section for which we don't know a specific type it should have,
15797 regardless of name. Previously this was done only for the exact
15798 names ".init_array", ".fini_array", and ".preinit_array".
15799 (default_elf_asm_named_section): Add comment about
15800 relationship with default_section_type_flags and SECTION_NOTYPE.
15801 (get_section): Don't consider it a type conflict if one side has
15802 SECTION_NOTYPE and the other doesn't, as long as neither has the
15803 SECTION_BSS et al used in the default_section_type_flags logic.
15805 2018-05-05 Tom de Vries <tom@codesourcery.com>
15808 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
15809 (workaround_barsyncs): New function.
15810 (nvptx_reorg): Use workaround_barsyncs.
15811 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
15812 (define_expand "nvptx_membar_cta"): New define_expand.
15813 (define_insn "*nvptx_membar_cta"): New insn.
15815 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
15817 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
15818 To improve optimization opportunities.
15819 * builtin-types.def: The new needed builtin types for the above.
15821 2018-05-04 Richard Biener <rguenther@suse.de>
15823 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
15824 * gimple-ssa-store-merging.c
15825 (imm_store_chain_info::output_merged_store): Remove redundant create,
15826 release split_store vector contents on failure.
15827 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
15828 scalar stmt vector on cache hit.
15830 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
15832 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
15834 * config.gcc (powerpc-xilinx-eabi*): Remove.
15835 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
15837 (fusion_addis_mem_combo_load): Ditto.
15838 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
15840 (rs6000_cpu_cpp_builtins): Ditto.
15841 * config/rs6000/rs6000-linux.c
15842 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
15843 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
15844 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
15846 (rs6000_setup_reg_addr_masks): Ditto.
15847 (rs6000_init_hard_regno_mode_ok): Ditto.
15848 (rs6000_option_override_internal): Ditto.
15849 (legitimate_lo_sum_address_p): Ditto.
15850 (rs6000_legitimize_address): Ditto.
15851 (rs6000_legitimize_reload_address): Ditto.
15852 (rs6000_legitimate_address_p): Ditto.
15853 (abi_v4_pass_in_fpr): Ditto.
15854 (setup_incoming_varargs): Ditto.
15855 (rs6000_gimplify_va_arg): Ditto.
15856 (rs6000_split_multireg_move): Ditto.
15857 (rs6000_savres_strategy): Ditto.
15858 (rs6000_emit_prologue_components): Ditto.
15859 (rs6000_emit_epilogue_components): Ditto.
15860 (rs6000_emit_prologue): Ditto.
15861 (rs6000_emit_epilogue): Ditto.
15862 (rs6000_elf_file_end): Ditto.
15863 (rs6000_function_value): Ditto.
15864 (rs6000_libcall_value): Ditto.
15865 * config/rs6000/rs6000.h: Ditto.
15866 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
15867 (TARGET_MINMAX): ... this. New.
15868 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
15869 * config/rs6000/rs6000.md: Remove Xilinx FP support.
15870 (*movsi_internal1_single): Delete.
15871 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
15872 mfpu=, mxilinx-fpu): Delete.
15873 * config/rs6000/singlefp.h: Delete.
15874 * config/rs6000/sysv4.h: Remove Xilinx FP support.
15875 * config/rs6000/t-rs6000: Ditto.
15876 * config/rs6000/t-xilinx: Delete.
15877 * config/rs6000/titan.md: Adjust for fp_type removal.
15878 * config/rs6000/vsx.md: Remove Xilinx FP support.
15879 (VStype_simple): Delete.
15880 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
15881 * config/rs6000/xfpu.h: Delete.
15882 * config/rs6000/xfpu.md: Delete.
15883 * config/rs6000/xilinx.h: Delete.
15884 * config/rs6000/xilinx.opt: Delete.
15885 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove
15886 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
15888 2018-05-04 Tom de Vries <tom@codesourcery.com>
15891 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
15894 2018-05-04 Richard Biener <rguenther@suse.de>
15896 PR middle-end/85627
15897 * tree-complex.c (update_complex_assignment): We are always in SSA form.
15898 (expand_complex_div_wide): Likewise.
15899 (expand_complex_operations_1): Likewise.
15900 (expand_complex_libcall): Preserve EH info of the original stmt.
15901 (tree_lower_complex): Handle removed blocks.
15902 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
15903 on complex multiplication and division libcall builtins.
15905 2018-05-04 Richard Biener <rguenther@suse.de>
15907 PR middle-end/85574
15908 * fold-const.c (negate_expr_p): Restrict negation of operand
15909 zero of a division to when we know that can happen without
15911 (fold_negate_expr_1): Likewise.
15913 2018-05-04 Jakub Jelinek <jakub@redhat.com>
15916 * real.h (real_nextafter): Declare.
15917 * real.c (real_nextafter): New function.
15918 * fold-const-call.c (fold_const_nextafter): New function.
15919 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
15920 CASE_CFN_NEXTTOWARD.
15921 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
15922 even when arg1_mode is different from arg0_mode.
15924 2018-05-03 Nathan Sidwell <nathan@acm.org>
15926 * doc/extend.texi (Deprecated Features): Remove
15927 -ffriend-injection.
15928 (Backwards Compatibility): Likewise.
15929 * doc/invoke.texi (C++ Language Options): Likewise.
15930 (C++ Dialect Options): Likewise.
15932 2018-05-03 Jakub Jelinek <jakub@redhat.com>
15935 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
15936 _mm512_mask_mullox_epi64): New intrinsics.
15938 2018-05-03 Tom de Vries <tom@codesourcery.com>
15941 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
15942 dump files): Add offload-tree.
15944 2018-05-03 Richard Biener <rguenther@suse.de>
15946 PR tree-optimization/85615
15947 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
15948 to loops not nested in BBs loop father to avoid creating multi-entry
15951 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15953 PR tree-optimization/70291
15954 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
15955 arguments. Change return type to tree. Emit libcall as a new
15956 statement rather than replacing existing one when inplace_p is true.
15957 (expand_complex_multiplication_components): New function.
15958 (expand_complex_multiplication): Expand floating-point complex
15959 multiplication using the above.
15960 (expand_complex_division): Rename inner_type parameter to type.
15961 Update expand_complex_libcall call-site.
15962 (expand_complex_operations_1): Update expand_complex_multiplication
15963 and expand_complex_division call-sites.
15965 2018-05-02 Jakub Jelinek <jakub@redhat.com>
15968 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
15969 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
15970 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
15971 the highest significant bit of the shift count mask is clear. In
15972 check whether and[sq]i3 is needed verify that all significant bits
15973 of the shift count other than the highest are set.
15975 2018-05-02 Tom de Vries <tom@codesourcery.com>
15978 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
15979 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
15980 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
15981 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
15982 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
15983 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
15984 __builtin_goacc_parlevel_size.
15986 2018-05-02 Richard Biener <rguenther@suse.de>
15988 PR tree-optimization/85597
15989 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
15990 do not use split vect_get_vec_defs call but call vect_get_slp_defs
15993 2018-05-02 Tom de Vries <tom@codesourcery.com>
15996 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
15997 dump files): Add ltrans-tree.
15999 2018-05-02 Tom de Vries <tom@codesourcery.com>
16002 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
16003 dump files): Add wpa-ipa.
16005 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
16007 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
16008 powerpc*-*-linux*paired* target.
16009 * config/rs6000/750cl.h: Delete.
16010 * config/rs6000/paired.h: Delete.
16011 * config/rs6000/paired.md: Delete.
16012 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
16014 * config/rs6000/rs6000-builtin.def: Remove paired float support.
16015 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
16016 comment. Remove paired float support.
16017 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
16018 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
16020 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
16021 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
16023 * config/rs6000/rs6000.c: Remove paired float support.
16024 (paired_expand_vector_init, paired_expand_vector_move,
16025 paired_emit_vector_compare, paired_emit_vector_cond_expr,
16026 (paired_expand_lv_builtin, paired_expand_stv_builtin,
16027 paired_expand_builtin, paired_expand_predicate_builtin,
16028 paired_init_builtins): Delete.
16029 * config/rs6000/rs6000.h: Remove paired float support.
16030 * config/rs6000/rs6000.md: Remove paired float support.
16031 (move_from_CR_ov_bit): Delete.
16032 * config/rs6000/rs6000.opt (mpaired): Delete.
16033 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
16034 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
16036 2018-05-02 Richard Biener <rguenther@suse.de>
16038 PR middle-end/85567
16039 * gimplify.c (gimplify_save_expr): When in SSA form allow
16040 SAVE_EXPRs to compute to SSA vars.
16042 2018-05-02 Jakub Jelinek <jakub@redhat.com>
16045 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
16046 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
16047 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
16048 clobber operands[2], instead use a new pseudo. Formatting fixes.
16050 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
16052 PR tree-optimization/85586
16053 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
16054 exit early for statements in the same group if the accesses are
16057 2018-05-02 Tom de Vries <tom@codesourcery.com>
16060 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
16063 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
16065 PR tree-optimization/85143
16066 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
16068 2018-05-01 Tom de Vries <tom@codesourcery.com>
16071 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
16072 not found" error message.
16074 2018-05-01 Tom de Vries <tom@codesourcery.com>
16077 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
16078 * vec.c (test_ordered_remove_if): New function.
16079 (vec_c_tests): Call test_ordered_remove_if.
16080 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
16081 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
16082 * tree-vect-patterns.c (vect_pattern_recog_1): Use
16083 VEC_ORDERED_REMOVE_IF.
16085 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
16087 PR tree-optimization/82665
16088 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
16089 pointer subtraction where arguments come from a memchr call.
16091 2018-05-01 Jakub Jelinek <jakub@redhat.com>
16093 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
16094 --push-state --as-needed and --pop-state instead of --as-needed and
16095 --no-as-needed if ld supports it.
16096 * configure: Regenerated.
16099 * doc/install.texi2html: Replace _002d with - and _002a with * in
16100 generated html files using sed.
16102 2018-04-30 David Malcolm <dmalcolm@redhat.com>
16105 * gcc-rich-location.c (blank_line_before_p): New function.
16106 (use_new_line): New function.
16107 (gcc_rich_location::add_fixit_insert_formatted): New function.
16108 * gcc-rich-location.h
16109 (gcc_rich_location::add_fixit_insert_formatted): New function.
16111 2018-04-30 David Malcolm <dmalcolm@redhat.com>
16113 * selftest.c (assert_streq): Rename "expected" and "actual" to
16114 "val1" and "val2". Extend NULL-handling to cover both inputs
16115 symmetrically, while still requiring both to be non-NULL for a pass.
16116 * selftest.h (assert_streq): Rename "expected" and "actual" to
16118 (ASSERT_EQ): Likewise.
16119 (ASSERT_EQ_AT): Likewise.
16120 (ASSERT_KNOWN_EQ): Likewise.
16121 (ASSERT_KNOWN_EQ_AT): Likewise.
16122 (ASSERT_NE): Likewise.
16123 (ASSERT_MAYBE_NE): Likewise.
16124 (ASSERT_MAYBE_NE_AT): Likewise.
16125 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
16126 the assertion to pass.
16127 (ASSERT_STREQ_AT): Likewise.
16129 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
16131 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
16132 interaction with -pie.
16134 2018-04-30 David Malcolm <dmalcolm@redhat.com>
16136 * selftest.h: Fix alphabetization of per-source-file selftest
16139 2018-04-30 Jason Merrill <jason@redhat.com>
16141 PR c++/61982 - dead stores to destroyed objects.
16142 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
16145 2018-04-30 Jason Merrill <jason@redhat.com>
16147 * tree.c (build_clobber): New.
16148 * tree.h: Declare it.
16149 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
16151 2018-04-30 David Malcolm <dmalcolm@redhat.com>
16153 * diagnostic-show-locus.c (layout::layout): Update for
16154 location_get_source_line returning a char_span.
16155 (struct char_span): Move to input.h.
16156 (struct correction): Update for fields in char_span becoming
16158 (struct source_line): Update for location_get_source_line
16159 returning a char_span.
16160 (layout::print_line): Likewise.
16161 * edit-context.c (edited_file::print_content): Likewise.
16162 (edited_file::print_diff_hunk): Likewise.
16163 (edited_file::print_run_of_changed_lines): Likewise.
16164 (edited_file::get_num_lines): Likewise.
16165 (edited_line::edited_line): Likewise.
16166 * final.c (asm_show_source): Likewise.
16167 * input.c (location_get_source_line): Convert return type
16168 from const char * to char_span, losing the final "line_len"
16170 (dump_location_info): Update for the above.
16171 (get_substring_ranges_for_loc): Likewise. Use a char_span
16172 when handling the literal within the line.
16173 (test_reading_source_line): Update for location_get_source_line
16174 returning a char_span.
16175 * input.h (class char_span): Move here from
16176 diagnostic-show-locus.c, converting from a struct to a class.
16177 Make data members private.
16178 (char_span::operator bool): New.
16179 (char_span::length): New.
16180 (char_span::get_buffer): New.
16181 (char_span::operator[]): New.
16182 (char_span::subspan): Make const.
16183 (char_span::xstrdup): New.
16184 (location_get_source_line): Convert return type from const char *
16185 to char_span, losing the final "line_size" param.
16187 2018-04-30 Jan Hubicka <jh@suse.cz>
16189 * lto-wrapper.c (ltrans_priorities): New static var.
16190 (cmp_priority): New.
16191 (run_gcc): Read priorities and if doing parallel build order
16192 the Makefile by them.
16194 2018-04-30 David Malcolm <dmalcolm@redhat.com>
16196 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
16198 2018-04-30 Richard Biener <rguenther@suse.de>
16200 * tree-cfg.c (verify_address): Remove base argument, add
16201 flag whether to check TREE_ADDRESSABLE and do that.
16202 (verify_expr): Remove.
16203 (verify_types_in_gimple_reference): Add pieces from verify_expr.
16204 (verify_gimple_assign_single): Likewise.
16205 (verify_gimple_switch): Likewise.
16206 (verify_expr_location_1): Dereference tp once. Add (disabled)
16207 piece from verify_expr.
16208 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
16210 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
16212 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
16214 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
16216 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
16217 (small_data_pattern): Likewise.
16218 (arc_rewrite_small_data): Likewise.
16219 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
16220 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
16221 (get_symbol_alignment): New function.
16222 (legitimate_small_data_address_p): Likewise.
16223 (legitimate_scaled_address): Update, call
16224 legitimate_small_data_address_p.
16225 (output_sdata): New static variable.
16226 (arc_print_operand): Update how we handle small data operands.
16227 (arc_print_operand_address): Likewise.
16228 (arc_legitimate_address_p): Update, use
16229 legitimate_small_data_address_p.
16230 (arc_rewrite_small_data_p): Remove.
16231 (arc_rewrite_small_data_1): Likewise.
16232 (arc_rewrite_small_data): Likewise.
16233 (small_data_pattern): Likewise.
16234 (compact_sda_memory_operand): Update to use
16235 legitimate_small_data_address_p and get_symbol_alignment.
16236 (prepare_move_operands): Don't rewite sdata pattern.
16237 (prepare_extend_operands): Remove.
16238 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
16240 (zero_extendqisi2): Likewise.
16241 (zero_extendhisi2): Likewise.
16242 (extendqihi2): Likewise.
16243 (extendqisi2): Likewise.
16244 (extendhisi2): Likewise.
16245 (addsi3): Likewise.
16246 (subsi3): Likewise.
16247 (andsi3): Likewise.
16248 * config/arc/constraints.md (Usd): Change it to memory constraint.
16250 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
16252 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
16253 as source of std instructions.
16254 * config/arc/arc.md (movsi_insn): Update pattern predicate to
16255 allow 6-bit constants as source for store instructions.
16256 (movdi_insn): Update instruction pattern to allow 6-bit constants
16257 as source for store instructions.
16259 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
16261 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
16263 2018-04-30 Nathan Sidwell <nathan@acm.org>
16264 Sandra Loosemore <sandra@codesourcery.com>
16266 * dumpfile.c (dump_open): Allow '-' for stdout.
16267 * doc/invoke.texi (Developer Options): Document dump filename
16268 determination early. Document stdin/stdout selection.
16270 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
16272 Microblaze Target: PIC data text relative
16274 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
16275 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
16277 * config/microblaze/microblaze.h (microblaze_constant_address_p):
16278 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
16279 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
16280 New addressing mode for data-text relative position indepenedent code.
16281 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
16282 'ADDRESS_SYMBOLIC_TXT_REL'.
16283 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
16284 (microblaze_legitimate_pic_operand): Exclude function calls from
16285 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
16286 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
16288 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
16289 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
16290 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
16291 for 'address + offset'.
16292 (microblaze_expand_prologue): Add new function prologue call for
16294 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
16295 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
16296 table in case of TARGET_PIC_DATA_TEXT_REL.
16297 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
16298 * config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
16299 Add new macros 'UNSPEC_TEXT',
16300 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
16301 + exclude function calls from 'UNSPEC_PLT' in case of data text
16303 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
16304 new target hook for generating address diff vector tables in case of
16306 * doc/tm.texi : Regenerate.
16307 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
16308 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
16309 of addr diff vector generation.
16310 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
16311 target hook definition.
16312 * targhooks.h, targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
16313 Add default function for generate_pic_addr_diff_vec -> flag_pic.
16314 * doc/invoke.texi (Add new pic option): Add new microblaze pic
16315 option for data text relative.
16317 2018-04-30 Richard Biener <rguenther@suse.de>
16319 * tree-chrec.h (evolution_function_is_constant_p): Remove
16321 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
16323 2018-04-30 Richard Biener <rguenther@suse.de>
16326 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
16328 2018-04-30 Richard Biener <rguenther@suse.de>
16330 PR tree-optimization/28364
16331 PR tree-optimization/85275
16332 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
16333 copying first exit test.
16335 2018-04-28 Mark Wielaard <mark@klomp.org>
16337 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
16338 dwarf_version >= 5.
16339 (dwarf_AT): Handle DW_AT_addr_base.
16340 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
16342 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
16345 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
16346 (*ashl<dwi>3_doubleword_mask_1): Ditto.
16347 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
16348 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
16350 2018-04-28 Richard Biener <rguenther@suse.de>
16352 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
16353 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
16354 to reflect use. Only add interesting stmts.
16356 2018-04-27 Martin Jambor <mjambor@suse.cz>
16359 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
16360 the jump function allows for passing through aggregate values.
16362 2018-04-27 David Malcolm <dmalcolm@redhat.com>
16364 * input.h (in_system_header_at): Convert from macro to inline
16366 (from_macro_expansion_at): Likewise.
16367 (from_macro_definition_at): Likewise.
16369 2018-04-27 Jeff Law <law@redhat.com>
16371 * config.gcc: Mark tile* targets as deprecated/obsolete.
16373 2018-04-27 Richard Biener <rguenther@suse.de>
16375 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
16378 2018-04-27 Richard Biener <rguenther@suse.de>
16380 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
16382 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
16384 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
16385 with Yd constraint. Set "preferred_for_speed" attribute from
16386 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
16387 with Yd constraint.
16388 (*movdi_internal): Ditto.
16389 (movti_interunit splitters): Remove
16390 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
16391 (movdi_interunit splitters): Ditto.
16392 * config/i386/constraints.md (Ye): Remove.
16393 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
16395 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
16398 * config/aarch64/constraints.md (Usg): Limit to 31.
16399 (Usj): Limit to 63.
16401 2018-04-27 Jakub Jelinek <jakub@redhat.com>
16403 PR tree-optimization/85529
16404 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
16405 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
16406 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
16407 zero extension or masking of the MSB bit.
16408 (optimize_range_tests): Add FIRST_BB argument, pass it through
16409 to optimize_range_tests_var_bound.
16410 (maybe_optimize_range_tests, reassociate_bb): Adjust
16411 optimize_range_tests callers.
16413 2018-04-26 Richard Biener <rguenther@suse.de>
16414 Jakub Jelinek <jakub@redhat.com>
16416 * cgraph.h (symbol_table): Just declare debug method here.
16417 * symtab.c (symbol_table::debug): Define.
16419 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
16421 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
16423 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
16425 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
16426 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
16427 (*movdi_internal): Substitute Yi and Yj constraint with x
16428 and Ym and Yn constraint with y constraint. Update "isa"
16429 attribute and set "preferred_for_speed" attribute from
16430 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
16431 (*movsi_internal): Ditto.
16432 (*movdf_internal): Ditto.
16433 (*movsf_internal): Ditto.
16434 (*zero_extendsidi2): Ditto.
16435 * config/i386/sse.md (vec_set<mode>_0): Ditto.
16436 (sse2_loadld): Ditto.
16437 (*vec_extract<ssevecmodelower>_0): Ditto.
16438 (*vec_extractv4si_0_zext_sse4): Ditto.
16439 (vec_concatv2di): Ditto.
16440 (*vec_dup<mode>): Ditto.
16441 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
16442 * config/i386/constraints.md (Yi): Remove.
16447 2018-04-26 Nathan Sidwell <nathan@acm.org>
16449 * dumpfile.c (dump_open): New.
16450 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
16451 (dump_finish): Detect stdio/stderr by value not name.
16453 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
16455 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
16457 2018-04-26 Tom de Vries <tom@codesourcery.com>
16460 * config/nvptx/nvptx.c (verify_neutering_jumps)
16461 (verify_neutering_labels): New function
16462 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
16464 2018-04-26 Tom de Vries <tom@codesourcery.com>
16467 * config/nvptx/nvptx.c (needs_neutering_p): New function.
16468 (nvptx_single): Use needs_neutering_p to skip over insns that do not
16471 2018-04-26 Richard Biener <rguenther@suse.de>
16472 Tom de Vries <tom@codesourcery.com>
16475 * lto-streamer-out.c (output_function): Fixup loops if required to match
16476 discovery done in the reader.
16478 2018-04-26 Richard Biener <rguenther@suse.de>
16480 PR tree-optimization/85116
16481 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
16482 have a loop exit from the single latch predecessor. Remove
16483 case of header with just condition.
16484 (ch_base::copy_headers): Exclude infinite loops from any
16486 (pass_ch::execute): Record exits.
16488 2018-04-26 Richard Biener <rguenther@suse.de>
16490 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
16491 prologue cost vector and pass it to vect_get_load_cost.
16492 (vect_get_peeling_costs_all_drs): Likewise.
16493 (vect_peeling_hash_get_lowest_cost): Likewise.
16494 (vect_enhance_data_refs_alignment): Likewise.
16496 2018-04-26 Richard Biener <rguenther@suse.de>
16498 PR middle-end/85450
16499 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
16500 checking of integer<->pointer conversions.
16501 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
16502 sign-/zero-extending pointer types.
16503 (expand_omp_for_static_chunk): Likewise.
16505 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
16506 Jean Lee <xiaoyur347@gmail.com>
16508 * config/mips/mips.c (mips_asan_shadow_offset): New function.
16509 (TARGET_ASAN_SHADOW_OFFSET): Define.
16510 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
16511 true for -fsanitize=address.
16513 2018-04-25 Mark Wielaard <mark@klomp.org>
16515 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
16518 2018-04-25 Jakub Jelinek <jakub@redhat.com>
16520 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
16521 than "alu", remove explicit "memory" and "imm_disp" attributes.
16522 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
16524 PR middle-end/85414
16525 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
16526 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
16527 gen_lowpart_no_emit.
16529 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
16532 * config/i386/i386.c (ix86_expand_builtin): Change memory
16533 operand to XI, extend p0 to Pmode.
16534 * config/i386/i386.md: Change unspec volatile and operand
16535 1 mode to XI, change operand 0 mode to P.
16537 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
16539 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
16540 GET_MODE_MASK before any checking.
16541 (nds32_can_use_bset_p): Likewise.
16542 (nds32_can_use_btgl_p): Likewise.
16544 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
16546 * config/nds32/nds32-doubleword.md: New define_split pattern for
16547 illegal register number.
16549 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
16551 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
16553 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
16555 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
16557 2018-04-25 Richard Biener <rguenther@suse.de>
16559 * lto-streamer.h (LTO_major_version): Bump to 8.
16561 2018-04-25 Jakub Jelinek <jakub@redhat.com>
16563 * BASE-VER: Set to 9.0.0.
16565 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
16567 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
16568 in __abskf2 and __powikf2.
16570 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
16573 * config/aarch64/constraints.md (Usg, Usj): New constraints.
16574 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
16575 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
16576 Use the above on operand 2. Reindent.
16577 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
16579 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
16582 * common/config/i386/i386-common.c (ix86_handle_option): Don't
16584 * config/i386/i386.opt (mcet): Removed.
16585 * doc/install.texi: Remove -mcet documentation.
16586 * doc/invoke.texi: Likewise.
16588 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
16591 * doc/install.texi: Remove -mcet from bootstrap-cet.
16593 2018-04-24 Jakub Jelinek <jakub@redhat.com>
16596 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
16597 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
16601 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
16602 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
16603 containing a CONST_VECTOR.
16605 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
16607 * doc/install.texi: Update newlib dependency for nvptx.
16609 2018-04-24 Jakub Jelinek <jakub@redhat.com>
16612 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
16613 instead of INTVAL when shifting x left.
16615 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
16617 PR tree-optimization/85478
16618 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
16619 vect_grouped_store_supported for single element vectors.
16621 2018-04-24 Richard Biener <rguenther@suse.de>
16624 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
16625 load cost increase to the case of non-constant step.
16627 2018-04-24 Jakub Jelinek <jakub@redhat.com>
16630 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
16631 destination if any_malformed_asm.
16633 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
16635 PR middle-end/85496
16636 * expr.c (store_field): In the bitfield case, if the value comes from
16637 a function call and is returned in registers by means of a PARALLEL,
16638 do not change the mode of the temporary unless BLKmode and VOIDmode.
16640 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
16642 PR rtl-optimization/85423
16643 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
16644 dependencies to debug insns when the previous insn is non-debug.
16646 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
16648 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
16649 enums into a single definition.
16650 (fls): Fix predicates and printing.
16653 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
16655 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
16656 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
16657 and short u6 immediate.
16658 (check_if_valid_sleep_operand): Remove.
16659 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
16661 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
16663 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
16664 flag_always_save_lp condition.
16665 * config/nds32/nds32.opt (malways-save-lp): New option.
16667 2018-04-22 Shiva Chen <shiva0217@gmail.com>
16669 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
16670 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
16671 * config/nds32/nds32.h
16672 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
16673 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
16675 2018-04-22 Shiva Chen <shiva0217@gmail.com>
16677 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
16678 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
16680 2018-04-22 Shiva Chen <shiva0217@gmail.com>
16681 Chung-Ju Wu <jasonwucj@gmail.com>
16683 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
16685 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
16686 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
16688 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
16690 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
16692 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
16694 * config/nds32/nds32-protos.h (nds32_data_alignment,
16695 nds32_local_alignment): Declare.
16696 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
16697 nds32_local_alignment): New functions.
16698 (TARGET_CONSTANT_ALIGNMENT): Define.
16699 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
16701 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
16703 * config/nds32/nds32.c
16704 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
16705 (TARGET_MODES_TIEABLE_P): Likewise.
16707 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
16709 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
16710 level Ofast and Og.
16712 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
16713 Chung-Ju Wu <jasonwucj@gmail.com>
16715 * config/nds32/constants.md (unspec_volatile_element): Add enum values
16716 for unaligned access.
16717 * config/nds32/nds32-intrinsic.c: Implementation of expanding
16719 * config/nds32/nds32-intrinsic.md: Likewise.
16720 * config/nds32/nds32_intrinsic.h: Likewise.
16721 * config/nds32/nds32.h (nds32_builtins): Likewise.
16722 * config/nds32/nds32.opt (munaligned-access): New option.
16723 * config/nds32/nds32.c (nds32_asm_file_start): Display
16724 flag_unaligned_access status.
16726 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
16728 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
16729 -mno-relax is present.
16730 * config/riscv/linux.h (LINK_SPEC): Ditto.
16732 2018-04-20 Martin Sebor <msebor@redhat.com>
16735 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
16737 (gimple_fold_builtin_stxcpy_chk): Same.
16738 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
16740 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
16743 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
16744 __powikf2 when long double is IEEE 128-bit.
16746 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
16748 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
16749 step to make sure stack always aligned.
16751 2018-04-20 Carl Love <cel@us.ibm.com>
16754 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
16755 size check for arg0.
16757 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
16758 Tom de Vries <tom@codesourcery.com>
16761 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
16762 Emit insns for calls too.
16763 (nvptx_find_par): Always look for worker-level predecessor insn.
16764 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
16766 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
16767 (nvptx_process_pars): Propagate frames for calls.
16769 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
16772 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
16774 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
16775 (ix86_handle_option): Don't handle OPT_mibt.
16776 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
16778 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
16780 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
16781 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
16782 (ix86_target_macros): Define __CET__ with flag_cf_protection
16783 for -fcf-protection.
16784 * config/i386/i386.c (isa2_opts): Remove -mibt.
16785 * config/i386/i386.h (TARGET_IBT): Removed.
16786 (TARGET_IBT_P): Likewise.
16787 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
16788 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
16789 * config/i386/i386.opt (mcet): Update help message.
16790 (mshstk): Likewise.
16792 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
16793 -mcet as an alias for -mshstk.
16795 2018-04-20 Richard Biener <rguenther@suse.de>
16797 PR middle-end/85475
16798 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
16799 complexity by forcing a single use of the multiply operand.
16801 2018-04-20 Martin Jambor <mjambor@suse.cz>
16804 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
16805 recursion dependency to only apply to non-clones.
16807 2018-04-20 Martin Jambor <mjambor@suse.cz>
16810 * ipa-cp.c (create_specialized_node): Check that clones of
16811 self-recursive edges exist during IPA-CP.
16813 2018-04-19 Toon Moene <toon@moene.org>
16815 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
16818 2018-04-19 Jakub Jelinek <jakub@redhat.com>
16820 PR tree-optimization/85467
16821 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
16822 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
16823 VECTOR_CST element to type.
16825 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
16828 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
16829 * config/i386/i386.md (builtin_setjmp_setup): Removed.
16830 (builtin_longjmp): Likewise.
16831 (save_stack_nonlocal): New pattern.
16832 (restore_stack_nonlocal): Likewise.
16834 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
16837 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
16838 Replace ASM_OUTPUT_LABEL with fprintf.
16840 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
16843 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
16844 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
16845 * config/i386/i386-c.c (ix86_target_macros_internal): Also
16846 define __IBT__ and __SHSTK__ for -fcf-protection.
16847 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
16849 (ix86_trampoline_init): Likewise.
16850 (x86_output_mi_thunk): Likewise.
16851 (ix86_notrack_prefixed_insn_p): Likewise.
16852 (ix86_option_override_internal): Don't disallow -fcf-protection.
16853 * config/i386/i386.md (rdssp<mode>): Also enable for
16855 (incssp<mode>): Likewise.
16856 (nop_endbr): Likewise.
16857 * config/i386/i386.opt (mcet): Change help message to built-in
16860 (mshstk): Likewise.
16861 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
16862 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
16863 enable CET built-in functions.
16865 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
16867 * common/config/i386/i386-common.c
16868 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
16869 OPTION_MASK_ISA_MOVDIRI_UNSET,
16870 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
16871 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
16872 * config.gcc (movdirintrin.h): New header.
16873 * config/i386/cpuid.h (bit_MOVDIRI,
16874 bit_MOVDIR64B): New bits.
16875 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
16877 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
16878 (VOID, PVOID, PCVOID)): New function types.
16879 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
16880 __builtin_ia32_directstoreu_u64,
16881 __builtin_ia32_movdir64b): New builtins.
16882 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
16883 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
16885 (ix86_valid_target_attribute_inner_p): Ditto.
16886 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
16887 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
16888 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
16889 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
16890 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
16891 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
16892 (movdiri<mode>, movdir64b_<mode>): New.
16893 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
16894 * config/i386/immintrin.h: Include movdirintrin.h.
16895 * config/i386/movdirintrin.h: New file.
16896 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
16898 2018-04-19 Richard Biener <rguenther@suse.de>
16900 PR middle-end/85455
16901 * cfg.c (clear_bb_flags): When loop state says we have
16902 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
16904 2018-04-19 Richard Biener <rguenther@suse.de>
16906 PR tree-optimization/84737
16907 * tree-vect-data-refs.c (vect_copy_ref_info): New function
16908 copying restrict info.
16909 (vect_setup_realignment): Use it.
16910 * tree-vectorizer.h (vect_copy_ref_info): Declare.
16911 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
16912 the first DR to all generated stores.
16913 (vectorizable_load): Likewise for loads.
16915 2018-04-19 Jakub Jelinek <jakub@redhat.com>
16917 PR tree-optimization/85446
16918 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
16919 the integral and pointer types to have the same precision.
16921 * doc/install.texi: Document --disable-cet being the default and
16924 2018-04-18 Martin Liska <mliska@suse.cz>
16926 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
16929 2018-04-18 Martin Liska <mliska@suse.cz>
16932 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
16935 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
16936 arguments if they are comparable.
16938 2018-04-18 Martin Liska <mliska@suse.cz>
16941 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
16944 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
16947 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
16950 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
16951 ENDBR after calling __morestack.
16953 2018-04-18 David Malcolm <dmalcolm@redhat.com>
16956 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
16957 by using gcc_base_ver to generate a gcc_driver_version, and use
16958 it when generating GCC_DRIVER_NAME.
16959 * configure: Regenerate.
16961 2018-04-18 Jakub Jelinek <jakub@redhat.com>
16964 * config.gcc: Obsolete powerpc*-*-*spe*.
16966 2018-04-17 Jakub Jelinek <jakub@redhat.com>
16969 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
16970 (stabstr_D): Change type of unum from unsigned int to
16971 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
16974 2018-04-17 Jim Wilson <jimw@sifive.com>
16977 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
16978 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
16979 Set arg_pointer_offset after using pretend_args_size.
16981 2018-04-17 Jakub Jelinek <jakub@redhat.com>
16983 PR rtl-optimization/85431
16984 * dse.c (record_store): Ignore zero width stores.
16987 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
16988 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
16989 __builtin_stack_restore rather than after it.
16990 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
16991 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
16992 argument instead of virtual_dynamic_stack_rtx.
16994 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
16996 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
16998 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
16999 Add note to error message to explain internal mapping of overloaded
17000 built-in function name to non-overloaded built-in function name.
17001 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
17004 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
17007 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
17008 where the inputs overlap with the output.
17010 2018-04-17 Jakub Jelinek <jakub@redhat.com>
17012 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
17013 (=v, v) alternative and explicit "memory" attribute.
17014 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
17015 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
17017 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
17018 "sselog1" type instead of "sselog".
17019 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
17020 "sselog". Remove explicit "memory" attribute.
17021 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
17022 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
17024 (vec_extract_hi_v32hi): Merge all alternatives into one, use
17025 "sselog1" type instead of "sselog". Remove explicit "memory"
17027 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
17028 use "sselog1" type instead of "sselog". Remove explicit "memory"
17030 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
17031 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
17033 (vec_extract_hi_v64qi): Merge all alternatives into one, use
17034 "sselog1" type instead of "sselog". Remove explicit "memory"
17036 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
17037 use "sselog1" type instead of "sselog". Remove explicit "memory"
17041 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
17043 PR middle-end/85414
17044 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
17047 2018-04-17 Martin Jambor <mjambor@suse.cz>
17050 * ipa-cp.c (create_specialized_node): Call
17051 expand_all_artificial_thunks if necessary.
17053 2018-04-17 Martin Liska <mliska@suse.cz>
17056 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
17057 in message, remote space in between '_G' and '('.
17059 2018-04-17 Jakub Jelinek <jakub@redhat.com>
17062 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
17063 avx512f_vmcmp<mode>3<round_saeonly_name>,
17064 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
17065 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
17066 avx512f_rndscale<mode><round_saeonly_name>,
17067 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
17068 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
17069 Use %<iptr>2 instead of %2 for -masm=intel.
17070 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
17071 avx512f_vcvttss2usi<round_saeonly_name>,
17072 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
17074 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
17075 avx512f_vcvttsd2usi<round_saeonly_name>,
17076 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
17077 Use %q1 instead of %1 for -masm=intel.
17078 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
17079 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
17080 of %3 for -masm=intel.
17081 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
17083 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
17085 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
17087 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
17088 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
17090 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
17092 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
17093 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
17094 %g1 and one with %0 and %1.
17095 (avx512er_vmrcp28<mode><round_saeonly_name>,
17096 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
17097 %1 for -masm=intel.
17098 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
17099 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
17100 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
17101 of %0 and %{%4%} for -masm=intel.
17102 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
17103 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
17104 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
17105 order of %0 and %{%5%}%{z%} for -masm=intel.
17107 2018-04-17 Jan Hubicka <jh@suse.cz>
17110 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
17112 2018-04-17 Martin Liska <mliska@suse.cz>
17115 * multiple_target.c (create_dispatcher_calls): Set apostrophes
17116 for target_clone error message. Make default implementation
17117 clone to be a local declaration.
17118 (separate_attrs): Add new argument and check for an empty
17120 (expand_target_clones): Handle it.
17121 (ipa_target_clone): Make redirection just for target_clones
17124 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
17125 Tom de Vries <tom@codesourcery.com>
17127 PR middle-end/84955
17128 * omp-expand.c (expand_oacc_for): Add dummy false branch for
17129 tiled basic blocks without omp continue statements.
17131 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
17134 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
17135 vec_extract expression as having side effects to make sure it gets
17138 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
17141 * config/i386/i386.c (get_builtin_code_for_version): Check
17144 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
17147 * config.gcc: Support "skylake".
17148 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
17150 * config/i386/i386.c (m_SKYLAKE): Define.
17151 (processor_target_table): Add "skylake".
17152 (ix86_option_override_internal): Add "skylake".
17153 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
17154 PROCESSOR_CANNONLAKE.
17155 (get_builtin_code_for_version): Fix priority for
17156 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
17157 PROCESSOR_SKYLAKE-AVX512.
17158 * config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
17159 (processor_type): Add PROCESSOR_SKYLAKE.
17161 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
17162 Jason Merrill <jason@redhat.com>
17165 * convert.c (convert_to_integer_1): Use direct recursion for
17166 enumeral types and types with a precision less than the number
17167 of bits in their mode.
17169 2018-04-16 Julia Koval <julia.koval@intel.com>
17172 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
17173 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
17175 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
17178 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
17179 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
17180 and -mno-direct-move.
17182 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
17185 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
17186 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
17187 Ensure negative shifts result in {0}.
17189 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
17191 PR rtl-optimization/79916
17192 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
17193 regs (if any) to define how to gnerate SD moves when LRA is in
17196 2018-04-13 Jakub Jelinek <jakub@redhat.com>
17198 PR rtl-optimization/85393
17199 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
17200 * except.c (expand_dw2_landing_pad_for_region): Make static.
17201 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
17202 a label and unconditional jump to old_bb, rather than
17203 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
17206 PR rtl-optimization/85376
17207 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
17208 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
17209 instead of a specific value.
17211 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
17212 Bin Cheng <bin.cheng@arm.com>
17214 PR tree-optimization/82965
17215 PR tree-optimization/83991
17216 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
17217 by_profile_only parameter.
17218 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
17219 information if the loop was predicted to iterate too many times.
17220 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
17222 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
17225 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
17228 2018-04-13 Martin Liska <mliska@suse.cz>
17229 Jakub Jelinek <jakub@redhat.com>
17231 PR middle-end/81657
17232 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
17233 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
17234 * builtins.c (expand_builtin_memory_copy_args): Use
17235 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
17236 handle dest_addr == pc_rtx.
17238 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
17241 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
17242 asked to not generate direct moves.
17243 (fix_trunc<mode>si2_stfiwx): Similar.
17244 (fix_trunc<mode>si2_internal): Similar.
17246 2018-04-12 Jakub Jelinek <jakub@redhat.com>
17249 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
17250 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
17251 lookup if dest in some wider mode is known to be const0_rtx and
17252 if so, record permanent equivalence for it to be ZERO_EXTEND of
17253 the narrower mode destination.
17255 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
17257 * lto-streamer-out.c (output_function): Revert 259346.
17258 * omp-expand.c (expand_oacc_for): Likewise.
17260 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
17262 PR rtl-optimization/85354
17263 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
17264 * sel-sched.c (sel_global_init): ... here.
17266 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
17269 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
17270 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
17271 mode for PE-COFF targets.
17272 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
17273 (i386_pe_asm_lto_end): Likewise.
17274 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
17275 (TARGET_ASM_LTO_END): Likewise.
17276 * config/i386/winnt.c (saved_debug_info_level): New static variable.
17277 (i386_pe_asm_lto_start): New function.
17278 (i386_pe_asm_lto_end): Likewise.
17280 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
17281 Richard Biener <rguenther@suse.de>
17283 PR middle-end/84955
17284 * lto-streamer-out.c (output_function): Fix CFG loop state before
17286 * omp-expand.c (expand_oacc_for): Handle calls to internal
17287 functions like regular functions.
17289 2018-04-12 Richard Biener <rguenther@suse.de>
17292 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
17293 for the early LTO debug to properly generate references to it
17294 during DIE emission. Do not re-use that for the skeleton for
17296 (dwarf2out_early_finish): Likewise.
17298 2018-04-12 Jakub Jelinek <jakub@redhat.com>
17301 * config/i386/sse.md
17302 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
17303 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
17304 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
17305 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
17306 and output is a reg, avoid creating invalid lowpart subreg, but
17307 instead split into a 512-bit move. Don't split if not AVX512VL,
17308 input is xmm16+ reg and output is a mem.
17309 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
17310 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
17311 xmm16+ reg and output is a mem.
17313 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
17315 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
17316 also for flag_dwarf2_cfi_asm.
17318 2018-04-12 Jakub Jelinek <jakub@redhat.com>
17320 PR rtl-optimization/85342
17321 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
17322 a bool scalar var inside of the loop instead. Don't try to update
17323 recog_data.operand after failed apply_change_group.
17325 2018-04-12 Tom de Vries <tom@codesourcery.com>
17328 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
17329 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
17330 array with flexible array member as array without given dimension.
17331 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
17332 argument for undefined param to true.
17334 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
17337 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
17338 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
17339 from PowerPC section.
17340 * config/rs6000/sysv4.opt (mcall-): Improve help text.
17341 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
17342 help text that is too long.
17343 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
17344 help text that is too long.
17345 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
17346 help text that is too long.
17348 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
17350 * config/alpha/alpha.md (stack_probe_internal): Rename
17351 from "probe_stack". Update all callers.
17353 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
17355 PR rtl-optimization/84566
17356 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
17357 sched_macro_fuse_insns.
17359 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
17362 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
17363 (compute_block_dependences): ... from here.
17365 2018-04-11 Jakub Jelinek <jakub@redhat.com>
17367 PR tree-optimization/85331
17368 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
17369 from int to HOST_WIDE_INT.
17371 2018-04-11 Martin Jambor <mjambor@suse.cz>
17374 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
17375 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
17376 not the same as the source val.
17377 (cgraph_edge_brings_value_p): New parameter.
17378 (gather_edges_for_value): Pass destination value to
17379 cgraph_edge_brings_value_p.
17380 (perhaps_add_new_callers): Likewise.
17381 (get_info_about_necessary_edges): Likewise and exclude values brought
17382 only by self-recursive edges.
17383 (create_specialized_node): Redirect only clones of self-calling edges.
17384 (+self_recursive_pass_through_p): New function.
17385 (find_more_scalar_values_for_callers_subset): Use it.
17386 (find_aggregate_values_for_callers_subset): Likewise.
17387 (known_aggs_to_agg_replacement_list): Removed.
17388 (decide_whether_version_node): Re-calculate known constants for all
17389 remaining context clones.
17391 2018-04-11 Richard Biener <rguenther@suse.de>
17394 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
17395 from early DWARF output.
17396 (dwarf2out_early_finish): Output line info unconditionally into
17397 early DWARF and add reference to it.
17399 2018-04-11 Jakub Jelinek <jakub@redhat.com>
17402 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
17403 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
17404 other than V2DFmode using iptr mode attribute.
17405 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
17407 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
17409 PR rtl-optimization/84659
17410 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
17412 2018-04-11 Jakub Jelinek <jakub@redhat.com>
17415 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
17417 (output_loc_list): Pass address of a dummy size variable even in the
17418 locview handling loop.
17419 (index_location_lists): Add comment on why skip_loc_list_entry can't
17422 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
17425 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
17428 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
17431 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
17432 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
17433 and -mstring-compare-inline-limit.
17435 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
17438 * config/rs6000/rs6000.md (allocate_stack): Put the residual size
17439 for stack clash protection in a register whenever we need it to be in
17442 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
17444 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
17445 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
17447 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
17450 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
17452 (mlong-double-): Ditto.
17453 * config/rs6000/sysv4.opt (msdata=): Ditto.
17454 (mtls-size=): Ditto.
17456 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
17458 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
17459 erroneous entries for
17460 "vector int vec_ldl (int, long int *)", and
17461 "vector unsigned int vec_ldl (int, unsigned long int *)".
17462 Add comments and entries for
17463 "vector bool char vec_ldl (int, bool char *)",
17464 "vector bool short vec_ldl (int, bool short *)",
17465 "vector bool int vec_ldl (int, bool int *)",
17466 "vector bool long long vec_ldl (int, bool long long *)",
17467 "vector pixel vec_ldl (int, pixel *)",
17468 "vector long long vec_ldl (int, long long *)",
17469 "vector unsigned long long vec_ldl (int, unsigned long long *)".
17470 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
17471 type tree bool_long_long_type_node and correct definition of
17472 bool_V2DI_type_node to make reference to this new type tree.
17473 (rs6000_mangle_type): Replace erroneous reference to
17474 bool_long_type_node with bool_long_long_type_node.
17475 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
17476 comments to emphasize sign distinctions for char and int types and
17477 replace RS6000_BTI_bool_long constant with
17478 RS6000_BTI_bool_long_long constant. Also add comment to restrict
17479 use of RS6000_BTI_pixel.
17480 (bool_long_type_node): Remove this macro definition.
17481 (bool_long_long_type_node): New macro definition
17483 2018-04-10 Jakub Jelinek <jakub@redhat.com>
17485 PR rtl-optimization/85300
17486 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
17487 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
17488 simplify_unary_operation fails.
17490 2018-04-10 Martin Liska <mliska@suse.cz>
17492 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
17493 cgraph_edge and ipa_ref.
17495 2018-04-10 Jakub Jelinek <jakub@redhat.com>
17499 * config/i386/sse.md
17500 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
17501 computation of the VEC_MERGE selector from mask.
17502 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
17503 Fix decoding of the VEC_MERGE selector into mask.
17505 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
17507 PR tree-optimization/85286
17508 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
17510 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
17512 * final.c (final_1): Set insn_last_address as well as
17513 insn_current_address.
17515 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
17518 * explow.c (emit_stack_probe): Call validize_mem on memory location
17519 before passing it to gen_probe_stack. Create address operand and
17520 legitimize it for the probe_stack_address case.
17522 2018-04-09 Jan Hubicka <jh@suse.cz>
17525 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
17526 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
17527 * tree.c (free_lang_data_in_type): Fix handling of binfos;
17529 (free_lang_data): Rebuild type inheritance graph.
17531 2018-04-09 Martin Sebor <msebor@redhat.com>
17533 * invoke.texi (-finline-small-functions): Mention other optimization
17535 (-findirect-inlining, -fpartial-inlining): Same.
17536 (-finline-functions-called-once): Same.
17537 (-freorder-blocks-and-partition): Same.
17539 2018-04-09 Jan Hubicka <jh@suse.cz>
17542 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
17543 jumps; choose last target that matches the criteria (i.e.
17544 no partition changes for non-crossing jumps).
17545 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
17546 support for redirecting crossing jumps to non-crossing.
17548 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
17550 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
17551 also for naked functions.
17553 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
17555 * config/arc/arc.md (add_shift): New pattern.
17556 (add_shift2): Likewise.
17557 (sub_shift): Likewise.
17558 (sub_shift_cmp0_noout): Likewise.
17559 (compare_si_ashiftsi): Likewise.
17560 (xbfu_cmp0_noout): New combine pattern.
17561 (xbfu_cmp0"): Likewise.
17562 (movsi_set_cc_insn): Place the predicable variant first.
17563 (commutative_binary_cmp0_noout): Remove clobber.
17564 (commutative_binary_cmp0): New pattern.
17565 (noncommutative_binary_cmp0): Likewise.
17566 (noncommutative_binary_cmp0_noout): Likewise.
17567 (noncommutative_binary_comparison_result_used): Removed.
17568 (rsub_cmp0): New pattern.
17569 (rsub_cmp0_noout): Likewise.
17570 (extzvsi): Changed, keep only meaningful variants.
17571 (SQH, SEZ): New iterators.
17572 (SQH_postfix): New mode attribute.
17573 (SEZ_prefix): New code attribute.
17574 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
17575 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
17576 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
17577 of numerical value.
17578 (noncommutative_operator): Check the availability of barrel
17581 2018-04-09 Richard Biener <rguenther@suse.de>
17583 PR tree-optimization/85284
17584 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
17585 Only use the niter constraining form of simple_iv when the exit
17586 is always executed.
17588 2018-04-09 Tom de Vries <tom@codesourcery.com>
17591 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
17592 (define_expand "*memory_barrier"): New define_expand.
17593 (define_insn "memory_barrier"): New insn.
17595 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
17597 PR rtl-optimization/80463
17598 PR rtl-optimization/83972
17599 PR rtl-optimization/83480
17601 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
17602 correct producer for the insn.
17603 (tidy_control_flow): Fixup seqnos in case of debug insns.
17605 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
17607 PR rtl-optimization/83913
17609 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
17610 different sched-times when merging exprs.
17612 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
17614 PR rtl-optimization/83962
17616 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
17617 tidy_fallthru_edge and tidy_control_flow.
17619 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
17621 PR rtl-optimization/83530
17623 * sel-sched.c (force_next_insn): New global variable.
17624 (remove_insn_for_debug): When force_next_insn is true, also leave only
17625 next insn in the ready list.
17626 (sel_sched_region): When the region wasn't scheduled, make another pass
17627 over it with force_next_insn set to 1.
17629 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
17631 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
17633 * config/nds32/constants.md (unspec_volatile_element): Add enum values
17634 for interrupt control.
17635 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
17636 functions for interrupt control.
17637 * config/nds32/nds32-intrinsic.md: Likewise.
17638 * config/nds32/nds32_intrinsic.h: Likewise.
17639 * config/nds32/nds32.h (nds32_builtins): Likewise.
17641 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
17643 * config/nds32/nds32.c (nds32_init_machine_status,
17644 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
17645 strict_aligned_p field.
17646 (nds32_expand_to_rtl_hook): New function.
17647 (TARGET_EXPAND_TO_RTL_HOOK): Define.
17648 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
17650 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
17651 Chung-Ju Wu <jasonwucj@gmail.com>
17653 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
17654 * config/nds32/nds32-n7.md: New file.
17655 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
17656 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
17658 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
17659 * config/nds32/nds32.md (pipeline_model): Add n7.
17660 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
17661 * config/nds32/pipelines.md: Include n7 settings.
17663 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
17664 Chung-Ju Wu <jasonwucj@gmail.com>
17666 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
17667 * config/nds32/nds32-e8.md: New file.
17668 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
17669 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
17671 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
17672 * config/nds32/nds32.md (pipeline_model): Add e8.
17673 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
17674 * config/nds32/pipelines.md: Include e8 settings.
17676 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
17677 Chung-Ju Wu <jasonwucj@gmail.com>
17679 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
17680 * config/nds32/nds32-n8.md: New file.
17681 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
17682 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
17684 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
17685 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
17686 * config/nds32/nds32.md (pipeline_model): Add n8.
17687 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
17688 * config/nds32/pipelines.md: Include n8 settings.
17690 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
17691 Chung-Ju Wu <jasonwucj@gmail.com>
17693 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
17694 * config/nds32/nds32-n9-2r1w.md: New file.
17695 * config/nds32/nds32-n9-3r2w.md: New file.
17696 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
17697 nds32_register_ports): New or modify for cpu n9.
17698 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
17700 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
17701 * config/nds32/nds32-utils.c: New file.
17702 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
17703 TARGET_MUL_SLOW): Define.
17704 * config/nds32/nds32.md (pipeline_model): New attribute.
17705 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
17706 New options that support cpu n9.
17707 * config/nds32/pipelines.md: Include n9 settings.
17708 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
17710 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
17712 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
17713 information if necessary.
17714 (output_cond_branch_compare_zero): Likewise.
17715 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
17716 (nds32_target_alignment): Refine for alignment.
17717 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
17718 (FUNCTION_BOUNDARY): Modify.
17719 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
17721 * config/nds32/nds32.opt (malways-align, malign-functions): New.
17723 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
17725 * config/nds32/constants.md (unspec_volatile_element): Add values for
17726 TLB operation and data prefetch.
17727 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
17728 functions for TLB operation and data prefetch.
17729 * config/nds32/nds32-intrinsic.md: Likewise.
17730 * config/nds32/nds32_intrinsic.h: Likewise.
17731 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
17732 (nds32_print_operand): Likewise.
17733 * config/nds32/nds32.h (nds32_builtins): Likewise.
17735 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
17736 Andrew Pinski <pinsika@gcc.gnu.org>
17738 PR middle-end/82976
17739 * match.pd: Use constant_boolean_node of correct type instead of
17740 boolean_true_node or boolean_false_node for simplifying
17741 pointer comparisons to zero.
17743 2018-04-07 Jakub Jelinek <jakub@redhat.com>
17745 PR tree-optimization/80021
17746 * tree.c (verify_type_variant): Make error call in verify_variant_match
17747 translatable and remove final full stop.
17749 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
17751 * config/nds32/constants.md (unspec_volatile_element): Add
17752 UNSPEC_VOLATILE_EH_RETURN.
17753 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
17754 nds32_output_stack_pop): Support dwarf exception handling process.
17755 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
17756 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
17757 exception handling process.
17758 (nds32_compute_stack_frame): Likewise.
17759 (nds32_return_addr_rtx): Likewise.
17760 (nds32_initial_elimination_offset): Likewise.
17761 (nds32_expand_prologue): Likewise.
17762 (nds32_expand_epilogue): Likewise.
17763 (nds32_dynamic_chain_address): New function.
17764 * config/nds32/nds32.h (machine_function): Add fields for dwarf
17765 exception handling.
17766 (DYNAMIC_CHAIN_ADDRESS): Define.
17767 (EH_RETURN_DATA_REGNO): Define.
17768 (EH_RETURN_STACKADJ_RTX): Define.
17769 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
17770 patterns for dwarf exception handling.
17772 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
17774 * config/nds32/nds32.h: Clean up obsolete macros.
17776 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
17778 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
17779 Add enum values for particular instructions.
17780 * config/nds32/nds32-intrinsic.c: Implementation of expanding
17781 particular intrinsic functions.
17782 * config/nds32/nds32-intrinsic.md: Likewise.
17783 * config/nds32/nds32_intrinsic.h: Likewise.
17784 * config/nds32/nds32.h (nds32_builtins): Likewise.
17785 * config/nds32/nds32.md (type): Add pbsad and pbsada.
17786 (btst, ave): New patterns for particular instructions.
17788 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
17790 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
17791 Add enum values for atomic load/store and memory sync.
17792 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
17794 * config/nds32/nds32-intrinsic.md: Likewise.
17795 * config/nds32/nds32_intrinsic.h: Likewise.
17796 * config/nds32/nds32.h (nds32_builtins): Likewise.
17798 2018-04-07 Jakub Jelinek <jakub@redhat.com>
17800 PR tree-optimization/85257
17801 * fold-const.c (native_encode_vector): If not all elts could fit
17802 and off is -1, return 0 rather than offset.
17803 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
17804 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
17805 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
17806 adjust buffer in native_interpret_expr call.
17808 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
17810 * config/nds32/constants.md (unspec_volatile_element): Add cache
17811 control enum values.
17812 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
17813 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
17814 * config/nds32/nds32.c (nds32_cctl_names): New.
17815 (nds32_print_operand): Handle cache control register names.
17816 * config/nds32/nds32.h (nds32_builtins): New enum values.
17817 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
17819 * config/nds32/nds32.md (type): Add mmu.
17820 * config/nds32/pipelines.md (simple_insn): Add mmu.
17822 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
17824 * config/nds32/nds32.md (type): Remove call.
17825 * config/nds32/pipelines.md (simple_insn): Likewise.
17827 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
17829 * config/nds32/constants.md (unspec_volatile_element): Add
17830 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
17831 UNSPEC_VOLATILE_FMFCFG.
17832 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
17833 description for fmfcfg and fmfcsr.
17834 (bdesc_1arg): Add fmtcsr.
17835 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
17836 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
17837 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
17838 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
17839 unspec_fmfcfg): New patterns.
17840 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
17841 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
17842 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
17843 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
17844 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
17845 __nds32__fmfcfg): Define.
17847 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
17849 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
17850 intrinsic register names.
17851 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
17852 intrinsic register enum values and macros.
17854 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
17856 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
17857 for load/store addressing form.
17858 (nds32_print_operand_address): Likewise.
17860 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
17863 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
17864 based on LABEL_REF. Remove useless assertion.
17865 (pic_address_needs_scratch): Fix formatting.
17866 (sparc_legitimize_pic_address): Minor tweaks.
17867 (sparc_delegitimize_address): Adjust assertion accordingly.
17868 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
17869 into symbolic_operand.
17870 (movsi_high_pic_label_ref): Likewise.
17871 (movsi_lo_sum_pic_label_ref): Likewise.
17872 (movdi_pic_label_ref): Likewise.
17873 (movdi_high_pic_label_ref): Likewise.
17874 (movdi_lo_sum_pic_label_ref): Likewise.
17876 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
17878 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
17879 custom LIB_SPEC setup.
17881 2018-04-06 Ruslan Bukin <br@bsdpad.com>
17882 Kito Cheng <kito.cheng@gmail.com>
17884 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
17885 * config/riscv/freebsd.h: New.
17887 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
17889 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
17890 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
17893 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
17894 Kito Cheng <kito.cheng@gmail.com>
17896 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
17897 nds32_output_call, nds32_symbol_binds_local_p): New functions.
17898 * config/nds32/nds32-protos.h (nds32_output_call,
17899 nds32_output_return): Declare.
17900 * config/nds32/nds32.md: Refine all the call and return patterns.
17902 2018-04-06 Jakub Jelinek <jakub@redhat.com>
17905 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
17906 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
17908 PR rtl-optimization/84872
17909 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
17910 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
17911 EDGE_CROSSING edge.
17913 2018-04-06 Tamar Christina <tamar.christina@arm.com>
17915 * expr.c (copy_blkmode_to_reg): Revert 254862.
17916 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
17918 2018-04-06 Richard Biener <rguenther@suse.de>
17920 PR middle-end/85244
17921 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
17922 after seeing a component reference with an adjacent field. Treat
17923 refs to arrays at struct end of external decls similar to
17924 refs to unconstrained commons.
17926 2018-04-06 Jakub Jelinek <jakub@redhat.com>
17929 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
17930 look through SAVE_EXPRs with non-side-effects argument. Adjust
17932 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
17935 2018-04-06 Richard Biener <rguenther@suse.de>
17937 PR middle-end/85180
17938 * alias.c (find_base_term): New wrapper around find_base_term
17939 unwinding CSELIB_VAL_PTR changes.
17940 (find_base_term): Do not restore CSELIB_VAL_PTR during the
17943 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
17945 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
17947 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
17948 constant definitions.
17949 ("nop"): lr 0,0 -> nopr r0
17950 ("nop_lr0", "nop_lr1"): New insn definitions.
17952 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
17954 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
17955 NDS32_V3PUSH_AVAILABLE_P macro.
17957 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
17958 Chung-Ju Wu <jasonwucj@gmail.com>
17960 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
17961 (nds32*-*-*): Add float and fpu_config into supported_defaults.
17962 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
17963 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
17964 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
17965 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
17966 * config/nds32/constraints.md: New constraints and checking for hard
17967 float configuration.
17968 * config/nds32/iterators.md: New mode iterator and attribute for hard
17969 float configuration.
17970 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
17972 * config/nds32/nds32-fpu.md: New file.
17973 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
17974 deal with hard float code generation.
17975 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
17977 (abi_type, float_reg_number): New enum type.
17978 * config/nds32/nds32-predicates.c: New predicates for hard float.
17979 * config/nds32/nds32-protos.h: Declare functions for hard float.
17980 * config/nds32/nds32.c: Implementation for hard float configuration.
17981 * config/nds32/nds32.h: Definitions for hard float configuration.
17982 * config/nds32/nds32.md: Include hard float machine description and
17983 modify patterns for hard float configuration.
17984 * config/nds32/nds32.opt: New options for hard float configuration.
17985 * config/nds32/predicates.md: New predicates for hard float
17988 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
17990 * common/config/nds32/nds32-common.c
17991 (nds32_option_optimization_table): Enable -mreleax-hint by default.
17993 2018-04-05 Jakub Jelinek <jakub@redhat.com>
17995 PR middle-end/85195
17996 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
17997 CONSTRUCTOR_ELT (ctor, ...)->value.
17999 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
18002 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
18004 2018-04-05 Tom de Vries <tom@codesourcery.com>
18007 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
18010 2018-04-05 Shiva Chen <shiva0217@gmail.com>
18011 Kito Cheng <kito.cheng@gmail.com>
18013 * config/nds32/constraints.md (U33): Fine-tune checking condition.
18014 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
18015 * config/nds32/nds32.h (nds32_16bit_address_type): Add
18016 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
18018 2018-04-05 Shiva Chen <shiva0217@gmail.com>
18019 Kito Cheng <kito.cheng@gmail.com>
18021 * config/nds32/constraints.md (Ufe): New memory constraint.
18022 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
18023 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
18024 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
18026 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
18027 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
18029 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
18031 * config/nds32/nds32.md: Use optimize_size in the condition for
18032 alu-shift instructions.
18034 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
18036 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
18038 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
18040 * config/nds32/nds32.md (negsi2): Refine pattern.
18042 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
18043 Chung-Ju Wu <jasonwucj@gmail.com>
18045 * config/nds32/iterators.md (shift_rotate): New code iterator.
18046 (shift): New code attribute.
18047 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
18048 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
18049 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
18050 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
18051 bit-wise operations.
18052 (andsi3, *andsi3): Ditto.
18053 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
18054 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
18055 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
18056 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
18057 nds32_ior_operand, nds32_xor_operand): New predicates.
18059 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
18061 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
18062 (addsi3, subsi3): ... this.
18064 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
18066 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
18068 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
18070 * config/nds32/nds32.md: Adjust indention.
18072 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
18074 * config/nds32/nds32.md (feature): New attribute.
18076 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
18078 * config/nds32/nds32.md (subtype): New attribute.
18080 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
18083 * config/arm/arm-builtins.c (arm_expand_builtin): Change
18084 expansion to perform a bitwise AND of the argument followed by a
18085 boolean negation of the result.
18087 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
18089 PR rtl-optimization/84878
18090 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
18091 the basic block. Assert the use reference is not artificial and that
18092 it has an associated insn.
18094 2018-04-04 Michael Matz <matz@suse.de>
18096 * builtins.c (compute_objsize): Pass correct operand
18097 to array_at_struct_end_p.
18099 2018-04-04 Richard Biener <rguenther@suse.de>
18102 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
18103 from contexts for DINFO_LEVEL_TERSE and below.
18105 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
18107 * config/nds32/nds32-doubleword.md (move_<mode>): Require
18108 resiter_operand condition.
18109 * config/nds32/nds32.md (*move<mode>): Ditto.
18111 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
18112 Monk Chiang <sh.chiang04@gmail.com>
18114 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
18116 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
18118 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
18120 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
18121 Kito Cheng <kito.cheng@gmail.com>
18123 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
18124 nds32_cond_code_str, output_cond_branch,
18125 output_cond_branch_compare_zero, nds32_expand_cbranch,
18126 nds32_expand_cstore, nds32_expand_movcc,
18127 nds32_output_cbranchsi4_equality_zero,
18128 nds32_output_cbranchsi4_equality_reg,
18129 nds32_output_cbranchsi4_equality_reg_or_const_int,
18130 nds32_output_cbranchsi4_greater_less_zero: New functions.
18131 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
18132 nds32_expand_cstore, nds32_expand_movcc,
18133 nds32_output_cbranchsi4_equality_zero,
18134 nds32_output_cbranchsi4_equality_reg,
18135 nds32_output_cbranchsi4_equality_reg_or_const_int,
18136 nds32_output_cbranchsi4_greater_less_zero): Declare.
18137 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
18138 nds32_rimm11s_operand): New predicates.
18139 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
18140 * config/nds32/nds32.md: Rewrite all the branch and conditional move
18143 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
18145 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
18146 * config/nds32/nds32.md: Ditto.
18147 * config/nds32/pipelines.md: Ditto.
18149 2018-04-04 Richard Biener <rguenther@suse.de>
18151 PR tree-optimization/85168
18152 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
18153 propagating abnormals.
18155 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
18157 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
18159 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
18160 Kito Cheng <kito.cheng@gmail.com>
18162 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
18163 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
18164 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
18165 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
18166 * config/nds32/nds32.md (sibcall_internal): New.
18167 (sibcall_register): Remove.
18168 (sibcall_immediate): Remove.
18169 (sibcall_value_internal): New.
18170 (sibcall_value_register): Remove.
18171 (sibcall_value_immediate): Remove.
18172 * config/nds32/predicates.md (nds32_general_register_operand): New.
18173 (nds32_call_address_operand): New.
18175 2018-04-03 Jakub Jelinek <jakub@redhat.com>
18177 PR rtl-optimization/85167
18178 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
18179 bb_defs if *split_p, instead preinitialize it to NULL.
18181 PR tree-optimization/85156
18182 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
18183 evaluating the argument multiple times.
18185 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
18187 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
18189 (_mm_cvtpd_ps): Likewise.
18190 (_mm_cvttpd_epi32): Likewise.
18191 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
18192 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
18193 vector, pixel, and bool following altivec.h include.
18195 2018-04-03 Martin Sebor <msebor@redhat.com>
18197 * doc/extend.texi (Common Function Attributes): Clarify.
18198 (const attribute): Likewise.
18199 (pure attribute): Likewise.
18201 2018-04-03 Jakub Jelinek <jakub@redhat.com>
18204 * config/i386/i386.c (ix86_expand_vector_set): Use
18205 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
18207 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
18209 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
18210 instructions when changing rounding bits to preserve precision bits
18211 in the x87 control word.
18213 2018-04-03 Martin Liska <mliska@suse.cz>
18215 PR tree-optimization/82491
18216 * rtl.h (strip_offset_and_add): Replace += suboffset with
18217 poly_uint64 () + suboffset.
18219 2018-03-29 Martin Liska <mliska@suse.cz>
18220 Martin Jambor <mjambor@suse.cz>
18223 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
18224 param_type is not an integral or pointer type.
18226 2018-04-03 Richard Biener <rguenther@suse.de>
18228 * sese.h (recompute_all_dominators): Remove.
18230 2018-04-02 Martin Sebor <msebor@redhat.com>
18232 * doc/invoke.texi (-Wrestrict): Fix typos.
18234 2018-04-02 Jim Wilson <jimw@sifive.com>
18236 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
18237 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
18238 (<optab>di3, <optab>si3_extend): Likewise.
18239 (<optab>si3_mask, <optab>si3_mask_1): New.
18240 (<optab>di3_mask, <optab>di3_mask_1): New.
18241 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
18242 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
18243 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
18245 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
18247 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
18250 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
18252 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
18253 (nds32_canonicalize_comparison): New function.
18255 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
18256 Kito Cheng <kito.cheng@gmail.com>
18257 Kuan-Lin Chen <kuanlinchentw@gmail.com>
18259 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
18260 * config/nds32/constants.md (unspec_volatile_element): Add
18261 UNSPEC_VOLATILE_RELAX_GROUP.
18262 * config/nds32/nds32-relax-opt.c: New file.
18263 * config/nds32/nds32-predicates.c
18264 (nds32_symbol_load_store_p): New function.
18265 * config/nds32/nds32-protos.h
18266 (nds32_symbol_load_store_p): Declare function.
18267 (make_pass_nds32_relax_opt): Declare new rtl pass function.
18268 * config/nds32/nds32.c
18269 (nds32_register_pass): New function to register pass.
18270 (nds32_register_passes): New function to register passes.
18271 * config/nds32/nds32.md (relax_group): New pattern.
18272 * config/nds32/nds32.opt (mrelax-hint): New option.
18273 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
18275 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
18277 * config/nds32/t-nds32: Modify files dependency.
18279 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
18281 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
18282 (PROFILE_HOOK): Define its implementation.
18284 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
18286 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
18287 type and 32-bit size.
18289 2018-04-01 Jakub Jelinek <jakub@redhat.com>
18291 PR middle-end/85090
18292 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
18293 (V_128_256): New mode iterator.
18294 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
18295 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
18296 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
18298 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
18299 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
18301 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
18304 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
18305 NaN inputs correctly.
18307 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
18310 * config/rs6000/vsx.md (??r): New mode attribute.
18311 (*vsx_mov<mode>_64bit): Use it.
18312 (*vsx_mov<mode>_32bit): Likewise.
18314 2018-03-30 Martin Sebor <msebor@redhat.com>
18316 PR tree-optimization/84818
18317 * builtins.c (check_access): Use warning_n.
18319 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
18322 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
18324 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
18327 2018-03-30 Julia Koval <julia.koval@intel.com>
18330 * x86-tune.def (movx, partial_reg_dependency): Enable for
18333 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
18335 PR inline-asm/84985
18336 * lra-constraints.c (process_alt_operands): Move setting
18337 this_alternative_matches below.
18339 2018-03-29 Martin Liska <mliska@suse.cz>
18342 * doc/invoke.texi: Document how LTO works with debug info.
18343 Describe auto-load support of binutils. Mention 'x86-64'
18344 as valid option value of -march option.
18346 2018-03-29 Jakub Jelinek <jakub@redhat.com>
18348 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
18351 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
18352 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
18353 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
18356 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
18359 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
18360 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
18361 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
18362 for RS6000_BTM_POWERPC64.
18363 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
18364 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
18365 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
18368 (DIVDEU): Likewise.
18370 2018-03-28 Carl Love <cel@us.ibm.com>
18373 2017-09-27 Carl Love <cel@us.ibm.com>
18375 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
18376 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
18377 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
18380 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
18382 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
18383 instead of __vector bool.
18384 (_mm_max_pu8): Likewise.
18385 (_mm_min_pi16): Likewise.
18387 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
18390 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
18391 (DIVWEUO): Likewise.
18392 (DIVDEO): Likewise.
18393 (DIVDEUO): Likewise.
18394 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
18395 DIVWEUO and DIVDEUO.
18396 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
18397 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
18398 (div_extend): Likewise.
18399 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
18401 (__builtin_divweuo): Likewise.
18402 (__builtin_divdeo): Likewise.
18403 (__builtin_divdeuo): Likewise.
18405 2018-03-28 Jakub Jelinek <jakub@redhat.com>
18408 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
18409 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
18411 PR tree-optimization/82004
18412 * gimple-match-head.c (optimize_pow_to_exp): New function.
18413 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
18414 Don't fold to exp if optimize_pow_to_exp is false.
18416 2018-03-28 Martin Liska <mliska@suse.cz>
18419 * calls.c (initialize_argument_information): Fix trailing space.
18420 * common.opt: Fix typo and provide better explanation for
18421 -fsanitize-coverage option.
18422 * config/i386/i386.opt: Fix typo.
18424 2018-03-28 Jakub Jelinek <jakub@redhat.com>
18425 Martin Liska <mliska@suse.cz>
18428 * gimplify.c (asan_poison_variable): Don't do the check for
18429 gimplify_omp_ctxp here.
18430 (gimplify_decl_expr): Do it here.
18431 (gimplify_target_expr): Likewise.
18433 2018-03-28 Martin Liska <mliska@suse.cz>
18436 * config/i386/i386.c (ix86_function_arg_advance): Do not call
18437 chkp_type_bounds_count if MPX is not enabled.
18439 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
18441 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
18443 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
18446 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
18447 function to create the function decl for complex long double
18448 multiply and divide for -mabi=ieeelongdouble.
18449 (init_float128_ieee): Call it.
18451 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
18454 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
18455 -fcf-protection=branch -mibt.
18456 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
18458 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
18461 * config/arm/arm.c (arm_valid_symbolic_address): Handle
18462 arm_word_relocations.
18464 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
18467 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
18468 extern array declarations.
18470 2018-03-27 Richard Biener <rguenther@suse.de>
18472 PR middle-end/84067
18473 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
18474 explicit single_use checks.
18476 2018-03-27 Richard Biener <rguenther@suse.de>
18478 PR tree-optimization/85082
18479 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
18482 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
18484 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
18485 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
18486 Turn on fasynchronous-unwind-tables and funwind-tables.
18488 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
18491 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
18492 (*bmi_blsr_<mode>_ccz): Ditto.
18494 2018-03-26 Tom de Vries <tom@codesourcery.com>
18496 PR tree-optimization/85063
18497 * omp-general.c (offloading_function_p): New function. Factor out
18499 * omp-offload.c (pass_omp_target_link::gate): ... here.
18500 * omp-general.h (offloading_function_p): Declare.
18501 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
18502 with attribute omp declare target for offloading functions.
18504 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
18506 PR tree-optimization/84005
18507 * tree-data-ref.h (get_base_for_alignment): Declare.
18508 * tree-data-ref.c (get_base_for_alignment_1): New function.
18509 (get_base_for_alignment): Likewise.
18510 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
18511 get_base_for_alignment to find a suitable base object, instead
18512 of always using drb->base_address.
18514 2018-03-23 Jakub Jelinek <jakub@redhat.com>
18516 PR inline-asm/85022
18517 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
18518 known size by default.
18520 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
18522 PR inline-asm/85030
18523 * lra-constraints.c (process_alt_operands): Don't match BLKmode
18524 and non BLKmode operands.
18526 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
18529 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
18530 Clean up attributes.
18532 2018-03-23 Richard Biener <rguenther@suse.de>
18535 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
18536 we are going to emit early debug for LTO.
18538 2018-03-23 Jakub Jelinek <jakub@redhat.com>
18540 PR inline-asm/85034
18541 * function.c (match_asm_constraints_1): Don't optimize if input
18542 doesn't satisfy general_operand predicate for output's mode.
18544 PR inline-asm/85022
18545 * alias.c (write_dependence_p): Don't require for x_canonicalized
18546 non-VOIDmode if x has VOIDmode.
18549 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
18550 just don't try to optimize it rather than assert it never happens.
18552 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
18554 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
18555 macro expansions for definition of ST_INTERNAL_<mode> and
18556 LD_INTERNAL_<mode> builtins.
18557 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
18559 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
18561 (altivec_expand_st_builtin): Likewise.
18562 (altivec_expand_builtin): Remove calls to deleted functions.
18563 (rs6000_address_for_altivec): Delete this function.
18564 * config/rs6000/vector.md: Remove expands for
18565 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
18567 2018-03-22 Sudakshina Das <sudi.das@arm.com>
18570 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
18571 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
18572 re-computing once computed.
18573 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
18574 (arm_init_machine_status): Initialize
18575 machine->static_chain_stack_bytes.
18577 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
18580 * doc/extend.texi: Add four new prototypes for vec_ld.
18581 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
18582 definitions for more logical presentation.
18583 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
18584 entries for V1TI variants of __builtin_altivec_ld builtin.
18585 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
18586 handling of V1TI variant of LVX icode pattern.
18587 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
18588 (rs6000_gimple_fold_builtin): Likewise.
18589 (altivec_init_builtins): Add code to define
18590 __builtin_altivec_lvx_v1ti function.
18592 2018-03-22 Jakub Jelinek <jakub@redhat.com>
18594 PR inline-asm/84941
18595 * function.c (match_asm_constraints_1): Don't do the optimization
18596 if input isn't a REG, SUBREG, MEM or constant.
18598 2018-03-22 Tom de Vries <tom@codesourcery.com>
18600 PR tree-optimization/84956
18601 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
18602 bb_has_abnormal_pred.
18604 2018-03-22 Jakub Jelinek <jakub@redhat.com>
18607 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
18608 DECL_INITIAL (decl) to decl at the end.
18609 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
18610 adjust the comment.
18612 2018-03-21 Joseph Myers <joseph@codesourcery.com>
18614 * doc/extend.texi (__builtin_tgmath): Document when complex
18615 integer types are treated as _Complex _Float64.
18617 2018-03-21 Tom de Vries <tom@codesourcery.com>
18619 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
18621 2018-03-21 Jakub Jelinek <jakub@redhat.com>
18623 PR tree-optimization/84960
18624 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
18625 if it is ENTRY block, move them into single succ of ENTRY in that case.
18627 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
18629 PR tree-optimization/84811
18630 * poly-int.h (poly_span_traits): Remove the T3 parameter and
18631 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
18632 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
18633 (known_subrange_p): Update accordingly. Cast each value involved
18634 in the size comparison, rather than casting the result of the
18637 2018-03-21 Jakub Jelinek <jakub@redhat.com>
18639 PR tree-optimization/84982
18640 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
18641 by flipping the least significant bit rather than all bits from
18642 bitpos to bitpos + bitsize - 1.
18644 2018-03-21 Nathan Sidwell <nathan@acm.org>
18646 * doc/extend.texi (Deprecated Features): Remove mention of
18647 long-deleted deprecations.
18649 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
18652 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
18653 * configure: Regenerate.
18655 2018-03-21 Tom de Vries <tom@codesourcery.com>
18657 PR tree-optimization/83126
18658 * tree-parloops.c (num_phis): New function.
18659 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
18661 2018-03-21 Nathan Sidwell <nathan@acm.org>
18663 * doc/extend.texi (Deprecated Features): Update deprecated flags,
18664 mention anon-struct/union members and trailing attributes.
18666 2018-03-21 Bin Cheng <bin.cheng@arm.com>
18668 PR tree-optimization/84969
18669 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
18670 builtin memset partitions if they set different rhs values.
18672 2018-03-21 Jakub Jelinek <jakub@redhat.com>
18674 PR rtl-optimization/84989
18675 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
18676 VEC_DUPLICATE with scalar result mode.
18678 2018-03-21 Martin Liska <mliska@suse.cz>
18681 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
18682 not intended return statement.
18684 2018-03-21 Martin Liska <mliska@suse.cz>
18687 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
18688 (chkp_find_bound_slots_1): Limit number of iterations.
18690 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
18693 * Minor grammar fixes for x86 options.
18695 2018-03-20 Jakub Jelinek <jakub@redhat.com>
18698 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
18699 holding REG_CFA_RESTORE notes, instead turn them into a USE.
18701 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
18704 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
18705 (altivec_lvx_<mode>_1op): Likewise.
18706 (altivec_stvx_<mode>_2op): Likewise.
18707 (altivec_stvx_<mode>_1op): Likewise.
18708 (altivec_lvx_<VM2:mode>): New define_expand.
18709 (altivec_stvx_<VM2:mode>): Likewise.
18710 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
18711 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
18712 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
18713 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
18714 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
18715 (rs6000_gen_lvx): Likewise.
18716 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
18717 (altivec_expand_stv_builtin): Likewise.
18718 (altivec_expand_builtin): Likewise.
18719 * config/rs6000/vector.md: Likewise.
18721 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
18724 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
18727 2018-03-20 Richard Biener <rguenther@suse.de>
18730 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
18731 sign-conversions as zero, fall back to standard scalar_stmt
18734 2018-03-20 Martin Liska <mliska@suse.cz>
18737 * predict.c (rebuild_frequencies): Handle case when we have
18738 PROFILE_ABSENT, but flag_guess_branch_prob is false.
18740 2018-03-20 Jakub Jelinek <jakub@redhat.com>
18743 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
18744 flag_section_anchors.
18745 * varasm.c (use_blocks_for_decl_p): Remove hack for
18746 dw2_force_const_mem.
18749 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
18751 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
18752 be created, use lowpart_subreg of operands[0] rather than operands[0]
18754 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
18755 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
18756 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
18757 and n constraint instead of aarch64_shift_imm_di and Usd.
18758 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
18759 (*aarch64_<optab>_reg_minus<mode>3): ... this.
18761 2018-03-20 Sudakshina Das <sudi.das@arm.com>
18764 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
18765 to favor GPR over NEON registers.
18766 (<shift>di3_neon): Likewise.
18768 2018-03-20 Tom de Vries <tom@codesourcery.com>
18771 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
18772 (nvptx_process_pars): Emit bar.sync asap and alap.
18774 2018-03-20 Tom de Vries <tom@codesourcery.com>
18777 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
18778 seen_label if seen_label is already set.
18780 2018-03-20 Jakub Jelinek <jakub@redhat.com>
18783 * config/i386/i386.c (fold_builtin_cpu): For features above 31
18784 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
18785 Use 1U instead of 1. Formatting fixes.
18788 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
18789 instead of TREE_TYPE (s1) for the return value.
18791 2018-03-19 Jakub Jelinek <jakub@redhat.com>
18793 PR tree-optimization/84946
18794 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
18795 bitsize + bitsize in poly_uint64 rather than poly_int64.
18798 * dwarf2asm.c: Include fold-const.c.
18799 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
18800 of decl rather than decl itself.
18802 PR rtl-optimization/84643
18803 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
18805 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
18808 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
18809 calling assemble_variable.
18811 2018-03-19 Sudakshina Das <sudi.das@arm.com>
18814 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
18815 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
18817 2018-03-19 Jim Wilson <jimw@sifive.com>
18820 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
18821 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
18822 (riscv_first_stack_step): Likewise.
18823 (riscv_option_override): Use STACK_BOUNDARY instead of
18824 MIN_STACK_BOUNDARY.
18825 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
18826 MIN_STACK_BOUNDARY.
18827 (BIGGEST_ALIGNMENT): Set to 128.
18828 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
18829 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
18832 2018-03-19 Richard Biener <rguenther@suse.de>
18834 PR tree-optimization/84933
18835 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
18836 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
18838 2018-03-19 Richard Biener <rguenther@suse.de>
18840 PR tree-optimization/84859
18841 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
18842 (cond_if_else_store_replacement): Perform sinking operation on
18843 single-store BBs regardless of MAX_STORES_TO_SINK setting.
18844 Generalize what a BB with a single eligible store is.
18846 2018-03-19 Richard Biener <rguenther@suse.de>
18848 PR tree-optimization/84929
18849 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
18850 chrec_is_positive against non-chrec arg.
18852 2018-03-19 Tamar Christina <tamar.christina@arm.com>
18855 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
18857 2018-03-18 Martin Liska <mliska@suse.cz>
18859 PR rtl-optimization/84635
18860 * regrename.c (build_def_use): Use matches_mode only when
18863 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
18865 PR tree-optimization/84913
18866 * tree-vect-loop.c (vectorizable_reduction): Don't try to
18867 vectorize chains of COND_EXPRs.
18869 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
18871 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
18873 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
18875 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
18877 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
18879 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
18881 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
18882 Kito Cheng <kito.cheng@gmail.com>
18884 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
18885 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
18886 (nds32_adjust_reg_alloc_order): New function.
18887 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
18889 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
18891 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
18892 nds32_print_operand, nds32_print_operand_address): Use
18893 HOST_WIDE_INT_PRINT_DEC instead.
18895 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
18897 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
18899 2018-03-17 Jakub Jelinek <jakub@redhat.com>
18902 * config/i386/i386.c (initial_ix86_tune_features,
18903 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
18904 unsigned long long.
18905 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
18906 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
18907 rather than 1u << ix86_tune. Formatting fix.
18908 (ix86_option_override_internal): Change ix86_arch_mask from
18909 unsigned int to unsigned HOST_WIDE_INT, initialize to
18910 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
18911 (ix86_function_specific_restore): Likewise.
18913 2018-03-16 Jakub Jelinek <jakub@redhat.com>
18916 * postreload.c (reload_combine_recognize_pattern): Perform
18917 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
18918 truncate_int_for_mode the result for the destination's mode.
18921 * hsa-gen.c (mem_type_for_type): Fix comment typo.
18922 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
18924 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
18927 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
18930 * lra-assigns.c (lra_split_hard_reg_for): Don't use
18931 regno_allocno_class_array and sorted_pseudos.
18932 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
18933 insns where regno is used.
18935 2018-03-16 Martin Liska <mliska@suse.cz>
18938 * multiple_target.c (create_dispatcher_calls): Redirect
18939 reference in the symbol table.
18941 2018-03-16 Martin Liska <mliska@suse.cz>
18944 * multiple_target.c (create_dispatcher_calls): Redirect also
18947 2018-03-16 Jakub Jelinek <jakub@redhat.com>
18951 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
18952 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
18953 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
18955 2018-03-16 Julia Koval <julia.koval@intel.com>
18957 * doc/invoke.texi (Skylake Server): Add CLWB.
18958 Cannonlake): Remove CLWB.
18960 2018-03-16 Jakub Jelinek <jakub@redhat.com>
18962 PR tree-optimization/84841
18963 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
18965 (FLOAT_ONE_CONST_TYPE): Define.
18966 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
18967 (sort_by_operand_rank): Put entries with higher constant_type last
18968 rather than first to match comments.
18970 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
18972 * config/nios2/nios2.md (movsi_internal): Fix thinko in
18975 2018-03-15 Jakub Jelinek <jakub@redhat.com>
18978 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
18979 check and use address of target always.
18981 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
18984 * config/i386/i386.c (indirect_thunk_needed): Update comments.
18985 (indirect_thunk_bnd_needed): Likewise.
18986 (indirect_thunks_used): Likewise.
18987 (indirect_thunks_bnd_used): Likewise.
18988 (indirect_return_needed): New.
18989 (indirect_return_bnd_needed): Likewise.
18990 (output_indirect_thunk_function): Add a bool argument for
18992 (output_indirect_thunk_function): Don't generate alias for
18993 function return thunk.
18994 (ix86_code_end): Call output_indirect_thunk_function to generate
18995 function return thunks.
18996 (ix86_output_function_return): Set indirect_return_bnd_needed
18997 and indirect_return_needed instead of indirect_thunk_bnd_needed
18998 and indirect_thunk_needed.
19000 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
19002 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
19003 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
19004 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
19006 2018-03-15 David Malcolm <dmalcolm@redhat.com>
19007 Paul Hua <paul.hua.gm@gmail.com>
19010 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
19012 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
19014 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
19015 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
19016 resp. SFmode cases.
19018 2018-03-15 Tamar Christina <tamar.christina@arm.com>
19021 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
19022 instead of GET_MODE_SIZE when comparing Units.
19024 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
19027 * varasm.c (hash_section): Return an unchangeble hash value
19028 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
19029 Return !aarch64_can_use_per_function_literal_pools_p ().
19031 2018-03-15 Jakub Jelinek <jakub@redhat.com>
19034 * optabs.c (emit_conditional_move): Pass address of cmode's copy
19035 rather than address of cmode as last argument to prepare_cmp_insn.
19037 2018-03-15 Julia Koval <julia.koval@intel.com>
19039 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
19040 F_AVX512VNNI, F_AVX512BITALG): New.
19042 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
19045 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
19046 insn for floating-point loads and stores.
19048 2018-03-14 Carl Love <cel@us.ibm.com>
19050 * config/rs6000/rs6000-c.c: Add macro definitions for
19051 ALTIVEC_BUILTIN_VEC_PERMXOR.
19052 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
19053 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
19054 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
19055 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
19057 * config/doc/extend.texi: Add prototypes for vec_permxor.
19059 2018-03-14 David Malcolm <dmalcolm@redhat.com>
19062 * diagnostic-show-locus.c (class layout_point): Convert m_line
19063 from int to linenum_type.
19064 (line_span::comparator): Use linenum "compare" function when
19065 comparing line numbers.
19066 (test_line_span): New function.
19067 (layout_range::contains_point): Convert param "row" from int to
19069 (layout_range::intersects_line_p): Likewise.
19070 (layout::will_show_line_p): Likewise.
19071 (layout::print_source_line): Likewise.
19072 (layout::should_print_annotation_line_p): Likewise.
19073 (layout::print_annotation_line): Likewise.
19074 (layout::print_leading_fixits): Likewise.
19075 (layout::annotation_line_showed_range_p): Likewise.
19076 (struct line_corrections): Likewise for field m_row.
19077 (line_corrections::line_corrections): Likewise for param "row".
19078 (layout::print_trailing_fixits): Likewise.
19079 (layout::get_state_at_point): Likewise.
19080 (layout::get_x_bound_for_row): Likewise.
19081 (layout::print_line): Likewise.
19082 (diagnostic_show_locus): Likewise for locals "last_line" and "row".
19083 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
19084 * input.c (selftest::test_linenum_comparisons): New function.
19085 (selftest::input_c_tests): Call it.
19086 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
19087 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
19088 * selftest.h (ASSERT_GT): New macro.
19089 (ASSERT_GT_AT): New macro.
19090 (ASSERT_LT): New macro.
19091 (ASSERT_LT_AT): New macro.
19093 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
19095 PR rtl-optimization/84780
19096 * combine.c (distribute_links): Don't make a link based on pc_rtx.
19098 2018-03-14 Martin Liska <mliska@suse.cz>
19100 * tree.c (record_node_allocation_statistics): Use
19101 get_stats_node_kind.
19102 (get_stats_node_kind): New function extracted from
19103 record_node_allocation_statistics.
19104 (free_node): Use get_stats_node_kind.
19106 2018-03-14 Richard Biener <rguenther@suse.de>
19108 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
19109 that the value-set of ANTIC_IN doesn't grow.
19112 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
19114 (BB_VISITED_WITH_VISITED_SUCCS): New define.
19115 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
19117 2018-03-14 Julia Koval <julia.koval@intel.com>
19119 * config.gcc (icelake-client, icelake-server): New.
19121 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
19122 (initial_ix86_arch_features): Ditto.
19123 (PTA_SKYLAKE): Add SGX.
19124 (PTA_ICELAKE): Remove.
19125 (PTA_ICELAKE_CLIENT): New.
19126 (PTA_ICELAKE_SERVER): New.
19127 (ix86_option_override_internal): Split up icelake on icelake client and
19129 (get_builtin_code_for_version): Ditto.
19130 (fold_builtin_cpu): Ditto.
19131 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
19132 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
19133 * config/i386/i386.h (processor_type): Ditto.
19134 * doc/invoke.texi: Ditto.
19136 2018-03-14 Jakub Jelinek <jakub@redhat.com>
19139 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
19140 INTEGER_CST offset, add it together with bitpos / 8 and
19141 sign extend based on POINTER_SIZE.
19145 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
19148 * config/i386/constraints.md (Yc): New register constraint.
19149 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
19150 Use Yc constraint for alternative 2 of operand 0. Remove
19151 preferred_for_speed attribute.
19153 2018-03-14 Richard Biener <rguenther@suse.de>
19155 PR tree-optimization/84830
19156 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
19157 with the old one to avoid oscillations.
19159 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
19162 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
19164 (assign_by_spills): Return a flag of reload assignment failure.
19165 Do not process the reload assignment failures. Do not spill other
19166 reload pseudos if they has the same reg class. Update n if
19168 (lra_assign): Add a return arg. Set up from the result of
19169 assign_by_spills call.
19170 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
19171 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
19172 usage_insns if it is not NULL.
19173 (spill_hard_reg_in_range): New function.
19174 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
19175 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
19176 function prototypes.
19177 (lra_assign): Change prototype.
19178 * lra.c (lra): Add code to deal with fails by splitting hard reg
19181 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
19183 * config/riscv/riscv.opt (mrelax): New option.
19184 * config/riscv/riscv.c (riscv_file_start): Emit ".option
19185 "norelax" when riscv_mrelax is disabled.
19186 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
19188 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
19191 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
19192 reassociation for int modes.
19194 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
19196 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
19197 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
19199 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
19200 * config/aarch64/aarch64-sve.md
19201 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
19202 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
19203 (*extend<mode><Vwide>2): Rename to...
19204 (aarch64_sve_extend<mode><Vwide>2): ...this.
19205 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
19206 renaming the old pattern to...
19207 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
19209 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
19210 define_expand, renaming the old pattern to...
19211 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
19212 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
19213 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
19214 account when deciding which SVE instruction the optab should use.
19215 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
19217 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
19219 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
19220 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
19221 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
19222 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
19223 (tlsdesc_small_<mode>): Turn a define_expand and use
19224 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
19225 (tlsdesc_small_advsimd_<mode>): ...this.
19226 (tlsdesc_small_sve_<mode>): New pattern.
19228 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
19230 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
19231 (UNSPEC_UMUL_HIGHPART): New constants.
19232 (MUL_HIGHPART): New int iteraor.
19233 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
19234 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
19236 (*<su>mul<mode>3_highpart): New define_insn.
19238 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
19241 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
19244 2018-03-13 Martin Liska <mliska@suse.cz>
19247 * (sem_item_optimizer::sem_item_optimizer): Initialize new
19249 (sem_item_optimizer::~sem_item_optimizer): Release it.
19250 (sem_item_optimizer::merge_classes): Register variable aliases.
19251 (sem_item_optimizer::fixup_pt_set): New function.
19252 (sem_item_optimizer::fixup_points_to_sets): Likewise.
19253 * ipa-icf.h: Declare new variables and functions.
19255 2018-03-13 Jakub Jelinek <jakub@redhat.com>
19257 PR middle-end/84834
19258 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
19259 integer_pow2p@2 and test integer_pow2p in condition.
19260 (A < 0 ? C : 0): Similarly for @1.
19262 PR middle-end/84831
19263 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
19264 characters starting at p contain '\0' character, don't look beyond
19268 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
19269 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
19272 * reg-stack.c (change_stack): Change update_end var from int to
19273 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
19274 also call set_block_for_insn on the newly added insns and rescan.
19277 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
19278 on the last operand.
19281 * tree.c (stabilize_reference_1): Return save_expr (e) for
19282 STATEMENT_LIST even if it doesn't have side-effects.
19284 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
19286 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
19288 2018-03-12 Renlin Li <renlin.li@arm.com>
19290 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
19291 aarch64_output_scalar_simd_mov_immediate.
19293 2018-03-12 Martin Sebor <msebor@redhat.com>
19295 PR tree-optimization/83456
19296 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
19297 for perfectly overlapping calls to memcpy.
19298 (gimple_fold_builtin_memory_chk): Same.
19299 (gimple_fold_builtin_strcpy): Handle no-warning.
19300 (gimple_fold_builtin_stxcpy_chk): Same.
19301 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
19303 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
19305 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
19306 parameter. Use it for SFmode.
19307 (rs6000_function_arg_advance_1): Adjust.
19308 (rs6000_function_arg): Adjust.
19309 (rs6000_gimplify_va_arg): Pass false for that new parameter.
19311 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
19313 PR rtl-optimization/84169
19314 PR rtl-optimization/84780
19315 * combine.c (can_combine_p): Check for a 2-insn combination whether
19316 the destination register is used between the two insns, too.
19318 2018-03-12 Richard Biener <rguenther@suse.de>
19320 PR tree-optimization/84803
19321 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
19322 for refs DR analysis didn't process.
19324 2018-03-12 Richard Biener <rguenther@suse.de>
19326 PR tree-optimization/84777
19327 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
19328 force-vectorize loops ignore whether we are optimizing for size.
19330 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
19332 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
19333 (TARGET_MD_ASM_ADJUST): Define.
19335 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
19336 Kito Cheng <kito.cheng@gmail.com>
19337 Chung-Ju Wu <jasonwucj@gmail.com>
19339 * config/nds32/nds32.c (nds32_compute_stack_frame,
19340 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
19341 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
19342 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
19343 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
19344 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
19345 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
19346 * config/nds32/nds32.md (prologue, epilogue): Use macro
19347 NDS32_V3PUSH_AVAILABLE_P to do checking.
19349 2018-03-11 Jakub Jelinek <jakub@redhat.com>
19352 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
19353 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
19354 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
19355 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
19356 addition of most attributes on !orig_type_die or the attribute not
19357 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
19359 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
19360 Chung-Ju Wu <jasonwucj@gmail.com>
19362 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
19363 __NDS32_VH__ macro.
19364 * config/nds32/nds32.opt (mvh): New option.
19366 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
19367 Chung-Ju Wu <jasonwucj@gmail.com>
19369 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
19371 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
19372 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
19375 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
19376 Chung-Ju Wu <jasonwucj@gmail.com>
19378 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
19380 * config/nds32/nds32-multiple.md (strlensi): New pattern.
19381 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
19383 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
19384 Kito Cheng <kito.cheng@gmail.com>
19385 Chung-Ju Wu <jasonwucj@gmail.com>
19387 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
19388 UNSPEC_FFMISM and UNSPEC_FLMISM.
19389 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
19390 for ffb, ffmism and flmism.
19391 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
19392 (unspec_ffmism): Ditto.
19393 (unspec_flmism): Ditto.
19394 (nds32_expand_builtin_impl): Check if string extension is available.
19395 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
19396 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
19398 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
19401 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
19404 * lra-assigns.c (assign_by_spills): Return a flag of reload
19405 assignment failure. Do not process the reload assignment
19406 failures. Do not spill other reload pseudos if they has the same
19408 (lra_assign): Add a return arg. Set up from the result of
19409 assign_by_spills call.
19410 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
19411 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
19412 usage_insns if it is not NULL.
19413 (spill_hard_reg_in_range): New function.
19414 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
19415 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
19416 function prototypes.
19417 (lra_assign): Change prototype.
19418 * lra.c (lra): Add code to deal with fails by splitting hard reg
19421 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
19424 * config/i386/i386.opt: Replace Enforcment with Enforcement.
19426 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
19429 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
19430 (dw_val_node): Add val_symbolic_view.
19431 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
19432 (symview_upper_bound): New.
19433 (new_line_info_table): Initialize symviews_since_reset.
19434 (dwarf2out_source_line): Count symviews_since_reset and set
19435 symview_upper_bound.
19436 (dw_val_equal_p): Handle symview.
19437 (add_AT_symview): New.
19438 (print_dw_val): Handle symview.
19439 (attr_checksum, attr_checksum_ordered): Likewise.
19440 (same_dw_val_p, size_of_die): Likewise.
19441 (value_format, output_die): Likewise.
19442 (add_high_low_attributes): Use add_AT_symview for entry_view.
19443 (dwarf2out_finish): Reset symview_upper_bound, clear
19446 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
19449 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
19450 Add strict argument and use it.
19451 (rs6000_split_multireg_move): Update for new strict argument.
19452 (mem_operand_gpr): Disallow all non-offsettable addresses.
19453 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
19455 2018-03-09 Jakub Jelinek <jakub@redhat.com>
19458 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
19459 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
19460 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
19463 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
19464 decl, use remap_type if we want to use the type.
19466 2018-03-09 Martin Sebor <msebor@redhat.com>
19468 PR tree-optimization/84526
19469 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
19471 (builtin_access::generic_overlap): Be prepared to handle non-array
19474 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
19476 PR rtl-optimization/84682
19477 * lra-constraints.c (process_address_1): Check is_address flag
19478 for address constraints.
19479 (process_alt_operands): Likewise.
19480 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
19481 preprocess_constraints.
19482 * recog.h (preprocess_constraints): Add oploc parameter.
19484 * recog.c (preprocess_constraints): Test address_operand for
19485 CT_ADDRESS constraints.
19487 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
19490 * lra-assigns.c (assign_by_spills): Return a flag of reload
19491 assignment failure. Do not process the reload assignment
19492 failures. Do not spill other reload pseudos if they has the same
19494 (lra_assign): Add a return arg. Set up from the result of
19495 assign_by_spills call.
19496 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
19497 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
19498 usage_insns if it is not NULL.
19499 (spill_hard_reg_in_range): New function.
19500 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
19501 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
19502 function prototypes.
19503 (lra_assign): Change prototype.
19504 * lra.c (lra): Add code to deal with fails by splitting hard reg
19507 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
19510 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
19511 Accept complain bool parameter. Only emit errors if it is true.
19512 (arm_parse_cpu_option_name): Likewise.
19513 (arm_target_thumb_only): Adjust callers of the above.
19514 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
19515 prototype to take a default true bool parameter.
19516 (arm_parse_arch_option_name): Likewise.
19518 2018-03-09 David Malcolm <dmalcolm@redhat.com>
19519 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
19523 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
19524 * configure: Regenerate.
19525 * configure.ac ("linker --version-script option"): New.
19526 ("linker soname option"): New.
19528 2018-03-09 Richard Biener <rguenther@suse.de>
19530 PR tree-optimization/84775
19531 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
19532 immediate uses of predicate stmts and mark them modified.
19535 PR tree-optimization/84178
19536 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
19538 (version_loop_for_if_conversion): Delay update_ssa call.
19539 (tree_if_conversion): Delay update_ssa until after predicate
19542 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
19545 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
19546 when the function accesses prior frames.
19548 2018-03-08 Jakub Jelinek <jakub@redhat.com>
19551 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
19552 gen_llsym, otherwise call maybe_gen_llsym.
19554 PR inline-asm/84742
19555 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
19556 has ',' character inside of it.
19558 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
19561 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
19562 as clobbering CC_REGNUM.
19564 2018-03-08 Richard Biener <rguenther@suse.de>
19566 PR middle-end/84552
19567 * tree-scalar-evolution.c: Include tree-into-ssa.h.
19568 (follow_copies_to_constant): Do not follow SSA names registered
19571 2018-03-08 Richard Biener <rguenther@suse.de>
19573 PR tree-optimization/84178
19574 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
19576 (version_loop_for_if_conversion): Delay update_ssa call.
19577 (tree_if_conversion): Delay update_ssa until after predicate
19580 2018-03-08 David Malcolm <dmalcolm@redhat.com>
19582 PR tree-optimization/84178
19583 * tree-if-conv.c (release_bb_predicate): Remove the
19584 the assertion that the stmts have NULL use_ops.
19585 Discard the statements, asserting that they haven't
19586 yet been added to a BB.
19588 2018-03-08 Richard Biener <rguenther@suse.de>
19590 PR tree-optimization/84746
19591 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
19592 (phi_translate): Pass in destination ANTIC_OUT set.
19593 (phi_translate_1): Likewise. For a simplified result lookup
19594 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
19595 (phi_translate_set): Adjust.
19596 (do_pre_regular_insertion): Likewise.
19597 (do_pre_partial_partial_insertion): Likewise.
19599 2018-03-08 Martin Liska <mliska@suse.cz>
19601 PR gcov-profile/84735
19602 * doc/gcov.texi: Document usage of profile files.
19603 * gcov-io.h: Document changes in the format.
19605 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
19609 * dwarf2out.c (struct dw_line_info_table): Update comments for
19611 (FORCE_RESET_NEXT_VIEW): New.
19612 (FORCE_RESETTING_VIEW_P): New.
19613 (RESETTING_VIEW_P): Check for -1 too.
19614 (ZERO_VIEW_P): Likewise.
19615 (new_line_info_table): Force-reset next view.
19616 (dwarf2out_begin_function): Likewise.
19617 (dwarf2out_source_line): Simplify zero_view_p initialization.
19618 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
19619 view directly. Omit view when omitting .loc at line 0.
19621 2018-03-08 Jakub Jelinek <jakub@redhat.com>
19623 PR tree-optimization/84740
19624 * tree-switch-conversion.c (process_switch): Call build_constructors
19625 only if info.phi_count is non-zero.
19627 PR tree-optimization/84739
19628 * tree-tailcall.c (find_tail_calls): Check call arguments against
19629 DECL_ARGUMENTS (current_function_decl) rather than
19630 DECL_ARGUMENTS (func) when checking for tail recursion.
19632 2018-03-07 Jakub Jelinek <jakub@redhat.com>
19634 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
19635 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
19636 Volker Reichelt's entry and add entries for people that perform
19637 GCC fuzzy testing and report numerous bugs.
19639 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
19642 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
19643 readonly data in sdata, if that is disabled.
19644 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
19645 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
19646 -mreadonly-in-sdata option.
19648 2018-03-07 Martin Sebor <msebor@redhat.com>
19650 PR tree-optimization/84468
19651 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
19652 basic block when looking for nul assignment.
19654 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
19657 * except.h (output_function_exception_table): Adjust prototype.
19658 * except.c (output_function_exception_table): Remove FNNAME parameter
19659 and add SECTION parameter. Ouput one part of the table at a time.
19660 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
19661 the first part of the exception table and emit unwind directives.
19662 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
19663 (i386_pe_seh_cold_init): Likewise.
19664 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
19665 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
19666 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
19667 (ix86_output_call_insn): Emit a nop in one more case for SEH.
19668 * config/i386/winnt.c: Include except.h.
19669 (struct seh_frame_state): Add reg_offset, after_prologue and
19670 in_cold_section fields.
19671 (i386_pe_seh_end_prologue): Set seh->after_prologue.
19672 (i386_pe_seh_cold_init): New function.
19673 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
19674 to seh->in_cold_section.
19675 (seh_emit_push): Record the offset of the push.
19676 (seh_emit_save): Record the offet of the save.
19677 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
19678 Test seh->after_prologue to disregard the epilogue.
19679 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
19680 (i386_pe_end_cold_function): New function.
19682 2018-03-07 Jakub Jelinek <jakub@redhat.com>
19685 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
19686 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
19689 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
19691 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
19692 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
19694 PR middle-end/84723
19695 * multiple_target.c: Include tree-inline.h and intl.h.
19696 (expand_target_clones): Diagnose and fail if node->definition and
19697 !tree_versionable_function_p (node->decl).
19699 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
19701 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
19703 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
19704 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
19705 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
19707 2018-03-06 Jakub Jelinek <jakub@redhat.com>
19710 * combine.c (try_combine): Use reg_or_subregno instead of handling
19711 just paradoxical SUBREGs and REGs.
19713 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
19715 * config/arc/arc.c (arc_finalize_pic): Remove function.
19716 (arc_must_save_register): We use single base PIC register, remove
19717 checks to save/restore the PIC register.
19718 (arc_expand_prologue): Likewise.
19719 * config/arc/arc-protos.h (arc_set_default_type_attributes):
19721 (arc_verify_short): Likewise.
19722 (arc_attr_type): Likewise.
19723 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
19724 (walk_stores): Likewise.
19725 (arc_address_cost): Make it static.
19726 (arc_verify_short): Likewise.
19727 (branch_dest): Likewise.
19728 (arc_attr_type): Likewise.
19729 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
19730 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
19731 (arc_final_prescan_insn): Remove inserting the nops due to
19732 hardware hazards. It is done in reorg step.
19733 (insn_length_variant_t): Remove.
19734 (insn_length_parameters_t): Likewise.
19735 (arc_insn_length_parameters): Likewise.
19736 (arc_get_insn_variants): Likewise.
19737 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
19739 2018-03-06 Jakub Jelinek <jakub@redhat.com>
19741 PR inline-asm/84683
19742 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
19745 PR tree-optimization/84687
19746 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
19748 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
19750 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
19752 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
19753 Rename to ppc_speculation_barrier.
19754 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
19755 __builtin_ppc_speculation_barrier.
19757 2018-03-05 Jakub Jelinek <jakub@redhat.com>
19760 * combine.c (combine_simplify_rtx): Don't try to simplify if
19761 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
19764 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
19766 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
19767 to 32 bytes when compiling for POWER9.
19769 2018-03-05 Jakub Jelinek <jakub@redhat.com>
19772 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
19773 regparm >= 3 with no arg reg available also for calls with
19774 flag_force_indirect_call. Pass decl to ix86_function_regparm.
19777 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
19779 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
19781 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
19784 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
19786 2018-03-05 Richard Biener <rguenther@suse.de>
19788 PR tree-optimization/84486
19789 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
19790 When inserting a __builtin_assume_aligned call set the LHS
19791 SSA name alignment info accordingly.
19793 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
19795 PR tree-optimization/84114
19796 * config/aarch64/aarch64.c (aarch64_reassociation_width)
19797 Avoid reassociation of FLOAT_MODE addition.
19799 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
19801 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
19802 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
19803 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
19804 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
19805 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
19806 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
19807 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
19809 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
19810 __builtin_ia32_wbinvd): New builtins.
19811 (SPECIAL_ARGS2): New.
19812 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
19813 (SPECIAL_ARGS2): New.
19814 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
19815 (ix86_valid_target_attribute_inner_p): Ditto.
19816 (ix86_init_mmx_sse_builtins): Add special_args2.
19817 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
19818 TARGET_WBNOINVD_P): New.
19819 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
19820 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
19821 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
19822 * config/i386/immintrin.h (_wbinvd): New intrinsic.
19823 * config/i386/pconfigintrin.h: New file.
19824 * config/i386/wbnoinvdintrin.h: Ditto.
19825 * config/i386/x86intrin.h: Add headers pconfigintrin.h and
19827 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
19829 2018-03-05 Richard Biener <rguenther@suse.de>
19831 PR tree-optimization/84670
19832 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
19834 (BB_VISITED_WITH_VISITED_SUCCS): New define.
19835 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
19836 (compute_antic_aux): Only assert the number of values in ANTIC_IN
19837 doesn't grow if all successors (recursively) were visited at least
19840 2018-03-05 Richard Biener <rguenther@suse.de>
19842 PR tree-optimization/84650
19843 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
19844 if executed in the loop pipeline.
19846 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
19848 * doc/configfiles.texi (Configuration Files): Move info about
19849 conditionalizing $target-protos.h to...
19850 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
19851 differs from $target-protos.h.
19853 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
19854 Chung-Ju Wu <jasonwucj@gmail.com>
19856 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
19857 * config/nds32/nds32-multiple.md (setmemsi): Define.
19858 * config/nds32/nds32-memory-manipulation.c
19859 (nds32_gen_dup_4_byte_to_word_value): New.
19860 (emit_setmem_word_loop): New.
19861 (emit_setmem_byte_loop): New.
19862 (nds32_expand_setmem_loop): New.
19863 (nds32_expand_setmem_loop_v3m): New.
19864 (nds32_expand_setmem_unroll): New.
19865 (nds32_expand_setmem): New.
19867 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
19868 Chung-Ju Wu <jasonwucj@gmail.com>
19870 * config/nds32/nds32-memory-manipulation.c
19871 (nds32_emit_load_store): New.
19872 (nds32_emit_post_inc_load_store): New.
19873 (nds32_emit_mem_move): New.
19874 (nds32_emit_mem_move_block): New.
19875 (nds32_expand_movmemsi_loop_unknown_size): New.
19876 (nds32_expand_movmemsi_loop_known_size): New.
19877 (nds32_expand_movmemsi_loop): New.
19878 (nds32_expand_movmemsi_unroll): New.
19879 (nds32_expand_movmemqi): Rename ...
19880 (nds32_expand_movmemsi): ... to this.
19881 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
19882 (movmemsi): ... to this.
19883 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
19884 (nds32_expand_movmemsi): ... to this.
19886 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
19887 Monk Chiang <sh.chiang04@gmail.com>
19888 Chung-Ju Wu <jasonwucj@gmail.com>
19890 * config/nds32/nds32-protos.h
19891 (nds32_expand_load_multiple): New arguments.
19892 (nds32_expand_store_multiple): Ditto.
19893 (nds32_valid_multiple_load_store): Rename ...
19894 (nds32_valid_multiple_load_store_p): ... to this.
19895 * config/nds32/nds32-memory-manipulation.c
19896 (nds32_expand_load_multiple): Refine implementation.
19897 (nds32_expand_store_multiple): Ditto.
19898 * config/nds32/nds32-multiple.md
19899 (load_multiple): Update nds32_expand_load_multiple interface.
19900 (store_multiple): Update nds32_expand_store_multiple interface.
19901 * config/nds32/nds32-predicates.c
19902 (nds32_valid_multiple_load_store): Rename ...
19903 (nds32_valid_multiple_load_store_p): ... to this and refine
19905 * config/nds32/predicates.md
19906 (nds32_load_multiple_and_update_address_operation): New predicate.
19907 (nds32_store_multiple_and_update_address_operation): New predicate.
19909 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
19910 Chung-Ju Wu <jasonwucj@gmail.com>
19912 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
19913 (combo): New attribute.
19914 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
19916 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
19918 * config/nds32/nds32.opt: Change -mcmodel= default value.
19920 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
19921 Monk Chiang <sh.chiang04@gmail.com>
19922 Chung-Ju Wu <jasonwucj@gmail.com>
19924 * config/nds32/constants.md (unspec_element): New enum.
19925 * config/nds32/constraints.md (Umw): New constraint.
19926 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
19927 * config/nds32/nds32-intrinsic.md: Likewise.
19928 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
19929 (nds32_valid_smw_lwm_base_p): New.
19930 (nds32_output_smw_single_word): New.
19931 (nds32_output_lmw_single_word): New.
19932 (nds32_expand_unaligned_load): New.
19933 (nds32_expand_unaligned_store): New.
19934 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
19935 (nds32_output_smw_single_word): Declare.
19936 (nds32_output_lmw_single_word): Declare.
19937 (nds32_expand_unaligned_load): Declare.
19938 (nds32_expand_unaligned_store): Declare.
19939 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
19940 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
19941 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
19942 NDS32_BUILTIN_UASTORE_DW.
19943 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
19946 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
19947 Kito Cheng <kito.cheng@gmail.com>
19948 Chung-Ju Wu <jasonwucj@gmail.com>
19950 * config/nds32/nds32-intrinsic.c
19951 (nds32_expand_builtin_null_ftype_reg): Delete.
19952 (nds32_expand_builtin_reg_ftype_imm): Ditto.
19953 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
19954 (nds32_read_argument): New.
19955 (nds32_legitimize_target): Ditto.
19956 (nds32_legitimize_argument): Ditto.
19957 (nds32_check_constant_argument): Ditto.
19958 (nds32_expand_unop_builtin): Ditto.
19959 (nds32_expand_unopimm_builtin): Ditto.
19960 (nds32_expand_binop_builtin): Ditto.
19961 (nds32_builtin_decl_impl): Ditto.
19962 (builtin_description): Ditto.
19963 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
19964 (nds32_init_builtins_impl): Ditto.
19965 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
19966 (nds32_builtin_decl): New.
19967 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
19968 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
19970 2018-03-02 Jeff Law <law@redhat.com>
19972 * reorg.c (stop_search_p): Handle DEBUG_INSN.
19973 (redundant_insn, fill_simple_delay_slots): Likewise.
19974 (fill_slots_from_thread): Likewise.
19975 * resource.c (mark_referenced_resources): Likewise.
19976 (mark_set_resources, find_dead_or_set_registers): Likewise.
19978 2018-03-02 Jakub Jelinek <jakub@redhat.com>
19980 * substring-locations.h (format_warning_va): Formatting fix for
19981 ATTRIBUTE_GCC_DIAG.
19982 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
19984 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
19985 * substring-locations.c: Include intl.h.
19986 (format_warning_va): Turned into small wrapper around
19987 format_warning_n_va, renamed to ...
19988 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
19989 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
19991 (format_warning_at_substring_n): New function.
19992 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
19993 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
19994 format_warning_at_substring with just a shorter name instead of
19995 const function pointer.
19996 (fmtwarn_n): New function.
19997 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
19998 appropriate, get rid of all the fmtstr temporaries, move conditionals
19999 with G_() wrapped string literals directly into fmtwarn arguments,
20000 cast dir.len to (int), formatting fixes.
20002 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
20004 * doc/invoke.texi: Remove "Cilk Plus" references.
20006 2018-03-02 Jakub Jelinek <jakub@redhat.com>
20007 Richard Biener <rguenther@suse.de>
20010 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
20011 for error or warning attributes if CALL_FROM_THUNK_P is set.
20014 2018-03-02 Jakub Jelinek <jakub@redhat.com>
20017 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
20018 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
20021 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
20022 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
20024 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
20025 instead of -1U in last predictors element's probability member.
20027 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
20030 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
20031 arguments if they are comparable.
20033 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
20035 PR tree-optimization/84634
20036 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
20037 masks and masked_loop_p with a single loop_masks, making sure it's
20038 null for bb vectorization.
20040 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
20042 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
20043 (vect_analyze_data_ref_access): Use loop->safe_len rather than
20044 loop->force_vectorize to check whether there is no alias.
20046 2018-03-02 Jakub Jelinek <jakub@redhat.com>
20049 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
20051 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
20053 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
20054 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
20055 instead of a loop around prev_real_insn.
20056 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
20059 PR inline-asm/84625
20060 * config/i386/i386.c (ix86_print_operand): Use conditional
20061 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
20064 2018-03-02 Richard Biener <rguenther@suse.de>
20066 PR tree-optimization/84427
20067 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
20068 (bitmap_set_subtract_values): Rewrite to handle multiple
20071 (prune_clobbered_mems): Likewise.
20072 (phi_translate): Take edge instead of pred/phiblock.
20073 (phi_translate_1): Likewise.
20074 (phi_translate_set): Likewise. Insert all translated
20075 exprs for a value into the set, keeping possibly multiple
20076 expressions per value.
20077 (compute_antic_aux): Adjust for phi_translate changes.
20078 When intersecting union the expressions and prune those
20079 not in the final value set, keeping possibly multiple
20080 expressions per value. Do not use value-insertion
20081 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
20082 all expressions. Add verification that the value-sets
20083 only shrink during iteration.
20084 (compute_partial_antic_aux): Adjust for the phi_translate changes.
20085 (do_pre_regular_insertion): Likewise.
20086 (do_pre_partial_partial_insertion): Likewise.
20088 2018-03-02 Richard Biener <rguenther@suse.de>
20091 * config/darwin.c (saved_debug_info_level): New static global.
20092 (darwin_asm_lto_start): Disable debug info generation for LTO out.
20093 (darwin_asm_lto_end): Restore debug info generation settings.
20095 2018-03-01 Martin Liska <mliska@suse.cz>
20098 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
20099 volatile arguments.
20101 2018-03-01 Richard Biener <rguenther@suse.de>
20104 * dwarf2out.c (gen_variable_die): Properly handle late VLA
20105 type annotation with LTO when debug was disabled at compile-time.
20107 2018-03-01 Matthew Fortune <mfortune@gmail.com>
20109 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
20111 (mips_final_postscan_insn): Likewise.
20113 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
20115 PR rtl-optimization/84528
20116 * alias.c (init_alias_target): Add commentary.
20117 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
20118 a unique base value if the frame pointer is not eliminated
20119 to the stack pointer.
20121 2018-03-01 Tom de Vries <tom@codesourcery.com>
20123 PR rtl-optimization/83327
20124 * lra-int.h (hard_regs_spilled_into): Declare.
20125 * lra.c (hard_regs_spilled_into): Define.
20126 (init_reg_info): Init hard_regs_spilled_into.
20127 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
20128 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
20129 (process_bb_lives): Handle hard_regs_spilled_into.
20130 (lra_create_live_ranges_1): Before doing liveness propagation, clear
20131 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
20133 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
20135 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
20136 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
20137 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
20138 * config/rs6000/aix72.h: New file.
20140 2018-02-28 Jakub Jelinek <jakub@redhat.com>
20142 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
20143 instead of warning_at with conditional singular and plural messages
20147 * stor-layout.c (update_alignment_for_field): For
20148 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
20149 && !DECL_PACKED (field), do the alignment update, just use
20150 only desired_align instead of MAX (type_align, desired_align)
20152 (place_field): Don't do known_align < desired_align handling
20153 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
20154 is non-NULL, instead do it after rli->prev_field handling and
20155 only if not within a bitfield word. For DECL_PACKED (field)
20156 use type_align of BITS_PER_UNIT.
20158 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
20160 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
20161 superfluous parentheses and trailing spaces.
20163 2018-02-28 Richard Biener <rguenther@suse.de>
20165 PR tree-optimization/84584
20166 * graphite-scop-detection.c (scop_detection::add_scop): Discard
20167 SCoPs with fake exit edge.
20169 2018-02-28 Martin Liska <mliska@suse.cz>
20172 * timevar.c (timer::print): Fix format to properly print 100%
20175 2018-02-28 Richard Biener <rguenther@suse.de>
20177 PR middle-end/84607
20178 * genmatch.c (capture_info::walk_match): Do not mark
20179 captured expressions without operands as expr_p given
20180 they act more like predicates and should be subject to
20181 "lost tail" side-effect preserving.
20183 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
20185 PR rtl-optimization/81611
20186 * auto-inc-dec.c (attempt_change): Move dead note from
20187 mem_insn if it's the next use of regno
20188 (find_address): Take address use of reg holding
20189 non-incremented value. Add parm to limit search to the named
20191 (merge_in_block): Attempt to use a mem insn that is the next
20192 use of the original regno.
20194 2018-02-27 Martin Sebor <msebor@redhat.com>
20197 * doc/invoke.texi (-Wmissing-attributes): New option.
20198 * print-tree.c (print_node): Handle DECL_UNINLINABLE.
20200 2018-02-27 Martin Sebor <msebor@redhat.com>
20202 PR translation/84207
20203 * diagnostic-core.h (warning_n, error_n, inform_n): Change
20204 n argument to unsigned HOST_WIDE_INT.
20205 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
20206 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
20207 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
20208 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
20210 2018-02-27 Richard Biener <rguenther@suse.de>
20212 PR tree-optimization/84512
20213 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
20214 Do not use the estimate returned from record_stmt_cost for
20215 the scalar iteration cost but sum properly using add_stmt_cost.
20217 2018-02-27 Richard Biener <rguenther@suse.de>
20219 PR tree-optimization/84466
20220 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
20221 Adjust last change to less strictly validate use operands.
20223 2018-02-27 Martin Liska <mliska@suse.cz>
20225 PR gcov-profile/84548
20226 * gcov.c (process_file): Allow partial overlap and consider it
20227 also as group functions.
20228 (output_lines): Properly calculate range of lines for a group.
20230 2018-02-27 Martin Liska <mliska@suse.cz>
20232 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
20233 'ggc' suffixes. Change first column width.
20234 (timer::print): Fix formatting of the column.
20236 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
20238 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
20239 preserve inline entry blocks for the sake of debug inline
20240 entry point markers alone.
20241 (remove_unused_locals): Suggest in comments a better place to
20242 force the preservation of inline entry blocks that are
20243 otherwise unused, but do not preserve them.
20245 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
20247 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
20249 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
20252 * config/i386/constraints.md (Bs): Replace
20253 ix86_indirect_branch_register with
20254 TARGET_INDIRECT_BRANCH_REGISTER.
20256 * config/i386/i386.md (indirect_jump): Likewise.
20257 (tablejump): Likewise.
20258 (*sibcall_memory): Likewise.
20259 (*sibcall_value_memory): Likewise.
20260 Peepholes of indirect call and jump via memory: Likewise.
20261 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
20262 (*sibcall_value_GOT_32): Likewise.
20263 * config/i386/predicates.md (indirect_branch_operand): Likewise.
20264 (GOT_memory_operand): Likewise.
20265 (call_insn_operand): Likewise.
20266 (sibcall_insn_operand): Likewise.
20267 (GOT32_symbol_operand): Likewise.
20268 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
20270 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
20272 PR rtl-optimization/83496
20273 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
20274 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
20275 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
20276 redundant insn, if any.
20277 (relax_delay_slots): Likewise.
20278 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
20280 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
20282 PR tree-optimization/83965
20283 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
20284 that grouped statements are part of a reduction chain. Return
20285 true if the statement is not marked as a reduction itself but
20286 is part of a group.
20287 (vect_recog_dot_prod_pattern): Don't check whether the statement
20288 is part of a group here.
20289 (vect_recog_sad_pattern): Likewise.
20290 (vect_recog_widen_sum_pattern): Likewise.
20292 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
20295 * final.c (rest_of_clean_state): Also look for calls inside sequences.
20297 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
20300 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
20302 (ix86_output_indirect_function_return): New prototype.
20303 (ix86_split_simple_return_pop_internal): Likewise.
20304 * config/i386/i386.c (indirect_return_via_cx): New.
20305 (indirect_return_via_cx_bnd): Likewise.
20306 (indirect_thunk_name): Handle return va CX_REG.
20307 (output_indirect_thunk_function): Create alias for
20308 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
20309 (ix86_output_indirect_jmp): Remove the bool argument.
20310 (ix86_output_indirect_function_return): New function.
20311 (ix86_split_simple_return_pop_internal): Likewise.
20312 * config/i386/i386.md (*indirect_jump): Don't pass false
20313 to ix86_output_indirect_jmp.
20314 (*tablejump_1): Likewise.
20315 (simple_return_pop_internal): Change it to define_insn_and_split.
20316 Call ix86_split_simple_return_pop_internal to split it for
20317 -mfunction-return=.
20318 (simple_return_indirect_internal): Call
20319 ix86_output_indirect_function_return instead of
20320 ix86_output_indirect_jmp.
20322 2018-02-26 Jakub Jelinek <jakub@redhat.com>
20325 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
20326 memset and value initialization afterwards.
20328 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
20330 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
20332 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
20335 * common/config/aarch64/aarch64-common.c
20336 (aarch_option_optimization_table[]): Switch
20337 off fomit-frame-pointer
20339 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
20340 Chung-Ju Wu <jasonwucj@gmail.com>
20342 * config/nds32/nds32-multiple.md (load_multiple): Disallow
20344 (store_multiple): Ditto.
20346 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
20348 * config.gcc: Add --with-cpu support for nds32 target.
20349 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
20350 * config/nds32/nds32.opt: Add -mcpu= option.
20352 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
20354 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
20355 isel=yes): Warn for these deprecated options.
20357 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
20359 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
20360 ISA_2_5_MASKS_EMBEDDED.
20362 2018-02-23 Jakub Jelinek <jakub@redhat.com>
20364 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
20365 p->max as pointers rather than using iterative_hash_expr.
20367 2018-02-23 Carl Love <cel@us.ibm.com>
20369 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
20370 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
20372 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
20373 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
20374 P8V_BUILTIN_VEC_VUNSIGNED2.
20376 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
20379 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
20380 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
20382 * lra-constraints.c (curr_insn_transform): Set up
20383 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
20385 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
20386 * lra-eliminations.c (spill_pseudos): Ditto.
20387 (process_insn_for_elimination): Ditto.
20388 * lra-lives.c (reg_early_clobber_p): Use the new macros.
20389 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
20390 LRA_NON_CLOBBERED_ALT.
20392 2018-02-22 Martin Sebor <msebor@redhat.com>
20394 PR tree-optimization/84480
20395 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
20396 to maybe_diag_stxncpy_trunc. Call it.
20397 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
20398 from gimple_fold_builtin_strcpy. Print inlining stack.
20399 (handle_builtin_stxncpy): Print inlining stack.
20400 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
20402 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
20405 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
20406 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
20407 and -fcheck-pointer-bounds are used together.
20408 (indirect_thunk_prefix): New enum.
20409 (indirect_thunk_need_prefix): New function.
20410 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
20411 "_nt" instead of "_bnd" for NOTRACK prefix.
20412 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
20413 (output_indirect_thunk_function): Likewise.
20415 (ix86_code_end): Update output_indirect_thunk_function calls.
20416 (ix86_output_indirect_branch_via_reg): Replace
20417 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
20418 (ix86_output_indirect_branch_via_push): Likewise.
20419 (ix86_output_function_return): Likewise.
20420 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
20421 incompatible with -fcf-protection=branch and
20422 -fcheck-pointer-bounds.
20424 2018-02-22 Steve Ellcey <sellcey@cavium.com>
20427 * config/aarch64/aarch64.c (aarch64_print_address_internal):
20428 Change gcc_assert call to output_operand_lossage.
20430 2018-02-22 Steve Ellcey <sellcey@cavium.com>
20432 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
20434 2018-02-22 DJ Delorie <dj@redhat.com>
20435 Sebastian Perta <sebastian.perta@renesas.com>
20436 Oleg Endo <olegendo@gcc.gnu.org>
20438 * config/rx/rx.c (rx_rtx_costs): New function.
20439 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
20441 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
20443 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
20445 2018-02-22 Martin Liska <mliska@suse.cz>
20448 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
20449 Add "native" as a possible value.
20451 2018-02-22 Martin Liska <mliska@suse.cz>
20454 * config/i386/i386.c (ix86_option_override_internal):
20455 Add "native" as a possible value for -march and -mtune.
20457 2018-02-22 Jakub Jelinek <jakub@redhat.com>
20460 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
20461 to all type variants.
20463 PR tree-optimization/84503
20464 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
20465 width as info->bitpos + info->bitsize - start.
20466 (merged_store_group::merge_overlapping): Simplify width computation.
20467 (check_no_overlap): New function.
20468 (imm_store_chain_info::try_coalesce_bswap): Compute expected
20469 start + width and last_order of the group, fail if check_no_overlap
20471 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
20472 to group if check_no_overlap fails.
20474 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
20476 * config/rs6000/altivec.md: Delete contraint arguments to
20477 define_expand, define_split, and define_peephole2, and in
20478 define_insn_and_split if always unused.
20479 * config/rs6000/darwin.md: Ditto.
20480 * config/rs6000/dfp.md: Ditto.
20481 * config/rs6000/rs6000.md: Ditto.
20482 * config/rs6000/sync.md: Ditto.
20483 * config/rs6000/vector.md: Ditto.
20484 * config/rs6000/vsx.md: Ditto.
20486 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
20488 * config/rs6000/altivec.md: Write output control strings as braced
20489 blocks instead of double-quoted strings.
20490 * config/rs6000/darwin.md: Ditto.
20491 * config/rs6000/rs6000.md: Ditto.
20492 * config/rs6000/vector.md: Ditto.
20493 * config/rs6000/vsx.md: Ditto.
20495 2018-02-21 Jason Merrill <jason@redhat.com>
20497 PR c++/84314 - ICE with templates and fastcall attribute.
20498 * attribs.c (build_type_attribute_qual_variant): Remove assert.
20500 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
20502 * ipa-cp.c (determine_versionability): Fix comment typos.
20504 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
20507 * ipa-cp.c (determine_versionability): Do not version functions caling
20510 2018-02-21 Martin Liska <mliska@suse.cz>
20513 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
20514 Add "native" as a possible value.
20515 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
20516 the macro when native cpu detection is available.
20518 2018-02-21 Martin Liska <mliska@suse.cz>
20521 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
20522 Add "native" as a possible value.
20523 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
20524 when native cpu detection is available.
20526 2018-02-21 Jakub Jelinek <jakub@redhat.com>
20527 Martin Sebor <msebor@redhat.com>
20529 PR tree-optimization/84478
20530 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
20532 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
20533 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
20534 support which is conservatively correct, for 2 only stay conservative
20535 for maxlen. Formatting and comment capitalization fixes. Add STRICT
20536 argument to the 2 argument get_range_strlen, adjust 6 arg
20537 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
20539 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
20540 (gimple_fold_builtin_strlen): Pass true as last argument to
20543 2018-02-20 Martin Sebor <msebor@redhat.com>
20545 PR middle-end/84095
20546 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
20547 (builtin_memref::set_base_and_offset): Same. Handle inner references.
20548 (builtin_memref::builtin_memref): Factor out parts into
20549 set_base_and_offset and call it.
20551 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
20553 PR middle-end/84406
20554 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
20555 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
20556 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
20557 search at the associated MODE_INT.
20559 2018-02-20 Jeff Law <law@redhat.com>
20561 PR middle-end/82123
20562 PR tree-optimization/81592
20563 PR middle-end/79257
20564 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
20565 for range data rather than using global data.
20566 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
20567 range data rather than using global data.
20568 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
20569 pass it to children as needed.
20570 (struct directive::fmtresult): Similarly.
20571 (struct directive::set_width): Similarly.
20572 (struct directive::set_precision): Similarly.
20573 (format_integer, format_directive, parse_directive): Similarly.
20574 (format_none): Accept unnamed vr_values parameter.
20575 (format_percent, format_floating, format_character): Similarly.
20576 (format_string, format_plain): Similarly.
20577 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
20578 the EVRP range analyzer for range data rather than using global data.
20579 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
20580 gimple-ssa-evrp-analyze.h
20581 (class sprintf_dom_walker): Add after_dom_children member function.
20582 Add evrp_range_analyzer member.
20583 (sprintf_dom_walker::before_dom_children): Call into the EVRP
20584 range analyzer as needed.
20585 (sprintf_dom_walker::after_dom_children): New member function.
20586 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
20588 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
20589 (evrp_range_analyzer::pop_to_marker): Likewise.
20591 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
20593 PR tree-optimization/84419
20594 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
20595 with the required type if its current type is compatible but
20598 2018-02-20 Jakub Jelinek <jakub@redhat.com>
20600 PR middle-end/82004
20601 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
20602 after vectorization.
20604 2018-02-20 Martin Liska <mliska@suse.cz>
20607 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
20608 possible values if we don't have a hint.
20610 2018-02-20 Martin Liska <mliska@suse.cz>
20614 * final.c (shorten_branches): Build align_tab array with one
20616 * opts.c (finish_options): Add alignment option limit check.
20617 (MAX_CODE_ALIGN): Likewise.
20618 (MAX_CODE_ALIGN_VALUE): Likewise.
20619 * doc/invoke.texi: Document maximum allowed option value for
20620 all -falign-* options.
20622 2018-02-19 Jakub Jelinek <jakub@redhat.com>
20625 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
20626 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
20627 * var-tracking.c (emit_note_insn_var_location): Remove all references
20628 to NOTE_INSN_CALL_ARG_LOCATION.
20629 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
20630 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
20631 Use copy_rtx_if_shared.
20632 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
20633 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
20634 (dwarf2out_var_location): Remove handling of
20635 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
20637 * final.c (final_scan_insn): Remove all references to
20638 NOTE_INSN_CALL_ARG_LOCATION.
20639 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
20640 before dumping final insns.
20641 * except.c (emit_note_eh_region_end): Remove all references to
20642 NOTE_INSN_CALL_ARG_LOCATION.
20643 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
20644 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
20645 * config/arc/arc.c (hwloop_optimize): Likewise.
20646 * config/arm/arm.c (create_fix_barrier): Likewise.
20647 * config/s390/s390.c (s390_chunkify_start): Likewise.
20648 * config/sh/sh.c (find_barrier): Likewise.
20649 * config/i386/i386.c (rest_of_insert_endbranch,
20650 ix86_seh_fixup_eh_fallthru): Likewise.
20651 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
20652 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
20653 * config/frv/frv.c (frv_function_prologue): Likewise.
20654 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
20656 (note_outside_basic_block_p): Remove all references to
20657 NOTE_INSN_CALL_ARG_LOCATION.
20658 * gengtype.c (adjust_field_rtx_def): Likewise.
20659 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
20661 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
20662 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
20665 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
20668 PR tree-optimization/84452
20669 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
20670 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
20673 2018-02-19 Martin Liska <mliska@suse.cz>
20676 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
20678 2018-02-19 Martin Liska <mliska@suse.cz>
20679 Richard Sandiford <richard.sandiford@linaro.org>
20681 PR tree-optimization/82491
20682 * gimple-fold.c (get_base_constructor): Make earlier bail out
20685 2018-02-19 Carl Love <cel@us.ibm.com>
20687 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
20688 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
20690 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
20691 P8V_BUILTIN_VEC_NEG.
20693 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
20695 * config/rl78/rl78.md (movdf): New define expand.
20697 2018-02-19 Martin Liska <mliska@suse.cz>
20700 * doc/invoke.texi: Fix typo.
20701 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
20703 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
20705 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
20706 handle rs6000_single_float and rs6000_double_float specially for
20709 2018-02-16 Jeff Law <law@redhat.com>
20711 * config/rx/rx.c (add_pop_cfi_notes): New function.;
20712 (pop_regs): Use it.
20714 2018-02-16 Jakub Jelinek <jakub@redhat.com>
20717 * ipa-inline.c (inline_small_functions): Fix a typo.
20719 2018-02-16 Nathan Sidwell <nathan@acm.org>
20721 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
20723 2018-02-16 Carl Love <cel@us.ibm.com>
20725 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
20726 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
20727 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
20728 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
20729 expansion to P8V_BUILTIN_VEC_FLOAT2.
20731 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
20733 PR rtl-optimization/70023
20734 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
20735 src_regno into account.
20737 2018-02-16 Carl Love <cel@us.ibm.com>
20739 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
20740 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
20741 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
20742 * config/rs6000/rs6000.c: Remove case statements for
20743 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
20744 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
20745 and P9V_BUILTIN_VEC_VINSERT4B.
20746 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
20747 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
20748 * config/rs6000/vsx.md:
20749 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
20752 2018-02-16 Carl Love <cel@us.ibm.com>
20754 * config/rs6000/altivec.h: Add builtin names vec_extract4b
20756 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
20758 * config/rs6000/rs6000-c.c: Add the definitions for
20759 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
20760 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
20761 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
20762 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
20763 definition for insert4b and define insn *insert3b_internal.
20764 * doc/extend.texi: Add documentation for vec_extract4b.
20766 2018-02-16 Nathan Sidwell <nathan@acm.org>
20768 * doc/extend.texi (Backwards Compatibility): Mention friend
20769 injection. Note for-scope is deprecated.
20770 * doc/invoke.texi (-ffriend-injection): Deprecate.
20772 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
20774 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
20775 that moved to I2, also allow destinations that are a paradoxical
20776 subreg (instead of a normal reg).
20778 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
20781 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
20784 2018-02-16 Richard Biener <rguenther@suse.de>
20786 PR tree-optimization/84037
20787 PR tree-optimization/84016
20789 * config/i386/i386.c (ix86_builtin_vectorization_cost):
20790 Adjust vec_construct for the fact we need additional higher latency
20791 128bit inserts for AVX256 and AVX512 vector builds.
20792 (ix86_add_stmt_cost): Scale vector construction cost for
20795 2018-02-16 Richard Biener <rguenther@suse.de>
20797 PR tree-optimization/84417
20798 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
20799 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
20800 (non_rewritable_lvalue_p): Likewise, use poly-ints.
20802 2018-02-16 Martin Liska <mliska@suse.cz>
20805 * internal-fn.def (ASAN_CHECK): Set proper flags.
20806 (ASAN_MARK): Likewise.
20808 2018-02-16 Julia Koval <julia.koval@intel.com>
20810 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
20811 from PTA_CANNONLAKE.
20813 2018-02-16 Jakub Jelinek <jakub@redhat.com>
20816 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
20817 Use ++iter rather than iter++ for std::list iterators.
20818 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
20819 defer deleting them until all nodes in the forest are processed. Do
20820 free even leaf nodes. Change to_process into auto_vec.
20823 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
20824 * vec.h (vec_default_construct): Use memset instead of placement new
20825 if BROKEN_VALUE_INITIALIZATION is defined.
20826 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
20827 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
20830 PR rtl-optimization/83723
20831 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
20832 * lra.c (lra_substitute_pseudo): Likewise. If true, use
20833 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
20835 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
20837 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
20839 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
20841 PR rtl-optimization/81443
20842 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
20843 from inner REGs to paradoxical SUBREGs.
20845 2018-02-16 Richard Biener <rguenther@suse.de>
20847 PR tree-optimization/84399
20848 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
20849 For operands we can analyze at their definition make sure we can
20850 analyze them at each use as well.
20852 2018-02-16 Richard Biener <rguenther@suse.de>
20854 PR tree-optimization/84190
20855 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
20856 volatile accesses if the decl isn't volatile.
20858 2018-02-15 Jason Merrill <jason@redhat.com>
20860 PR c++/84314 - ICE with templates and fastcall attribute.
20861 * attribs.c (build_type_attribute_qual_variant): Don't clobber
20862 TYPE_CANONICAL on an existing type.
20864 2018-02-15 Jakub Jelinek <jakub@redhat.com>
20866 PR tree-optimization/84383
20867 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
20868 dstoff nor call operand_equal_p if dstbase is NULL.
20870 PR tree-optimization/84334
20871 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
20872 also a CONSTANT_CLASS_P, punt.
20874 2018-02-14 Jim Wilson <jimw@sifive.com>
20876 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
20877 first SMALL_OPERAND check. New local min_second_step. Move assert
20878 to where locals are set. Add TARGET_RVC support.
20879 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
20881 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
20883 * doc/invoke.texi: Correct -Wformat-overflow code sample.
20885 2018-02-14 Martin Sebor <msebor@redhat.com>
20887 PR tree-optimization/83698
20888 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
20889 arrays constrain the offset range to their bounds.
20890 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
20891 (builtin_access::overlap): Avoid setting the size of overlap if it's
20893 (maybe_diag_overlap): Also consider arrays when deciding what values
20894 of offsets to include in diagnostics.
20896 2018-02-14 Martin Sebor <msebor@redhat.com>
20899 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
20900 that correspond to the kind of a declaration.
20902 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
20905 * config/pa/pa.md: Load address of PIC label using the linkage table
20906 if the label is nonlocal.
20908 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
20910 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
20911 warning message if user requests -maltivec=be.
20912 * doc/invoke.texi: Document deprecation of -maltivec=be.
20914 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
20917 * config/rs6000/rs6000-c.c: Update definitions for
20918 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
20919 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
20921 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
20924 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
20925 add _get_ssp intrinsics. Remove argument from
20926 __builtin_ia32_rdssp[d|q].
20927 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
20928 * config/i386/i386-builtin.def: Remove argument from
20929 __builtin_ia32_rdssp[d|q].
20930 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
20931 ix86_expand_special_args_builtin for _rdssp[d|q].
20932 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
20933 Clear register before usage.
20934 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
20935 Add documentation for new _get_ssp and _inc_ssp intrinsics.
20937 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
20939 PR tree-optimization/84357
20940 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
20941 operand 1 of an ARRAY_REF too.
20943 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
20946 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
20947 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
20949 (set_of_reg): New struct.
20950 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
20951 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
20952 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
20954 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
20955 Split into bitclr, bitset, bitinvert patterns if appropriate.
20956 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
20957 use rx_fuse_in_memory_bitop.
20958 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
20959 to named insn, correct maximum insn length.
20961 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
20964 * machmode.def: Define a complex mode for PARTIAL_INT.
20965 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
20967 * doc/rtl.texi: Document CSPImode.
20968 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
20970 (msp430_hard_regno_nregs_with_padding): Likewise.
20972 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
20975 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
20977 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
20979 PR rtl-optimization/84169
20980 * combine.c (try_combine): New variable split_i2i3. Set it to true if
20981 we generated a parallel as new i3 and we split that to new i2 and i3
20982 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
20983 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
20984 those to i2, not i1. Partially rewrite this scan code.
20986 2018-02-13 Jakub Jelinek <jakub@redhat.com>
20989 * stor-layout.c (place_field): For variable length fields, adjust
20990 offset_align afterwards not just based on the field's alignment,
20991 but also on the size.
20993 PR middle-end/84309
20994 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
20995 of exps and logs in the use_exp2 case.
20997 2018-02-13 Jeff Law <law@redhat.com>
20999 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
21000 entry for "vector".
21002 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
21005 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
21009 * common.opt (gas-loc-support, gas-locview-support): New.
21010 (ginline-points, ginternal-reset-location-views): New.
21011 * doc/invoke.texi: Document them. Use @itemx where intended.
21012 (gvariable-location-views): Adjust.
21013 * target.def (reset_location_view): New.
21014 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
21015 (TARGET_RESET_LOCATION_VIEW): New.
21016 * doc/tm.texi: Rebuilt.
21017 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
21018 (dwarf2out_default_as_locview_support): New.
21019 (output_asm_line_debug_info): Use option variables.
21020 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
21021 (output_loc_list): Likewise.
21022 (add_high_low_attributes): Check option variables.
21023 Don't output entry view attribute in strict mode.
21024 (gen_inlined_subroutine_die): Check option variables.
21025 (dwarf2out_inline_entry): Likewise.
21026 (init_sections_and_labels): Likewise.
21027 (dwarf2out_early_finish): Likewise.
21028 (maybe_reset_location_view): New, from...
21029 (dwarf2out_var_location): ... here. Call it.
21030 * debug.h (dwarf2out_default_as_loc_support): Declare.
21031 (dwarf2out_default_as_locview_support): Declare.
21032 * hooks.c (hook_int_rtx_insn_0): New.
21033 * hooks.h (hook_int_rtx_insn_0): Declare.
21034 * toplev.c (process_options): Take -gas-loc-support and
21035 -gas-locview-support from dwarf2out. Enable
21036 -gvariable-location-views by default only with locview
21037 assembler support. Enable -ginternal-reset-location-views by
21038 default only if the target defines the corresponding hook.
21039 Enable -ginline-points by default if location views are
21040 enabled; force it disabled if statement frontiers are
21042 * tree-inline.c (expand_call_inline): Check option variables.
21043 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
21045 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
21047 PR tree-optimization/84321
21048 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
21049 handling. Also check whether the anti-range contains any values
21050 that satisfy the mask; switch to a VR_RANGE if not.
21052 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
21055 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
21057 2018-02-13 Martin Jambor <mjambor@suse.cz>
21060 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
21061 of call statements, also set location of a load to a temporary.
21063 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
21065 * config/rl78/rl78.c (add_vector_labels): New function.
21066 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
21067 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
21068 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
21069 which checks that no arguments are passed.
21070 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
21071 * doc/extend.texi: Documentation for the new attribute.
21073 2018-02-13 Andreas Schwab <schwab@suse.de>
21075 * config/riscv/linux.h (CPP_SPEC): Define.
21077 2018-02-13 Jakub Jelinek <jakub@redhat.com>
21080 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
21081 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
21082 OPTION_MASK_ISA_AES as first argument to def_builtin_const
21083 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
21084 instead of OPTION_MASK_ISA_PCLMUL as first argument to
21085 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
21086 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
21087 temporarily for AES and PCLMUL builtins.
21089 PR tree-optimization/84339
21090 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
21091 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
21094 PR middle-end/84309
21095 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
21096 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
21097 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
21099 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
21101 * omp-simd-clone.h: New file.
21102 * omp-simd-clone.c: Include omp-simd-clone.h.
21103 (expand_simd_clones): No longer static.
21104 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
21105 cgraph.h and omp-simd-clone.h.
21106 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
21107 (vect_recog_widen_shift_pattern): Formatting fix.
21108 (vect_pattern_recog_1): Don't check optab for calls.
21111 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
21112 operands[2] into a REG before using gen_lowpart on it.
21114 2018-02-12 Jeff Law <law@redhat.com>
21117 * config/sh/sh.c (find_barrier): Consider a sibling call
21120 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
21121 successfully back substituting a reg.
21123 2018-02-12 Richard Biener <rguenther@suse.de>
21125 PR tree-optimization/84037
21126 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
21127 parameter, move visited init to caller.
21128 (vect_slp_analyze_operations): Separate cost from validity
21129 check, initialize visited once for all instances.
21130 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
21132 * tree-vect-stmts.c (vect_model_simple_cost): Make early
21134 (vect_model_promotion_demotion_cost): Likewise.
21135 (vectorizable_bswap): Guard cost modeling with !slp_node
21136 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
21138 (vectorizable_call): Likewise.
21139 (vectorizable_conversion): Likewise.
21140 (vectorizable_assignment): Likewise.
21141 (vectorizable_shift): Likewise.
21142 (vectorizable_operation): Likewise.
21143 (vectorizable_store): Likewise.
21144 (vectorizable_load): Likewise.
21145 (vectorizable_condition): Likewise.
21146 (vectorizable_comparison): Likewise.
21148 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
21151 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
21152 (ASAN_MARK): Fix fnspec to account for return value, change pointer
21153 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
21155 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
21157 PR middle-end/83665
21158 * params.def (inline-min-speedup): Increase from 8 to 15.
21159 (max-inline-insns-auto): Decrease from 40 to 30.
21160 * ipa-split.c (consider_split): Add some buffer for function to
21161 be considered inlining candidate.
21162 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
21165 2018-02-12 Richard Biener <rguenther@suse.de>
21167 PR tree-optimization/84037
21168 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
21169 matched stmts if we cannot swap the non-matched ones.
21171 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
21173 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
21174 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
21175 _mm_maskz_scalef_round_ss): New intrinsics.
21176 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
21177 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
21178 __builtin_ia32_scalefss_round): Remove.
21179 (__builtin_ia32_scalefsd_mask_round,
21180 __builtin_ia32_scalefss_mask_round): New intrinsics.
21181 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
21182 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
21183 ((match_operand:VF_128 2 "<round_nimm_predicate>"
21184 "<round_constraint>")): Changed to ...
21185 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
21186 "<round_scalar_constraint>")): ... this.
21187 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
21188 %0, %1, %2<round_op3>}"): Changed to ...
21189 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
21190 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
21191 %2<round_scalar_mask_op3>}"): ... this.
21192 * config/i386/subst.md (round_scalar_nimm_predicate): New.
21194 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
21196 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
21197 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
21198 (_mm_maskz_sqrt_round_ss): New intrinsics.
21199 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
21200 (__builtin_ia32_sqrtsd_mask_round)
21201 (__builtin_ia32_sqrtss_mask_round): New builtins.
21202 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
21203 (__builtin_ia32_sqrtss_round): Remove.
21204 (__builtin_ia32_sqrtsd_mask_round)
21205 (__builtin_ia32_sqrtss_mask_round): New builtins.
21206 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
21207 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
21208 ((match_operand:VF_128 1 "vector_operand"
21209 "xBm,<round_constraint>")): Changed to ...
21210 ((match_operand:VF_128 1 "vector_operand"
21211 "xBm,<round_scalar_constraint>")): ... this.
21212 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
21213 %0, %2, %<iptr>1<round_op3>}): Changed to ...
21214 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
21215 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
21216 %<iptr>1<round_scalar_mask_op3>}): ... this.
21217 ((set_attr "prefix" "<round_prefix>")): Changed to ...
21218 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
21220 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
21223 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
21224 Cast vec_cmpeq result to correct type.
21225 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
21226 Cast vec_cmpgt result to correct type.
21228 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
21230 * final.c (final_scan_insn_1): Renamed from...
21231 (final_scan_insn): ... this. New wrapper, to recover
21232 seen from the outermost call in recursive ones.
21233 * config/sparc/sparc.c (output_return): Drop seen from call.
21234 (output_sibcall): Likewise.
21235 * config/visium/visium.c (output_branch): Likewise.
21237 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
21239 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
21242 2018-02-10 Alan Modra <amodra@gmail.com>
21245 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
21246 Specify LR as an input.
21248 2018-02-10 Jakub Jelinek <jakub@redhat.com>
21251 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
21252 remove_member_access_dummy_vars): New functions.
21253 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
21254 lower_omp_1, execute_lower_omp): Use them.
21256 PR rtl-optimization/84308
21257 * shrink-wrap.c (spread_components): Release todo vector.
21259 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
21261 PR rtl-optimization/57193
21262 * ira-color.c (struct allocno_color_data): Add member
21263 conflict_allocno_hard_prefs.
21264 (update_conflict_allocno_hard_prefs): New.
21265 (bucket_allocno_compare_func): Add a preference based on
21266 conflict_allocno_hard_prefs.
21267 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
21268 (color_allocnos): Remove a dead code. Initiate
21269 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
21271 2018-02-09 Jakub Jelinek <jakub@redhat.com>
21274 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
21275 constraint from =wa to wa. Avoid a subreg on the output operand,
21276 instead use a pseudo and subreg it in a move.
21277 (p9_xxbrd_<mode>): Changed to ...
21278 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
21279 (p9_xxbrd_v2df): New expander.
21280 (p9_xxbrw_<mode>): Changed to ...
21281 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
21282 (p9_xxbrw_v4sf): New expander.
21284 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
21286 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
21288 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
21291 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
21292 multiply in 32-bit mode.
21293 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
21294 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
21297 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
21299 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
21300 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
21301 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
21302 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
21304 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
21307 * dwarf2out.c (is_trivial_indirect_ref): New function.
21308 (dwarf2out_late_global_decl): Do not generate a location
21309 attribute for variables that have a non-trivial DECL_VALUE_EXPR
21310 and that are not defined in the current unit.
21312 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
21314 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
21315 instead of a libcall for UNORDERED.
21317 2018-02-09 Tamar Christina <tamar.christina@arm.com>
21320 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
21321 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
21323 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
21326 * config/s390/s390.c (s390_set_current_function): Invoke
21327 s390_indirect_branch_settings also if fndecl didn't change.
21329 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
21331 * config/rs6000/rs6000.md (blockage): Set length to zero.
21333 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
21335 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
21337 2018-02-09 Jakub Jelinek <jakub@redhat.com>
21340 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
21341 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
21345 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
21346 PARALLEL incoming that failed vt_get_decl_and_offset check.
21348 PR middle-end/84237
21349 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
21350 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
21352 (get_variable_section): For decls in named .bss* sections pass true as
21353 second argument to bss_initializer_p.
21355 2018-02-09 Marek Polacek <polacek@redhat.com>
21356 Jakub Jelinek <jakub@redhat.com>
21359 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
21360 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
21361 Sync some changes from cxx_fold_indirect_ref.
21363 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
21365 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
21367 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
21368 (BLOCK_INLINE_ENTRY_LABEL): New.
21369 (dwarf2out_var_location): Disregard inline entry markers.
21370 (inline_entry_data): New struct.
21371 (inline_entry_data_hasher): New hashtable type.
21372 (inline_entry_data_hasher::hash): New.
21373 (inline_entry_data_hasher::equal): New.
21374 (inline_entry_data_table): New variable.
21375 (add_high_low_attributes): Add DW_AT_entry_pc and
21376 DW_AT_GNU_entry_view attributes if a pending entry is found
21377 in inline_entry_data_table. Add old entry_pc attribute only
21378 if debug nonbinding markers are disabled.
21379 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
21380 markers are enabled.
21381 (block_within_block_p, dwarf2out_inline_entry): New.
21382 (dwarf2out_finish): Check that no entries remained in
21383 inline_entry_data_table.
21384 * final.c (reemit_insn_block_notes): Handle inline entry notes.
21385 (final_scan_insn, notice_source_line): Likewise.
21386 (rest_of_clean_state): Skip inline entry markers.
21387 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
21389 * gimple.c (gimple_build_debug_inline_entry): New.
21390 * gimple.h (enum gimple_debug_subcode): Add
21391 GIMPLE_DEBUG_INLINE_ENTRY.
21392 (gimple_build_debug_inline_entry): Declare.
21393 (gimple_debug_inline_entry_p): New.
21394 (gimple_debug_nonbind_marker_p): Adjust.
21395 * insn-notes.def (INLINE_ENTRY): New.
21396 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
21397 inline entry marker notes.
21398 (print_insn): Likewise.
21399 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
21400 (INSN_DEBUG_MARKER_KIND): Likewise.
21401 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
21402 * tree-inline.c (expand_call_inline): Build and insert
21403 debug_inline_entry stmt.
21404 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
21405 inline entry blocks early, if nonbind markers are enabled.
21406 (dump_scope_block): Dump fragment info.
21407 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
21408 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
21409 (gimple_build_debug_inline_entry): New.
21410 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
21411 Enable/disable inline entry points too.
21412 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
21413 (DEBUG_INSN): Describe inline entry markers.
21415 * common.opt (gvariable-location-views): New.
21416 (gvariable-location-views=incompat5): New.
21417 * config.in: Rebuilt.
21418 * configure: Rebuilt.
21419 * configure.ac: Test assembler for view support.
21420 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
21421 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
21422 * dwarf2out.c (var_loc_view): New typedef.
21423 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
21424 (dwarf2out_locviews_in_attribute): New.
21425 (dwarf2out_locviews_in_loclist): New.
21426 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
21427 (enum dw_line_info_opcode): Add LI_adv_address.
21428 (struct dw_line_info_table): Add view.
21429 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
21430 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
21431 (zero_view_p): New variable.
21432 (ZERO_VIEW_P): New macro.
21433 (output_asm_line_debug_info): New.
21434 (struct var_loc_node): Add view.
21435 (add_AT_view_list, AT_loc_list): New.
21436 (add_var_loc_to_decl): Add view param. Test it against last.
21437 (new_loc_list): Add view params. Record them.
21438 (AT_loc_list_ptr): Handle loc and view lists.
21439 (view_list_to_loc_list_val_node): New.
21440 (print_dw_val): Handle dw_val_class_view_list.
21441 (size_of_die): Likewise.
21442 (value_format): Likewise.
21443 (loc_list_has_views): New.
21444 (gen_llsym): Set vl_symbol too.
21445 (maybe_gen_llsym, skip_loc_list_entry): New.
21446 (dwarf2out_maybe_output_loclist_view_pair): New.
21447 (output_loc_list): Output view list or entries too.
21448 (output_view_list_offset): New.
21449 (output_die): Handle dw_val_class_view_list.
21450 (output_dwarf_version): New.
21451 (output_compilation_unit_header): Use it.
21452 (output_skeleton_debug_sections): Likewise.
21453 (output_rnglists, output_line_info): Likewise.
21454 (output_pubnames, output_aranges): Update version comments.
21455 (output_one_line_info_table): Output view numbers in asm comments.
21456 (dw_loc_list): Determine current endview, pass it to new_loc_list.
21457 Call maybe_gen_llsym.
21458 (loc_list_from_tree_1): Adjust.
21459 (add_AT_location_description): Create view list attribute if
21460 needed, check it's absent otherwise.
21461 (convert_cfa_to_fb_loc_list): Adjust.
21462 (maybe_emit_file): Call output_asm_line_debug_info for test.
21463 (dwarf2out_var_location): Reset views as needed. Precompute
21464 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
21465 attribute. Set view.
21466 (new_line_info_table): Reset next view.
21467 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
21468 (dwarf2out_source_line): Likewise. Output view resets and labels to
21469 the assembler, or select appropriate line info opcodes.
21470 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
21471 (optimize_string_length): Catch it. Adjust.
21472 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
21473 dw_val_class_view_list, and remove it if no longer needed.
21474 (hash_loc_list): Hash view numbers.
21475 (loc_list_hasher::equal): Compare them.
21476 (optimize_location_lists): Check whether a view list symbol is
21477 needed, and whether the locview attribute is present, and
21478 whether they match. Remove the locview attribute if no longer
21480 (index_location_lists): Call skip_loc_list_entry for test.
21481 (dwarf2out_finish): Call output_asm_line_debug_info for test.
21482 Use output_dwarf_version.
21483 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
21484 (struct dw_val_node): Add val_view_list.
21485 * final.c (SEEN_NEXT_VIEW): New.
21486 (set_next_view_needed): New.
21487 (clear_next_view_needed): New.
21488 (maybe_output_next_view): New.
21489 (final_start_function): Rename to...
21490 (final_start_function_1): ... this. Take pointer to FIRST,
21491 add SEEN parameter. Emit param bindings in the initial view.
21492 (final_start_function): Reintroduce SEEN-less interface.
21493 (final): Rename to...
21494 (final_1): ... this. Take SEEN parameter. Output final pending
21495 next view at the end.
21496 (final): Reintroduce seen-less interface.
21497 (final_scan_insn): Output pending next view before switching
21498 sections or ending a block. Mark the next view as needed when
21499 outputting variable locations. Notify debug backend of section
21500 changes, and of location view changes.
21501 (rest_of_handle_final): Adjust.
21502 * toplev.c (process_options): Autodetect value for debug variable
21503 location views option. Warn on incompat5 without -gdwarf-5.
21504 * doc/invoke.texi (gvariable-location-views): New.
21505 (gvariable-location-views=incompat5): New.
21506 (gno-variable-location-views): New.
21508 2018-02-08 David Malcolm <dmalcolm@redhat.com>
21510 PR tree-optimization/84136
21511 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
21512 that the result of find_edge is non-NULL.
21514 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
21517 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
21518 storing integer register in SImode. Fix cost of 256 and 512
21519 byte aligned SSE register store.
21521 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
21523 * config/i386/i386.c (ix86_multiplication_cost): Fix
21524 multiplication cost for TARGET_AVX512DQ.
21526 2018-02-08 Marek Polacek <polacek@redhat.com>
21528 PR tree-optimization/84238
21529 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
21532 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
21534 PR tree-optimization/84265
21535 * tree-vect-stmts.c (vectorizable_store): Don't treat
21536 VMAT_CONTIGUOUS accesses as grouped.
21537 (vectorizable_load): Likewise.
21539 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
21541 PR tree-optimization/81635
21542 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
21543 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
21544 (test_round_for_mask): New functions.
21545 (wide_int_cc_tests): Call test_round_for_mask.
21546 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
21547 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
21548 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
21549 range returned by get_range_info.
21551 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
21554 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
21555 * symtab.c: Include builtins.h
21556 (symtab_node::output_to_lto_symbol_table_p): Move here
21557 from lto-streamer-out.c:output_symbol_p.
21558 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
21559 (output_symbol_p): Move all logic to symtab.c
21560 (produce_symtab): Update.
21562 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
21564 * config/s390/s390-opts.h (enum indirect_branch): Define.
21565 * config/s390/s390-protos.h (s390_return_addr_from_memory)
21566 (s390_indirect_branch_via_thunk)
21567 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
21568 (enum s390_indirect_branch_type): Define.
21569 * config/s390/s390.c (struct s390_frame_layout, struct
21570 machine_function): Remove.
21571 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
21572 (indirect_branch_table_label_no, indirect_branch_table_name):
21574 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
21575 (enum s390_indirect_branch_option): Define.
21576 (s390_return_addr_from_memory): New function.
21577 (s390_handle_string_attribute): New function.
21578 (s390_attribute_table): Add new attribute handler.
21579 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
21580 (s390_indirect_branch_via_thunk): New function.
21581 (s390_indirect_branch_via_inline_thunk): New function.
21582 (s390_function_ok_for_sibcall): When jumping via thunk disallow
21583 sibling call optimization for non z10 compiles.
21584 (s390_emit_call): Force indirect branch target to be a single
21585 register. Add r1 clobber for non-z10 compiles.
21586 (s390_emit_epilogue): Emit return jump via return_use expander.
21587 (s390_reorg): Handle JUMP_INSNs as execute targets.
21588 (s390_option_override_internal): Perform validity checks for the
21589 new command line options.
21590 (s390_indirect_branch_attrvalue): New function.
21591 (s390_indirect_branch_settings): New function.
21592 (s390_set_current_function): Invoke s390_indirect_branch_settings.
21593 (s390_output_indirect_thunk_function): New function.
21594 (s390_code_end): Implement target hook.
21595 (s390_case_values_threshold): Implement target hook.
21596 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
21598 * config/s390/s390.h (struct s390_frame_layout)
21599 (struct machine_function): Move here from s390.c.
21600 (TARGET_INDIRECT_BRANCH_NOBP_RET)
21601 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
21602 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
21603 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
21604 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
21605 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
21606 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
21607 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
21608 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
21609 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
21610 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
21611 (mnemonic attribute): Add values which aren't recognized
21613 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
21614 pattern for branch conversion. Fix mnemonic attribute.
21615 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
21616 indirect branch via thunk if requested.
21617 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
21618 ("*indirect_jump"): Disable for branch conversion using out of
21620 ("indirect_jump_via_thunk<mode>_z10")
21621 ("indirect_jump_via_thunk<mode>")
21622 ("indirect_jump_via_inlinethunk<mode>_z10")
21623 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
21624 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
21625 ("casesi_jump_via_inlinethunk<mode>_z10")
21626 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
21627 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
21628 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
21629 ("*indirect2_jump"): Disable for branch conversion.
21630 ("casesi_jump"): Turn into expander and expand patterns for branch
21632 ("return_use"): New expander.
21633 ("*return"): Emit return via thunk and rename it to ...
21634 ("*return<mode>"): ... this one.
21635 * config/s390/s390.opt: Add new options and and enum for the
21638 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
21640 * lra-constraints.c (match_reload): Unconditionally use
21641 gen_lowpart_SUBREG, rather than selecting between that
21642 and equivalent gen_rtx_SUBREG code.
21644 2018-02-08 Richard Biener <rguenther@suse.de>
21646 PR tree-optimization/84233
21647 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
21648 changed flag instead of boguously re-using phi_inserted.
21650 2018-02-08 Martin Jambor <mjambor@suse.cz>
21652 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
21653 static local variables.
21655 2018-02-08 Richard Biener <rguenther@suse.de>
21657 PR tree-optimization/84278
21658 * tree-vect-stmts.c (vectorizable_store): When looking for
21659 smaller vector types to perform grouped strided loads/stores
21660 make sure the mode is supported by the target.
21661 (vectorizable_load): Likewise.
21663 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
21665 * config/aarch64/aarch64.c (aarch64_components_for_bb):
21666 Increase LDP/STP opportunities by adding adjacent callee-saves.
21668 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
21670 PR rtl-optimization/84068
21671 PR rtl-optimization/83459
21672 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
21674 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
21676 PR tree-optimization/84224
21677 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
21678 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
21679 non-zero arguments.
21681 2018-02-07 Iain Sandoe <iain@codesourcery.com>
21684 * config/rs6000/altivec.md (*restore_world): Remove LR use.
21685 * config/rs6000/predicates.md (restore_world_operation): Adjust op
21686 count, remove one USE.
21688 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
21690 * doc/install.texi (Configuration): Document the
21691 --with-long-double-format={ibm,ieee} PowerPC configuration
21695 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
21696 Convert from define_expand to be define_insn_and_split. Rework
21697 float/double/_Float128 conversions to QI/HI/SImode to work with
21698 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
21699 conversions to QI/HImode types did a store and then a load to
21700 truncate the value. For conversions to VSX registers, don't split
21701 the insn, instead emit the code directly. Use the code iterator
21702 any_fix to combine signed and unsigned conversions.
21703 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
21704 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
21705 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
21706 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
21707 (fix_<mode>di2_hw): Likewise.
21708 (fixuns_<mode>di2_hw): Likewise.
21709 (fix_<mode>si2_hw): Likewise.
21710 (fixuns_<mode>si2_hw): Likewise.
21711 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
21712 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
21713 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
21714 fix<uns>_trunc<SFDF:mode>si2_p8.
21715 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer used.
21716 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
21717 (fix<uns>_<mode>_mem): Likewise.
21718 (fctiw<u>z_<mode>_mem): Likewise.
21719 (fix<uns>_<mode>_mem): Likewise.
21720 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
21721 the register allocator from doing a direct move to the GPRs to do
21722 a store, and instead use the ISA 3.0 store byte/half-word from
21723 vector register instruction. For IEEE 128-bit floating point,
21724 also optimize stores of 32-bit ints.
21725 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
21727 2018-02-07 Alan Hayward <alan.hayward@arm.com>
21729 * genextract.c (push_pathstr_operand): New function to support
21731 (walk_rtx): Call push_pathstr_operand.
21732 (print_path): Support [a-zA-Z].
21734 2018-02-07 Richard Biener <rguenther@suse.de>
21736 PR tree-optimization/84037
21737 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
21738 (cse_and_gimplify_to_preheader): Declare.
21739 (vect_get_place_in_interleaving_chain): Likewise.
21740 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
21742 (_loop_vec_info::~_loop_vec_info): Delete it.
21743 (cse_and_gimplify_to_preheader): New function.
21744 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
21745 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
21746 (vectorizable_load): Likewise. For grouped stores always base
21747 the IV on the first element.
21748 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
21749 condition before gimplifying.
21751 2018-02-07 Jakub Jelinek <jakub@redhat.com>
21753 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
21754 *DIV_EXPR and *MOD_EXPR.
21756 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
21759 * config/i386/i386.c (ix86_option_override_internal): Mask out
21760 the CF_SET bit when checking -fcf-protection.
21762 2018-02-07 Tom de Vries <tom@codesourcery.com>
21765 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
21768 2018-02-07 Richard Biener <rguenther@suse.de>
21770 PR tree-optimization/84204
21771 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
21774 PR tree-optimization/84205
21775 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
21776 special-case isl_ast_op_zdiv_r.
21778 PR tree-optimization/84223
21779 * graphite-scop-detection.c (gather_bbs::before_dom_children):
21780 Only add conditions from within the region.
21781 (gather_bbs::after_dom_children): Adjust.
21783 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
21786 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
21787 * config/avr/avr.md: Only post-reload split REG-REG moves if
21788 either register is GENERAL_REG_P.
21790 2018-02-07 Jakub Jelinek <jakub@redhat.com>
21792 PR tree-optimization/84235
21793 * tree-ssa-scopedtables.c
21794 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
21795 if the subtraction is performed in floating point type where NaNs are
21796 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
21797 build 1. Formatting fix.
21799 2018-02-06 Jakub Jelinek <jakub@redhat.com>
21802 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
21803 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
21804 and skip it regardless of bb boundaries. Use CALL_P macro,
21805 don't test INSN_P (insn) together with CALL_P or JUMP_P check
21806 unnecessarily, formatting fix.
21808 2018-02-06 Michael Collison <michael.collison@arm.com>
21810 * config/arm/thumb2.md:
21811 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
21812 (*thumb_mov_notscc): Ditto.
21814 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
21817 * config/rs6000/rs6000.md (su code attribute): Use "u" for
21818 unsigned_fix, not "s".
21820 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
21822 * configure.ac (gcc_fn_eh_frame_ro): New function.
21823 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
21824 correct .eh_frame permissions.
21825 * configure: Regenerate.
21827 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
21829 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
21830 irrelevant options.
21832 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
21834 * config/rs6000/rs6000.c (rs6000_option_override_internal):
21835 Display warning message for -mno-speculate-indirect-jumps.
21837 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
21839 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
21841 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
21843 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
21845 PR tree-optimization/84225
21846 * tree-eh.c (find_trapping_overflow): Only call
21847 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
21849 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
21852 * config/i386/i386.c: Reimplement the check of possible options
21853 -mibt/-mshstk conbination. Change error messages.
21854 * doc/invoke.texi: Fix a typo: remove extra '='.
21856 2018-02-06 Marek Polacek <polacek@redhat.com>
21858 PR tree-optimization/84228
21859 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
21861 2018-02-06 Tamar Christina <tamar.christina@arm.com>
21864 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
21865 emitted arch directives.
21866 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
21867 __ARM_FEATURE_COPROC before changing architectures.
21869 2018-02-06 Richard Biener <rguenther@suse.de>
21871 * config/i386/i386.c (print_reg): Fix typo.
21872 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
21874 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
21876 * configure: Regenerate.
21878 2018-02-05 Martin Sebor <msebor@redhat.com>
21880 PR tree-optimization/83369
21881 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
21884 2018-02-05 Martin Liska <mliska@suse.cz>
21886 * doc/invoke.texi: Cherry-pick upstream r323995.
21888 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
21890 * ira.c (ira_init_register_move_cost): Adjust comment.
21892 2018-02-05 Martin Liska <mliska@suse.cz>
21894 PR gcov-profile/84137
21895 * doc/gcov.texi: Fix typo in documentation.
21897 2018-02-05 Martin Liska <mliska@suse.cz>
21899 PR gcov-profile/83879
21900 * doc/gcov.texi: Document necessity of --dynamic-list-data when
21901 using dlopen functionality.
21903 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
21905 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
21906 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
21907 _mm_maskz_range_ss, _mm_mask_range_round_ss,
21908 _mm_maskz_range_round_ss): New intrinsics.
21909 (__builtin_ia32_rangesd128_round)
21910 (__builtin_ia32_rangess128_round): Remove.
21911 (__builtin_ia32_rangesd128_mask_round,
21912 __builtin_ia32_rangess128_mask_round): New builtins.
21913 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
21914 __builtin_ia32_rangess128_round): Remove.
21915 (__builtin_ia32_rangesd128_mask_round,
21916 __builtin_ia32_rangess128_mask_round): New builtins.
21917 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
21918 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
21919 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
21920 "<round_saeonly_constraint>")): Changed to ...
21921 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
21922 "<round_saeonly_scalar_constraint>")): ... this.
21923 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
21924 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
21925 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
21926 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
21927 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
21929 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
21931 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
21933 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
21934 Remove all values except native, 8540 and 8548.
21936 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
21938 * config/i386/i386.c (ix86_output_function_return): Pass
21939 INVALID_REGNUM, instead of -1, as invalid register number to
21940 indirect_thunk_name and output_indirect_thunk.
21942 2018-02-02 Julia Koval <julia.koval@intel.com>
21944 * config.gcc: Add -march=icelake.
21945 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
21946 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
21947 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
21948 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
21949 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
21950 (processor_target_table): Add icelake.
21951 (ix86_option_override_internal): Handle new PTAs.
21952 (get_builtin_code_for_version): Handle icelake.
21953 (M_INTEL_COREI7_ICELAKE): New.
21954 (fold_builtin_cpu): Handle icelake.
21955 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
21956 * doc/invoke.texi: Add -march=icelake.
21958 2018-02-02 Julia Koval <julia.koval@intel.com>
21960 * config/i386/i386.c (ix86_option_override_internal): Change flags type
21961 to wide_int_bitmask.
21962 * wide-int-bitmask.h: New.
21964 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
21967 * config/i386/i386.md: Replace Pmode with word_mode in
21968 builtin_setjmp_setup and builtin_longjmp to support x32.
21970 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
21974 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
21976 (rs6000_supported_cpu_names): New static variable.
21977 (linux_cpu_translation_table): Likewise.
21978 (elf_platform) <cpu>: Define new static variable and use it.
21979 Translate kernel AT_PLATFORM name to canonical name if needed.
21980 Error if platform name is unknown.
21982 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
21985 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
21987 2018-02-01 Jeff Law <law@redhat.com>
21990 * config/i386/i386.c (release_scratch_register_on_entry): Add new
21991 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
21992 the scratch if RELEASE_VIA_POP is false.
21993 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
21994 If we have to save a temporary register, decrement SIZE appropriately.
21995 Pass new arguments to release_scratch_register_on_entry.
21996 (ix86_adjust_stack_and_probe): Likewise.
21997 (ix86_emit_probe_stack_range): Pass new arguments to
21998 release_scratch_register_on_entry.
22000 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
22002 PR rtl-optimization/84157
22003 * combine.c (change_zero_ext): Use REG_P predicate in
22004 front of HARD_REGISTER_P predicate.
22006 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
22008 * config/avr/avr.c (avr_option_override): Move disabling of
22009 -fdelete-null-pointer-checks to...
22010 * common/config/avr/avr-common.c (avr_option_optimization_table):
22013 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
22015 PR tree-optimization/81635
22016 * tree-data-ref.c (split_constant_offset_1): For types that
22017 wrap on overflow, try to use range info to prove that wrapping
22020 2018-02-01 Renlin Li <renlin.li@arm.com>
22023 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
22024 TAILCALL_ADDR_REGS.
22025 (aarch64_register_move_cost): Likewise.
22026 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
22027 TAILCALL_ADDR_REGS.
22028 (REG_CLASS_NAMES): Likewise.
22029 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
22030 TAILCALL_ADDR_REGS. Remove IP registers.
22031 * config/aarch64/aarch64.md (Ucs): Update register constraint.
22033 2018-02-01 Richard Biener <rguenther@suse.de>
22035 * domwalk.h (dom_walker::dom_walker): Add additional constructor
22036 for specifying RPO order and allow NULL for that.
22037 * domwalk.c (dom_walker::dom_walker): Likewise.
22038 (dom_walker::walk): Handle NULL RPO order.
22039 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
22041 (rewrite_update_dom_walker): Likewise.
22042 (mark_def_dom_walker): Likewise.
22044 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
22046 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
22047 (aarch64_maybe_expand_sve_subreg_move): Declare.
22048 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
22049 * config/aarch64/predicates.md (aarch64_any_register_operand): New
22051 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
22052 that are semantically a reverse operation.
22053 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
22054 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
22055 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
22057 (aarch64_can_change_mode_class): For big-endian, forbid changes
22058 between two SVE modes if they have different element sizes.
22060 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
22062 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
22063 the TImode handling for big-endian targets.
22065 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
22067 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
22068 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
22070 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
22071 Remove BSWAP handing for big-endian targets and use the form of
22072 LD1RQ appropariate for the mode.
22074 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
22076 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
22077 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
22078 duplicated element.
22080 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
22083 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
22084 check for operands that need to go through aarch64_sve_reload_be.
22086 2018-02-01 Jakub Jelinek <jakub@redhat.com>
22088 PR tree-optimization/81661
22089 PR tree-optimization/84117
22090 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
22091 * tree-eh.c: Include gimplify.h.
22092 (find_trapping_overflow, replace_trapping_overflow,
22093 rewrite_to_non_trapping_overflow): New functions.
22094 * tree-vect-loop.c: Include tree-eh.h.
22095 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
22096 * tree-data-ref.c: Include tree-eh.h.
22097 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
22099 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
22101 PR rtl-optimization/84123
22102 * combine.c (change_zero_ext): Check if hard register satisfies
22103 can_change_dest_mode before calling gen_lowpart_SUBREG.
22105 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
22108 * ira.c (ira_init_register_move_cost): Remove assert.
22110 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
22112 PR rtl-optimization/84071
22113 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
22114 * doc/tm.texi: Regenerate.
22116 2018-01-31 Richard Biener <rguenther@suse.de>
22118 PR tree-optimization/84132
22119 * tree-data-ref.c (analyze_miv_subscript): Properly
22120 check whether evolution_function_is_affine_multivariate_p
22121 before calling gcd_of_steps_may_divide_p.
22123 2018-01-31 Julia Koval <julia.koval@intel.com>
22126 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
22127 * config/i386/i386.md (rdpid_rex64) New.
22128 (rdpid): Make 32bit only.
22130 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
22133 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
22134 an IDENTIFIER_NODE for FUNCTION_TYPE's.
22136 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
22139 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
22141 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
22143 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
22145 PR rtl-optimization/84071
22146 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
22147 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
22149 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
22151 * config/arc/arc.c (arc_handle_aux_attribute): New function.
22152 (arc_attribute_table): Add 'aux' attribute.
22153 (arc_in_small_data_p): Consider aux like variables.
22154 (arc_is_aux_reg_p): New function.
22155 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
22156 (arc_get_aux_arg): New function.
22157 (prepare_move_operands): Handle aux-register access.
22158 (arc_handle_aux_attribute): New function.
22159 * doc/extend.texi (ARC Variable attributes): Add subsection.
22161 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
22163 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
22164 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
22165 (arc_attribute_table): Add 'uncached' attribute.
22166 (arc_print_operand): Print '.di' flag for uncached memory
22168 (arc_in_small_data_p): Do not consider for small data the uncached
22170 (arc_is_uncached_mem_p): New function.
22171 * config/arc/predicates.md (compact_store_memory_operand): Check
22172 for uncached memory accesses.
22173 (nonvol_nonimm_operand): Likewise.
22174 * doc/extend.texi (ARC Type Attribute): New subsection.
22176 2018-01-31 Jakub Jelinek <jakub@redhat.com>
22179 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
22180 falign-loops=): Add Optimization flag.
22182 2018-01-30 Jeff Law <law@redhat.com>
22185 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
22186 INT_REGISTERS_SAVED. Check it prior to calling
22187 get_scratch_register_on_entry.
22188 (ix86_adjust_stack_and_probe): Similarly.
22189 (ix86_emit_probe_stack_range): Similarly.
22190 (ix86_expand_prologue): Corresponding changes.
22192 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
22195 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
22196 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
22198 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
22201 * lra-constraints.c (curr_insn_transform): Process AND in the
22204 2018-01-30 Jakub Jelinek <jakub@redhat.com>
22206 PR rtl-optimization/83986
22207 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
22208 dependence against last_pending_memory_flush in addition to
22209 pending_jump_insns.
22211 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
22213 PR tree-optimization/81611
22214 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
22217 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
22220 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
22223 2018-01-30 Richard Biener <rguenther@suse.de>
22224 Jakub Jelinek <jakub@redhat.com>
22226 PR tree-optimization/84111
22227 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
22228 inner loops added during recursion, as they don't have up-to-date
22231 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
22234 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
22235 (can_inline_edge_by_limits_p): ... here.
22236 (can_early_inline_edge_p, check_callers,
22237 update_caller_keys, update_callee_keys, recursive_inlining,
22238 add_new_edges_to_heap, speculation_useful_p,
22239 inline_small_functions,
22240 inline_small_functions, flatten_function,
22241 inline_to_all_callers_1): Update.
22243 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
22245 * profile-count.c (profile_count::combine_with_ipa_count): Handle
22248 2018-01-30 Richard Biener <rguenther@suse.de>
22250 PR tree-optimization/83008
22251 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
22252 invariant and constant vector uses in stmts when they need
22253 more than one stmt.
22255 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
22258 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
22259 * configure: Regenerate.
22261 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
22263 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
22265 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
22266 Use gen_rtx_REG rather than gen_lowpart.
22268 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
22270 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
22271 rather than 0 when creating partial subregs.
22273 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
22275 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
22278 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
22281 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
22282 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
22283 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
22284 flags. This restores the settings used before the 2017-07-24.
22285 Turning off pre increment/decrement/modify allows IVOPTS to
22286 optimize DF/SF loops where the index is an int.
22288 2018-01-29 Richard Biener <rguenther@suse.de>
22289 Kelvin Nilsen <kelvin@gcc.gnu.org>
22292 * tree-vect-stmts.c (vectorizable_call): Don't call
22293 targetm.vectorize_builtin_md_vectorized_function if callee is
22296 2018-01-22 Carl Love <cel@us.ibm.com>
22298 * doc/extend.tex: Fix typo in second arg in
22299 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
22301 2018-01-29 Richard Biener <rguenther@suse.de>
22303 PR tree-optimization/84086
22304 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
22305 (flush_ssaname_freelist): When SSA names were released reset
22306 the SCEV hash table.
22308 2018-01-29 Richard Biener <rguenther@suse.de>
22310 PR tree-optimization/84057
22311 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
22312 removed paths when removing edges.
22314 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
22316 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
22317 -mfunction-return=@var{choice}.
22319 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
22321 PR diagnostic/84034
22322 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
22323 Handle CR like TAB.
22324 (layout::print_source_line): Likewise.
22325 (test_get_line_width_without_trailing_whitespace): Add test cases.
22327 2018-01-27 Jakub Jelinek <jakub@redhat.com>
22329 PR middle-end/84040
22330 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
22333 2018-01-26 Jim Wilson <jimw@sifive.com>
22335 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
22337 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
22340 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22342 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
22343 and CMP + SUB-immediate -> SUBS.
22345 2018-01-26 Martin Sebor <msebor@redhat.com>
22347 PR tree-optimization/83896
22348 * tree-ssa-strlen.c (get_string_len): Rename...
22349 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
22350 Avoid assuming length is constant.
22351 (handle_char_store): Use HOST_WIDE_INT for string length.
22353 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
22356 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
22357 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
22359 2018-01-26 Richard Biener <rguenther@suse.de>
22361 PR rtl-optimization/84003
22362 * dse.c (record_store): Only record redundant stores when
22363 the earlier store aliases at least all accesses the later one does.
22365 2018-01-26 Jakub Jelinek <jakub@redhat.com>
22367 PR rtl-optimization/83985
22368 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
22369 REG_CFA_RESTORE insns.
22370 (delete_unmarked_insns): Don't ignore separate shrink wrapping
22371 REG_CFA_RESTORE insns here.
22374 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
22375 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
22377 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
22379 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
22380 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
22381 (arc_init): Likewise.
22382 (arc_override_options): Likewise.
22383 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
22385 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
22387 * config/arc/arc.h (TARGET_DBNZ): Define.
22388 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
22389 properly set the tune attribute.
22390 (dbnz): Use TARGET_DBNZ guard.
22391 * config/arc/arc.opt (mtune): Add core3 option.
22393 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
22395 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
22396 recognize new pic like addresses.
22397 (arc_delegitimize_address): Clean up.
22399 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
22401 * config/arc/arc-arches.def: Option mrf16 valid for all
22403 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
22404 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
22405 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
22406 * config/arc/arc-tables.opt: Regenerate.
22407 * config/arc/arc.c (arc_conditional_register_usage): Handle
22408 reduced register file case.
22409 (arc_file_start): Set must have build attributes.
22410 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
22411 mrf16 option value.
22412 * config/arc/arc.opt (mrf16): Add new option.
22413 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
22414 * config/arc/genmultilib.awk: Handle new mrf16 option.
22415 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
22416 * config/arc/t-multilib: Regenerate.
22417 * doc/invoke.texi (ARC Options): Document mrf16 option.
22419 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
22421 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
22422 * config/arc/arc.c (arc_handle_secure_attribute): New function.
22423 (arc_attribute_table): Add 'secure_call' attribute.
22424 (arc_print_operand): Print secure call operand.
22425 (arc_function_ok_for_sibcall): Don't optimize tail calls when
22427 (arc_is_secure_call_p): New function. * config/arc/arc.md
22428 (call_i): Add support for sjli instruction.
22429 (call_value_i): Likewise.
22430 * config/arc/constraints.md (Csc): New constraint.
22432 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
22433 John Eric Martin <John.Martin@emmicro-us.com>
22435 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
22436 * config/arc/arc.c (_arc_jli_section): New struct.
22437 (arc_jli_section): New type.
22438 (rc_jli_sections): New static variable.
22439 (arc_handle_jli_attribute): New function.
22440 (arc_attribute_table): Add jli_always and jli_fixed attribute.
22441 (arc_file_end): New function.
22442 (TARGET_ASM_FILE_END): Define.
22443 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
22444 (arc_add_jli_section): New function.
22445 (jli_call_scan): Likewise.
22446 (arc_reorg): Call jli_call_scan.
22447 (arc_output_addsi): Remove 'S' from printing asm operand.
22448 (arc_is_jli_call_p): New function.
22449 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
22451 (movhi_insn): Likewise.
22452 (movsi_insn): Likewise.
22453 (movsi_set_cc_insn): Likewise.
22454 (loadqi_update): Likewise.
22455 (load_zeroextendqisi_update): Likewise.
22456 (load_signextendqisi_update): Likewise.
22457 (loadhi_update): Likewise.
22458 (load_zeroextendhisi_update): Likewise.
22459 (load_signextendhisi_update): Likewise.
22460 (loadsi_update): Likewise.
22461 (loadsf_update): Likewise.
22462 (movsicc_insn): Likewise.
22463 (bset_insn): Likewise.
22464 (bxor_insn): Likewise.
22465 (bclr_insn): Likewise.
22466 (bmsk_insn): Likewise.
22467 (bicsi3_insn): Likewise.
22468 (cmpsi_cc_c_insn): Likewise.
22469 (movsi_ne): Likewise.
22470 (movsi_cond_exec): Likewise.
22471 (clrsbsi2): Likewise.
22472 (norm_f): Likewise.
22482 (call_i): Remove 'S' asm letter, add jli instruction.
22483 (call_value_i): Likewise.
22484 * config/arc/arc.op (mjli-always): New option.
22485 * config/arc/constraints.md (Cji): New constraint.
22486 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
22488 (subsf3_fpx): Likewise.
22489 (mulsf3_fpx): Likewise.
22490 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
22492 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
22493 function attrbutes.
22494 * doc/invoke.texi (ARC): Document mjli-always option.
22496 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
22498 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
22499 avoid addition with 0 and use incw and decw where possible.
22501 2018-01-26 Richard Biener <rguenther@suse.de>
22503 PR tree-optimization/81082
22504 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
22505 association if it requires casting to unsigned.
22506 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
22507 from fold_plusminus_mult_expr to catch important cases late when
22508 range info is available.
22510 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
22512 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
22513 * configure.ac (hidden_linkonce): New test.
22514 * configure: Regenerate.
22515 * config.in: Regenerate.
22517 2018-01-26 Julia Koval <julia.koval@intel.com>
22519 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
22520 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
22521 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
22522 _mm_mask_bitshuffle_epi64_mask): Fix type.
22523 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
22524 USI_FTYPE_V4DI_V4DI_USI): Remove.
22525 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
22526 __builtin_ia32_vpshufbitqmb256_mask,
22527 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
22528 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
22529 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
22531 2018-01-26 Alan Modra <amodra@gmail.com>
22534 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
22535 UNSPEC_VBPERMQ. Sort other unspecs.
22537 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
22539 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
22541 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
22543 PR middle-end/83055
22544 * predict.c (drop_profile): Do not push/pop cfun; update also
22546 (handle_missing_profiles): Fix logic looking for zero profiles.
22548 2018-01-25 Jakub Jelinek <jakub@redhat.com>
22550 PR middle-end/83977
22551 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
22552 on functions with #pragma omp declare simd or functions with simd
22554 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
22555 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
22556 Remove trailing \n from warning_at calls.
22558 2018-01-25 Tom de Vries <tom@codesourcery.com>
22561 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
22562 for neutered workers.
22564 2018-01-24 Joseph Myers <joseph@codesourcery.com>
22567 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
22568 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
22570 2018-01-24 Jeff Law <law@redhat.com>
22573 * i386.c (get_probe_interval): Move to earlier point.
22574 (ix86_compute_frame_layout): If -fstack-clash-protection and
22575 the frame is larger than the probe interval, then use pushes
22576 to save registers rather than reg->mem moves.
22577 (ix86_expand_prologue): Remove conditional for int_registers_saved
22580 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
22583 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
22584 min/max for never referenced object.
22586 2018-01-24 Jakub Jelinek <jakub@redhat.com>
22588 PR middle-end/83977
22589 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
22591 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
22592 attributes from DECL_ATTRIBUTES (decl) without affecting
22593 DECL_ATTRIBUTES (current_function_decl).
22594 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
22595 functions with non-NULL DECL_ABSTRACT_ORIGIN.
22597 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
22599 PR tree-optimization/83979
22600 * fold-const.c (fold_comparison): Use constant_boolean_node
22601 instead of boolean_{true,false}_node.
22603 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
22605 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
22608 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
22610 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
22611 Simplify the clause that sets the length attribute.
22612 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
22613 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
22614 clause that sets the length attribute.
22615 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
22617 2018-01-24 Tom de Vries <tom@codesourcery.com>
22620 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
22621 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
22622 Add strict parameter.
22623 (prevent_branch_around_nothing): Insert dummy insn between branch to
22624 label and label with no ptx insn inbetween.
22625 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
22627 2018-01-24 Tom de Vries <tom@codesourcery.com>
22630 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
22631 for neutered threads in warp.
22632 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
22634 2018-01-24 Richard Biener <rguenther@suse.de>
22636 PR tree-optimization/83176
22637 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
22640 2018-01-24 Richard Biener <rguenther@suse.de>
22642 PR tree-optimization/82819
22643 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
22644 code generating pluses that are no-ops in the target precision.
22646 2018-01-24 Richard Biener <rguenther@suse.de>
22648 PR middle-end/84000
22649 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
22651 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
22653 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
22654 to merge probabilities.
22655 * predict.c (probably_never_executed): Also mark as cold functions
22656 with global 0 profile and guessed local profile.
22657 * profile-count.c (profile_probability::combine_with_count): New
22659 * profile-count.h (profile_probability::operator*,
22660 profile_probability::operator*=, profile_probability::operator/,
22661 profile_probability::operator/=): Reduce precision to adjusted
22662 and set value to guessed on contradictory divisions.
22663 (profile_probability::combine_with_freq): Remove.
22664 (profile_probability::combine_wiht_count): Declare.
22665 (profile_count::force_nonzero):: Set to adjusted.
22666 (profile_count::probability_in):: Set quality to adjusted.
22667 * tree-ssa-tail-merge.c (replace_block_by): Use
22668 combine_with_count.
22670 2018-01-23 Andrew Waterman <andrew@sifive.com>
22671 Jim Wilson <jimw@sifive.com>
22673 * config/riscv/riscv.c (riscv_stack_boundary): New.
22674 (riscv_option_override): Set riscv_stack_boundary. Handle
22675 riscv_preferred_stack_boundary_arg.
22676 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
22677 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
22678 (STACK_BOUNDARY): Set to riscv_stack_boundary.
22679 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
22680 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
22681 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
22683 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
22686 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
22687 of struct ix86_frame.
22688 (ix86_expand_epilogue): Likewise. Add a local variable for
22689 the reg_save_offset field in struct ix86_frame.
22691 2018-01-23 Bin Cheng <bin.cheng@arm.com>
22693 PR tree-optimization/82604
22694 * tree-loop-distribution.c (enum partition_kind): New enum item
22695 PKIND_PARTIAL_MEMSET.
22696 (partition_builtin_p): Support above new enum item.
22697 (generate_code_for_partition): Ditto.
22698 (compute_access_range): Differentiate cases that equality can be
22699 proven at all loops, the innermost loops or no loops.
22700 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
22701 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
22702 (finalize_partitions, distribute_loop): Don't fuse partition of
22703 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
22704 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
22705 parloop is enabled.
22707 2018-01-23 Martin Liska <mliska@suse.cz>
22709 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
22710 order to ignore the predictor.
22711 (PRED_POLYMORPHIC_CALL): Likewise.
22712 (PRED_RECURSIVE_CALL): Likewise.
22714 2018-01-23 Martin Liska <mliska@suse.cz>
22716 * tree-profile.c (tree_profiling): Print function header to
22717 aware reader which function we are working on.
22718 * value-prof.c (gimple_find_values_to_profile): Do not print
22719 not interesting value histograms.
22721 2018-01-23 Martin Liska <mliska@suse.cz>
22723 * profile-count.h (enum profile_quality): Add
22724 profile_uninitialized as the first value. Do not number values
22725 as they are zero based.
22726 (profile_count::verify): Update sanity check.
22727 (profile_probability::verify): Likewise.
22729 2018-01-23 Nathan Sidwell <nathan@acm.org>
22731 * doc/invoke.texi (ffor-scope): Deprecate.
22733 2018-01-23 David Malcolm <dmalcolm@redhat.com>
22735 PR tree-optimization/83510
22736 * domwalk.c (set_all_edges_as_executable): New function.
22737 (dom_walker::dom_walker): Convert bool param
22738 "skip_unreachable_blocks" to enum reachability. Move setup of
22739 edge flags to set_all_edges_as_executable and only do it when
22740 reachability is REACHABLE_BLOCKS.
22741 * domwalk.h (enum dom_walker::reachability): New enum.
22742 (dom_walker::dom_walker): Convert bool param
22743 "skip_unreachable_blocks" to enum reachability.
22744 (set_all_edges_as_executable): New decl.
22745 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
22746 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
22748 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
22749 but converting true to REACHABLE_BLOCKS.
22750 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
22752 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
22753 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
22754 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
22756 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
22757 if check_all_array_refs will be called.
22759 2018-01-23 David Malcolm <dmalcolm@redhat.com>
22761 * tree.c (selftest::test_location_wrappers): Add more test
22764 2018-01-23 David Malcolm <dmalcolm@redhat.com>
22766 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
22767 (selftest::test_bit_in_range): Likewise.
22769 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
22772 * doc/sourcebuild.texi (vect_float): Say that the selector
22773 only describes the situation when -funsafe-math-optimizations is on.
22774 (vect_float_strict): Document.
22776 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
22778 PR tree-optimization/83965
22779 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
22780 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
22781 instead of checking only for a reduction.
22782 (vect_recog_widen_sum_pattern): Likewise.
22784 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
22786 * predict.c (probably_never_executed): Only use precise profile info.
22787 (compute_function_frequency): Skip after inlining hack since we now
22788 have quality checking.
22790 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
22792 * profile-count.h (profile_probability::very_unlikely,
22793 profile_probability::unlikely, profile_probability::even): Set
22794 precision to guessed.
22796 2018-01-23 Richard Biener <rguenther@suse.de>
22798 PR tree-optimization/83963
22799 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
22800 Properly terminate dominator walk when crossing the exit edge not
22801 when visiting its source block.
22803 2018-01-23 Jakub Jelinek <jakub@redhat.com>
22806 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
22807 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
22809 2018-01-22 Jakub Jelinek <jakub@redhat.com>
22811 PR tree-optimization/83957
22812 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
22813 semicolon after for body surrounded by braces.
22815 PR tree-optimization/83081
22816 * profile-count.h (profile_probability::split): New method.
22817 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
22818 Use profile_probability::split.
22819 (do_compare_rtx_and_jump): Fix adjustment of probabilities
22820 when splitting a single conditional jump into 2.
22822 2018-01-22 David Malcolm <dmalcolm@redhat.com>
22824 PR tree-optimization/69452
22825 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
22828 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
22830 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
22831 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
22832 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
22834 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
22836 * config/rl78/rl78-protos.h (rl78_split_movdi): New function
22838 * config/rl78/rl78.md (movdi): New define_expand.
22839 * config/rl78/rl78.c (rl78_split_movdi): New function.
22841 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
22844 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
22846 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
22847 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
22848 128-bit to produce an UNSPEC move to get the double word with the
22849 signbit and then a shift directly to do signbit.
22850 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
22851 implementation with a new version that just does either a direct
22852 move or a regular move. Move memory interface to separate insns.
22853 Move insns so they are next to the expander.
22854 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
22855 with signbit move. Split big and little endian case.
22856 (signbit<mode>2_dm_mem_le): Likewise.
22857 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
22858 (signbit<mode>2_dm2): Likewise.
22860 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
22862 * config/rl78/rl78.md (anddi3): New define_expand.
22864 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
22866 * config/rl78/rl78.md (umindi3): New define_expand.
22868 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
22870 * config/rl78/rl78.md (smindi3): New define_expand.
22872 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
22874 * config/rl78/rl78.md (smaxdi3): New define_expand.
22876 2018-01-22 Carl Love <cel@us.ibm.com>
22878 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
22879 LVX_V1TI): Add macro expansion.
22880 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
22881 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
22882 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
22883 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
22884 Change check to determine if the instruction is a byte reversing
22885 entry. Fix typo in comment.
22886 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
22887 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
22888 Add def_builtin calls for new builtins.
22889 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
22890 Add define_insn expansion.
22892 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
22894 * config/rl78/rl78.md (umaxdi3): New define_expand.
22896 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
22898 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
22899 for non-QImode registers.
22901 2018-01-22 Richard Biener <rguenther@suse.de>
22903 PR tree-optimization/83963
22904 * graphite-scop-detection.c (scop_detection::get_sese): Delay
22905 including the loop exit block.
22906 (scop_detection::merge_sese): Likewise.
22907 (scop_detection::add_scop): Do it here instead.
22909 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22911 * doc/sourcebuild.texi (arm_softfloat): Document.
22913 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
22916 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
22917 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
22918 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
22920 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
22921 David Edelsohn <dje.gcc@gmail.com>
22924 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
22925 Change "crset eq" to "crset 2".
22926 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
22927 (*call_indirect_aix<mode>_nospec): Likewise.
22928 (*call_value_indirect_aix<mode>_nospec): Likewise.
22929 (*call_indirect_elfv2<mode>_nospec): Likewise.
22930 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
22931 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
22932 change assembly output from . to $.
22933 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
22934 (indirect_jump<mode>_nospec): Change assembly output from . to $.
22935 (*tablejump<mode>_internal1_nospec): Likewise.
22937 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
22940 * config/sh/sh_optimize_sett_clrt.cc:
22941 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
22943 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
22945 PR tree-optimization/83940
22946 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
22947 offset_dt to vect_constant_def rather than vect_unknown_def_type.
22948 (vect_check_load_store_mask): Add a mask_dt_out parameter and
22949 use it to pass back the definition type.
22950 (vect_check_store_rhs): Likewise rhs_dt_out.
22951 (vect_build_gather_load_calls): Add a mask_dt argument and use
22952 it instead of a call to vect_is_simple_use.
22953 (vectorizable_store): Update calls to vect_check_load_store_mask
22954 and vect_check_store_rhs. Use the dt returned by the latter instead
22955 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
22956 instead of calls to vect_is_simple_use. Pass the scalar rather
22957 than the vector operand to vect_is_simple_use when handling
22958 second and subsequent copies of an rhs value.
22959 (vectorizable_load): Update calls to vect_check_load_store_mask
22960 and vect_build_gather_load_calls. Use the cached mask_dt and
22961 gs_info.offset_dt instead of calls to vect_is_simple_use.
22963 2018-01-20 Jakub Jelinek <jakub@redhat.com>
22965 PR middle-end/83945
22966 * tree-emutls.c: Include gimplify.h.
22967 (lower_emutls_2): New function.
22968 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
22969 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
22970 it before further processing.
22973 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
22974 UINTVAL (trueop1) instead of INTVAL (op1).
22976 2018-01-19 Jakub Jelinek <jakub@redhat.com>
22980 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
22981 INCOMING_FRAME_SP_OFFSET if not defined.
22982 (scan_trace): Add ENTRY argument. If true and
22983 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
22984 emit a note to adjust the CFA offset.
22985 (create_cfi_notes): Adjust scan_trace callers.
22986 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
22987 INCOMING_FRAME_SP_OFFSET in the CIE.
22988 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
22989 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
22991 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
22992 * doc/tm.texi: Regenerated.
22994 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
22996 PR rtl-optimization/83147
22997 * lra-constraints.c (remove_inheritance_pseudos): Use
22998 lra_substitute_pseudo_within_insn.
23000 2018-01-19 Tom de Vries <tom@codesourcery.com>
23001 Cesar Philippidis <cesar@codesourcery.com>
23004 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
23006 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
23009 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
23010 spaces for function labels.
23012 2018-01-19 Martin Liska <mliska@suse.cz>
23014 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
23015 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
23016 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
23017 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
23018 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
23019 (PRED_CONST_RETURN): Change from 69 to 65.
23020 (PRED_NULL_RETURN): Change from 91 to 71.
23021 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
23022 (PRED_LOOP_GUARD): Change from 66 to 73.
23024 2018-01-19 Martin Liska <mliska@suse.cz>
23026 * predict.c (predict_insn_def): Add new assert.
23027 (struct branch_predictor): Change type to signed integer.
23028 (test_prediction_value_range): Amend test to cover
23029 PROB_UNINITIALIZED.
23030 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
23031 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
23032 (PRED_LOOP_ITERATIONS_MAX): Likewise.
23033 (PRED_LOOP_IV_COMPARE): Likewise.
23034 * predict.h (PROB_UNINITIALIZED): Define new constant.
23036 2018-01-19 Martin Liska <mliska@suse.cz>
23038 * predict.c (dump_prediction): Add new format for
23039 analyze_brprob.py script which is enabled with -details
23041 * profile-count.h (precise_p): New function.
23043 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
23045 PR tree-optimization/83922
23046 * tree-vect-loop.c (vect_verify_full_masking): Return false if
23047 there are no statements that need masking.
23048 (vect_active_double_reduction_p): New function.
23049 (vect_analyze_loop_operations): Use it when handling phis that
23050 are not in the loop header.
23052 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
23054 PR tree-optimization/83914
23055 * tree-vect-loop.c (vectorizable_induction): Don't convert
23056 init_expr or apply the peeling adjustment for inductions
23057 that are nested within the vectorized loop.
23059 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23061 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
23064 2018-01-18 Jakub Jelinek <jakub@redhat.com>
23068 * function.h (gimplify_parameters): Add gimple_seq * argument.
23069 * function.c: Include gimple.h and options.h.
23070 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
23071 for the added local temporaries if needed.
23072 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
23073 if there are any parameter cleanups, wrap whole body into a
23074 try/finally with the cleanups.
23076 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
23079 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
23080 Use GET_MODE_CLASS for scalar floating point.
23082 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
23086 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
23087 Fix call of call_cgraph_insertion_hooks.
23089 2018-01-18 Martin Sebor <msebor@redhat.com>
23091 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
23093 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
23096 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
23099 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
23102 * common.opt: (-ffile-prefix-map): New option.
23103 * opts.c (common_handle_option): Defer it.
23104 * opts-global.c (handle_common_deferred_options): Handle it.
23105 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
23106 * file-prefix-map.h: New file.
23107 (remap_debug_filename, add_debug_prefix_map): ...here.
23108 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
23109 * final.c (debug_prefix_map, add_debug_prefix_map
23110 remap_debug_filename): Move to...
23111 * file-prefix-map.c: New file.
23112 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
23113 generalize, get rid of alloca(), use strrchr() instead of strchr().
23114 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
23115 Implement in terms of add_prefix_map().
23116 (remap_macro_filename, remap_debug_filename): Implement in term of
23118 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
23119 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
23120 * dbxout.c: Include file-prefix-map.h.
23121 * varasm.c: Likewise.
23122 * vmsdbgout.c: Likewise.
23123 * xcoffout.c: Likewise.
23124 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
23125 * doc/cppopts.texi (-fmacro-prefix-map): Document.
23126 * doc/invoke.texi (-ffile-prefix-map): Document.
23127 (-fdebug-prefix-map): Update description.
23129 2018-01-18 Martin Liska <mliska@suse.cz>
23131 * config/i386/i386.c (indirect_thunk_name): Document that also
23133 (output_indirect_thunk): Document why both instructions
23134 (pause and lfence) are generated.
23136 2018-01-18 Richard Biener <rguenther@suse.de>
23138 PR tree-optimization/83887
23139 * graphite-scop-detection.c
23140 (scop_detection::get_nearest_dom_with_single_entry): Remove.
23141 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
23142 (scop_detection::merge_sese): Re-implement with a flood-fill
23143 algorithm that properly finds a SESE region if it exists.
23145 2018-01-18 Jakub Jelinek <jakub@redhat.com>
23148 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
23149 pointer_diff optimizations use view_convert instead of convert.
23151 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
23153 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
23154 Generate different code for -mno-speculate-indirect-jumps.
23155 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
23156 (*call_indirect_aix<mode>): Disable for
23157 -mno-speculate-indirect-jumps.
23158 (*call_indirect_aix<mode>_nospec): New define_insn.
23159 (*call_value_indirect_aix<mode>): Disable for
23160 -mno-speculate-indirect-jumps.
23161 (*call_value_indirect_aix<mode>_nospec): New define_insn.
23162 (*sibcall_nonlocal_sysv<mode>): Generate different code for
23163 -mno-speculate-indirect-jumps.
23164 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
23166 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
23168 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
23169 long double type, set the flags for noting the default long double
23170 type, even if we don't pass or return a long double type.
23172 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
23175 * ipa-inline.c (flatten_function): Do not overwrite final inlining
23178 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
23180 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
23181 support for merge[hl].
23182 (fold_mergehl_helper): New helper function.
23183 (tree-vector-builder.h): New #include for tree_vector_builder usage.
23184 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
23185 (altivec_vmrglw_direct): Add xxmrglw insn.
23187 2018-01-17 Andrew Waterman <andrew@sifive.com>
23189 * config/riscv/riscv.c (riscv_conditional_register_usage): If
23190 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
23192 2018-01-17 David Malcolm <dmalcolm@redhat.com>
23195 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
23196 call the lto_location_cache before reading the
23197 DECL_SOURCE_LOCATION of the types.
23199 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
23200 Richard Sandiford <richard.sandiford@linaro.org>
23202 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
23203 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
23204 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
23205 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
23206 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
23208 * config/aarch64/constraints.md (aarch64_movti_operand):
23210 * config/aarch64/predicates.md (Uti): Add new constraint.
23212 2018-01-17 Carl Love <cel@us.ibm.com>
23214 * config/rs6000/vsx.md (define_expand xl_len_r,
23215 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
23216 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
23218 (define_expand, define_insn): Move the shift left from the
23219 define_insn to the define_expand for lxvl and stxvl instructions.
23220 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
23221 and XL_LEN_R definitions to PURE.
23223 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
23225 * config/i386/i386.c (indirect_thunk_name): Declare regno
23226 as unsigned int. Compare regno with INVALID_REGNUM.
23227 (output_indirect_thunk): Ditto.
23228 (output_indirect_thunk_function): Ditto.
23229 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
23230 in the call to output_indirect_thunk_function.
23232 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
23234 PR middle-end/83884
23235 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
23236 rather than the size of inner_type to determine the stack slot size
23237 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
23239 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
23242 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
23245 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
23247 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
23248 endian Linux systems to optionally enable multilibs for selecting
23249 the long double type if the user configured an explicit type.
23250 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
23251 have no long double multilibs if not defined.
23252 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
23253 warn if the user used -mabi={ieee,ibm}longdouble and we built
23254 multilibs for long double.
23255 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
23256 appropriate multilib option.
23257 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
23259 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
23260 for building long double multilibs.
23261 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
23263 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
23265 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
23268 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
23270 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
23273 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
23276 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
23279 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
23281 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
23282 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
23284 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
23286 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
23287 different rtl trees depending on TARGET_64BIT.
23288 (rs6000_gen_lvx): Likewise.
23290 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
23292 * config/visium/visium.md (nop): Tweak comment.
23293 (hazard_nop): Likewise.
23295 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
23297 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
23298 -mspeculate-indirect-jumps.
23299 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
23300 for -mno-speculate-indirect-jumps.
23301 (*call_indirect_elfv2<mode>_nospec): New define_insn.
23302 (*call_value_indirect_elfv2<mode>): Disable for
23303 -mno-speculate-indirect-jumps.
23304 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
23305 (indirect_jump): Emit different RTL for
23306 -mno-speculate-indirect-jumps.
23307 (*indirect_jump<mode>): Disable for
23308 -mno-speculate-indirect-jumps.
23309 (*indirect_jump<mode>_nospec): New define_insn.
23310 (tablejump): Emit different RTL for
23311 -mno-speculate-indirect-jumps.
23312 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
23313 (tablejumpsi_nospec): New define_expand.
23314 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
23315 (tablejumpdi_nospec): New define_expand.
23316 (*tablejump<mode>_internal1): Disable for
23317 -mno-speculate-indirect-jumps.
23318 (*tablejump<mode>_internal1_nospec): New define_insn.
23319 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
23322 2018-01-16 Artyom Skrobov tyomitch@gmail.com
23324 * caller-save.c (insert_save): Drop unnecessary parameter. All
23327 2018-01-16 Jakub Jelinek <jakub@redhat.com>
23328 Richard Biener <rguenth@suse.de>
23331 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
23332 return early, inline manually is_gimple_sizepos. Make sure if we
23333 call gimplify_expr we don't end up with a gimple constant.
23334 * tree.c (variably_modified_type_p): Don't return true for
23335 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
23336 * gimplify.h (is_gimple_sizepos): Remove.
23338 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
23340 PR tree-optimization/83857
23341 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
23342 vectorizable_live_operation for pure SLP statements.
23343 (vectorizable_live_operation): Handle PHIs.
23345 2018-01-16 Richard Biener <rguenther@suse.de>
23347 PR tree-optimization/83867
23348 * tree-vect-stmts.c (vect_transform_stmt): Precompute
23349 nested_in_vect_loop_p since the scalar stmt may get invalidated.
23351 2018-01-16 Jakub Jelinek <jakub@redhat.com>
23354 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
23355 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
23356 If off is not INTEGER_CST, issue a may not be aligned warning
23357 rather than isn't aligned. Use isn%'t rather than isn't.
23358 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
23360 <case MULT_EXPR>: Improve the case when bottom and one of the
23361 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
23362 operand, in that case check if the other operand is multiple of
23363 bottom divided by the INTEGER_CST operand.
23365 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
23368 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
23369 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
23370 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
23371 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
23372 * config/pa/pa.c (pa_function_arg_advance): Likewise.
23373 (pa_function_arg, pa_arg_partial_bytes): Likewise.
23374 (pa_function_arg_size): New function.
23376 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
23378 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
23379 in a separate statement.
23381 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
23383 PR tree-optimization/83847
23384 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
23385 group gathers and scatters.
23387 2018-01-16 Jakub Jelinek <jakub@redhat.com>
23389 PR rtl-optimization/86620
23390 * params.def (max-sched-ready-insns): Bump minimum value to 1.
23392 PR rtl-optimization/83213
23393 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
23394 to last if both are JUMP_INSNs.
23396 PR tree-optimization/83843
23397 * gimple-ssa-store-merging.c
23398 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
23399 store_immediate_info for bswap/nop orig_stores.
23401 2018-01-15 Andrew Waterman <andrew@sifive.com>
23403 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
23405 <UDIV>: Increase cost if !TARGET_DIV.
23407 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
23409 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
23410 (define_attr "cr_logical_3op"): New.
23411 (cceq_ior_compare): Adjust.
23412 (cceq_ior_compare_complement): Adjust.
23413 (*cceq_rev_compare): Adjust.
23414 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
23415 (is_cracked_insn): Adjust.
23416 (insn_must_be_first_in_group): Adjust.
23417 * config/rs6000/40x.md: Adjust.
23418 * config/rs6000/440.md: Adjust.
23419 * config/rs6000/476.md: Adjust.
23420 * config/rs6000/601.md: Adjust.
23421 * config/rs6000/603.md: Adjust.
23422 * config/rs6000/6xx.md: Adjust.
23423 * config/rs6000/7450.md: Adjust.
23424 * config/rs6000/7xx.md: Adjust.
23425 * config/rs6000/8540.md: Adjust.
23426 * config/rs6000/cell.md: Adjust.
23427 * config/rs6000/e300c2c3.md: Adjust.
23428 * config/rs6000/e500mc.md: Adjust.
23429 * config/rs6000/e500mc64.md: Adjust.
23430 * config/rs6000/e5500.md: Adjust.
23431 * config/rs6000/e6500.md: Adjust.
23432 * config/rs6000/mpc.md: Adjust.
23433 * config/rs6000/power4.md: Adjust.
23434 * config/rs6000/power5.md: Adjust.
23435 * config/rs6000/power6.md: Adjust.
23436 * config/rs6000/power7.md: Adjust.
23437 * config/rs6000/power8.md: Adjust.
23438 * config/rs6000/power9.md: Adjust.
23439 * config/rs6000/rs64.md: Adjust.
23440 * config/rs6000/titan.md: Adjust.
23442 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
23444 * config/i386/predicates.md (indirect_branch_operand): Rewrite
23445 ix86_indirect_branch_register logic.
23447 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
23449 * config/i386/constraints.md (Bs): Update
23450 ix86_indirect_branch_register check. Don't check
23451 ix86_indirect_branch_register with GOT_memory_operand.
23453 * config/i386/predicates.md (GOT_memory_operand): Don't check
23454 ix86_indirect_branch_register here.
23455 (GOT32_symbol_operand): Likewise.
23457 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
23459 * config/i386/predicates.md (constant_call_address_operand):
23460 Rewrite ix86_indirect_branch_register logic.
23461 (sibcall_insn_operand): Likewise.
23463 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
23465 * config/i386/constraints.md (Bs): Replace
23466 ix86_indirect_branch_thunk_register with
23467 ix86_indirect_branch_register.
23469 * config/i386/i386.md (indirect_jump): Likewise.
23470 (tablejump): Likewise.
23471 (*sibcall_memory): Likewise.
23472 (*sibcall_value_memory): Likewise.
23473 Peepholes of indirect call and jump via memory: Likewise.
23474 * config/i386/i386.opt: Likewise.
23475 * config/i386/predicates.md (indirect_branch_operand): Likewise.
23476 (GOT_memory_operand): Likewise.
23477 (call_insn_operand): Likewise.
23478 (sibcall_insn_operand): Likewise.
23479 (GOT32_symbol_operand): Likewise.
23481 2018-01-15 Jakub Jelinek <jakub@redhat.com>
23483 PR middle-end/83837
23484 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
23485 type rather than type addr's type points to.
23486 (expand_omp_atomic_mutex): Likewise.
23487 (expand_omp_atomic): Likewise.
23489 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
23492 * config/i386/i386.c (output_indirect_thunk_function): Use
23493 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
23494 for __x86_return_thunk.
23496 2018-01-15 Richard Biener <rguenther@suse.de>
23498 PR middle-end/83850
23499 * expmed.c (extract_bit_field_1): Fix typo.
23501 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23504 * config/arm/iterators.md (VF): New mode iterator.
23505 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
23506 Remove integer-related logic from pattern.
23507 (neon_vabd<mode>_3): Likewise.
23509 2018-01-15 Jakub Jelinek <jakub@redhat.com>
23511 PR middle-end/82694
23512 * common.opt (fstrict-overflow): No longer an alias.
23513 (fwrapv-pointer): New option.
23514 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
23515 also for pointer types based on flag_wrapv_pointer.
23516 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
23517 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
23518 opts->x_flag_wrapv got set.
23519 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
23520 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
23521 POINTER_TYPE_OVERFLOW_UNDEFINED.
23522 * match.pd: Likewise in address comparison pattern.
23523 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
23525 2018-01-15 Richard Biener <rguenther@suse.de>
23528 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
23529 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
23530 Reset type names to their identifier if their TYPE_DECL doesn't
23531 have linkage (and thus is used for ODR and devirt).
23532 (save_debug_info_for_decl): Remove.
23533 (save_debug_info_for_type): Likewise.
23534 (add_tree_to_fld_list): Adjust.
23535 * tree-pretty-print.c (dump_generic_node): Make dumping of
23536 type names more robust.
23538 2018-01-15 Richard Biener <rguenther@suse.de>
23540 * BASE-VER: Bump to 8.0.1.
23542 2018-01-14 Martin Sebor <msebor@redhat.com>
23545 * builtins.c (check_access): Avoid warning when the no-warning bit
23548 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
23550 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
23551 * ira-color (allocno_hard_regs_compare): Likewise.
23553 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
23556 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
23557 Use .pushsection/.popsection.
23559 2018-01-14 Martin Sebor <msebor@redhat.com>
23562 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
23564 2018-01-14 Jakub Jelinek <jakub@redhat.com>
23566 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
23567 entry from extra_headers.
23568 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
23569 extra_headers, make the list bitwise identical to the i?86-*-* one.
23571 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
23573 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
23574 -mcmodel=large with -mindirect-branch=thunk,
23575 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
23576 -mfunction-return=thunk-extern.
23577 * doc/invoke.texi: Document -mcmodel=large is incompatible with
23578 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
23579 -mfunction-return=thunk and -mfunction-return=thunk-extern.
23581 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
23583 * config/i386/i386.c (print_reg): Print the name of the full
23584 integer register without '%'.
23585 (ix86_print_operand): Handle 'V'.
23586 * doc/extend.texi: Document 'V' modifier.
23588 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
23590 * config/i386/constraints.md (Bs): Disallow memory operand for
23591 -mindirect-branch-register.
23593 * config/i386/predicates.md (indirect_branch_operand): Likewise.
23594 (GOT_memory_operand): Likewise.
23595 (call_insn_operand): Likewise.
23596 (sibcall_insn_operand): Likewise.
23597 (GOT32_symbol_operand): Likewise.
23598 * config/i386/i386.md (indirect_jump): Call convert_memory_address
23599 for -mindirect-branch-register.
23600 (tablejump): Likewise.
23601 (*sibcall_memory): Likewise.
23602 (*sibcall_value_memory): Likewise.
23603 Disallow peepholes of indirect call and jump via memory for
23604 -mindirect-branch-register.
23605 (*call_pop): Replace m with Bw.
23606 (*call_value_pop): Likewise.
23607 (*sibcall_pop_memory): Replace m with Bs.
23608 * config/i386/i386.opt (mindirect-branch-register): New option.
23609 * doc/invoke.texi: Document -mindirect-branch-register option.
23611 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
23613 * config/i386/i386-protos.h (ix86_output_function_return): New.
23614 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
23615 set function_return_type.
23616 (indirect_thunk_name): Add ret_p to indicate thunk for function
23618 (output_indirect_thunk_function): Pass false to
23619 indirect_thunk_name.
23620 (ix86_output_indirect_branch_via_reg): Likewise.
23621 (ix86_output_indirect_branch_via_push): Likewise.
23622 (output_indirect_thunk_function): Create alias for function
23623 return thunk if regno < 0.
23624 (ix86_output_function_return): New function.
23625 (ix86_handle_fndecl_attribute): Handle function_return.
23626 (ix86_attribute_table): Add function_return.
23627 * config/i386/i386.h (machine_function): Add
23628 function_return_type.
23629 * config/i386/i386.md (simple_return_internal): Use
23630 ix86_output_function_return.
23631 (simple_return_internal_long): Likewise.
23632 * config/i386/i386.opt (mfunction-return=): New option.
23633 (indirect_branch): Mention -mfunction-return=.
23634 * doc/extend.texi: Document function_return function attribute.
23635 * doc/invoke.texi: Document -mfunction-return= option.
23637 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
23639 * config/i386/i386-opts.h (indirect_branch): New.
23640 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
23641 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
23642 with local indirect jump when converting indirect call and jump.
23643 (ix86_set_indirect_branch_type): New.
23644 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
23645 (indirectlabelno): New.
23646 (indirect_thunk_needed): Likewise.
23647 (indirect_thunk_bnd_needed): Likewise.
23648 (indirect_thunks_used): Likewise.
23649 (indirect_thunks_bnd_used): Likewise.
23650 (INDIRECT_LABEL): Likewise.
23651 (indirect_thunk_name): Likewise.
23652 (output_indirect_thunk): Likewise.
23653 (output_indirect_thunk_function): Likewise.
23654 (ix86_output_indirect_branch_via_reg): Likewise.
23655 (ix86_output_indirect_branch_via_push): Likewise.
23656 (ix86_output_indirect_branch): Likewise.
23657 (ix86_output_indirect_jmp): Likewise.
23658 (ix86_code_end): Call output_indirect_thunk_function if needed.
23659 (ix86_output_call_insn): Call ix86_output_indirect_branch if
23661 (ix86_handle_fndecl_attribute): Handle indirect_branch.
23662 (ix86_attribute_table): Add indirect_branch.
23663 * config/i386/i386.h (machine_function): Add indirect_branch_type
23664 and has_local_indirect_jump.
23665 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
23667 (tablejump): Likewise.
23668 (*indirect_jump): Use ix86_output_indirect_jmp.
23669 (*tablejump_1): Likewise.
23670 (simple_return_indirect_internal): Likewise.
23671 * config/i386/i386.opt (mindirect-branch=): New option.
23672 (indirect_branch): New.
23675 (thunk-inline): Likewise.
23676 (thunk-extern): Likewise.
23677 * doc/extend.texi: Document indirect_branch function attribute.
23678 * doc/invoke.texi: Document -mindirect-branch= option.
23680 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
23683 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
23685 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
23687 * ipa-inline.c (want_inline_small_function_p): Return false if
23688 inlining has already failed with CIF_FINAL_ERROR.
23689 (update_caller_keys): Call want_inline_small_function_p before
23691 (update_callee_keys): Likewise.
23693 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
23695 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
23697 (rs6000_quadword_masked_address_p): Likewise.
23698 (quad_aligned_load_p): Likewise.
23699 (quad_aligned_store_p): Likewise.
23700 (const_load_sequence_p): Add comment to describe the outer-most loop.
23701 (mimic_memory_attributes_and_flags): New function.
23702 (rs6000_gen_stvx): Likewise.
23703 (replace_swapped_aligned_store): Likewise.
23704 (rs6000_gen_lvx): Likewise.
23705 (replace_swapped_aligned_load): Likewise.
23706 (replace_swapped_load_constant): Capitalize argument name in
23707 comment describing this function.
23708 (rs6000_analyze_swaps): Add a third pass to search for vector loads
23709 and stores that access quad-word aligned addresses and replace
23710 with stvx or lvx instructions when appropriate.
23711 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
23712 New function prototype.
23713 (rs6000_quadword_masked_address_p): Likewise.
23714 (rs6000_gen_lvx): Likewise.
23715 (rs6000_gen_stvx): Likewise.
23716 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
23717 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
23718 when memory address is aligned.
23719 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
23720 this split to select lvx instruction when memory address is aligned.
23721 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
23722 instruction when memory address is aligned.
23723 (*vsx_le_perm_load_v16qi): Likewise.
23724 (four unnamed splitters): Modify to select the stvx instruction
23725 when memory is aligned.
23727 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
23729 * predict.c (determine_unlikely_bbs): Handle correctly BBs
23730 which appears in the queue multiple times.
23732 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
23733 Alan Hayward <alan.hayward@arm.com>
23734 David Sherwood <david.sherwood@arm.com>
23736 * tree-vectorizer.h (vec_lower_bound): New structure.
23737 (_loop_vec_info): Add check_nonzero and lower_bounds.
23738 (LOOP_VINFO_CHECK_NONZERO): New macro.
23739 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
23740 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
23741 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
23742 fields. Make seg_len the distance travelled, not including the
23744 (dr_direction_indicator): Declare.
23745 (dr_zero_step_indicator): Likewise.
23746 (dr_known_forward_stride_p): Likewise.
23747 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
23749 (runtime_alias_check_p): Allow runtime alias checks with
23751 (operator ==): Compare access_size and align.
23752 (prune_runtime_alias_test_list): Rework for new distinction between
23753 the access_size and seg_len.
23754 (create_intersect_range_checks_index): Likewise. Cope with polynomial
23756 (get_segment_min_max): New function.
23757 (create_intersect_range_checks): Use it.
23758 (dr_step_indicator): New function.
23759 (dr_direction_indicator): Likewise.
23760 (dr_zero_step_indicator): Likewise.
23761 (dr_known_forward_stride_p): Likewise.
23762 * tree-loop-distribution.c (data_ref_segment_size): Return
23763 DR_STEP * (niters - 1).
23764 (compute_alias_check_pairs): Update call to the dr_with_seg_len
23766 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
23767 (vect_preserves_scalar_order_p): New function, split out from...
23768 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
23769 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
23770 (vect_vfa_access_size): New function.
23771 (vect_vfa_align): Likewise.
23772 (vect_compile_time_alias): Take access_size_a and access_b arguments.
23773 (dump_lower_bound): New function.
23774 (vect_check_lower_bound): Likewise.
23775 (vect_small_gap_p): Likewise.
23776 (vectorizable_with_step_bound_p): Likewise.
23777 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
23778 depencies if the vectorization factor is 1. Convert the checks
23779 for nonzero steps into checks on the bounds of DR_STEP. Try using
23780 a bunds check for variable steps if the minimum required step is
23781 relatively small. Update calls to the dr_with_seg_len
23782 constructor and to vect_compile_time_alias.
23783 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
23785 (vect_loop_versioning): Call it.
23786 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
23788 (vect_estimate_min_profitable_iters): Account for any bounds checks.
23790 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
23791 Alan Hayward <alan.hayward@arm.com>
23792 David Sherwood <david.sherwood@arm.com>
23794 * doc/sourcebuild.texi (vect_scatter_store): Document.
23795 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
23797 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
23799 * genopinit.c (main): Add supports_vec_scatter_store and
23800 supports_vec_scatter_store_cached to target_optabs.
23801 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
23802 IFN_MASK_SCATTER_STORE.
23803 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
23805 * internal-fn.h (internal_store_fn_p): Declare.
23806 (internal_fn_stored_value_index): Likewise.
23807 * internal-fn.c (scatter_store_direct): New macro.
23808 (expand_scatter_store_optab_fn): New function.
23809 (direct_scatter_store_optab_supported_p): New macro.
23810 (internal_store_fn_p): New function.
23811 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
23812 IFN_MASK_SCATTER_STORE.
23813 (internal_fn_mask_index): Likewise.
23814 (internal_fn_stored_value_index): New function.
23815 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
23816 for scatter stores.
23817 * optabs-query.h (supports_vec_scatter_store_p): Declare.
23818 * optabs-query.c (supports_vec_scatter_store_p): New function.
23819 * tree-vectorizer.h (vect_get_store_rhs): Declare.
23820 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
23821 true for scatter stores.
23822 (vect_gather_scatter_fn_p): Handle scatter stores too.
23823 (vect_check_gather_scatter): Consider using scatter stores if
23824 supports_vec_scatter_store_p.
23825 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
23826 scatter stores too.
23827 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
23828 internal_fn_stored_value_index.
23829 (check_load_store_masking): Handle scatter stores too.
23830 (vect_get_store_rhs): Make public.
23831 (vectorizable_call): Use internal_store_fn_p.
23832 (vectorizable_store): Handle scatter store internal functions.
23833 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
23834 when deciding whether the end of the group has been reached.
23835 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
23836 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
23837 (mask_scatter_store<mode>): New insns.
23839 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
23840 Alan Hayward <alan.hayward@arm.com>
23841 David Sherwood <david.sherwood@arm.com>
23843 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
23844 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
23845 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
23847 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
23848 Use vect_truncate_gather_scatter_offset if we can't treat the
23849 operation as a normal gather load or scatter store.
23850 (get_group_load_store_type): Take the gather_scatter_info
23851 as argument. Try using a gather load or scatter store for
23852 single-element groups.
23853 (get_load_store_type): Update calls to get_group_load_store_type
23854 and vect_use_strided_gather_scatters_p.
23856 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
23857 Alan Hayward <alan.hayward@arm.com>
23858 David Sherwood <david.sherwood@arm.com>
23860 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
23861 optional tree argument.
23862 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
23864 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
23865 but continue to use the current value as a fallback.
23866 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
23867 to compare the updates.
23868 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
23869 (get_load_store_type): Use it when handling a strided access.
23870 (vect_get_strided_load_store_ops): New function.
23871 (vect_get_data_ptr_increment): Likewise.
23872 (vectorizable_load): Handle strided gather loads. Always pass
23873 a step to vect_create_data_ref_ptr and bump_vector_ptr.
23875 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
23876 Alan Hayward <alan.hayward@arm.com>
23877 David Sherwood <david.sherwood@arm.com>
23879 * doc/md.texi (gather_load@var{m}): Document.
23880 (mask_gather_load@var{m}): Likewise.
23881 * genopinit.c (main): Add supports_vec_gather_load and
23882 supports_vec_gather_load_cached to target_optabs.
23883 * optabs-tree.c (init_tree_optimization_optabs): Use
23884 ggc_cleared_alloc to allocate target_optabs.
23885 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
23886 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
23888 * internal-fn.h (internal_load_fn_p): Declare.
23889 (internal_gather_scatter_fn_p): Likewise.
23890 (internal_fn_mask_index): Likewise.
23891 (internal_gather_scatter_fn_supported_p): Likewise.
23892 * internal-fn.c (gather_load_direct): New macro.
23893 (expand_gather_load_optab_fn): New function.
23894 (direct_gather_load_optab_supported_p): New macro.
23895 (direct_internal_fn_optab): New function.
23896 (internal_load_fn_p): Likewise.
23897 (internal_gather_scatter_fn_p): Likewise.
23898 (internal_fn_mask_index): Likewise.
23899 (internal_gather_scatter_fn_supported_p): Likewise.
23900 * optabs-query.c (supports_at_least_one_mode_p): New function.
23901 (supports_vec_gather_load_p): Likewise.
23902 * optabs-query.h (supports_vec_gather_load_p): Declare.
23903 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
23904 and memory_type field.
23905 (NUM_PATTERNS): Bump to 15.
23906 * tree-vect-data-refs.c: Include internal-fn.h.
23907 (vect_gather_scatter_fn_p): New function.
23908 (vect_describe_gather_scatter_call): Likewise.
23909 (vect_check_gather_scatter): Try using internal functions for
23910 gather loads. Recognize existing calls to a gather load function.
23911 (vect_analyze_data_refs): Consider using gather loads if
23912 supports_vec_gather_load_p.
23913 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
23914 (vect_get_gather_scatter_offset_type): Likewise.
23915 (vect_convert_mask_for_vectype): Likewise.
23916 (vect_add_conversion_to_patterm): Likewise.
23917 (vect_try_gather_scatter_pattern): Likewise.
23918 (vect_recog_gather_scatter_pattern): New pattern recognizer.
23919 (vect_vect_recog_func_ptrs): Add it.
23920 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
23921 internal_fn_mask_index and internal_gather_scatter_fn_p.
23922 (check_load_store_masking): Take the gather_scatter_info as an
23923 argument and handle gather loads.
23924 (vect_get_gather_scatter_ops): New function.
23925 (vectorizable_call): Check internal_load_fn_p.
23926 (vectorizable_load): Likewise. Handle gather load internal
23928 (vectorizable_store): Update call to check_load_store_masking.
23929 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
23930 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
23931 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
23932 (aarch64_gather_scale_operand_d): New predicates.
23933 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
23934 (mask_gather_load<mode>): New insns.
23936 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
23937 Alan Hayward <alan.hayward@arm.com>
23938 David Sherwood <david.sherwood@arm.com>
23940 * optabs.def (fold_left_plus_optab): New optab.
23941 * doc/md.texi (fold_left_plus_@var{m}): Document.
23942 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
23943 * internal-fn.c (fold_left_direct): Define.
23944 (expand_fold_left_optab_fn): Likewise.
23945 (direct_fold_left_optab_supported_p): Likewise.
23946 * fold-const-call.c (fold_const_fold_left): New function.
23947 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
23948 * tree-parloops.c (valid_reduction_p): New function.
23949 (gather_scalar_reductions): Use it.
23950 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
23951 (vect_finish_replace_stmt): Declare.
23952 * tree-vect-loop.c (fold_left_reduction_fn): New function.
23953 (needs_fold_left_reduction_p): New function, split out from...
23954 (vect_is_simple_reduction): ...here. Accept reductions that
23955 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
23956 (vect_force_simple_reduction): Also store the reduction type in
23957 the assignment's STMT_VINFO_REDUC_TYPE.
23958 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
23959 (merge_with_identity): New function.
23960 (vect_expand_fold_left): Likewise.
23961 (vectorize_fold_left_reduction): Likewise.
23962 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
23963 scalar phi in place for it. Check for target support and reject
23964 cases that would reassociate the operation. Defer the transform
23965 phase to vectorize_fold_left_reduction.
23966 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
23967 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
23968 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
23970 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
23972 * tree-if-conv.c (predicate_mem_writes): Remove redundant
23973 call to ifc_temp_var.
23975 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
23976 Alan Hayward <alan.hayward@arm.com>
23977 David Sherwood <david.sherwood@arm.com>
23979 * target.def (legitimize_address_displacement): Take the original
23980 offset as a poly_int.
23981 * targhooks.h (default_legitimize_address_displacement): Update
23983 * targhooks.c (default_legitimize_address_displacement): Likewise.
23984 * doc/tm.texi: Regenerate.
23985 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
23986 as an argument, moving assert of ad->disp == ad->disp_term to...
23987 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
23988 Try calling targetm.legitimize_address_displacement before expanding
23989 the address rather than afterwards, and adjust for the new interface.
23990 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
23991 Match the new hook interface. Handle SVE addresses.
23992 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
23993 new hook interface.
23995 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
23997 * Makefile.in (OBJS): Add early-remat.o.
23998 * target.def (select_early_remat_modes): New hook.
23999 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
24000 * doc/tm.texi: Regenerate.
24001 * targhooks.h (default_select_early_remat_modes): Declare.
24002 * targhooks.c (default_select_early_remat_modes): New function.
24003 * timevar.def (TV_EARLY_REMAT): New timevar.
24004 * passes.def (pass_early_remat): New pass.
24005 * tree-pass.h (make_pass_early_remat): Declare.
24006 * early-remat.c: New file.
24007 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
24009 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
24011 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
24012 Alan Hayward <alan.hayward@arm.com>
24013 David Sherwood <david.sherwood@arm.com>
24015 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
24016 vfm1 with a bound_epilog parameter.
24017 (vect_do_peeling): Update calls accordingly, and move the prologue
24018 call earlier in the function. Treat the base bound_epilog as 0 for
24019 fully-masked loops and retain vf - 1 for other loops. Add 1 to
24020 this base when peeling for gaps.
24021 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
24022 with fully-masked loops.
24023 (vect_estimate_min_profitable_iters): Handle the single peeled
24024 iteration in that case.
24026 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
24027 Alan Hayward <alan.hayward@arm.com>
24028 David Sherwood <david.sherwood@arm.com>
24030 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
24031 single-element interleaving even if the size is not a power of 2.
24032 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
24033 accesses for single-element interleaving if the group size is
24036 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
24037 Alan Hayward <alan.hayward@arm.com>
24038 David Sherwood <david.sherwood@arm.com>
24040 * doc/md.texi (fold_extract_last_@var{m}): Document.
24041 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
24042 * optabs.def (fold_extract_last_optab): New optab.
24043 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
24044 * internal-fn.c (fold_extract_direct): New macro.
24045 (expand_fold_extract_optab_fn): Likewise.
24046 (direct_fold_extract_optab_supported_p): Likewise.
24047 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
24048 * tree-vect-loop.c (vect_model_reduction_cost): Handle
24049 EXTRACT_LAST_REDUCTION.
24050 (get_initial_def_for_reduction): Do not create an initial vector
24051 for EXTRACT_LAST_REDUCTION reductions.
24052 (vectorizable_reduction): Leave the scalar phi in place for
24053 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
24054 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
24055 epilogue code for EXTRACT_LAST_REDUCTION and defer the
24056 transform phase to vectorizable_condition.
24057 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
24059 (vect_finish_stmt_generation): ...here.
24060 (vect_finish_replace_stmt): New function.
24061 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
24062 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
24064 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
24066 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
24067 Alan Hayward <alan.hayward@arm.com>
24068 David Sherwood <david.sherwood@arm.com>
24070 * doc/md.texi (extract_last_@var{m}): Document.
24071 * optabs.def (extract_last_optab): New optab.
24072 * internal-fn.def (EXTRACT_LAST): New internal function.
24073 * internal-fn.c (cond_unary_direct): New macro.
24074 (expand_cond_unary_optab_fn): Likewise.
24075 (direct_cond_unary_optab_supported_p): Likewise.
24076 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
24077 loops using EXTRACT_LAST.
24078 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
24079 (extract_last_<mode>): ...this optab.
24080 (vec_extract<mode><Vel>): Update accordingly.
24082 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
24083 Alan Hayward <alan.hayward@arm.com>
24084 David Sherwood <david.sherwood@arm.com>
24086 * target.def (empty_mask_is_expensive): New hook.
24087 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
24088 * doc/tm.texi: Regenerate.
24089 * targhooks.h (default_empty_mask_is_expensive): Declare.
24090 * targhooks.c (default_empty_mask_is_expensive): New function.
24091 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
24092 if the target says that empty masks are expensive.
24093 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
24095 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
24097 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
24098 Alan Hayward <alan.hayward@arm.com>
24099 David Sherwood <david.sherwood@arm.com>
24101 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
24102 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
24103 (vect_use_loop_mask_for_alignment_p): New function.
24104 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
24105 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
24106 niters_skip argument. Make sure that the first niters_skip elements
24107 of the first iteration are inactive.
24108 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
24109 Update call to vect_set_loop_masks_directly.
24110 (get_misalign_in_elems): New function, split out from...
24111 (vect_gen_prolog_loop_niters): ...here.
24112 (vect_update_init_of_dr): Take a code argument that specifies whether
24113 the adjustment should be added or subtracted.
24114 (vect_update_init_of_drs): Likewise.
24115 (vect_prepare_for_masked_peels): New function.
24116 (vect_do_peeling): Skip prologue peeling if we're using a mask
24117 instead. Update call to vect_update_inits_of_drs.
24118 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
24120 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
24121 alignment. Do not include the number of peeled iterations in
24122 the minimum threshold in that case.
24123 (vectorizable_induction): Adjust the start value down by
24124 LOOP_VINFO_MASK_SKIP_NITERS iterations.
24125 (vect_transform_loop): Call vect_prepare_for_masked_peels.
24126 Take the number of skipped iterations into account when calculating
24128 * tree-vect-stmts.c (vect_gen_while_not): New function.
24130 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
24131 Alan Hayward <alan.hayward@arm.com>
24132 David Sherwood <david.sherwood@arm.com>
24134 * doc/sourcebuild.texi (vect_fully_masked): Document.
24135 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
24136 default value to 0.
24137 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
24139 (vect_analyze_loop_2): ...here. Don't check the vectorization
24140 factor against the number of loop iterations if the loop is
24143 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
24144 Alan Hayward <alan.hayward@arm.com>
24145 David Sherwood <david.sherwood@arm.com>
24147 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
24148 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
24149 (dump_groups): Update accordingly.
24150 (iv_use::mem_type): New member variable.
24151 (address_p): New function.
24152 (record_use): Add a mem_type argument and initialize the new
24154 (record_group_use): Add a mem_type argument. Use address_p.
24155 Remove obsolete null checks of base_object. Update call to record_use.
24156 (find_interesting_uses_op): Update call to record_group_use.
24157 (find_interesting_uses_cond): Likewise.
24158 (find_interesting_uses_address): Likewise.
24159 (get_mem_type_for_internal_fn): New function.
24160 (find_address_like_use): Likewise.
24161 (find_interesting_uses_stmt): Try find_address_like_use before
24162 calling find_interesting_uses_op.
24163 (addr_offset_valid_p): Use the iv mem_type field as the type
24164 of the addressed memory.
24165 (add_autoinc_candidates): Likewise.
24166 (get_address_cost): Likewise.
24167 (split_small_address_groups_p): Use address_p.
24168 (split_address_groups): Likewise.
24169 (add_iv_candidate_for_use): Likewise.
24170 (autoinc_possible_for_pair): Likewise.
24171 (rewrite_groups): Likewise.
24172 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
24173 (determine_group_iv_cost): Update after split of USE_ADDRESS.
24174 (get_alias_ptr_type_for_ptr_address): New function.
24175 (rewrite_use_address): Rewrite address uses in calls that were
24176 identified by find_address_like_use.
24178 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
24179 Alan Hayward <alan.hayward@arm.com>
24180 David Sherwood <david.sherwood@arm.com>
24182 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
24184 * gimple-expr.h (is_gimple_addressable: Likewise.
24185 * gimple-expr.c (is_gimple_address): Likewise.
24186 * internal-fn.c (expand_call_mem_ref): New function.
24187 (expand_mask_load_optab_fn): Use it.
24188 (expand_mask_store_optab_fn): Likewise.
24190 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
24191 Alan Hayward <alan.hayward@arm.com>
24192 David Sherwood <david.sherwood@arm.com>
24194 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
24195 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
24196 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
24197 (cond_umax@var{mode}): Document.
24198 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
24199 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
24200 (cond_umin_optab, cond_umax_optab): New optabs.
24201 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
24202 (COND_IOR, COND_XOR): New internal functions.
24203 * internal-fn.h (get_conditional_internal_fn): Declare.
24204 * internal-fn.c (cond_binary_direct): New macro.
24205 (expand_cond_binary_optab_fn): Likewise.
24206 (direct_cond_binary_optab_supported_p): Likewise.
24207 (get_conditional_internal_fn): New function.
24208 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
24209 Cope with reduction statements that are vectorized as calls rather
24211 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
24212 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
24213 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
24214 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
24215 (UNSPEC_COND_EOR): New unspecs.
24216 (optab): Add mappings for them.
24217 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
24218 (sve_int_op, sve_fp_op): New int attributes.
24220 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
24221 Alan Hayward <alan.hayward@arm.com>
24222 David Sherwood <david.sherwood@arm.com>
24224 * optabs.def (while_ult_optab): New optab.
24225 * doc/md.texi (while_ult@var{m}@var{n}): Document.
24226 * internal-fn.def (WHILE_ULT): New internal function.
24227 * internal-fn.h (direct_internal_fn_supported_p): New override
24228 that takes two types as argument.
24229 * internal-fn.c (while_direct): New macro.
24230 (expand_while_optab_fn): New function.
24231 (convert_optab_supported_p): Likewise.
24232 (direct_while_optab_supported_p): New macro.
24233 * wide-int.h (wi::udiv_ceil): New function.
24234 * tree-vectorizer.h (rgroup_masks): New structure.
24235 (vec_loop_masks): New typedef.
24236 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
24237 and fully_masked_p.
24238 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
24239 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
24240 (vect_max_vf): New function.
24241 (slpeel_make_loop_iterate_ntimes): Delete.
24242 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
24243 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
24244 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
24245 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
24246 internal-fn.h, stor-layout.h and optabs-query.h.
24247 (vect_set_loop_mask): New function.
24248 (add_preheader_seq): Likewise.
24249 (add_header_seq): Likewise.
24250 (interleave_supported_p): Likewise.
24251 (vect_maybe_permute_loop_masks): Likewise.
24252 (vect_set_loop_masks_directly): Likewise.
24253 (vect_set_loop_condition_masked): Likewise.
24254 (vect_set_loop_condition_unmasked): New function, split out from
24255 slpeel_make_loop_iterate_ntimes.
24256 (slpeel_make_loop_iterate_ntimes): Rename to..
24257 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
24258 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
24259 (vect_do_peeling): Update call accordingly.
24260 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
24262 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
24263 mask_compare_type, can_fully_mask_p and fully_masked_p.
24264 (release_vec_loop_masks): New function.
24265 (_loop_vec_info): Use it to free the loop masks.
24266 (can_produce_all_loop_masks_p): New function.
24267 (vect_get_max_nscalars_per_iter): Likewise.
24268 (vect_verify_full_masking): Likewise.
24269 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
24270 retries, and free the mask rgroups before retrying. Check loop-wide
24271 reasons for disallowing fully-masked loops. Make the final decision
24272 about whether use a fully-masked loop or not.
24273 (vect_estimate_min_profitable_iters): Do not assume that peeling
24274 for the number of iterations will be needed for fully-masked loops.
24275 (vectorizable_reduction): Disable fully-masked loops.
24276 (vectorizable_live_operation): Likewise.
24277 (vect_halve_mask_nunits): New function.
24278 (vect_double_mask_nunits): Likewise.
24279 (vect_record_loop_mask): Likewise.
24280 (vect_get_loop_mask): Likewise.
24281 (vect_transform_loop): Handle the case in which the final loop
24282 iteration might handle a partial vector. Call vect_set_loop_condition
24283 instead of slpeel_make_loop_iterate_ntimes.
24284 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
24285 (check_load_store_masking): New function.
24286 (prepare_load_store_mask): Likewise.
24287 (vectorizable_store): Handle fully-masked loops.
24288 (vectorizable_load): Likewise.
24289 (supportable_widening_operation): Use vect_halve_mask_nunits for
24291 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
24292 (vect_gen_while): New function.
24293 * config/aarch64/aarch64.md (umax<mode>3): New expander.
24294 (aarch64_uqdec<mode>): New insn.
24296 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
24297 Alan Hayward <alan.hayward@arm.com>
24298 David Sherwood <david.sherwood@arm.com>
24300 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
24301 (reduc_xor_scal_optab): New optabs.
24302 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
24303 (reduc_xor_scal_@var{m}): Document.
24304 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
24305 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
24306 internal functions.
24307 * fold-const-call.c (fold_const_call): Handle them.
24308 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
24309 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
24310 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
24311 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
24312 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
24313 (UNSPEC_XORV): New unspecs.
24314 (optab): Add entries for them.
24315 (BITWISEV): New int iterator.
24316 (bit_reduc_op): New int attributes.
24318 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
24319 Alan Hayward <alan.hayward@arm.com>
24320 David Sherwood <david.sherwood@arm.com>
24322 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
24323 * internal-fn.def (VEC_SHL_INSERT): New internal function.
24324 * optabs.def (vec_shl_insert_optab): New optab.
24325 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
24326 (duplicate_and_interleave): Likewise.
24327 * tree-vect-loop.c: Include internal-fn.h.
24328 (neutral_op_for_slp_reduction): New function, split out from
24329 get_initial_defs_for_reduction.
24330 (get_initial_def_for_reduction): Handle option 2 for variable-length
24331 vectors by loading the neutral value into a vector and then shifting
24332 the initial value into element 0.
24333 (get_initial_defs_for_reduction): Replace the code argument with
24334 the neutral value calculated by neutral_op_for_slp_reduction.
24335 Use gimple_build_vector for constant-length vectors.
24336 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
24337 but the first group_size elements have a neutral value.
24338 Use duplicate_and_interleave otherwise.
24339 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
24340 Update call to get_initial_defs_for_reduction. Handle SLP
24341 reductions for variable-length vectors by creating one vector
24342 result for each scalar result, with the elements associated
24343 with other scalar results stubbed out with the neutral value.
24344 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
24345 Require IFN_VEC_SHL_INSERT for double reductions on
24346 variable-length vectors, or SLP reductions that have
24347 a neutral value. Require can_duplicate_and_interleave_p
24348 support for variable-length unchained SLP reductions if there
24349 is no neutral value, such as for MIN/MAX reductions. Also require
24350 the number of vector elements to be a multiple of the number of
24351 SLP statements when doing variable-length unchained SLP reductions.
24352 Update call to vect_create_epilog_for_reduction.
24353 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
24354 and remove initial values.
24355 (duplicate_and_interleave): Make public.
24356 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
24357 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
24359 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
24360 Alan Hayward <alan.hayward@arm.com>
24361 David Sherwood <david.sherwood@arm.com>
24363 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
24364 (can_duplicate_and_interleave_p): New function.
24365 (vect_get_and_check_slp_defs): Take the vector of statements
24366 rather than just the current one. Remove excess parentheses.
24367 Restriction rejectinon of vect_constant_def and vect_external_def
24368 for variable-length vectors to boolean types, or types for which
24369 can_duplicate_and_interleave_p is false.
24370 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
24371 (duplicate_and_interleave): New function.
24372 (vect_get_constant_vectors): Use gimple_build_vector for
24373 constant-length vectors and suitable variable-length constant
24374 vectors. Use duplicate_and_interleave for other variable-length
24375 vectors. Don't defer the update when inserting new statements.
24377 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
24378 Alan Hayward <alan.hayward@arm.com>
24379 David Sherwood <david.sherwood@arm.com>
24381 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
24382 min_profitable_iters doesn't go negative.
24384 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
24385 Alan Hayward <alan.hayward@arm.com>
24386 David Sherwood <david.sherwood@arm.com>
24388 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
24389 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
24390 * optabs.def (vec_mask_load_lanes_optab): New optab.
24391 (vec_mask_store_lanes_optab): Likewise.
24392 * internal-fn.def (MASK_LOAD_LANES): New internal function.
24393 (MASK_STORE_LANES): Likewise.
24394 * internal-fn.c (mask_load_lanes_direct): New macro.
24395 (mask_store_lanes_direct): Likewise.
24396 (expand_mask_load_optab_fn): Handle masked operations.
24397 (expand_mask_load_lanes_optab_fn): New macro.
24398 (expand_mask_store_optab_fn): Handle masked operations.
24399 (expand_mask_store_lanes_optab_fn): New macro.
24400 (direct_mask_load_lanes_optab_supported_p): Likewise.
24401 (direct_mask_store_lanes_optab_supported_p): Likewise.
24402 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
24404 (vect_load_lanes_supported): Likewise.
24405 * tree-vect-data-refs.c (strip_conversion): New function.
24406 (can_group_stmts_p): Likewise.
24407 (vect_analyze_data_ref_accesses): Use it instead of checking
24408 for a pair of assignments.
24409 (vect_store_lanes_supported): Take a masked_p parameter.
24410 (vect_load_lanes_supported): Likewise.
24411 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
24412 vect_store_lanes_supported and vect_load_lanes_supported.
24413 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
24414 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
24415 parameter. Don't allow gaps for masked accesses.
24416 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
24417 and vect_load_lanes_supported.
24418 (get_load_store_type): Take a masked_p parameter and update
24419 call to get_group_load_store_type.
24420 (vectorizable_store): Update call to get_load_store_type.
24421 Handle IFN_MASK_STORE_LANES.
24422 (vectorizable_load): Update call to get_load_store_type.
24423 Handle IFN_MASK_LOAD_LANES.
24425 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
24426 Alan Hayward <alan.hayward@arm.com>
24427 David Sherwood <david.sherwood@arm.com>
24429 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
24431 * config/aarch64/aarch64-protos.h
24432 (aarch64_sve_struct_memory_operand_p): Declare.
24433 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
24434 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
24435 (VPRED, vpred): Handle SVE structure modes.
24436 * config/aarch64/constraints.md (Utx): New constraint.
24437 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
24438 (aarch64_sve_struct_nonimmediate_operand): New predicates.
24439 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
24440 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
24441 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
24442 structure modes. Split into pieces after RA.
24443 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
24444 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
24446 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
24447 SVE structure modes.
24448 (aarch64_classify_address): Likewise.
24449 (sizetochar): Move earlier in file.
24450 (aarch64_print_operand): Handle SVE register lists.
24451 (aarch64_array_mode): New function.
24452 (aarch64_sve_struct_memory_operand_p): Likewise.
24453 (TARGET_ARRAY_MODE): Redefine.
24455 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
24456 Alan Hayward <alan.hayward@arm.com>
24457 David Sherwood <david.sherwood@arm.com>
24459 * target.def (array_mode): New target hook.
24460 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
24461 * doc/tm.texi: Regenerate.
24462 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
24463 * hooks.c (hook_optmode_mode_uhwi_none): New function.
24464 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
24465 targetm.array_mode.
24466 * stor-layout.c (mode_for_array): Likewise. Support polynomial
24469 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
24470 Alan Hayward <alan.hayward@arm.com>
24471 David Sherwood <david.sherwood@arm.com>
24473 * fold-const.c (fold_binary_loc): Check the argument types
24474 rather than the result type when testing for a vector operation.
24476 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
24478 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
24479 * doc/tm.texi: Regenerate.
24481 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
24482 Alan Hayward <alan.hayward@arm.com>
24483 David Sherwood <david.sherwood@arm.com>
24485 * doc/invoke.texi (-msve-vector-bits=): Document new option.
24486 (sve): Document new AArch64 extension.
24487 * doc/md.texi (w): Extend the description of the AArch64
24488 constraint to include SVE vectors.
24489 (Upl, Upa): Document new AArch64 predicate constraints.
24490 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
24492 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
24493 (msve-vector-bits=): New option.
24494 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
24495 SVE when these are disabled.
24496 (sve): New extension.
24497 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
24498 modes. Adjust their number of units based on aarch64_sve_vg.
24499 (MAX_BITSIZE_MODE_ANY_MODE): Define.
24500 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
24501 aarch64_addr_query_type.
24502 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
24503 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
24504 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
24505 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
24506 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
24507 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
24508 (aarch64_simd_imm_zero_p): Delete.
24509 (aarch64_check_zero_based_sve_index_immediate): Declare.
24510 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
24511 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
24512 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
24513 (aarch64_sve_float_mul_immediate_p): Likewise.
24514 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
24515 rather than an rtx.
24516 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
24517 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
24518 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
24519 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
24520 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
24521 (aarch64_regmode_natural_size): Likewise.
24522 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
24523 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
24525 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
24526 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
24527 for VG and the SVE predicate registers.
24528 (V_ALIASES): Add a "z"-prefixed alias.
24529 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
24530 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
24531 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
24532 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
24533 (REG_CLASS_NAMES): Add entries for them.
24534 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
24535 and the predicate registers.
24536 (aarch64_sve_vg): Declare.
24537 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
24538 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
24539 (REGMODE_NATURAL_SIZE): Define.
24540 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
24542 * config/aarch64/aarch64.c: Include cfgrtl.h.
24543 (simd_immediate_info): Add a constructor for series vectors,
24544 and an associated step field.
24545 (aarch64_sve_vg): New variable.
24546 (aarch64_dbx_register_number): Handle VG and the predicate registers.
24547 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
24548 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
24549 (VEC_ANY_DATA, VEC_STRUCT): New constants.
24550 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
24551 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
24552 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
24553 (aarch64_get_mask_mode): New functions.
24554 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
24555 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
24556 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
24557 predicate modes and predicate registers. Explicitly restrict
24558 GPRs to modes of 16 bytes or smaller. Only allow FP registers
24559 to store a vector mode if it is recognized by
24560 aarch64_classify_vector_mode.
24561 (aarch64_regmode_natural_size): New function.
24562 (aarch64_hard_regno_caller_save_mode): Return the original mode
24564 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
24565 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
24566 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
24567 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
24569 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
24570 does not overlap dest if the function is frame-related. Handle
24572 (aarch64_split_add_offset): New function.
24573 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
24574 them aarch64_add_offset.
24575 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
24576 and update call to aarch64_sub_sp.
24577 (aarch64_add_cfa_expression): New function.
24578 (aarch64_expand_prologue): Pass extra temporary registers to the
24579 functions above. Handle the case in which we need to emit new
24580 DW_CFA_expressions for registers that were originally saved
24581 relative to the stack pointer, but now have to be expressed
24582 relative to the frame pointer.
24583 (aarch64_output_mi_thunk): Pass extra temporary registers to the
24585 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
24586 IP0 and IP1 values for SVE frames.
24587 (aarch64_expand_vec_series): New function.
24588 (aarch64_expand_sve_widened_duplicate): Likewise.
24589 (aarch64_expand_sve_const_vector): Likewise.
24590 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
24591 Handle SVE constants. Use emit_move_insn to move a force_const_mem
24592 into the register, rather than emitting a SET directly.
24593 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
24594 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
24595 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
24596 (offset_9bit_signed_scaled_p): New functions.
24597 (aarch64_replicate_bitmask_imm): New function.
24598 (aarch64_bitmask_imm): Use it.
24599 (aarch64_cannot_force_const_mem): Reject expressions involving
24600 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
24601 (aarch64_classify_index): Handle SVE indices, by requiring
24602 a plain register index with a scale that matches the element size.
24603 (aarch64_classify_address): Handle SVE addresses. Assert that
24604 the mode of the address is VOIDmode or an integer mode.
24605 Update call to aarch64_classify_symbol.
24606 (aarch64_classify_symbolic_expression): Update call to
24607 aarch64_classify_symbol.
24608 (aarch64_const_vec_all_in_range_p): New function.
24609 (aarch64_print_vector_float_operand): Likewise.
24610 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
24611 "vN" for FP registers with SVE modes. Handle (const ...) vectors
24612 and the FP immediates 1.0 and 0.5.
24613 (aarch64_print_address_internal): Handle SVE addresses.
24614 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
24615 (aarch64_regno_regclass): Handle predicate registers.
24616 (aarch64_secondary_reload): Handle big-endian reloads of SVE
24618 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
24619 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
24620 (aarch64_convert_sve_vector_bits): New function.
24621 (aarch64_override_options): Use it to handle -msve-vector-bits=.
24622 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
24623 rather than an rtx.
24624 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
24625 Handle SVE vector and predicate modes. Accept VL-based constants
24626 that need only one temporary register, and VL offsets that require
24627 no temporary registers.
24628 (aarch64_conditional_register_usage): Mark the predicate registers
24629 as fixed if SVE isn't available.
24630 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
24631 Return true for SVE vector and predicate modes.
24632 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
24633 rather than an unsigned int. Handle SVE modes.
24634 (aarch64_preferred_simd_mode): Update call accordingly. Handle
24636 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
24638 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
24639 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
24640 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
24641 (aarch64_sve_float_mul_immediate_p): New functions.
24642 (aarch64_sve_valid_immediate): New function.
24643 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
24644 Explicitly reject structure modes. Check for INDEX constants.
24645 Handle PTRUE and PFALSE constants.
24646 (aarch64_check_zero_based_sve_index_immediate): New function.
24647 (aarch64_simd_imm_zero_p): Delete.
24648 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
24649 vector modes. Accept constants in the range of CNT[BHWD].
24650 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
24651 ask for an Advanced SIMD mode.
24652 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
24653 (aarch64_simd_vector_alignment): Handle SVE predicates.
24654 (aarch64_vectorize_preferred_vector_alignment): New function.
24655 (aarch64_simd_vector_alignment_reachable): Use it instead of
24657 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
24658 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
24660 (MAX_VECT_LEN): Delete.
24661 (expand_vec_perm_d): Add a vec_flags field.
24662 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
24663 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
24664 (aarch64_evpc_ext): Don't apply a big-endian lane correction
24666 (aarch64_evpc_rev): Rename to...
24667 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
24668 (aarch64_evpc_rev_global): New function.
24669 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
24670 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
24672 (aarch64_evpc_sve_tbl): New function.
24673 (aarch64_expand_vec_perm_const_1): Update after rename of
24674 aarch64_evpc_rev. Handle SVE permutes too, trying
24675 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
24676 than aarch64_evpc_tbl.
24677 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
24678 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
24679 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
24680 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
24681 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
24682 (aarch64_expand_sve_vcond): New functions.
24683 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
24684 of aarch64_vector_mode_p.
24685 (aarch64_dwarf_poly_indeterminate_value): New function.
24686 (aarch64_compute_pressure_classes): Likewise.
24687 (aarch64_can_change_mode_class): Likewise.
24688 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
24689 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
24690 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
24691 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
24692 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
24693 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
24694 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
24695 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
24697 (Dn, Dl, Dr): Accept const as well as const_vector.
24698 (Dz): Likewise. Compare against CONST0_RTX.
24699 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
24700 of "vector" where appropriate.
24701 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
24702 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
24703 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
24704 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
24705 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
24706 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
24707 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
24708 (v_int_equiv): Extend to SVE modes.
24709 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
24711 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
24712 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
24713 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
24714 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
24715 (SVE_COND_FP_CMP): New int iterators.
24716 (perm_hilo): Handle the new unpack unspecs.
24717 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
24719 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
24720 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
24721 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
24722 (aarch64_equality_operator, aarch64_constant_vector_operand)
24723 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
24724 (aarch64_sve_nonimmediate_operand): Likewise.
24725 (aarch64_sve_general_operand): Likewise.
24726 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
24727 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
24728 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
24729 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
24730 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
24731 (aarch64_sve_float_arith_immediate): Likewise.
24732 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
24733 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
24734 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
24735 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
24736 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
24737 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
24738 (aarch64_sve_float_arith_operand): Likewise.
24739 (aarch64_sve_float_arith_with_sub_operand): Likewise.
24740 (aarch64_sve_float_mul_operand): Likewise.
24741 (aarch64_sve_vec_perm_operand): Likewise.
24742 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
24743 (aarch64_mov_operand): Accept const_poly_int and const_vector.
24744 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
24745 as well as const_vector.
24746 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
24747 in file. Use CONST0_RTX and CONSTM1_RTX.
24748 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
24749 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
24750 Use aarch64_simd_imm_zero.
24751 * config/aarch64/aarch64-sve.md: New file.
24752 * config/aarch64/aarch64.md: Include it.
24753 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
24754 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
24755 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
24756 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
24757 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
24758 (sve): New attribute.
24759 (enabled): Disable instructions with the sve attribute unless
24761 (movqi, movhi): Pass CONST_POLY_INT operaneds through
24762 aarch64_expand_mov_immediate.
24763 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
24764 CNT[BHSD] immediates.
24765 (movti): Split CONST_POLY_INT moves into two halves.
24766 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
24767 Split additions that need a temporary here if the destination
24768 is the stack pointer.
24769 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
24770 (*add<mode>3_poly_1): New instruction.
24771 (set_clobber_cc): New expander.
24773 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
24775 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
24776 parameter and use it instead of GET_MODE_SIZE (innermode). Use
24777 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
24778 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
24779 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
24780 Change innermode from fixed_mode_size to machine_mode.
24781 (simplify_subreg): Update call accordingly. Handle a constant-sized
24782 subreg of a variable-length CONST_VECTOR.
24784 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
24785 Alan Hayward <alan.hayward@arm.com>
24786 David Sherwood <david.sherwood@arm.com>
24788 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
24789 (add_offset_to_base): New function, split out from...
24790 (create_mem_ref): ...here. When handling a scale other than 1,
24791 check first whether the address is valid without the offset.
24792 Add it into the base if so, leaving the index and scale as-is.
24794 2018-01-12 Jakub Jelinek <jakub@redhat.com>
24797 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
24798 fold_for_warn before checking if arg2 is INTEGER_CST.
24800 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
24802 * config/rs6000/predicates.md (load_multiple_operation): Delete.
24803 (store_multiple_operation): Delete.
24804 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
24805 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
24806 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
24807 guarded by TARGET_STRING.
24808 (rs6000_output_load_multiple): Delete.
24809 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
24810 OPTION_MASK_STRING / TARGET_STRING handling.
24811 (print_operand) <'N', 'O'>: Add comment that these are unused now.
24812 (const rs6000_opt_masks) <"string">: Change mask to 0.
24813 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
24814 (MASK_STRING): Delete.
24815 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
24817 (load_multiple): Delete.
24824 (store_multiple): Delete.
24831 (movmemsi_8reg): Delete.
24832 (corresponding unnamed define_insn): Delete.
24833 (movmemsi_6reg): Delete.
24834 (corresponding unnamed define_insn): Delete.
24835 (movmemsi_4reg): Delete.
24836 (corresponding unnamed define_insn): Delete.
24837 (movmemsi_2reg): Delete.
24838 (corresponding unnamed define_insn): Delete.
24839 (movmemsi_1reg): Delete.
24840 (corresponding unnamed define_insn): Delete.
24841 * config/rs6000/rs6000.opt (mno-string): New.
24842 (mstring): Replace by deprecation warning stub.
24843 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
24845 2018-01-12 Jakub Jelinek <jakub@redhat.com>
24847 * regrename.c (regrename_do_replace): If replacing the same
24848 reg multiple times, try to reuse last created gen_raw_REG.
24851 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
24852 main to workaround a bug in GDB.
24854 2018-01-12 Tom de Vries <tom@codesourcery.com>
24857 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
24859 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
24861 PR rtl-optimization/80481
24862 * ira-color.c (get_cap_member): New function.
24863 (allocnos_conflict_by_live_ranges_p): Use it.
24864 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
24865 (setup_slot_coalesced_allocno_live_ranges): Ditto.
24867 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
24870 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
24871 (*saddl_se_1): Ditto.
24872 (*ssubsi_1): Ditto.
24873 (*ssubl_se_1): Ditto.
24875 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
24877 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
24878 rather than wi::to_widest for DR_INITs.
24879 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
24880 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
24881 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
24883 (vect_analyze_group_access_1): Note that here.
24885 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
24887 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
24888 polynomial type sizes.
24890 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
24892 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
24893 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
24894 (gimple_add_tmp_var): Likewise.
24896 2018-01-12 Martin Liska <mliska@suse.cz>
24898 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
24899 (gimple_alloc_sizes): Likewise.
24900 (dump_gimple_statistics): Use PRIu64 in printf format.
24901 * gimple.h: Change uint64_t to int.
24903 2018-01-12 Martin Liska <mliska@suse.cz>
24905 * tree-core.h: Use uint64_t instead of int.
24906 * tree.c (tree_node_counts): Likewise.
24907 (tree_node_sizes): Likewise.
24908 (dump_tree_statistics): Use PRIu64 in printf format.
24910 2018-01-12 Martin Liska <mliska@suse.cz>
24912 * Makefile.in: As qsort_chk is implemented in vec.c, add
24913 vec.o to linkage of gencfn-macros.
24914 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
24915 passing the info to record_node_allocation_statistics.
24916 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
24918 * ggc-common.c (struct ggc_usage): Add operator== and use
24919 it in operator< and compare function.
24920 * mem-stats.h (struct mem_usage): Likewise.
24921 * vec.c (struct vec_usage): Remove operator< and compare
24922 function. Can be simply inherited.
24924 2018-01-12 Martin Jambor <mjambor@suse.cz>
24927 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
24928 * tree-ssa-math-opts.c: Include domwalk.h.
24929 (convert_mult_to_fma_1): New function.
24930 (fma_transformation_info): New type.
24931 (fma_deferring_state): Likewise.
24932 (cancel_fma_deferring): New function.
24933 (result_of_phi): Likewise.
24934 (last_fma_candidate_feeds_initial_phi): Likewise.
24935 (convert_mult_to_fma): Added deferring logic, split actual
24936 transformation to convert_mult_to_fma_1.
24937 (math_opts_dom_walker): New type.
24938 (math_opts_dom_walker::after_dom_children): New method, body moved
24939 here from pass_optimize_widening_mul::execute, added deferring logic
24941 (pass_optimize_widening_mul::execute): Moved most of code to
24942 math_opts_dom_walker::after_dom_children.
24943 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
24944 * config/i386/i386.c (ix86_option_override_internal): Added
24945 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
24947 2018-01-12 Richard Biener <rguenther@suse.de>
24950 * dwarf2out.c (gen_variable_die): Do not reset old_die for
24951 inline instance vars.
24953 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
24956 * config/rx/rx.c (rx_is_restricted_memory_address):
24957 Handle SUBREG case.
24959 2018-01-12 Richard Biener <rguenther@suse.de>
24961 PR tree-optimization/80846
24962 * target.def (split_reduction): New target hook.
24963 * targhooks.c (default_split_reduction): New function.
24964 * targhooks.h (default_split_reduction): Declare.
24965 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
24966 target requests first reduce vectors by combining low and high
24968 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
24969 (get_vectype_for_scalar_type_and_size): Export.
24970 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
24971 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
24972 * doc/tm.texi: Regenerate.
24973 * config/i386/i386.c (ix86_split_reduction): Implement
24974 TARGET_VECTORIZE_SPLIT_REDUCTION.
24976 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
24979 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
24980 in PIC mode except for TARGET_VXWORKS_RTP.
24981 * config/sparc/sparc.c: Include cfgrtl.h.
24982 (TARGET_INIT_PIC_REG): Define.
24983 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
24984 (sparc_pic_register_p): New predicate.
24985 (sparc_legitimate_address_p): Use it.
24986 (sparc_legitimize_pic_address): Likewise.
24987 (sparc_delegitimize_address): Likewise.
24988 (sparc_mode_dependent_address_p): Likewise.
24989 (gen_load_pcrel_sym): Remove 4th parameter.
24990 (load_got_register): Adjust call to above. Remove obsolete stuff.
24991 (sparc_expand_prologue): Do not call load_got_register here.
24992 (sparc_flat_expand_prologue): Likewise.
24993 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
24994 (sparc_use_pseudo_pic_reg): New function.
24995 (sparc_init_pic_reg): Likewise.
24996 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
24997 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
24999 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
25001 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
25002 Add item for branch_cost.
25004 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
25006 PR rtl-optimization/83565
25007 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
25008 not extend the result to a larger mode for rotate operations.
25009 (num_sign_bit_copies1): Likewise.
25011 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
25014 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
25016 Use values-Xc.o for -pedantic.
25017 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
25019 2018-01-12 Martin Liska <mliska@suse.cz>
25022 * ipa-devirt.c (final_warning_record::grow_type_warnings):
25024 (possible_polymorphic_call_targets): Use it.
25025 (ipa_devirt): Likewise.
25027 2018-01-12 Martin Liska <mliska@suse.cz>
25029 * profile-count.h (enum profile_quality): Use 0 as invalid
25030 enum value of profile_quality.
25032 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
25034 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
25035 -mext-string options.
25037 2018-01-12 Richard Biener <rguenther@suse.de>
25039 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
25040 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
25041 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
25043 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
25045 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
25047 * configure.ac (--with-long-double-format): Add support for the
25048 configuration option to change the default long double format on
25050 * config.gcc (powerpc*-linux*-*): Likewise.
25051 * configure: Regenerate.
25052 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
25053 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
25054 used without modification.
25056 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
25058 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
25059 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
25060 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
25061 MISC_BUILTIN_SPEC_BARRIER.
25062 (rs6000_init_builtins): Likewise.
25063 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
25065 (speculation_barrier): New define_insn.
25066 * doc/extend.texi: Document __builtin_speculation_barrier.
25068 2018-01-11 Jakub Jelinek <jakub@redhat.com>
25071 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
25072 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
25073 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
25075 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
25076 integral modes instead of "ss" and "sd".
25077 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
25078 vectors with 32-bit and 64-bit elements.
25079 (vecdupssescalarmodesuffix): New mode attribute.
25080 (vec_dup<mode>): Use it.
25082 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
25085 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
25086 frame if argument is passed on stack.
25088 2018-01-11 Jakub Jelinek <jakub@redhat.com>
25091 * ree.c (combine_reaching_defs): Optimize also
25092 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
25093 reg2=any_extend(exp); reg1=reg2;, formatting fix.
25095 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
25097 PR middle-end/83189
25098 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
25100 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
25102 PR middle-end/83718
25103 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
25104 after they are computed.
25106 2018-01-11 Bin Cheng <bin.cheng@arm.com>
25108 PR tree-optimization/83695
25109 * gimple-loop-linterchange.cc
25110 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
25111 reset cached scev information after interchange.
25112 (pass_linterchange::execute): Remove call to scev_reset_htab.
25114 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
25116 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
25117 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
25118 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
25119 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
25120 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
25121 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
25122 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
25123 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
25124 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
25125 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
25126 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
25127 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
25128 (V_lane_reg): Likewise.
25129 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
25131 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
25132 (vfmal_lane_low<mode>_intrinsic,
25133 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
25134 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
25135 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
25136 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
25137 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
25138 vfmsl_lane_high<mode>_intrinsic): New define_insns.
25140 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
25142 * config/arm/arm-cpus.in (fp16fml): New feature.
25143 (ALL_SIMD): Add fp16fml.
25144 (armv8.2-a): Add fp16fml as an option.
25145 (armv8.3-a): Likewise.
25146 (armv8.4-a): Add fp16fml as part of fp16.
25147 * config/arm/arm.h (TARGET_FP16FML): Define.
25148 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
25150 * config/arm/arm-modes.def (V2HF): Define.
25151 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
25152 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
25153 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
25154 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
25155 vfmsl_low, vfmsl_high): New set of builtins.
25156 * config/arm/iterators.md (PLUSMINUS): New code iterator.
25157 (vfml_op): New code attribute.
25158 (VFMLHALVES): New int iterator.
25159 (VFML, VFMLSEL): New mode attributes.
25160 (V_reg): Define mapping for V2HF.
25161 (V_hi, V_lo): New mode attributes.
25162 (VF_constraint): Likewise.
25163 (vfml_half, vfml_half_selector): New int attributes.
25164 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
25166 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
25167 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
25169 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
25170 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
25171 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
25172 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
25174 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
25175 Document new effective target and option set.
25177 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
25179 * config/arm/arm-cpus.in (armv8_4): New feature.
25180 (ARMv8_4a): New fgroup.
25181 (armv8.4-a): New arch.
25182 * config/arm/arm-tables.opt: Regenerate.
25183 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
25184 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
25185 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
25186 Add matching rules for -march=armv8.4-a and extensions.
25187 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
25189 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
25192 * config/rx/rx.md (BW): New mode attribute.
25193 (sync_lock_test_and_setsi): Add mode suffix to insn output.
25195 2018-01-11 Richard Biener <rguenther@suse.de>
25197 PR tree-optimization/83435
25198 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
25199 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
25200 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
25202 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
25203 Alan Hayward <alan.hayward@arm.com>
25204 David Sherwood <david.sherwood@arm.com>
25206 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
25208 (aarch64_classify_address): Initialize it. Track polynomial offsets.
25209 (aarch64_print_address_internal): Use it to check for a zero offset.
25211 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
25212 Alan Hayward <alan.hayward@arm.com>
25213 David Sherwood <david.sherwood@arm.com>
25215 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
25216 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
25217 Return a poly_int64 rather than a HOST_WIDE_INT.
25218 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
25219 rather than a HOST_WIDE_INT.
25220 * config/aarch64/aarch64.h (aarch64_frame): Protect with
25221 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
25222 hard_fp_offset, frame_size, initial_adjust, callee_offset and
25223 final_offset from HOST_WIDE_INT to poly_int64.
25224 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
25225 to_constant when getting the number of units in an Advanced SIMD
25227 (aarch64_builtin_vectorized_function): Check for a constant number
25229 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
25231 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
25232 attribute instead of GET_MODE_NUNITS.
25233 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
25234 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
25235 GET_MODE_SIZE for fixed-size registers.
25236 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
25237 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
25238 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
25239 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
25240 (aarch64_print_operand, aarch64_print_address_internal)
25241 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
25242 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
25243 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
25244 Handle polynomial GET_MODE_SIZE.
25245 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
25246 wider than SImode without modification.
25247 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
25248 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
25249 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
25250 passing and returning SVE modes.
25251 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
25252 rather than GEN_INT.
25253 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
25254 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
25255 (aarch64_allocate_and_probe_stack_space): Likewise.
25256 (aarch64_layout_frame): Cope with polynomial offsets.
25257 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
25258 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
25259 polynomial offsets.
25260 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
25261 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
25262 poly_int64 rather than a HOST_WIDE_INT.
25263 (aarch64_get_separate_components, aarch64_process_components)
25264 (aarch64_expand_prologue, aarch64_expand_epilogue)
25265 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
25266 (aarch64_anchor_offset): New function, split out from...
25267 (aarch64_legitimize_address): ...here.
25268 (aarch64_builtin_vectorization_cost): Handle polynomial
25269 TYPE_VECTOR_SUBPARTS.
25270 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
25272 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
25273 number of elements from the PARALLEL rather than the mode.
25274 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
25275 rather than GET_MODE_BITSIZE.
25276 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
25277 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
25278 (aarch64_expand_vec_perm_const_1): Handle polynomial
25279 d->perm.length () and d->perm elements.
25280 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
25281 Apply to_constant to d->perm elements.
25282 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
25283 polynomial CONST_VECTOR_NUNITS.
25284 (aarch64_move_pointer): Take amount as a poly_int64 rather
25286 (aarch64_progress_pointer): Avoid temporary variable.
25287 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
25288 the mode attribute instead of GET_MODE.
25290 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
25291 Alan Hayward <alan.hayward@arm.com>
25292 David Sherwood <david.sherwood@arm.com>
25294 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
25295 x exists before using it.
25296 (aarch64_add_constant_internal): Rename to...
25297 (aarch64_add_offset_1): ...this. Replace regnum with separate
25298 src and dest rtxes. Handle the case in which they're different,
25299 including when the offset is zero. Replace scratchreg with an rtx.
25300 Use 2 additions if there is no spare register into which we can
25301 move a 16-bit constant.
25302 (aarch64_add_constant): Delete.
25303 (aarch64_add_offset): Replace reg with separate src and dest
25304 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
25305 Use aarch64_add_offset_1.
25306 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
25307 an rtx rather than an int. Take the delta as a poly_int64
25308 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
25309 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
25310 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
25311 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
25312 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
25313 and aarch64_add_sp.
25314 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
25315 aarch64_add_constant.
25317 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
25319 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
25320 Use scalar_float_mode.
25322 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
25324 * config/aarch64/aarch64-simd.md
25325 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
25326 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
25327 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
25328 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
25329 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
25330 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
25331 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
25332 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
25333 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
25334 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
25336 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
25339 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
25340 targ_options->x_arm_arch_string is non NULL.
25342 2018-01-11 Tamar Christina <tamar.christina@arm.com>
25344 * config/aarch64/aarch64.h
25345 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
25347 2018-01-11 Sudakshina Das <sudi.das@arm.com>
25350 * expmed.c (emit_store_flag_force): Swap if const op0
25351 and change VOIDmode to mode of op0.
25353 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
25355 PR rtl-optimization/83761
25356 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
25357 than bytes to mode_for_size.
25359 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
25361 PR middle-end/83189
25362 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
25363 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
25366 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
25368 PR middle-end/83575
25369 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
25370 when in layout mode.
25371 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
25372 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
25375 2018-01-10 Michael Collison <michael.collison@arm.com>
25377 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
25378 * config/aarch64/aarch64-option-extension.def: Add
25379 AARCH64_OPT_EXTENSION of 'fp16fml'.
25380 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
25381 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
25382 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
25383 * config/aarch64/constraints.md (Ui7): New constraint.
25384 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
25385 (VFMLA_SEL_W): Ditto.
25388 (VFMLA16_LOW): New int iterator.
25389 (VFMLA16_HIGH): Ditto.
25390 (UNSPEC_FMLAL): New unspec.
25391 (UNSPEC_FMLSL): Ditto.
25392 (UNSPEC_FMLAL2): Ditto.
25393 (UNSPEC_FMLSL2): Ditto.
25394 (f16mac): New code attribute.
25395 * config/aarch64/aarch64-simd-builtins.def
25396 (aarch64_fmlal_lowv2sf): Ditto.
25397 (aarch64_fmlsl_lowv2sf): Ditto.
25398 (aarch64_fmlalq_lowv4sf): Ditto.
25399 (aarch64_fmlslq_lowv4sf): Ditto.
25400 (aarch64_fmlal_highv2sf): Ditto.
25401 (aarch64_fmlsl_highv2sf): Ditto.
25402 (aarch64_fmlalq_highv4sf): Ditto.
25403 (aarch64_fmlslq_highv4sf): Ditto.
25404 (aarch64_fmlal_lane_lowv2sf): Ditto.
25405 (aarch64_fmlsl_lane_lowv2sf): Ditto.
25406 (aarch64_fmlal_laneq_lowv2sf): Ditto.
25407 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
25408 (aarch64_fmlalq_lane_lowv4sf): Ditto.
25409 (aarch64_fmlsl_lane_lowv4sf): Ditto.
25410 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
25411 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
25412 (aarch64_fmlal_lane_highv2sf): Ditto.
25413 (aarch64_fmlsl_lane_highv2sf): Ditto.
25414 (aarch64_fmlal_laneq_highv2sf): Ditto.
25415 (aarch64_fmlsl_laneq_highv2sf): Ditto.
25416 (aarch64_fmlalq_lane_highv4sf): Ditto.
25417 (aarch64_fmlsl_lane_highv4sf): Ditto.
25418 (aarch64_fmlalq_laneq_highv4sf): Ditto.
25419 (aarch64_fmlsl_laneq_highv4sf): Ditto.
25420 * config/aarch64/aarch64-simd.md:
25421 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
25422 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
25423 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
25424 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
25425 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
25426 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
25427 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
25428 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
25429 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
25430 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
25431 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
25432 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
25433 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
25434 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
25435 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
25436 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
25437 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
25438 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
25439 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
25440 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
25441 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
25442 (vfmlsl_low_u32): Ditto.
25443 (vfmlalq_low_u32): Ditto.
25444 (vfmlslq_low_u32): Ditto.
25445 (vfmlal_high_u32): Ditto.
25446 (vfmlsl_high_u32): Ditto.
25447 (vfmlalq_high_u32): Ditto.
25448 (vfmlslq_high_u32): Ditto.
25449 (vfmlal_lane_low_u32): Ditto.
25450 (vfmlsl_lane_low_u32): Ditto.
25451 (vfmlal_laneq_low_u32): Ditto.
25452 (vfmlsl_laneq_low_u32): Ditto.
25453 (vfmlalq_lane_low_u32): Ditto.
25454 (vfmlslq_lane_low_u32): Ditto.
25455 (vfmlalq_laneq_low_u32): Ditto.
25456 (vfmlslq_laneq_low_u32): Ditto.
25457 (vfmlal_lane_high_u32): Ditto.
25458 (vfmlsl_lane_high_u32): Ditto.
25459 (vfmlal_laneq_high_u32): Ditto.
25460 (vfmlsl_laneq_high_u32): Ditto.
25461 (vfmlalq_lane_high_u32): Ditto.
25462 (vfmlslq_lane_high_u32): Ditto.
25463 (vfmlalq_laneq_high_u32): Ditto.
25464 (vfmlslq_laneq_high_u32): Ditto.
25465 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
25466 (AARCH64_FL_FOR_ARCH8_4): New.
25467 (AARCH64_ISA_F16FML): New ISA flag.
25468 (TARGET_F16FML): New feature flag for fp16fml.
25469 (doc/invoke.texi): Document new fp16fml option.
25471 2018-01-10 Michael Collison <michael.collison@arm.com>
25473 * config/aarch64/aarch64-builtins.c:
25474 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
25475 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
25476 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
25477 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
25478 (AARCH64_ISA_SHA3): New ISA flag.
25479 (TARGET_SHA3): New feature flag for sha3.
25480 * config/aarch64/iterators.md (sha512_op): New int attribute.
25481 (CRYPTO_SHA512): New int iterator.
25482 (UNSPEC_SHA512H): New unspec.
25483 (UNSPEC_SHA512H2): Ditto.
25484 (UNSPEC_SHA512SU0): Ditto.
25485 (UNSPEC_SHA512SU1): Ditto.
25486 * config/aarch64/aarch64-simd-builtins.def
25487 (aarch64_crypto_sha512hqv2di): New builtin.
25488 (aarch64_crypto_sha512h2qv2di): Ditto.
25489 (aarch64_crypto_sha512su0qv2di): Ditto.
25490 (aarch64_crypto_sha512su1qv2di): Ditto.
25491 (aarch64_eor3qv8hi): Ditto.
25492 (aarch64_rax1qv2di): Ditto.
25493 (aarch64_xarqv2di): Ditto.
25494 (aarch64_bcaxqv8hi): Ditto.
25495 * config/aarch64/aarch64-simd.md:
25496 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
25497 (aarch64_crypto_sha512su0qv2di): Ditto.
25498 (aarch64_crypto_sha512su1qv2di): Ditto.
25499 (aarch64_eor3qv8hi): Ditto.
25500 (aarch64_rax1qv2di): Ditto.
25501 (aarch64_xarqv2di): Ditto.
25502 (aarch64_bcaxqv8hi): Ditto.
25503 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
25504 (vsha512h2q_u64): Ditto.
25505 (vsha512su0q_u64): Ditto.
25506 (vsha512su1q_u64): Ditto.
25507 (veor3q_u16): Ditto.
25508 (vrax1q_u64): Ditto.
25509 (vxarq_u64): Ditto.
25510 (vbcaxq_u16): Ditto.
25511 * config/arm/types.md (crypto_sha512): New type attribute.
25512 (crypto_sha3): Ditto.
25513 (doc/invoke.texi): Document new sha3 option.
25515 2018-01-10 Michael Collison <michael.collison@arm.com>
25517 * config/aarch64/aarch64-builtins.c:
25518 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
25519 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
25520 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
25521 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
25522 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
25523 (AARCH64_ISA_SM4): New ISA flag.
25524 (TARGET_SM4): New feature flag for sm4.
25525 * config/aarch64/aarch64-simd-builtins.def
25526 (aarch64_sm3ss1qv4si): Ditto.
25527 (aarch64_sm3tt1aq4si): Ditto.
25528 (aarch64_sm3tt1bq4si): Ditto.
25529 (aarch64_sm3tt2aq4si): Ditto.
25530 (aarch64_sm3tt2bq4si): Ditto.
25531 (aarch64_sm3partw1qv4si): Ditto.
25532 (aarch64_sm3partw2qv4si): Ditto.
25533 (aarch64_sm4eqv4si): Ditto.
25534 (aarch64_sm4ekeyqv4si): Ditto.
25535 * config/aarch64/aarch64-simd.md:
25536 (aarch64_sm3ss1qv4si): Ditto.
25537 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
25538 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
25539 (aarch64_sm4eqv4si): Ditto.
25540 (aarch64_sm4ekeyqv4si): Ditto.
25541 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
25542 (sm3part_op): Ditto.
25543 (CRYPTO_SM3TT): Ditto.
25544 (CRYPTO_SM3PART): Ditto.
25545 (UNSPEC_SM3SS1): New unspec.
25546 (UNSPEC_SM3TT1A): Ditto.
25547 (UNSPEC_SM3TT1B): Ditto.
25548 (UNSPEC_SM3TT2A): Ditto.
25549 (UNSPEC_SM3TT2B): Ditto.
25550 (UNSPEC_SM3PARTW1): Ditto.
25551 (UNSPEC_SM3PARTW2): Ditto.
25552 (UNSPEC_SM4E): Ditto.
25553 (UNSPEC_SM4EKEY): Ditto.
25554 * config/aarch64/constraints.md (Ui2): New constraint.
25555 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
25556 * config/arm/types.md (crypto_sm3): New type attribute.
25557 (crypto_sm4): Ditto.
25558 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
25559 (vsm3tt1aq_u32): Ditto.
25560 (vsm3tt1bq_u32): Ditto.
25561 (vsm3tt2aq_u32): Ditto.
25562 (vsm3tt2bq_u32): Ditto.
25563 (vsm3partw1q_u32): Ditto.
25564 (vsm3partw2q_u32): Ditto.
25565 (vsm4eq_u32): Ditto.
25566 (vsm4ekeyq_u32): Ditto.
25567 (doc/invoke.texi): Document new sm4 option.
25569 2018-01-10 Michael Collison <michael.collison@arm.com>
25571 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
25572 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
25573 (AARCH64_FL_FOR_ARCH8_4): New.
25574 (AARCH64_FL_V8_4): New flag.
25575 (doc/invoke.texi): Document new armv8.4-a option.
25577 2018-01-10 Michael Collison <michael.collison@arm.com>
25579 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
25580 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
25581 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
25582 * config/aarch64/aarch64-option-extension.def: Add
25583 AARCH64_OPT_EXTENSION of 'sha2'.
25584 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
25585 (crypto): Disable sha2 and aes if crypto disabled.
25586 (crypto): Enable aes and sha2 if enabled.
25587 (simd): Disable sha2 and aes if simd disabled.
25588 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
25590 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
25591 (TARGET_SHA2): New feature flag for sha2.
25592 (TARGET_AES): New feature flag for aes.
25593 * config/aarch64/aarch64-simd.md:
25594 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
25595 conditional on TARGET_AES.
25596 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
25597 (aarch64_crypto_sha1hsi): Make pattern conditional
25599 (aarch64_crypto_sha1hv4si): Ditto.
25600 (aarch64_be_crypto_sha1hv4si): Ditto.
25601 (aarch64_crypto_sha1su1v4si): Ditto.
25602 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
25603 (aarch64_crypto_sha1su0v4si): Ditto.
25604 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
25605 (aarch64_crypto_sha256su0v4si): Ditto.
25606 (aarch64_crypto_sha256su1v4si): Ditto.
25607 (doc/invoke.texi): Document new aes and sha2 options.
25609 2018-01-10 Martin Sebor <msebor@redhat.com>
25611 PR tree-optimization/83781
25612 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
25615 2018-01-11 Martin Sebor <msebor@gmail.com>
25616 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
25618 PR tree-optimization/83501
25619 PR tree-optimization/81703
25621 * tree-ssa-strlen.c (get_string_cst): Rename...
25622 (get_string_len): ...to this. Handle global constants.
25623 (handle_char_store): Adjust.
25625 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
25626 Jim Wilson <jimw@sifive.com>
25628 * config/riscv/riscv-protos.h (riscv_output_return): New.
25629 * config/riscv/riscv.c (struct machine_function): New naked_p field.
25630 (riscv_attribute_table, riscv_output_return),
25631 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
25632 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
25633 (riscv_compute_frame_info): Only compute frame->mask if not a naked
25635 (riscv_expand_prologue): Add early return for naked function.
25636 (riscv_expand_epilogue): Likewise.
25637 (riscv_function_ok_for_sibcall): Return false for naked function.
25638 (riscv_set_current_function): New.
25639 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
25640 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
25641 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
25642 * doc/extend.texi (RISC-V Function Attributes): New.
25644 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
25646 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
25647 check for 128-bit long double before checking TCmode.
25648 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
25649 128-bit long doubles before checking TFmode or TCmode.
25650 (FLOAT128_IBM_P): Likewise.
25652 2018-01-10 Martin Sebor <msebor@redhat.com>
25654 PR tree-optimization/83671
25655 * builtins.c (c_strlen): Unconditionally return zero for the empty
25657 Use -Warray-bounds for warnings.
25658 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
25659 for non-constant array indices with COMPONENT_REF, arrays of
25660 arrays, and pointers to arrays.
25661 (gimple_fold_builtin_strlen): Determine and set length range for
25662 non-constant character arrays.
25664 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
25666 PR middle-end/81897
25667 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
25670 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
25672 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
25674 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
25677 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
25678 VECTOR_MEM_ALTIVEC_OR_VSX_P.
25679 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
25680 indexed_or_indirect_operand predicate.
25681 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
25682 (*vsx_le_perm_load_v8hi): Likewise.
25683 (*vsx_le_perm_load_v16qi): Likewise.
25684 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
25685 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
25686 (*vsx_le_perm_store_v8hi): Likewise.
25687 (*vsx_le_perm_store_v16qi): Likewise.
25688 (eight unnamed splitters): Likewise.
25690 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
25692 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
25693 * config/rs6000/emmintrin.h: Likewise.
25694 * config/rs6000/mmintrin.h: Likewise.
25695 * config/rs6000/xmmintrin.h: Likewise.
25697 2018-01-10 David Malcolm <dmalcolm@redhat.com>
25700 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
25702 * tree.c (tree_nop_conversion): Return true for location wrapper
25704 (maybe_wrap_with_location): New function.
25705 (selftest::check_strip_nops): New function.
25706 (selftest::test_location_wrappers): New function.
25707 (selftest::tree_c_tests): Call it.
25708 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
25709 (maybe_wrap_with_location): New decl.
25710 (EXPR_LOCATION_WRAPPER_P): New macro.
25711 (location_wrapper_p): New inline function.
25712 (tree_strip_any_location_wrapper): New inline function.
25714 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
25717 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
25718 stack_realign_offset for the largest alignment of stack slot
25720 (ix86_find_max_used_stack_alignment): New function.
25721 (ix86_finalize_stack_frame_flags): Use it. Set
25722 max_used_stack_alignment if we don't realign stack.
25723 * config/i386/i386.h (machine_function): Add
25724 max_used_stack_alignment.
25726 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
25728 * config/arm/arm.opt (-mbranch-cost): New option.
25729 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
25732 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
25735 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
25736 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
25738 2018-01-10 Richard Biener <rguenther@suse.de>
25741 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
25742 early out so it also covers the case where we have a non-NULL
25745 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
25747 PR tree-optimization/83753
25748 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
25749 for non-strided grouped accesses if the number of elements is 1.
25751 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
25754 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
25755 * i386.h (TARGET_USE_GATHER): Define.
25756 * x86-tune.def (X86_TUNE_USE_GATHER): New.
25758 2018-01-10 Martin Liska <mliska@suse.cz>
25761 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
25762 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
25764 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
25765 CLEANUP_NO_PARTITIONING is not set.
25767 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
25769 * doc/rtl.texi: Remove documentation of (const ...) wrappers
25770 for vectors, as a partial revert of r254296.
25771 * rtl.h (const_vec_p): Delete.
25772 (const_vec_duplicate_p): Don't test for vector CONSTs.
25773 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
25774 * expmed.c (make_tree): Likewise.
25777 * common.md (E, F): Use CONSTANT_P instead of checking for
25779 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
25780 checking for CONST_VECTOR.
25782 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
25784 PR middle-end/83575
25785 * predict.c (force_edge_cold): Handle in more sane way edges
25786 with no prediction.
25788 2018-01-09 Carl Love <cel@us.ibm.com>
25790 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
25792 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
25793 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
25794 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
25795 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
25796 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
25797 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
25798 * config/rs6000/rs6000-protos.h: Add extern defition for
25799 rs6000_generate_float2_double_code.
25800 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
25802 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
25803 (float2_v2df): Add define_expand.
25805 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
25808 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
25809 op_mode in the force_to_mode call.
25811 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
25813 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
25814 instead of checking each element individually.
25815 (aarch64_evpc_uzp): Likewise.
25816 (aarch64_evpc_zip): Likewise.
25817 (aarch64_evpc_ext): Likewise.
25818 (aarch64_evpc_rev): Likewise.
25819 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
25820 instead of checking each element individually. Return true without
25822 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
25823 whether all selected elements come from the same input, instead of
25824 checking each element individually. Remove calls to gen_rtx_REG,
25825 start_sequence and end_sequence and instead assert that no rtl is
25828 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
25830 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
25831 order of HIGH and CONST checks.
25833 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
25835 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
25836 if the destination isn't an SSA_NAME.
25838 2018-01-09 Richard Biener <rguenther@suse.de>
25840 PR tree-optimization/83668
25841 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
25843 (canonicalize_loop_form): ... here, renamed from ...
25844 (canonicalize_loop_closed_ssa_form): ... this and amended to
25845 swap successor edges for loop exit blocks to make us use
25846 the RPO order we need for initial schedule generation.
25848 2018-01-09 Joseph Myers <joseph@codesourcery.com>
25850 PR tree-optimization/64811
25851 * match.pd: When optimizing comparisons with Inf, avoid
25852 introducing or losing exceptions from comparisons with NaN.
25854 2018-01-09 Martin Liska <mliska@suse.cz>
25857 * asan.c (shadow_mem_size): Add gcc_assert.
25859 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
25861 Don't save registers in main().
25864 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
25865 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
25866 * config/avr/avr.c (avr_set_current_function): Don't error if
25867 naked, OS_task or OS_main are specified at the same time.
25868 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
25870 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
25872 * common/config/avr/avr-common.c (avr_option_optimization_table):
25873 Switch on -mmain-is-OS_task for optimizing compilations.
25875 2018-01-09 Richard Biener <rguenther@suse.de>
25877 PR tree-optimization/83572
25878 * graphite.c: Include cfganal.h.
25879 (graphite_transform_loops): Connect infinite loops to exit
25880 and remove fake edges at the end.
25882 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
25884 * ipa-inline.c (edge_badness): Revert accidental checkin.
25886 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
25889 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
25890 symbols; not inline clones.
25892 2018-01-09 Jakub Jelinek <jakub@redhat.com>
25895 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
25896 hard registers. Formatting fixes.
25898 PR preprocessor/83722
25899 * gcc.c (try_generate_repro): Pass
25900 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
25901 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
25904 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
25905 Kito Cheng <kito.cheng@gmail.com>
25907 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
25908 (riscv_leaf_function_p): Delete.
25909 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
25911 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
25913 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
25915 (do_ifelse): New function.
25916 (do_isel): New function.
25917 (do_sub3): New function.
25918 (do_add3): New function.
25919 (do_load_mask_compare): New function.
25920 (do_overlap_load_compare): New function.
25921 (expand_compare_loop): New function.
25922 (expand_block_compare): Call expand_compare_loop() when appropriate.
25923 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
25924 option description.
25925 (-mblock-compare-inline-loop-limit): New option.
25927 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
25930 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
25931 Reverse order of second and third operands in first alternative.
25932 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
25933 of first and second elements in UNSPEC_VPERMR vector.
25934 (altivec_expand_vec_perm_le): Likewise.
25936 2018-01-08 Jeff Law <law@redhat.com>
25938 PR rtl-optimizatin/81308
25939 * tree-switch-conversion.c (cfg_altered): New file scoped static.
25940 (process_switch): If group_case_labels makes a change, then set
25942 (pass_convert_switch::execute): If a switch is converted, then
25943 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
25945 PR rtl-optimization/81308
25946 * recog.c (split_all_insns): Conditionally cleanup the CFG after
25949 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
25951 PR target/83663 - Revert r255946
25952 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
25953 generation for cases where splatting a value is not useful.
25954 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
25955 across a vec_duplicate and a paradoxical subreg forming a vector
25956 mode to a vec_concat.
25958 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
25960 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
25961 -march=armv8.3-a variants.
25962 * config/arm/t-multilib: Likewise.
25963 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
25965 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
25967 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
25969 (cceq_ior_compare_complement): Give it a name so I can use it, and
25970 change boolean_or_operator predicate to boolean_operator so it can
25971 be used to generate a crand.
25972 (eqne): New code iterator.
25973 (bd/bd_neg): New code_attrs.
25974 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
25975 a single define_insn.
25976 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
25977 decrement (bdnzt/bdnzf/bdzt/bdzf).
25978 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
25979 with the new names of the branch decrement patterns, and added the
25980 names of the branch decrement conditional patterns.
25982 2018-01-08 Richard Biener <rguenther@suse.de>
25984 PR tree-optimization/83563
25985 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
25988 2018-01-08 Richard Biener <rguenther@suse.de>
25990 PR middle-end/83713
25991 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
25993 2018-01-08 Richard Biener <rguenther@suse.de>
25995 PR tree-optimization/83685
25996 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
25997 references to abnormals.
25999 2018-01-08 Richard Biener <rguenther@suse.de>
26002 * dwarf2out.c (output_indirect_strings): Handle empty
26003 skeleton_debug_str_hash.
26004 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
26006 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
26008 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
26009 (emit_store_direct): Likewise.
26010 (arc_trampoline_adjust_address): Likewise.
26011 (arc_asm_trampoline_template): New function.
26012 (arc_initialize_trampoline): Use asm_trampoline_template.
26013 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
26014 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
26015 * config/arc/arc.md (flush_icache): Delete pattern.
26017 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
26019 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
26020 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
26023 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
26026 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
26027 by not USED_FOR_TARGET.
26028 (make_pass_resolve_sw_modes): Likewise.
26030 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
26032 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
26035 2018-01-08 Richard Biener <rguenther@suse.de>
26037 PR middle-end/83580
26038 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
26040 2018-01-08 Richard Biener <rguenther@suse.de>
26042 PR middle-end/83517
26043 * match.pd ((t * 2) / 2) -> t): Add missing :c.
26045 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
26047 PR middle-end/81897
26048 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
26049 basic blocks with a small number of successors.
26050 (convert_control_dep_chain_into_preds): Improve handling of
26052 (dump_predicates): Split apart into...
26053 (dump_pred_chain): ...here...
26054 (dump_pred_info): ...and here.
26055 (can_one_predicate_be_invalidated_p): Add debugging printfs.
26056 (can_chain_union_be_invalidated_p): Improve check for invalidation
26058 (uninit_uses_cannot_happen): Avoid unnecessary if
26059 convert_control_dep_chain_into_preds yielded nothing.
26061 2018-01-06 Martin Sebor <msebor@redhat.com>
26063 PR tree-optimization/83640
26064 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
26065 subtracting negative offset from size.
26066 (builtin_access::overlap): Adjust offset bounds of the access to fall
26067 within the size of the object if possible.
26069 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
26071 PR rtl-optimization/83699
26072 * expmed.c (extract_bit_field_1): Restrict the vector usage of
26073 extract_bit_field_as_subreg to cases in which the extracted
26074 value is also a vector.
26076 * lra-constraints.c (process_alt_operands): Test for the equivalence
26077 substitutions when detecting a possible reload cycle.
26079 2018-01-06 Jakub Jelinek <jakub@redhat.com>
26082 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
26083 by default if flag_selective_schedling{,2}. Formatting fixes.
26085 PR rtl-optimization/83682
26086 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
26087 if it has non-VECTOR_MODE element mode.
26088 (vec_duplicate_p): Likewise.
26090 PR middle-end/83694
26091 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
26092 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
26094 2018-01-05 Jakub Jelinek <jakub@redhat.com>
26097 * config/i386/i386-builtin.def
26098 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
26099 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
26100 Require also OPTION_MASK_ISA_AVX512F in addition to
26101 OPTION_MASK_ISA_GFNI.
26102 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
26103 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
26104 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
26105 to OPTION_MASK_ISA_GFNI.
26106 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
26107 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
26108 OPTION_MASK_ISA_AVX512BW.
26109 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
26110 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
26111 addition to OPTION_MASK_ISA_GFNI.
26112 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
26113 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
26114 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
26115 to OPTION_MASK_ISA_GFNI.
26116 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
26117 a requirement for all ISAs rather than any of them with a few
26119 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
26121 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
26122 bitmasks to be enabled with 3 exceptions, instead of requiring any
26123 enabled ISA with lots of exceptions.
26124 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
26125 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
26126 Change avx512bw in isa attribute to avx512f.
26127 * config/i386/sgxintrin.h: Add license boilerplate.
26128 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
26129 to __AVX512F__ and __AVX512VL to __AVX512VL__.
26130 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
26131 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
26133 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
26134 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
26135 temporarily sse2 rather than sse if not enabled already.
26138 * config/i386/sse.md (VI248_VLBW): Rename to ...
26139 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
26140 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
26141 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
26142 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
26143 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
26144 mode iterator instead of VI248_VLBW.
26146 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
26148 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
26149 (record_modified): Skip clobbers; add debug output.
26150 (param_change_prob): Use sreal frequencies.
26152 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
26154 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
26155 punt for user-aligned variables.
26157 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
26159 * tree-chrec.c (chrec_contains_symbols): Return true for
26162 2018-01-05 Sudakshina Das <sudi.das@arm.com>
26165 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
26166 of (x|y) == x for BICS pattern.
26168 2018-01-05 Jakub Jelinek <jakub@redhat.com>
26170 PR tree-optimization/83605
26171 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
26172 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
26175 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
26177 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
26178 * config/epiphany/rtems.h: New file.
26180 2018-01-04 Jakub Jelinek <jakub@redhat.com>
26181 Uros Bizjak <ubizjak@gmail.com>
26184 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
26185 QIreg_operand instead of register_operand predicate.
26186 * config/i386/i386.c (ix86_rop_should_change_byte_p,
26187 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
26188 comments instead of -fmitigate[-_]rop.
26190 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
26193 * cgraphunit.c (symbol_table::compile): Switch to text_section
26194 before calling assembly_start debug hook.
26195 * run-rtl-passes.c (run_rtl_passes): Likewise.
26198 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
26200 * tree-vrp.c (extract_range_from_binary_expr_1): Check
26201 range_int_cst_p rather than !symbolic_range_p before calling
26202 extract_range_from_multiplicative_op_1.
26204 2018-01-04 Jeff Law <law@redhat.com>
26206 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
26207 redundant test in assertion.
26209 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
26211 * doc/rtl.texi: Document machine_mode wrapper classes.
26213 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
26215 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
26216 using tree_to_uhwi.
26218 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
26220 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
26221 the VEC_PERM_EXPR fold to fail.
26223 2018-01-04 Jakub Jelinek <jakub@redhat.com>
26226 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
26227 to switched_sections.
26229 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
26232 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
26233 test for d.testing.
26235 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
26238 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
26239 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
26241 2018-01-04 Jakub Jelinek <jakub@redhat.com>
26244 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
26245 is BLKmode and bitpos not zero or mode change is needed.
26247 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
26250 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
26253 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
26256 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
26257 instead of MULT rtx. Update all corresponding splitters.
26258 (*saddl_se): Ditto.
26259 (*ssub<modesuffix>): Ditto.
26260 (*ssubl_se): Ditto.
26261 (*cmp_sadd_di): Update split patterns.
26262 (*cmp_sadd_si): Ditto.
26263 (*cmp_sadd_sidi): Ditto.
26264 (*cmp_ssub_di): Ditto.
26265 (*cmp_ssub_si): Ditto.
26266 (*cmp_ssub_sidi): Ditto.
26267 * config/alpha/predicates.md (const23_operand): New predicate.
26268 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
26269 Look for ASHIFT, not MULT inner operand.
26270 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
26272 2018-01-04 Martin Liska <mliska@suse.cz>
26274 PR gcov-profile/83669
26275 * gcov.c (output_intermediate_file): Add version to intermediate
26277 * doc/gcov.texi: Document new field 'version' in intermediate
26278 file format. Fix location of '-k' option of gcov command.
26280 2018-01-04 Martin Liska <mliska@suse.cz>
26283 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
26285 2018-01-04 Jakub Jelinek <jakub@redhat.com>
26287 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
26289 2018-01-03 Martin Sebor <msebor@redhat.com>
26291 PR tree-optimization/83655
26292 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
26293 checking calls with invalid arguments.
26295 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26297 * tree-vect-stmts.c (vect_get_store_rhs): New function.
26298 (vectorizable_mask_load_store): Delete.
26299 (vectorizable_call): Return false for masked loads and stores.
26300 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
26301 instead of gimple_assign_rhs1.
26302 (vectorizable_load): Handle IFN_MASK_LOAD.
26303 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
26305 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26307 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
26309 (vectorizable_mask_load_store): ...here.
26310 (vectorizable_load): ...and here.
26312 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26314 * tree-vect-stmts.c (vect_build_all_ones_mask)
26315 (vect_build_zero_merge_argument): New functions, split out from...
26316 (vectorizable_load): ...here.
26318 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26320 * tree-vect-stmts.c (vect_check_store_rhs): New function,
26322 (vectorizable_mask_load_store): ...here.
26323 (vectorizable_store): ...and here.
26325 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26327 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
26329 (vectorizable_mask_load_store): ...here.
26331 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26333 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
26334 (vect_model_store_cost): Take a vec_load_store_type instead of a
26336 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
26337 (vect_model_store_cost): Take a vec_load_store_type instead of a
26339 (vectorizable_mask_load_store): Update accordingly.
26340 (vectorizable_store): Likewise.
26341 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
26343 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26345 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
26346 IFN_MASK_LOAD calls here rather than...
26347 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
26349 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26350 Alan Hayward <alan.hayward@arm.com>
26351 David Sherwood <david.sherwood@arm.com>
26353 * expmed.c (extract_bit_field_1): For vector extracts,
26354 fall back to extract_bit_field_as_subreg if vec_extract
26357 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26358 Alan Hayward <alan.hayward@arm.com>
26359 David Sherwood <david.sherwood@arm.com>
26361 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
26362 they are variable or constant sized.
26363 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
26364 slots for constant-sized data.
26366 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26367 Alan Hayward <alan.hayward@arm.com>
26368 David Sherwood <david.sherwood@arm.com>
26370 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
26371 handling COND_EXPRs with boolean comparisons, try to find a better
26372 basis for the mask type than the boolean itself.
26374 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26376 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
26377 is calculated and how it can be overridden.
26378 * genmodes.c (max_bitsize_mode_any_mode): New variable.
26379 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
26381 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
26384 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26385 Alan Hayward <alan.hayward@arm.com>
26386 David Sherwood <david.sherwood@arm.com>
26388 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
26389 Remove the mode argument.
26390 (aarch64_simd_valid_immediate): Remove the mode and inverse
26392 * config/aarch64/iterators.md (bitsize): New iterator.
26393 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
26394 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
26395 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
26396 aarch64_simd_valid_immediate.
26397 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
26398 (aarch64_reg_or_bic_imm): Likewise.
26399 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
26400 with an insn_type enum and msl with a modifier_type enum.
26401 Replace element_width with a scalar_mode. Change the shift
26402 to unsigned int. Add constructors for scalar_float_mode and
26403 scalar_int_mode elements.
26404 (aarch64_vect_float_const_representable_p): Delete.
26405 (aarch64_can_const_movi_rtx_p)
26406 (aarch64_simd_scalar_immediate_valid_for_move)
26407 (aarch64_simd_make_constant): Update call to
26408 aarch64_simd_valid_immediate.
26409 (aarch64_advsimd_valid_immediate_hs): New function.
26410 (aarch64_advsimd_valid_immediate): Likewise.
26411 (aarch64_simd_valid_immediate): Remove mode and inverse
26412 arguments. Rewrite to use the above. Use const_vec_duplicate_p
26413 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
26414 and aarch64_float_const_representable_p on the result.
26415 (aarch64_output_simd_mov_immediate): Remove mode argument.
26416 Update call to aarch64_simd_valid_immediate and use of
26417 simd_immediate_info.
26418 (aarch64_output_scalar_simd_mov_immediate): Update call
26421 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26422 Alan Hayward <alan.hayward@arm.com>
26423 David Sherwood <david.sherwood@arm.com>
26425 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
26426 (mode_nunits): Likewise CONST_MODE_NUNITS.
26427 * machmode.def (ADJUST_NUNITS): Document.
26428 * genmodes.c (mode_data::need_nunits_adj): New field.
26429 (blank_mode): Update accordingly.
26430 (adj_nunits): New variable.
26431 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
26433 (emit_mode_size_inline): Set need_bytesize_adj for all modes
26434 listed in adj_nunits.
26435 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
26436 listed in adj_nunits. Don't emit case statements for such modes.
26437 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
26438 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
26439 nothing if adj_nunits is nonnull.
26440 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
26441 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
26442 (emit_mode_fbit): Update use of print_maybe_const_decl.
26443 (emit_move_size): Likewise. Treat the array as non-const
26445 (emit_mode_adjustments): Handle adj_nunits.
26447 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26449 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
26450 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
26451 (VECTOR_MODES): Use it.
26452 (make_vector_modes): Take the prefix as an argument.
26454 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26455 Alan Hayward <alan.hayward@arm.com>
26456 David Sherwood <david.sherwood@arm.com>
26458 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
26459 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
26460 for MODE_VECTOR_BOOL.
26461 * machmode.def (VECTOR_BOOL_MODE): Document.
26462 * genmodes.c (VECTOR_BOOL_MODE): New macro.
26463 (make_vector_bool_mode): New function.
26464 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
26466 * lto-streamer-in.c (lto_input_mode_table): Likewise.
26467 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
26469 * stor-layout.c (int_mode_for_mode): Likewise.
26470 * tree.c (build_vector_type_for_mode): Likewise.
26471 * varasm.c (output_constant_pool_2): Likewise.
26472 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
26473 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
26474 for MODE_VECTOR_BOOL.
26475 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
26476 of mode class checks.
26477 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
26478 instead of a list of mode class checks.
26479 (expand_vector_scalar_condition): Likewise.
26480 (type_for_widest_vector_mode): Handle BImode as an inner mode.
26482 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26483 Alan Hayward <alan.hayward@arm.com>
26484 David Sherwood <david.sherwood@arm.com>
26486 * machmode.h (mode_size): Change from unsigned short to
26488 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
26489 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
26490 or if measurement_type is not polynomial.
26491 (fixed_size_mode::includes_p): Check for constant-sized modes.
26492 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
26493 return a poly_uint16 rather than an unsigned short.
26494 (emit_mode_size): Change the type of mode_size from unsigned short
26495 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
26496 (emit_mode_adjustments): Cope with polynomial vector sizes.
26497 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
26499 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
26501 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
26502 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
26503 * caller-save.c (setup_save_areas): Likewise.
26504 (replace_reg_with_saved_mem): Likewise.
26505 * calls.c (emit_library_call_value_1): Likewise.
26506 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
26507 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
26508 (gen_lowpart_for_combine): Likewise.
26509 * convert.c (convert_to_integer_1): Likewise.
26510 * cse.c (equiv_constant, cse_insn): Likewise.
26511 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
26512 (cselib_subst_to_values): Likewise.
26513 * dce.c (word_dce_process_block): Likewise.
26514 * df-problems.c (df_word_lr_mark_ref): Likewise.
26515 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
26516 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
26517 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
26518 (rtl_for_decl_location): Likewise.
26519 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
26520 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
26521 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
26522 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
26523 (expand_expr_real_1): Likewise.
26524 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
26525 (pad_below): Likewise.
26526 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
26527 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
26528 * ira.c (get_subreg_tracking_sizes): Likewise.
26529 * ira-build.c (ira_create_allocno_objects): Likewise.
26530 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
26531 (ira_sort_regnos_for_alter_reg): Likewise.
26532 * ira-costs.c (record_operand_costs): Likewise.
26533 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
26534 (resolve_simple_move): Likewise.
26535 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
26536 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
26537 (lra_constraints): Likewise.
26538 (CONST_POOL_OK_P): Reject variable-sized modes.
26539 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
26540 (add_pseudo_to_slot, lra_spill): Likewise.
26541 * omp-low.c (omp_clause_aligned_alignment): Likewise.
26542 * optabs-query.c (get_best_extraction_insn): Likewise.
26543 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
26544 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
26545 (expand_mult_highpart, valid_multiword_target_p): Likewise.
26546 * recog.c (offsettable_address_addr_space_p): Likewise.
26547 * regcprop.c (maybe_mode_change): Likewise.
26548 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
26549 * regrename.c (build_def_use): Likewise.
26550 * regstat.c (dump_reg_info): Likewise.
26551 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
26552 (find_reloads, find_reloads_subreg_address): Likewise.
26553 * reload1.c (eliminate_regs_1): Likewise.
26554 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
26555 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
26556 (simplify_binary_operation_1, simplify_subreg): Likewise.
26557 * targhooks.c (default_function_arg_padding): Likewise.
26558 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
26559 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
26560 (verify_gimple_assign_ternary): Likewise.
26561 * tree-inline.c (estimate_move_cost): Likewise.
26562 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
26563 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
26564 (get_address_cost_ainc): Likewise.
26565 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
26566 (vect_supportable_dr_alignment): Likewise.
26567 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
26568 (vectorizable_reduction): Likewise.
26569 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
26570 (vectorizable_operation, vectorizable_load): Likewise.
26571 * tree.c (build_same_sized_truth_vector_type): Likewise.
26572 * valtrack.c (cleanup_auto_inc_dec): Likewise.
26573 * var-tracking.c (emit_note_insn_var_location): Likewise.
26574 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
26575 (ADDR_VEC_ALIGN): Likewise.
26577 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26578 Alan Hayward <alan.hayward@arm.com>
26579 David Sherwood <david.sherwood@arm.com>
26581 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
26583 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
26584 or if measurement_type is polynomial.
26585 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
26586 * combine.c (make_extraction): Likewise.
26587 * dse.c (find_shift_sequence): Likewise.
26588 * dwarf2out.c (mem_loc_descriptor): Likewise.
26589 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
26590 (extract_bit_field, extract_low_bits): Likewise.
26591 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
26592 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
26593 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
26594 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
26595 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
26596 * reload.c (find_reloads): Likewise.
26597 * reload1.c (alter_reg): Likewise.
26598 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
26599 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
26600 * tree-if-conv.c (predicate_mem_writes): Likewise.
26601 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
26602 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
26603 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
26604 * valtrack.c (dead_debug_insert_temp): Likewise.
26605 * varasm.c (mergeable_constant_section): Likewise.
26606 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
26608 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26609 Alan Hayward <alan.hayward@arm.com>
26610 David Sherwood <david.sherwood@arm.com>
26612 * expr.c (expand_assignment): Cope with polynomial mode sizes
26613 when assigning to a CONCAT.
26615 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26616 Alan Hayward <alan.hayward@arm.com>
26617 David Sherwood <david.sherwood@arm.com>
26619 * machmode.h (mode_precision): Change from unsigned short to
26621 (mode_to_precision): Return a poly_uint16 rather than an unsigned
26623 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
26624 or if measurement_type is not polynomial.
26625 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
26626 in which the mode is already known to be a scalar_int_mode.
26627 * genmodes.c (emit_mode_precision): Change the type of mode_precision
26628 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
26630 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
26631 for GET_MODE_PRECISION.
26632 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
26633 for GET_MODE_PRECISION.
26634 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
26636 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
26637 (expand_field_assignment, make_extraction): Likewise.
26638 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
26639 (get_last_value): Likewise.
26640 * convert.c (convert_to_integer_1): Likewise.
26641 * cse.c (cse_insn): Likewise.
26642 * expr.c (expand_expr_real_1): Likewise.
26643 * lra-constraints.c (simplify_operand_subreg): Likewise.
26644 * optabs-query.c (can_atomic_load_p): Likewise.
26645 * optabs.c (expand_atomic_load): Likewise.
26646 (expand_atomic_store): Likewise.
26647 * ree.c (combine_reaching_defs): Likewise.
26648 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
26649 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
26650 * tree.h (type_has_mode_precision_p): Likewise.
26651 * ubsan.c (instrument_si_overflow): Likewise.
26653 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26654 Alan Hayward <alan.hayward@arm.com>
26655 David Sherwood <david.sherwood@arm.com>
26657 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
26658 polynomial numbers of units.
26659 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
26660 (valid_vector_subparts_p): New function.
26661 (build_vector_type): Remove temporary shim and take the number
26662 of units as a poly_uint64 rather than an int.
26663 (build_opaque_vector_type): Take the number of units as a
26664 poly_uint64 rather than an int.
26665 * tree.c (build_vector_from_ctor): Handle polynomial
26666 TYPE_VECTOR_SUBPARTS.
26667 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
26668 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
26669 (build_vector_from_val): If the number of units is variable,
26670 use build_vec_duplicate_cst for constant operands and
26671 VEC_DUPLICATE_EXPR otherwise.
26672 (make_vector_type): Remove temporary is_constant ().
26673 (build_vector_type, build_opaque_vector_type): Take the number of
26674 units as a poly_uint64 rather than an int.
26675 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
26677 * cfgexpand.c (expand_debug_expr): Likewise.
26678 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
26679 (store_constructor, expand_expr_real_1): Likewise.
26680 (const_scalar_mask_from_tree): Likewise.
26681 * fold-const-call.c (fold_const_reduction): Likewise.
26682 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
26683 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
26684 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
26685 (fold_relational_const): Likewise.
26686 (native_interpret_vector): Likewise. Change the size from an
26687 int to an unsigned int.
26688 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
26689 TYPE_VECTOR_SUBPARTS.
26690 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
26691 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
26692 duplicating a non-constant operand into a variable-length vector.
26693 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
26694 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
26695 * ipa-icf.c (sem_variable::equals): Likewise.
26696 * match.pd: Likewise.
26697 * omp-simd-clone.c (simd_clone_subparts): Likewise.
26698 * print-tree.c (print_node): Likewise.
26699 * stor-layout.c (layout_type): Likewise.
26700 * targhooks.c (default_builtin_vectorization_cost): Likewise.
26701 * tree-cfg.c (verify_gimple_comparison): Likewise.
26702 (verify_gimple_assign_binary): Likewise.
26703 (verify_gimple_assign_ternary): Likewise.
26704 (verify_gimple_assign_single): Likewise.
26705 * tree-pretty-print.c (dump_generic_node): Likewise.
26706 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
26707 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
26708 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
26709 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
26710 (vect_shift_permute_load_chain): Likewise.
26711 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
26712 (expand_vector_condition, optimize_vector_constructor): Likewise.
26713 (lower_vec_perm, get_compute_type): Likewise.
26714 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
26715 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
26716 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
26717 (vect_recog_mask_conversion_pattern): Likewise.
26718 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
26719 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
26720 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
26721 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
26722 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
26723 (vectorizable_shift, vectorizable_operation, vectorizable_store)
26724 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
26725 (supportable_widening_operation): Likewise.
26726 (supportable_narrowing_operation): Likewise.
26727 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
26729 * varasm.c (output_constant): Likewise.
26731 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26732 Alan Hayward <alan.hayward@arm.com>
26733 David Sherwood <david.sherwood@arm.com>
26735 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
26736 so that both the length == 3 and length != 3 cases set up their
26737 own permute vectors. Add comments explaining why we know the
26738 number of elements is constant.
26739 (vect_permute_load_chain): Likewise.
26741 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26742 Alan Hayward <alan.hayward@arm.com>
26743 David Sherwood <david.sherwood@arm.com>
26745 * machmode.h (mode_nunits): Change from unsigned char to
26747 (ONLY_FIXED_SIZE_MODES): New macro.
26748 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
26749 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
26750 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
26752 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
26753 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
26754 or if measurement_type is not polynomial.
26755 * genmodes.c (ZERO_COEFFS): New macro.
26756 (emit_mode_nunits_inline): Make mode_nunits_inline return a
26758 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
26759 Use ZERO_COEFFS when emitting initializers.
26760 * data-streamer.h (bp_pack_poly_value): New function.
26761 (bp_unpack_poly_value): Likewise.
26762 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
26763 for GET_MODE_NUNITS.
26764 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
26765 for GET_MODE_NUNITS.
26766 * tree.c (make_vector_type): Remove temporary shim and make
26767 the real function take the number of units as a poly_uint64
26768 rather than an int.
26769 (build_vector_type_for_mode): Handle polynomial nunits.
26770 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
26771 * emit-rtl.c (const_vec_series_p_1): Likewise.
26772 (gen_rtx_CONST_VECTOR): Likewise.
26773 * fold-const.c (test_vec_duplicate_folding): Likewise.
26774 * genrecog.c (validate_pattern): Likewise.
26775 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
26776 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
26777 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
26778 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
26779 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
26780 * rtlanal.c (subreg_get_info): Likewise.
26781 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
26782 (vect_grouped_load_supported): Likewise.
26783 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
26784 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
26785 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
26786 (simplify_const_unary_operation, simplify_binary_operation_1)
26787 (simplify_const_binary_operation, simplify_ternary_operation)
26788 (test_vector_ops_duplicate, test_vector_ops): Likewise.
26789 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
26790 instead of CONST_VECTOR_NUNITS.
26791 * varasm.c (output_constant_pool_2): Likewise.
26792 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
26793 explicit-encoded elements in the XVEC for variable-length vectors.
26795 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26797 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
26799 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26800 Alan Hayward <alan.hayward@arm.com>
26801 David Sherwood <david.sherwood@arm.com>
26803 * coretypes.h (fixed_size_mode): Declare.
26804 (fixed_size_mode_pod): New typedef.
26805 * builtins.h (target_builtins::x_apply_args_mode)
26806 (target_builtins::x_apply_result_mode): Change type to
26807 fixed_size_mode_pod.
26808 * builtins.c (apply_args_size, apply_result_size, result_vector)
26809 (expand_builtin_apply_args_1, expand_builtin_apply)
26810 (expand_builtin_return): Update accordingly.
26812 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26814 * cse.c (hash_rtx_cb): Hash only the encoded elements.
26815 * cselib.c (cselib_hash_rtx): Likewise.
26816 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
26817 CONST_VECTOR encoding.
26819 2018-01-03 Jakub Jelinek <jakub@redhat.com>
26820 Jeff Law <law@redhat.com>
26823 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
26824 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
26825 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
26826 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
26829 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
26830 explicitly probe *sp in a noreturn function if there were any callee
26831 register saves or frame pointer is needed.
26833 2018-01-03 Jakub Jelinek <jakub@redhat.com>
26836 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
26837 BLKmode for ternary, binary or unary expressions.
26840 * var-tracking.c (delete_vta_debug_insn): New inline function.
26841 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
26842 insns from get_insns () to NULL instead of each bb separately.
26843 Use delete_vta_debug_insn. No longer static.
26844 (vt_debug_insns_local, variable_tracking_main_1): Adjust
26845 delete_vta_debug_insns callers.
26846 * rtl.h (delete_vta_debug_insns): Declare.
26847 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
26848 instead of variable_tracking_main.
26850 2018-01-03 Martin Sebor <msebor@redhat.com>
26852 PR tree-optimization/83603
26853 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
26854 arguments past the endof the argument list in functions declared
26855 without a prototype.
26856 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
26857 Avoid checking when arguments are null.
26859 2018-01-03 Martin Sebor <msebor@redhat.com>
26862 * doc/extend.texi (attribute const): Fix a typo.
26863 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
26864 issuing -Wsuggest-attribute for void functions.
26866 2018-01-03 Martin Sebor <msebor@redhat.com>
26868 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
26869 offset_int::from instead of wide_int::to_shwi.
26870 (maybe_diag_overlap): Remove assertion.
26871 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
26872 * gimple-ssa-sprintf.c (format_directive): Same.
26873 (parse_directive): Same.
26874 (sprintf_dom_walker::compute_format_length): Same.
26875 (try_substitute_return_value): Same.
26877 2018-01-03 Jeff Law <law@redhat.com>
26879 PR middle-end/83654
26880 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
26881 non-constant residual for zero at runtime and avoid probing in
26882 that case. Reorganize code for trailing problem to mirror handling
26885 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
26887 PR tree-optimization/83501
26888 * tree-ssa-strlen.c (get_string_cst): New.
26889 (handle_char_store): Call get_string_cst.
26891 2018-01-03 Martin Liska <mliska@suse.cz>
26893 PR tree-optimization/83593
26894 * tree-ssa-strlen.c: Include tree-cfg.h.
26895 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
26896 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
26897 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
26899 (strlen_dom_walker::before_dom_children): Call
26900 gimple_purge_dead_eh_edges. Dump tranformation with details
26902 (strlen_dom_walker::before_dom_children): Update call by adding
26903 new argument cleanup_eh.
26904 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
26906 2018-01-03 Martin Liska <mliska@suse.cz>
26909 * cif-code.def (VARIADIC_THUNK): New enum value.
26910 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
26913 2018-01-03 Jan Beulich <jbeulich@suse.com>
26915 * sse.md (mov<mode>_internal): Tighten condition for when to use
26916 vmovdqu<ssescalarsize> for TI and OI modes.
26918 2018-01-03 Jakub Jelinek <jakub@redhat.com>
26920 Update copyright years.
26922 2018-01-03 Martin Liska <mliska@suse.cz>
26925 * ipa-visibility.c (function_and_variable_visibility): Skip
26926 functions with noipa attribure.
26928 2018-01-03 Jakub Jelinek <jakub@redhat.com>
26930 * gcc.c (process_command): Update copyright notice dates.
26931 * gcov-dump.c (print_version): Ditto.
26932 * gcov.c (print_version): Ditto.
26933 * gcov-tool.c (print_version): Ditto.
26934 * gengtype.c (create_file): Ditto.
26935 * doc/cpp.texi: Bump @copying's copyright year.
26936 * doc/cppinternals.texi: Ditto.
26937 * doc/gcc.texi: Ditto.
26938 * doc/gccint.texi: Ditto.
26939 * doc/gcov.texi: Ditto.
26940 * doc/install.texi: Ditto.
26941 * doc/invoke.texi: Ditto.
26943 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26945 * vector-builder.h (vector_builder::m_full_nelts): Change from
26946 unsigned int to poly_uint64.
26947 (vector_builder::full_nelts): Update prototype accordingly.
26948 (vector_builder::new_vector): Likewise.
26949 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
26950 (vector_builder::operator ==): Likewise.
26951 (vector_builder::finalize): Likewise.
26952 * int-vector-builder.h (int_vector_builder::int_vector_builder):
26953 Take the number of elements as a poly_uint64 rather than an
26955 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
26956 from unsigned int to poly_uint64.
26957 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
26958 (vec_perm_indices::new_vector): Likewise.
26959 (vec_perm_indices::length): Likewise.
26960 (vec_perm_indices::nelts_per_input): Likewise.
26961 (vec_perm_indices::input_nelts): Likewise.
26962 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
26963 number of elements per input as a poly_uint64 rather than an
26964 unsigned int. Use the original encoding for variable-length
26965 vectors, rather than clamping each individual element.
26966 For the second and subsequent elements in each pattern,
26967 clamp the step and base before clamping their sum.
26968 (vec_perm_indices::series_p): Handle polynomial element counts.
26969 (vec_perm_indices::all_in_range_p): Likewise.
26970 (vec_perm_indices_to_tree): Likewise.
26971 (vec_perm_indices_to_rtx): Likewise.
26972 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
26973 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
26974 (tree_vector_builder::new_binary_operation): Handle polynomial
26975 element counts. Return false if we need to know the number
26976 of elements at compile time.
26977 * fold-const.c (fold_vec_perm): Punt if the number of elements
26978 isn't known at compile time.
26980 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26982 * vec-perm-indices.h (vec_perm_builder): Change element type
26983 from HOST_WIDE_INT to poly_int64.
26984 (vec_perm_indices::element_type): Update accordingly.
26985 (vec_perm_indices::clamp): Handle polynomial element_types.
26986 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
26987 (vec_perm_indices::all_in_range_p): Likewise.
26988 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
26990 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
26991 polynomial vec_perm_indices element types.
26992 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
26993 * fold-const.c (fold_vec_perm): Likewise.
26994 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
26995 * tree-vect-generic.c (lower_vec_perm): Likewise.
26996 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
26997 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
26998 element type to HOST_WIDE_INT.
27000 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27001 Alan Hayward <alan.hayward@arm.com>
27002 David Sherwood <david.sherwood@arm.com>
27004 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
27005 rather than an int. Use plus_constant.
27006 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
27007 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
27009 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27010 Alan Hayward <alan.hayward@arm.com>
27011 David Sherwood <david.sherwood@arm.com>
27013 * calls.c (emit_call_1, expand_call): Change struct_value_size from
27014 a HOST_WIDE_INT to a poly_int64.
27016 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27017 Alan Hayward <alan.hayward@arm.com>
27018 David Sherwood <david.sherwood@arm.com>
27020 * calls.c (load_register_parameters): Cope with polynomial
27021 mode sizes. Require a constant size for BLKmode parameters
27022 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
27023 forces a parameter to be padded at the lsb end in order to
27024 fill a complete number of words, require the parameter size
27025 to be ordered wrt UNITS_PER_WORD.
27027 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27028 Alan Hayward <alan.hayward@arm.com>
27029 David Sherwood <david.sherwood@arm.com>
27031 * reload1.c (spill_stack_slot_width): Change element type
27032 from unsigned int to poly_uint64_pod.
27033 (alter_reg): Treat mode sizes as polynomial.
27035 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27036 Alan Hayward <alan.hayward@arm.com>
27037 David Sherwood <david.sherwood@arm.com>
27039 * reload.c (complex_word_subreg_p): New function.
27040 (reload_inner_reg_of_subreg, push_reload): Use it.
27042 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27043 Alan Hayward <alan.hayward@arm.com>
27044 David Sherwood <david.sherwood@arm.com>
27046 * lra-constraints.c (process_alt_operands): Reject matched
27047 operands whose sizes aren't ordered.
27048 (match_reload): Refer to this check here.
27050 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27051 Alan Hayward <alan.hayward@arm.com>
27052 David Sherwood <david.sherwood@arm.com>
27054 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
27055 that the mode size is in the set {1, 2, 4, 8, 16}.
27057 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27058 Alan Hayward <alan.hayward@arm.com>
27059 David Sherwood <david.sherwood@arm.com>
27061 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
27062 Use plus_constant instead of gen_rtx_PLUS.
27064 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27065 Alan Hayward <alan.hayward@arm.com>
27066 David Sherwood <david.sherwood@arm.com>
27068 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
27069 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
27070 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
27071 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
27072 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
27073 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
27074 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
27075 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
27076 * config/i386/i386.c (ix86_push_rounding): ...this new function.
27077 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
27079 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
27080 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
27081 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
27082 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
27083 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
27084 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
27085 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
27086 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
27087 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
27088 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
27090 * expr.c (emit_move_resolve_push): Treat the input and result
27091 of PUSH_ROUNDING as a poly_int64.
27092 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
27093 (emit_push_insn): Likewise.
27094 * lra-eliminations.c (mark_not_eliminable): Likewise.
27095 * recog.c (push_operand): Likewise.
27096 * reload1.c (elimination_effects): Likewise.
27097 * rtlanal.c (nonzero_bits1): Likewise.
27098 * calls.c (store_one_arg): Likewise. Require the padding to be
27099 known at compile time.
27101 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27102 Alan Hayward <alan.hayward@arm.com>
27103 David Sherwood <david.sherwood@arm.com>
27105 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
27106 Use plus_constant instead of gen_rtx_PLUS.
27108 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27109 Alan Hayward <alan.hayward@arm.com>
27110 David Sherwood <david.sherwood@arm.com>
27112 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
27113 rather than an int.
27115 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27116 Alan Hayward <alan.hayward@arm.com>
27117 David Sherwood <david.sherwood@arm.com>
27119 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
27120 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
27121 via stack temporaries. Treat the mode size as polynomial too.
27123 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27124 Alan Hayward <alan.hayward@arm.com>
27125 David Sherwood <david.sherwood@arm.com>
27127 * expr.c (expand_expr_real_2): When handling conversions involving
27128 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
27129 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
27130 as a poly_uint64 too.
27132 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27133 Alan Hayward <alan.hayward@arm.com>
27134 David Sherwood <david.sherwood@arm.com>
27136 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
27138 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27139 Alan Hayward <alan.hayward@arm.com>
27140 David Sherwood <david.sherwood@arm.com>
27142 * combine.c (can_change_dest_mode): Handle polynomial
27143 REGMODE_NATURAL_SIZE.
27144 * expmed.c (store_bit_field_1): Likewise.
27145 * expr.c (store_constructor): Likewise.
27146 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
27147 and polynomial REGMODE_NATURAL_SIZE.
27148 (gen_lowpart_common): Likewise.
27149 * reginfo.c (record_subregs_of_mode): Likewise.
27150 * rtlanal.c (read_modify_subreg_p): Likewise.
27152 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27153 Alan Hayward <alan.hayward@arm.com>
27154 David Sherwood <david.sherwood@arm.com>
27156 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
27157 numbers of elements.
27159 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27160 Alan Hayward <alan.hayward@arm.com>
27161 David Sherwood <david.sherwood@arm.com>
27163 * match.pd: Cope with polynomial numbers of vector elements.
27165 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27166 Alan Hayward <alan.hayward@arm.com>
27167 David Sherwood <david.sherwood@arm.com>
27169 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
27170 in a POINTER_PLUS_EXPR.
27172 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27173 Alan Hayward <alan.hayward@arm.com>
27174 David Sherwood <david.sherwood@arm.com>
27176 * omp-simd-clone.c (simd_clone_subparts): New function.
27177 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
27178 (ipa_simd_modify_function_body): Likewise.
27180 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27181 Alan Hayward <alan.hayward@arm.com>
27182 David Sherwood <david.sherwood@arm.com>
27184 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
27185 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
27186 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
27187 (expand_vector_condition, vector_element): Likewise.
27188 (subparts_gt): New function.
27189 (get_compute_type): Use subparts_gt.
27190 (count_type_subparts): Delete.
27191 (expand_vector_operations_1): Use subparts_gt instead of
27192 count_type_subparts.
27194 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27195 Alan Hayward <alan.hayward@arm.com>
27196 David Sherwood <david.sherwood@arm.com>
27198 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
27199 (vect_compile_time_alias): ...this new function. Do the calculation
27200 on poly_ints rather than trees.
27201 (vect_prune_runtime_alias_test_list): Update call accordingly.
27203 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27204 Alan Hayward <alan.hayward@arm.com>
27205 David Sherwood <david.sherwood@arm.com>
27207 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
27209 (vect_schedule_slp_instance): Likewise.
27211 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27212 Alan Hayward <alan.hayward@arm.com>
27213 David Sherwood <david.sherwood@arm.com>
27215 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
27216 constant and extern definitions for variable-length vectors.
27217 (vect_get_constant_vectors): Note that the number of units
27218 is known to be constant.
27220 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27221 Alan Hayward <alan.hayward@arm.com>
27222 David Sherwood <david.sherwood@arm.com>
27224 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
27225 of units as polynomial. Choose between WIDE and NARROW based
27228 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27229 Alan Hayward <alan.hayward@arm.com>
27230 David Sherwood <david.sherwood@arm.com>
27232 * tree-vect-stmts.c (simd_clone_subparts): New function.
27233 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
27235 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27236 Alan Hayward <alan.hayward@arm.com>
27237 David Sherwood <david.sherwood@arm.com>
27239 * tree-vect-stmts.c (vectorizable_call): Treat the number of
27240 vectors as polynomial. Use build_index_vector for
27241 IFN_GOMP_SIMD_LANE.
27243 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27244 Alan Hayward <alan.hayward@arm.com>
27245 David Sherwood <david.sherwood@arm.com>
27247 * tree-vect-stmts.c (get_load_store_type): Treat the number of
27248 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
27249 for variable-length vectors.
27250 (vectorizable_mask_load_store): Treat the number of units as
27251 polynomial, asserting that it is constant if the condition has
27252 already been enforced.
27253 (vectorizable_store, vectorizable_load): Likewise.
27255 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27256 Alan Hayward <alan.hayward@arm.com>
27257 David Sherwood <david.sherwood@arm.com>
27259 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
27260 of units as polynomial. Punt if we can't tell at compile time
27261 which vector contains the final result.
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-loop.c (vectorizable_induction): Treat the number
27268 of units as polynomial. Punt on SLP inductions. Use an integer
27269 VEC_SERIES_EXPR for variable-length integer reductions. Use a
27270 cast of such a series for variable-length floating-point
27273 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27274 Alan Hayward <alan.hayward@arm.com>
27275 David Sherwood <david.sherwood@arm.com>
27277 * tree.h (build_index_vector): Declare.
27278 * tree.c (build_index_vector): New function.
27279 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
27280 of units as polynomial, forcibly converting it to a constant if
27281 vectorizable_reduction has already enforced the condition.
27282 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
27283 to create a {1,2,3,...} vector.
27284 (vectorizable_reduction): Treat the number of units as polynomial.
27285 Choose vectype_in based on the largest scalar element size rather
27286 than the smallest number of units. Enforce the restrictions
27289 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27290 Alan Hayward <alan.hayward@arm.com>
27291 David Sherwood <david.sherwood@arm.com>
27293 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
27294 number of units as polynomial.
27296 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27297 Alan Hayward <alan.hayward@arm.com>
27298 David Sherwood <david.sherwood@arm.com>
27300 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
27301 * target.def (autovectorize_vector_sizes): Return the vector sizes
27302 by pointer, using vector_sizes rather than a bitmask.
27303 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
27304 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
27305 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
27307 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
27308 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
27309 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
27310 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
27311 * omp-general.c (omp_max_vf): Likewise.
27312 * omp-low.c (omp_clause_aligned_alignment): Likewise.
27313 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
27314 * tree-vect-loop.c (vect_analyze_loop): Likewise.
27315 * tree-vect-slp.c (vect_slp_bb): Likewise.
27316 * doc/tm.texi: Regenerate.
27317 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
27319 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
27320 the vector size as a poly_uint64 rather than an unsigned int.
27321 (current_vector_size): Change from an unsigned int to a poly_uint64.
27322 (get_vectype_for_scalar_type): Update accordingly.
27323 * tree.h (build_truth_vector_type): Take the size and number of
27324 units as a poly_uint64 rather than an unsigned int.
27325 (build_vector_type): Add a temporary overload that takes
27326 the number of units as a poly_uint64 rather than an unsigned int.
27327 * tree.c (make_vector_type): Likewise.
27328 (build_truth_vector_type): Take the number of units as a poly_uint64
27329 rather than an unsigned int.
27331 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27332 Alan Hayward <alan.hayward@arm.com>
27333 David Sherwood <david.sherwood@arm.com>
27335 * target.def (get_mask_mode): Take the number of units and length
27336 as poly_uint64s rather than unsigned ints.
27337 * targhooks.h (default_get_mask_mode): Update accordingly.
27338 * targhooks.c (default_get_mask_mode): Likewise.
27339 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
27340 * doc/tm.texi: Regenerate.
27342 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27343 Alan Hayward <alan.hayward@arm.com>
27344 David Sherwood <david.sherwood@arm.com>
27346 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
27347 * omp-general.c (omp_max_vf): Likewise.
27348 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
27349 (expand_omp_simd): Handle polynomial safelen.
27350 * omp-low.c (omplow_simd_context): Add a default constructor.
27351 (omplow_simd_context::max_vf): Change from int to poly_uint64.
27352 (lower_rec_simd_input_clauses): Update accordingly.
27353 (lower_rec_input_clauses): Likewise.
27355 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27356 Alan Hayward <alan.hayward@arm.com>
27357 David Sherwood <david.sherwood@arm.com>
27359 * tree-vectorizer.h (vect_nunits_for_cost): New function.
27360 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
27361 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
27362 (vect_analyze_slp_cost): Likewise.
27363 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
27364 (vect_model_load_cost): Likewise.
27366 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27367 Alan Hayward <alan.hayward@arm.com>
27368 David Sherwood <david.sherwood@arm.com>
27370 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
27371 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
27372 from an unsigned int * to a poly_uint64_pod *.
27373 (calculate_unrolling_factor): New function.
27374 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
27376 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27377 Alan Hayward <alan.hayward@arm.com>
27378 David Sherwood <david.sherwood@arm.com>
27380 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
27381 from an unsigned int to a poly_uint64.
27382 (_loop_vec_info::slp_unrolling_factor): Likewise.
27383 (_loop_vec_info::vectorization_factor): Change from an int
27385 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
27386 (vect_get_num_vectors): New function.
27387 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
27388 (vect_get_num_copies): Use vect_get_num_vectors.
27389 (vect_analyze_data_ref_dependences): Change max_vf from an int *
27390 to an unsigned int *.
27391 (vect_analyze_data_refs): Change min_vf from an int * to a
27393 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
27394 than an unsigned HOST_WIDE_INT.
27395 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
27396 (vect_analyze_data_ref_dependence): Change max_vf from an int *
27397 to an unsigned int *.
27398 (vect_analyze_data_ref_dependences): Likewise.
27399 (vect_compute_data_ref_alignment): Handle polynomial vf.
27400 (vect_enhance_data_refs_alignment): Likewise.
27401 (vect_prune_runtime_alias_test_list): Likewise.
27402 (vect_shift_permute_load_chain): Likewise.
27403 (vect_supportable_dr_alignment): Likewise.
27404 (dependence_distance_ge_vf): Take the vectorization factor as a
27405 poly_uint64 rather than an unsigned HOST_WIDE_INT.
27406 (vect_analyze_data_refs): Change min_vf from an int * to a
27408 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
27409 vfm1 as a poly_uint64 rather than an int. Make the same change
27410 for the returned bound_scalar.
27411 (vect_gen_vector_loop_niters): Handle polynomial vf.
27412 (vect_do_peeling): Likewise. Update call to
27413 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
27414 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
27416 * tree-vect-loop.c (vect_determine_vectorization_factor)
27417 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
27418 (vect_get_known_peeling_cost): Likewise.
27419 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
27420 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
27421 (vect_transform_loop): Likewise. Use the lowest possible VF when
27422 updating the upper bounds of the loop.
27423 (vect_min_worthwhile_factor): Make static. Return an unsigned int
27424 rather than an int.
27425 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
27426 polynomial unroll factors.
27427 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
27428 (vect_make_slp_decision): Likewise.
27429 (vect_supported_load_permutation_p): Likewise, and polynomial
27431 (vect_analyze_slp_cost): Handle polynomial vf.
27432 (vect_slp_analyze_node_operations): Likewise.
27433 (vect_slp_analyze_bb_1): Likewise.
27434 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
27435 than an unsigned HOST_WIDE_INT.
27436 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
27437 (vectorizable_load): Handle polynomial vf.
27438 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
27440 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
27442 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27443 Alan Hayward <alan.hayward@arm.com>
27444 David Sherwood <david.sherwood@arm.com>
27446 * match.pd: Handle bit operations involving three constants
27447 and try to fold one pair.
27449 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27451 * tree-vect-loop-manip.c: Include gimple-fold.h.
27452 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
27453 niters_maybe_zero parameters. Handle other cases besides a step of 1.
27454 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
27455 Add a path that uses a step of VF instead of 1, but disable it
27457 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
27458 and niters_no_overflow parameters. Update calls to
27459 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
27460 Create a new SSA name if the latter choses to use a ste other
27461 than zero, and return it via niters_vector_mult_vf_var.
27462 * tree-vect-loop.c (vect_transform_loop): Update calls to
27463 vect_do_peeling, vect_gen_vector_loop_niters and
27464 slpeel_make_loop_iterate_ntimes.
27465 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
27466 (vect_gen_vector_loop_niters): Update declarations after above changes.
27468 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
27470 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
27471 128-bit round to integer instructions.
27472 (ceil<mode>2): Likewise.
27473 (btrunc<mode>2): Likewise.
27474 (round<mode>2): Likewise.
27476 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
27478 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
27479 unaligned VSX load/store on P8/P9.
27480 (expand_block_clear): Allow the use of unaligned VSX
27481 load/store on P8/P9.
27483 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
27485 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
27487 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
27488 swap associated with both a load and a store.
27490 2018-01-02 Andrew Waterman <andrew@sifive.com>
27492 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
27493 * config/riscv/riscv.md (clear_cache): Use it.
27495 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
27497 * web.c: Remove out-of-date comment.
27499 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
27501 * expr.c (fixup_args_size_notes): Check that any existing
27502 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
27503 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
27504 (emit_single_push_insn): ...here.
27506 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
27508 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
27509 (const_vector_encoded_nelts): New function.
27510 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
27511 (const_vector_int_elt, const_vector_elt): Declare.
27512 * emit-rtl.c (const_vector_int_elt_1): New function.
27513 (const_vector_elt): Likewise.
27514 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
27515 of CONST_VECTOR_ELT.
27517 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
27519 * expr.c: Include rtx-vector-builder.h.
27520 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
27521 directly on the tree encoding.
27522 (const_vector_from_tree): Likewise.
27523 * optabs.c: Include rtx-vector-builder.h.
27524 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
27525 sequence of "u" values.
27526 * vec-perm-indices.c: Include rtx-vector-builder.h.
27527 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
27528 directly on the vec_perm_indices encoding.
27530 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
27532 * doc/rtl.texi (const_vector): Describe new encoding scheme.
27533 * Makefile.in (OBJS): Add rtx-vector-builder.o.
27534 * rtx-vector-builder.h: New file.
27535 * rtx-vector-builder.c: Likewise.
27536 * rtl.h (rtx_def::u2): Add a const_vector field.
27537 (CONST_VECTOR_NPATTERNS): New macro.
27538 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
27539 (CONST_VECTOR_DUPLICATE_P): Likewise.
27540 (CONST_VECTOR_STEPPED_P): Likewise.
27541 (CONST_VECTOR_ENCODED_ELT): Likewise.
27542 (const_vec_duplicate_p): Check for a duplicated vector encoding.
27543 (unwrap_const_vec_duplicate): Likewise.
27544 (const_vec_series_p): Check for a non-duplicated vector encoding.
27545 Say that the function only returns true for integer vectors.
27546 * emit-rtl.c: Include rtx-vector-builder.h.
27547 (gen_const_vec_duplicate_1): Delete.
27548 (gen_const_vector): Call gen_const_vec_duplicate instead of
27549 gen_const_vec_duplicate_1.
27550 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
27551 (gen_const_vec_duplicate): Use rtx_vector_builder.
27552 (gen_const_vec_series): Likewise.
27553 (gen_rtx_CONST_VECTOR): Likewise.
27554 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
27555 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
27556 Build a new vector rather than modifying a CONST_VECTOR in-place.
27557 (handle_special_swappables): Update call accordingly.
27558 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
27559 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
27560 Build a new vector rather than modifying a CONST_VECTOR in-place.
27561 (handle_special_swappables): Update call accordingly.
27563 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
27565 * simplify-rtx.c (simplify_const_binary_operation): Use
27566 CONST_VECTOR_ELT instead of XVECEXP.
27568 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
27570 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
27571 the selector elements to be different from the data elements
27572 if the selector is a VECTOR_CST.
27573 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
27574 ssizetype for the selector.
27576 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
27578 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
27579 before testing each element individually.
27580 * tree-vect-generic.c (lower_vec_perm): Likewise.
27582 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
27584 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
27585 * selftest-run-tests.c (selftest::run_tests): Call it.
27586 * vector-builder.h (vector_builder::operator ==): New function.
27587 (vector_builder::operator !=): Likewise.
27588 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
27589 (vec_perm_indices::all_from_input_p): New function.
27590 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
27591 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
27592 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
27593 instead of reading the VECTOR_CST directly. Detect whether both
27594 vector inputs are the same before constructing the vec_perm_indices,
27595 and update the number of inputs argument accordingly. Use the
27596 utility functions added above. Only construct sel2 if we need to.
27598 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
27600 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
27601 the broadcast of the low byte.
27602 (expand_mult_highpart): Use an explicit encoding for the permutes.
27603 * optabs-query.c (can_mult_highpart_p): Likewise.
27604 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
27605 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
27606 (vectorizable_bswap): Likewise.
27607 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
27608 explicit encoding for the power-of-2 permutes.
27609 (vect_permute_store_chain): Likewise.
27610 (vect_grouped_load_supported): Likewise.
27611 (vect_permute_load_chain): Likewise.
27613 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
27615 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
27616 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
27617 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
27618 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
27619 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
27620 (vect_gen_perm_mask_any): Likewise.
27622 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
27624 * int-vector-builder.h: New file.
27625 * vec-perm-indices.h: Include int-vector-builder.h.
27626 (vec_perm_indices): Redefine as an int_vector_builder.
27627 (auto_vec_perm_indices): Delete.
27628 (vec_perm_builder): Redefine as a stand-alone class.
27629 (vec_perm_indices::vec_perm_indices): New function.
27630 (vec_perm_indices::clamp): Likewise.
27631 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
27632 (vec_perm_indices::new_vector): New function.
27633 (vec_perm_indices::new_expanded_vector): Update for new
27634 vec_perm_indices class.
27635 (vec_perm_indices::rotate_inputs): New function.
27636 (vec_perm_indices::all_in_range_p): Operate directly on the
27637 encoded form, without computing elided elements.
27638 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
27639 encoding. Update for new vec_perm_indices class.
27640 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
27641 the given vec_perm_builder.
27642 (expand_vec_perm_var): Update vec_perm_builder constructor.
27643 (expand_mult_highpart): Use vec_perm_builder instead of
27644 auto_vec_perm_indices.
27645 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
27646 vec_perm_indices instead of auto_vec_perm_indices. Use a single
27647 or double series encoding as appropriate.
27648 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
27649 vec_perm_indices instead of auto_vec_perm_indices.
27650 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
27651 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
27652 (vect_permute_store_chain): Likewise.
27653 (vect_grouped_load_supported): Likewise.
27654 (vect_permute_load_chain): Likewise.
27655 (vect_shift_permute_load_chain): Likewise.
27656 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
27657 (vect_transform_slp_perm_load): Likewise.
27658 (vect_schedule_slp_instance): Likewise.
27659 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
27660 (vectorizable_mask_load_store): Likewise.
27661 (vectorizable_bswap): Likewise.
27662 (vectorizable_store): Likewise.
27663 (vectorizable_load): Likewise.
27664 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
27665 vec_perm_indices instead of auto_vec_perm_indices. Use
27666 tree_to_vec_perm_builder to read the vector from a tree.
27667 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
27668 vec_perm_builder instead of a vec_perm_indices.
27669 (have_whole_vector_shift): Use vec_perm_builder and
27670 vec_perm_indices instead of auto_vec_perm_indices. Leave the
27671 truncation to calc_vec_perm_mask_for_shift.
27672 (vect_create_epilog_for_reduction): Likewise.
27673 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
27674 from auto_vec_perm_indices to vec_perm_indices.
27675 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
27676 instead of changing individual elements.
27677 (aarch64_vectorize_vec_perm_const): Use new_vector to install
27678 the vector in d.perm.
27679 * config/arm/arm.c (expand_vec_perm_d::perm): Change
27680 from auto_vec_perm_indices to vec_perm_indices.
27681 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
27682 instead of changing individual elements.
27683 (arm_vectorize_vec_perm_const): Use new_vector to install
27684 the vector in d.perm.
27685 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
27686 Update vec_perm_builder constructor.
27687 (rs6000_expand_interleave): Likewise.
27688 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
27689 (rs6000_expand_interleave): Likewise.
27691 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
27693 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
27694 to qimode could truncate the indices.
27695 * optabs.c (expand_vec_perm_var): Likewise.
27697 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
27699 * Makefile.in (OBJS): Add vec-perm-indices.o.
27700 * vec-perm-indices.h: New file.
27701 * vec-perm-indices.c: Likewise.
27702 * target.h (vec_perm_indices): Replace with a forward class
27704 (auto_vec_perm_indices): Move to vec-perm-indices.h.
27705 * optabs.h: Include vec-perm-indices.h.
27706 (expand_vec_perm): Delete.
27707 (selector_fits_mode_p, expand_vec_perm_var): Declare.
27708 (expand_vec_perm_const): Declare.
27709 * target.def (vec_perm_const_ok): Replace with...
27710 (vec_perm_const): ...this new hook.
27711 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
27712 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
27713 * doc/tm.texi: Regenerate.
27714 * optabs.def (vec_perm_const): Delete.
27715 * doc/md.texi (vec_perm_const): Likewise.
27716 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
27717 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
27718 expand_vec_perm for constant permutation vectors. Assert that
27719 the mode of variable permutation vectors is the integer equivalent
27720 of the mode that is being permuted.
27721 * optabs-query.h (selector_fits_mode_p): Declare.
27722 * optabs-query.c: Include vec-perm-indices.h.
27723 (selector_fits_mode_p): New function.
27724 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
27725 is defined, instead of checking whether the vec_perm_const_optab
27726 exists. Use targetm.vectorize.vec_perm_const instead of
27727 targetm.vectorize.vec_perm_const_ok. Check whether the indices
27728 fit in the vector mode before using a variable permute.
27729 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
27730 vec_perm_indices instead of an rtx.
27731 (expand_vec_perm): Replace with...
27732 (expand_vec_perm_const): ...this new function. Take the selector
27733 as a vec_perm_indices rather than an rtx. Also take the mode of
27734 the selector. Update call to shift_amt_for_vec_perm_mask.
27735 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
27736 Use vec_perm_indices::new_expanded_vector to expand the original
27737 selector into bytes. Check whether the indices fit in the vector
27738 mode before using a variable permute.
27739 (expand_vec_perm_var): Make global.
27740 (expand_mult_highpart): Use expand_vec_perm_const.
27741 * fold-const.c: Includes vec-perm-indices.h.
27742 * tree-ssa-forwprop.c: Likewise.
27743 * tree-vect-data-refs.c: Likewise.
27744 * tree-vect-generic.c: Likewise.
27745 * tree-vect-loop.c: Likewise.
27746 * tree-vect-slp.c: Likewise.
27747 * tree-vect-stmts.c: Likewise.
27748 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
27750 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
27751 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
27752 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
27753 (aarch64_vectorize_vec_perm_const): ...this new function.
27754 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
27755 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
27756 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
27757 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
27758 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
27759 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
27760 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
27762 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
27763 check for NEON modes.
27764 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
27765 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
27766 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
27767 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
27769 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
27770 the old VEC_PERM_CONST conditions.
27771 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
27772 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
27773 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
27774 (ia64_vectorize_vec_perm_const_ok): Merge into...
27775 (ia64_vectorize_vec_perm_const): ...this new function.
27776 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
27777 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
27778 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
27779 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
27780 * config/mips/mips.c (mips_expand_vec_perm_const)
27781 (mips_vectorize_vec_perm_const_ok): Merge into...
27782 (mips_vectorize_vec_perm_const): ...this new function.
27783 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
27784 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
27785 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
27786 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
27787 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
27788 (rs6000_expand_vec_perm_const): Delete.
27789 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
27791 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
27792 (altivec_expand_vec_perm_const_le): Take each operand individually.
27793 Operate on constant selectors rather than rtxes.
27794 (altivec_expand_vec_perm_const): Likewise. Update call to
27795 altivec_expand_vec_perm_const_le.
27796 (rs6000_expand_vec_perm_const): Delete.
27797 (rs6000_vectorize_vec_perm_const_ok): Delete.
27798 (rs6000_vectorize_vec_perm_const): New function.
27799 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
27800 an element count and rtx array.
27801 (rs6000_expand_extract_even): Update call accordingly.
27802 (rs6000_expand_interleave): Likewise.
27803 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
27804 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
27805 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
27806 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
27807 (rs6000_expand_vec_perm_const): Delete.
27808 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
27809 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
27810 (altivec_expand_vec_perm_const_le): Take each operand individually.
27811 Operate on constant selectors rather than rtxes.
27812 (altivec_expand_vec_perm_const): Likewise. Update call to
27813 altivec_expand_vec_perm_const_le.
27814 (rs6000_expand_vec_perm_const): Delete.
27815 (rs6000_vectorize_vec_perm_const_ok): Delete.
27816 (rs6000_vectorize_vec_perm_const): New function. Remove stray
27817 reference to the SPE evmerge intructions.
27818 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
27819 an element count and rtx array.
27820 (rs6000_expand_extract_even): Update call accordingly.
27821 (rs6000_expand_interleave): Likewise.
27822 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
27823 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
27825 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
27827 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
27829 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
27830 vector mode and that that mode matches the mode of the data
27832 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
27833 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
27834 directly using expand_vec_perm_1 when forcing selectors into
27836 (expand_vec_perm_var): New function, split out from expand_vec_perm.
27838 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
27840 * optabs-query.h (can_vec_perm_p): Delete.
27841 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
27842 * optabs-query.c (can_vec_perm_p): Split into...
27843 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
27844 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
27845 particular selector is valid.
27846 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
27847 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
27848 (vect_grouped_load_supported): Likewise.
27849 (vect_shift_permute_load_chain): Likewise.
27850 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
27851 (vect_transform_slp_perm_load): Likewise.
27852 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
27853 (vectorizable_bswap): Likewise.
27854 (vect_gen_perm_mask_checked): Likewise.
27855 * fold-const.c (fold_ternary_loc): Likewise. Don't take
27856 implementations of variable permutation vectors into account
27857 when deciding which selector to use.
27858 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
27859 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
27860 with a false third argument.
27861 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
27862 to test whether the constant selector is valid and can_vec_perm_var_p
27863 to test whether a variable selector is valid.
27865 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
27867 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
27868 * optabs-query.c (can_vec_perm_p): Likewise.
27869 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
27870 instead of vec_perm_indices.
27871 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
27872 (vect_gen_perm_mask_checked): Likewise,
27873 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
27874 (vect_gen_perm_mask_checked): Likewise,
27876 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
27878 * optabs-query.h (qimode_for_vec_perm): Declare.
27879 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
27880 (qimode_for_vec_perm): ...this new function.
27881 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
27883 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
27885 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
27886 does not have a conditional at the top.
27888 2018-01-02 Richard Biener <rguenther@suse.de>
27890 * ipa-inline.c (big_speedup_p): Fix expression.
27892 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
27895 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
27898 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
27902 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
27903 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
27904 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
27905 cond_taken_branch_cost 3->4.
27907 2018-01-01 Jakub Jelinek <jakub@redhat.com>
27909 PR tree-optimization/83581
27910 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
27911 TODO_cleanup_cfg if any changes have been made.
27913 PR middle-end/83608
27914 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
27915 convert_modes if target mode has the right side, but different mode
27918 PR middle-end/83609
27919 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
27920 last argument when extracting from CONCAT. If either from_real or
27921 from_imag is NULL, use expansion through memory. If result is not
27922 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
27923 the parts directly to inner mode, if even that fails, use expansion
27926 PR middle-end/83623
27927 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
27928 check for bswap in mode rather than HImode and use that in expand_unop
27931 Copyright (C) 2018 Free Software Foundation, Inc.
27933 Copying and distribution of this file, with or without modification,
27934 are permitted in any medium without royalty provided the copyright
27935 notice and this notice are preserved.