1 2018-10-15 David Malcolm <dmalcolm@redhat.com>
3 * gcc-rich-location.h (gcc_rich_location::add_location_if_nearby):
4 Fix usage of "error_at_rich_loc" in the comment.
6 2018-10-15 Renlin Li <renlin.li@arm.com>
9 * tree-vectorizer.c (try_vectorize_loop_1): Don't use
10 if-conversioned loop when it contains ifn with types not
12 * internal-fn.c (expand_direct_optab_fn): Add an assert.
13 (direct_internal_fn_supported_p): New helper function.
14 * internal-fn.h (direct_internal_fn_supported_p): Declare.
16 2018-10-15 Jakub Jelinek <jakub@redhat.com>
19 * common/config/i386/i386-common.c (OPTION_MASK_ISA_AVX512F_UNSET):
20 Remove OPTION_MASK_ISA_AVX5124FMAPS_UNSET and
21 OPTION_MASK_ISA_AVX5124VNNIW_UNSET.
23 2018-10-15 Bin Cheng <bin.cheng@linux.alibaba.com>
25 PR tree-optimization/87022
26 * tree-loop-distribution.c (pg_add_dependence_edges): Check all
27 bits in dist vector rather than the first one.
29 2018-10-15 Richard Biener <rguenther@suse.de>
32 * tree-ssa-structalias.c (struct vls_data): Add escaped_p member.
33 (visit_loadstore): When a used restrict tag escaped verify that
34 the points-to solution of "other" pointers do not include
36 (compute_dependence_clique): If a used restrict tag escaped
37 communicated that down to visit_loadstore.
39 2018-10-15 Andreas Krebbel <krebbel@linux.ibm.com>
41 * config/s390/s390.c (s390_expand_vec_init): Force vector element
42 into reg if it isn't a general operand.
44 2018-10-14 H.J. Lu <hongjiu.lu@intel.com>
47 * config/i386/sse.md (*vec_dupv2di): Add register source to
50 2018-10-14 H.J. Lu <hongjiu.lu@intel.com>
53 * common/config/i386/i386-common.c (OPTION_MASK_ISA_AVX512F_UNSET):
54 Add OPTION_MASK_ISA_AVX512IFMA_UNSET,
55 OPTION_MASK_ISA_AVX5124FMAPS_UNSET and
56 OPTION_MASK_ISA_AVX5124VNNIW_UNSET.
58 2018-10-13 Eric Botcazou <ebotcazou@adacore.com>
60 * dwarf2cfi.c (struct dw_trace_info): Add args_size_defined_for_eh.
61 (notice_args_size): Set it in the current trace if no insn that can
62 throw internally has been seen yet.
63 (connect_traces): When connecting args_size between traces, allow the
64 incoming values not to match if there is an insn setting it before the
65 first insn that can throw internally; in that case, force the creation
66 of a CFI note on this latter insn.
68 2018-10-13 Jonathan Wakely <jwakely@redhat.com>
70 * opt-problem.h (opt_wrapper): Use template-argument-list when naming
71 the base class, because using the injected-class-name was not clearly
72 specified until DR 176.
74 2018-10-12 Paul Koning <ni1d@arrl.net>
76 * config/pdp11/pdp11.md (doloop_end): New expander.
77 (doloop_end_insn): renamed from "doloop_end".
78 (addqi3): New pattern.
79 (subqi3): New pattern.
80 * config/pdp11/predicates.md (incdec_operand): New predicate.
82 2018-10-12 Yury Gribov <tetra2005@gmail.com>
85 * real.c (format_helper::can_represent_integral_type_p): New function
86 * real.h (format_helper::can_represent_integral_type_p): Ditto.
87 * match.pd: New pattern.
89 2018-10-12 Alexandre Oliva <oliva@adacore.com>
91 * configure.ac: Introduce --enable-large-address-aware
92 to define MINGW_DEFAULT_LARGE_ADDR_AWARE.
93 * doc/install.texi: Document it.
94 * configure, config.in: Rebuilt.
95 * config/i386/mingw32.h (LINK_SPEC_LARGE_ADDR_AWARE): Define,
96 based on MINGW_DEFAULT_LARGE_ADDR_AWARE.
97 (LINK_SPEC): Insert it.
98 * config/i386/mingw-264.h: Likewise.
100 * cygming.h (HAVE_GAS_ALIGNED_COMM): Fallback-define.
102 2018-10-12 Peter Bergner <bergner@linux.ibm.com>
104 PR rtl-optimization/87600
105 * ira-lives (non_conflicting_reg_copy_p): Disable for non LRA targets.
107 2018-10-12 Paul Koning <ni1d@arrl.net>
109 * doc/md.texi (doloop_end): Document that the pattern code may
110 need to check operand mode.
112 2018-10-12 Wilco Dijkstra <wdijkstr@arm.com>
114 * config/aarch64/aarch64.md (zero_extendsidi2_aarch64): Add alternatives
115 to zero-extend between int and floating-point registers.
116 (load_pair_zero_extendsidi2_aarch64): Add alternative for zero-extended
117 ldp into floating-point registers. Add type and arch attributes.
118 (zero_extend<SHORT:mode><GPI:mode>2_aarch64): Add arch attribute.
119 Use f_loads for type attribute.
121 2018-10-11 Martin Sebor <msebor@redhat.com>
123 * doc/extend.texi (attribute packed): Correct typos.
125 2018-10-11 Martin Sebor <msebor@redhat.com>
127 * doc/extend.texi (attribute flatten): Mention interaction with
130 2018-10-11 Jan Hubicka <hubicka@ucw.cz>
133 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
134 Set new_decl virtual flag to zero.
136 2018-10-11 Martin Sebor <msebor@redhat.com>
139 * doc/extend.texi (attribute format_arg): Discuss using multiple
140 attributes on a single function.
142 2018-10-11 Giuliano Belinassi <giuliano.belinassi@usp.br>
144 PR tree-optimization/86829
145 * match.pd (sin (atan (x))): New simplification rules.
146 (cos (atan (x))): Likewise.
147 * real.c (build_sinatan_real): New function.
148 * real.h (build_sinatan_real): Prototype.
150 2018-10-11 Will Schmidt <will_schmidt@vnet.ibm.com>
152 * config/rs6000/rs6000.c (map_to_integral_tree_type): New helper
154 (fold_mergeeo_helper): New helper function.
155 (rs6000_gimple_fold_builtin): Add hooks for vec_mergee and vec_mergeo
156 intrinsics. Correct some whitespace indentation issues.
158 2018-10-11 Wilco Dijkstra <wdijkstr@arm.com>
161 * config/aarch64/aarch64.c (aarch64_mask_and_shift_for_ubfiz_p):
162 Use HOST_WIDE_INT_1U for shift.
164 2018-10-11 Doug Rupp <rupp@adacore.com>
165 Olivier Hainque <hainque@adacore.com>
167 * config/rs6000/vxworks.h (VXWORKS_RELAX_LINK_SPEC): New macro.
168 Pass --relax to the linker for RTPs.
169 (LINK_SPEC): Append VXWORKS_RELAX_LINK_SPEC.
171 2018-10-11 Andrew Stubbs <ams@codesourcery.com>
172 Jan Hubicka <jh@suse.cz>
173 Martin Jambor <mjambor@suse.cz>
175 * print-rtl.c (print_rtx_operand_codes_E_and_V): Print how many times
176 the same elements are repeated rather than printing all of them.
177 * read-rtl.c (rtx_reader::read_rtx_operand): Recognize and expand
179 * read-rtl-function.c (test_loading_repeat): New function.
180 (read_rtl_function_c_tests): Call test_loading_repeat.
181 * rtl-tests.c (test_dumping_repeat): New function.
182 (rtl_tests_c_tests): Call test_dumping_repeat.
184 2018-10-11 Richard Biener <rguenther@suse.de>
186 * config/i386/x86-tune-costs.h (bdver?_memcpy, bdver?_memset,
187 bdver?_cost): Unify to ...
188 (bdver_memcpy, bdver_memset, bdver_cost): ... this.
189 * config/i386/i386.c (processor_cost_table): Adjust.
191 2018-10-10 Eric Botcazou <ebotcazou@adacore.com>
194 * cgraphunit.c (cgraph_node::expand_thunk): Force DECL_IGNORED_P on
195 the thunk when expanding to GIMPLE.
197 2018-10-10 Bernd Edlinger <bernd.edlinger@hotmail.de>
199 * varasm.c (mergeable_string_section): Don't try to move zero-length
200 strings to the merge section.
202 2018-10-10 Uros Bizjak <ubizjak@gmail.com>
205 * config/i386/mmx.md (const_vector 0 -> mem splitter): New splitter.
207 2018-10-10 Jakub Jelinek <jakub@redhat.com>
210 * config/i386/i386-builtin.def (IX86_BUILTIN_RDPMC): Move from args set
213 2018-10-10 Richard Biener <rguenther@suse.de>
215 * config/i386/sse.md (reduc_plus_scal_v8df, reduc_plus_scal_v4df,
216 reduc_plus_scal_v2df, reduc_plus_scal_v16sf, reduc_plus_scal_v8sf,
217 reduc_plus_scal_v4sf): Merge into pattern reducing to half width
218 and recursing and pattern terminating the recursion on SSE
219 vector width using ix86_expand_reduc.
220 (reduc_sminmax_scal_<mode>): Split into part reducing to half
221 width and recursing and SSE2 vector variant doing the final
222 reduction with ix86_expand_reduc.
223 (reduc_uminmax_scal_<mode>): Likewise for the AVX512 variants
224 with terminating the recursion at AVX level, splitting that
227 2018-10-09 David Malcolm <dmalcolm@redhat.com>
229 * genmatch.c (error_cb): Rename to...
230 (diagnostic_cb): ...this, converting int params to enums.
231 (fatal_at): Update for renaming.
232 (warning_at): Likewise.
234 * input.c (selftest::ebcdic_execution_charset::apply):
235 Update for renaming of...
236 (selftest::ebcdic_execution_charset::on_error): ...this, renaming
238 (selftest::ebcdic_execution_charset::on_diagnostic): ...this,
239 converting level and reason to enums.
240 (class selftest::lexer_error_sink): Rename to...
241 (class selftest::lexer_test_options): ...this, renaming field
242 "m_errors" to "m_diagnostics".
243 (selftest::lexer_test_options::apply): Update for renaming of...
244 (selftest::lexer_test_options::on_error): ...this, renaming to...
245 (selftest::lexer_test_options::on_diagnostic): ...this
246 converting level and reason to enums.
247 (selftest::test_lexer_string_locations_raw_string_unterminated):
248 Update for renamings.
249 * opth-gen.awk (struct cpp_reason_option_codes_t): Use enum for
252 2018-10-09 Paul A. Clarke <pc@us.ibm.com>
254 * config.gcc (powerpc*-*-*): Add pmmintrin.h to extra_headers.
255 * config/rs6000/pmmintrin.h: New file.
257 2018-10-09 Eric Botcazou <ebotcazou@adacore.com>
259 PR tree-optimization/86659
260 * gimple-match.h (gimple_match_op constructors): Initialize reverse.
262 2018-10-09 Richard Biener <rguenther@suse.de>
264 PR tree-optimization/63155
265 * tree-ssa-structalias.c: Include tree-ssa.h.
266 (get_constraint_for_ssa_var): For undefs return nothing_id.
267 (find_func_aliases): Cleanup PHI handling.
269 2018-10-09 Richard Biener <rguenther@suse.de>
271 * tree-vectorizer.c (dump_stmt_cost): Fix cut&paste missing
274 2018-10-09 Martin Liska <mliska@suse.cz>
276 * asan.c (asan_emit_stack_protection): If a stack variable
277 is located in a same file as current function, then emit
278 line info into variable definition string.
280 2018-10-08 Eric Botcazou <ebotcazou@adacore.com>
282 * print-rtl.c (rtx_writer::print_rtx_operand_code_i): Print column
285 2018-10-08 Eric Botcazou <ebotcazou@adacore.com>
287 * cgraphunit.c (cgraph_node::expand_thunk): Do not force DECL_IGNORED_P
290 2018-10-08 Eric Botcazou <ebotcazou@adacore.com>
292 PR tree-optimization/86659
293 * gimple-match.h (struct gimple_match_op): Add reverse field.
294 (gimple_match_op::set_op): New overloaded method.
295 * gimple-match-head.c (maybe_build_generic_op) <BIT_FIELD_REF>: Set
296 the REF_REVERSE_STORAGE_ORDER flag on the value.
297 (gimple_simplify) <GIMPLE_ASSIGN>: For BIT_FIELD_REF, propagate the
298 REF_REVERSE_STORAGE_ORDER flag and avoid simplifying if it is set.
300 2018-10-08 Richard Sandiford <richard.sandiford@arm.com>
303 * gimple-ssa-backprop.c (backprop::intersect_uses): Use
304 FOR_EACH_IMM_USE_FAST instead of FOR_EACH_IMM_USE_STMT.
306 2018-10-08 H.J. Lu <hongjiu.lu@intel.com>
309 * config/i386/avx512fintrin.h (_mm512_mask_fmaddsub_round_pd):
310 Defined with __builtin_ia32_vfmaddsubpd512_mask.
312 2018-10-08 Richard Biener <rguenther@suse.de>
314 * config/i386/x86-tune-costs.h (znver1_cost): Make AVX256 vector loads
315 cost the same as AVX128 ones.
317 2018-10-08 Paul Koning <ni1d@arrl.net>
319 * config/pdp11/pdp11-protos.h (output_block_move): Remove.
320 (expand_block_move): New function.
321 * config/pdp11/pdp11.c (output_block_move): Remove.
322 (expand_block_move): New function.
323 * config/pdp11/pdp11.h (MOVE_RATIO): New definition.
324 * config/pdp11/pdp11.md (movmemhi): Use expand_block_move.
325 (*movmemhi1): Remove.
327 2018-10-08 Robin Dapp <rdapp@linux.ibm.com>
329 * config/s390/2827.md: Increase latencies for some FP instructions.
331 2018-10-08 Richard Biener <rguenther@suse.de>
333 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
335 * tree-vectorizer.c (dump_stmt_cost): Add cost param and dump it.
336 * tree-vectorizer.h (dump_stmt_cost): Adjust.
337 (add_stmt_cost): Dump return value of the hook.
339 2018-10-08 Richard Biener <rguenther@suse.de>
341 PR tree-optimization/63155
342 * tree-ssa-propagate.c (add_ssa_edge): Do cheap check first.
343 (ssa_propagation_engine::ssa_propagate): Remove redundant
346 2018-10-05 Peter Bergner <bergner@linux.ibm.com>
348 PR rtl-optimization/86939
349 PR rtl-optimization/87479
350 * ira.h (non_conflicting_reg_copy_p): New prototype.
351 * ira-lives.c (ignore_reg_for_conflicts): New static variable.
352 (make_hard_regno_dead): Don't add conflicts for register
353 ignore_reg_for_conflicts.
354 (make_object_dead): Likewise.
355 (non_conflicting_reg_copy_p): New function.
356 (process_bb_node_lives): Set ignore_reg_for_conflicts for copies.
357 Remove special conflict handling of REAL_PIC_OFFSET_TABLE_REGNUM.
358 * lra-lives.c (ignore_reg_for_conflicts): New static variable.
359 (make_hard_regno_dead): Don't add conflicts for register
360 ignore_reg_for_conflicts. Remove special conflict handling of
361 REAL_PIC_OFFSET_TABLE_REGNUM. Remove now unused argument
362 check_pic_pseudo_p and update callers.
363 (mark_pseudo_dead): Don't add conflicts for register
364 ignore_reg_for_conflicts.
365 (process_bb_lives): Set ignore_reg_for_conflicts for copies.
367 2018-10-05 Andrew Waterman <andrew@sifive.com>
368 Jim Wilson <jimw@sifive.com>
370 * config/riscv/riscv.md (f<quiet_pattern>_quiet<ANYF:mode><X:mode>4):
371 Add define_expand. Add ! HONOR_SNANS check to current pattern. Add
372 new pattern using HONOR_SNANS that emits one extra instruction.
374 2018-10-05 Segher Boessenkool <segher@kernel.crashing.org>
376 * config/rs6000/rs6000.md (unnamed mfcr scc_comparison_operator
377 patterns): Merge SI and DI patterns to a GPR pattern.
378 (unnamed define_insn and define_split for record form of that): Merge
379 to a single define_insn_and_split pattern.
381 2018-10-05 David Malcolm <dmalcolm@redhat.com>
384 * input.c (expand_location_to_spelling_point): Add param "aspect"
385 and use rather than hardcoding LOCATION_ASPECT_CARET.
386 (get_substring_ranges_for_loc): Handle the case of a single token
387 within a macro expansion.
388 * input.h (expand_location_to_spelling_point): Add "aspect" param,
389 defaulting to LOCATION_ASPECT_CARET.
391 2018-10-05 Paul Koning <ni1d@arrl.net>
393 * config/pdp11/pdp11.c (TARGET_CXX_GUARD_TYPE): Define.
394 (TARGET_CXX_CLASS_DATA_ALWAYS_COMDAT): Define.
395 (TARGET_CXX_LIBRARY_RTTI_COMDAT): Define.
396 (pdp11_guard_type): New function.
398 2018-10-05 Paul Koning <ni1d@arrl.net>
400 * config/pdp11/pdp11.h (FLOAT_TYPE_SIZE): Always 32.
401 * config/pdp11/pdp11.opt (mfloat32): Remove.
403 * doc/invoke.texi (pdp11 -mfloat32): Remove:
404 (pdp11 -mfloat64): Remove.
406 2018-10-05 Uros Bizjak <ubizjak@gmail.com>
408 * config/i386/i386.md (*cmpxf_cc_i387): Remove pattern.
409 (*cmp<mode>_cc_i387): Ditto.
410 (*cmpu<mode>_cc_i387): Ditto.
411 (*cmp<X87MODEF:mode>_<SWI24:mode>_cc_i387): Ditto.
412 * config/i386/i386.c (ix86_expand_fp_compare): Remove
414 <case IX86_FPCMP_SAHF>: Do not generate pattern with HImode clobber.
415 Emit x86_sahf_1 pattern.
416 (ix86_expand_compare): Update call to ix86_expand_fp_compare.
417 (ix86_expand_carry_flag_compare): Ditto.
419 2018-10-05 Uros Bizjak <ubizjak@gmail.com>
421 * config/i386/i386.md (*cmpxf_i387): Change operand 2 predicate
422 to reg_or_0_operand. Add "C" constraint.
423 (*cmpxf_cc_i387): Ditto.
424 (*cmp<mode>_i387): Change operand 2 predicate
425 to nonimm_or_0_operand. Add "C" constraint.
426 (*cmp<mode>_cc_i387): Ditto.
427 (*cmp<mode>_0_i387): Remove insn pattern.
428 (*cmp<mode>_0_cc_i387): Ditto.
430 2018-10-05 Uros Bizjak <ubizjak@gmail.com>
432 * config/i386/constraints.md ("C"): Do not depend on TARGET_SSE.
433 * config/i386/predicates.md (nonimm_or_0_operand): Rename
434 from vector_move_operand. Update all uses.
436 2018-10-05 Martin Sebor <msebor@redhat.com>
438 PR tree-optimization/87490
439 * builtins.c (expand_builtin_strnlen): Handle a null data.decl
442 2018-10-05 Richard Biener <rguenther@suse.de>
444 PR tree-optimization/63155
445 * tree-ssa-ccp.c (ccp_propagate::visit_phi): Avoid excess
446 vertical space in dumpfiles.
447 * tree-ssa-propagate.h
448 (ssa_propagation_engine::process_ssa_edge_worklist): Remove.
449 * tree-ssa-propagate.c (cfg_blocks_back): New global.
450 (ssa_edge_worklist_back): Likewise.
451 (curr_order): Likewise.
452 (cfg_blocks_get): Remove abstraction.
453 (cfg_blocks_add): Likewise.
454 (cfg_blocks_empty_p): Likewise.
455 (add_ssa_edge): Add to current or next worklist based on
457 (add_control_edge): Likewise.
458 (ssa_propagation_engine::process_ssa_edge_worklist): Fold
460 (ssa_propagation_engine::ssa_propagate): ... here. Unify
461 iteration from CFG and SSA edge worklist so we process
462 everything in RPO order, prioritizing forward progress
464 (ssa_prop_init): Allocate new worklists, do not dump
466 (ssa_prop_fini): Free new worklists.
468 2018-10-05 Richard Biener <rguenther@suse.de>
470 * tree-core.h (tree_block::abstract_flag): Remove.
471 (tree_block::block_num): Make full 32bits.
472 * tree.def (BLOCK): Remove docs about BLOCK_ABSTRACT.
473 * tree.h (BLOCK_ABSTRACT): Remove.
474 * dwarf2out.c (gen_lexical_block_die): Remove dead code
475 resulting from BLOCK_ABSTRACT being always false.
476 (gen_inlined_subroutine_die): Likewise.
477 (gen_block_die): Likewise.
478 * tree.c (block_ultimate_origin): Likewise.
479 * tree-pretty-print.c (dump_block_node): Remove code dealing
481 * tree-ssa-live.c (dump_scope_block): Likewise.
482 * tree-streamer-in.c (unpack_ts_block_value_fields): Likewise.
483 * tree-streamer-out.c (pack_ts_block_value_fields): Likewise.
485 2018-10-05 Richard Biener <rguenther@suse.de>
487 * config/i386/i386.c (ix86_add_stmt_cost): When scalar cost
488 is asked for initialize mode to the component mode of the
491 2018-10-05 H.J. Lu <hongjiu.lu@intel.com>
494 * config/i386/gnu-user.h (ASM_SPEC): Don't pass -msse2avx to
496 * config/i386/gnu-user64.h (ASM_SPEC): Likewise.
498 2018-10-05 Segher Boessenkool <segher@kernel.crashing.org>
501 * config/rs6000/rs6000-builtin.def (RS6000_BUILTIN_SET_FPSCR_DRN): Use
503 * config/rs6000/rs6000.md (rs6000_set_fpscr_rn): Require the operand
504 to be DImode. When using mffscrn, force the operand to a register.
506 2018-10-04 Uros Bizjak <ubizjak@gmail.com>
508 * config/i386/i386.md (*fop_<X87MODEF:mode>_2_i387): Macroize insn
509 from *fop_<MODEF:mode>_2_i387 and *fop_xf_2_i387 using
510 X87MODEF mode iterator.
511 (*fop_<X87MODEF:mode>_3_i387): Macroize insn from
512 *fop_<MODEF:mode>_3_i387 and *fop_xf_3_i387 using
513 X87MODEF mode iterator.
515 2018-10-04 Vinay Kumar <vinay.kumar@blackfigtech.com>
517 * doc/invoke.texi (-Wno-prio-ctor-dtor): Document new warning
520 2018-10-04 David Malcolm <dmalcolm@redhat.com>
522 * Makefile.in (OBJS): Add opt-problem.o.
523 * dump-context.h: Include "selftest.h.
524 (selftest::temp_dump_context): New forward decl.
525 (class dump_context): Make friend of class
526 selftest::temp_dump_context.
527 (dump_context::dump_loc_immediate): New decl.
528 (class dump_pretty_printer): Move here from dumpfile.c.
529 (class temp_dump_context): Move to namespace selftest.
530 (temp_dump_context::temp_dump_context): Add param
531 "forcibly_enable_dumping".
532 (selftest::verify_dumped_text):
533 (ASSERT_DUMPED_TEXT_EQ): Move here from dumpfile.c.
534 (selftest::verify_item):
535 (ASSERT_IS_TEXT): Move here from dumpfile.c.
536 (ASSERT_IS_TREE): Likewise.
537 (ASSERT_IS_GIMPLE): Likewise.
538 * dumpfile.c (dump_context::dump_loc): Move immediate dumping
540 (dump_context::dump_loc_immediate): ...this new function.
541 (class dump_pretty_printer): Move to dump-context.h.
542 (dump_switch_p_1): Don't enable MSG_PRIORITY_REEMITTED.
543 (opt_info_switch_p_1): Enable MSG_PRIORITY_REEMITTED.
544 (temp_dump_context::temp_dump_context): Move to "selftest"
545 namespace. Add param "forcibly_enable_dumping", and use it to
546 conditionalize the use of m_pp;
547 (selftest::verify_dumped_text): Make non-static.
548 (ASSERT_DUMPED_TEXT_EQ): Move to dump-context.h.
549 (selftest::verify_item): Make non-static.
550 (ASSERT_IS_TEXT): Move to dump-context.h.
551 (ASSERT_IS_TREE): Likewise.
552 (ASSERT_IS_GIMPLE): Likewise.
553 (selftest::test_capture_of_dump_calls): Pass "true" for new
554 param of temp_dump_context.
555 * dumpfile.h (enum dump_flag): Add MSG_PRIORITY_REEMITTED, adding
556 it to MSG_ALL_PRIORITIES. Update values of TDF_COMPARE_DEBUG and
558 * opt-problem.cc: New file.
559 * opt-problem.h: New file.
560 * optinfo-emit-json.cc
561 (selftest::test_building_json_from_dump_calls): Pass "true" for
562 new param of temp_dump_context.
563 * optinfo.cc (optinfo_kind_to_dump_flag): New function.
564 (optinfo::emit_for_opt_problem): New function.
565 (optinfo::emit): Clarity which emit_item is used.
566 * optinfo.h (optinfo::get_dump_location): New accessor.
567 (optinfo::emit_for_opt_problem): New decl.
568 (optinfo::emit): Make const.
569 * selftest-run-tests.c (selftest::run_tests): Call
570 selftest::opt_problem_cc_tests.
571 * selftest.h (selftest::opt_problem_cc_tests): New decl.
572 * tree-data-ref.c (dr_analyze_innermost): Convert return type from
573 bool to opt_result, converting fprintf messages to
574 opt_result::failure_at calls. Add "stmt" param for use by the
576 (create_data_ref): Pass "stmt" to the dr_analyze_innermost call.
577 (runtime_alias_check_p): Convert return type from bool to
578 opt_result, converting dump_printf calls to
579 opt_result::failure_at, using the statement DDR_A for their
581 (find_data_references_in_stmt): Convert return type from bool to
582 opt_result, converting "return false" to opt_result::failure_at
584 * tree-data-ref.h: Include "opt-problem.h".
585 (dr_analyze_innermost): Convert return type from bool to opt_result,
586 and add a const gimple * param.
587 (find_data_references_in_stmt): Convert return type from bool to
589 (runtime_alias_check_p): Likewise.
590 * tree-predcom.c (find_looparound_phi): Pass "init_stmt" to
591 dr_analyze_innermost.
592 * tree-vect-data-refs.c (vect_mark_for_runtime_alias_test):
593 Convert return type from bool to opt_result, adding a message for
594 the PARAM_VECT_MAX_VERSION_FOR_ALIAS_CHECKS zero case.
595 (vect_analyze_data_ref_dependence): Convert return type from bool
596 to opt_result. Change sense of return type from "false"
597 effectively meaning "no problems" to "false" meaning a problem,
598 so that "return false" becomes "return opt_result::success".
599 Convert "return true" calls to opt_result::failure_at, using
600 the location of statement A rather than vect_location.
601 (vect_analyze_data_ref_dependences): Convert return type from bool
603 (verify_data_ref_alignment): Likewise, converting dump_printf_loc
604 calls to opt_result::failure_at, using the stmt location rather
606 (vect_verify_datarefs_alignment): Convert return type from bool
608 (vect_enhance_data_refs_alignment): Likewise. Split local "stat"
609 into multiple more-tightly-scoped copies.
610 (vect_analyze_data_refs_alignment): Convert return type from bool
612 (vect_analyze_data_ref_accesses): Likewise, converting a
613 "return false" to a "return opt_result::failure_at", adding a
615 (vect_prune_runtime_alias_test_list): Convert return type from
616 bool to opt_result, converting dump_printf_loc to
617 opt_result::failure_at. Add a %G to show the pertinent statement,
618 and use the stmt's location rather than vect_location.
619 (vect_find_stmt_data_reference): Convert return type from
620 bool to opt_result, converting dump_printf_loc to
621 opt_result::failure_at, using stmt's location.
622 (vect_analyze_data_refs): Convert return type from bool to
623 opt_result. Convert "return false" to "return
624 opt_result::failure_at", adding messages as needed.
625 * tree-vect-loop.c (vect_determine_vf_for_stmt_1): Convert return
626 type from bool to opt_result.
627 (vect_determine_vf_for_stmt): Likewise.
628 (vect_determine_vectorization_factor): Likewise, converting
629 dump_printf_loc to opt_result::failure_at, using location of phi
630 rather than vect_location.
631 (vect_analyze_loop_form_1): Convert return type from bool to
632 opt_result, converting dump_printf_loc calls, retaining the use of
634 (vect_analyze_loop_form): Convert return type from loop_vec_info
635 to opt_loop_vec_info.
636 (vect_analyze_loop_operations): Convert return type from bool to
637 opt_result, converting dump_printf_loc calls, using the location
638 of phi/stmt rather than vect_location where available. Convert
639 various "return false" to "return opt_result::failure_at" with
640 "unsupported phi" messages.
641 (vect_get_datarefs_in_loop): Convert return type from bool to
642 opt_result. Add a message for the
643 PARAM_LOOP_MAX_DATAREFS_FOR_DATADEPS failure.
644 (vect_analyze_loop_2): Convert return type from bool to
645 opt_result. Ensure "ok" is set to a opt_result::failure_at before
646 each "goto again;", adding new messages where needed.
647 Add "unsupported grouped {store|load}" messages.
648 (vect_analyze_loop): Convert return type from loop_vec_info to
650 * tree-vect-slp.c (vect_analyze_slp): Convert return type from
652 * tree-vect-stmts.c (process_use): Likewise, converting
653 dump_printf_loc call and using stmt location, rather than
655 (vect_mark_stmts_to_be_vectorized): Likeise.
656 (vect_analyze_stmt): Likewise, adding a %G.
657 (vect_get_vector_types_for_stmt): Convert return type from bool to
658 opt_result, converting dump_printf_loc calls and using stmt
659 location, rather than vect_location.
660 (vect_get_mask_type_for_stmt): Convert return type from tree to
661 opt_tree, converting dump_printf_loc calls and using stmt location.
662 * tree-vectorizer.c: Include "opt-problem.h.
663 (try_vectorize_loop_1): Flag "Analyzing loop at" dump message as
664 MSG_PRIORITY_INTERNALS. Convert local "loop_vinfo" from
665 loop_vec_info to opt_loop_vec_info. If if fails, and dumping is
666 enabled, use it to report at the top level "couldn't vectorize
667 loop" followed by the problem.
668 * tree-vectorizer.h (opt_loop_vec_info): New typedef.
669 (vect_mark_stmts_to_be_vectorized): Convert return type from bool
671 (vect_analyze_stmt): Likewise.
672 (vect_get_vector_types_for_stmt): Likewise.
673 (tree vect_get_mask_type_for_stmt): Likewise.
674 (vect_analyze_data_ref_dependences): Likewise.
675 (vect_enhance_data_refs_alignment): Likewise.
676 (vect_analyze_data_refs_alignment): Likewise.
677 (vect_verify_datarefs_alignment): Likewise.
678 (vect_analyze_data_ref_accesses): Likewise.
679 (vect_prune_runtime_alias_test_list): Likewise.
680 (vect_find_stmt_data_reference): Likewise.
681 (vect_analyze_data_refs): Likewise.
682 (vect_analyze_loop): Convert return type from loop_vec_info to
684 (vect_analyze_loop_form): Likewise.
685 (vect_analyze_slp): Convert return type from bool to opt_result.
687 2018-10-04 David Malcolm <dmalcolm@redhat.com>
689 * doc/invoke.texi (-fopt-info): Document new "internals"
691 * dump-context.h (dump_context::apply_dump_filter_p): New decl.
692 * dumpfile.c (dump_options): Update for renaming of MSG_ALL to
694 (optinfo_verbosity_options): Add "internals".
695 (kind_as_string): Update for renaming of MSG_ALL to MSG_ALL_KINDS.
696 (dump_context::apply_dump_filter_p): New member function.
697 (dump_context::dump_loc): Use apply_dump_filter_p rather than
698 explicitly masking the dump_kind.
699 (dump_context::begin_scope): Increment the scope depth first. Use
700 apply_dump_filter_p rather than explicitly masking the dump_kind.
701 (dump_context::emit_item): Use apply_dump_filter_p rather than
702 explicitly masking the dump_kind.
703 (dump_dec): Likewise.
704 (dump_hex): Likewise.
705 (dump_switch_p_1): Default to MSG_ALL_PRIORITIES.
706 (opt_info_switch_p_1): Default to MSG_PRIORITY_USER_FACING.
707 (opt_info_switch_p): Update handling of default
708 MSG_OPTIMIZED_LOCATIONS to cope with default of
709 MSG_PRIORITY_USER_FACING.
710 (dump_basic_block): Use apply_dump_filter_p rather than explicitly
711 masking the dump_kind.
712 (selftest::test_capture_of_dump_calls): Update test_dump_context
713 instances to use MSG_ALL_KINDS | MSG_PRIORITY_USER_FACING rather
714 than MSG_ALL. Generalize scope test to be run at all four
715 combinations of with/without MSG_PRIORITY_USER_FACING and
716 MSG_PRIORITY_INTERNALS, adding examples of explicit priority
717 for each of the two values.
718 * dumpfile.h (enum dump_flag): Add comment about the MSG_* flags.
719 Rename MSG_ALL to MSG_ALL_KINDS. Add MSG_PRIORITY_USER_FACING,
720 MSG_PRIORITY_INTERNALS, and MSG_ALL_PRIORITIES, updating the
721 values for TDF_COMPARE_DEBUG and TDF_ALL_VALUES.
722 (AUTO_DUMP_SCOPE): Add a note to the comment about the interaction
724 * tree-vect-loop-manip.c (vect_loop_versioning): Mark versioning
725 dump messages as MSG_PRIORITY_USER_FACING.
726 * tree-vectorizer.h (DUMP_VECT_SCOPE): Add a note to the comment
727 about the interaction with MSG_PRIORITY_*.
729 2018-10-04 Bernd Edlinger <bernd.edlinger@hotmail.de>
731 * varasm.c (output_constant): Add new parameter merge_strings.
732 Make strings properly zero terminated in merge string sections.
733 (mergeable_string_section): Don't fail if the last char is non-zero.
734 (assemble_variable_contents): Handle merge string sections.
735 (assemble_variable): Likewise.
736 (assemble_constant_contents): Likewise.
737 (output_constant_def_contents): Likewise.
738 (output_constructor_array_range,
739 output_constructor_regular_field): Adjust call to output_constant.
740 (output_object_block): Adjust call to assemble_constant_contents
741 and assemble_variable_contents.
743 2018-10-04 Martin Liska <mliska@suse.cz>
746 * cgraphunit.c (process_function_and_variable_attributes):
747 Warn about a function with alias attribute and a body.
749 2018-10-04 Martin Liska <mliska@suse.cz>
752 * multiple_target.c (redirect_to_specific_clone): New function.
753 (ipa_target_clone): Use it.
754 * tree-inline.c: Fix comment.
756 2018-10-04 David Malcolm <dmalcolm@redhat.com>
758 * dumpfile.c (gcc::dump_manager::dump_manager): Initialize new
760 (gcc::dump_manager::~dump_manager): Free m_optinfo_filename.
761 (gcc::dump_manager::register_pass): New member function, adapted
762 from loop body in gcc::pass_manager::register_pass, adding a
763 call to update_dfi_for_opt_info.
764 (gcc::dump_manager::opt_info_enable_passes): Store the
765 -fopt-info options into the new fields. Move the loop
767 (gcc::dump_manager::update_dfi_for_opt_info): ...this new member
769 * dumpfile.h (struct opt_pass): New forward decl.
770 (gcc::dump_manager::register_pass): New decl.
771 (gcc::dump_manager::update_dfi_for_opt_info): New decl.
772 (class gcc::dump_manager): Add fields "m_optgroup_flags",
773 "m_optinfo_flags", and "m_optinfo_filename".
774 * passes.c (gcc::pass_manager::register_pass): Move all of the
775 dump-handling code to gcc::dump_manager::register_pass.
777 2018-10-04 Peter Bergner <bergner@linux.ibm.com>
779 PR rtl-optimization/87466
780 * target.def (setjmp_preserves_nonvolatile_regs_p): New target hook.
781 * doc/tm.texi.in (TARGET_SETJMP_PRESERVES_NONVOLATILE_REGS_P): New hook.
782 * doc/tm.texi: Regenerate.
783 * ira-lives.c (process_bb_node_lives): Use the new target hook.
784 * lra-lives.c (process_bb_lives): Likewise.
785 * config/rs6000/rs6000.c (TARGET_SETJMP_PRESERVES_NONVOLATILE_REGS_P):
788 2018-10-04 Tamar Christina <tamar.christina@arm.com>
790 * params.c (add_params): Fix initialization.
792 2018-10-04 Martin Liska <mliska@suse.cz>
794 PR gcov-profile/84107
795 * tree-profile.c (init_ic_make_global_vars):
796 Remove ic_void_ptr_var and ic_gcov_type_ptr_var.
797 Come up with new ic_tuple* variables. Emit
798 __gcov_indirect_call{,_topn} variables.
799 (gimple_gen_ic_profiler): Access the variable
801 (gimple_gen_ic_func_profiler): Access
802 __gcov_indirect_call.callee field.
803 (gimple_init_gcov_profiler): Use ptr_type_node.
804 * value-prof.c (gimple_ic): Use ptr_type_node.
806 2018-10-04 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
808 PR tree-optimization/85787
809 * ipa-pure-const.c (malloc_candidate_p_1): Move most of malloc_candidate_p
810 into this function and add support for detecting multiple phis.
811 (DUMP_AND_RETURN): Move from malloc_candidate_p into top-level macro.
813 2018-10-04 Martin Liska <mliska@suse.cz>
816 * ipa-inline.c (inline_to_all_callers_1):
817 Call ultimate_alias_target for node being inlined.
819 2018-10-03 Jeff Law <law@redhat.com>
821 * gimple-ssa-sprintf.c (format_string): Do not hardcode size of
823 * tree.c (get_typenode_from_name): Moved from fortran/trans-types.c.
824 * tree.h (get_typenode_from_name): Prototype.
826 2018-10-03 Uros Bizjak <ubizjak@gmail.com>
828 * config/i386/i386.md (*cmp<X87MODEF:mode>_<SWI24:mode>_i387):
829 Change operand 2 predicate to nonimmediate_operand.
830 (*cmp<X87MODEF:mode>_<SWI24:mode>_cc_i387): Ditto.
832 2018-10-03 Martin Sebor <msebor@redhat.com>
833 Jeff Law <law@redhat.com>
835 * gimple-ssa-sprintf.c (struct fmtresult): Add new member and
837 (get_string_length): Detect unterminated arrays.
838 (format_string): Same.
839 (format_directive): Warn about unterminated arrays.
840 (handle_gimple_call): Mark statements with no_warning as needed.
842 2018-10-03 Jim Wilson <jimw@sifive.com>
844 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): For ABI_ILP32E,
845 also define __riscv_abi_rve. Delete trailing white space.
847 2018-10-03 Paul Koning <ni1d@arrl.net>
849 Enable LRA register allocator for PDP11.
850 * config/pdp11/constraints.md (Q): Use define_memory_constraint.
853 * config/pdp11/pdp11.c (pdp11_lra_p): New function.
854 * config/pdp11/pdp11.opt (-mlra): New option.
855 * doc/invoke.texi (PDP-11 Options): Document -mlra.
857 2018-10-03 Uros Bizjak <ubizjak@gmail.com>
859 * config/i386/i386.md (*<absneg:code>extendsfdf2): Remove.
860 (*<absneg:code>extend<mode>xf2): Ditto.
862 2018-10-03 Aldy Hernandez <aldyh@redhat.com>
864 PR tree-optimization/87415
865 * tree-vrp.c (set_value_range_with_overflow): Special case one bit
868 2018-10-02 Jeff Law <law@redhat.com>
870 * gimple-fold.c (get_range_strlen): Only set *nonstr when
871 an unterminated string is discovered. Bubble up range
872 even for unterminated strings.
873 (gimple_fold_builtin_strlen): Do not fold if get_range_strlen
874 indicates the string was not terminated via NONSTR.
876 2018-10-03 Aldy Hernandez <aldyh@redhat.com>
878 * tree-vrp.c (extract_range_from_unary_expr): Special case all
880 Do not do anything special for anti-ranges.
882 2018-10-03 Jérôme Lambourg <lambourg@adacore.com>
884 * config/arm/vxworks.h (ARM_TARGET2_DWARF_FORMAT): Adjust to
885 DW_EH_PE_pcrel | DW_EH_PE_indirect for RTPs.
887 2018-10-03 Martin Liska <mliska@suse.cz>
889 PR gcov-profile/86109
890 * coverage.c (coverage_begin_function): Do not
891 mark lambdas as artificial.
892 * tree-core.h (struct GTY): Remove tm_clone_flag
893 and introduce new lambda_function.
894 * tree.h (DECL_LAMBDA_FUNCTION): New macro.
896 2018-10-02 Aaron Sawdey <acsawdey@linux.ibm.com>
899 * config/rs6000/rs6000-string.c (expand_strn_compare): Check that both
900 P8_VECTOR and VSX are enabled.
902 2018-10-02 Andreas Krebbel <krebbel@linux.ibm.com>
904 * config/s390/driver-native.c (s390_host_detect_local_cpu): Add
905 0x3907 as CPU model number.
907 2018-10-02 Andreas Krebbel <krebbel@linux.ibm.com>
909 * common/config/s390/s390-common.c: Rename PF_ARCH12 to PF_Z14.
910 * config/s390/s390.h (enum processor_flags): Rename PF_ARCH12 to
911 PF_Z14. Rename TARGET_CPU_ARCH12 to TARGET_CPU_Z14,
912 TARGET_CPU_ARCH12_P to TARGET_CPU_Z14_P, TARGET_ARCH12 to
913 TARGET_Z14, and TARGET_ARCH12_P to TARGET_Z14_P.
914 * config/s390/s390.md: Likewise. Rename also the cpu attribute
915 value from arch12 to z14.
917 2018-10-02 Uros Bizjak <ubizjak@gmail.com>
919 * config/i386/i386.md (fxam<mode>2_i387_with_temp): Remove.
921 (isinf<mode>2): Ditto.
923 2018-10-02 Uros Bizjak <ubizjak@gmail.com>
925 * config/i386/i386.c (ix86_emit_i387_round): Extend op1 to XFmode
926 before emitting fxam. Perform calculations in XFmode.
928 2018-10-02 Marc Glisse <marc.glisse@inria.fr>
930 * match.pd (((X /[ex] A) +- B) * A): New transformation.
932 2018-10-02 Marc Glisse <marc.glisse@inria.fr>
935 * fold-const.c (fold_plusminus_mult_expr): Handle complex and vectors.
936 * tree.c (signed_or_unsigned_type_for): Handle complex.
938 2018-10-02 Jeff Law <law@redhat.com>
940 * gimple-fold.c (get_range_strlen): Remove dead code.
942 2018-10-02 Martin Sebor <msebor@redhat.com>
943 Jeff Law <law@redhat.com>
945 * builtins.c (unterminated_array): Add new arguments.
946 If argument is not terminated, bubble up size and exact
948 (expand_builtin_strnlen): Detect, avoid expanding
949 and diagnose unterminated arrays.
950 (c_strlen): Fill in offset of start of unterminated strings.
951 * builtins.h (unterminated_array): Update prototype.
953 2018-10-02 Richard Biener <rguenther@suse.de>
955 * config/i386/sse.md (reduc_plus_scal_v4df): Avoid the use
956 of haddv4df, first reduce to SSE width and exploit the fact
957 that we only need element zero with the reduction result.
958 (reduc_plus_scal_v2df): Likewise.
960 2018-10-02 Eric Botcazou <ebotcazou@adacore.com>
962 * dojump.h (do_jump): Delete.
963 (do_jump_1): Likewise.
964 (split_comparison): Move around.
965 * dojump.c (do_jump): Make static.
966 (do_jump_1): Likewise.
967 (jumpifnot): Move around.
968 (jumpifnot_1): Likewise.
970 (jumpif_1): Likewise.
971 * expr.c (expand_expr_real_1): Call jumpif[not] instead of do_jump.
973 2018-10-02 Eric Botcazou <ebotcazou@adacore.com>
975 * reorg.c (make_return_insns): Use emit_copy_of_insn_after for the
976 insns in the delay slot and add_insn_after for the jump insn.
978 2018-10-02 Richard Biener <rguenther@suse.de>
980 * tree-inline.c (expand_call_inline): Use the location of
981 the callee declaration for the inline-entry marker.
982 * final.c (notice_source_line): Remove special-casing of
983 NOTE_INSN_INLINE_ENTRY.
985 2018-10-01 Carl Love <cel@us.ibm.com>
988 * config/rs6000/rs6000-builtin.def (__builtin_mffsl): New.
989 (__builtin_mtfsb0): New.
990 (__builtin_mtfsb1): New.
991 ( __builtin_set_fpscr_rn): New.
992 (__builtin_set_fpscr_drn): New.
993 * config/rs6000/rs6000.c (rs6000_expand_mtfsb_builtin): Add.
994 (rs6000_expand_set_fpscr_rn_builtin): Add.
995 (rs6000_expand_set_fpscr_drn_builtin): Add.
996 (rs6000_expand_builtin): Add case statement entries for
997 RS6000_BUILTIN_MTFSB0, RS6000_BUILTIN_MTFSB1,
998 RS6000_BUILTIN_SET_FPSCR_RN, RS6000_BUILTIN_SET_FPSCR_DRN,
999 RS6000_BUILTIN_MFFSL.
1000 (rs6000_init_builtins): Add ftype initialization and def_builtin
1001 calls for __builtin_mffsl, __builtin_mtfsb0, __builtin_mtfsb1,
1002 __builtin_set_fpscr_rn, __builtin_set_fpscr_drn.
1003 * config/rs6000.md (rs6000_mtfsb0, rs6000_mtfsb1, rs6000_mffscrn,
1004 rs6000_mffscdrn): Add define_insn.
1005 (rs6000_set_fpscr_rn, rs6000_set_fpscr_drn): Add define_expand.
1006 * doc/extend.texi: Add documentation for the builtins.
1008 2018-10-01 Richard Biener <rguenther@suse.de>
1010 PR tree-optimization/87465
1011 * tree-ssa-loop-ivcanon.c (tree_estimate_loop_size): Fix typo
1012 causing branch miscounts.
1014 2018-10-01 Tamar Christina <tamar.christina@arm.com>
1016 * common/config/aarch64/aarch64-common.c (TARGET_OPTION_DEFAULT_PARAM,
1017 aarch64_option_default_param): New.
1018 (params.h): Include.
1019 (TARGET_OPTION_VALIDATE_PARAM, aarch64_option_validate_param): New.
1020 * config/aarch64/aarch64.c (aarch64_override_options_internal): Simplify
1021 stack-clash protection validation code.
1023 2018-10-01 Tamar Christina <tamar.christina@arm.com>
1025 * params.c (validate_param): New.
1026 (add_params): Use it.
1027 (set_param_value): Refactor param validation into validate_param.
1028 (diagnostic.h): Include.
1029 * diagnostic.h (diagnostic_ready_p): New.
1031 2018-10-01 Tamar Christina <tamar.christina@arm.com>
1033 * params.c (set_param_value):
1034 Add index of parameter being validated.
1035 * common/common-target.def (option_validate_param): New.
1036 * common/common-targhooks.h (default_option_validate_param): New.
1037 * common/common-targhooks.c (default_option_validate_param): New.
1038 * doc/tm.texi.in (TARGET_OPTION_VALIDATE_PARAM): New.
1039 * doc/tm.texi: Regenerate.
1041 2018-10-01 Tamar Christina <tamar.christina@arm.com>
1044 * config/aarch64/aarch64.c (aarch64_override_options_internal):
1045 Add validation for stack-clash parameters and set defaults.
1047 2018-10-01 Tamar Christina <tamar.christina@arm.com>
1050 * configure.ac: Add stack-clash-protection-guard-size.
1051 * doc/install.texi: Document it.
1052 * config.in (DEFAULT_STK_CLASH_GUARD_SIZE): New.
1053 * params.def: Update comment for guard-size.
1054 (PARAM_STACK_CLASH_PROTECTION_GUARD_SIZE,
1055 PARAM_STACK_CLASH_PROTECTION_PROBE_INTERVAL): Update description.
1056 * configure: Regenerate.
1058 2018-10-01 Tamar Christina <tamar.christina@arm.com>
1061 * config/aarch64/aarch64.h (STACK_CLASH_MIN_BYTES_OUTGOING_ARGS,
1062 STACK_DYNAMIC_OFFSET): New.
1063 * config/aarch64/aarch64.c (aarch64_layout_frame):
1064 Update outgoing args size.
1065 (aarch64_stack_clash_protection_alloca_probe_range,
1066 TARGET_STACK_CLASH_PROTECTION_ALLOCA_PROBE_RANGE): New.
1068 2018-10-01 Tamar Christina <tamar.christina@arm.com>
1071 * explow.c (anti_adjust_stack_and_probe_stack_clash): Support custom
1073 * target.def (stack_clash_protection_alloca_probe_range): New.
1074 (stack_clash_protection_final_dynamic_probe): Remove.
1075 * targhooks.h (default_stack_clash_protection_alloca_probe_range) New.
1076 (default_stack_clash_protection_final_dynamic_probe): Remove.
1077 * targhooks.c: Likewise.
1078 * doc/tm.texi.in (TARGET_STACK_CLASH_PROTECTION_ALLOCA_PROBE_RANGE): New.
1079 (TARGET_STACK_CLASH_PROTECTION_FINAL_DYNAMIC_PROBE): Remove.
1080 * doc/tm.texi: Regenerate.
1082 2018-10-01 Tamar Christina <tamar.christina@arm.com>
1085 * config/aarch64/aarch64-protos.h (aarch64_output_probe_sve_stack_clash): New.
1086 * config/aarch64/aarch64.c (aarch64_output_probe_sve_stack_clash,
1087 aarch64_clamp_to_uimm12_shift): New.
1088 (aarch64_allocate_and_probe_stack_space): Add SVE specific section.
1089 * config/aarch64/aarch64.md (probe_sve_stack_clash): New.
1091 2018-10-01 Tamar Christina <tamar.christina@arm.com>
1094 * config/aarch64/aarch64.c (aarch64_layout_frame): Add assert.
1096 2018-10-01 Jeff Law <law@redhat.com>
1097 Richard Sandiford <richard.sandiford@linaro.org>
1098 Tamar Christina <tamar.christina@arm.com>
1101 * config/aarch64/aarch64.md
1102 (probe_stack_range): Add k (SP) constraint.
1103 * config/aarch64/aarch64.h (STACK_CLASH_CALLER_GUARD,
1104 STACK_CLASH_MAX_UNROLL_PAGES): New.
1105 * config/aarch64/aarch64.c (aarch64_output_probe_stack_range): Emit
1106 stack probes for stack clash.
1107 (aarch64_allocate_and_probe_stack_space): New.
1108 (aarch64_expand_prologue): Use it.
1109 (aarch64_expand_epilogue): Likewise and update IP regs re-use criteria.
1110 (aarch64_sub_sp): Add emit_move_imm optional param.
1112 2018-10-01 MCC CS <deswurstes@users.noreply.github.com>
1114 PR tree-optimization/87261
1115 * match.pd: Remove trailing whitespace.
1116 Add (x & y) | ~(x | y) -> ~(x ^ y),
1117 (~x | y) ^ (x ^ y) -> x | ~y and (x ^ y) | ~(x | y) -> ~(x & y)
1119 2018-10-01 Claudiu Zissulescu <claziss@synopsys.com>
1121 * config/arc/arc.md (*add_n): Clean up pattern, update instruction
1123 (ashlsi3_insn): Update instruction constraints.
1124 (ashrsi3_insn): Likewise.
1125 (rotrsi3): Likewise.
1126 (add_shift): Likewise.
1127 * config/arc/constraints.md (Csz): New 32 bit constraint. It
1128 avoids placing in the limm field small constants which, otherwise,
1129 could end into a small instruction.
1131 2018-10-01 Claudiu Zissulescu <claziss@synopsys.com>
1133 * config/arc/arc.md (maddsidi4_split): Don't use dmac if the
1134 destination register is not odd-even.
1135 (umaddsidi4_split): Likewise.
1137 2018-10-01 Richard Biener <rguenther@suse.de>
1139 * tree-inline.c (expand_call_inline): Store origin of fn
1140 in BLOCK_ABSTRACT_ORIGIN for the inline BLOCK.
1141 * tree.c (block_ultimate_origin): Simplify and do some
1144 2018-09-30 Uros Bizjak <ubizjak@gmail.com>
1146 * config/i386/mmx.md (EMMS): New int iterator.
1147 (emms): New int attribute.
1148 (mmx_<emms>): Macroize insn from *mmx_emms and *mmx_femms using
1149 EMMS int iterator. Explicitly declare clobbers.
1150 (mmx_emms): Remove expander.
1152 * config/i386/predicates.md (emms_operation): Remove predicate.
1153 (vzeroall_pattern): New predicate.
1154 (vzeroupper_pattern): Rename from vzeroupper_operation.
1155 * config/i386/i386.c (ix86_avx_u128_mode_after): Use
1156 vzeroupper_pattern and vzeroall_pattern predicates.
1158 2018-09-30 Peter Bergner <bergner@linux.ibm.com>
1160 PR rtl-optimization/86939
1161 * ira-lives.c (make_hard_regno_born): Rename from this...
1162 (make_hard_regno_live): ... to this. Remove update to conflict
1163 information. Update function comment.
1164 (make_hard_regno_dead): Add conflict information update. Update
1166 (make_object_born): Rename from this...
1167 (make_object_live): ... to this. Remove update to conflict information.
1168 Update function comment.
1169 (make_object_dead): Add conflict information update. Update function
1171 (mark_pseudo_regno_live): Call make_object_live.
1172 (mark_pseudo_regno_subword_live): Likewise.
1173 (mark_hard_reg_dead): Update function comment.
1174 (mark_hard_reg_live): Call make_hard_regno_live.
1175 (process_bb_node_lives): Likewise.
1176 * lra-lives.c (make_hard_regno_born): Rename from this...
1177 (make_hard_regno_live): ... to this. Remove update to conflict
1178 information. Remove now uneeded check_pic_pseudo_p argument.
1179 Update function comment.
1180 (make_hard_regno_dead): Add check_pic_pseudo_p argument and add update
1181 to conflict information. Update function comment.
1182 (mark_pseudo_live): Remove update to conflict information. Update
1184 (mark_pseudo_dead): Add conflict information update.
1185 (mark_regno_live): Call make_hard_regno_live.
1186 (mark_regno_dead): Call make_hard_regno_dead with new arguement.
1187 (process_bb_lives): Call make_hard_regno_live and make_hard_regno_dead.
1189 2018-09-29 H.J. Lu <hongjiu.lu@intel.com>
1192 * config/i386/i386.c (construct_container): Use TImode for
1193 BLKmode values in 2 integer registers.
1195 2018-09-29 Jeff Law <law@redhat.com>
1197 * builtins.c (unterminated_array): Pass in c_strlen_data * to
1198 c_strlen rather than just a tree *.
1199 (c_strlen): Change NONSTR argument to a c_strlen_data pointer.
1200 Update recursive calls appropriately. If caller did not provide a
1201 suitable data pointer, create a local one. When a non-terminated
1202 string is discovered, bubble up information about the string via the
1203 c_strlen_data object.
1204 * builtins.h (c_strlen): Update prototype.
1205 (c_strlen_data): New structure.
1206 * gimple-fold.c (get_range_strlen): Update calls to c_strlen.
1207 For a type 2 call, if c_strlen indicates a non-terminated string
1208 use the length of the non-terminated string.
1209 (gimple_fold_builtin_stpcpy): Update calls to c_strlen.
1211 2018-09-29 Jakub Jelinek <jakub@redhat.com>
1214 * config/i386/avx512fintrin.h (_mm512_abs_pd, _mm512_mask_abs_pd): Use
1215 __m512d type for __A argument rather than __m512.
1217 2018-09-28 John David Anglin <danglin@gcc.gnu.org>
1219 * match.pd (simple_comparison): Don't optimize if either operand is
1220 a function pointer when target needs function pointer canonicalization.
1222 2018-09-28 Segher Boessenkool <segher@kernel.crashing.org>
1224 * config/rs6000/driver-rs6000.c (asm_names): Adjust the entries for
1225 power5 .. power9 to remove indirection.
1226 * config/rs6000/rs6000.h (ASM_CPU_POWER5_SPEC, ASM_CPU_POWER6_SPEC,
1227 ASM_CPU_POWER7_SPEC, ASM_CPU_POWER8_SPEC, ASM_CPU_POWER9_SPEC,
1228 ASM_CPU_476_SPEC): Delete.
1229 (ASM_CPU_SPEC): Adjust.
1230 (EXTRA_SPECS): Delete asm_cpu_power5, asm_cpu_power6, asm_cpu_power7,
1231 asm_cpu_power8, asm_cpu_power9, asm_cpu_476.
1233 2018-09-28 Segher Boessenkool <segher@kernel.crashing.org>
1235 * config.in: Delete HAVE_AS_DCI.
1236 * config/powerpcspe/powerpcspe.h: Treat HAVE_AS_DCI as always true.
1237 * config/rs6000/rs6000.h: Ditto.
1238 * configure.ac: Delete HAVE_AS_DCI.
1239 * configure: Regenerate.
1241 2018-09-28 Segher Boessenkool <segher@kernel.crashing.org>
1243 * config.in (HAVE_AS_LWSYNC): Delete.
1244 * config/powerpcspe/powerpcspe.h (TARGET_LWSYNC_INSTRUCTION): Delete.
1245 * config/powerpcspe/sync.md (*lwsync): Always generate lwsync, never
1247 * config/rs6000/rs6000.h (TARGET_LWSYNC_INSTRUCTION): Delete.
1248 * config/rs6000/sync.md (*lwsync): Always generate lwsync, never do it
1250 * configure.ac: Delete HAVE_AS_LWSYNC.
1251 * configure: Regenerate.
1253 2018-09-28 Eric Botcazou <ebotcazou@adacore.com>
1254 Pierre-Marie de Rodat <derodat@adacore.com>
1256 * calls.c (expand_call): Try to do a tail call for thunks at -O0 too.
1257 * cgraph.h (struct cgraph_thunk_info): Add indirect_offset.
1258 (cgraph_node::create_thunk): Add indirect_offset parameter.
1259 (thunk_adjust): Likewise.
1260 * cgraph.c (cgraph_node::create_thunk): Add indirect_offset parameter
1261 and initialize the corresponding field with it.
1262 (cgraph_node::dump): Dump indirect_offset field.
1263 * cgraphclones.c (duplicate_thunk_for_node): Deal with indirect_offset.
1264 * cgraphunit.c (cgraph_node::analyze): Be prepared for external thunks.
1265 (thunk_adjust): Add indirect_offset parameter and deal with it.
1266 (cgraph_node::expand_thunk): Deal with the indirect_offset field and
1267 pass it to thunk_adjust. Do not call the target hook if it's non-zero
1268 or if the thunk is external or local. Fix formatting. Do not chain
1269 the RESULT_DECL to BLOCK_VARS. Pass the static chain to the target,
1270 if any, in the GIMPLE representation.
1271 * ipa-icf.c (sem_function::equals_wpa): Deal with indirect_offset.
1272 * lto-cgraph.c (lto_output_node): Write indirect_offset field.
1273 (input_node): Read indirect_offset field.
1274 * tree-inline.c (expand_call_inline): Pass indirect_offset field in the
1275 call to thunk_adjust.
1276 * tree-nested.c (struct nesting_info): Add thunk_p field.
1277 (create_nesting_tree): Set it.
1278 (convert_all_function_calls): Copy static chain from targets to thunks.
1279 (finalize_nesting_tree_1): Return early for thunks.
1280 (unnest_nesting_tree_1): Do not finalize thunks.
1281 (gimplify_all_functions): Do not gimplify thunks.
1283 2018-09-28 David Malcolm <dmalcolm@redhat.com>
1285 * opt-suggestions.c (option_proposer::build_option_suggestions):
1286 Release "option_values".
1288 2018-09-28 David Malcolm <dmalcolm@redhat.com>
1290 * coverage.c (get_coverage_counts): Convert problem-reporting dump
1291 messages from MSG_OPTIMIZED_LOCATIONS to MSG_MISSED_OPTIMIZATION.
1292 * dumpfile.c (kind_as_string): New function.
1293 (dump_loc): Rather than a hardcoded prefix of "note: ", use
1294 kind_as_string to vary the prefix based on dump_kind.
1295 (selftest::test_capture_of_dump_calls): Update for above.
1297 2018-09-28 Uros Bizjak <ubizjak@gmail.com>
1299 * config/i386/i386.h (SSE_REGNO): Fix check for FIRST_REX_SSE_REG.
1300 (GET_SSE_REGNO): Rename from SSE_REGNO. Update all uses for rename.
1302 2018-09-28 Uros Bizjak <ubizjak@gmail.com>
1304 * config/i386/i386.h (CC_REGNO): Remove FPSR_REG.
1305 * config/i386/i386.c (ix86_fixed_condition_code_regs): Use
1306 INVALID_REGNUM instead of FPSR_REG.
1307 (ix86_md_asm_adjust): Do not clobber FPSR_REG.
1308 * config/i386/i386.md: Update comment of FP compares.
1309 (fldenv): Do not clobber FPSR_REG.
1311 2018-09-28 Richard Biener <rguenther@suse.de>
1313 * tree.h (BLOCK_ORIGIN): New.
1314 * omp-expand.c (grid_expand_target_grid_body): Assign
1315 BLOCK_ORIGIN to BLOCK_ABSTRACT_ORIGIN.
1316 * tree-inline.c (remap_block): Likewise.
1317 * auto-profile.c (get_function_decl_from_block): Simplify
1318 by eliding the BLOCK_ABSTRACT_ORIGIN chasing.
1319 * langhooks.c (lhd_print_error_function): Likewise.
1320 * optinfo-emit-json.cc (optrecord_json_writer::inlining_chain_to):
1322 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
1323 * tree.c (block_nonartificial_location): Likewise.
1324 (block_ultimate_origin): Likewise.
1325 * tree-pretty-print.c (percent_K_format): Likewise. Remove
1326 no longer needed LTO case.
1328 2018-09-28 Andrew Stubbs <ams@codesourcery.com>
1329 Jan Hubicka <jh@suse.cz>
1330 Martin Jambor <mjambor@suse.cz>
1332 * simplify-rtx.c (simplify_merge_mask): New function.
1333 (simplify_ternary_operation): Use it, also see if VEC_MERGEs with the
1334 same masks are used in op1 or op2.
1335 (test_vec_merge): New function.
1336 (test_vector_ops): Call test_vec_merge.
1338 2018-09-28 Eric Botcazou <ebotcazou@adacore.com>
1340 * config/sparc/sparc-protos.h (sparc_branch_cost): Declare.
1341 * config/sparc/sparc.h (BRANCH_COST): Call sparc_branch_cost.
1342 * config/sparc/sparc.c (struct processor_costs): Add branch_cost field.
1343 (cypress_costs): Set it.
1344 (supersparc_costs): Likewise.
1345 (hypersparc_costs): Likewise.
1346 (leon_cost): Likewise.
1347 (leon3_costs): Likewise.
1348 (sparclet_costs): Likewise.
1349 (ultrasparc_costs): Likewise.
1350 (ultrasparc_costs): Likewise.
1351 (niagara_costs): Likewise.
1352 (niagara2_costs): Likewise.
1353 (niagara3_costs): Likewise.
1354 (niagara4_costs): Likewise.
1355 (niagara7_costs): Likewise.
1356 (m8_costs): Likewise.
1357 (TARGET_CAN_FOLLOW_JUMP): Define.
1358 (pass_work_around_errata::gate): Minor tweak.
1359 (sparc_option_override): Remove MASK_FSMULD mask for V7 processors.
1360 Do not set both MASK_VIS4 and MASK_VIS4B for M8 processor.
1361 Automaitcally clear MASK_FSMULD mask for V7 processors.
1362 (sparc_can_follow_jump): New static function.
1363 (output_ubranch): Deal with CROSSING_JUMP_P.
1364 (sparc_use_sched_lookahead): Rewrite using switch statement.
1365 (sparc_issue_rate): Reorder.
1366 (sparc_branch_cost): New function.
1368 2018-09-27 Martin Sebor <msebor@redhat.com>
1370 * tree.h (tree_to_shwi): Add attribute nonnull and pure.
1371 (tree_to_poly_int64, tree_to_uhwi, tree_to_poly_uint64): Same.
1372 (int_fits_type_p): Same.
1374 2018-09-27 Uros Bizjak <ubizjak@gmail.com>
1376 * config/i386/i386.md (FPCR_REG): Remove.
1377 (UNSPEC_FLDCW): Remove.
1378 (x86_fnstcw_1): Use (const_int 0) instead of FPCR_REG.
1379 (x86_fldcw_1): Remove insn pattern.
1380 (fnstenv): Do not clobber FPCR_REG.
1382 * config/i386/i386.h (FIXED_REGISTERS) Remove fpsr register.
1383 (CALL_USED_REGISTERS): Ditto.
1384 (REG_ALLOC_ORDER): Ditto.
1385 (REG_CLASS_CONTENTS): Ditto.
1386 (HI_REGISTER_NAMES): Ditto.
1387 (ADDITIONAL_REGISTER_NAMES): Use defines instead
1388 of numerical constants.
1389 * config/i386/i386.c (regclass_map): Remove fpsr register.
1390 (dbx_register_map): Ditto.
1391 (dbx64_register_map): Ditto.
1392 (svr4_dbx_register_map): Ditto.
1393 (print_reg): Do not handle FPCR_REG.
1395 2018-09-27 Segher Boessenkool <segher@kernel.crashing.org>
1398 * config.in (HAVE_AS_CMPB, HAVE_AS_DFP, HAVE_AS_FPRND, HAVE_AS_MFPGPR,
1399 HAVE_AS_POPCNTB, HAVE_AS_POPCNTD, HAVE_AS_POWER8, HAVE_AS_POWER9):
1400 Delete, always treat as true.
1401 * config/powerpcspe/powerpcspe.c (rs6000_option_override_internal):
1402 Ditto. Simplify remaining code.
1403 * config/powerpcspe/powerpcspe.h: Ditto.
1404 * config/rs6000/rs6000.c (rs6000_option_override_internal): Ditto.
1405 Simplify remaining code.
1406 (rs6000_expand_builtin): Ditto.
1407 * config/rs6000/rs6000.h: Ditto.
1408 * configure.ac: Ditto.
1409 * configure: Regenerate.
1411 2018-09-27 Martin Liska <mliska@suse.cz>
1413 * coverage.c (get_coverage_counts): Revert the formatting
1414 of missing profile opt info.
1416 2018-09-27 Richard Biener <rguenther@suse.de>
1420 * dwarf2out.c (set_block_origin_self): Do not mark outermost
1421 block as we do not output that.
1422 (gen_inlined_subroutine_die): Elide the originally outermost
1423 block, matching what we do for concrete instances.
1424 (decls_for_scope): Add parameter specifying whether to recurse
1427 2018-09-27 Andrew Stubbs <ams@codesourcery.com>
1428 Tom de Vries <tom@codesourcery.com>
1432 * expmed.c (emit_cstore): Fix handling of result_mode == BImode and
1433 STORE_FLAG_VALUE == 1.
1435 2018-09-27 Andreas Krebbel <krebbel@linux.ibm.com>
1437 * config/s390/s390.md (PPA_TX_ABORT, PPA_OOO_BARRIER): New
1438 constant definitions.
1439 ("tx_assist"): Replace magic number with PPA_TX_ABORT.
1440 ("*ppa"): Enable pattern also for -march=zEC12 -mno-htm.
1441 ("speculation_barrier"): New expander definition.
1443 2018-09-26 Indu Bhagat <indu.bhagat@oracle.com>
1445 PR gcov-profile/86957
1446 * common.opt: New warning option -Wmissing-profile.
1447 * coverage.c (get_coverage_counts): Add warning for missing .gcda file.
1448 * doc/invoke.texi: Document -Wmissing-profile.
1450 2018-09-26 Jim Wilson <jimw@sifive.com>
1452 * config/riscv/riscv.md (subsi3_extended2): Add J constraint.
1453 (negdi2, negsi2, negsi2_extended, negsi2_extended2): New.
1455 2018-09-26 Martin Sebor <msebor@redhat.com>
1457 * tree.c (zerop): Change return type to bool.
1458 (integer_zerop, integer_onep, integer_each_onep): Same.
1459 (integer_all_onesp, integer_minus_onep, integer_pow2p): Same.
1460 (integer_nonzerop, integer_truep, tree_ctz, real_zerop): Same.
1461 (real_onep, real_minus_onep, chain_index): Same.
1462 (print_type_hash_statistics, type_list_equal): Same.
1463 * tree.h (zerop): Same.
1464 (zerop, integer_zerop, integer_onep, integer_each_onep): Same.
1465 (integer_all_onesp, integer_minus_onep, integer_pow2p): Same.
1466 (integer_nonzerop, integer_truep, tree_ctz, real_zerop): Same.
1467 (real_onep, real_minus_onep, chain_index): Same.
1468 (print_type_hash_statistics, type_list_equal): Same.
1470 2018-09-26 Jim Wilson <jimw@sifive.com>
1472 * config/riscv/riscv.h (FUNCTION_ARG_REGNO_P): Fix comment.
1474 2018-09-26 Jakub Jelinek <jakub@redhat.com>
1477 * config/i386/i386.c: Include debug.h and dwarf2out.h.
1478 (output_indirect_thunk): Emit DW_CFA_def_cfa_offset after the
1481 2018-09-25 Andrew Stubbs <ams@codesourcery.com>
1483 * builtins.c (get_builtin_sync_mem): Force address mode conversion.
1485 2018-09-26 Uros Bizjak <ubizjak@gmail.com>
1487 * config/i386/i386.h (enum reg_class): Remove FP_TOP_SSE_REGS
1488 and FP_SECOND_SSE_REGS.
1489 (REG_CLASS_NAMES): Ditto.
1490 (REG_CLASS_CONTENTS): Ditto.
1491 * config/i386/i386.c (ix86_preferred_reload_class) Do not handle
1492 FP_TOP_SSE_REGS and FP_SECOND_SSE_REGS classes.
1493 (ix86_preferred_output_reload_class): Ditto.
1494 * config/i386/i386.md (fix_trunc<mode>_i387_fisttp): Change "=&1f"
1495 clobber constraint to "=&f".
1496 (fix_truncdi_i387): Ditto.
1497 (lrintxfdi2): Ditto.
1498 (fistdi2_<rounding>): Ditto.
1499 (fpremxf4_i387): Change "=u" constraint to "=f".
1500 (fprem1xf4_i387): Ditto.
1502 (fptanxf4_i387): Ditto.
1503 (fxtractxf3_i387): Ditto.
1504 (fscalexf4_i387): Ditto.
1505 (atan2xf3): Change "u" constraint to "f".
1506 (fyl2xxf3_i387): Ditto.
1507 (fyl2xp1xf3_i387): Ditto.
1509 2018-09-26 Uros Bizjak <ubizjak@gmail.com>
1512 * config/i386/i386.h (NUM_MODES_FOR_MODE_SWITCHING): Update
1513 for removed I387_MASK_PM entity.
1516 2018-09-26 Jeff Law <law@redhat.com>
1518 2018-09-26 Alexey Neyman <stilor@att.net>
1520 * graphite.h: Include <isl/id.h> and <isl/space.h>; these
1521 headers are no longer pulled in by <isl/val.h>.
1523 2018-09-26 Richard Biener <rguenther@suse.de>
1526 * dwarf2out.c (gen_lexical_block_die): Do not equate inline
1527 or concrete instance DIE to the tree. Create abstract origin
1528 attributes also for concrete instances.
1530 2018-09-26 Alexey Neyman <stilor@att.net>
1532 * graphite.h: Include <isl/id.h> and <isl/space.h>; these
1533 headers are no longer pulled in by <isl/val.h>.
1535 2018-09-26 Matthew Malcomson <matthew.malcomson@arm.com>
1537 * config/arm/arm.c (arm_split_compare_and_swap, arm_split_atomic_op):
1538 Use new helper functions.
1539 * config/arm/sync.md (atomic_load<mode>, atomic_store<mode>):
1540 Use new helper functions.
1541 * config/arm/aarch-common-protos.h (aarch_mm_needs_acquire,
1542 aarch_mm_needs_release): New declarations.
1543 * config/arm/aarch-common.c (aarch_mm_needs_acquire,
1544 aarch_mm_needs_release): New.
1546 2018-09-26 Eric Botcazou <ebotcazou@adacore.com>
1548 * config/arm/arm.c (arm_reorg): Skip Thumb reorg pass for thunks.
1549 (arm32_output_mi_thunk): Deal with long calls.
1551 2018-09-26 Richard Biener <rguenther@suse.de>
1555 * tree-inline.c (expand_call_inline): When the location
1556 of the call is UNKNOWN_LOCATION use DECL_SOURCE_LOCATION
1557 or BUILTINS_LOCATION for the BLOCK_SOURCE_LOCATION of
1558 the inserted BLOCK to make inlined_function_outer_scope_p
1560 * dwarf2out.c (add_call_src_coords_attributes): Do not add
1561 coords for reserved locations.
1563 2018-09-25 Segher Boessenkool <segher@kernel.crashing.org>
1565 * config/rs6000/rs6000.md (*movcc_internal1): Use set_attr_alternative.
1566 (*call_indirect_nonlocal_sysv<mode>): Ditto.
1567 (*call_value_indirect_nonlocal_sysv<mode>): Ditto.
1568 (*sibcall_nonlocal_sysv<mode>): Ditto.
1569 (*sibcall_value_nonlocal_sysv<mode>): Ditto.
1570 (<bd>_<mode>): Ditto.
1571 (<bd>tf_<mode>): Ditto.
1573 2018-09-25 Segher Boessenkool <segher@kernel.crashing.org>
1575 * config/rs6000/altivec.md (*altivec_mov<mode>): Write the output
1576 control string as a list of templates instead of as C code.
1577 (*altivec_movti): Ditto.
1578 * config/rs6000/darwin.md (movdf_low_di): Ditto.
1580 2018-09-25 Jim Wilson <jimw@sifive.com>
1582 * config/riscv/riscv.c (riscv_split_symbol): Mark auipc label as weak
1583 when target symbol is weak.
1585 2018-09-25 Bernd Edlinger <bernd.edlinger@hotmail.de>
1588 * builtins.c (unterminated_array): Simplify.
1589 * expr.c (string_constant): Handle SSA_NAME. Add more exceptions
1590 where pointer arithmetic is safe.
1592 2018-09-25 Segher Boessenkool <segher@kernel.crashing.org>
1595 * config/rs6000/altivec.md (altivec_vspltb): Use
1596 const_0_to_15_operand instead of u5bit_cint_operand.
1597 (*altivec_vspltb_internal): Ditto.
1598 (altivec_vspltb_direct): Ditto.
1599 (altivec_vsplth): Use const_0_to_7_operand instead of
1601 (*altivec_vsplth_internal): Ditto.
1602 (altivec_vsplth_direct): Ditto.
1603 (altivec_vspltw): Use const_0_to_3_operand instead of
1605 (*altivec_vspltw_internal): Ditto.
1606 (altivec_vspltw_direct): Ditto.
1607 (altivec_vspltsf): Ditto.
1608 (*altivec_vspltsf_internal): Ditto.
1609 * config/rs6000/rs6000.c (rs6000_expand_binop_builtin): Handle the
1610 various splats with the proper size immediate. Reorder the various
1611 cases by ascending size of immediate, and put all such together.
1613 2018-09-25 Richard Biener <rguenther@suse.de>
1616 * dwarf2out.c (add_AT_external_die_ref): Remove now redundant
1618 (maybe_create_die_with_external_ref): Do not create
1619 DW_TAG_imported_unit here.
1620 (add_abstract_origin_attribute): Handle external BLOCK refs.
1621 (dwarf2out_abstract_function): Simplify LTO case.
1622 (dwarf2out_early_finish): Create DW_TAG_imported_unit explicitely
1623 rather than using maybe_create_die_with_external_ref.
1625 2018-09-25 Uros Bizjak <ubizjak@gmail.com>
1628 * config/i386/i386.md (frndintxf2_mask_pm): Remove.
1629 (frndintxf2_mask_pm_i387): Ditto.
1630 (nearbyintxf2): Rewrite expander pattern to match rintxf2.
1631 Enable for !flag_trapping_math.
1632 (nearbyint<mode>2): Enable x87 modes for !flag_trapping_math.
1633 Enable SSE modes for TARGET_SSE4_1 and expand them with round insn.
1634 Change operand 1 predicate to nonimmediate_operand.
1635 (attr "i387_cw"): Remove mask_pm.
1636 * config/i386/i386.h (enum ix86_stack_slot): Remove SLOT_CW_MASK_PM.
1637 (enum ix86_entity): Remove I387_MASK_PM.
1638 * config/i386/i386.c (ix86_i387_mode_needed): Do not
1639 handle I387_MASK_PM.
1640 (ix86_mode_needed): Ditto.
1641 (ix86_mode_after): Ditto.
1642 (ix86_mode_entry): Ditto.
1643 (ix86_mode_exit): Ditto.
1644 (emit_i387_cw_initialization): Do not handle I387_CW_MASK_PM.
1646 2018-09-25 Jakub Jelinek <jakub@redhat.com>
1648 * vr-values.c (vr_values::vr_values): Initialize to_remove_edges and
1649 to_update_switch_stmts to vNULL instead of calling create on them
1652 2018-09-25 Richard Biener <rguenther@suse.de>
1654 PR tree-optimization/87402
1655 * tree-ssa-sccvn.c (SSA_VISITED): Remove unused function.
1656 (visit_phi): Re-instantiate handling of supposed to be VARYING
1657 but non-VARYING backedge value.
1659 2018-09-25 Richard Biener <rguenther@suse.de>
1662 * dwarf2out.c (struct sym_off_pair): New.
1663 (external_die_map): New global.
1664 (lookup_decl_die): When in LTO create DIEs lazily from the
1666 (lookup_block_die): New function, create DIEs lazily in LTO.
1667 (equate_block_to_die): New function.
1668 (dwarf2out_die_ref_for_decl): During WPA get the association
1669 from the external DIE map.
1670 (dwarf2out_register_external_die): Record mapping into the
1672 (maybe_create_die_with_external_ref): New function split out from
1673 DIE generation part of old dwarf2out_register_external_die.
1674 (add_abstract_origin_attribute): Do not return the DIE. When
1675 in LTO reference externals directly.
1676 (dwarf2out_abstract_function): When in LTO ignore calls for
1677 decls with external DIEs (already present abstract instances).
1678 (gen_call_site_die): Adjust.
1679 (add_high_low_attributes): Likewise.
1680 (gen_lexical_block_die): Likewise.
1681 (gen_inlined_subroutine_die): Likewie.
1682 (gen_block_die): Likewise.
1683 (dwarf2out_inline_entry): Likewise.
1684 (dwarf2out_early_finish): In LTRANS phase create DW_TAG_imported_unit
1687 2018-09-25 Martin Liska <mliska@suse.cz>
1689 * ipa-fnsummary.c (estimate_node_size_and_time): Scale by two
1690 integers and not by a float value.
1692 2018-09-25 Martin Liska <mliska@suse.cz>
1695 * dbgcnt.c (dbg_cnt_process_single_pair): Return false
1697 * dumpfile.c (dump_enable_all): Remove extra parenthesis.
1698 * gcov-tool.c: Declare the function with ATTRIBUTE_NORETURN.
1699 * godump.c (go_format_type): Remove extra parenthesis.
1701 2018-09-25 Martin Liska <mliska@suse.cz>
1703 * alias.c (set_dest_equal_p): Remove unused function.
1704 * config/i386/i386.c (def_builtin_pure2): Likewise.
1705 * diagnostic-show-locus.c (class layout): Remove
1707 (layout::layout): Likewise here.
1708 * dump-context.h (class temp_dump_context): Likewise.
1709 * dwarf2out.c (add_AT_fde_ref): Remove unused function.
1710 (add_AT_loclistsptr): Likewise.
1711 (add_AT_offset): Likewise.
1712 (get_AT_hi_pc): Likewise.
1713 (is_comdat_die): Likewise.
1714 (type_is_enum): Likewise.
1715 (ceiling): Likewise.
1716 (add_AT_vms_delta): Likewise.
1717 (is_class_die): Likewise.
1718 * edit-context.c (class line_event): Remove unused field.
1719 * graphite-sese-to-poly.c (tree_int_to_gmp): Remove
1721 * ipa-cp.c (ipa_get_vr_lat): Likewise.
1722 * lra-constraints.c (ok_for_index_p_nonstrict): Likewise.
1723 (ok_for_base_p_nonstrict): Likewise.
1724 * tree-chrec.c (is_not_constant_evolution): Likewise.
1725 (chrec_fold_poly_cst): Likewise.
1726 * tree-if-conv.c (has_pred_critical_p): Likewise.
1727 * tree-ssa-coalesce.c (print_exprs): Likewise.
1728 * tree-ssa-pre.c (bitmap_set_contains_expr): Likewise.
1729 * tree-ssa-uninit.c (is_and_or_or_p): Likewise.
1730 * tree-vrp.c (value_ranges_intersect_p): Likewise.
1731 (value_range_nonnegative_p): Likewise.
1733 2018-09-25 Martin Liska <mliska@suse.cz>
1735 * config/powerpcspe/powerpcspe.c (rs6000_output_function_epilogue):
1736 Do not handle "GNU Pascal".
1737 * config/rs6000/rs6000.c (rs6000_output_function_epilogue):
1739 * config/sparc/sparc.c (sparc_pass_by_reference): Remove Pascal
1740 from documentation. Likewise.
1741 * dbxout.c (dbxout_range_type): Likewise.
1742 * doc/cpp.texi: Likewise.
1743 * doc/extend.texi: Likewise.
1744 * doc/frontends.texi: Likewise.
1745 * doc/invoke.texi: Remove Pascal entry.
1746 * tree.def (CLEANUP_POINT_EXPR): Likewise.
1747 * doc/rtl.texi (MODE_FUNCTION): Remove not used entry.
1749 2018-09-25 Martin Liska <mliska@suse.cz>
1752 * doc/invoke.texi: Document all parameters and remove default
1755 2018-09-25 Ilya Leoshkevich <iii@linux.ibm.com>
1758 * rtl.c (rtx_code_size): Take into account that EQ_ATTR_ALT
1759 contains HOST_WIDE_INTs when computing its size.
1761 2018-09-24 Jim Wilson <jimw@sifive.com>
1764 * config/riscv/riscv.h (STACK_BOUNDARY): Test riscv_abi == ABI_ILP32E
1766 (ABI_STACK_BOUNDARY, MAX_ARGS_IN_REGISTERS): Likewise.
1768 2018-09-24 Andrew Pinski <apinski@marvell.com>
1770 * config/aarch64/aarch64.c (aarch_macro_fusion_pair_p): Don't
1771 access prev before checking it for NULLness in the
1772 AARCH64_FUSE_CMP_BRANCH case.
1774 2018-09-24 H.J. Lu <hongjiu.lu@intel.com>
1777 * config/i386/i386.c (rest_of_insert_endbranch): Set
1778 endbr_queued_at_entrance to true and don't insert ENDBR if
1779 x86_function_profiler will be called.
1780 (x86_function_profiler): Insert ENDBR if endbr_queued_at_entrance
1782 * config/i386/i386.h (machine_function): Add
1783 endbr_queued_at_entrance.
1785 2018-09-24 Ilya Leoshkevich <iii@linux.ibm.com>
1787 * genattrtab.c (mk_attr_alt): Use alternative_mask.
1788 (attr_rtx_1): Adjust caching to match the new EQ_ATTR_ALT field
1790 (check_attr_test): Use alternative_mask.
1791 (get_attr_value): Likewise.
1792 (compute_alternative_mask): Use alternative_mask and XWINT.
1793 (make_alternative_compare): Use alternative_mask.
1794 (attr_alt_subset_p): Use XWINT.
1795 (attr_alt_subset_of_compl_p): Likewise.
1796 (attr_alt_intersection): Use alternative_mask and XWINT.
1797 (attr_alt_union): Likewise.
1798 (attr_alt_complement): Use HOST_WIDE_INT and XWINT.
1799 (mk_attr_alt): Use alternative_mask and HOST_WIDE_INT.
1800 (simplify_test_exp): Use alternative_mask and XWINT.
1801 (write_test_expr): Use alternative_mask and XWINT, adjust bit
1802 number calculation to support 64 bits. Generate code that
1803 checks 64-bit masks.
1804 (main): Use alternative_mask.
1805 * rtl.def (EQ_ATTR_ALT): Change field types from ii to ww.
1807 2018-09-24 Ilya Leoshkevich <iii@linux.ibm.com>
1810 * config/s390/s390.c (s390_emit_epilogue): Do not use PARALLEL
1811 RETURN+USE when returning via %r14.
1813 2018-09-24 Martin Liska <mliska@suse.cz>
1815 * gcov.c (output_lines): Print colorization legend
1816 for both flag_use_colors and flag_use_hotness_colors.
1819 2018-09-24 Martin Liska <mliska@suse.cz>
1821 * coverage.c (get_coverage_counts): Use warning_at
1822 with current_function_decl location. Use %qD in warning
1825 2018-09-24 Martin Liska <mliska@suse.cz>
1827 * memory-block.h (memory_block_pool::release): Annotate with
1828 valgrind that the memory is not accessible.
1830 2018-09-24 Martin Liska <mliska@suse.cz>
1833 * asan.c: Make asan_handled_variables extern.
1835 * cfgexpand.c (expand_stack_vars): Make sure
1836 a representative is unpoison if another
1837 variable in the partition is handled by
1838 use-after-scope sanitization.
1840 2018-09-24 Richard Biener <rguenther@suse.de>
1842 PR tree-optimization/63155
1843 * tree-ssa-propagate.c (add_ssa_edge): Avoid adding PHIs to
1844 the worklist when the edge of the respective argument isn't
1847 2018-09-23 Uros Bizjak <ubizjak@gmail.com>
1849 * config/i386/i386.h (enum reg_class): Rename MASK_REGS to
1850 ALL_MASK_REGS and MASK_EVEX_REGS to MASK_REGS.
1851 (MASK_CLASS_P): Update for rename.
1852 (MAYBE_MASK_CLASS_P): Ditto.
1853 (REG_CLASS_NAMES): Update.
1854 (REG_CLASS_CONTENT): Update.
1855 * config/i386/i386.c (regclass_map): Update for MASK_REG
1856 and ALL_MASK_REGS rename.
1857 * config/i386/constraints.md (Yk): Update for rename.
1860 2018-09-23 Uros Bizjak <ubizjak@gmail.com>
1862 * config/i386/i386.h (enum reg_class): Remove
1863 EVEX_SSE_REGS and MOD4_SSE_REGS.
1864 (REG_CLASS_NAMES): Update.
1865 (REG_CLASS_CONTENT): Update.
1866 * config/i386/i386.c (regclass_map): Declare AVX-512 SSE
1867 registers as ALL_SSE_REGS.
1868 (ix86_additional_allocno_class_p): Remove.
1869 (TARGET_ADDITIONAL_ALLOCNO_CLASS_P): Remove.
1870 (ix86_register_priority): Lower priority of EVEX SSE registers.
1871 Use IN_RANGE macro where appropriate.
1872 (ix86_hard_regno_mode_ok): Merge AVX-5124FMAPS and
1873 AVX-5124VNNIW checks.
1874 (ix86_modes_tieable_p): Tie 512-bit SSE modes.
1875 * config/i386/sse.md (avx5124fmaddps_4fmaddps)
1876 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddps_maskz)
1877 (avx5124fmaddps_4fmaddss, avx5124fmaddps_4fmaddss_mask)
1878 (avx5124fmaddps_4fmaddss_maskz, avx5124fmaddps_4fnmaddps)
1879 (avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddps_maskz)
1880 (avx5124fmaddps_4fnmaddss, avx5124fmaddps_4fnmaddss_mask)
1881 (avx5124fmaddps_4fnmaddss_maskz, avx5124vnniw_vp4dpwssd)
1882 (avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssd_maskz)
1883 (avx5124vnniw_vp4dpwssds, avx5124vnniw_vp4dpwssds_mask)
1884 (avx5124vnniw_vp4dpwssds_maskz): Use "v" instead of "Yh" constraint.
1885 * config/i386/constraints.md (Yh): Remove.
1887 2018-09-23 Uros Bizjak <ubizjak@gmail.com>
1889 * config/i386/i386.c (regclass_map): Declare integer REX registers
1892 2018-09-23 Gerald Pfeifer <gerald@pfeifer.com>
1894 * doc/service.texi (Service): Switch the fsf.org link to https.
1896 2018-09-22 Chung-Ju Wu <jasonwucj@gmail.com>
1899 * config/nds32/nds32.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
1900 Define to speculation_safe_value_not_needed.
1902 2018-09-21 Florian Weimer <fweimer@redhat.com>
1905 * doc/extend.texi (Common Function Attributes): Mention that
1906 noreturn suppresses tail call optimization.
1908 2018-09-21 Jeff Law <law@redhat.com>
1910 * gimple-ssa-evrp.c (evrp_dom_walker::cleanup): Call
1911 vr_values::cleanup_edges_and_switches.
1912 * tree-vrp.c (to_remove_edges, to_update_switch_stmts): Moved into
1914 (identify_jump_threads): Remove EDGE_IGNORE handling.
1915 (execute_vrp): Move handling of to_remove_edges and
1916 to_update_switch_stmts into vr_values class member functions.
1917 * tree-vrp.h (switch_update, to_remove_edges): Remove declarations.
1918 (to_update_switch_stmts): Likewise.
1919 * vr-values.c: Include cfghooks.h.
1920 (vr_values::vr_values): Initialize to_remove_edges and
1921 to_update_switch_stmts.
1922 (vr_values::~vr_values): Verify to_remove_edges and
1923 to_update_switch_stmts are empty.
1924 (vr_values::simplify_switch_using_ranges): Set EDGE_IGNORE as needed.
1925 (vr_values::cleanup_edges_and_switches): New member function.
1926 * vr-values.h (vr_values): Add cleanup_edges_and_switches member
1927 function. Add new data members.
1929 2018-09-21 David Malcolm <dmalcolm@redhat.com>
1931 PR tree-optimization/87309
1932 * dumpfile.c (dump_context::begin_scope): Filter the dump_loc
1933 calls with pflags and alt_flags.
1934 (selftest::test_capture_of_dump_calls): Add test of interaction of
1935 MSG_OPTIMIZED_LOCATIONS with AUTO_DUMP_SCOPE.
1937 2018-09-21 Olivier Hainque <hainque@adacore.com>
1939 * config.gcc: Factorize and comment inclusion of vxworks-dummy.h.
1941 2018-09-21 Olivier Hainque <hainque@adacore.com>
1943 * config/vxworks.h (CLEAR_INSN_CACHE): #define to 1.
1945 2018-09-21 Olivier Hainque <hainque@adacore.com>
1947 * config/vxworks.h (VXWORKS_LIBGCC_SPEC): Remove -lc_internal.
1948 Merge block comment with the one ahead of VXWORKS_LIBS_RTP. Then:
1949 (VXWORKS_LIBS_RTP): Minor reordering.
1951 2018-09-21 Olivier Hainque <hainque@adacore.com>
1953 * config/vxworks.h (STARTFILE_PREFIX_SPEC): Define.
1954 (VXWORKS_LIBS_DIR_RTP): Remove definition and use.
1956 2018-09-21 Olivier Hainque <hainque@adacore.com>
1958 * config/vxworks.h (SIZE_TYPE): Account for TARGET_VXWORKS64.
1959 (PTRDIFF_TYPE): Likewise.
1961 2018-09-21 Olivier Hainque <hainque@adacore.com>
1963 * config.gcc: Enforce def of TARGET_VXWORKS64 to 1 from
1964 triplet, similar to support for VxWorks7.
1965 * config/vxworks-dummy.h: Provide a default definition
1966 of TARGET_VXWORKS64 to 0.
1968 2018-09-21 Olivier Hainque <hainque@adacore.com>
1970 * config/vxworks.h (TARGET_VXWORKS7): Move default definition ...
1971 * config/vxworks-dummy.h: here.
1973 2018-09-21 Olivier Hainque <hainque@adacore.com>
1975 * config.gcc: Prepend vxworks-dummy.h to tm_file for powerpc*
1977 2018-09-21 Shaokun Zhang <zhangshaokun@hisilicon.com>
1978 Bo Zhou <zbo.zhou@hisilicon.com>
1980 * config/aarch64/aarch64-cores.def (tsv110): New CPU.
1981 * config/aarch64/aarch64-tune.md: Regenerated.
1982 * doc/invoke.texi (AArch64 Options/-mtune): Add "tsv110".
1983 * config/aarch64/aarch64.c (tsv110_tunings): New tuning table.
1984 * config/aarch64/aarch64-cost-tables.h: Add "tsv110" extra costs.
1986 2018-09-21 Andrew Stubbs <ams@codesourcery.com>
1987 Julian Brown <julian@codesourcery.com>
1989 * builtins.c (get_builtin_sync_mem): Handle address spaces.
1991 2018-09-21 Eric Botcazou <ebotcazou@adacore.com>
1993 * config/rs6000/rs6000.c (rs6000_function_ok_for_sibcall): Return false
1994 if the call takes a static chain.
1996 2018-09-21 Martin Liska <mliska@suse.cz>
1998 * auto-profile.c (autofdo_source_profile::read): Do not
2000 (read_profile): Do not add working sets.
2001 (read_autofdo_file): Remove sum_all.
2002 (afdo_callsite_hot_enough_for_early_inline): Remove const
2004 * coverage.c (struct counts_entry): Remove gcov_summary.
2005 (read_counts_file): Read new GCOV_TAG_OBJECT_SUMMARY,
2006 do not support GCOV_TAG_PROGRAM_SUMMARY.
2007 (get_coverage_counts): Remove summary and expected
2009 * coverage.h (get_coverage_counts): Likewise.
2010 * doc/gcov-dump.texi: Remove -w option.
2011 * gcov-dump.c (dump_working_sets): Remove.
2012 (main): Do not support '-w' option.
2013 (print_usage): Likewise.
2014 (tag_summary): Likewise.
2015 * gcov-io.c (gcov_write_summary): Do not dump
2017 (gcov_read_summary): Likewise.
2018 (gcov_histo_index): Remove.
2019 (gcov_histogram_merge): Likewise.
2020 (compute_working_sets): Likewise.
2021 * gcov-io.h (GCOV_TAG_OBJECT_SUMMARY): Mark
2023 (GCOV_TAG_PROGRAM_SUMMARY): Mark it obsolete.
2024 (GCOV_TAG_SUMMARY_LENGTH): Adjust.
2025 (GCOV_HISTOGRAM_SIZE): Remove.
2026 (GCOV_HISTOGRAM_BITVECTOR_SIZE): Likewise.
2027 (struct gcov_summary): Simplify rapidly just
2028 to runs and sum_max fields.
2029 (gcov_histo_index): Remove.
2030 (NUM_GCOV_WORKING_SETS): Likewise.
2031 (compute_working_sets): Likewise.
2032 * gcov-tool.c (print_overlap_usage_message): Remove
2033 trailing empty line.
2034 * gcov.c (read_count_file): Read GCOV_TAG_OBJECT_SUMMARY.
2035 (output_lines): Remove program related line.
2036 * ipa-profile.c (ipa_profile): Do not consider GCOV histogram.
2037 * lto-cgraph.c (output_profile_summary): Do not stream GCOV
2039 (input_profile_summary): Do not read it.
2040 (merge_profile_summaries): And do not merge it.
2041 (input_symtab): Do not call removed function.
2042 * modulo-sched.c (sms_schedule): Do not print sum_max.
2043 * params.def (HOT_BB_COUNT_FRACTION): Reincarnate param that was
2044 removed when histogram method was invented.
2045 (HOT_BB_COUNT_WS_PERMILLE): Mention that it's used only in LTO
2047 * postreload-gcse.c (eliminate_partially_redundant_load): Fix
2049 * predict.c (get_hot_bb_threshold): Use HOT_BB_COUNT_FRACTION
2050 and dump selected value.
2051 * profile.c (add_working_set): Remove.
2052 (get_working_sets): Likewise.
2053 (find_working_set): Likewise.
2054 (get_exec_counts): Do not work with working sets.
2055 (read_profile_edge_counts): Do not inform as sum_max is removed.
2056 (compute_branch_probabilities): Likewise.
2057 (compute_value_histograms): Remove argument for call of
2058 get_coverage_counts.
2059 * profile.h: Do not make gcov_summary const.
2061 2018-09-21 Monk Chiang <sh.chiang04@gmail.com>
2063 * config.gcc (nds32*-*-*): Set TARGET_DEFAULT_TLSDESC_TRAMPOLINE=0.
2065 2018-09-21 Eric Botcazou <ebotcazou@adacore.com>
2067 PR tree-optimization/86990
2068 * gimple-ssa-store-merging.c (imm_store_chain_info:coalesce_immediate):
2069 Check that the entire merged store group is made of constants only for
2072 2018-09-20 Allan Sandfeld Jensen <allan.jensen@qt.io>
2074 * gcc.c (LINK_COMMAND_SPEC): Handle -r like -nostdlib.
2075 (VTABLE_VERIFICATION_SPEC): Likewise.
2076 (SANITIZER_EARLY_SPEC): Likewise.
2077 (SANITIZER_SPEC): Likewise.
2078 * config/darwin.h (LINK_COMMAND_SPEC): Likewise.
2079 * doc/invoke.texi (Link Options): Document -r.
2081 2018-09-20 Richard Biener <rguenther@suse.de>
2084 * gimplify.c (gimplify_expr): Retain alignment of
2085 addressable lvalue in dereference.
2087 2018-09-20 Alexandre Oliva <aoliva@redhat.com>
2090 * configure.ac: Check for .loc is_stmt support.
2091 * configure, config.in: Rebuilt.
2092 * dwarf2out.c (dwarf2out_source_line): Skip is_stmt
2095 2018-09-20 Segher Boessenkool <segher@kernel.crashing.org>
2097 * config/rs6000/rs6000.opt (misel=no, misel=yes): Delete.
2098 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -misel=yes and
2101 2018-09-20 Segher Boessenkool <segher@kernel.crashing.org>
2103 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
2105 * config/rs6000/rs6000.c (rs6000_debug_vector_unit): Delete
2108 2018-09-20 Marek Polacek <polacek@redhat.com>
2110 * doc/invoke.texi: Add -Wno-init-list-lifetime to C++ Language Options.
2112 2018-09-20 Richard Sandiford <richard.sandiford@arm.com>
2114 PR tree-optimization/87288
2115 * tree-vect-loop.c (vect_analyze_loop_2): Take PEELING_FOR_GAPS
2116 into account when determining PEELING_FOR_NITERS.
2118 2018-09-20 Richard Sandiford <richard.sandiford@arm.com>
2120 PR tree-optimization/86877
2121 * tree-vect-loop.c (vect_analyze_loop_2): Call
2122 vect_verify_datarefs_alignment.
2124 2018-09-19 Marek Polacek <polacek@redhat.com>
2126 * doc/invoke.texi: Document -Wclass-conversion.
2128 2018-09-19 John David Anglin <danglin@gcc.gnu.org>
2130 * config/pa/pa.c (pa_adjust_priority): Delete.
2131 (TARGET_SCHED_ADJUST_PRIORITY): Delete define.
2133 * config/pa/pa.md (atomic_storeqi): Restore deleted expander.
2134 (atomic_storehi): Likewise.
2135 (atomic_storesi): Likewise.
2136 (atomic_loaddi): Restore compare and swap exchange loop code.
2138 2018-09-19 Segher Boessenkool <segher@kernel.crashing.org>
2140 PR rtl-optimization/86902
2141 * combine.c (try_combine): When changing the CC mode used, don't change
2142 an unrelated mode in other_insn to that new CC mode.
2144 2018-09-19 David Malcolm <dmalcolm@redhat.com>
2146 * tree-data-ref.c (runtime_alias_check_p): Use formatted printing
2147 with %T in place of calls to dump_generic_expr.
2148 (prune_runtime_alias_test_list): Likewise.
2149 (create_runtime_alias_checks): Likewise.
2150 * tree-vect-data-refs.c (vect_check_nonzero_value): Likewise.
2151 (vect_analyze_data_ref_dependence): Likewise.
2152 (vect_slp_analyze_data_ref_dependence): Likewise.
2153 (vect_record_base_alignment): Likewise. Use %G in place of call
2154 to dump_gimple_stmt.
2155 (vect_compute_data_ref_alignment): Likewise.
2156 (verify_data_ref_alignment): Likewise.
2157 (vect_find_same_alignment_drs): Likewise.
2158 (vect_analyze_group_access_1): Likewise.
2159 (vect_analyze_data_ref_accesses): Likewise.
2160 (dependence_distance_ge_vf): Likewise.
2161 (dump_lower_bound): Likewise.
2162 (vect_prune_runtime_alias_test_list): Likewise.
2163 (vect_find_stmt_data_reference): Likewise.
2164 (vect_analyze_data_refs): Likewise.
2165 (vect_create_addr_base_for_vector_ref): Likewise.
2166 (vect_create_data_ref_ptr): Likewise.
2167 * tree-vect-loop-manip.c (vect_set_loop_condition): Likewise.
2168 (vect_can_advance_ivs_p): Likewise.
2169 (vect_update_ivs_after_vectorizer): Likewise.
2170 (vect_gen_prolog_loop_niters): Likewise.
2171 (vect_prepare_for_masked_peels): Likewise.
2172 * tree-vect-loop.c (vect_determine_vf_for_stmt): Likewise.
2173 (vect_determine_vectorization_factor): Likewise.
2174 (vect_is_simple_iv_evolution): Likewise.
2175 (vect_analyze_scalar_cycles_1): Likewise.
2176 (vect_analyze_loop_operations): Likewise.
2177 (report_vect_op): Likewise.
2178 (vect_is_slp_reduction): Likewise.
2179 (check_reduction_path): Likewise.
2180 (vect_is_simple_reduction): Likewise.
2181 (vect_create_epilog_for_reduction): Likewise.
2182 (vect_finalize_reduction:): Likewise.
2183 (vectorizable_induction): Likewise.
2184 (vect_transform_loop_stmt): Likewise.
2185 (vect_transform_loop): Likewise.
2186 (optimize_mask_stores): Likewise.
2187 * tree-vect-patterns.c (vect_pattern_detected): Likewise.
2188 (vect_split_statement): Likewise.
2189 (vect_recog_over_widening_pattern): Likewise.
2190 (vect_recog_average_pattern): Likewise.
2191 (vect_determine_min_output_precision_1): Likewise.
2192 (vect_determine_precisions_from_range): Likewise.
2193 (vect_determine_precisions_from_users): Likewise.
2194 (vect_mark_pattern_stmts): Likewise.
2195 (vect_pattern_recog_1): Likewise.
2196 * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
2197 (vect_record_max_nunits): Likewise.
2198 (vect_build_slp_tree_1): Likewise.
2199 (vect_build_slp_tree_2): Likewise.
2200 (vect_print_slp_tree): Likewise.
2201 (vect_analyze_slp_instance): Likewise.
2202 (vect_detect_hybrid_slp_stmts): Likewise.
2203 (vect_detect_hybrid_slp_1): Likewise.
2204 (vect_slp_analyze_operations): Likewise.
2205 (vect_slp_analyze_bb_1): Likewise.
2206 (vect_transform_slp_perm_load): Likewise.
2207 (vect_schedule_slp_instance): Likewise.
2208 * tree-vect-stmts.c (vect_mark_relevant): Likewise.
2209 (vect_mark_stmts_to_be_vectorized): Likewise.
2210 (vect_init_vector_1): Likewise.
2211 (vect_get_vec_def_for_operand): Likewise.
2212 (vect_finish_stmt_generation_1): Likewise.
2213 (vect_check_load_store_mask): Likewise.
2214 (vectorizable_call): Likewise.
2215 (vectorizable_conversion): Likewise.
2216 (vectorizable_operation): Likewise.
2217 (vectorizable_load): Likewise.
2218 (vect_analyze_stmt): Likewise.
2219 (vect_is_simple_use): Likewise.
2220 (vect_get_vector_types_for_stmt): Likewise.
2221 (vect_get_mask_type_for_stmt): Likewise.
2222 * tree-vectorizer.c (increase_alignment): Likewise.
2224 2018-09-19 Andrew Stubbs <ams@codesourcery.com>
2226 * doc/rtl.texi: Adjust vec_select description.
2227 * simplify-rtx.c (simplify_binary_operation_1): Allow VEC_SELECT to use
2228 non-constant selectors.
2230 2018-09-19 Matthew Malcomson <matthew.malcomson@arm.com>
2232 * config/aarch64/aarch64-protos.h
2233 (aarch64_offset_9bit_signed_unscaled_p): New declaration.
2234 * config/aarch64/aarch64.md (arches): New "rcpc8_4" attribute value.
2235 (arch_enabled): Add check for "rcpc8_4" attribute value of "arch".
2236 * config/aarch64/aarch64.h (AARCH64_FL_RCPC8_4): New bitfield.
2237 (AARCH64_FL_FOR_ARCH8_4): Include AARCH64_FL_RCPC8_4.
2238 (AARCH64_FL_PROFILE): Move index so flags are ordered.
2239 (AARCH64_ISA_RCPC8_4): New flag.
2240 * config/aarch64/aarch64.c (offset_9bit_signed_unscaled_p): Renamed
2241 to aarch64_offset_9bit_signed_unscaled_p.
2242 * config/aarch64/atomics.md (atomic_store<mode>): Allow offset
2244 * config/aarch64/constraints.md (Ust): New constraint.
2245 * config/aarch64/predicates.md.
2246 (aarch64_9bit_offset_memory_operand): New predicate.
2247 (aarch64_rcpc_memory_operand): New predicate.
2249 2018-09-19 Eric Botcazou <ebotcazou@adacore.com>
2251 PR rtl-optimization/87361
2252 * rtlanal.c (nonzero_bits1): Revert accidental change.
2254 2018-09-19 Richard Biener <rguenther@suse.de>
2256 PR tree-optimization/87349
2257 PR tree-optimization/87342
2258 * tree-ssa-sccvn.c (do_rpo_vn): Iterate max_rpo computation.
2260 2018-09-18 Marek Polacek <polacek@redhat.com>
2262 P1064R0 - Allowing Virtual Function Calls in Constant Expressions
2263 * gimple-fold.c (gimple_get_virt_method_for_vtable): Adjust assert.
2265 2018-09-18 Segher Boessenkool <segher@kernel.crashing.org>
2267 * config/rs6000/rs6000.md: Remove old "Cygnus sibcall" comment.
2269 2018-09-18 Segher Boessenkool <segher@kernel.crashing.org>
2271 PR rtl-optimization/86882
2272 * rtlanal.c (reg_overlap_mentioned_p): Handle CLOBBER.
2274 2018-09-18 Uros Bizjak <ubizjak@gmail.com>
2276 * config/i386/i386.md (*<code>extend<mode>xf2): Macroize insn from
2277 *<code>extendsfxf2 and *<code>extenddfxf2 using MODEF mode iterator.
2279 2018-09-18 Jonathan Wakely <jwakely@redhat.com>
2282 * doc/invoke.texi (Link Options): Fix formatting and grammar.
2284 2018-09-18 Richard Biener <rguenther@suse.de>
2287 * tree-ssa-coalesce.c (tree_int_map_hasher): Remove.
2288 (compute_samebase_partition_bases): Likewise.
2289 (coalesce_ssa_name): Always use compute_optimized_partition_bases.
2290 (gimple_can_coalesce_p): Simplify.
2292 2018-09-18 Hans-Peter Nilsson <hp@bitrange.com>
2294 Handle a library implementation of ffs calling __builtin_ffs.
2295 * config/mmix/mmix.c (TARGET_INIT_LIBFUNCS): Override with...
2296 (mmix_init_libfuncs): New function: make __builtin_ffs expand
2299 2018-09-17 David Malcolm <dmalcolm@redhat.com>
2301 * diagnostic-show-locus.c (class layout_range): Add field
2303 (layout_range::layout_range): Add "original_idx" param and use it
2304 to initialize new field.
2305 (make_range): Use 0 for original_idx.
2306 (layout::layout): Pass in index to calls to
2307 maybe_add_location_range.
2308 (layout::maybe_add_location_range): Add param "original_idx" and
2309 pass it on to layout_range.
2310 (layout::print_any_labels): Pass on range->m_original_idx to
2312 (gcc_rich_location::add_location_if_nearby): Use 0 for
2314 * gcc-rich-location.h (text_range_label::get_text): Update for new
2316 (range_label_for_type_mismatch::get_text): Likewise.
2318 2018-09-17 Uros Bizjak <ubizjak@gmail.com>
2320 * config/i386/i386.c (ix86_emit_i387_log1p): Emit fldln2 earlier.
2322 2018-09-17 David Malcolm <dmalcolm@redhat.com>
2324 * gimple-ssa-sprintf.c (fmtwarn): Update for introduction of
2325 format_string_diagnostic_t.
2326 (fmtwarn_n): Likewise.
2327 * substring-locations.c
2328 (format_string_diagnostic_t::format_string_diagnostic_t) New ctor.
2329 (format_warning_n_va): Convert to...
2330 (format_string_diagnostic_t::emit_warning_n_va): ...this.
2331 (format_warning_va): Convert to...
2332 (format_string_diagnostic_t::emit_warning_va): ...this.
2333 (format_warning_at_substring): Convert to...
2334 (format_string_diagnostic_t::emit_warning): ...this.
2335 (format_warning_at_substring_n): Convert to...
2336 (format_string_diagnostic_t::emit_warning_n): ...this.
2337 * substring-locations.h (class format_string_diagnostic_t): New
2339 (format_warning_va): Convert to
2340 format_string_diagnostic_t::emit_warning_va.
2341 (format_warning_n_va): Convert to
2342 format_string_diagnostic_t::emit_warning_n_va.
2343 (format_warning_at_substring): Convert to
2344 format_string_diagnostic_t::emit_warning.
2345 (format_warning_at_substring_n): Convert to
2346 format_string_diagnostic_t::emit_warning_n.
2348 2018-09-17 Cesar Philippidis <cesar@codesourcery.com>
2349 Bernd Schmidt <bernds_cb1@t-online.de>
2351 * config/nvptx/nvptx.md (atomic_fetch_<logic><mode>): Enable with
2354 2018-09-17 Uros Bizjak <ubizjak@gmail.com>
2356 * config/i386/i386.md (truncxf<mode>2_i387_noop_unspec): Change
2357 operand 0 predicate to nonimmediate operand.
2358 (rint<mode>2_frndint): Remove insn pattern.
2359 (rint<mode>2): Change operand 1 predicate to general_operand.
2360 Extend operand 1 to XFmode and generate rintxf2 insn.
2361 (frndintxf2_<rounding>): Rename from frndint<mode>2_<rounding>.
2362 Do not use X87MODEF mode macro.
2363 (frndintxf2_<rounding>_i387): Rename from
2364 frndint<mode>2_<rounding>_i387. Do not use X87MODEF mode macro.
2365 (<rounding_insn><mode>2): For non-SSE modes, extend operand 1
2366 to XFmode and generate significandxf3 insn.
2368 2018-09-17 Richard Biener <rguenther@suse.de>
2370 PR tree-optimization/87328
2371 * tree-ssa-sccvn.c (process_bb): Remove assertion about not
2372 visiting unexecutable backedges when not iterating.
2373 (do_rpo_vn): Mark all edges not executable even when not
2376 2018-09-17 Martin Jambor <mjambor@suse.cz>
2379 * doc/invoke.texi (Warning Options): Likewise.
2381 2018-09-17 Richard Biener <rguenther@suse.de>
2383 PR tree-optimization/87301
2384 * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_cleanup): Properly
2385 clean EH info from leftover copy assignments.
2387 2018-09-17 Martin Liska <mliska@suse.cz>
2389 PR gcov-profile/85871
2390 * gcov.c (output_intermediate_file): Fix out of bounds
2393 2018-09-17 Vineet Gupta <vgupta@synopsys.com>
2395 * config/arc/arc.c: Object attributes for core4 not reflected
2397 * config/arc/arc.h: Don't restrict DBNZ to core3 (core4 includes
2400 2018-09-17 Alexey Brodkin <abrodkin@synopsys.com>
2402 * config/arc/linux.h (LINK_EH_SPEC): Add missing space.
2404 2018-09-17 Martin Liska <mliska@suse.cz>
2406 * doc/gcov.texi: Document new option --use-hotness-colors.
2407 * gcov.c (struct source_info): Declare new field.
2408 (source_info::source_info): Set default for maximum_count.
2409 (print_usage): Add new -q option.
2410 (process_args): Process it.
2411 (accumulate_line_info): Save src->maximum_count.
2412 (output_line_beginning): Make color line number if
2413 flag_use_hotness_colors is set.
2414 (output_line_details): Pass default argument value.
2415 (output_lines): Pass src->maximum_count.
2417 2018-09-17 Martin Liska <mliska@suse.cz>
2419 * common/config/i386/i386-common.c (ix86_get_valid_option_values):
2420 Use processor_names table.
2421 * config/i386/i386.c (ix86_default_align): Use
2422 processor_cost_table for alignment values.
2423 (ix86_option_override_internal): Use processor_names.
2424 (ix86_function_specific_print): Likewise.
2425 * config/i386/i386.h (struct processor_costs):
2426 Add alignment values.
2427 (struct ptt): Remove and replace with const char *.
2428 * config/i386/x86-tune-costs.h (struct processor_costs):
2429 Declare default alignments for all costs.
2431 2018-09-17 Aldy Hernandez <aldyh@redhat.com>
2433 * tree-vrp.c (extract_range_from_unary_expr): Do not special case
2434 symbolics or VR_VARYING ranges for ABS_EXPR.
2435 * wide-int-range.cc (wide_int_range_abs): Return positive numbers
2436 when range will wrap.
2438 2018-09-15 Eric Botcazou <ebotcazou@adacore.com>
2441 * cfgexpand.c (expand_gimple_basic_block): Be prepared for a BARRIER
2442 before and after a JUMP_TABLE_DATA.
2444 2018-09-14 John David Anglin <danglin@gcc.gnu.org>
2447 * dojump.c (do_compare_and_jump): Canonicalize function pointers
2448 when one operand is a function pointer. Use POINTER_TYPE_P and
2449 FUNC_OR_METHOD_TYPE_P.
2450 * expr.c (do_store_flag): Use POINTER_TYPE_P and FUNC_OR_METHOD_TYPE_P.
2451 * fold-const.c (build_range_check): Likewise.
2452 * match.pd (simple_comparison): Likewise.
2454 2018-09-14 David Malcolm <dmalcolm@redhat.com>
2457 * spellcheck.c (get_edit_distance_cutoff): New function.
2458 (selftest::test_edit_distance_unit_test_oneway): Rename to...
2459 (selftest::test_get_edit_distance_one_way): ...this.
2460 (selftest::test_get_edit_distance_unit): Rename to...
2461 (selftest::test_get_edit_distance_both_ways): ...this.
2462 (selftest::test_edit_distances): Move tests to this new function,
2463 and test some more pairs of strings. Update for above renaming.
2464 (selftest::get_old_cutoff): New function.
2465 (selftest::test_get_edit_distance_cutoff): New function.
2466 (selftest::assert_suggested_for): New function.
2467 (ASSERT_SUGGESTED_FOR): New macro.
2468 (selftest::assert_not_suggested_for): New function.
2469 (ASSERT_NOT_SUGGESTED_FOR): New macro.
2470 (selftest::test_suggestions): New function.
2471 (selftest::spellcheck_c_tests): Move test_get_edit_distance_unit
2472 tests to selftest::test_edit_distances and call it. Add calls to
2473 selftest::test_get_edit_distance_cutoff and
2474 selftest::test_suggestions.
2475 * spellcheck.h (get_edit_distance_cutoff): New function declaration.
2476 (best_match::consider): Replace hard-coded cutoff calculation with
2478 (best_match::get_cutoff): New declaration.
2479 (best_match::get_best_meaningful_candidate): Likewise.
2481 2018-09-14 Bernd Edlinger <bernd.edlinger@hotmail.de>
2483 * builtins.c (fold_builtin_strlen): Remove TODO comment.
2485 2018-09-14 Bernd Edlinger <bernd.edlinger@hotmail.de>
2488 2018-07-30 Bernd Edlinger <bernd.edlinger@hotmail.de>
2490 * tree-ssa-forwprop.c (simplify_builtin_call): Don't create a not NUL
2491 terminated string literal.
2493 2018-09-14 Martin Sebor <msebor@redhat.com>
2495 * builtins.c (unterminated_array): Handle ARRAY_REF.
2496 (expand_builtin_stpcpy_1): Detect unterminated char arrays.
2497 * builtins.h (unterminated_array): Declare extern.
2498 * gimple-fold.c (gimple_fold_builtin_stpcpy): Detect unterminated
2500 (gimple_fold_builtin_sprintf): Propagate NO_WARNING to transformed
2503 2018-09-14 Martin Sebor <msebor@redhat.com>
2504 Jeff Law <law@redhat.com>
2506 * builtins.c (unterminated_array): New.
2507 (expand_builtin_strcpy): Adjust.
2508 (expand_builtin_strcpy_args): Detect unterminated arrays.
2509 * gimple-fold.c (get_maxval_strlen): Add argument. Detect
2510 unterminated arrays.
2511 * gimple-fold.h (get_maxval_strlen): Add argument.
2512 (gimple_fold_builtin_strcpy): Detec unterminated arrays.
2514 * gimple-fold.c (get_range_strlen): Add argument.
2515 (get_maxval_strlen): Adjust.
2516 * gimple-fold.h (get_range_strlen): Add argument.
2518 2018-09-14 Wei Xiao <wei3.xiao@intel.com>
2520 * config/i386/movdirintrin.h: Fix copyright year.
2522 2018-09-14 Uros Bizjak <ubizjak@gmail.com>
2524 * reg-stack.c: Include regs.h.
2525 (replace_reg): Assert that mode is MODE_FLOAT or MODE_COMPLEX_FLOAT.
2526 (emit_pop_insn): Default pop insn mode to the reg_raw_mode of
2527 FIRST_STACK_REG, not DFmode.
2528 (emit_swap_insn): Default swap insn mode to the reg_raw_mode of
2529 FIRST_STACK_REG, not XFmode. Explicitly construct swap RTX.
2530 (change stack): Default register mode to the reg_raw_mode of
2531 FIRST_STACK_REG, not DFmode.
2532 * config/i386/i386.md (*swap<mode>): Remove insn pattern.
2533 (*swapxf): Rename from swapxf.
2535 2018-09-14 Carl Love <cel@us.ibm.com>
2537 * config/rs6000/emmintrin.h: Add _MM_SHUFFLE2.
2538 * config/rs6000/xmmintrin.h: Add _MM_SHUFFLE.
2540 2018-09-14 Segher Boessenkool <segher@kernel.crashing.org>
2543 * config/rs6000/rs6000.md (*mov<mode>_hardfloat64): Add Z to the Y
2546 2018-09-14 Sam Tebbs <sam.tebbs@arm.com>
2549 * config/aarch64/aarch64.md (*aarch64_bfxilsi_uxtw): Define.
2551 2018-09-14 Jason Merrill <jason@redhat.com>
2553 Fix --enable-gather-detailed-mem-stats.
2554 * hash-table.c (hash_table_usage): Change from variable to function.
2555 * hash-table.h: Adjust.
2556 * Makefile.in: Add missing dependencies on hash-table.h.
2558 2018-09-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2560 PR tree-optimization/87259
2562 (pass_cse_reciprocals::execute): Run optimize_recip_sqrt after
2563 execute_cse_reciprocals_1 has tried transforming.
2565 2018-09-14 Aldy Hernandez <aldyh@redhat.com>
2567 * tree-vrp.c (extract_range_from_binary_expr_1): Normalize
2568 VR_VARYING for PLUS/MINUS_EXPR.
2570 2018-09-14 Ilya Leoshkevich <iii@linux.ibm.com>
2572 * config/s390/s390-passes.def (INSERT_PASS_BEFORE): Improve
2575 2018-09-14 Richard Biener <rguenther@suse.de>
2578 * tree-ssa-coalesce.c (ssa_conflicts_merge): Remove conflict
2579 bits for the merged partition.
2581 2018-09-13 Martin Sebor <msebor@redhat.com>
2582 Bernd Edlinger <bernd.edlinger@hotmail.de>
2584 * builtins.h (c_srlen): Add argument.
2585 * builtins.c (warn_string_no_nul): New function.
2586 (c_strlen): Add argument and use it. Update recursive calls.
2587 Pass DECL argument to string_constant to get info on non
2588 terminated strings. Update *NONSTR as needed.
2589 (fold_builtin_strlen): Add argument to calls to c_strlen.
2590 Warn for unterminated arrays.
2591 (warn_string_no_null): Add prototype.
2592 * expr.c (string_constant): Update arguments. Update recursive
2593 calls appropriately. Detect missing NUL terminator and outermost
2594 declaration its missing in.
2595 Improve checks for arrays with nonzero lower bound or elements
2596 that are not a single byte. Simplify offset computation.
2597 Simplify checks for non-NUL terminated strings.
2598 * gimple-fold.c (get_range_strlen): Add argument to c_strlen call.
2599 * gimple-ssa-sprintf.c (get_string_length): Remove unnecessary code.
2601 2018-09-13 Bernd Edlinger <bernd.edlinger@hotmail.de>
2603 * builtins.c (c_strlen): Handle not zero terminated STRING_CSTs
2605 * fold-const.c (c_getstr): Fix function comment. Remove unused third
2606 argument. Fix range checks.
2607 * fold-const.h (c_getstr): Adjust protoype.
2608 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid folding when
2609 string is constant but contains no NUL byte.
2611 * expr.c (string_constant): Adjust function comment.
2612 Remove bogus check for zero termination.
2614 * fold-const.c (c_getstr): Clamp STRING_LENGTH to STRING_SIZE.
2616 * varasm.c (compare_constant): Compare type size of STRING_CSTs.
2617 (get_constant_size): Don't make STRING_CSTs larger than they are.
2618 (check_string_literal): New check function for STRING_CSTs.
2619 (output_constant): Use it.
2621 2018-09-13 Eric Botcazou <ebotcazou@adacore.com>
2624 * config/visium/visium.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define.
2626 2018-09-13 Richard Biener <rguenther@suse.de>
2628 PR tree-optimization/87263
2629 * tree-ssa-sccvn.c (visit_phi): Revert some earlier changes.
2630 (struct unwind_state): Add max_rpo field.
2631 (do_rpo_vn): Allow up-to-date loop state to be used when not iterating.
2632 Compute max_rpo, the max RPO number a block can be backwards reached
2633 from. Re-write non-iterating mode to a RPO ordered worklist approach,
2634 separating it from the iterating mode.
2636 2018-09-13 Vlad Lazar <vlad.lazar@arm.com>
2638 * haifa-sched.c (rank_for_schedule): Schedule by INSN_COST.
2639 (rfs_decision): New scheduling decision.
2641 2018-09-13 Richard Biener <rguenther@suse.de>
2644 * tree-ssa-sccvn.c (vn_nary_op_insert_into): Fix assert.
2645 (vn_nary_op_insert_pieces_predicated): Do not write useless
2646 valid_dominated_by_p entry outside of the allocated storage.
2648 2018-09-13 Omar Sandoval <osandov@osandov.com>
2649 Tom de Vries <tdevries@suse.de>
2652 * dwarf2out.c (is_c): New function.
2653 (add_subscript_info): Add DW_AT_count of 0 for C zero-length arrays.
2655 2018-09-13 Sam Tebbs <sam.tebbs@arm.com>
2658 * config/aarch64/aarch64.md (*aarch64_bfxil):
2660 * config/aarch64/constraints.md (Ulc): Define.
2661 * config/aarch64/aarch64-protos.h (aarch64_high_bits_all_ones_p):
2663 * config/aarch64/aarch64.c (aarch64_high_bits_all_ones_p):
2666 2018-09-13 Vlad Lazar <vlad.lazar@arm.com>
2668 * config/aarch64/aarch64.h (TARGET_COMPUTE_FRAME_LAYOUT): Define.
2669 * config/aarch64/aarch64.c (aarch64_expand_prologue): Remove
2670 aarch64_layout_frame call.
2671 (aarch64_expand_epilogue): Likewise.
2672 (aarch64_initial_elimination_offset): Likewise.
2673 (aarch64_get_separate_components): Likewise.
2674 (aarch64_use_return_insn_p): Likewise.
2675 (aarch64_layout_frame): Remove unneeded check.
2677 2018-09-13 Jakub Jelinek <jakub@redhat.com>
2679 * configure.ac: Only append
2680 " : (reconfigured) $TOPLEVEL_CONFIGURE_ARGUMENTS" to
2681 gcc_config_arguments if it was never reconfigured or last reconfigure
2682 was with different arguments.
2683 * configure: Regenerated.
2685 2018-09-13 Jakub Jelinek <jakub@redhat.com>
2686 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2689 * expr.c (maybe_optimize_pow2p_mod_cmp): New function.
2690 (maybe_optimize_mod_cmp): Use it if integer_pow2p treeop1.
2692 2018-09-13 Jakub Jelinek <jakub@redhat.com>
2694 PR tree-optimization/87287
2695 * fold-const.c (fold_binary_loc) <case EQ_EXPR>: Move signed modulo
2696 X % C == 0 to X % (unsigned) C == 0 optimization to ...
2697 * match.pd (X % C == 0): ... here. New optimization.
2699 2018-09-12 Jakub Jelinek <jakub@redhat.com>
2702 * expr.h (maybe_optimize_mod_cmp): Declare.
2703 * expr.c (mod_inv): New function.
2704 (maybe_optimize_mod_cmp): New function.
2705 (do_store_flag): Use it.
2706 * cfgexpand.c (expand_gimple_cond): Likewise.
2708 2018-09-09 Cesar Philippidis <cesar@codesourcery.com>
2709 Julian Brown <julian@codesourcery.com>
2712 * gimplify.c (gimplify_scan_omp_clauses): Set
2713 target_firstprivatize_array_bases in OpenACC parallel and kernels
2714 region contexts. Remove GOMP_MAP_FIRSTPRIVATE_REFERENCE clauses from
2715 OpenACC data regions.
2717 2018-09-12 Uros Bizjak <ubizjak@gmail.com>
2719 * config/i386/i386.md (sqrt_extend<mode>xf3_i387): Remove.
2720 (sqrt<mode>2): Extend operand 1 to XFmode and generate sqrtxf3 insn.
2722 2018-09-12 Richard Biener <rguenther@suse.de>
2724 PR tree-optimization/87280
2725 * tree-ssa-sccvn.c (process_bb): Handle the case of executable
2726 edge but unreachable target.
2727 (do_rpo_vn): For conservatively handling a PHI only mark
2728 the backedge executable but not the block reachable.
2730 2018-09-12 Richard Biener <rguenther@suse.de>
2732 PR tree-optimization/87266
2733 * tree-ssa-sccvn.c (do_rpo_vn): Always iterate to not yet
2736 2018-09-12 Andreas Krebbel <krebbel@linux.ibm.com>
2738 * config/s390/s390.md (PFPO_RND_MODE_DFP, PFPO_RND_MODE_BFP): New
2740 ("trunc<BFP:mode><DFP_ALL:mode>2")
2741 ("trunc<DFP_ALL:mode><BFP:mode>2")
2742 ("extend<BFP:mode><DFP_ALL:mode>2")
2743 ("extend<DFP_ALL:mode><BFP:mode>2"): Set proper rounding mode
2744 according to the target operand type.
2746 2018-09-12 Jakub Jelinek <jakub@redhat.com>
2747 Andreas Krebbel <krebbel@linux.ibm.com>
2749 PR tree-optimization/86844
2750 * gimple-ssa-store-merging.c
2751 (imm_store_chain_info::coalesce_immediate): For overlapping stores, if
2752 there are any overlapping stores in between them, make sure they are
2753 also coalesced or we give up completely.
2755 2018-09-12 Jakub Jelinek <jakub@redhat.com>
2758 * fold-const.c (fold_ternary_loc) <case COND_EXPR>: Verify also that
2759 BIT_AND_EXPR's second operand is a power of two. Formatting fix.
2761 2018-09-12 Tom de Vries <tdevries@suse.de>
2763 * common.opt (gdescribe-dies): Add option.
2764 * dwarf2out.c (add_name_and_src_coords_attributes): Add description
2765 attribute for artifical and nameless decls.
2766 (dwarf2out_register_external_die): Add description attribute to
2767 external reference die.
2768 (add_desc_attribute): New functions.
2769 (gen_subprogram_die): Add description attribute to
2770 DW_TAG_call_site_parameter.
2771 * tree-pretty-print.c (print_generic_expr_to_str): New function.
2772 * tree-pretty-print.h (print_generic_expr_to_str): Declare.
2773 * doc/invoke.texi (@item Debugging Options): Add -gdescribe-dies and
2775 (@item -gdescribe-dies): Add.
2777 2018-09-12 Aldy Hernandez <aldyh@redhat.com>
2779 * tree-vrp.c (vrp_shift_undefined_p): Remove.
2780 (extract_range_from_binary_expr_1: Call
2781 wide_int_range_shift_undefined_p instead of vrp_shift_undefined_p.
2782 * wide-int-range.h (wide_int_range_shift_undefined_p): Do not
2785 2018-09-12 Aldy Hernandez <aldyh@redhat.com>
2787 * gimple-ssa-warn-alloca.c
2788 (alloca_type_and_limit::alloca_type_and_limit): Initialize limit
2789 field for ALLOCA_BOUND_*_LARGE.
2791 2018-09-11 Nathan Sidwell <nathan@acm.org>
2793 * gcc.c (load_specs, execute, run_attempt): Use %qs not '%s'.
2795 2018-09-11 Uros Bizjak <ubizjak@gmail.com>
2797 * reg-stack.c (subst_asm_stack_regs): Call replace_reg also
2798 for clobbers. Remove obsolete comment.
2800 2018-09-11 Uros Bizjak <ubizjak@gmail.com>
2802 * config/i386/i386.md (define_attr "type"): Remove mpxmov, mpxmk,
2803 mpxchk, mpxld and mpxst types.
2804 (define_attr length_immediate): Remove all processing of mpx types.
2805 (define_attr prefix_0f): Ditto.
2806 (define_attr memory): Ditto.
2808 2018-09-11 Uros Bizjak <ubizjak@gmail.com>
2810 * config/i386/i386.md (fyl2x_extend<mode>xf3_i387): Remove.
2811 (log<mode>2): Change operand 1 predicate to general_operand.
2812 Extend operand 1 to XFmode and generate logxf3 insn.
2813 (log10<mode>2): Change operand 1 predicate to general_operand.
2814 Extend operand 1 to XFmode and generate log10xf3 insn.
2815 (log2<mode>2): Change operand 1 predicate to general_operand.
2816 Extend operand 1 to XFmode and generate log2xf3 insn.
2817 (fyl2xp1_extend<mode>xf3_i387): Remove.
2818 (log1p<mode>2): Change operand 1 predicate to general_operand.
2819 Extend operand 1 to XFmode and generate log1pxf3 insn.
2820 (fxtract_extend<mode>xf3_i387): Remove.
2821 (logb<mode>2): Change operand 1 predicate to general_operand.
2822 Extend operand 1 to XFmode and generate logbxf3 insn.
2823 (ilogb<mode>2): Change operand 1 predicate to general_operand.
2824 Extend operand 1 to XFmode and generate fxtractxf3_i387 insn.
2825 (significand<mode>2): Change operand 1 predicate to general_operand.
2826 Extend operand 1 to XFmode and generate significandxf3 insn.
2828 2018-09-11 Nathan Sidwell <nathan@acm.org>
2830 * gcc.c (perror_with_name, pfatal_with_name): Delete.
2831 (load_specs): Use fatal_error.
2832 (DELETE_IF_ORDINARY, process_command): Use error.
2833 (execute, run_attempt): Use fatal_error.
2835 2018-09-11 Andrew Stubbs <ams@codesourcery.com>
2837 * diagnostic-core.h (sorry_at): New prototype.
2838 * diagnostic.c (sorry_at): New function.
2840 2018-09-11 Aldy Hernandez <aldyh@redhat.com>
2842 * tree-vrp.c (extract_range_from_binary_expr_1): Treat all divisions
2843 by zero as VR_UNDEFINED.
2845 2018-09-10 Uros Bizjak <ubizjak@gmail.com>
2847 * config/i386/i386.md (<sincos>xf2): Rename from *<sincos>xf2_i387.
2848 (*<sincos>_extend<mode>xf2_i387): Remove insn pattern.
2849 (<sincos>mode2): New expander.
2850 (sincos_extend<mode>xf3_i387): Remove insn pattern.
2851 (sincos -> sin, cos splitters): Remove splitter patterns.
2852 (sincos<mode>3): Change operand 2 predicate to general_operand.
2853 Extend operand 2 to XFmode and generate sincosxf3 insn.
2854 (fptanxf4_i387): Change mode of operands 0 and 3 to SFmode.
2855 Change operand 3 predicate to const1_operand.
2856 (fptan_extend<mode>xf4_i387): Remove insn pattern.
2857 (tanxf2): Update operands in the call to fptanxf4_i387.
2858 (tan<mode>2): Change operand 1 predicate to general_operand.
2859 Extend operand 1 to XFmode and generate tanxf3 insn.
2860 (atan2xf3): Rename from *fpatanxf3_i387.
2861 (fpatan_extend<mode>xf3_i387): Remove insn pattern.
2862 (atan2xf3): Remove expander.
2863 (atan2<mode<3): Change operand 1 and 2 predicates to general_operand.
2864 Extend operands 1 and 2 to XFmode and generate atan2xf3 insn.
2865 (atan<mode>2): Change operand 1 predicate to general_operand.
2866 Extend operand 1 to XFmode and generate atanxf3 insn.
2868 2018-09-10 Uros Bizjak <ubizjak@gmail.com>
2870 * config/i386/i386.md (x87/SSE constant load splitter): Use
2871 memory_operand instead of nonimmediate_operand for input operand
2874 2018-09-09 Uros Bizjak <ubizjak@gmail.com>
2876 * config/i386/i386.md (float partial SSE register stall splitter): Move
2877 splitter near its instruction pattern.
2878 (float_extend partial SSE register stall splitter): Ditto.
2879 (float_truncate partial SSE register stall splitter): Ditto.
2881 2018-09-09 Hans-Peter Nilsson <hp@bitrange.com>
2884 * config/mmix/mmix.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
2885 to speculation_safe_value_not_needed.
2888 * config/mmix/mmix.c (mmix_assemble_integer): Handle byte-size
2889 non-CONST_INT rtx:es using assemble_integer_with_op ".byte".
2890 (MMIX_CFUN_NEEDS_SAVED_EH_RETURN_ADDRESS): Don't call
2891 leaf_function_p, instead use has_hard_reg_initial_val.
2893 2018-09-09 Nathan Sidwell <nathan@acm.org>
2895 * gcc.h (pfatal_with_name): Don't declare here.
2896 * gcc.c (pfatal_with_name): Make static.
2898 2018-09-09 Xianmiao Qu <xianmiao_qu@c-sky.com>
2900 * config/csky/csky.md (*cskyv2_adddi3, *cskyv2_subdi3): Add
2903 2018-09-08 John David Anglin <danglin@gcc.gnu.org>
2905 PR rtl-optimization/85458
2906 * sel-sched.c (sel_target_adjust_priority): Allow backend adjust
2907 priority hook to reduce the priority of EXPR.
2909 2018-09-07 Uros Bizjak <ubizjak@gmail.com>
2911 * config/i386/i386.md (float<SWI48x:mode><MODEF:mode>2) Enable
2912 DImode for x87 on 32bit targets. Conditionally disable x87 modes
2913 with X87_ENABLE_FLOAT. Remove preparation code.
2914 (*float<SWI48:mode><MODEF:mode>2): Rename from
2915 *float<SWI48:mode><MODEF:mode>2_mixed. Handle x87, SSE and mixed
2916 math using "enabled" attribute.
2917 (*floatdi<MODEF:mode>2_i387): Rename from
2918 *float<SWI48x:mode><MODEF:mode>2_i387. Handle only DImode and
2919 enable for 32bit targets only.
2920 (floatdi<X87MODEF:mode>2_i387_with_xmm pre-reload splitter): New
2922 (floatdi<X87MODEF:mode>2_i387_with_xmm): Use register_operand
2923 as operand 1 predicate. Rewrite as define_insn_and_split.
2924 (floatdi<X87MODEF:mode>2_i387_with_xmm memory input splitter): Remove.
2926 2018-09-06 Uros Bizjak <ubizjak@gmail.com>
2928 * reg-stack.c (get_true_reg) <case FLOAT_TRUNCATE>: Reorder
2929 to fallthru to FLOAT case.
2931 2018-09-06 Will Schmidt <will_schmidt@vnet.ibm.com>
2934 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Update logic
2935 around folding of vec_sl to handle out of range shift values.
2937 2018-09-06 Uros Bizjak <ubizjak@gmail.com>
2939 * config/i386/i386.md (fix_trunc<mode>_fisttp_i387_1): Remove.
2940 Update callers to gen_fix_trunc<mode>_i387_fisttp
2941 (fix_trunc<mode>_i387_fisttp): Change operand 0 predicate to
2942 nonimmediate_operand.
2943 (fix_trunc<mode>_i387_fisttp_with_temp): Remove insn pattern
2944 and corresponding splitters.
2945 (*fix_trunc<mode>_i387_1): Always expand with fix_trunc<mode>_i387.
2946 (fix_truncdi_i387): Change operand 0 predicate to nonimmediate_operand.
2947 (fix_truncdi_i387_with_temp): Remove insn pattern
2948 and corresponding splitters.
2949 (fix_trunc<mode>_i387): Change operand 0 predicate to
2950 nonimmediate_operand.
2951 (fix_trunc<mode>_i387_with_temp): Remove insn pattern
2952 and corresponding splitters.
2953 (*fistdi2_1): Remove.
2955 (fistdi2_with_temp): Remove insn pattern and corresponding splitters.
2956 (lrintxfdi2): Remove expander. Reimplement as define_insn.
2957 (*fist<mode>2_1): Remove.
2958 (fist<mode>2): Ditto.
2959 (fist<mode>2_with_temp): Remove insn pattern and corresponding
2961 (lrintxf<mode>2): Remove expander. Reimplement as define_insn.
2962 (*fist<mode>2_<rounding>_1): Always expand with fist<mode>2_<rounding>.
2963 (fistdi2_<rounding>): Change operand 0 predicate to
2964 nonimmediate_operand.
2965 (fistdi2_<rounding>_with_temp): Remove insn pattern
2966 and corresponding splitters.
2967 (fist<mode>2_<rounding>): Change operand 0 predicate to
2968 nonimmediate_operand.
2969 (fist<mode>2_<rounding>_with_temp): Remove insn pattern
2970 and corresponding splitters.
2972 (*fixuns_trunc<mode>si2_avx512f_zext): Depend on TARGET_SSE_MATH.
2974 2018-09-06 Bernd Edlinger <bernd.edlinger@hotmail.de>
2976 * varasm.c (output_constructor_regular_field): Check TYPE_SIZE_UNIT of
2979 2018-09-06 Will Schmidt <will_schmidt@vnet.ibm.com>
2981 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support for
2982 early gimple folding of vec_splat().
2983 * tree-vect-generic.c: Remove static from tree_vec_extract() definition.
2984 * gimple-fold.h: Add an extern define for tree_vec_extract().
2986 2018-09-06 Will Schmidt <will_schmidt@vnet.ibm.com>
2988 * config/rs6000/rs6000.c (fold_mergehl_helper): Add types_compatible_p
2989 wrappers around TREE_TYPE comparisons.
2991 2018-09-06 Ilya Leoshkevich <iii@linux.ibm.com>
2994 * config/s390/predicates.md: Add nonsym_memory_operand.
2995 * config/s390/s390.c (s390_legitimize_cs_operand): If operand
2996 contains a SYMBOL_REF, load it into an intermediate pseudo.
2997 (s390_emit_compare_and_swap): Legitimize operand.
2998 * config/s390/s390.md: Use the new nonsym_memory_operand
2999 with UNSPECV_CAS patterns.
3001 2018-09-06 Ilya Leoshkevich <iii@linux.ibm.com>
3004 * config/s390/s390-passes.def: New file.
3005 * config/s390/s390-protos.h (class rtl_opt_pass): Add forward
3007 (make_pass_s390_early_mach): Add declaration.
3008 * config/s390/s390.c (make_pass_s390_early_mach):
3009 (s390_option_override): Remove dynamic registration.
3010 * config/s390/t-s390: Add s390-passes.def.
3012 2018-09-06 Ilya Leoshkevich <iii@linux.ibm.com>
3014 * config/s390/s390.c (s390_decompose_constant_pool_ref):
3015 Remove UNSPEC_LTREL_BASE check.
3016 (annotate_constant_pool_refs): Likewise.
3017 (find_constant_pool_ref): Likewise.
3018 (find_ltrel_base): Removed.
3019 (replace_ltrel_base): Removed.
3020 (s390_mainpool_finish): Remove replace_ltrel_base call.
3021 (s390_chunkify_start): Remove pending LTREL_BASE logic.
3022 (s390_chunkify_finish): Remove replace_ltrel_base call.
3023 * config/s390/s390.md: Remove UNSPEC_LTREL_BASE.
3025 2018-09-06 Hans-Peter Nilsson <hp@axis.com>
3028 * config/cris/cris.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
3029 to speculation_safe_value_not_needed.
3031 2018-09-05 Cesar Philippidis <cesar@codesourcery.com>
3032 Bernd Schmidt <bernds_cb1@t-online.de>
3034 * config/nvptx/nvptx-opts.h: New file.
3035 * config/nvptx/nvptx.c (nvptx_file_start): Print the correct .target.
3036 * config/nvptx/nvptx.h: Include "nvptx-opts.h".
3038 (TARGET_SM35): New macro.
3039 * config/nvptx/nvptx.md (atomic_fetch_<logic><mode>): Enable with the
3041 * config/nvptx/nvptx.opt (ptx_isa, sm_30, sm_35): New enum and its
3043 (misa=): New option.
3044 * doc/invoke.texi (Nvidia PTX Options): Document -misa.
3046 2018-09-05 Uros Bizjak <ubizjak@gmail.com>
3048 * config/i386/i386.md (truncdfsf2): Remove expander.
3049 (truncdfsf2_with_temp): Ditto.
3050 (truncxf<mode>2): Ditto.
3051 (*truncdfsf_fast_mixed): Remove insn pattern.
3052 (*truncdfsf_fast_i387): Ditto.
3053 (*truncdfsf_mixed): Ditto.
3054 (*truncdfsf_i387): Ditto.
3055 (*truncdfsf2_i387_1): Ditto.
3056 (*truncxfsf2_mixed): Ditto.
3057 (*truncxfdf2_mixed): Ditto.
3058 (*truncxf<mode>2_i387_noop): Ditto. Update callers
3059 to call gen_truncxf<mode>2 instead.
3060 (*truncxf<mode>2_i387): Remove.
3061 (reg->reg splitters): Remove splitter pattern.
3062 (reg->mem splitters): Ditto.
3064 (truncdfsf2): New insn pattern.
3065 (truncxf<mode>2): Ditto.
3067 2018-09-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3069 * tree-ssa-math-opts.c (is_mult_by): New function.
3070 (is_square_of): Use the above.
3071 (optimize_recip_sqrt): New function.
3072 (pass_cse_reciprocals::execute): Use the above.
3074 2018-09-05 Richard Biener <rguenther@suse.de>
3077 * tree-ssa-sccvn.c (rpo_elim::eliminate_push_avail): Make sure
3078 to zero-init the emplaced vec.
3080 2018-09-05 Martin Liska <mliska@suse.cz>
3082 PR tree-optimization/87205
3083 * tree-switch-conversion.c (pass_lower_switch::execute):
3084 Group cases for switch statements.
3086 2018-09-05 Richard Biener <rguenther@suse.de>
3088 PR tree-optimization/87217
3089 * tree-ssa-sccvn.c (vuse_valueize): New.
3090 (vn_reference_lookup_pieces): Use it.
3091 (vn_reference_lookup): Likewise.
3093 2018-09-05 Nathan Sidwell <nathan@acm.org>
3096 * stor-layout.c (place_field): Scan forwards to check last
3097 bitfield when ms_bitfield_placement is in effect.
3099 2018-09-05 Richard Biener <rguenther@suse.de>
3102 * tree-vect-stmts.c (vectorizable_simd_clone_call): Fix bogus
3105 2018-09-05 Siddhesh Poyarekar <siddhesh@sourceware.org>
3106 Kugan Vivekanandarajah <kugan.vivekanandarajah@linaro.org>
3108 * config/aarch64/falkor-tag-collision-avoidance.c: New file.
3109 * config.gcc (extra_objs): Build it.
3110 * config/aarch64/t-aarch64 (falkor-tag-collision-avoidance.o):
3112 * config/aarch64/aarch64-passes.def
3113 (pass_tag_collision_avoidance): New pass.
3114 * config/aarch64/aarch64.c (qdf24xx_tunings): Add
3115 AARCH64_EXTRA_TUNE_RENAME_LOAD_REGS to tuning_flags.
3116 (aarch64_classify_address): Remove static qualifier.
3117 (aarch64_address_info, aarch64_address_type): Move to...
3118 * config/aarch64/aarch64-protos.h: ... here.
3119 (make_pass_tag_collision_avoidance): New function.
3120 * config/aarch64/aarch64-tuning-flags.def (rename_load_regs):
3123 2018-09-05 Martin Liska <mliska@suse.cz>
3125 * doc/gcov.texi: Update documentation of humar
3127 * gcov.c (format_count): Print one decimal place, it provides
3128 more fine number of situations like '1G' vs. '1.4G'.
3130 2018-09-05 Martin Liska <mliska@suse.cz>
3133 * config/rs6000/rs6000.opt: Mark the option as Deprecated.
3134 * optc-gen.awk: Allow 'Var' for Deprecated options in order
3135 to generate a MASK value.
3137 2018-09-04 H.J. Lu <hongjiu.lu@intel.com>
3140 * dwarf2out.c (based_loc_descr): Allow hard frame pointer even
3141 if frame pointer isn't used.
3142 (compute_frame_pointer_to_fb_displacement): Likewise.
3144 2018-09-04 Jakub Jelinek <jakub@redhat.com>
3147 * common/config/i386/i386-common.c (OPTION_MASK_ISA_XSAVEOPT_SET,
3148 OPTION_MASK_ISA_XSAVES_SET, OPTION_MASK_ISA_XSAVEC_SET): Use
3149 OPTION_MASK_ISA_XSAVE_SET instead of OPTION_MASK_ISA_XSAVE.
3150 (OPTION_MASK_ISA_XSAVE_UNSET): Add OPTION_MASK_ISA_XSAVES_UNSET
3151 and OPTION_MASK_ISA_XSAVEC_UNSET.
3153 2018-09-04 Max Filippov <jcmvbkbc@gmail.com>
3155 * config/xtensa/xtensa.c (xtensa_expand_atomic): Reorder AND and
3156 XOR operations in NAND case.
3158 2018-09-04 Aldy Hernandez <aldyh@redhat.com>
3160 * wide-int-range.cc (wide_int_range_convert): New.
3161 * wide-int-range.h (wide_int_range_convert): New.
3162 * tree-vrp.c (extract_range_from_unary_expr): Abstract wide int
3163 code into wide_int_range_convert.
3164 (extract_range_into_wide_ints): Do not munge anti range constants
3165 into the entire domain. Just return the range back.
3167 2018-09-04 Martin Liska <mliska@suse.cz>
3169 * genmatch.c (output_line_directive): Add new argument
3171 (dt_simplify::gen_1): Encapsulate dump within __builtin_expect.
3173 2018-09-04 Jonathan Wakely <jwakely@redhat.com>
3175 * doc/invoke.texi (Option Summary): Add whitespace.
3177 * doc/invoke.texi (Option Summary): Add -Waligned-new.
3179 2018-09-04 Richard Biener <rguenther@suse.de>
3181 PR tree-optimization/87211
3182 * tree-ssa-sccvn.c (visit_phi): When value-numbering to a
3183 backedge value we're supposed to treat as VARYING also number
3184 the PHI to VARYING in case it got a different value-number already.
3186 2018-09-04 Aldy Hernandez <aldyh@redhat.com>
3188 * tree-vrp.c (vrp_can_optimize_bit_op): Remove.
3189 (extract_range_from_binary_expr_1): Do not call
3190 vrp_can_optimize_bit_op.
3191 * wide-int-range.cc (wide_int_range_can_optimize_bit_op): Make
3193 (wide_int_range_get_mask_and_bounds): New.
3194 (wide_int_range_optimize_bit_op): New.
3195 (wide_int_range_bit_ior): Call wide_int_range_optimize_bit_op.
3196 (wide_int_range_bit_and): Same.
3197 * wide-int-range.h (wide_int_range_can_optimize_bit_op): Remove.
3198 (wide_int_range_optimize_bit_op): New.
3199 (wide_int_range_get_mask_and_bounds): New.
3201 2018-09-04 Richard Biener <rguenther@suse.de>
3203 PR tree-optimization/87176
3204 * tree-ssa-sccvn.c (visit_phi): Remove redundant allsame
3205 variable. When value-numbering a virtual PHI node make sure
3206 to not value-number to the backedge value.
3208 2018-09-04 Jonathan Wakely <jwakely@redhat.com>
3210 * doc/extend.texi (Long Long, Hex Floats): Document support for
3211 long long and hex floats in more recent versions of ISO C++.
3213 2018-09-03 Richard Biener <rguenther@suse.de>
3215 PR tree-optimization/87177
3216 * tree-ssa-sccvn.c (vuse_ssa_val): Revert previous change, keep
3219 2018-09-03 Alexander Monakov <amonakov@ispras.ru>
3221 * bb-reorder.c (edge_order): Convert to C-qsort-style
3222 tri-state comparator.
3223 (reorder_basic_blocks_simple): Change std::stable_sort to
3226 2018-09-03 Alexander Monakov <amonakov@ispras.ru>
3228 * tree-loop-distribution.c (offset_cmp): Convert to C-qsort-style
3229 tri-state comparator.
3230 (fuse_memset_builtins): Change std::stable_sort to gcc_stablesort.
3232 2018-09-03 Alexander Monakov <amonakov@ispras.ru>
3234 * sort.cc (struct sort_ctx): New field 'nlim'. Use it...
3235 (mergesort): ... here as maximum count for using netsort.
3236 (gcc_qsort): Set nlim to 3 if stable sort is requested.
3237 (gcc_stablesort): New.
3238 * system.h (gcc_stablesort): Declare.
3240 2018-09-03 Alexander Monakov <amonakov@ispras.ru>
3242 * sort.cc (gcc_qsort) [CHECKING_P]: Call qsort_chk.
3243 * system.h (qsort): Always redirect to gcc_qsort. Update comment.
3244 * vec.c (qsort_chk): Do not call gcc_qsort. Update comment.
3246 2018-09-03 Segher Boessenkool <segher@kernel.crashing.org>
3248 * config/rs6000/rs6000.md (*mov<mode>_hardfloat32): Remove %U from the
3249 lxsdx and stxsdx alternatives.
3250 (*mov<mode>_hardfloat64): Ditto.
3251 * config/rs6000/vsx.md (*vsx_extract_<mode>_store): Ditto.
3253 2018-09-03 Richard Biener <rguenther@suse.de>
3255 PR tree-optimization/87200
3256 * tree-ssa-sccvn.c (vn_nary_build_or_lookup_1): Valueize a
3259 2018-09-03 Martin Liska <mliska@suse.cz>
3261 PR tree-optimization/87201
3262 * tree-switch-conversion.c (switch_decision_tree::balance_case_nodes):
3263 Fix parenthesis in an expression.
3265 2018-09-03 Richard Biener <rguenther@suse.de>
3267 PR tree-optimization/87197
3268 * tree-ssa-sccvn.c (vn_nary_build_or_lookup_1): Mark the new def
3269 visited. CSE the VN_INFO hashtable lookup.
3271 PR tree-optimization/87169
3272 * tree-ssa-sccvn.c (do_rpo_vn): When marking loops for not
3273 iterating make sure there's no extra backedges from irreducible
3274 regions feeding the header. Mark the destination block
3277 2018-09-03 Martin Liska <mliska@suse.cz>
3280 * common/common-target.def: Add TARGET_GET_VALID_OPTION_VALUES.
3281 * common/common-targhooks.c (default_get_valid_option_values):
3283 * common/common-targhooks.h (default_get_valid_option_values):
3285 * common/config/i386/i386-common.c: Move processor_target_table
3287 (ix86_get_valid_option_values): New function.
3288 (TARGET_GET_VALID_OPTION_VALUES): New macro.
3289 * config/i386/i386.c (struct ptt): Move to i386-common.c.
3290 (PTA_*): Move all defined masks into i386-common.c.
3291 (ix86_function_specific_restore): Use new processor_cost_table.
3292 * config/i386/i386.h (struct ptt): Moved from i386.c.
3293 (struct pta): Likewise.
3294 * doc/tm.texi: Document new TARGET_GET_VALID_OPTION_VALUES.
3295 * doc/tm.texi.in: Likewise.
3296 * opt-suggestions.c (option_proposer::suggest_option):
3297 Pass prefix to build_option_suggestions.
3298 (option_proposer::get_completions): Likewise.
3299 (option_proposer::build_option_suggestions): Use the new target
3301 * opts.c (struct option_help_tuple): New struct.
3302 (print_filtered_help): Use the new target hook.
3304 2018-09-03 Martin Liska <mliska@suse.cz>
3307 * predict.c (set_even_probabilities): Add likely_edges
3308 argument and handle cases where we have precisely one
3310 (combine_predictions_for_bb): Catch also likely_edges.
3311 (tree_predict_by_opcode): Handle gswitch statements.
3312 * tree-cfg.h (find_case_label_for_value): New declaration.
3313 (find_taken_edge_switch_expr): Likewise.
3314 * tree-switch-conversion.c (switch_decision_tree::balance_case_nodes):
3315 Find pivot in decision tree based on probabily, not by number of
3318 2018-09-02 Gerald Pfeifer <gerald@pfeifer.com>
3320 * doc/standards.texi (Standards): Update Objective-C reference.
3322 2018-09-01 Gerald Pfeifer <gerald@pfeifer.com>
3324 * doc/install.texi (Prerequisites): Update link for MPC.
3326 2018-09-01 Michael Matz <matz@suse.de>
3328 PR tree-optimization/87074
3329 * gimple-loop-jam.c (unroll_jam_possible_p): Check loop exit
3330 PHIs for outer-loop uses.
3332 2018-09-01 Gerald Pfeifer <gerald@pfeifer.com>
3334 * doc/generic.texi (OpenMP): Adjust link to openmp.org.
3335 * doc/invoke.texi (C Dialect Options): Ditto.
3337 2018-09-01 Gerald Pfeifer <gerald@pfeifer.com>
3339 * doc/install.texi (Prerequisites): Adjust link mpfr.org.
3341 2018-08-31 Richard Biener <rguenther@suse.de>
3343 PR tree-optimization/87168
3344 * tree-ssa-sccvn.c (SSA_VAL): Add visited output parameter.
3345 (rpo_elim::eliminate_avail): When OP was not visited it must
3348 2018-08-31 David Malcolm <dmalcolm@redhat.com>
3350 * tree-vrp.c (copy_value_range): Convert param "from" from
3351 "value_range *" to "const value_range *".
3352 (range_is_null): Likewise for param "vr".
3353 (range_int_cst_p): Likewise.
3354 (range_int_cst_singleton_p): Likewise.
3355 (symbolic_range_p): Likewise.
3356 (value_ranges_intersect_p): Likewise for both params.
3357 (value_range_nonnegative_p): Likewise for param "vr".
3358 (value_range_constant_singleton): Likewise.
3359 (vrp_set_zero_nonzero_bits): Likewise for param "ar".
3360 (extract_range_into_wide_ints): Likewise for param "vr".
3361 (extract_range_from_multiplicative_op): Likewise for params "vr0"
3363 (vrp_can_optimize_bit_op): Likewise.
3364 (extract_range_from_binary_expr_1): Likewise for params "vr0_" and
3366 (extract_range_from_unary_expr): Likewise.
3367 (debug_value_range): Likewise for param "vr".
3368 (value_range::dump): Add "const" qualifier.
3369 (vrp_prop::check_array_ref): Convert local "vr" from
3370 "value_range *" to "const value_range *".
3371 (vrp_prop::check_mem_ref): Likewise.
3372 (vrp_prop::visit_stmt): Likewise for local "old_vr".
3373 (vrp_intersect_ranges_1): Likewise for param "vr_1".
3374 (vrp_intersect_ranges): Likewise.
3375 (simplify_stmt_for_jump_threading): Likewise for local "vr".
3376 (vrp_prop::vrp_finalize): Likewise.
3377 * tree-vrp.h (value_range::dump): Add "const" qualifier.
3378 (vrp_intersect_ranges): Add "const" qualifier to params as above.
3379 (extract_range_from_unary_expr): Likewise.
3380 (value_range_constant_singleton): Likewise.
3381 (symbolic_range_p): Likewise.
3382 (copy_value_range): Likewise.
3383 (extract_range_from_binary_expr_1): Likewise.
3384 (range_int_cst_p): Likewise.
3385 (vrp_set_zero_nonzero_bits): Likewise.
3386 (range_int_cst_singleton_p): Likewise.
3388 2018-08-31 Vlad Lazar <vlad.lazar@arm.com>
3390 * config/aarch64/arm_neon.h (vabsd_s64): New.
3391 (vnegd_s64): Likewise.
3393 2018-08-31 Martin Jambor <mjambor@suse.cz>
3395 * ipa-cp.c (estimate_local_effects): Replace wrong MAX with MIN.
3397 2018-08-31 Martin Liska <mliska@suse.cz>
3399 * ipa-icf.c (sem_item::add_type): Use
3400 sem_item::m_type_hash_cache.
3401 * ipa-icf.h: Move the cache from sem_item_optimizer
3404 2018-08-31 Nathan Sidwell <nathan@acm.org>
3406 * doc/extend.texi (Backwards Compatibility): Remove implicit
3407 extern C leeway of () being (...).
3409 2018-08-31 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3411 * ipa-inline.c (can_inline_edge_by_limits_p): Fix typos in comment.
3413 2018-08-31 Segher Boessenkool <segher@kernel.crashing.org>
3417 * config/rs6000/rs6000.md (lround<mode>di2): Gate on TARGET_FPRND.
3419 2018-08-31 Jakub Jelinek <jakub@redhat.com>
3422 * expmed.c (expand_mult_const): Use immed_wide_int_const instead of
3423 gen_int_mode. Formatting fixes.
3425 2018-08-30 Sandra Loosemore <sandra@codesourcery.com>
3427 * target.def (custom_function_descriptors): Improve documentation.
3428 * doc/tm.texi.in (Trampolines): Expand discussion of function
3429 descriptors and move TARGET_CUSTOM_FUNCTION_DESCRIPTORS to the
3430 beginning of the section.
3431 * doc/tm.texi: Regenerated.
3433 2018-08-30 Jose E. Marchesi <jose.marchesi@oracle.com>
3435 * cfg.h (class auto_edge_flag): Spell out the template-id of the
3436 base class in the initializer list. This is a workaround for
3437 building with older GCC.
3438 (class auto_bb_flag): Likewise.
3440 2018-08-30 Aaron Sawdey <acsawdey@linux.ibm.com>
3442 * config/rs6000/altivec.md (altivec_eq<mode>): Remove star.
3443 (altivec_vcmpequ<VI_char>_p): Remove star.
3444 * config/rs6000/rs6000-string.c (do_load_for_compare): Support
3446 (expand_strncmp_vec_sequence): New function.
3447 (emit_final_str_compare_vec): New function.
3448 (expand_strn_compare): Add support for vector strncmp.
3449 * config/rs6000/rs6000.opt (-mstring-compare-inline-limit): Change
3450 length specification to bytes.
3451 * config/rs6000/vsx.md (vsx_ld_elemrev_v16qi_internal): Remove star.
3452 (vcmpnezb_p): New pattern.
3453 * doc/invoke.texi (RS/6000 and PowerPC Options): Update documentation
3454 for option -mstring-compare-inline-limit.
3456 2018-08-30 Thiago Macieira <thiago.macieira@intel.com>
3458 * config/i386/i386.c (PTA_WESTMERE): Remove PTA_AES.
3459 (PTA_SKYLAKE): Add PTA_AES.
3460 (PTA_GOLDMONT): Likewise.
3462 2018-08-29 Jan Hubicka <jh@suse.cz>
3465 * lto-opts.c (lto_write_options): Always stream PIC/PIE mode.
3466 * lto-wrapper.c (merge_and_complain): Fix merging of PIC/PIE.
3468 2018-08-29 Jan Hubicka <jh@suse.cz>
3470 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not follow
3472 (hash_tree): Do not visit TYPE_STUB_DECL.
3473 * tree-streamer-out.c (write_ts_type_common_tree_pointers): Do not
3474 stream TYPE_STUB_DECL.
3475 * tree-streamer-in.c (lto_input_ts_type_common_tree_pointers): Likewise.
3476 * ipa-utils.h (type_with_linkage_p): Do not rely on TYPE_STUB_DECL
3477 after free_lang_data.
3478 (type_in_anonymous_namespace_p): Likewise.
3480 2018-08-29 Jan Hubicka <jh@suse.cz>
3482 * sreal.h (SREAL_PART_BITS): Change to 31; remove seemingly unnecessary
3483 comment that it has to be even number.
3484 (class sreal): Change m_sig type to int32_t.
3485 * sreal.c (sreal::dump, sreal::to_int, opreator+, operator-): Use
3486 int64_t for temporary calculations.
3487 (sreal_verify_basics): Drop one bit from minimum and maximum.
3489 2018-08-30 Richard Biener <rguenther@suse.de>
3491 PR tree-optimization/87147
3492 * tree-ssa-sccvn.c (SSA_VISITED): New function.
3493 (visit_phi): When the degenerate result is from the backedge and
3494 we didn't visit its definition yet drop to VARYING.
3495 (do_rpo_vn): Properly mark blocks with incoming backedges as executable.
3497 2018-08-29 Jan Hubicka <jh@suse.cz>
3499 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not walk
3501 (hash_tree): Likewise.
3503 2018-08-29 Jan Hubicka <jh@suse.cz>
3505 * tree.c (find_decls_types_r): Walk also TYPE_NEXT_PTR_TO
3506 and TYPE_NEXT_REF_TO.
3508 2018-08-29 Jan Hubicka <jh@suse.cz>
3510 * sreal.h (SREAL_PART_BITS): Change to 31; remove seemingly unnecessary
3511 comment that it has to be even number.
3512 (class sreal): Change m_sig type to int32_t.
3513 * sreal.c (sreal::dump, sreal::to_int, opreator+, operator-): Use
3514 int64_t for temporary calculations.
3515 (sreal_verify_basics): Drop one bit from minimum and maximum.
3517 2018-08-30 Tamar Christina <tamar.christina@arm.com>
3519 * config/aarch64/aarch64.c (aarch64_expand_movmem): Set TImode max.
3521 2018-08-30 Vlad Lazar <vlad.lazar@arm.com>
3524 * expmed.c (canonicalize_comparison): Use wi::sub instead of wi::add
3525 if to_add is negative.
3527 2018-08-29 Bernd Edlinger <bernd.edlinger@hotmail.de>
3530 * builtins.c (c_strlen): Improve range checks.
3532 2018-08-29 Martin Sebor <msebor@redhat.com>
3533 Jeff Law <law@redhat.com>
3535 PR tree-optimization/86714
3536 PR tree-optimization/86711
3537 * builtins.c (c_strlen): Add arguments to call to string_constant.
3538 * expr.c (string_constant): Add argument. Detect missing nul
3539 terminator and outermost declaration it's missing in.
3540 * expr.h (string_constant): Add argument.
3541 * fold-const.c (read_from_constant_string): Add arguments to call to
3543 (c_getstr): Likewise.
3544 * tree-ssa-forwprop.c (simplify_builtin_call): Likewise.
3546 * tree-ssa-strlen.c (get_stridx): Likewise.
3548 2018-08-29 Jan Hubicka <jh@suse.cz>
3550 * tree-streamer-in.c (lto_input_ts_function_decl_tree_pointers):
3551 Do not stream DECL_VINDEX.
3552 * tree-streamer-out.c (write_ts_function_decl_tree_pointers): Likewise.
3553 * tree.c (free_lang_data_in_decl): Clear DECL_VINDEX.
3554 (decl_function_context): Use DECL_VIRTUAL_P rather than DECL_VINDEX.
3556 2018-08-29 Richard Biener <rguenther@suse.de>
3558 * tree-ssa-sccvn.c (vuse_ssa_val): Return NULL for unvisited
3559 virtual operands that are not default defs to honor region
3561 (rpo_vn_valueize): Remove ineffective code here.
3563 2018-08-29 Richard Biener <rguenther@suse.de>
3565 PR tree-optimization/87132
3566 * tree-ssa-alias.c (get_continuation_for_phi): Do not translate
3567 when skipping defs reachable over backedges.
3569 2018-08-29 Richard Biener <rguenther@suse.de>
3571 * tree-core.h: Document use of deprecated_flag in SSA_NAME.
3572 * tree.h (SSA_NAME_POINTS_TO_READONLY_MEMORY): Define.
3573 * tree-into-ssa.c (pass_build_ssa::execute): Initialize
3574 function parameters SSA_NAME_POINTS_TO_READONLY_MEMORY from fnspec.
3575 * tree-ssa-sccvn.c (const_parms, init_const_parms): Remove.
3576 (vn_reference_lookup_3): Remove use of const_parms.
3577 (free_rpo_vn): Do not free const_parms.
3578 (do_rpo_vn): Do not call init_const_parms.
3579 * tree-ssa-alias.c (refs_may_alias_p_1): Honor
3580 SSA_NAME_POINTS_TO_READONLY_MEMORY.
3581 (call_may_clobber_ref_p_1): Likewise.
3583 2018-08-29 Alexander Monakov <amonakov@ispras.ru>
3586 * invoke.texi (Optimization Options): List -ftree-scev-cprop.
3588 (-ftree-scev-cprop): Document.
3590 2018-08-29 Jan Hubicka <jh@suse.cz>
3592 * sreal.h (normalize, normalize_up, normalize_down): Add new_sig/new_exp
3594 (sreal constructor): Update.
3595 * sreal.c (sreal:operator+, sreal:operator-, sreal:operator*,
3596 sreal:operator/): Update.
3598 2018-08-29 Martin Liska <mliska@suse.cz>
3600 * tree-switch-conversion.c (switch_conversion::expand):
3601 Strenghten assumption about gswitch statements.
3603 2018-08-29 Richard Biener <rguenther@suse.de>
3605 PR tree-optimization/87117
3606 * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_stmt): Only
3607 re-value-number released SSA VDEFs.
3609 2018-08-29 Richard Biener <rguenther@suse.de>
3611 PR tree-optimization/87126
3612 * tree-ssa-sccvn.c (vn_reference_insert): Remove assert.
3614 2018-08-28 Jim Wilson <jimw@sifive.com>
3616 * config/riscv/pic.md: Rewrite.
3617 * config/riscv/riscv.c (riscv_address_insns): Return cost of 3 for
3619 * config/riscv/riscv.md (ZERO_EXTEND_LOAD): Delete.
3620 (SOFTF, default_load, softload, softstore): New.
3622 2018-08-28 Jeff Law <law@redhat.com>
3624 * fold-const.c (fold_binary_loc): Remove recently added assert.
3626 2018-08-28 Joern Rennecke <joern.rennecke@riscy-ip.com>
3628 * genpreds.c (write_predicate_subfunction): Also add ATTRIBUTE_UNUSED
3629 to OP parmeter of generated function.
3631 2018-08-28 MCC CS <deswurstes@users.noreply.github.com>
3633 PR tree-optimization/87009
3634 * match.pd: Add boolean optimizations.
3636 2018-08-28 Martin Sebor <msebor@redhat.com>
3639 * calls.c (alloc_max_size): Treat HOST_WIDE_INT special.
3640 * gimple-ssa-warn-alloca.c (adjusted_warn_limit): New function.
3641 (pass_walloca::gate): Use it.
3642 (alloca_call_type): Same.
3643 (pass_walloca::execute): Same.
3644 * stor-layout.c (layout_decl): Treat HOST_WIDE_INT special.
3646 2018-08-28 David Malcolm <dmalcolm@redhat.com>
3648 * dumpfile.h (ATTRIBUTE_GCC_DUMP_PRINTF): Change version check on
3649 GCC_VERSION for usage of "__gcc_dump_printf__" format from
3652 2018-08-28 Richard Biener <rguenther@suse.de>
3654 PR tree-optimization/87124
3655 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Guard against
3656 constants before looking up avail.
3658 2018-08-28 Jakub Jelinek <jakub@redhat.com>
3661 * calls.c (maybe_warn_nonstring_arg): Punt early if
3662 warn_stringop_overflow is zero. Don't call get_range_strlen
3663 on 3rd argument, keep iterating until lenrng[1] is INTEGER_CST.
3664 Swap comparison operands to have constants on rhs. Only use
3665 lenrng[1] if non-NULL and INTEGER_CST. Don't uselessly
3666 increment lenrng[0].
3668 2018-08-28 Richard Sandiford <richard.sandiford@arm.com>
3670 * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Fix unguarded
3671 use of tree_to_shwi. Remove duplicated test for the size being
3672 a whole number of bytes.
3674 2018-08-28 Richard Biener <rguenther@suse.de>
3676 PR tree-optimization/87117
3677 * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_cleanup):
3678 Handle removed stmt without LHS (GIMPLE_NOP).
3680 2018-08-28 Richard Biener <rguenther@suse.de>
3682 PR tree-optimization/87117
3683 * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Exclude
3684 void which is is_gimple_reg_type by checking for COMPLETE_TYPE_P.
3686 2018-08-28 Richard Biener <rguenther@suse.de>
3688 PR tree-optimization/87117
3689 * tree-ssa-pre.c (compute_avail): Do not make expressions
3690 with predicated values available.
3691 (get_expr_value_id): Assert we do not run into predicated value
3694 2018-08-28 Richard Biener <rguenther@suse.de>
3696 PR tree-optimization/87117
3697 * tree-ssa-operands.c (add_stmt_operand): STRING_CST may
3698 get virtual operands.
3699 (get_expr_operands): Handle STRING_CST like other decls.
3701 2018-08-28 Martin Liska <mliska@suse.cz>
3703 * tree.h: Update documentation of fndecl_built_in_p
3707 2018-08-27 Jeff Law <law@redhat.com>
3708 PR tree-optimization/87110
3709 * tree-ssa-dse.c (compute_trims): Handle non-constant
3712 2018-08-27 Martin Sebor <msebor@redhat.com>
3714 PR tree-optimization/86914
3715 * tree-ssa-strlen.c (maybe_set_strlen_range): Avoid MEM_REF.
3717 2018-08-27 Martin Sebor <msebor@redhat.com>
3719 PR tree-optimization/87112
3720 * builtins.c (expand_builtin_strnlen): Convert c_strlen result to
3721 the type of the bound argument.
3723 2018-08-27 Jeff Law <law@redhat.com>
3725 * tree-ssa-dse.c (compute_trims): Handle case where the reference's
3726 type does not have a TYPE_SIZE_UNIT.
3728 2018-08-27 Steve Ellcey <sellcey@cavium.com>
3730 * config/aarch64/aarch64-speculation.cc: Replace include of cfg.h
3731 with include of backend.h.
3733 2018-08-27 Richard Biener <rguenther@suse.de>
3735 PR tree-optimization/86927
3736 * tree-vect-loop.c (vect_create_epilog_for_reduction): Properly
3737 use const cond reduction code.
3739 2018-08-27 Alexander Monakov <amonakov@ispras.ru>
3741 PR tree-optimization/85758
3742 * match.pd ((X & Y) ^ Y): Add :s qualifier to inner expression.
3744 2018-08-27 David Malcolm <dmalcolm@redhat.com>
3747 * diagnostic-show-locus.c (class layout_range): Update for
3748 conversion of show_caret_p to a tri-state.
3749 (layout_range::layout_range): Likewise.
3750 (make_range): Likewise.
3751 (layout::maybe_add_location_range): Likewise.
3752 (layout::should_print_annotation_line_p): Don't show annotation
3753 lines for ranges that are SHOW_LINES_WITHOUT_RANGE.
3754 (layout::get_state_at_point): Update for conversion of
3755 show_caret_p to a tri-state. Bail out early for
3756 SHOW_LINES_WITHOUT_RANGE, so that such ranges don't affect
3757 underlining or source colorization.
3758 (gcc_rich_location::add_location_if_nearby): Update for conversion
3759 of show_caret_p to a tri-state.
3760 (selftest::test_one_liner_multiple_carets_and_ranges): Likewise.
3761 (selftest::test_one_liner_fixit_replace_equal_secondary_range):
3763 (selftest::test_one_liner_labels): Likewise.
3764 * gcc-rich-location.c (gcc_rich_location::add_expr): Update for
3765 conversion of show_caret_p to a tri-state.
3766 * pretty-print.c (text_info::set_location): Likewise.
3767 * pretty-print.h (text_info::set_location): Likewise.
3768 * substring-locations.c (format_warning_n_va): Likewise.
3769 * tree-diagnostic.c (default_tree_printer): Likewise.
3770 * tree-pretty-print.c (newline_and_indent): Likewise.
3772 2018-08-27 David Malcolm <dmalcolm@redhat.com>
3775 * diagnostic-show-locus.c (get_line_span_for_fixit_hint): Show the
3776 line above for line-insertion fix-it hints.
3777 (selftest::test_fixit_insert_containing_newline): Update the
3778 expected results, and add a test with line-numbering enabled.
3780 2018-08-27 Martin Liska <mliska@suse.cz>
3783 * sanopt.c (sanitize_rewrite_addressable_params): Ignore
3784 params with DECL_HAS_VALUE_EXPR_P.
3786 2018-08-27 Martin Liska <mliska@suse.cz>
3788 * config/i386/i386.c (ix86_expand_set_or_movmem): Dump
3789 selected expansion strategy.
3791 2018-08-27 Martin Liska <mliska@suse.cz>
3793 * builtins.h (is_builtin_fn): Remove and fndecl_built_in_p.
3794 * builtins.c (is_builtin_fn): Likewise.
3795 * attribs.c (diag_attr_exclusions): Use new function
3796 fndecl_built_in_p and remove check for FUNCTION_DECL if
3798 (builtin_mathfn_code): Likewise.
3799 (fold_builtin_expect): Likewise.
3800 (fold_call_expr): Likewise.
3801 (fold_builtin_call_array): Likewise.
3802 (fold_call_stmt): Likewise.
3803 (set_builtin_user_assembler_name): Likewise.
3804 (is_simple_builtin): Likewise.
3805 * calls.c (gimple_alloca_call_p): Likewise.
3806 (maybe_warn_nonstring_arg): Likewise.
3807 * cfgexpand.c (expand_call_stmt): Likewise.
3808 * cgraph.c (cgraph_update_edges_for_call_stmt_node): Likewise.
3809 (cgraph_edge::verify_corresponds_to_fndecl): Likewise.
3810 (cgraph_node::verify_node): Likewise.
3811 * cgraphclones.c (build_function_decl_skip_args): Likewise.
3812 (cgraph_node::create_clone): Likewise.
3813 * config/arm/arm.c (arm_insert_attributes): Likewise.
3814 * config/i386/i386.c (ix86_gimple_fold_builtin): Likewise.
3815 * dse.c (scan_insn): Likewise.
3816 * expr.c (expand_expr_real_1): Likewise.
3817 * fold-const.c (operand_equal_p): Likewise.
3818 (fold_binary_loc): Likewise.
3819 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Likewise.
3820 * gimple-low.c (lower_stmt): Likewise.
3821 * gimple-pretty-print.c (dump_gimple_call): Likewise.
3822 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
3824 * gimple.c (gimple_build_call_from_tree): Likewise.
3825 (gimple_call_builtin_p): Likewise.
3826 (gimple_call_combined_fn): Likewise.
3827 * gimplify.c (gimplify_call_expr): Likewise.
3828 (gimple_boolify): Likewise.
3829 (gimplify_modify_expr): Likewise.
3830 (gimplify_addr_expr): Likewise.
3831 * hsa-gen.c (gen_hsa_insns_for_call): Likewise.
3832 * ipa-cp.c (determine_versionability): Likewise.
3833 * ipa-fnsummary.c (compute_fn_summary): Likewise.
3834 * ipa-param-manipulation.c (ipa_modify_formal_parameters): Likewise.
3835 * ipa-split.c (visit_bb): Likewise.
3836 (split_function): Likewise.
3837 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
3838 * lto-cgraph.c (input_node): Likewise.
3839 * lto-streamer-out.c (write_symbol): Likewise.
3840 * omp-low.c (setjmp_or_longjmp_p): Likewise.
3841 (lower_omp_1): Likewise.
3842 * predict.c (strip_predict_hints): Likewise.
3843 * print-tree.c (print_node): Likewise.
3844 * symtab.c (symtab_node::output_to_lto_symbol_table_p): Likewise.
3845 * trans-mem.c (is_tm_irrevocable): Likewise.
3846 (is_tm_load): Likewise.
3847 (is_tm_simple_load): Likewise.
3848 (is_tm_store): Likewise.
3849 (is_tm_simple_store): Likewise.
3850 (is_tm_abort): Likewise.
3851 (tm_region_init_1): Likewise.
3852 * tree-call-cdce.c (gen_shrink_wrap_conditions): Likewise.
3853 * tree-cfg.c (verify_gimple_call): Likewise.
3854 (move_stmt_r): Likewise.
3855 (stmt_can_terminate_bb_p): Likewise.
3856 * tree-eh.c (lower_eh_constructs_2): Likewise.
3857 * tree-if-conv.c (if_convertible_stmt_p): Likewise.
3858 * tree-inline.c (remap_gimple_stmt): Likewise.
3859 (copy_bb): Likewise.
3860 (estimate_num_insns): Likewise.
3861 (fold_marked_statements): Likewise.
3862 * tree-sra.c (scan_function): Likewise.
3863 * tree-ssa-ccp.c (surely_varying_stmt_p): Likewise.
3864 (optimize_stack_restore): Likewise.
3865 (pass_fold_builtins::execute): Likewise.
3866 * tree-ssa-dce.c (mark_stmt_if_obviously_necessary): Likewise.
3867 (mark_all_reaching_defs_necessary_1): Likewise.
3868 * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Likewise.
3869 * tree-ssa-forwprop.c (simplify_builtin_call): Likewise.
3870 (pass_forwprop::execute): Likewise.
3871 * tree-ssa-loop-im.c (stmt_cost): Likewise.
3872 * tree-ssa-math-opts.c (pass_cse_reciprocals::execute): Likewise.
3873 * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Likewise.
3874 * tree-ssa-strlen.c (get_string_length): Likewise.
3875 * tree-ssa-structalias.c (handle_lhs_call): Likewise.
3876 (find_func_aliases_for_call): Likewise.
3877 * tree-ssa-ter.c (find_replaceable_in_bb): Likewise.
3878 * tree-stdarg.c (optimize_va_list_gpr_fpr_size): Likewise.
3879 * tree-tailcall.c (find_tail_calls): Likewise.
3880 * tree.c (need_assembler_name_p): Likewise.
3881 (free_lang_data_in_decl): Likewise.
3882 (get_call_combined_fn): Likewise.
3883 * ubsan.c (is_ubsan_builtin_p): Likewise.
3884 * varasm.c (incorporeal_function_p): Likewise.
3885 * tree.h (DECL_BUILT_IN): Remove and replace with
3887 (DECL_BUILT_IN_P): Transfort to fndecl_built_in_p.
3888 (fndecl_built_in_p): New.
3890 2018-08-27 Martin Liska <mliska@suse.cz>
3892 PR tree-optimization/86847
3893 * tree-switch-conversion.c (switch_decision_tree::dump_case_nodes):
3894 Dump also subtree probability.
3895 (switch_decision_tree::do_jump_if_equal): New function.
3896 (switch_decision_tree::emit_case_nodes): Handle special
3897 situations in balanced tree that can be emitted much simpler.
3898 Fix calculation of probabilities that happen in tree expansion.
3899 * tree-switch-conversion.h (struct cluster): Add
3901 (struct simple_cluster): Likewise.
3902 (struct case_tree_node): Add new function has_child.
3903 (do_jump_if_equal): New.
3905 2018-08-27 Martin Liska <mliska@suse.cz>
3907 * tree-switch-conversion.c (bit_test_cluster::find_bit_tests):
3908 Add new argument to bit_test_cluster constructor.
3909 (bit_test_cluster::emit): Set bits really number of values
3911 (bit_test_cluster::hoist_edge_and_branch_if_true): Add
3912 probability argument.
3913 * tree-switch-conversion.h (struct bit_test_cluster):
3914 Add m_handles_entire_switch.
3916 2018-08-27 Martin Liska <mliska@suse.cz>
3918 PR tree-optimization/86702
3919 * tree-switch-conversion.c (jump_table_cluster::emit):
3920 Make probabilities even for values in jump table
3921 according to number of cases handled.
3922 (switch_decision_tree::compute_cases_per_edge): Pass
3923 argument to reset_out_edges_aux function.
3924 (switch_decision_tree::analyze_switch_statement): Likewise.
3925 * tree-switch-conversion.h (switch_decision_tree::reset_out_edges_aux):
3928 2018-08-27 Martin Liska <mliska@suse.cz>
3930 * cfgexpand.c (expand_asm_stmt): Use label_to_block and pass
3931 cfun argument explicitly.
3932 * gimple-pretty-print.c (dump_gimple_switch): Likewise.
3933 * hsa-gen.c (gen_hsa_insns_for_switch_stmt): Use new
3934 function gimple_switch_default_bb.
3935 (convert_switch_statements):
3937 * ipa-fnsummary.c (set_switch_stmt_execution_predicate):
3938 * stmt.c (label_to_block_fn): Use label_to_block and pass
3939 cfun argument explicitly and use gimple_switch_label_bb.
3940 (expand_case): Likewise.
3941 * tree-cfg.c (lower_phi_internal_fn): Use label_to_block and pass
3942 cfun argument explicitly. Likewise.
3943 (make_edges_bb): Likewise.
3944 (make_cond_expr_edges): Likewise.
3945 (get_cases_for_edge): Likewise.
3946 (make_gimple_switch_edges): Likewise.
3947 (label_to_block_fn): Likewise.
3948 (label_to_block): Likewise.
3949 (make_goto_expr_edges): Likewise.
3950 (make_gimple_asm_edges): Likewise.
3951 (main_block_label): Likewise.
3952 (group_case_labels_stmt): Likewise.
3953 (find_taken_edge_computed_goto): Likewise.
3954 (find_taken_edge_switch_expr): Likewise.
3955 (gimple_verify_flow_info): Likewise.
3956 (gimple_redirect_edge_and_branch): Likewise.
3957 (gimple_switch_label_bb): New function.
3958 (gimple_switch_default_bb): Likewise.
3959 (gimple_switch_edge): Likewise.
3960 (gimple_switch_default_edge): Likewise.
3961 * tree-cfg.h (label_to_block_fn): Remove and replace ...
3962 (label_to_block): ... with this.
3963 (gimple_switch_label_bb): New.
3964 (gimple_switch_default_bb): Likewise.
3965 (gimple_switch_edge): Likewise.
3966 (gimple_switch_default_edge): Likewise.
3967 * tree-cfgcleanup.c (convert_single_case_switch): Use
3968 new gimple functions and pass new argument to label_to_block.
3969 (cleanup_control_flow_bb):
3970 * tree-eh.c (make_eh_dispatch_edges): Use label_to_block and pass
3971 cfun argument explicitly.
3972 (make_eh_edges): Likewise.
3973 (redirect_eh_dispatch_edge): Likewise.
3974 (lower_resx): Likewise.
3975 (lower_eh_dispatch): Likewise.
3976 (maybe_remove_unreachable_handlers): Likewise.
3977 (unsplit_eh): Likewise.
3978 (cleanup_empty_eh): Likewise.
3979 (verify_eh_edges): Likewise.
3980 (verify_eh_dispatch_edge): Likewise.
3981 * tree-ssa-dom.c (record_edge_info): Likewise.
3982 * tree-ssa-forwprop.c (simplify_gimple_switch_label_vec): Likewise.
3983 * tree-ssa-threadedge.c (thread_around_empty_blocks): Likewise.
3984 (thread_through_normal_block): Likewise.
3985 * tree-ssa-uncprop.c (associate_equivalences_with_edges): Likewise.
3986 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds):
3987 * tree-switch-conversion.c (switch_conversion::collect): Use new
3989 (switch_conversion::check_final_bb): Likewise.
3990 (switch_conversion::gather_default_values): Pass new argument
3992 (switch_conversion::build_constructors): Likewise.
3993 (switch_decision_tree::compute_cases_per_edge): Use new
3994 gimple_switch_edge function.
3995 (switch_decision_tree::analyze_switch_statement): Pass new argument
3997 (switch_decision_tree::try_switch_expansion): Use
3998 gimple_switch_default_edge.
3999 * tree-vrp.c (find_switch_asserts): Pass new argument
4001 * vr-values.c (vr_values::vrp_visit_switch_stmt): Likewise.
4002 (vr_values::simplify_switch_using_ranges): Likewise.
4004 2018-08-27 Richard Biener <rguenther@suse.de>
4006 * cfganal.h (rev_post_order_and_mark_dfs_back_seme): Declare.
4007 * cfganal.c (rev_post_order_and_mark_dfs_back_seme): New function.
4009 * tree-ssa-sccvn.h (struct vn_pval): New structure.
4010 (struct vn_nary_op_s): Add unwind_to member. Add
4011 predicated_values flag and put result into a union together
4012 with a linked list of vn_pval.
4013 (struct vn_ssa_aux): Add name member to make maintaining
4014 a map of SSA name to vn_ssa_aux possible. Remove no longer
4015 needed info, dfsnum, low, visited, on_sccstack, use_processed
4016 and range_info_anti_range_p members.
4017 (run_scc_vn, vn_eliminate, free_scc_vn, vn_valueize): Remove.
4018 (do_rpo_vn, run_rpo_vn, eliminate_with_rpo_vn, free_rpo_vn):
4020 (vn_valueize): New global.
4021 (vn_context_bb): Likewise.
4022 (VN_INFO_RANGE_INFO, VN_INFO_ANTI_RANGE_P, VN_INFO_RANGE_TYPE,
4023 VN_INFO_PTR_INFO): Remove.
4024 * tree-ssa-sccvn.c: ... (rewrite)
4025 (pass_fre::execute): For -O2+ initialize loops and run
4026 RPO VN in optimistic mode (iterating). For -O1 and -Og
4027 run RPO VN in non-optimistic mode.
4028 * params.def (PARAM_SCCVN_MAX_SCC_SIZE): Remove.
4029 (PARAM_RPO_VN_MAX_LOOP_DEPTH): Add.
4030 * doc/invoke.texi (sccvn-max-scc-size): Remove.
4031 (rpo-vn-max-loop-depth): Document.
4032 * tree-ssa-alias.c (walk_non_aliased_vuses): Stop walking
4033 when valuezing the VUSE signals we walked out of the region.
4034 * tree-ssa-pre.c (phi_translate_1): Ignore predicated values.
4035 (phi_translate): Set VN context block to use for availability
4037 (compute_avail): Likewise.
4038 (pre_valueize): New function.
4039 (pass_pre::execute): Adjust to the RPO VN API.
4041 * tree-ssa-loop-ivcanon.c: Include tree-ssa-sccvn.h.
4042 (propagate_constants_for_unrolling): Remove.
4043 (tree_unroll_loops_completely): Perform value-numbering
4044 on the unrolled bodies loop parent.
4046 2018-08-27 Richard Biener <rguenther@suse.de>
4048 * tree-ssa-pre.c (compute_antic): Re-use inverted postorder
4049 for partial antic compute.
4051 2018-08-27 Jakub Jelinek <jakub@redhat.com>
4053 PR rtl-optimization/87065
4054 * combine.c (simplify_if_then_else): Formatting fix.
4055 (if_then_else_cond): Guard MULT optimization with SCALAR_INT_MODE_P
4057 (known_cond): Don't return const_true_rtx for vector modes. Use
4058 CONST0_RTX instead of const0_rtx. Formatting fixes.
4060 2018-08-27 Martin Liska <mliska@suse.cz>
4062 PR gcov-profile/87069
4063 * gcov.c (process_file): Record files already processed
4064 and warn about a file being processed multiple times.
4066 2018-08-27 Martin Liska <mliska@suse.cz>
4069 * config/aarch64/aarch64.c (aarch64_override_options_internal):
4070 Set default values for x_aarch64_*_string strings.
4071 * config/aarch64/aarch64.opt: Remove --{march,mcpu,mtune}==
4072 prefix. For -mabi do not print '=ABI' in help and use
4073 <option_value> format for -msve-vector-bits and -moverride
4076 2018-08-26 Jeff Law <law@redhat.com>
4078 * config/mips/frame-header-opt.c: Include "backend.h" rather than
4081 2018-08-26 Marek Polacek <polacek@redhat.com>
4083 PR c++/87029, Implement -Wredundant-move.
4084 * doc/invoke.texi: Document -Wredundant-move.
4086 2018-08-25 Martin Sebor <msebor@redhat.com>
4088 PR tree-optimization/87059
4089 * builtins.c (expand_builtin_strncmp): Convert MIN_EXPR operand
4090 to the same type as the other.
4091 * fold-const.c (fold_binary_loc): Assert expectation.
4093 2018-08-25 Iain Sandoe <iain@sandoe.co.uk>
4095 * config/darwin.c (machopic_legitimize_pic_address): Clean up
4096 extraneous parentheses, dead code section and formatting.
4098 2018-08-24 David Malcolm <dmalcolm@redhat.com>
4101 * diagnostic-show-locus.c (layout::layout): Ensure the margin is
4102 wide enough for jumps in the line-numbering to be visible.
4103 (layout::print_gap_in_line_numbering): New member function.
4104 (layout::calculate_line_spans): When using line numbering, merge
4105 line spans that are only 1 line apart.
4106 (diagnostic_show_locus): When printing line numbers, show gaps in
4107 line numbering directly, rather than printing headers.
4108 (selftest::test_diagnostic_show_locus_fixit_lines): Add test of
4109 line-numbering with multiple line spans.
4110 (selftest::test_fixit_insert_containing_newline_2): Add test of
4111 line-numbering, in which the spans are close enough to be merged.
4113 2018-08-24 Aldy Hernandez <aldyh@redhat.com>
4115 * gimple-ssa-evrp-analyze.c (set_ssa_range_info): Pass value_range
4116 to range_includes_zero_p. Do not special case VR_ANTI_RANGE.
4117 * tree-vrp.c (range_is_nonnull): Remove.
4118 (range_includes_zero_p): Accept value_range instead of min/max.
4119 (extract_range_from_binary_expr_1): Do not early bail on
4121 Use range_includes_zero_p instead of range_is_nonnull.
4122 (extract_range_from_unary_expr): Use range_includes_zero_p instead
4123 of range_is_nonnull.
4124 (vrp_meet_1): Pass value_range to range_includes_zero_p. Do not
4125 special case VR_ANTI_RANGE.
4126 (vrp_finalize): Same.
4127 * tree-vrp.h (range_includes_zero_p): Pass value_range as argument
4129 (range_is_nonnull): Remove.
4130 * vr-values.c (vrp_stmt_computes_nonzero): Use
4131 range_includes_zero_p instead of range_is_nonnull.
4132 (extract_range_basic): Pass value_range to range_includes_zero_p
4133 instead of range_is_nonnull.
4135 2018-08-24 Uros Bizjak <ubizjak@gmail.com>
4137 * emit-rtl.c (init_emit_once): Do not emit MODE_POINTER_BOUNDS RTXes.
4138 * emit-rtl.h (rtl_data): Remove return_bnd.
4139 * explow.c (trunc_int_for_mode): Do not handle POINTER_BOUNDS_MODE_P.
4140 * function.c (diddle_return_value): Do not handle crtl->return_bnd.
4141 * genmodes.c (complete_mode): Do not handle MODE_POINTER_BOUNDS.
4142 (POINTER_BOUNDS_MODE): Remove definition.
4143 (make_pointer_bounds_mode): Remove.
4144 (get_mode_class): Do not handle MODE_POINTER_BOUNDS.
4145 * machmode.h (POINTER_BOUNDS_MODE_P): Remove definition.
4146 (scalare_mode::includes_p): Do not handle MODE_POINTER_BOUNDS.
4147 * mode-classes.def: Do not define MODE_POINTER_BOUNDS.
4148 * stor-layout.c (int_mode_for_mode): Do not handle MODE_POINTER_BOUNDS.
4149 * tree-core.h (enum tree_index): Remove TI_POINTER_BOUNDS_TYPE.
4150 * varasm.c (output_constant_pool_2): Do not handle MODE_POINTER_BOUNDS.
4152 * config/i386/i386-modes.def (BND32, BND64): Remove.
4153 * config/i386/i386.c (dbx_register_map): Remove bound registers.
4154 (dbx64_register_map): Ditto.
4155 (svr4_dbx_register_map): Ditto.
4156 (indirect_thunk_bnd_needed): Remove.
4157 (indirect_thunks_bnd_used): Ditto.
4158 (indirect_return_bnd_needed): Ditto.
4159 (indirect_return_via_cx_bnd): Ditto.
4160 (enum indirect_thunk_prefix): Remove indirect_thunk_prefix_bnd.
4161 (indirect_thunk_name): Remove handling of indirect_thunk_prefix_bnd.
4162 (output_indirect_thunk): Ditto. Remove need_prefix argument.
4163 (output_indirect_thunk_function): Remove handling of
4164 indirect_return_bnd_needed, indirect_return_via_cx_bnd,
4165 indirect_thunk_bnd_needed and indirect_thunks_bnd_used variables.
4166 (ix86_save_reg): Remove handling of crtl->return_bnd.
4167 (ix86_legitimate_constant_p): Remove handling of POINTER_BOUNDS_MODE_P.
4168 (ix86_print_operand_address_as): Remove handling of UNSPEC_BNDMK_ADDR
4169 and UNSPEC_BNDLX_ADDR.
4170 (ix86_output_indirect_branch_via_reg): Remove handling of
4171 indirect_thunk_prefix_bnd.
4172 (ix86_output_indirect_branch_via_push): Ditto.
4173 (ix86_output_function_return): Ditto.
4174 (ix86_output_indirect_function_return): Ditto.
4175 (avoid_func_arg_motion): Do not handle UNSPEC_BNDSTX.
4176 * config/i386/i386.h (FIXED_REGISTERS): Remove bound registers.
4177 (CALL_USED_REGISTERS): Ditto.
4178 (REG_ALLOC_ORDER): Update for removal of bound registers.
4179 (HI_REGISTER_NAMES): Ditto.
4180 * config/i386/i386.md (UNSPEC_BNDMK, UNSPEC_BNDMK_ADDR, UNSPEC_BNDSTX)
4181 (UNSPEC_BNDLDX, UNSPEC_BNDLDX_ADDR, UNSPEC_BNDCL, UNSPEC_BNDCU)
4182 (UNSPEC_BNDCN, UNSPEC_MPX_FENCE): Remove.
4183 (BND0_REG, BND1_REG, BND2_REG, BND3_REG): Remove
4184 (FIRST_PSEUDO_REG): Update.
4185 (BND): Remove mode iterator.
4186 * config/i386/predicates.md (bnd_mem_operator): Remove.
4188 2018-08-24 Richard Sandiford <richard.sandiford@arm.com>
4190 * tree-vect-stmts.c (vectorizable_bswap): Handle variable-length
4193 2018-08-24 Richard Sandiford <richard.sandiford@arm.com>
4195 * tree-vect-slp.c (vect_transform_slp_perm_load): Separate out
4196 the case in which the permute needs only a single element and
4197 repeats for every vector of the result. Extend that case to
4198 handle variable-length vectors.
4199 * tree-vect-stmts.c (vectorizable_load): Update accordingly.
4201 2018-08-24 H.J. Lu <hongjiu.lu@intel.com>
4204 * dwarf2out.c (save_macinfo_strings): Call set_indirect_string
4205 on DW_MACINFO_start_file for -gsplit-dwarf -g3.
4207 2018-08-24 Richard Biener <rguenther@suse.de>
4209 * cfg.h (struct control_flow_graph): Add edge_flags_allocated and
4210 bb_flags_allocated members.
4211 (auto_flag): New RAII class for allocating flags.
4212 (auto_edge_flag): New RAII class for allocating edge flags.
4213 (auto_bb_flag): New RAII class for allocating bb flags.
4214 * cfgloop.c (verify_loop_structure): Allocate temporary edge
4216 * cfganal.c (dfs_enumerate_from): Remove use of visited sbitmap
4217 in favor of temporarily allocated BB flag.
4218 * hsa-brig.c: Re-order includes.
4219 * hsa-dump.c: Likewise.
4220 * hsa-regalloc.c: Likewise.
4221 * print-rtl.c: Likewise.
4222 * profile-count.c: Likewise.
4224 2018-08-24 Segher Boessenkool <segher@kernel.crashing.org>
4227 * config/rs6000/rs6000.c (toc_relative_expr_p): Check that the base is
4230 2018-08-24 Aldy Hernandez <aldyh@redhat.com>
4233 * wide-int-range.cc (wide_int_range_div): Do not ignore result
4234 from wide_int_range_multiplicative_op.
4236 2018-08-23 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
4238 * tree-vect-data-refs.c (vect_grouped_store_supported): Fix typo
4241 2018-08-23 Giuliano Belinassi <giuliano.belinassi@usp.br>
4243 * genmatch.c (parser::parse_operation): Fix typo 'exapnded'
4246 2018-08-23 Alexander Monakov <amonakov@ispras.ru>
4248 * tree-scalar-evolution.c (final_value_replacement_loop): Dump
4249 full GENERIC expression used for replacement.
4251 2018-08-23 Aldy Hernandez <aldyh@redhat.com>
4253 * tree-vrp.c (abs_extent_range): Remove.
4254 (extract_range_into_wide_ints): Pass wide ints by reference.
4255 (extract_range_from_binary_expr_1): Rewrite the *DIV_EXPR code.
4256 Pass wide ints by reference in all calls to
4257 extract_range_into_wide_ints.
4258 * wide-int-range.cc (wide_int_range_div): New.
4259 * wide-int-range.h (wide_int_range_div): New.
4260 (wide_int_range_includes_zero_p): New.
4261 (wide_int_range_zero_p): New.
4263 2018-08-23 Matthew Malcomson <matthew.malcomson@arm.com>
4265 * config/aarch64/aarch64.md (arches): New enum.
4266 (arch): New enum attr.
4267 (arch_enabled): New attr.
4268 (enabled): Now uses arch_enabled only.
4269 (simd, sve, fp16): Removed attribute.
4270 (fp): Attr now defined in terms of 'arch'.
4271 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64, *movti_aarch64,
4272 *movhf_aarch64, <optab><fcvt_target><GPF:mode>2,
4273 <FCVT_F2FIXED:fcvt_fixed_insn><GPF:mode>3,
4274 <FCVT_FIXED2F:fcvt_fixed_insn><GPI:mode>3): Merge 'fp' and 'simd'
4275 attributes into 'arch'.
4276 (*movsf_aarch64, *movdf_aarch64, *movtf_aarch64, *add<mode>3_aarch64,
4277 subdi3, neg<mode>2, <optab><mode>3, one_cmpl<mode>2,
4278 *<NLOGICAL:optab>_one_cmpl<mode>3, *xor_one_cmpl<mode>3,
4279 *aarch64_ashl_sisd_or_int_<mode>3, *aarch64_lshr_sisd_or_int_<mode>3,
4280 *aarch64_ashr_sisd_or_int_<mode>3, *aarch64_sisd_ushl): Convert use of
4281 'simd' attribute into 'arch'.
4282 (load_pair_sw_<SX:mode><SX2:mode>, load_pair_dw_<DX:mode><DX2:mode>,
4283 store_pair_sw_<SX:mode><SX2:mode>, store_pair_dw_<DX:mode><DX2:mode>):
4284 Convert use of 'fp' attribute to 'arch'.
4285 * config/aarch64/aarch64-simd.md (move_lo_quad_internal_<mode>,
4286 move_lo_quad_internal_<mode>): (different modes) Merge 'fp' and 'simd'
4288 (move_lo_quad_internal_be_<mode>, move_lo_quad_internal_be_<mode>):
4289 (different modes) Merge 'fp' and 'simd' into 'arch'.
4290 (*aarch64_combinez<mode>, *aarch64_combinez_be<mode>): Merge 'fp' and
4293 2018-08-23 Segher Boessenkool <segher@kernel.crashing.org>
4295 PR rtl-optimization/87026
4296 * expmed.c (canonicalize_comparison): If we can no longer create
4297 pseudoregisters, don't.
4299 2018-08-23 Richard Earnshaw <rearnsha@arm.com>
4302 * config/arm/arm-protos.h (arm_emit_speculation_barrier): New
4304 * config/arm/arm.c (speculation_barrier_libfunc): New static
4306 (arm_init_libfuncs): Initialize it.
4307 (arm_emit_speculation_barrier): New function.
4308 * config/arm/arm.md (speculation_barrier): Call
4309 arm_emit_speculation_barrier for architectures that do not have
4311 (speculation_barrier_insn): Only match on Armv7 or later.
4313 2018-08-23 Richard Biener <rguenther@suse.de>
4316 * tree-inline.c (copy_bb): Drop unused __builtin_va_arg_pack_len
4319 2018-08-23 Richard Sandiford <richard.sandiford@arm.com>
4321 * config/aarch64/aarch64.c (aarch64_evpc_sve_tbl): Fix handling
4322 of single-vector TBLs.
4323 (aarch64_vectorize_vec_perm_const): Set one_vector_p when only
4326 2018-08-23 Richard Sandiford <richard.sandiford@arm.com>
4329 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const_1): Fix
4330 aarch64_evpc_tbl guard.
4332 2018-08-22 Bernd Edlinger <bernd.edlinger@hotmail.de>
4334 * tree-ssa-dse.c (compute_trims): Avoid folding away undefined
4337 2018-08-22 Martin Sebor <msebor@redhat.com>
4340 * tree-pretty-print.c (pretty_print_string): Add argument.
4341 (dump_generic_node): Call to pretty_print_string with string size.
4343 2018-08-22 Segher Boessenkool <segher@kernel.crashing.org>
4345 PR rtl-optimization/86771
4346 * combine.c (try_combine): Do not allow splitting a resulting PARALLEL
4347 of two SETs into those two SETs, one to be placed at i2, if that SETs
4348 destination is modified between i2 and i3.
4350 2018-08-22 Richard Sandiford <richard.sandiford@arm.com>
4352 PR tree-optimization/86725
4353 * tree-vect-loop.c (vect_inner_phi_in_double_reduction_p): New
4355 (vect_analyze_scalar_cycles_1): Check it.
4357 2018-08-22 Richard Sandiford <richard.sandiford@arm.com>
4359 PR tree-optimization/86725
4360 * tree-vect-loop.c (vect_is_simple_reduction): When treating
4361 an outer loop phi as a double reduction, make sure that the
4362 single user of the phi result is an inner loop phi.
4364 2018-08-22 Richard Sandiford <richard.sandiford@arm.com>
4366 * tree-vect-data-refs.c (vect_analyze_group_access_1): Convert
4367 grouped stores with gaps to a strided group.
4369 2018-08-22 Richard Sandiford <richard.sandiford@arm.com>
4371 * tree-vect-stmts.c (get_group_load_store_type)
4372 (get_load_store_type): Only test STMT_VINFO_STRIDED_P for the
4373 first statement in a group.
4375 2018-08-22 Iain Sandoe <iain@sandoe.co.uk>
4377 * config/darwin.h (LINK_COMMAND_SPEC_A): Sync LTO options with
4378 the sequence used in gcc/gcc.c.
4380 2018-08-22 Iain Sandoe <iain@sandoe.co.uk>
4383 * gcc-ar.c (main): Don’t try to invoke the plug-in if we’re not
4386 2018-08-22 Iain Sandoe <iain@sandoe.co.uk>
4388 * config/darwin10.h (LINK_GCC_C_SEQUENCE_SPEC): Adjust to use the
4389 Darwin10-specific unwinder-shim.
4390 * config/darwin12.h (LINK_GCC_C_SEQUENCE_SPEC): Remove.
4391 * config/rs6000/darwin.h (DARWIN_CRT1_SPEC, DARWIN_DYLIB1_SPEC):
4392 New to cater for Darwin10 Rosetta.
4394 2018-08-22 Iain Sandoe <iain@sandoe.co.uk>
4396 * config/i386/i386.c (ix86_output_addr_diff_elt): Move the MACH-O
4397 specific test before the one for HAVE_AS_GOTOFF_IN_DATA.
4399 2018-08-22 Iain Sandoe <iain@sandoe.co.uk>
4404 * gcc/dwarf2out.c (FUNC_SECOND_SECT_LABEL): New.
4405 (dwarf2out_switch_text_section): Generate a local label for the
4406 second function sub-section and apply it as the second FDE start
4408 * gcc/final.c (final_scan_insn_1): Emit second FDE label after the
4409 second sub-section start.
4411 2018-08-22 Richard Biener <rguenther@suse.de>
4413 PR tree-optimization/86988
4414 * tree-vrp.c (vrp_prop::check_mem_ref): Bail out on VLAs.
4416 2018-08-22 Richard Biener <rguenther@suse.de>
4418 PR tree-optimization/86945
4419 * tree-cfg.c (generate_range_test): Use unsigned arithmetic.
4421 2018-08-22 Alexandre Oliva <oliva@adacore.com>
4423 * config/rs6000/rs6000.c (SMALL_DATA_RELOC, SMALL_DATA_REG): Add
4424 a comment about how uses of r2 for .sdata2 come about.
4426 2018-08-22 Alexandre Oliva <aoliva@redhat.com>
4428 * tree-ssa-reassoc.c (is_reassociable_op): Fix cut&pasto.
4430 2018-08-21 Marek Polacek <polacek@redhat.com>
4432 PR c++/86981, Implement -Wpessimizing-move.
4433 * doc/invoke.texi: Document -Wpessimizing-move.
4435 2018-08-21 Jan Hubicka <jh@suse.cz>
4437 * tree.c (find_decls_types_r): Do not check for redundant typedefs.
4438 * tree.h (is_redundant_typedef): Remove.
4439 * dwarf2out.c (is_redundant_typedef): Turn into static function.
4441 2018-08-21 Jan Hubicka <jh@suse.cz>
4443 * tree.c (free_lang_data_in_decl): Remove types from DECL_CONTEXT
4446 2018-08-21 Tamar Christina <tamar.christina@arm.com>
4448 * expmed.c (extract_low_bits): Reject invalid subregs early.
4450 2018-08-21 Bernd Edlinger <bernd.edlinger@hotmail.de>
4453 * tree-ssa-strlen.c (adjust_last_stmt): Avoid folding away undefined
4456 2018-08-21 Rasmus Villemoes <rv@rasmusvillemoes.dk>
4458 * config/vxworks.h: Guard vxworks_asm_out_constructor and
4459 vxworks_asm_out_destructor by !HAVE_INITFINI_ARRAY_SUPPORT
4460 * config/vxworks.c: Likewise.
4462 2018-08-21 Rasmus Villemoes <rv@rasmusvillemoes.dk>
4464 * config/vxworks.c: Set targetm.have_ctors_dtors
4465 if HAVE_INITFINI_ARRAY_SUPPORT.
4466 * config/vxworks.h: Set SUPPORTS_INIT_PRIORITY
4467 if HAVE_INITFINI_ARRAY_SUPPORT.
4469 2018-08-21 Rasmus Villemoes <rv@rasmusvillemoes.dk>
4471 * config/vxworks.h: Add $(WIND_BASE)/target/h/wrn/coreip to
4472 default search path for VxWorks < 7.
4474 2018-08-21 Bernd Edlinger <bernd.edlinger@hotmail.de>
4476 * gimple-ssa-sprintf.c (decl_constant_value): Remove.
4477 (get_format_string): Refer to c_getstr.
4479 2018-08-21 Tom de Vries <tdevries@suse.de>
4481 * cgraph.h (debuginfo_early_init, debuginfo_init, debuginfo_fini)
4482 (debuginfo_start, debuginfo_stop, debuginfo_early_start)
4483 (debuginfo_early_stop): Declare.
4484 * cgraphunit.c (debuginfo_early_init, debuginfo_init, debuginfo_fini)
4485 (debuginfo_start, debuginfo_stop, debuginfo_early_start)
4486 (debuginfo_early_stop): New function.
4487 (symbol_table::finalize_compilation_unit): Call debuginfo_early_start
4488 and debuginfo_early_stop.
4489 * dwarf2out.c (dwarf2out_finish, dwarf2out_early_finish): Dump dwarf.
4490 * toplev.c (compile_file): Call debuginfo_start and debuginfo_stop.
4491 (general_init): Call debuginfo_early_init.
4492 (finalize): Call debuginfo_fini.
4493 (do_compile): Call debuginfo_init.
4494 * doc/invoke.texi (@gccoptlist): Add -fdump-debug and
4496 (@item -fdump-debug, @item -fdump-earlydebug): Add.
4498 2018-08-21 Tom de Vries <tdevries@suse.de>
4500 * dwarf2out.c (print_dw_val, print_loc_descr, print_die): Handle
4501 flag_dump_noaddr and flag_dump_unnumbered.
4503 2018-08-21 Aldy Hernandez <aldyh@redhat.com>
4505 * wide-int-range.cc (wide_int_range_abs): New.
4506 (wide_int_range_order_set): Rename from wide_int_range_min_max.
4507 * wide-int-range.h (wide_int_range_abs): New.
4508 (wide_int_range_min_max): New.
4509 * tree-vrp.c (extract_range_from_unary_expr): Rewrite ABS_EXPR
4510 case to call wide_int_range_abs.
4511 Rewrite MIN/MAX_EXPR to call wide_int_range_min_max.
4512 (extract_range_from_abs_expr): Delete.
4514 2018-08-20 Michael Meissner <meissner@linux.ibm.com>
4517 * config/rs6000/rs6000.md (extendsi<mode>2): Change constraints
4518 from 'Y' to 'YZ' to enable the LWAX instruction to be generated
4521 2018-08-20 Nathan Sidwell <nathan@acm.org>
4522 Jeff Law <law@redhat.com>
4524 * config/s390/s390-c (s390_macro_to_expand): Use cpp_macro_p.
4525 * config/spu/spu-c.c (spu_macro_to_expand): Likewise.
4527 2018-08-20 David Malcolm <dmalcolm@redhat.com>
4530 * attribs.c (diag_attr_exclusions): Add auto_diagnostic_group instance.
4531 (decl_attributes): Likewise.
4532 * calls.c (maybe_warn_nonstring_arg): Add auto_diagnostic_group
4534 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
4535 * diagnostic-core.h (class auto_diagnostic_group): New class.
4536 * diagnostic.c (diagnostic_initialize): Initialize the new fields.
4537 (diagnostic_report_diagnostic): Handle the first diagnostics within
4539 (emit_diagnostic): Add auto_diagnostic_group instance.
4541 (inform_n): Likewise.
4542 (warning): Likewise.
4543 (warning_at): Likewise.
4544 (warning_n): Likewise.
4545 (pedwarn): Likewise.
4546 (permerror): Likewise.
4548 (error_n): Likewise.
4549 (error_at): Likewise.
4551 (fatal_error): Likewise.
4552 (internal_error): Likewise.
4553 (internal_error_no_backtrace): Likewise.
4554 (auto_diagnostic_group::auto_diagnostic_group): New ctor.
4555 (auto_diagnostic_group::~auto_diagnostic_group): New dtor.
4556 * diagnostic.h (struct diagnostic_context): Add fields
4557 "diagnostic_group_nesting_depth",
4558 "diagnostic_group_emission_count", "begin_group_cb",
4560 * gimple-ssa-isolate-paths.c (find_implicit_erroneous_behavior):
4561 Add auto_diagnostic_group instance(s).
4562 (find_explicit_erroneous_behavior): Likewise.
4563 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Likewise.
4564 * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Likewise.
4565 * gimplify.c (warn_implicit_fallthrough_r): Likewise.
4566 (gimplify_va_arg_expr): Likewise.
4567 * hsa-gen.c (HSA_SORRY_ATV): Likewise.
4568 (HSA_SORRY_AT): Likewise.
4569 * ipa-devirt.c (compare_virtual_tables): Likewise.
4570 (warn_odr): Likewise.
4571 * multiple_target.c (expand_target_clones): Likewise.
4572 * opts-common.c (cmdline_handle_error): Likewise.
4573 * reginfo.c (globalize_reg): Likewise.
4574 * substring-locations.c (format_warning_n_va): Likewise.
4575 * tree-inline.c (expand_call_inline): Likewise.
4576 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Likewise.
4577 * tree-ssa-loop-niter.c
4578 (do_warn_aggressive_loop_optimizations): Likewise.
4579 * tree-ssa-uninit.c (warn_uninit): Likewise.
4580 * tree.c (warn_deprecated_use): Likewise.
4582 2018-08-20 H.J. Lu <hongjiu.lu@intel.com>
4585 * config/i386/i386.md (eh_return): Always update EH return
4586 address in word_mode.
4588 2018-08-20 Chung-Lin Tang <cltang@codesourcery.com>
4590 * targhooks.c (std_gimplify_va_arg_expr): Properly handle case of when
4591 TARGET_SPLIT_COMPLEX_ARG is defined.
4593 2018-08-20 Bernd Edlinger <bernd.edlinger@hotmail.de>
4595 * expr.c (store_field): Change gcc_assert to gcc_checking_assert.
4597 2018-08-20 Bernd Edlinger <bernd.edlinger@hotmail.de>
4600 * expr.c (expand_assignment): Assert that bitpos is positive.
4601 (store_field): Likewise
4602 (expand_expr_real_1): Make sure that bitpos is positive.
4603 * config/alpha/alpha.h (CONSTANT_ADDRESS_P): Avoid signed
4606 2018-08-20 Nathan Sidwell <nathan@acm.org>
4608 * Makefile.in (CPP_ID_DATA_H): Delete.
4609 (CPP_INTERNAL_H): Don't add it.
4610 (GTFILES): Replace CPP_ID_DATA_H with CPPLIB_H.
4611 * gengtype.c (open_base_files): Replace cpp-id-data.h with cpplib.h
4613 2018-08-20 Richard Biener <rguenther@suse.de>
4615 PR tree-optimization/78655
4616 * tree-vrp.c (extract_range_from_binary_expr_1): Make
4617 pointer + offset nonnull if either operand is nonnull work.
4619 2018-08-20 Tom de Vries <tdevries@suse.de>
4621 * dwarf2out.c (add_scalar_info): Don't add reference to existing die
4622 unless the referenced die describes the added property using
4623 DW_AT_location or DW_AT_const_value. Fall back to exprloc case.
4624 Otherwise, add a DW_AT_location to the referenced die.
4626 2018-08-19 Uros Bizjak <ubizjak@gmail.com>
4629 * config/i386/i386.c (ix86_rtx_costs) [case SET]: Check source for
4630 register_operand when calling ix86_set_reg_reg_cost.
4631 [case CONST_INT, case CONST, case LABEL_REF, case SYMBOL_REF]:
4632 Set *total to 0 for operands that satisfy x86_64_immediate_operand
4633 predicate and to 1 otherwise.
4635 2018-08-18 Iain Sandoe <iain@sandoe.co.uk>
4637 * config/darwin.c (darwin_override_options): If -gsplit-dwarf is set,
4638 emit a diagnostic that it is not supported and reset the option.
4639 * config/darwin.h (DRIVER_SELF_SPECS): Note that gsplit-dwarf is not
4640 supported and consume the option. (ASM_FINAL_SPEC): New.
4642 2018-08-17 Segher Boessenkool <segher@kernel.crashing.org>
4644 * doc/md.texi (Patterns): Use @ref instead of @xref in the middle of
4647 2018-08-17 Sandra Loosemore <sandra@codesourcery.com>
4649 C-SKY port: Documentation
4651 * doc/extend.texi (C-SKY Function Attributes): New section.
4652 * doc/invoke.texi (Option Summary): Add C-SKY options.
4653 (C-SKY Options): New section.
4654 * doc/md.texi (Machine Constraints): Document C-SKY constraints.
4656 2018-08-17 Jojo <jijie_rong@c-sky.com>
4657 Huibin Wang <huibin_wang@c-sky.com>
4658 Sandra Loosemore <sandra@codesourcery.com>
4659 Chung-Lin Tang <cltang@codesourcery.com>
4661 C-SKY port: Backend implementation
4663 * config/csky/*: New.
4664 * common/config/csky/*: New.
4666 2018-08-17 Jojo <jijie_rong@c-sky.com>
4667 Huibin Wang <huibin_wang@c-sky.com>
4668 Sandra Loosemore <sandra@codesourcery.com>
4669 Chung-Lin Tang <cltang@codesourcery.com>
4670 Andrew Jenner <andrew@codesourcery.com>
4672 C-SKY port: Configury
4674 * config.gcc (csky-*-*): New.
4675 * configure.ac: Add csky to targets for dwarf2 debug_line support.
4676 * configure: Regenerated.
4678 2018-08-17 David Malcolm <dmalcolm@redhat.com>
4680 * dump-context.h: Include "dumpfile.h".
4681 (dump_context::dump_printf_va): Convert final param from va_list
4682 to va_list *. Convert from ATTRIBUTE_PRINTF to
4683 ATTRIBUTE_GCC_DUMP_PRINTF.
4684 (dump_context::dump_printf_loc_va): Likewise.
4685 * dumpfile.c: Include "stringpool.h".
4686 (make_item_for_dump_printf_va): Delete.
4687 (make_item_for_dump_printf): Delete.
4688 (class dump_pretty_printer): New class.
4689 (dump_pretty_printer::dump_pretty_printer): New ctor.
4690 (dump_pretty_printer::emit_items): New member function.
4691 (dump_pretty_printer::emit_any_pending_textual_chunks): New member
4693 (dump_pretty_printer::emit_item): New member function.
4694 (dump_pretty_printer::stash_item): New member function.
4695 (dump_pretty_printer::format_decoder_cb): New member function.
4696 (dump_pretty_printer::decode_format): New member function.
4697 (dump_context::dump_printf_va): Reimplement in terms of
4698 dump_pretty_printer.
4699 (dump_context::dump_printf_loc_va): Convert final param from va_list
4701 (dump_context::begin_scope): Reimplement call to
4702 make_item_for_dump_printf.
4703 (dump_printf): Update for change to dump_printf_va.
4704 (dump_printf_loc): Likewise.
4705 (selftest::test_capture_of_dump_calls): Convert "stmt" from
4706 greturn * to gimple *. Add a test_decl. Add tests of dump_printf
4707 with %T, %E, and %G.
4708 * dumpfile.h (ATTRIBUTE_GCC_DUMP_PRINTF): New macro.
4709 (dump_printf): Replace ATTRIBUTE_PRINTF_2 with
4710 ATTRIBUTE_GCC_DUMP_PRINTF (2, 3).
4711 (dump_printf_loc): Replace ATTRIBUTE_PRINTF_3 with
4712 ATTRIBUTE_GCC_DUMP_PRINTF (3, 0).
4713 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Convert
4714 use of HOST_WIDE_INT_PRINT_DEC on unsigned HOST_WIDE_INT "count"
4715 within a dump_printf_loc call to "%wu".
4716 (vector_alignment_reachable_p): Merge two dump_printf[_loc] calls,
4717 converting a use of HOST_WIDE_INT_PRINT_DEC to "%wd". Add a
4718 missing space after "=".
4719 * tree-vect-loop.c (vect_analyze_loop_2) Within a dump_printf
4720 call, convert use of HOST_WIDE_INT_PRINT_DEC to "%wd".
4721 * tree-vect-slp.c (vect_slp_bb): Within a dump_printf_loc call,
4722 convert use of HOST_WIDE_INT_PRINT_UNSIGNED to "%wu".
4723 * tree-vectorizer.c (try_vectorize_loop_1): Likewise. Remove
4724 duplicate "vectorized" from message.
4726 2018-08-17 Szabolcs Nagy <szabolcs.nagy@arm.com>
4728 * config/arm/arm-builtins.c (arm_init_simd_builtin_types): Clear
4729 polyNxK_t element's TYPE_STRING_FLAG.
4731 2018-08-17 Segher Boessenkool <segher@kernel.crashing.org>
4733 * config/rs6000/rs6000.md (*cbranch, *creturn): Name these patterns
4734 (they were unnamed before). Fix comments.
4736 2018-08-17 Nathan Sidwell <nathan@acm.org>
4738 * cppbuiltin.c: Include "cpplib.h", not "cpp-id-data.h".
4740 2018-08-17 Richard Biener <rguenther@suse.de>
4742 PR tree-optimization/86841
4743 * wide-int-range.cc (wide_int_range_lshift): Use to_uhwi.
4745 2018-08-17 Martin Liska <mliska@suse.cz>
4747 * common.opt: Remove Warn, Init and Report for options with
4748 Ignore/Deprecated flag. Warning is done automatically for
4750 * config/i386/i386.opt: Likewise.
4751 * config/ia64/ia64.opt: Likewise.
4752 * config/rs6000/rs6000.opt: Likewise.
4753 * cppbuiltin.c (define_builtin_macros_for_compilation_flags):
4754 Remove usage of flag_check_pointer_bounds.
4755 * lto-wrapper.c (merge_and_complain): Do not handle
4756 OPT_fcheck_pointer_bounds.
4757 (append_compiler_options): Likewise.
4758 * opt-functions.awk: Do not handle Deprecated.
4759 * optc-gen.awk: Check that Var, Report and Init are not
4760 used for an option with Ignore/Deprecated flag.
4761 * opts-common.c (decode_cmdline_option): Do not report
4763 (read_cmdline_option): Report warning for OPT_SPECIAL_deprecated
4765 * opts.h (struct cl_option): Remove cl_deprecated flag.
4766 (CL_ERR_DEPRECATED): Remove error enum value.
4768 2018-08-17 Richard Biener <rguenther@suse.de>
4771 * tree-inline.c (copy_bb): When inlining __builtin_va_arg_pack_len ()
4772 across a va-arg-pack using call adjust its return value accordingly.
4774 2018-08-16 Martin Sebor <msebor@redhat.com>
4776 PR tree-optimization/86853
4777 * gimple-ssa-sprintf.c (struct format_result): Rename member.
4778 (struct fmtresult): Add member and initialize it in ctors.
4779 (format_character): Handle %C. Extend range to NUL. Set MAYFAIL.
4780 (format_string): Handle %S the same as %ls. Set MAYFAIL.
4781 (format_directive): Set POSUNDER4K when MAYFAIL is set.
4782 (parse_directive): Handle %C same as %c.
4783 (sprintf_dom_walker::compute_format_length): Adjust.
4784 (is_call_safe): Adjust.
4786 2018-08-16 Bernd Edlinger <bernd.edlinger@hotmail.de>
4788 * builtins.c (c_strlen): Add new parameter eltsize. Use it
4789 for determining how to count the elements.
4790 * builtins.h (c_strlen): Adjust prototype.
4791 * expr.c (string_constant): Add new parameter mem_size.
4792 Set *mem_size appropriately.
4793 * expr.h (string_constant): Adjust protoype.
4794 * gimple-fold.c (get_range_strlen): Add new parameter eltsize.
4795 * gimple-fold.h (get_range_strlen): Adjust prototype.
4796 * gimple-ssa-sprintf.c (get_string_length): Add new parameter eltsize.
4797 (format_string): Call get_string_length with eltsize.
4799 2018-08-16 David Malcolm <dmalcolm@redhat.com>
4801 * diagnostic.c (default_diagnostic_start_span_fn): Call pp_string
4802 to emit the span, rather than setting it as the prefix.
4804 2018-08-16 David Malcolm <dmalcolm@redhat.com>
4806 * diagnostic-show-locus.c (layout::start_annotation_line): Add
4807 "margin_char" parameter, defaulting to space. Use it in place
4808 of pp_space for the initial part of the margin.
4809 (layout::print_leading_fixits): Use '+' when filling the margin
4810 of line-insertion fix-it hints.
4812 2018-08-16 Segher Boessenkool <segher@kernel.crashing.org>
4814 * config/rs6000/rs6000.md (two unnamed define_insn and define_split):
4817 2018-08-16 Segher Boessenkool <segher@kernel.crashing.org>
4819 * config/rs6000/altivec.md: Don't set length attribute to the default
4821 * config/rs6000/darwin.md: Ditto.
4822 * config/rs6000/dfp.md: Ditto.
4823 * config/rs6000/htm.md: Ditto.
4824 * config/rs6000/rs6000.md: Ditto.
4825 * config/rs6000/sync.md: Ditto.
4826 * config/rs6000/vsx.md: Ditto.
4828 2018-08-16 Segher Boessenkool <segher@kernel.crashing.org>
4830 * config/rs6000/altivec.md: Don't set length attribute to the default
4831 value, for branch instructions.
4832 * config/rs6000/darwin.md: Ditto.
4833 * config/rs6000/rs6000.md: Ditto.
4835 2018-08-16 Segher Boessenkool <segher@kernel.crashing.org>
4837 * config/rs6000/rs6000.md (length): Always define as const_int 4.
4838 (unnamed conditional branch define_insn): Set length to 4 or 8
4839 depending on offset.
4840 (<bd>_<mode>): Similar, for alternative 0.
4841 (<bd>tf_<mode>): Ditto.
4843 2018-08-16 Tamar Christina <tamar.christina@arm.com>
4845 * expr.c (copy_blkmode_to_reg): Perform larger copies when safe.
4847 2018-08-16 Matthew Malcomson <matthew.malcomson@arm.com>
4849 * doc/rtl.texi: Replace old RTX class names with new names.
4852 2018-08-16 Vlad Lazar <vlad.lazar@arm.com>
4854 * expmed.h (canonicalize_comparison): New declaration.
4855 * expmed.c (canonicalize_comparison, equivalent_cmp_code): New function.
4856 * expmed.c (emit_store_flag_1): Add call to canonicalize_comparison.
4857 * optabs.c (prepare_cmp_insn): Likewise.
4858 * rtl.h (unsigned_condition_p): New function which checks if a
4859 comparison operator is unsigned.
4861 2018-08-16 Nathan Sidwell <nathan@acm.org>
4863 * config/rs6000/rs6000-c.c (rs6000_macro_to_expend): Use cpp_macro_p.
4864 * config/powerpcspc/powerpcspe-c.c (rs6000_macro_to_expend): Likewise.
4866 2018-08-16 Tamar Christina <tamar.christina@arm.com>
4869 * config/arm/arm.c (arm_can_change_mode_class): Disallow subreg.
4870 * config/arm/neon.md (movv4hf, movv8hf): Refactored to..
4871 (mov<mov>): ..this and enable unconditionally.
4873 2018-08-16 Tamar Christina <tamar.christina@arm.com>
4875 * config/arm/neon.md (*neon_mov<mode>): Remove reg-to-reg alternative.
4877 2018-08-16 Sam Tebbs <sam.tebbs@arm.com>
4879 * config/aarch64/aarch64.opt (mlow-precision-recip-sqrt)
4880 (mlow-precision-sqrt, mlow-precision-div, mverbose-cost-dump): Replace
4881 "Common" with "Target".
4883 2018-08-15 Uros Bizjak <ubizjak@gmail.com>
4885 * config/i386/i386.opt (mmitigate-rop): Mark as deprecated.
4886 * doc/invoke.texi (mmitigate-rop): Remove.
4887 * config/i386/i386.c: Do not include "regrename.h".
4888 (ix86_rop_should_change_byte_p, reg_encoded_number)
4889 (ix86_get_modrm_for_rop, set_rop_modrm_reg_bits, ix86_mitigate_rop):
4891 (ix86_reorg): Remove call to ix86_mitigate_rop.
4892 * config/i386/i386.md (attr "modrm_class"): Remove.
4893 (cmp<mode>_ccno_1, mov<mode>_xor, movstrict<mode>_xor)
4894 (x86_mov<mode>cc_0_m1. x86_mov<mode>cc_0_m1_se)
4895 (x86_mov<mode>cc_0_m1_neg): Remove modrm_class attribute override.
4897 2018-08-15 Will Schmidt <will_schmidt@vnet.ibm.com>
4899 * config/rs6000/rs600.c (rs6000_gimple_fold_builtin): Add entries to
4900 allow folding of mergeh() and mergel() for the float and double types.
4901 (fold_mergehl_helper): Rework to handle building a permute tree
4904 2018-08-15 Uros Bizjak <ubizjak@gmail.com>
4906 * config/i386/i386.c (expand_vec_perm_movs): Enable V4SFmode
4909 2018-08-15 David Malcolm <dmalcolm@redhat.com>
4911 * common.opt (fdiagnostics-show-labels): New option.
4912 * diagnostic-show-locus.c (class layout_range): Add field
4914 (class layout): Add field "m_show_labels_p".
4915 (layout_range::layout_range): Add param "label" and use it to
4917 (make_range): Pass in NULL for new "label" param of layout_range's
4919 (layout::layout): Initialize m_show_labels_p.
4920 (layout::maybe_add_location_range): Pass in loc_range->m_label
4921 when constructing layout_range instances.
4922 (struct line_label): New struct.
4923 (layout::print_any_labels): New member function.
4924 (layout::print_line): Call it if label-printing is enabled.
4925 (selftest::test_one_liner_labels): New test.
4926 (selftest::test_diagnostic_show_locus_one_liner): Call it.
4927 * diagnostic.c (diagnostic_initialize): Initialize
4928 context->show_labels_p.
4929 * diagnostic.h (struct diagnostic_context): Add field
4931 * doc/invoke.texi (Diagnostic Message Formatting Options): Add
4932 -fno-diagnostics-show-labels.
4933 * dwarf2out.c (gen_producer_string): Add
4934 OPT_fdiagnostics_show_labels to the ignored options.
4935 * gcc-rich-location.c (gcc_rich_location::add_expr): Add "label"
4937 (gcc_rich_location::maybe_add_expr): Likewise.
4938 * gcc-rich-location.h (gcc_rich_location::gcc_rich_location): Add
4939 label" param, defaulting to NULL.
4940 (gcc_rich_location::add_expr): Add "label" param.
4941 (gcc_rich_location::maybe_add_expr): Likewise.
4942 (class text_range_label): New class.
4943 (class range_label_for_type_mismatch): New class.
4944 * gimple-ssa-sprintf.c (fmtwarn): Pass NULL for new label params
4945 of format_warning_va.
4946 (fmtwarn_n): Likewise for new params of format_warning_n_va.
4947 * lto-wrapper.c (merge_and_complain): Add
4948 OPT_fdiagnostics_show_labels to the "pick one setting" options.
4949 (append_compiler_options): Likewise to the dropped options.
4950 (append_diag_options): Likewise to the passed-on options.
4951 * opts.c (common_handle_option): Handle the new option.
4952 * selftest-diagnostic.c
4953 (test_diagnostic_context::test_diagnostic_context): Enable
4955 * substring-locations.c: Include "gcc-rich-location.h".
4956 (format_warning_n_va): Add "fmt_label" and "param_label" params
4957 and use them as appropriate.
4958 (format_warning_va): Add "fmt_label" and "param_label" params,
4959 passing them on to format_warning_n_va.
4960 (format_warning_at_substring): Likewise.
4961 (format_warning_at_substring_n): Likewise.
4962 * substring-locations.h (format_warning_va): Add "fmt_label" and
4963 "param_label" params.
4964 (format_warning_n_va): Likewise.
4965 (format_warning_at_substring): Likewise.
4966 (format_warning_at_substring_n): Likewise.
4967 * toplev.c (general_init): Initialize global_dc->show_labels_p.
4969 2018-08-15 Qing Zhao <qing.zhao@oracle.com>
4972 * builtins.c (expand_builtin_memcmp): Do not expand the call
4973 when overflow is detected.
4975 2018-08-15 Martin Sebor <msebor@redhat.com>
4977 PR tree-optimization/71625
4978 * config/aarch64/aarch64-builtins.c
4979 (aarch64_init_simd_builtin_types): Clear Poly8_t's TYPE_STRING_FLAG.
4981 2018-08-15 Ilya Leoshkevich <iii@linux.ibm.com>
4983 * config/s390/s390.c (s390_reorg): Remove loop.
4985 2018-08-15 Iain Sandoe <iain@sandoe.co.uk>
4988 (darwin_function_switched_text_sections): Delete.
4989 * gcc/config/darwin.h
4990 (TARGET_ASM_FUNCTION_SWITCHED_TEXT_SECTIONS): Likewise.
4992 2018-08-15 Iain Sandoe <iain@sandoe.co.uk>
4995 * config/darwin.h: (DEBUG_STR_OFFSETS_SECTION, DEBUG_LOCLISTS_SECTION,
4996 DEBUG_RNGLISTS_SECTION) new macros. (DEBUG_PUBNAMES_SECTION,
4997 DEBUG_PUBTYPES_SECTION) update to include GNU variant.
4999 2018-08-15 Martin Liska <mliska@suse.cz>
5001 PR tree-optimization/86925
5002 * predict.c (expr_expected_value_1): When taking
5003 later predictor, assign also probability.
5004 Use fold_build2_initializer_loc in order to fold
5005 the expression in -frounding-math.
5007 2018-08-14 Allan Sandfeld Jensen <allan.jensen@qt.io>
5009 * config/i386/i386.c (expand_vec_perm_movs): New method matching movs
5011 (expand_vec_perm_1): Try the new method.
5013 2018-08-14 Ilya Leoshkevich <iii@linux.ibm.com>
5016 * lra-lives.c (remove_some_program_points_and_update_live_ranges):
5017 Check whether lra_live_max_point is 0 before dividing.
5019 2018-08-14 Martin Sebor <msebor@redhat.com>
5021 PR tree-optimization/86650
5022 * tree-vrp.c (vrp_prop::check_array_ref): Print an inform message.
5023 (vrp_prop::check_mem_ref): Same.
5025 2018-08-13 Liu Hao <lh_mouse@126.com>
5027 * pretty-print.c (eat_esc_sequence): Swap the foreground and
5028 background colors if the COMMON_LVB_REVERSE_VIDEO flag is set,
5029 and clear it thereafter, as it only works for DBCS.
5031 2018-08-13 Liu Hao <lh_mouse@126.com>
5033 * pretty-print.c (mingw_ansi_fputs): Do not call _close() on the
5034 handle returned by _get_osf_handle().
5036 2018-08-13 Will Schmidt <will_schmidt@vnet.ibm.com>
5038 * gcc/config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support
5039 for folding vec_perm.
5041 2018-08-13 Will Schmidt <will_schmidt@vnet.ibm.com>
5043 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin):
5044 Add support for gimple-folding of vec_pack() and vec_unpack()
5047 2018-08-13 Will Schmidt <will_schmidt@vnet.ibm.com>
5049 * config/rs6000/rs6000.c (rs6000_builtin_valid_without_lhs): Add
5050 vec_xst variants to the list.
5051 (rs6000_gimple_fold_builtin): Add support for folding unaligned
5052 vector loads and stores.
5054 2018-08-13 David Edelsohn <dje.gcc@gmail.com>
5056 * config.gcc (rs6000-ibm-aix4.x): Delete.
5057 (rs6000-ibm-aix5.1): Delete.
5058 (rs6000-ibm-aix5.2): Delete.
5059 (rs6000-ibm-aix5.3): Delete.
5060 * config/rs6000/aix43.h: Delete.
5061 * config/rs6000/aix51.h: Delete.
5062 * config/rs6000/aix52.h: Delete.
5063 * config/rs6000/t-aix43: Delete.
5065 2018-08-13 Ilya Leoshkevich <iii@linux.ibm.com>
5067 * config/s390/s390.c (s390_decompose_constant_pool_ref):
5069 (s390_decompose_address): Factor out constant pool ref
5072 2018-08-12 Chung-Ju Wu <jasonwucj@gmail.com>
5074 * config/nds32/nds32-predicates.c
5075 (nds32_can_use_bclr_p): Change return type as bool.
5076 (nds32_can_use_bset_p): Ditto.
5077 (nds32_can_use_btgl_p): Ditto.
5078 (nds32_can_use_bitci_p): Ditto.
5079 * config/nds32/nds32-protos.h
5080 (nds32_can_use_bclr_p): Change declaration.
5081 (nds32_can_use_bset_p): Ditto.
5082 (nds32_can_use_btgl_p): Ditto.
5083 (nds32_can_use_bitci_p): Ditto.
5085 2018-08-12 Chung-Ju Wu <jasonwucj@gmail.com>
5087 * config/nds32/nds32.c (nds32_expand_prologue, nds32_expand_epilogue):
5088 Support -msched-prolog-epilog option.
5089 * config/nds32/nds32.opt (msched-prolog-epilog): New option.
5091 2018-08-12 Chung-Ju Wu <jasonwucj@gmail.com>
5093 * common/config/nds32/nds32-common.c
5094 (nds32_option_optimization_table): Enalbe -malways-align.
5096 2018-08-12 Chung-Ju Wu <jasonwucj@gmail.com>
5098 * config.gcc (nds32*): Add nds32_isr.h and nds32_init.inc in
5100 * common/config/nds32/nds32-common.c (nds32_handle_option): Handle
5101 OPT_misr_secure_ case.
5102 * config/nds32/nds32-isr.c: Implementation of backward compatibility.
5103 * config/nds32/nds32-protos.h (nds32_isr_function_critical_p): New.
5104 * config/nds32/nds32.c (nds32_attribute_table): Add critical and
5106 * config/nds32/nds32.h (nds32_isr_nested_type): Add NDS32_CRITICAL.
5107 (nds32_isr_info): New field security_level.
5108 (TARGET_ISR_VECTOR_SIZE_4_BYTE): New macro.
5109 * config/nds32/nds32.md (return_internal): Consider critical attribute.
5110 * config/nds32/nds32.opt (misr-secure): New option.
5111 * config/nds32/nds32_init.inc: New file.
5112 * config/nds32/nds32_isr.h: New file.
5114 2018-08-11 John David Anglin <danglin@gcc.gnu.org>
5116 * config/pa/pa.md (UNSPEC_MEMORY_BARRIER): New unspec enum.
5117 Update comment for atomic instructions.
5118 (atomic_storeqi, atomic_storehi, atomic_storesi, atomic_storesf,
5119 atomic_loaddf, atomic_loaddf_1, atomic_storedf, atomic_storedf_1):
5121 (atomic_loaddi): Revise fence expansion to only emit fence prior to
5122 load for __ATOMIC_SEQ_CST model.
5123 (atomic_loaddi_1): Remove float register target.
5124 (atomic_storedi): Handle CONST_INT values.
5125 (atomic_storedi_1): Remove float register source. Add special case
5127 (memory_barrier): New expander and insn.
5129 2018-08-11 Jakub Jelinek <jakub@redhat.com>
5131 PR tree-optimization/86835
5132 * tree-ssa-math-opts.c (insert_reciprocals): Even when inserting
5133 new_stmt after def_gsi, make sure to insert new_square_stmt after
5134 that stmt, not 2 stmts before it.
5136 2018-08-10 Alexander Monakov <amonakov@ispras.ru>
5139 * config/i386/i386.md (<s>mul<mode>3_highpart): Use DWIH mode iterator
5142 2018-08-10 Martin Liska <mliska@suse.cz>
5145 * builtin-types.def (BT_FN_LONG_LONG_LONG_DOUBLE): Add new
5147 * builtins.c (expand_builtin_expect_with_probability):
5149 (expand_builtin_expect_with_probability): New function.
5150 (build_builtin_expect_predicate): Add new argumnet probability
5151 for BUILT_IN_EXPECT_WITH_PROBABILITY.
5152 (fold_builtin_expect):
5155 * builtins.def (BUILT_IN_EXPECT_WITH_PROBABILITY):
5156 * builtins.h (fold_builtin_expect): Set new argument.
5157 * doc/extend.texi: Document __builtin_expect_with_probability.
5158 * doc/invoke.texi: Likewise.
5159 * gimple-fold.c (gimple_fold_call): Pass new argument.
5160 * ipa-fnsummary.c (find_foldable_builtin_expect): Handle
5161 also BUILT_IN_EXPECT_WITH_PROBABILITY.
5162 * predict.c (get_predictor_value): New function.
5163 (expr_expected_value): Add new argument probability. Assume
5164 that predictor and probability are always non-null.
5165 (expr_expected_value_1): Likewise. For __builtin_expect and
5166 __builtin_expect_with_probability set probability. Handle
5167 combination in binary expressions.
5168 (tree_predict_by_opcode): Simplify code by simply calling
5169 get_predictor_value.
5170 (pass_strip_predict_hints::execute): Add handling of
5171 BUILT_IN_EXPECT_WITH_PROBABILITY.
5172 * predict.def (PRED_BUILTIN_EXPECT_WITH_PROBABILITY): Add
5174 * tree.h (DECL_BUILT_IN_P): New function.
5176 2018-08-10 Martin Liska <mliska@suse.cz>
5178 PR tree-optimization/85799
5179 * passes.def: Add argument for pass_strip_predict_hints.
5180 * predict.c (class pass_strip_predict_hints): Add new argument
5182 (strip_predictor_early): New function.
5183 (pass_strip_predict_hints::execute): Call the function to
5185 (strip_predict_hints): New function.
5186 * predict.def: Fix comment.
5188 2018-08-10 Thomas Preud'homme <thomas.preudhomme@linaro.org>
5190 * Makefile.in: Clarify which tm.texi to copy over to assert the
5191 right to grant a GFDL license for all.
5193 2018-08-09 Jeff Law <law@redhat.com>
5195 * config/m68k/m68k.c (m68k_adjust_decorated_operand): Remove
5198 2018-08-09 Andreas Schwab <schwab@linux-m68k.org>
5200 * config/m68k/m68k-protos.h (m68k_final_prescan_insn): Remove
5203 2018-08-09 Richard Sandiford <richard.sandiford@arm.com>
5205 * tree-vect-loop.c (vectorizable_reduction): Allow inner-loop
5206 reductions for variable-length vectors.
5208 2018-08-09 David Malcolm <dmalcolm@redhat.com>
5211 * common.opt (fdiagnostics-show-line-numbers): New option.
5212 * diagnostic-show-locus.c (class layout): Add fields
5213 "m_show_line_numbers_p" and "m_linenum_width";
5214 (num_digits): New function.
5215 (test_num_digits): New function.
5216 (layout::layout): Initialize new fields. Update m_x_offset
5217 logic to handle any left margin.
5218 (layout::print_source_line): Print line number when requested.
5219 (layout::start_annotation_line): New member function.
5220 (layout::print_annotation_line): Call it.
5221 (layout::print_leading_fixits): Likewise.
5222 (layout::print_trailing_fixits): Likewise. Update calls to
5223 move_to_column for new parameter.
5224 (layout::get_x_bound_for_row): Add "add_left_margin" param and use
5225 it to potentially call start_annotation_line.
5226 (layout::show_ruler): Call start_annotation_line.
5227 (selftest::test_line_numbers_multiline_range): New selftest.
5228 (selftest::diagnostic_show_locus_c_tests): Call test_num_digits
5229 and selftest::test_line_numbers_multiline_range.
5230 * diagnostic.c (diagnostic_initialize): Initialize
5231 show_line_numbers_p.
5232 * diagnostic.h (struct diagnostic_context): Add field
5233 "show_line_numbers_p".
5234 * doc/invoke.texi (Diagnostic Message Formatting Options): Add
5235 -fno-diagnostics-show-line-numbers.
5236 * dwarf2out.c (gen_producer_string): Add
5237 OPT_fdiagnostics_show_line_numbers to the ignored options.
5238 * lto-wrapper.c (merge_and_complain): Likewise to the "pick
5239 one setting" options.
5240 (append_compiler_options): Likewise to the dropped options.
5241 (append_diag_options): Likewise to the passed-on options.
5242 * opts.c (common_handle_option): Handle the new option.
5243 * toplev.c (general_init): Set up global_dc->show_line_numbers_p.
5245 2018-08-09 Kelvin Nilsen <kelvin@gcc.gnu.org>
5247 * doc/extend.texi (PowerPC AltiVec Built-in Functions Available on
5248 ISA 2.07): Correct spelling of bcdsub to be __builtin_bcdsub. Add
5249 third argument of type "const signed char" to descriptions of
5250 __builtin_bcdadd, __builtin_bcdadd_lt, __builtin_bcdadd_eq,
5251 __builtin_bcdadd_gt, __builtin_bcdadd_ov, __builtin_bcdsub,
5252 __builtin_bcdsub_lt, __builtin_bcdsub_eq, __builtin_bcdsub_gt,
5253 __builtin_bcdsub_ov functions.
5255 2018-08-09 Richard Sandiford <richard.sandiford@arm.com>
5257 PR tree-optimization/86858
5258 * tree-vect-loop.c (vect_is_simple_reduction): Restore
5259 flow_bb_inside_loop_p calls.
5261 2018-08-09 Richard Sandiford <richard.sandiford@arm.com>
5263 PR tree-optimization/86871
5264 * tree-vect-stmts.c (vect_transform_stmt): Use gimple_get_lhs
5265 instead of gimple_assign_lhs.
5267 2018-08-09 Richard Earnshaw <rearnsha@arm.com>
5270 * config/aarch64/aarch64.md (add<mode>3_carryinC_zero): Add missing
5271 register constraint to operand 0.
5272 (add<mode>3_carryinC): Likewise.
5273 (add<mode>3_carryinV_zero, add<mode>3_carryinV): Likewise.
5275 2018-08-09 Martin Liska <mliska@suse.cz>
5278 * common.opt: Remove extra line.
5280 2018-08-09 Martin Liska <mliska@suse.cz>
5282 * params.def (PARAM_ALIGN_LOOP_ITERATIONS): Remove double dots
5283 at the end of a line, make first letter capital and end up
5284 a sentence with a dot.
5285 (PARAM_LOOP_INTERCHANGE_STRIDE_RATIO): Likewise.
5286 (PARAM_LOOP_BLOCK_TILE_SIZE): Likewise.
5287 (PARAM_GRAPHITE_MAX_NB_SCOP_PARAMS): Likewise.
5288 (PARAM_GRAPHITE_MAX_ARRAYS_PER_SCOP): Likewise.
5289 (PARAM_MAX_ISL_OPERATIONS): Likewise.
5290 (PARAM_GRAPHITE_ALLOW_CODEGEN_ERRORS): Likewise.
5291 (PARAM_PROFILE_FUNC_INTERNAL_ID): Likewise.
5292 (PARAM_INDIR_CALL_TOPN_PROFILE): Likewise.
5293 (PARAM_SLP_MAX_INSNS_IN_BB): Likewise.
5294 (PARAM_IPA_CP_EVAL_THRESHOLD): Likewise.
5295 (PARAM_IPA_CP_RECURSION_PENALTY): Likewise.
5296 (PARAM_IPA_CP_SINGLE_CALL_PENALTY): Likewise.
5297 (PARAM_IPA_CP_LOOP_HINT_BONUS): Likewise.
5298 (PARAM_IPA_CP_ARRAY_INDEX_HINT_BONUS): Likewise.
5299 (PARAM_TREE_REASSOC_WIDTH): Likewise.
5300 (PARAM_HSA_GEN_DEBUG_STORES): Likewise.
5301 (PARAM_MAX_SPECULATIVE_DEVIRT_MAYDEFS): Likewise.
5302 (PARAM_MAX_VRP_SWITCH_ASSERTIONS): Likewise.
5304 2018-08-09 Andreas Krebbel <krebbel@linux.ibm.com>
5307 * config/s390/s390.c (s390_option_override_internal): Reduce the
5308 stack-clash-protection-probe-interval param if it would be too big
5311 2018-08-08 Andreas Schwab <schwab@linux-m68k.org>
5314 * config/m68k/m68k.h (FINAL_PRESCAN_INSN): Don't define.
5315 * config/m68k/m68k.c (handle_move_double): Don't call
5316 m68k_final_prescan_insn.
5317 (m68k_adjust_decorated_operand): Renamed from
5318 m68k_final_prescan_insn, remove first and third operand and
5320 (print_operand): Call it.
5321 (print_operand_address): Call it.
5323 2018-08-08 Nathan Sidwell <nathan@acm.org>
5325 * diagnostic.c (diagnostic_report_current_module): Use
5326 linemap_included_from & linemap_included_from_linemap.
5328 2018-08-08 Hongbo Zhang <hongbo.zhang@linaro.org>
5330 * config/aarch64/aarch64-cores.def: Add phecda core.
5331 * config/aarch64/aarch64-tune.md: Regenerate.
5332 * doc/invoke.texi: Add phecda core.
5334 2018-08-08 Andreas Krebbel <krebbel@linux.ibm.com>
5337 * config/s390/constraints.md ("NxHD0", "NxSD0"): New constraint
5339 * config/s390/s390.md ("movti"): Add more alternatives for
5340 constant to GPR copies.
5342 2018-08-08 Andreas Krebbel <krebbel@linux.ibm.com>
5344 * config/s390/s390.c: Fix whitespace damage throughout the file.
5345 * config/s390/s390.h: Likewise.
5346 * config/s390/tpf.h: Likewise.
5348 2018-08-08 Ilya Leoshkevich <iii@linux.ibm.com>
5350 * config/s390/s390.c (s390_loadrelative_operand_p):
5351 Remove TARGET_CPU_ZARCH usages.
5352 (s390_rtx_costs): Likewise.
5353 (s390_legitimate_constant_p): Likewise.
5354 (s390_cannot_force_const_mem): Likewise.
5355 (legitimate_reload_constant_p): Likewise.
5356 (s390_preferred_reload_class): Likewise.
5357 (legitimize_pic_address): Likewise.
5358 (legitimize_tls_address): Likewise.
5359 (s390_split_branches): Removed.
5360 (s390_add_execute): Removed.
5361 (s390_dump_pool): Remove TARGET_CPU_ZARCH usages.
5362 (s390_mainpool_start): Likewise.
5363 (s390_mainpool_finish): Likewise.
5364 (s390_mainpool_cancel): Removed.
5365 (s390_chunkify_start): Remove TARGET_CPU_ZARCH usages.
5366 (s390_chunkify_cancel): Likewise.
5367 (s390_return_addr_rtx): Likewise.
5368 (s390_register_info): Remove split_branches_pending_p uages.
5369 (s390_optimize_register_info): Likewise.
5370 (s390_init_frame_layout): Remove TARGET_CPU_ZARCH and
5371 split_branches_pending_p usages.
5372 (s390_can_eliminate): Remove TARGET_CPU_ZARCH usages.
5373 (s390_load_got): Likewise.
5374 (s390_expand_split_stack_prologue): Likewise.
5375 (output_asm_nops): Likewise.
5376 (s390_function_profiler): Likewise.
5377 (s390_emit_call): Likewise.
5378 (s390_conditional_register_usage): Likewise.
5379 (s390_optimize_prologue): Likewise.
5380 (s390_reorg): Remove TARGET_CPU_ZARCH and
5381 split_branches_pending_p usages.
5382 (s390_option_override_internal): Remove TARGET_CPU_ZARCH
5384 (s390_output_indirect_thunk_function): Likewise.
5385 * config/s390/s390.h (TARGET_CPU_ZARCH): Removed.
5386 (TARGET_CPU_ZARCH_P): Removed.
5387 (struct machine_function): Remove split_branches_pending_p.
5388 * config/s390/s390.md: Remove TARGET_CPU_ZARCH usages.
5390 2018-08-08 Ilya Leoshkevich <iii@linux.ibm.com>
5392 * common/config/s390/s390-common.c (processor_flags_table):
5394 * config.gcc: Remove with_arch/with_tune support.
5395 * config/s390/2064.md: Remove cpu attribute comparisons.
5396 * config/s390/driver-native.c (s390_host_detect_local_cpu):
5398 * config/s390/linux.h (ASM_SPEC):
5399 Remove -march support.
5400 * config/s390/s390-c.c (s390_cpu_cpp_builtins_internal):
5401 Use a table to get an arch level.
5402 * config/s390/s390-opts.h (enum processor_type):
5404 * config/s390/s390.c
5405 (processor_table): Remove entries, add arch_level values.
5406 (s390_issue_rate): Remove cases.
5407 (s390_option_override): Adjust
5408 s390_option_override_internal() call.
5409 (s390_option_override_internal): Remove deprecation warning.
5410 (s390_valid_target_attribute_tree): Adjust
5411 s390_option_override_internal() call.
5412 * config/s390/s390.h (struct s390_processor):
5413 Share with s390-c.c, add arch_level field.
5414 * config/s390/s390.md:
5415 Remove occurrences in cpu attribute.
5416 * config/s390/s390.opt: Remove -march/-mtune support.
5417 * config/s390/tpf.h (ASM_SPEC): Remove -march support.
5418 * doc/invoke.texi: Remove deprecation warning.
5420 2018-08-08 Luis Machado <luis.machado@linaro.org>
5422 * config/aarch64/aarch64.c (qdf24xx_vector_cost): New static
5424 (qdf24xx_tunings): Set vector cost structure to
5425 qdf24xx_vector_cost.
5427 * config/aarch64/aarch64.c (qdf24xx_addrcost_table)
5428 <register_sextend>: Set to 3.
5430 2018-08-07 Richard Sandiford <richard.sandiford@arm.com>
5433 * config/aarch64/iterators.md (FRECP, frecp_suffix): Delete.
5434 * config/aarch64/aarch64-simd.md
5435 (aarch64_frecp<FRECP:frecp_suffix><mode>): Fold FRECPE into...
5436 (@aarch64_frecpe<mode>): ...here and the move FRECPX to...
5437 (aarch64_frecpx<mode>): ...this new pattern.
5438 * config/aarch64/aarch64-simd-builtins.def: Remove comment
5439 about aarch64_frecp<FRECP:frecp_suffix><mode>.
5441 2018-08-07 Martin Liska <mliska@suse.cz>
5444 * predict.c (expr_expected_value_1): Handle DECL_IS_MALLOC,
5445 BUILT_IN_REALLOC and DECL_IS_OPERATOR_NEW.
5446 * predict.def (PRED_MALLOC_NONNULL): New predictor.
5447 * doc/extend.texi: Document that malloc attribute adds
5450 2018-08-06 John David Anglin <danglin@gcc.gnu.org>
5453 * config/pa/pa.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
5454 Define to speculation_safe_value_not_needed.
5456 2018-08-06 Jeff Law <law@redhat.com>
5458 * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Pass down
5459 the vr_values instance to cprop_into_stmt.
5460 (cprop_into_stmt): Pass vr_values instance down to cprop_operand.
5461 (cprop_operand): Also query EVRP to determine if OP is a constant.
5463 2018-08-06 Nathan Sidwell <nathan@acm.org>
5465 * diagnostic.c (diagnostic_report_current_module): Reroll
5466 included-at loop. Translate text.
5468 2018-08-06 David Malcolm <dmalcolm@redhat.com>
5470 * function-tests.c (selftest::test_expansion_to_rtl): Call
5471 free_after_compilation.
5473 2018-08-06 Alan Hayward <alan.hayward@arm.com>
5475 * config/aarch64/aarch64.md: Add clobber highs to tls_desc.
5477 2018-08-06 Andreas Krebbel <krebbel@linux.ibm.com>
5479 * config/s390/s390.c (s390_loop_unroll_adjust): Prevent small
5480 loops with memory block operations from getting unrolled.
5482 2018-08-06 Ulrich Weigand <uweigand@de.ibm.com>
5485 * config/spu/spu.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
5486 Define to speculation_safe_value_not_needed.
5488 2018-08-06 Jeff Law <law@redhat.com>
5490 * reload1.c (forget_old_reloads_1): Adjust CLOBBER_HIGH
5493 2018-08-06 Jozef Lawrynowicz <jozef.l@mittosystems.com>
5496 * gcc/tree.c (build_common_tree_nodes): Initialize integer_types array
5497 with all enabled __intN types.
5499 * gcc/testsuite/gcc.target/msp430/pr86662.c: New test.
5501 2018-08-06 Alan Hayward <alan.hayward@arm.com>
5503 * alias.c (record_set): Check for clobber high.
5504 * cfgexpand.c (expand_gimple_stmt): Likewise.
5505 * combine-stack-adj.c (single_set_for_csa): Likewise.
5506 * combine.c (find_single_use_1): Likewise.
5507 (set_nonzero_bits_and_sign_copies): Likewise.
5508 (get_combine_src_dest): Likewise.
5509 (is_parallel_of_n_reg_sets): Likewise.
5510 (try_combine): Likewise.
5511 (record_dead_and_set_regs_1): Likewise.
5512 (reg_dead_at_p_1): Likewise.
5513 (reg_dead_at_p): Likewise.
5514 * dce.c (deletable_insn_p): Likewise.
5515 (mark_nonreg_stores_1): Likewise.
5516 (mark_nonreg_stores_2): Likewise.
5517 * df-scan.c (df_find_hard_reg_defs): Likewise.
5518 (df_uses_record): Likewise.
5519 (df_get_call_refs): Likewise.
5520 * dwarf2out.c (mem_loc_descriptor): Likewise.
5521 * haifa-sched.c (haifa_classify_rtx): Likewise.
5522 * ira-build.c (create_insn_allocnos): Likewise.
5523 * ira-costs.c (scan_one_insn): Likewise.
5524 * ira.c (equiv_init_movable_p): Likewise.
5525 (rtx_moveable_p): Likewise.
5526 (interesting_dest_for_shprep): Likewise.
5527 * jump.c (mark_jump_label_1): Likewise.
5528 * postreload-gcse.c (record_opr_changes): Likewise.
5529 * postreload.c (reload_cse_simplify): Likewise.
5530 (struct reg_use): Add source expr.
5531 (reload_combine): Check for clobber high.
5532 (reload_combine_note_use): Likewise.
5533 (reload_cse_move2add): Likewise.
5534 (move2add_note_store): Likewise.
5535 * print-rtl.c (print_pattern): Likewise.
5536 * recog.c (decode_asm_operands): Likewise.
5537 (store_data_bypass_p): Likewise.
5538 (if_test_bypass_p): Likewise.
5539 * regcprop.c (kill_clobbered_value): Likewise.
5540 (kill_set_value): Likewise.
5541 * reginfo.c (reg_scan_mark_refs): Likewise.
5542 * reload1.c (maybe_fix_stack_asms): Likewise.
5543 (eliminate_regs_1): Likewise.
5544 (elimination_effects): Likewise.
5545 (mark_not_eliminable): Likewise.
5546 (scan_paradoxical_subregs): Likewise.
5547 (forget_old_reloads_1): Likewise.
5548 * reorg.c (find_end_label): Likewise.
5549 (try_merge_delay_insns): Likewise.
5550 (redundant_insn): Likewise.
5551 (own_thread_p): Likewise.
5552 (fill_simple_delay_slots): Likewise.
5553 (fill_slots_from_thread): Likewise.
5554 (dbr_schedule): Likewise.
5555 * resource.c (update_live_status): Likewise.
5556 (mark_referenced_resources): Likewise.
5557 (mark_set_resources): Likewise.
5558 * rtl.c (copy_rtx): Likewise.
5559 * rtlanal.c (reg_referenced_p): Likewise.
5560 (single_set_2): Likewise.
5561 (noop_move_p): Likewise.
5562 (note_stores): Likewise.
5563 * sched-deps.c (sched_analyze_reg): Likewise.
5564 (sched_analyze_insn): Likewise.
5566 2018-08-06 Alan Hayward <alan.hayward@arm.com>
5568 * cse.c (invalidate_reg): New function extracted from...
5569 (invalidate): ...here.
5570 (canonicalize_insn): Check for clobber high.
5571 (invalidate_from_clobbers): invalidate clobber highs.
5572 (invalidate_from_sets_and_clobbers): Likewise.
5573 (count_reg_usage): Check for clobber high.
5574 (insn_live_p): Likewise.
5575 * cselib.c (cselib_expand_value_rtx_1):Likewise.
5576 (cselib_invalidate_regno): Check for clobber in setter.
5577 (cselib_invalidate_rtx): Pass through setter.
5578 (cselib_invalidate_rtx_note_stores):
5579 (cselib_process_insn): Check for clobber high.
5580 * cselib.h (cselib_invalidate_rtx): Add operand.
5582 2018-08-06 Alan Hayward <alan.hayward@arm.com>
5584 * lra-eliminations.c (lra_eliminate_regs_1): Check for clobber high.
5585 (mark_not_eliminable): Likewise.
5586 * lra-int.h (struct lra_insn_reg): Add clobber high marker.
5587 * lra-lives.c (process_bb_lives): Check for clobber high.
5588 * lra.c (new_insn_reg): Remember clobber highs.
5589 (collect_non_operand_hard_regs): Check for clobber high.
5590 (lra_set_insn_recog_data): Likewise.
5591 (add_regs_to_insn_regno_info): Likewise.
5592 (lra_update_insn_regno_info): Likewise.
5594 2018-08-06 Alan Hayward <alan.hayward@arm.com>
5596 * rtl.h (reg_is_clobbered_by_clobber_high): Add declarations.
5597 * rtlanal.c (reg_is_clobbered_by_clobber_high): Add function.
5599 2018-08-06 Alan Hayward <alan.hayward@arm.com>
5601 * emit-rtl.c (verify_rtx_sharing): Check for CLOBBER_HIGH.
5602 (copy_insn_1): Likewise.
5603 (gen_hard_reg_clobber_high): New gen function.
5604 * genconfig.c (walk_insn_part): Check for CLOBBER_HIGH.
5605 * genemit.c (gen_exp): Likewise.
5606 (gen_emit_seq): Pass through info.
5607 (gen_insn): Check for CLOBBER_HIGH.
5608 (gen_expand): Pass through info.
5609 (gen_split): Likewise.
5610 (output_add_clobbers): Likewise.
5611 * genrecog.c (validate_pattern): Check for CLOBBER_HIGH.
5612 (remove_clobbers): Likewise.
5613 * rtl.h (gen_hard_reg_clobber_high): New declaration.
5615 2018-08-06 Alan Hayward <alan.hayward@arm.com>
5617 * doc/rtl.texi (clobber_high): Add.
5618 (parallel): Add in clobber high
5619 * rtl.c (rtl_check_failed_code3): Add function.
5620 * rtl.def (CLOBBER_HIGH): Add expression.
5621 * rtl.h (RTL_CHECKC3): Add macro.
5622 (rtl_check_failed_code3): Add declaration.
5623 (XC3EXP): Add macro.
5625 2018-08-05 H.J. Lu <hongjiu.lu@intel.com>
5628 * config/i386/i386.c (ix86_finalize_stack_frame_flags): Set
5629 cfun->machine->max_used_stack_alignment if needed.
5631 2018-08-04 Martin Sebor <msebor@redhat.com>
5633 PR tree-optimization/86571
5634 * gimple-ssa-sprintf.c (format_floating): Extend upper bound of
5637 2018-08-03 Sandra Loosemore <sandra@codesourcery.com>
5640 * config/nios2/nios2.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
5643 2018-08-03 Jeff Law <law@redhat.com>
5646 * config/mn10300/mn10300.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
5647 Define to speculation_safe_value_not_needed.
5649 2018-08-03 David Malcolm <dmalcolm@redhat.com>
5651 * doc/gcov.texi (-x): Remove duplicate "to".
5652 * doc/invoke.texi (-Wnoexcept-type): Remove duplicate "calls".
5653 (-Wif-not-aligned): Remove duplicate "is".
5654 (-flto): Remove duplicate "the".
5655 (MicroBlaze Options): In examples of "-mcpu=cpu-type", remove
5656 duplicate "v5.00.b".
5657 (MSP430 Options): Remove duplicate "and" from the description
5658 of "-mgprel-sec=regexp".
5659 (x86 Options): Remove duplicate copies of "vmldLog102" and
5660 vmlsLog104 from description of "-mveclibabi=type".
5662 2018-08-03 Richard Sandiford <richard.sandiford@arm.com>
5664 * internal-fn.h (first_commutative_argument): Declare.
5665 * internal-fn.c (first_commutative_argument): New function.
5666 * tree-vect-slp.c (vect_get_and_check_slp_defs): Remove extra
5667 restrictions for pattern statements. Use first_commutative_argument
5668 to look for commutative operands in calls to internal functions.
5670 2018-08-03 Aldy Hernandez <aldyh@redhat.com>
5672 * Makefile.in (wide-int-range.o): New.
5673 * tree-vrp.c: Move all the wide_int_* functions to...
5674 * wide-int-range.cc: ...here.
5675 * tree-vrp.h: Move all the wide_int_* prototypes to...
5676 * wide-int-range.h: ...here.
5678 2018-08-03 Tom de Vries <tdevries@suse.de>
5680 * common/config/nvptx/nvptx-common.c (nvptx_except_unwind_info): Return
5682 * config/nvptx/nvptx.c (TARGET_ASM_BYTE_OP): Remove define.
5683 * except.c (output_function_exception_table): Do early exit if
5684 targetm_common.except_unwind_info (&global_options) == UI_NONE.
5686 2018-08-03 Martin Liska <mliska@suse.cz>
5688 * predict.c (dump_prediction): Change to 2 digits
5691 2018-08-03 Siddhesh Poyarekar <siddhesh@sourceware.org>
5693 * config/aarch64/falkor.md (falkor_am_1_vxvy_vxvy): Move
5695 (falkor_am_1_gtov_gtov): ... a new insn reservation.
5697 2018-07-19 Ilya Leoshkevich <iii@linux.ibm.com>
5699 * config/nds32/nds32.c (nds32_hard_regno_mode_ok): Replace > with >=.
5700 * df-problems.c (df_remove_dead_eq_notes): Replace > with >=.
5701 * dwarf2out.c (mem_loc_descriptor): Replace > with >=.
5702 * lra-constraints.c (spill_hard_reg_in_range): Replace <= with <.
5703 * lra-remat.c (call_used_input_regno_present_p): Replace <= with <.
5705 2018-08-02 David Malcolm <dmalcolm@redhat.com>
5707 * diagnostic-show-locus.c (diagnostic_show_locus): Use
5708 pp_take_prefix when saving the existing prefix.
5709 * diagnostic.c (diagnostic_append_note): Likewise.
5710 * langhooks.c (lhd_print_error_function): Likewise.
5711 * pretty-print.c (pp_set_prefix): Drop the "const" from "prefix"
5712 param's type. Free the existing prefix.
5713 (pp_take_prefix): New function.
5714 (pretty_printer::pretty_printer): Drop the prefix parameter.
5715 Rename the length parameter to match the comment.
5716 (pretty_printer::~pretty_printer): Free the prefix.
5717 * pretty-print.h (pretty_printer::pretty_printer): Drop the prefix
5719 (struct pretty_printer): Drop the "const" from "prefix" field's
5720 type and clarify memory management.
5721 (pp_set_prefix): Drop the "const" from the 2nd param.
5722 (pp_take_prefix): New decl.
5724 2018-08-02 Aaron Sawdey <acsawdey@linux.ibm.com>
5726 * config/rs6000/rs6000-string.c (select_block_compare_mode): Move test
5727 for word_mode_ok here instead of passing as argument.
5728 (expand_block_compare): Change select_block_compare_mode() call.
5729 (expand_strncmp_gpr_sequence): New function.
5730 (expand_strn_compare): Make use of expand_strncmp_gpr_sequence.
5732 2018-08-02 Jeff Law <law@redhat.com>
5735 * config/m68k/m68k.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
5736 Define to speculation_safe_value_not_needed.
5739 * config/h8300/h8300.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
5740 Define to speculation_safe_value_not_needed.
5742 2018-08-02 Tom de Vries <tdevries@suse.de>
5745 * common/config/nvptx/nvptx-common.c (nvptx_except_unwind_info): New
5746 function. Return UI_TARGET unconditionally.
5747 (TARGET_EXCEPT_UNWIND_INFO): Redefine to nvptx_except_unwind_info.
5748 * config/nvptx/nvptx.c (TARGET_ASM_BYTE_OP): Emit commented out '.byte'.
5750 2018-08-02 Richard Sandiford <richard.sandiford@arm.com>
5752 * genemit.c (print_overload_test): Fix typo.
5754 2018-08-02 Richard Biener <rguenther@suse.de>
5756 PR tree-optimization/86816
5757 * tree-ssa-tail-merge.c (tail_merge_valueize): New function
5758 which checks for value availability before querying it.
5759 (gvn_uses_equal): Use it.
5760 (same_succ_hash): Likewise.
5761 (gimple_equal_p): Likewise.
5763 2018-08-02 Nick Clifton <nickc@redhat.com>
5766 * config/stormy16/stormy16.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
5767 Define to speculation_safe_value_not_needed.
5770 * config/v850/v850.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
5771 Define to speculation_safe_value_not_needed.
5774 * config/v850/v850.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
5775 Define to speculation_safe_value_not_needed.
5778 * config/rx/rx.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
5779 Define to speculation_safe_value_not_needed.
5782 * config/msp430/msp430.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
5783 Define to speculation_safe_value_not_needed.
5786 * config/mcore/mcore.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
5787 Define to speculation_safe_value_not_needed.
5790 * config/m32r/m32r.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
5791 Define to speculation_safe_value_not_needed.
5794 * config/iq2000/iq2000.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
5795 Define to speculation_safe_value_not_needed.
5798 * config/frv/frv.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define to
5799 speculation_safe_value_not_needed.
5802 * config/fr30/fr30.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define
5803 to speculation_safe_value_not_needed.
5805 2018-08-02 Richard Sandiford <richard.sandiford@arm.com>
5807 * doc/md.texi: Expand the documentation of instruction names
5808 to mention port-local uses. Document '@' in pattern names.
5809 * read-md.h (overloaded_instance, overloaded_name): New structs.
5811 (md_reader::handle_overloaded_name): New member function.
5812 (md_reader::get_overloads): Likewise.
5813 (md_reader::m_first_overload): New member variable.
5814 (md_reader::m_next_overload_ptr): Likewise.
5815 (md_reader::m_overloads_htab): Likewise.
5816 * read-md.c (md_reader::md_reader): Initialize m_first_overload,
5817 m_next_overload_ptr and m_overloads_htab.
5818 * read-rtl.c (iterator_group): Add "type" and "get_c_token" fields.
5819 (get_mode_token, get_code_token, get_int_token): New functions.
5820 (map_attr_string): Add an optional argument that passes back
5821 the associated iterator.
5822 (overloaded_name_hash, overloaded_name_eq_p, named_rtx_p):
5823 (md_reader::handle_overloaded_name, add_overload_instance): New
5825 (apply_iterators): Handle '@' names. Report an error if '@'
5826 is used without iterators.
5827 (initialize_iterators): Initialize the new iterator_group fields.
5828 * genopinit.c (handle_overloaded_code_for)
5829 (handle_overloaded_gen): New functions.
5830 (main): Use them to print declarations of maybe_code_for_* and
5831 maybe_gen_* functions, and inline definitions of code_for_* and gen_*.
5832 * genemit.c (print_overload_arguments, print_overload_test)
5833 (handle_overloaded_code_for, handle_overloaded_gen): New functions.
5834 (main): Use it to print definitions of maybe_code_for_* and
5835 maybe_gen_* functions.
5836 * config/aarch64/aarch64.c (aarch64_split_128bit_move): Use
5837 gen_aarch64_mov{low,high}_di and gen_aarch64_movdi_{low,high}
5838 instead of explicit mode checks.
5839 (aarch64_split_simd_combine): Likewise gen_aarch64_simd_combine.
5840 (aarch64_split_simd_move): Likewise gen_aarch64_split_simd_mov.
5841 (aarch64_emit_load_exclusive): Likewise gen_aarch64_load_exclusive.
5842 (aarch64_emit_store_exclusive): Likewise gen_aarch64_store_exclusive.
5843 (aarch64_expand_compare_and_swap): Likewise
5844 gen_aarch64_compare_and_swap and gen_aarch64_compare_and_swap_lse
5845 (aarch64_gen_atomic_cas): Likewise gen_aarch64_atomic_cas.
5846 (aarch64_emit_atomic_swap): Likewise gen_aarch64_atomic_swp.
5847 (aarch64_constant_pool_reload_icode): Delete.
5848 (aarch64_secondary_reload): Use code_for_aarch64_reload_movcp
5849 instead of aarch64_constant_pool_reload_icode. Use
5850 code_for_aarch64_reload_mov instead of explicit mode checks.
5851 (rsqrte_type, get_rsqrte_type, rsqrts_type, get_rsqrts_type): Delete.
5852 (aarch64_emit_approx_sqrt): Use gen_aarch64_rsqrte instead of
5853 get_rsqrte_type and gen_aarch64_rsqrts instead of gen_rqrts_type.
5854 (recpe_type, get_recpe_type, recps_type, get_recps_type): Delete.
5855 (aarch64_emit_approx_div): Use gen_aarch64_frecpe instead of
5856 get_recpe_type and gen_aarch64_frecps instead of get_recps_type.
5857 (aarch64_atomic_load_op_code): Delete.
5858 (aarch64_emit_atomic_load_op): Likewise.
5859 (aarch64_gen_atomic_ldop): Use UNSPECV_ATOMIC_* instead of
5860 aarch64_atomic_load_op_code. Use gen_aarch64_atomic_load
5861 instead of aarch64_emit_atomic_load_op.
5862 * config/aarch64/aarch64.md (aarch64_reload_movcp<GPF_TF:mode><P:mode>)
5863 (aarch64_reload_movcp<VALL:mode><P:mode>, aarch64_reload_mov<mode>)
5864 (aarch64_movdi_<mode>low, aarch64_movdi_<mode>high)
5865 (aarch64_mov<mode>high_di, aarch64_mov<mode>low_di): Add a '@'
5866 character before the pattern name.
5867 * config/aarch64/aarch64-simd.md (aarch64_split_simd_mov<mode>)
5868 (aarch64_rsqrte<mode>, aarch64_rsqrts<mode>)
5869 (aarch64_simd_combine<mode>, aarch64_frecpe<mode>)
5870 (aarch64_frecps<mode>): Likewise.
5871 * config/aarch64/atomics.md (atomic_compare_and_swap<mode>)
5872 (aarch64_compare_and_swap<mode>, aarch64_compare_and_swap<mode>_lse)
5873 (aarch64_load_exclusive<mode>, aarch64_store_exclusive<mode>)
5874 (aarch64_atomic_swp<mode>, aarch64_atomic_cas<mode>)
5875 (aarch64_atomic_load<atomic_ldop><mode>): Likewise.
5877 2018-08-02 Richard Sandiford <richard.sandiford@arm.com>
5879 * config/aarch64/aarch64.c (aarch64_float_const_representable_p):
5880 Allow HFmode constants if TARGET_FP_F16INST.
5882 2018-08-02 Jackson Woodruff <jackson.woodruff@arm.com>
5885 * config/aarch64/aarch64.c (aarch64_operands_adjust_ok_for_ldpstp):
5886 No longer check last store for clobber of address register.
5888 2018-08-02 Martin Liska <mliska@suse.cz>
5890 PR gcov-profile/86817
5891 * gcov.c (process_all_functions): New function.
5893 (process_file): Move functions processing to
5894 process_all_functions.
5896 2018-08-02 David Malcolm <dmalcolm@redhat.com>
5898 * dumpfile.c (dump_user_location_t::dump_user_location_t): Add
5899 "const" to the "gimple *" and "rtx_insn *" parameters.
5900 * dumpfile.h (dump_user_location_t::dump_user_location_t):
5902 (dump_location_t::dump_location_t): Likewise.
5904 2018-08-01 Martin Sebor <msebor@redhat.com>
5906 PR tree-optimization/86650
5907 * gimple-pretty-print.c (percent_G_format): Accept a "gimple *"
5908 rather than a "gcall *". Directly pass the data of interest
5909 to percent_K_format, rather than building a temporary CALL_EXPR
5911 * gimple-fold.c (gimple_fold_builtin_strncpy): Adjust.
5912 (gimple_fold_builtin_strncat): Adjust.
5913 * gimple-ssa-warn-restrict.h (check_bounds_or_overlap): Replace
5914 gcall* argument with gimple*.
5915 * gimple-ssa-warn-restrict.c (check_call): Same.
5916 (wrestrict_dom_walker::before_dom_children): Same.
5917 (builtin_access::builtin_access): Same.
5918 (check_bounds_or_overlap): Same
5919 (maybe_diag_overlap): Same.
5920 (maybe_diag_offset_bounds): Same.
5921 * tree-diagnostic.c (default_tree_printer): Move usage of
5922 EXPR_LOCATION (t) and TREE_BLOCK (t) from within percent_K_format
5924 * tree-pretty-print.c (percent_K_format): Add argument.
5925 * tree-pretty-print.h: Add argument.
5926 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Adjust.
5927 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Adjust.
5928 (maybe_diag_stxncpy_trunc): Same.
5929 (handle_builtin_stxncpy): Same.
5930 (handle_builtin_strcat): Same.
5932 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
5934 * match.pd: Optimise pointer range checks.
5936 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
5938 PR tree-optimization/86758
5939 * tree-vect-stmts.c (vectorizable_simd_clone_call): Don't try
5940 to remove pattern statements.
5942 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
5944 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Use the
5945 result of dfs_enumerate_from when constructing stmt_vec_infos,
5946 instead of additionally calling get_loop_body.
5948 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
5950 * tree-vectorizer.h (vect_create_data_ref_ptr): Remove inv_p
5952 * tree-vect-data-refs.c (vect_create_data_ref_ptr): Likewise.
5953 When creating an iv, assert that the step is not known to be zero.
5954 (vect_setup_realignment): Update call accordingly.
5955 * tree-vect-stmts.c (vectorizable_store): Likewise.
5956 (vectorizable_load): Likewise. Handle VMAT_INVARIANT separately.
5958 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
5960 * tree-vectorizer.h (vect_stmt_to_vectorize): New function.
5961 * tree-vect-loop.c (vect_update_vf_for_slp): Use it.
5962 (vectorizable_reduction): Likewise.
5963 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
5964 (vect_detect_hybrid_slp_stmts): Likewise.
5965 * tree-vect-stmts.c (vect_is_simple_use): Likewise.
5967 2018-08-01 Aldy Hernandez <aldyh@redhat.com>
5969 * tree-vrp (zero_nonzero_bits_from_bounds): Rename to...
5970 (wide_int_set_zero_nonzero_bits): ...this.
5971 (zero_nonzero_bits_from_vr): Rename to...
5972 (vrp_set_zero_nonzero_bits): ...this.
5973 (extract_range_from_multiplicative_op_1): Abstract wide int
5975 (wide_int_range_multiplicative_op): ...here.
5976 (extract_range_from_binary_expr_1): Extract wide int binary
5977 operations into their own functions.
5978 (wide_int_range_lshift): New.
5979 (wide_int_range_can_optimize_bit_op): New.
5980 (wide_int_range_shift_undefined_p): New.
5981 (wide_int_range_bit_xor): New.
5982 (wide_int_range_bit_ior): New.
5983 (wide_int_range_bit_and): New.
5984 (wide_int_range_trunc_mod): New.
5985 (extract_range_into_wide_ints): New.
5986 (vrp_shift_undefined_p): New.
5987 (extract_range_from_multiplicative_op): New.
5988 (vrp_can_optimize_bit_op): New.
5989 * tree-vrp.h (value_range::dump): New.
5990 (wide_int_range_multiplicative_op): New.
5991 (wide_int_range_lshift):New.
5992 (wide_int_range_shift_undefined_p): New.
5993 (wide_int_range_bit_xor): New.
5994 (wide_int_range_bit_ior): New.
5995 (wide_int_range_bit_and): New.
5996 (wide_int_range_trunc_mod): New.
5997 (zero_nonzero_bits_from_bounds): Rename to...
5998 (wide_int_set_zero_nonzero_bits): ...this.
5999 (zero_nonzero_bits_from_vr): Rename to...
6000 (vrp_set_zero_nonzero_bits): ...this.
6001 (range_easy_mask_min_max): Rename to...
6002 (wide_int_range_can_optimize_bit_op): this.
6003 * vr-values.c (simplify_bit_ops_using_ranges): Rename
6004 zero_nonzero_bits_from_vr into vrp_set_zero_nonzero_bits.
6006 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
6008 * tree-vectorizer.h (vect_orig_stmt): New function.
6009 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Use it.
6010 * tree-vect-loop.c (vect_model_reduction_cost): Likewise.
6011 (vect_create_epilog_for_reduction): Likewise.
6012 (vectorizable_live_operation): Likewise.
6013 * tree-vect-slp.c (vect_find_last_scalar_stmt_in_slp): Likewise.
6014 (vect_detect_hybrid_slp_stmts, vect_schedule_slp): Likewise.
6015 * tree-vect-stmts.c (vectorizable_call): Likewise.
6016 (vectorizable_simd_clone_call, vect_remove_stores): Likewise.
6018 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
6020 * tree-vectorizer.h (vect_transform_stmt): Remove grouped_store
6022 * tree-vect-stmts.c (vect_transform_stmt): Likewise.
6023 * tree-vect-loop.c (vect_transform_loop_stmt): Update call accordingly.
6024 (vect_transform_loop): Likewise.
6025 * tree-vect-slp.c (vect_schedule_slp_instance): Likewise.
6027 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
6029 * tree-vectorizer.h (vect_schedule_slp): Return void.
6030 * tree-vect-slp.c (vect_schedule_slp_instance): Likewise.
6031 (vect_schedule_slp): Likewise.
6033 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
6035 * tree-vect-loop.c (vect_transform_loop_stmt): Remove slp_scheduled
6037 (vect_transform_loop): Update calls accordingly. Schedule SLP
6038 instances before the main loop, if any exist.
6040 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
6042 PR tree-optimization/86749
6043 * tree-vect-patterns.c (vect_determine_min_output_precision_1):
6044 If the lhs is used in a COND_EXPR, check that it is being used
6045 as the "then" or "else" value.
6047 2018-08-01 Tom de Vries <tdevries@suse.de>
6050 * config/nvptx/nvptx.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define to
6051 speculation_safe_value_not_needed.
6053 2018-08-01 Richard Biener <rguenther@suse.de>
6055 * tree-ssa-sccvn.c (visit_phi): Compare invariant addresses
6058 2018-08-01 Martin Liska <mliska@suse.cz>
6060 * value-prof.c (gimple_divmod_fixed_value_transform): Unify
6061 format how successful transformation is dumped.
6062 (gimple_mod_pow2_value_transform): Likewise.
6063 (gimple_mod_subtract_transform): Likewise.
6064 (gimple_stringops_transform): Likewise.
6066 2018-08-01 Martin Liska <mliska@suse.cz>
6069 * value-prof.c (interesting_stringop_to_profile_p):
6070 Simplify the code and add BUILT_IN_MEMMOVE.
6071 (gimple_stringops_transform): Likewise.
6073 2018-08-01 Sam Tebbs <sam.tebbs@arm.com>
6075 * config/aarch64/aarch64-simd.md
6076 (*aarch64_get_lane_zero_extendsi<mode>): Rename to...
6077 (*aarch64_get_lane_zero_extend<GPI:mode><VDQQH:mode>): ... This and
6078 use GPI iterator instead of SI mode.
6080 2018-08-01 Richard Earnshaw <rearnsha@arm.com>
6082 * config/rs6000/rs6000.md (speculation_barrier): Renamed from
6083 rs6000_speculation_barrier.
6084 * config/rs6000/rs6000.c (rs6000_expand_builtin): Adjust for
6085 new barrier pattern name.
6087 2018-08-01 Richard Earnshaw <rearnsha@arm.com>
6089 * config/i386/i386.md (unspecv): Add UNSPECV_SPECULATION_BARRIER.
6090 (speculation_barrier): New insn.
6092 2018-08-01 Richard Biener <rguenther@suse.de>
6095 * graphite.h: Include isl/id.h and isl/space.h to allow build
6098 2018-08-01 Jan Willem Jagersma <jwjagersma@gmail.com>
6101 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
6102 mode for COFF targets.
6103 * defaults.h (TARGET_COFF): Define.
6104 * config/i386/djgpp.h (TARGET_ASM_LTO_START, TARGET_ASM_LTO_END,
6105 TARGET_COFF): Define.
6106 (i386_djgpp_asm_lto_start, i386_djgpp_asm_lto_end): Declare.
6107 * config/i386/djgpp.c (saved_debug_info_level): New static variable.
6108 (i386_djgpp_asm_lto_start, i386_djgpp_asm_lto_end): New functions.
6110 2018-07-31 Alexandre Oliva <oliva@adacore.com>
6112 * gimple-streamer-in.c (input_bb): Restore BB discriminator.
6113 * gimple-streamer-out.c (output_bb): Save it.
6114 * lto-streamer-in.c (input_struct_function_base): Restore
6115 instance discriminator if available. Create map on demand.
6116 * lto-streamer-out.c (output_struct_function_base): Save it if
6118 * final.c (decl_to_instance_map): Document LTO strategy.
6120 2018-07-31 Alexandre Oliva <oliva@adacore.com>
6121 Olivier Hainque <hainque@adacore.com>
6123 * debug.h (decl_to_instance_map_t): New type.
6124 (decl_to_instance_map): Declare.
6125 (maybe_create_decl_to_instance_map): New inline function.
6126 * final.c (bb_discriminator, last_bb_discriminator): New statics,
6127 to track basic block discriminators.
6128 (final_start_function_1): Initialize them.
6129 (final_scan_insn_1): On NOTE_INSN_BASIC_BLOCK, track
6131 (decl_to_instance_map): New variable.
6132 (map_decl_to_instance, maybe_set_discriminator): New functions.
6133 (notice_source_line): Set discriminator.
6135 2018-07-31 Ian Lance Taylor <iant@golang.org>
6137 * targhooks.c (default_have_speculation_safe_value): Add
6140 2018-07-31 David Malcolm <dmalcolm@redhat.com>
6142 * dump-context.h: Include "pretty-print.h".
6143 (dump_context::refresh_dumps_are_enabled): New decl.
6144 (dump_context::emit_item): New decl.
6145 (class dump_context): Add fields "m_test_pp" and
6147 (temp_dump_context::temp_dump_context): Add param "test_pp_flags".
6148 (temp_dump_context::get_dumped_text): New decl.
6149 (class temp_dump_context): Add field "m_pp".
6150 * dumpfile.c (refresh_dumps_are_enabled): Convert to...
6151 (dump_context::refresh_dumps_are_enabled): ...and add a test for
6153 (set_dump_file): Update for above change.
6154 (set_alt_dump_file): Likewise.
6155 (dump_loc): New overload, taking a pretty_printer *.
6156 (dump_context::dump_loc): Call end_any_optinfo. Dump the location
6157 to any test pretty-printer.
6158 (make_item_for_dump_gimple_stmt): New function, adapted from
6159 optinfo::add_gimple_stmt.
6160 (dump_context::dump_gimple_stmt): Call it, and use the result,
6161 eliminating the direct usage of dump_file and alt_dump_file in
6162 favor of indirectly using them via emit_item.
6163 (make_item_for_dump_gimple_expr): New function, adapted from
6164 optinfo::add_gimple_expr.
6165 (dump_context::dump_gimple_expr): Call it, and use the result,
6166 eliminating the direct usage of dump_file and alt_dump_file in
6167 favor of indirectly using them via emit_item.
6168 (make_item_for_dump_generic_expr): New function, adapted from
6170 (dump_context::dump_generic_expr): Call it, and use the result,
6171 eliminating the direct usage of dump_file and alt_dump_file in
6172 favor of indirectly using them via emit_item.
6173 (make_item_for_dump_printf_va): New function, adapted from
6174 optinfo::add_printf_va.
6175 (make_item_for_dump_printf): New function.
6176 (dump_context::dump_printf_va): Call make_item_for_dump_printf_va,
6177 and use the result, eliminating the direct usage of dump_file and
6178 alt_dump_file in favor of indirectly using them via emit_item.
6179 (make_item_for_dump_dec): New function.
6180 (dump_context::dump_dec): Call it, and use the result,
6181 eliminating the direct usage of dump_file and alt_dump_file in
6182 favor of indirectly using them via emit_item.
6183 (make_item_for_dump_symtab_node): New function, adapted from
6184 optinfo::add_symtab_node.
6185 (dump_context::dump_symtab_node): Call it, and use the result,
6186 eliminating the direct usage of dump_file and alt_dump_file in
6187 favor of indirectly using them via emit_item.
6188 (dump_context::begin_scope): Reimplement, avoiding direct usage
6189 of dump_file and alt_dump_file in favor of indirectly using them
6191 (dump_context::emit_item): New member function.
6192 (temp_dump_context::temp_dump_context): Add param "test_pp_flags".
6193 Set up test pretty-printer on the underlying context. Call
6194 refresh_dumps_are_enabled.
6195 (temp_dump_context::~temp_dump_context): Call
6196 refresh_dumps_are_enabled.
6197 (temp_dump_context::get_dumped_text): New member function.
6198 (selftest::verify_dumped_text): New function.
6199 (ASSERT_DUMPED_TEXT_EQ): New macro.
6200 (selftest::test_capture_of_dump_calls): Run all tests twice, with
6201 and then without optinfo enabled. Add uses of
6202 ASSERT_DUMPED_TEXT_EQ to all tests. Add test of nested scopes.
6203 * dumpfile.h: Update comment for the dump_* API.
6204 * optinfo-emit-json.cc
6205 (selftest::test_building_json_from_dump_calls): Update for new
6206 param for temp_dump_context ctor.
6207 * optinfo.cc (optinfo_item::optinfo_item): Remove "owned" param
6208 and "m_owned" field.
6209 (optinfo_item::~optinfo_item): Likewise.
6210 (optinfo::add_item): New member function.
6211 (optinfo::emit): Update comment.
6212 (optinfo::add_string): Delete.
6213 (optinfo::add_printf): Delete.
6214 (optinfo::add_printf_va): Delete.
6215 (optinfo::add_gimple_stmt): Delete.
6216 (optinfo::add_gimple_expr): Delete.
6217 (optinfo::add_tree): Delete.
6218 (optinfo::add_symtab_node): Delete.
6219 (optinfo::add_dec): Delete.
6220 * optinfo.h (class dump_context): New forward decl.
6221 (optinfo::add_item): New decl.
6222 (optinfo::add_string): Delete.
6223 (optinfo::add_printf): Delete.
6224 (optinfo::add_printf_va): Delete.
6225 (optinfo::add_gimple_stmt): Delete.
6226 (optinfo::add_gimple_expr): Delete.
6227 (optinfo::add_tree): Delete.
6228 (optinfo::add_symtab_node): Delete.
6229 (optinfo::add_dec): Delete.
6230 (optinfo::add_poly_int): Delete.
6231 (optinfo_item::optinfo_item): Remove "owned" param.
6232 (class optinfo_item): Remove field "m_owned".
6234 2018-07-31 Jozef Lawrynowicz <jozef.l@mittosystems.com>
6237 * gcc/cfgexpand.c (set_parm_rtl): Use the alignment of Pmode when
6238 MAX_SUPPORTED_STACK_ALIGNMENT would otherwise be exceeded by the
6239 requested variable alignment.
6240 (expand_one_ssa_partition): Likewise.
6241 (expand_one_var): Likewise.
6243 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
6245 * config/pdp11/pdp11.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
6246 to speculation_safe_value_not_needed.
6248 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
6250 * targhooks.h (speculation_safe_value_not_needed): New prototype.
6251 * targhooks.c (speculation_safe_value_not_needed): New function.
6252 * target.def (have_speculation_safe_value): Update documentation.
6253 * doc/tm.texi: Regenerated.
6255 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
6257 * config/aarch64/iterators.md (ALLI_TI): New iterator.
6258 * config/aarch64/aarch64.md (despeculate_copy<ALLI_TI:mode>): New
6260 (despeculate_copy<ALLI:mode>_insn): New insn.
6261 (despeculate_copyti_insn): New insn.
6262 (despeculate_simple<ALLI:mode>): New insn
6263 (despeculate_simpleti): New insn.
6264 * config/aarch64/aarch64.c (aarch64_speculation_safe_value): New
6266 (TARGET_SPECULATION_SAFE_VALUE): Redefine to
6267 aarch64_speculation_safe_value.
6268 (aarch64_print_operand): Handle const0_rtx in modifier 'H'.
6270 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
6272 * config/aarch64/aarch64-speculation.cc: New file.
6273 * config/aarch64/aarch64-passes.def (pass_track_speculation): Add
6274 before pass_reorder_blocks.
6275 * config/aarch64/aarch64-protos.h (make_pass_track_speculation): Add
6277 * config/aarch64/aarch64.c (aarch64_conditional_register_usage): Fix
6278 X14 and X15 when tracking speculation.
6279 * config/aarch64/aarch64.md (register name constants): Add
6280 SPECULATION_TRACKER_REGNUM and SPECULATION_SCRATCH_REGNUM.
6281 (unspec): Add UNSPEC_SPECULATION_TRACKER.
6282 (speculation_barrier): New insn attribute.
6283 (cmp<mode>): Allow SP in comparisons.
6284 (speculation_tracker): New insn.
6285 (speculation_barrier): Add speculation_barrier attribute.
6286 * config/aarch64/t-aarch64: Add make rule for aarch64-speculation.o.
6287 * config.gcc (aarch64*-*-*): Add aarch64-speculation.o to extra_objs.
6288 * doc/invoke.texi (AArch64 Options): Document -mtrack-speculation.
6290 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
6292 * config/aarch64/aarch64.md (cb<optab><mode>1): Disable when
6293 aarch64_track_speculation is true.
6294 (tb<optab><mode>1): Likewise.
6295 * config/aarch64/aarch64.c (aarch64_split_compare_regs): Do not
6296 generate CB[N]Z when tracking speculation.
6297 (aarch64_split_compare_and_swap): Likewise.
6298 (aarch64_split_atomic_op): Likewise.
6300 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
6302 * config/aarch64/aarch64.opt (mtrack-speculation): New target option.
6304 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
6306 * config/aarch64.md (unspecv): Add UNSPECV_SPECULAION_BARRIER.
6307 (speculation_barrier): New insn.
6309 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
6311 * config/arm/unspecs.md (unspecv): Add VUNSPEC_SPECULATION_BARRIER.
6312 * config/arm/arm.md (speculation_barrier): New expand.
6313 (speculation_barrier_insn): New pattern.
6315 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
6317 * builtin-types.def (BT_FN_PTR_PTR_VAR): New function type.
6318 (BT_FN_I1_I1_VAR, BT_FN_I2_I2_VAR, BT_FN_I4_I4_VAR): Likewise.
6319 (BT_FN_I8_I8_VAR, BT_FN_I16_I16_VAR): Likewise.
6320 * builtin-attrs.def (ATTR_NOVOPS_NOTHROW_LEAF_LIST): New attribute
6322 * builtins.def (BUILT_IN_SPECULATION_SAFE_VALUE_N): New builtin.
6323 (BUILT_IN_SPECULATION_SAFE_VALUE_PTR): New internal builtin.
6324 (BUILT_IN_SPECULATION_SAFE_VALUE_1): Likewise.
6325 (BUILT_IN_SPECULATION_SAFE_VALUE_2): Likewise.
6326 (BUILT_IN_SPECULATION_SAFE_VALUE_4): Likewise.
6327 (BUILT_IN_SPECULATION_SAFE_VALUE_8): Likewise.
6328 (BUILT_IN_SPECULATION_SAFE_VALUE_16): Likewise.
6329 * builtins.c (expand_speculation_safe_value): New function.
6330 (expand_builtin): Call it.
6331 * doc/cpp.texi: Document predefine __HAVE_SPECULATION_SAFE_VALUE.
6332 * doc/extend.texi: Document __builtin_speculation_safe_value.
6333 * doc/md.texi: Document "speculation_barrier" pattern.
6334 * doc/tm.texi.in: Pull in TARGET_SPECULATION_SAFE_VALUE and
6335 TARGET_HAVE_SPECULATION_SAFE_VALUE.
6336 * doc/tm.texi: Regenerated.
6337 * target.def (have_speculation_safe_value, speculation_safe_value): New
6339 * targhooks.c (default_have_speculation_safe_value): New function.
6340 (default_speculation_safe_value): New function.
6341 * targhooks.h (default_have_speculation_safe_value): Add prototype.
6342 (default_speculation_safe_value): Add prototype.
6344 2018-07-31 David Malcolm <dmalcolm@redhat.com>
6346 * dump-context.h (dump_context::dump_loc): New decl.
6347 * dumpfile.c (dump_context::dump_loc): New member function.
6348 (dump_context::dump_gimple_stmt_loc): Reimplement using dump_loc
6349 and dump_gimple_stmt.
6350 (dump_context::dump_gimple_expr_loc): Likewise, using
6352 (dump_context::dump_generic_expr_loc): Likewise, using
6354 (dump_context::dump_printf_loc_va): Likewise, using
6356 (dump_context::begin_scope): Explicitly using the global function
6357 "dump_loc", rather than the member function.
6359 2018-07-31 Martin Sebor <msebor@redhat.com>
6361 PR tree-optimization/86741
6362 * tree-vrp.c (vrp_prop::check_mem_ref): Avoid incomplete types.
6364 2018-07-31 Andreas Krebbel <krebbel@linux.ibm.com>
6366 * config/s390/s390.c (s390_expand_setmem): Make the unrolling to
6367 depend on whether prefetch instructions will be emitted or not.
6368 Use TARGET_SETMEM_PFD for checking whether prefetch instructions
6369 will be emitted or not.
6370 * config/s390/s390.h (TARGET_SETMEM_PREFETCH_DISTANCE)
6371 (TARGET_SETMEM_PFD): New macros.
6373 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6375 * tree-vectorizer.h (stmt_vec_info): Turn back into a typedef.
6376 (NULL_STMT_VEC_INFO): Delete.
6377 (stmt_vec_info::operator*): Likewise.
6378 (stmt_vec_info::operator gimple *): Likewise.
6379 * tree-vect-loop.c (vectorizable_reduction): Use NULL instead
6380 of NULL_STMT_VEC_INFO.
6381 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
6382 (vect_reassociating_reduction_p): Likewise.
6383 * tree-vect-stmts.c (vect_build_gather_load_calls): Likewise.
6384 (vectorizable_store): Likewise.
6385 * tree-vectorizer.c (vec_info::set_vinfo_for_stmt): Likewise.
6386 (vec_info::free_stmt_vec_infos): Likewise.
6388 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6390 * tree-vectorizer.h (vect_stmt_in_region_p): Delete.
6391 * tree-vectorizer.c (vect_stmt_in_region_p): Likewise.
6393 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6395 * tree-vectorizer.h (vec_info::new_vinfo_for_stmt)
6396 (vec_info::set_vinfo_for_stmt, vec_info::free_stmt_vec_infos)
6397 (vec_info::free_stmt_vec_info): New private member functions.
6398 (set_stmt_vec_info_vec, free_stmt_vec_infos, vinfo_for_stmt)
6399 (set_vinfo_for_stmt, new_stmt_vec_info, free_stmt_vec_info): Delete.
6400 * tree-parloops.c (gather_scalar_reductions): Remove calls to
6401 set_stmt_vec_info_vec and free_stmt_vec_infos.
6402 * tree-vect-loop.c (_loop_vec_info): Remove call to
6403 set_stmt_vec_info_vec.
6404 * tree-vect-stmts.c (new_stmt_vec_info, set_stmt_vec_info_vec)
6405 (free_stmt_vec_infos, free_stmt_vec_info): Delete in favor of...
6406 * tree-vectorizer.c (vec_info::new_stmt_vec_info)
6407 (vec_info::set_vinfo_for_stmt, vec_info::free_stmt_vec_infos)
6408 (vec_info::free_stmt_vec_info): ...these new functions. Remove
6409 assignments in {vec_info::,}new_stmt_vec_info that are redundant
6410 with the clearing in the xcalloc.
6411 (stmt_vec_info_vec): Delete.
6412 (vec_info::vec_info): Don't call set_stmt_vec_info_vec.
6413 (vectorize_loops): Likewise.
6414 (vec_info::~vec_info): Remove argument from call to
6415 free_stmt_vec_infos.
6416 (vec_info::add_stmt): Remove vinfo argument from call to
6419 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6421 * tree-vectorizer.h (free_stmt_vec_info): Take a stmt_vec_info
6422 rather than a gimple stmt.
6423 * tree-vect-stmts.c (free_stmt_vec_info): Likewise. Don't free
6424 information for pattern statements when passed the original
6425 statement; instead wait to be passed the pattern statement itself.
6426 Don't call set_vinfo_for_stmt here.
6427 (free_stmt_vec_infos): Update call to free_stmt_vec_info.
6428 * tree-vect-loop.c (_loop_vec_info::~loop_vec_info): Don't free
6429 stmt_vec_infos here.
6430 * tree-vect-slp.c (_bb_vec_info::~bb_vec_info): Likewise.
6431 * tree-vectorizer.c (vec_info::remove_stmt): Nullify the statement's
6432 stmt_vec_infos entry.
6434 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6436 * tree-vectorizer.h (vec_info::replace_stmt): Declare.
6437 * tree-vectorizer.c (vec_info::replace_stmt): New function.
6438 * tree-vect-slp.c (vect_remove_slp_scalar_calls): Use it.
6439 * tree-vect-stmts.c (vectorizable_call): Likewise.
6440 (vectorizable_simd_clone_call): Likewise.
6442 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6444 * tree-vectorizer.h (vec_info::remove_stmt): Declare.
6445 * tree-vectorizer.c (vec_info::remove_stmt): New function.
6446 * tree-vect-loop-manip.c (vect_set_loop_condition): Use it.
6447 * tree-vect-loop.c (vect_transform_loop): Likewise.
6448 * tree-vect-slp.c (vect_schedule_slp): Likewise.
6449 * tree-vect-stmts.c (vect_remove_stores): Likewise.
6451 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6453 * tree-vectorizer.h (vec_info::lookup_dr): New member function.
6454 (vect_dr_stmt): Delete.
6455 * tree-vectorizer.c (vec_info::lookup_dr): New function.
6456 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Use it instead
6458 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
6459 (vect_analyze_data_ref_dependence, vect_record_base_alignments)
6460 (vect_verify_datarefs_alignment, vect_peeling_supportable)
6461 (vect_analyze_data_ref_accesses, vect_prune_runtime_alias_test_list)
6462 (vect_analyze_data_refs): Likewise.
6463 (vect_slp_analyze_data_ref_dependence): Likewise. Take a vec_info
6465 (vect_find_same_alignment_drs): Likewise.
6466 (vect_slp_analyze_node_dependences): Update calls accordingly.
6467 (vect_analyze_data_refs_alignment): Likewise. Use vec_info::lookup_dr
6468 instead of DR_VECT_AUX.
6469 (vect_get_peeling_costs_all_drs): Take a loop_vec_info instead
6470 of a vector data references. Use vec_info::lookup_dr instead of
6472 (vect_peeling_hash_get_lowest_cost): Update calls accordingly.
6473 (vect_enhance_data_refs_alignment): Likewise. Use vec_info::lookup_dr
6474 instead of DR_VECT_AUX.
6476 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6478 * tree-vectorizer.h (_loop_vec_info::unaligned_dr): Change to
6480 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Update
6482 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
6483 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
6484 (vect_gen_prolog_loop_niters): Likewise.
6486 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6488 * tree-vectorizer.h (set_dr_misalignment, dr_misalignment)
6489 (DR_TARGET_ALIGNMENT, aligned_access_p, known_alignment_for_access_p)
6490 (vect_known_alignment_in_bytes, vect_dr_behavior)
6491 (vect_get_scalar_dr_size): Take references as dr_vec_infos
6492 instead of data_references. Update calls to other routines for
6493 which the same change has been made.
6494 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Take
6495 dr_vec_infos instead of stmt_vec_infos.
6496 (vect_analyze_data_ref_dependence): Update call accordingly.
6497 (vect_slp_analyze_data_ref_dependence)
6498 (vect_record_base_alignments): Use DR_VECT_AUX.
6499 (vect_calculate_target_alignment, vect_compute_data_ref_alignment)
6500 (vect_update_misalignment_for_peel, verify_data_ref_alignment)
6501 (vector_alignment_reachable_p, vect_get_data_access_cost)
6502 (vect_peeling_supportable, vect_analyze_group_access_1)
6503 (vect_analyze_group_access, vect_analyze_data_ref_access)
6504 (vect_vfa_segment_size, vect_vfa_access_size, vect_vfa_align)
6505 (vect_compile_time_alias, vect_small_gap_p)
6506 (vectorizable_with_step_bound_p, vect_duplicate_ssa_name_ptr_info):
6507 (vect_supportable_dr_alignment): Take references as dr_vec_infos
6508 instead of data_references. Update calls to other routines for
6509 which the same change has been made.
6510 (vect_verify_datarefs_alignment, vect_get_peeling_costs_all_drs)
6511 (vect_find_same_alignment_drs, vect_analyze_data_refs_alignment)
6512 (vect_slp_analyze_and_verify_node_alignment)
6513 (vect_analyze_data_ref_accesses, vect_prune_runtime_alias_test_list)
6514 (vect_create_addr_base_for_vector_ref, vect_create_data_ref_ptr)
6515 (vect_setup_realignment): Use dr_vec_infos. Update calls after
6517 (_vect_peel_info::dr): Replace with...
6518 (_vect_peel_info::dr_info): ...this new field.
6519 (vect_peeling_hash_get_most_frequent)
6520 (vect_peeling_hash_choose_best_peeling): Update accordingly.
6521 (vect_peeling_hash_get_lowest_cost):
6522 (vect_enhance_data_refs_alignment): Likewise. Update calls to other
6523 routines for which the same change has been made.
6524 (vect_peeling_hash_insert): Likewise. Take a dr_vec_info instead of a
6526 * tree-vect-loop-manip.c (get_misalign_in_elems)
6527 (vect_gen_prolog_loop_niters): Use dr_vec_infos. Update calls after
6529 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
6530 * tree-vect-stmts.c (vect_get_store_cost, vect_get_load_cost)
6531 (vect_truncate_gather_scatter_offset, compare_step_with_zero)
6532 (get_group_load_store_type, get_negative_load_store_type)
6533 (vect_get_data_ptr_increment, vectorizable_store)
6534 (vectorizable_load): Likewise.
6535 (ensure_base_align): Take a dr_vec_info instead of a data_reference.
6536 Update calls to other routines for which the same change has been made.
6538 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6540 * tree-vectorizer.h (vec_info::move_dr): New member function.
6541 (dataref_aux): Rename to...
6542 (dr_vec_info): ...this and add "dr" and "stmt" fields.
6543 (_stmt_vec_info::dr_aux): Update accordingly.
6544 (_stmt_vec_info::data_ref_info): Delete.
6545 (STMT_VINFO_GROUPED_ACCESS, DR_GROUP_FIRST_ELEMENT)
6546 (DR_GROUP_NEXT_ELEMENT, DR_GROUP_SIZE, DR_GROUP_STORE_COUNT)
6547 (DR_GROUP_GAP, DR_GROUP_SAME_DR_STMT, REDUC_GROUP_FIRST_ELEMENT):
6548 (REDUC_GROUP_NEXT_ELEMENT, REDUC_GROUP_SIZE): Use dr_aux.dr instead
6550 (STMT_VINFO_DATA_REF): Likewise. Turn into an lvalue.
6551 (STMT_VINFO_DR_INFO): New macro.
6552 (DR_VECT_AUX): Use STMT_VINFO_DR_INKFO and vect_dr_stmt.
6553 (set_dr_misalignment): Update after rename of dataref_aux.
6554 (vect_dr_stmt): Move earlier in file. Return dr_aux.stmt.
6555 * tree-vect-stmts.c (new_stmt_vec_info): Remove redundant
6556 initialization of STMT_VINFO_DATA_REF.
6557 * tree-vectorizer.c (vec_info::move_dr): New function.
6558 * tree-vect-patterns.c (vect_recog_bool_pattern)
6559 (vect_recog_mask_conversion_pattern)
6560 (vect_recog_gather_scatter_pattern): Use it.
6561 * tree-vect-data-refs.c (vect_analyze_data_refs): Initialize
6562 the "dr" and "stmt" fields of dr_vec_info instead of
6563 STMT_VINFO_DATA_REF.
6565 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6567 * tree-vectorizer.h (_stmt_vec_info::pattern_stmt_p): New field.
6568 (is_pattern_stmt_p): Use it.
6569 * tree-vect-patterns.c (vect_init_pattern_stmt): Set pattern_stmt_p
6570 on pattern statements.
6572 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6574 * tree-vect-patterns.c (vect_mark_pattern_stmts): Take the
6575 original stmt as a stmt_vec_info rather than a gimple stmt.
6576 (vect_pattern_recog_1): Take the statement directly as a
6577 stmt_vec_info, rather than via a gimple_stmt_iterator.
6578 Update call to vect_mark_pattern_stmts.
6579 (vect_pattern_recog): Update calls accordingly.
6581 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6583 * tree-vectorizer.h (vect_get_vec_defs_for_stmt_copy)
6584 (vect_get_vec_def_for_stmt_copy): Take a vec_info rather than
6585 a vect_def_type for the first argument.
6586 * tree-vect-stmts.c (vect_get_vec_defs_for_stmt_copy): Likewise.
6587 (vect_get_vec_def_for_stmt_copy): Likewise. Return the original
6588 operand if it isn't defined by a vectorized statement.
6589 (vect_build_gather_load_calls): Remove the mask_dt argument and
6590 update calls to vect_get_vec_def_for_stmt_copy.
6591 (vectorizable_bswap): Likewise the dt argument.
6592 (vectorizable_call): Update calls to vectorizable_bswap and
6593 vect_get_vec_def_for_stmt_copy.
6594 (vectorizable_simd_clone_call, vectorizable_assignment)
6595 (vectorizable_shift, vectorizable_operation, vectorizable_condition)
6596 (vectorizable_comparison): Update calls to
6597 vect_get_vec_def_for_stmt_copy.
6598 (vectorizable_store): Likewise. Remove now-unnecessary calls to
6600 (vect_get_loop_based_defs): Remove dt argument and update call
6601 to vect_get_vec_def_for_stmt_copy.
6602 (vectorizable_conversion): Update calls to vect_get_loop_based_defs
6603 and vect_get_vec_def_for_stmt_copy.
6604 (vectorizable_load): Update calls to vect_build_gather_load_calls
6605 and vect_get_vec_def_for_stmt_copy.
6606 * tree-vect-loop.c (vect_create_epilog_for_reduction)
6607 (vectorizable_reduction, vectorizable_live_operation): Update calls
6608 to vect_get_vec_def_for_stmt_copy.
6610 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6612 * tree-vect-data-refs.c (vect_record_base_alignment): Replace vec_info
6613 and gimple stmt arguments with a stmt_vec_info.
6614 (vect_record_base_alignments): Update calls accordingly.
6615 * tree-vect-slp.c (vect_record_max_nunits): Replace vec_info
6616 and gimple stmt arguments with a stmt_vec_info.
6617 (vect_build_slp_tree_1): Remove vinfo argument and update call
6618 to vect_record_max_nunits.
6619 (vect_build_slp_tree_2): Update calls to vect_build_slp_tree_1
6620 and vect_record_max_nunits.
6622 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6624 * tree-vectorizer.h (nested_in_vect_loop_p): Move further down
6625 file and take a stmt_vec_info instead of a gimple stmt.
6626 (supportable_widening_operation, vect_finish_replace_stmt)
6627 (vect_finish_stmt_generation, vect_get_store_rhs)
6628 (vect_get_vec_def_for_operand_1, vect_get_vec_def_for_operand)
6629 (vect_get_vec_defs, vect_init_vector, vect_transform_stmt)
6630 (vect_remove_stores, vect_analyze_stmt, vectorizable_condition)
6631 (vect_get_smallest_scalar_type, vect_check_gather_scatter)
6632 (vect_create_data_ref_ptr, bump_vector_ptr)
6633 (vect_permute_store_chain, vect_setup_realignment)
6634 (vect_transform_grouped_load, vect_record_grouped_load_vectors)
6635 (vect_create_addr_base_for_vector_ref, vectorizable_live_operation)
6636 (vectorizable_reduction, vectorizable_induction)
6637 (get_initial_def_for_reduction, is_simple_and_all_uses_invariant)
6638 (vect_get_place_in_interleaving_chain): Take stmt_vec_infos rather
6639 than gimple stmts as arguments.
6640 * tree-vect-data-refs.c (vect_get_smallest_scalar_type)
6641 (vect_preserves_scalar_order_p, vect_slp_analyze_node_dependences)
6642 (can_group_stmts_p, vect_check_gather_scatter)
6643 (vect_create_addr_base_for_vector_ref, vect_create_data_ref_ptr)
6644 (bump_vector_ptr, vect_permute_store_chain, vect_setup_realignment)
6645 (vect_permute_load_chain, vect_shift_permute_load_chain)
6646 (vect_transform_grouped_load)
6647 (vect_record_grouped_load_vectors): Likewise.
6648 * tree-vect-loop.c (vect_fixup_reduc_chain)
6649 (get_initial_def_for_reduction, vect_create_epilog_for_reduction)
6650 (vectorize_fold_left_reduction, is_nonwrapping_integer_induction)
6651 (vectorizable_reduction, vectorizable_induction)
6652 (vectorizable_live_operation, vect_loop_kill_debug_uses): Likewise.
6653 * tree-vect-patterns.c (type_conversion_p, adjust_bool_stmts)
6654 (vect_get_load_store_mask): Likewise.
6655 * tree-vect-slp.c (vect_get_place_in_interleaving_chain)
6656 (vect_analyze_slp_instance, vect_mask_constant_operand_p): Likewise.
6657 * tree-vect-stmts.c (vect_mark_relevant)
6658 (is_simple_and_all_uses_invariant)
6659 (exist_non_indexing_operands_for_use_p, process_use)
6660 (vect_init_vector_1, vect_init_vector, vect_get_vec_def_for_operand_1)
6661 (vect_get_vec_def_for_operand, vect_get_vec_defs)
6662 (vect_finish_stmt_generation_1, vect_finish_replace_stmt)
6663 (vect_finish_stmt_generation, vect_truncate_gather_scatter_offset)
6664 (compare_step_with_zero, vect_get_store_rhs, get_group_load_store_type)
6665 (get_negative_load_store_type, get_load_store_type)
6666 (vect_check_load_store_mask, vect_check_store_rhs)
6667 (vect_build_gather_load_calls, vect_get_strided_load_store_ops)
6668 (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
6669 (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
6670 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
6671 (get_group_alias_ptr_type, vectorizable_store, hoist_defs_of_uses)
6672 (vectorizable_load, vectorizable_condition, vectorizable_comparison)
6673 (vect_analyze_stmt, vect_transform_stmt, vect_remove_stores)
6674 (supportable_widening_operation): Likewise.
6676 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6678 * tree-vect-data-refs.c (vect_describe_gather_scatter_call): Take
6679 a stmt_vec_info instead of a gcall.
6680 (vect_check_gather_scatter): Update call accordingly.
6681 * tree-vect-loop-manip.c (iv_phi_p): Take a stmt_vec_info instead
6683 (vect_can_advance_ivs_p, vect_update_ivs_after_vectorizer)
6684 (slpeel_update_phi_nodes_for_loops):): Update calls accordingly.
6685 * tree-vect-loop.c (vect_transform_loop_stmt): Take a stmt_vec_info
6686 instead of a gimple stmt.
6687 (vect_transform_loop): Update calls accordingly.
6688 * tree-vect-slp.c (vect_split_slp_store_group): Take and return
6689 stmt_vec_infos instead of gimple stmts.
6690 (vect_analyze_slp_instance): Update use accordingly.
6691 * tree-vect-stmts.c (read_vector_array, write_vector_array)
6692 (vect_clobber_variable, vect_stmt_relevant_p, permute_vec_elements)
6693 (vect_use_strided_gather_scatters_p, vect_build_all_ones_mask)
6694 (vect_build_zero_merge_argument, vect_get_gather_scatter_ops)
6695 (vect_gen_widened_results_half, vect_get_loop_based_defs)
6696 (vect_create_vectorized_promotion_stmts, can_vectorize_live_stmts):
6697 Take a stmt_vec_info instead of a gimple stmt and pass stmt_vec_infos
6698 down to subroutines.
6700 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6702 * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Change the type
6703 of the worklist from a vector of gimple stmts to a vector of
6705 * tree-vect-stmts.c (vect_mark_relevant, process_use)
6706 (vect_mark_stmts_to_be_vectorized): Likewise
6708 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6710 * tree-vect-loop.c (vect_analyze_loop_operations): Look up the
6711 statement before passing it to vect_analyze_stmt.
6712 (vect_create_epilog_for_reduction): Use a stmt_vec_info to walk
6713 the chain of phi vector definitions. Track the exit phi via its
6715 (vectorizable_reduction): Set cond_stmt_vinfo directly from the
6716 STMT_VINFO_REDUC_DEF.
6717 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Use
6718 stmt_vec_infos to handle the statement chains.
6719 (vect_get_slp_defs): Record the first statement in the node
6720 using a stmt_vec_info.
6721 * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Look up
6722 statements here and pass their stmt_vec_info down to subroutines.
6723 (vect_init_vector_1): Hoist call to vinfo_for_stmt and pass it
6724 down to vect_finish_stmt_generation.
6725 (vect_init_vector, vect_get_vec_defs, vect_finish_replace_stmt)
6726 (vect_finish_stmt_generation): Call vinfo_for_stmt and pass
6727 stmt_vec_infos to subroutines.
6728 (vect_remove_stores): Use stmt_vec_infos to handle the statement
6731 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6733 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences):
6734 (vect_check_gather_scatter, vect_create_data_ref_ptr, bump_vector_ptr)
6735 (vect_permute_store_chain, vect_setup_realignment)
6736 (vect_permute_load_chain, vect_shift_permute_load_chain)
6737 (vect_transform_grouped_load): Use stmt_vec_info rather than gimple
6738 stmts internally, and when passing values to other vectorizer routines.
6739 * tree-vect-loop-manip.c (vect_can_advance_ivs_p): Likewise.
6740 * tree-vect-loop.c (vect_analyze_scalar_cycles_1)
6741 (vect_analyze_loop_operations, get_initial_def_for_reduction)
6742 (vect_create_epilog_for_reduction, vectorize_fold_left_reduction)
6743 (vectorizable_reduction, vectorizable_induction)
6744 (vectorizable_live_operation, vect_transform_loop_stmt)
6745 (vect_transform_loop): Likewise.
6746 * tree-vect-patterns.c (vect_reassociating_reduction_p)
6747 (vect_recog_widen_op_pattern, vect_recog_mixed_size_cond_pattern)
6748 (vect_recog_bool_pattern, vect_recog_gather_scatter_pattern): Likewise.
6749 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
6750 (vect_slp_analyze_node_operations_1): Likewise.
6751 * tree-vect-stmts.c (vect_mark_relevant, process_use)
6752 (exist_non_indexing_operands_for_use_p, vect_init_vector_1)
6753 (vect_mark_stmts_to_be_vectorized, vect_get_vec_def_for_operand)
6754 (vect_finish_stmt_generation_1, get_group_load_store_type)
6755 (get_load_store_type, vect_build_gather_load_calls)
6756 (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
6757 (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
6758 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
6759 (vectorizable_store, vectorizable_load, vectorizable_condition)
6760 (vectorizable_comparison, vect_analyze_stmt, vect_transform_stmt)
6761 (supportable_widening_operation): Likewise.
6762 (vect_get_vector_types_for_stmt): Likewise.
6763 * tree-vectorizer.h (vect_dr_behavior): Likewise.
6765 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6767 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
6768 (vect_slp_analyze_node_dependences, vect_analyze_data_ref_accesses)
6769 (vect_permute_store_chain, vect_permute_load_chain)
6770 (vect_shift_permute_load_chain, vect_transform_grouped_load): Avoid
6771 repeated stmt_vec_info lookups.
6772 * tree-vect-loop-manip.c (vect_can_advance_ivs_p): Likewise.
6773 (vect_update_ivs_after_vectorizer): Likewise.
6774 * tree-vect-loop.c (vect_is_simple_reduction): Likewise.
6775 (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
6776 * tree-vect-patterns.c (adjust_bool_stmts): Likewise.
6777 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
6778 (vect_bb_slp_scalar_cost): Likewise.
6779 * tree-vect-stmts.c (get_group_alias_ptr_type): Likewise.
6781 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6783 * tree-vect-data-refs.c (vect_check_gather_scatter): Pass the
6784 gcall rather than the generic gimple stmt to gimple_call_internal_fn.
6785 (vect_get_smallest_scalar_type, can_group_stmts_p): Use dyn_cast
6786 to get gassigns and gcalls, rather than operating on generc gimple
6788 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p)
6789 (vect_mark_stmts_to_be_vectorized, vectorizable_store)
6790 (vectorizable_load, vect_analyze_stmt): Likewise.
6791 * tree-vect-loop.c (vectorizable_reduction): Likewise gphi.
6793 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6795 * tree-vectorizer.h (get_earlier_stmt, get_later_stmt): Take and
6796 return stmt_vec_infos rather than gimple stmts. Do not accept
6798 (vect_find_last_scalar_stmt_in_slp): Return a stmt_vec_info instead
6800 * tree-vect-slp.c (vect_find_last_scalar_stmt_in_slp): Likewise.
6801 Update use of get_later_stmt.
6802 (vect_get_constant_vectors): Update call accordingly.
6803 (vect_schedule_slp_instance): Likewise
6804 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Likewise.
6805 (vect_slp_analyze_instance_dependence): Likewise.
6806 (vect_preserves_scalar_order_p): Update use of get_earlier_stmt.
6808 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6810 * tree-vectorizer.h (stmt_info_for_cost::stmt): Replace with...
6811 (stmt_info_for_cost::stmt_info): ...this new field.
6812 (add_stmt_costs): Update accordingly.
6813 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost)
6814 (vect_get_known_peeling_cost): Likewise.
6815 (vect_estimate_min_profitable_iters): Likewise.
6816 * tree-vect-stmts.c (record_stmt_cost): Likewise.
6818 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6820 * tree-vectorizer.h (_loop_vec_info::may_misalign_stmts): Change
6821 from an auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
6822 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Update
6824 * tree-vect-loop-manip.c (vect_create_cond_for_align_checks): Likewise.
6826 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6828 * tree-vectorizer.h (_stmt_vec_info::same_dr_stmt): Change from
6829 a gimple stmt to a stmt_vec_info.
6830 * tree-vect-stmts.c (vectorizable_load): Update accordingly.
6832 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6834 * tree-vectorizer.h (vec_info::grouped_stores): Change from
6835 an auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
6836 (_loop_vec_info::reduction_chains): Likewise.
6837 * tree-vect-loop.c (vect_fixup_scalar_cycles_with_patterns): Update
6839 * tree-vect-slp.c (vect_analyze_slp): Likewise.
6841 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6843 * tree-vectorizer.h (_stmt_vec_info::first_element): Change from
6844 a gimple stmt to a stmt_vec_info.
6845 (_stmt_vec_info::next_element): Likewise.
6846 * tree-vect-data-refs.c (vect_update_misalignment_for_peel)
6847 (vect_slp_analyze_and_verify_node_alignment)
6848 (vect_analyze_group_access_1, vect_analyze_group_access)
6849 (vect_small_gap_p, vect_prune_runtime_alias_test_list)
6850 (vect_create_data_ref_ptr, vect_record_grouped_load_vectors)
6851 (vect_supportable_dr_alignment): Update accordingly.
6852 * tree-vect-loop.c (vect_fixup_reduc_chain): Likewise.
6853 (vect_fixup_scalar_cycles_with_patterns, vect_is_slp_reduction)
6854 (vect_is_simple_reduction, vectorizable_reduction): Likewise.
6855 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
6856 * tree-vect-slp.c (vect_build_slp_tree_1)
6857 (vect_attempt_slp_rearrange_stmts, vect_supported_load_permutation_p)
6858 (vect_split_slp_store_group, vect_analyze_slp_instance)
6859 (vect_analyze_slp, vect_transform_slp_perm_load): Likewise.
6860 * tree-vect-stmts.c (vect_model_store_cost, vect_model_load_cost)
6861 (get_group_load_store_type, get_load_store_type)
6862 (get_group_alias_ptr_type, vectorizable_store, vectorizable_load)
6863 (vect_transform_stmt, vect_remove_stores): Likewise.
6865 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6867 * tree-vectorizer.h (vect_dr_stmt): Return a stmt_vec_info rather
6869 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
6870 (vect_slp_analyze_data_ref_dependence, vect_record_base_alignments)
6871 (vect_calculate_target_alignmentm, vect_compute_data_ref_alignment)
6872 (vect_update_misalignment_for_peel, vect_verify_datarefs_alignment)
6873 (vector_alignment_reachable_p, vect_get_data_access_cost)
6874 (vect_get_peeling_costs_all_drs, vect_peeling_hash_get_lowest_cost)
6875 (vect_peeling_supportable, vect_enhance_data_refs_alignment)
6876 (vect_find_same_alignment_drs, vect_analyze_data_refs_alignment)
6877 (vect_analyze_group_access_1, vect_analyze_group_access)
6878 (vect_analyze_data_ref_access, vect_analyze_data_ref_accesses)
6879 (vect_vfa_access_size, vect_small_gap_p, vect_analyze_data_refs)
6880 (vect_supportable_dr_alignment): Remove vinfo_for_stmt from the
6881 result of vect_dr_stmt and use the stmt_vec_info instead of
6882 the associated gimple stmt.
6883 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
6884 (vect_gen_prolog_loop_niters): Likewise.
6885 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
6887 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6889 * tree-vectorizer.h (_slp_tree::stmts): Change from a vec<gimple *>
6890 to a vec<stmt_vec_info>.
6891 * tree-vect-slp.c (vect_free_slp_tree): Update accordingly.
6892 (vect_create_new_slp_node): Take a vec<gimple *> instead of a
6894 (_slp_oprnd_info::def_stmts): Change from a vec<gimple *>
6895 to a vec<stmt_vec_info>.
6896 (bst_traits::value_type, bst_traits::value_type): Likewise.
6897 (bst_traits::hash): Update accordingly.
6898 (vect_get_and_check_slp_defs): Change the stmts parameter from
6899 a vec<gimple *> to a vec<stmt_vec_info>.
6900 (vect_two_operations_perm_ok_p, vect_build_slp_tree_1): Likewise.
6901 (vect_build_slp_tree): Likewise.
6902 (vect_build_slp_tree_2): Likewise. Update uses of
6903 SLP_TREE_SCALAR_STMTS.
6904 (vect_print_slp_tree): Update uses of SLP_TREE_SCALAR_STMTS.
6905 (vect_mark_slp_stmts, vect_mark_slp_stmts_relevant)
6906 (vect_slp_rearrange_stmts, vect_attempt_slp_rearrange_stmts)
6907 (vect_supported_load_permutation_p, vect_find_last_scalar_stmt_in_slp)
6908 (vect_detect_hybrid_slp_stmts, vect_slp_analyze_node_operations_1)
6909 (vect_slp_analyze_node_operations, vect_slp_analyze_operations)
6910 (vect_bb_slp_scalar_cost, vect_slp_analyze_bb_1)
6911 (vect_get_constant_vectors, vect_get_slp_defs)
6912 (vect_transform_slp_perm_load, vect_schedule_slp_instance)
6913 (vect_remove_slp_scalar_calls, vect_schedule_slp): Likewise.
6914 (vect_analyze_slp_instance): Build up a vec of stmt_vec_infos
6915 instead of gimple stmts.
6916 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Change
6917 the stores parameter for a vec<gimple *> to a vec<stmt_vec_info>.
6918 (vect_slp_analyze_instance_dependence): Update uses of
6919 SLP_TREE_SCALAR_STMTS.
6920 (vect_slp_analyze_and_verify_node_alignment): Likewise.
6921 (vect_slp_analyze_and_verify_instance_alignment): Likewise.
6922 * tree-vect-loop.c (neutral_op_for_slp_reduction): Likewise.
6923 (get_initial_defs_for_reduction): Likewise.
6924 (vect_create_epilog_for_reduction): Likewise.
6925 (vectorize_fold_left_reduction): Likewise.
6926 * tree-vect-stmts.c (vect_prologue_cost_for_slp_op): Likewise.
6927 (vect_model_simple_cost, vectorizable_shift, vectorizable_load)
6928 (can_vectorize_live_stmts): Likewise.
6930 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6932 * tree-vectorizer.h (_loop_vec_info::reductions): Change from an
6933 auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
6934 (vect_force_simple_reduction): Take and return stmt_vec_infos rather
6936 * tree-parloops.c (valid_reduction_p): Take a stmt_vec_info instead
6938 (gather_scalar_reductions): Update after above interface changes.
6939 * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Likewise.
6940 (vect_is_simple_reduction): Take and return stmt_vec_infos rather
6942 (vect_force_simple_reduction): Likewise.
6943 * tree-vect-patterns.c (vect_pattern_recog_1): Update use of
6944 LOOP_VINFO_REDUCTIONS.
6945 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
6947 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6949 * tree-vectorizer.h (_stmt_vec_info::reduc_def): Change from
6950 a gimple stmt to a stmt_vec_info.
6951 * tree-vect-loop.c (vect_active_double_reduction_p)
6952 (vect_force_simple_reduction, vectorizable_reduction): Update
6955 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6957 * tree-vectorizer.h (_slp_tree::vec_stmts): Change from a
6958 vec<gimple *> to a vec<stmt_vec_info>.
6959 * tree-vect-loop.c (vect_create_epilog_for_reduction): Change
6960 the reduction_phis argument from a vec<gimple *> to a
6962 (vectorizable_reduction): Likewise the phis local variable that
6963 is passed to vect_create_epilog_for_reduction. Update for new type
6964 of SLP_TREE_VEC_STMTS.
6965 (vectorizable_induction): Update for new type of SLP_TREE_VEC_STMTS.
6966 (vectorizable_live_operation): Likewise.
6967 * tree-vect-slp.c (vect_get_slp_vect_defs): Likewise.
6968 (vect_transform_slp_perm_load, vect_schedule_slp_instance): Likewise.
6970 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6972 * tree-vectorizer.h (_stmt_vec_info::vectorized_stmt): Change from
6973 a gimple stmt to a stmt_vec_info.
6974 (vectorizable_condition, vectorizable_live_operation)
6975 (vectorizable_reduction, vectorizable_induction): Pass back the
6976 vectorized statement as a stmt_vec_info.
6977 * tree-vect-data-refs.c (vect_record_grouped_load_vectors): Update
6978 use of STMT_VINFO_VEC_STMT.
6979 * tree-vect-loop.c (vect_create_epilog_for_reduction): Likewise,
6980 accumulating the inner phis that feed the STMT_VINFO_VEC_STMT
6981 as stmt_vec_infos rather than gimple stmts.
6982 (vectorize_fold_left_reduction): Change vec_stmt from a gimple stmt
6984 (vectorizable_live_operation): Likewise.
6985 (vectorizable_reduction, vectorizable_induction): Likewise,
6986 updating use of STMT_VINFO_VEC_STMT.
6987 * tree-vect-stmts.c (vect_get_vec_def_for_operand_1): Update use
6988 of STMT_VINFO_VEC_STMT.
6989 (vect_build_gather_load_calls, vectorizable_bswap, vectorizable_call)
6990 (vectorizable_simd_clone_call, vectorizable_conversion)
6991 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
6992 (vectorizable_store, vectorizable_load, vectorizable_condition)
6993 (vectorizable_comparison, can_vectorize_live_stmts): Change vec_stmt
6994 from a gimple stmt to a stmt_vec_info.
6995 (vect_transform_stmt): Update use of STMT_VINFO_VEC_STMT. Pass a
6996 pointer to a stmt_vec_info to the vectorizable_* routines.
6998 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
7000 * tree-vectorizer.h (_stmt_vec_info::related_stmt): Change from
7001 a gimple stmt to a stmt_vec_info.
7002 (is_pattern_stmt_p): Update accordingly.
7003 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Likewise.
7004 (vect_record_grouped_load_vectors): Likewise.
7005 * tree-vect-loop.c (vect_determine_vf_for_stmt): Likewise.
7006 (vect_fixup_reduc_chain, vect_update_vf_for_slp): Likewise.
7007 (vect_model_reduction_cost): Likewise.
7008 (vect_create_epilog_for_reduction): Likewise.
7009 (vectorizable_reduction, vectorizable_induction): Likewise.
7010 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
7011 Return the stmt_vec_info for the pattern statement.
7012 (vect_set_pattern_stmt): Update use of STMT_VINFO_RELATED_STMT.
7013 (vect_split_statement, vect_mark_pattern_stmts): Likewise.
7014 * tree-vect-slp.c (vect_detect_hybrid_slp_stmts): Likewise.
7015 (vect_detect_hybrid_slp, vect_get_slp_defs): Likewise.
7016 * tree-vect-stmts.c (vect_mark_relevant): Likewise.
7017 (vect_get_vec_def_for_operand_1, vectorizable_call): Likewise.
7018 (vectorizable_simd_clone_call, vect_analyze_stmt, new_stmt_vec_info)
7019 (free_stmt_vec_info, vect_is_simple_use): Likewise.
7021 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
7023 * tree-vectorizer.h (vect_finish_replace_stmt): Return a stmt_vec_info
7024 (vect_finish_stmt_generation): Likewise.
7025 * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
7026 (vect_finish_replace_stmt, vect_finish_stmt_generation): Likewise.
7027 (vect_build_gather_load_calls): Use the return value of the above
7028 functions instead of a separate call to vinfo_for_stmt. Use narrow
7029 scopes for the input gimple stmt and wider scopes for the associated
7030 stmt_vec_info. Use vec_info::lookup_def when setting these
7031 stmt_vec_infos from an SSA_NAME definition.
7032 (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
7033 (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
7034 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
7035 (vectorizable_store, vectorizable_load, vectorizable_condition)
7036 (vectorizable_comparison): Likewise.
7037 * tree-vect-loop.c (vectorize_fold_left_reduction): Likewise.
7038 (vectorizable_reduction): Likewise.
7040 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
7042 * tree-vectorizer.h (vect_is_simple_use): Add an optional
7043 stmt_vec_info * parameter before the optional gimple **.
7044 * tree-vect-stmts.c (vect_is_simple_use): Likewise.
7045 (process_use, vect_get_vec_def_for_operand_1): Update callers.
7046 (vect_get_vec_def_for_operand, vectorizable_shift): Likewise.
7047 * tree-vect-loop.c (vectorizable_reduction): Likewise.
7048 (vectorizable_live_operation): Likewise.
7049 * tree-vect-patterns.c (type_conversion_p): Likewise.
7050 (vect_look_through_possible_promotion): Likewise.
7051 (vect_recog_rotate_pattern): Likewise.
7052 * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
7054 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
7056 * tree-vectorizer.h (stmt_vec_info): Temporarily change from
7057 a typedef to a wrapper class.
7058 (NULL_STMT_VEC_INFO): New macro.
7059 (vec_info::stmt_infos): Change to vec<stmt_vec_info>.
7060 (stmt_vec_info::operator*): New function.
7061 (stmt_vec_info::operator gimple *): Likewise.
7062 (set_vinfo_for_stmt): Use NULL_STMT_VEC_INFO.
7063 (add_stmt_costs): Likewise.
7064 * tree-vect-loop-manip.c (iv_phi_p): Likewise.
7065 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost)
7066 (vect_get_known_peeling_cost): Likewise.
7067 (vect_estimate_min_profitable_iters): Likewise.
7068 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
7069 * tree-vect-slp.c (vect_remove_slp_scalar_calls): Likewise.
7070 * tree-vect-stmts.c (vect_build_gather_load_calls): Likewise.
7071 (vectorizable_store, free_stmt_vec_infos): Likewise.
7072 (new_stmt_vec_info): Change return type of xcalloc to
7075 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
7077 * tree-vectorizer.h (vec_info::lookup_single_use): Declare.
7078 * tree-vectorizer.c (vec_info::lookup_single_use): New function.
7079 * tree-vect-loop.c (vectorizable_reduction): Use it instead of
7080 a single_imm_use-based sequence.
7081 * tree-vect-stmts.c (supportable_widening_operation): Likewise.
7083 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
7085 * tree-vectorizer.h (vec_info::lookup_def): Declare.
7086 * tree-vectorizer.c (vec_info::lookup_def): New function.
7087 * tree-vect-patterns.c (vect_get_internal_def): Use it.
7088 (vect_widened_op_tree): Likewise.
7089 * tree-vect-stmts.c (vect_is_simple_use): Likewise.
7090 * tree-vect-loop.c (vect_analyze_loop_operations): Likewise.
7091 (vectorizable_reduction): Likewise.
7092 (vect_valid_reduction_input_p): Take a stmt_vec_info instead
7094 (vect_is_slp_reduction): Update calls accordingly. Use
7095 vec_info::lookup_def.
7096 (vect_is_simple_reduction): Likewise
7097 * tree-vect-slp.c (vect_detect_hybrid_slp_1): Use vec_info::lookup_def.
7099 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
7101 * tree-vectorizer.h (vec_info::lookup_stmt): Declare.
7102 * tree-vectorizer.c (vec_info::lookup_stmt): New function.
7103 * tree-vect-loop.c (vect_determine_vf_for_stmt): Use it instead
7105 (vect_determine_vectorization_factor, vect_analyze_scalar_cycles_1)
7106 (vect_compute_single_scalar_iteration_cost, vect_analyze_loop_form)
7107 (vect_update_vf_for_slp, vect_analyze_loop_operations)
7108 (vect_is_slp_reduction, vectorizable_induction)
7109 (vect_transform_loop_stmt, vect_transform_loop): Likewise.
7110 * tree-vect-patterns.c (vect_init_pattern_stmt):
7111 (vect_determine_min_output_precision_1, vect_determine_precisions)
7112 (vect_pattern_recog): Likewise.
7113 * tree-vect-stmts.c (vect_analyze_stmt, vect_transform_stmt): Likewise.
7114 * config/powerpcspe/powerpcspe.c (rs6000_density_test): Likewise.
7115 * config/rs6000/rs6000.c (rs6000_density_test): Likewise.
7116 * tree-vect-slp.c (vect_detect_hybrid_slp_stmts): Likewise.
7117 (vect_detect_hybrid_slp_1, vect_detect_hybrid_slp_2)
7118 (vect_detect_hybrid_slp): Likewise. Change the walk_stmt_info
7119 info field from a loop to a loop_vec_info.
7121 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
7123 * tree-vectorizer.h (stmt_vec_info): Move typedef earlier in file.
7124 (vec_info::add_stmt): Declare.
7125 * tree-vectorizer.c (vec_info::add_stmt): New function.
7126 * tree-vect-data-refs.c (vect_create_data_ref_ptr): Use it.
7127 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Likewise.
7128 (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
7129 (vectorizable_induction): Likewise.
7130 * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Likewise.
7131 * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
7132 (vectorizable_simd_clone_call, vectorizable_store): Likewise.
7133 (vectorizable_load): Likewise.
7134 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
7135 (vect_recog_bool_pattern, vect_recog_mask_conversion_pattern)
7136 (vect_recog_gather_scatter_pattern): Likewise.
7137 (append_pattern_def_seq): Likewise. Remove a check that is
7138 performed by add_stmt itself.
7140 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
7142 * tree-vect-loop.c (vectorizable_reduction): Fix an instance in
7143 which make_ssa_name was called with new_stmt before new_stmt
7146 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
7148 * tree-vect-loop.c (vect_valid_reduction_input_p): New function,
7150 (vect_is_slp_reduction): ...here...
7151 (vect_is_simple_reduction): ...and here. Remove repetition of tests
7152 that are already known to be false.
7154 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
7156 * tree-vectorizer.h (vect_free_slp_instance): Add a final_p parameter.
7157 * tree-vect-slp.c (vect_free_slp_tree): Likewise. Don't update
7158 STMT_VINFO_NUM_SLP_USES when it's true.
7159 (vect_free_slp_instance): Add a final_p parameter and pass it to
7161 (vect_build_slp_tree_2): Update call to vect_free_slp_instance.
7162 (vect_analyze_slp_instance): Likewise.
7163 (vect_slp_analyze_operations): Likewise.
7164 (vect_slp_analyze_bb_1): Likewise.
7165 * tree-vectorizer.c (vec_info): Likewise.
7166 * tree-vect-loop.c (vect_transform_loop): Likewise.
7168 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
7170 * tree-vect-loop.c (vectorizable_reduction): Assert that the
7171 function is not called for second and subsequent members of
7174 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
7176 * tree-vect-loop.c (get_initial_def_for_reduction): Move special
7177 cases for nested loops from here to ...
7178 (vect_create_epilog_for_reduction): ...here. Only call
7179 vect_is_simple_use for inner-loop reductions.
7181 2018-07-31 Martin Liska <mliska@suse.cz>
7183 PR gcov-profile/85338
7184 PR gcov-profile/85350
7185 PR gcov-profile/85372
7186 * profile.c (struct location_triplet): New.
7187 (struct location_triplet_hash): Likewise.
7188 (output_location): Do not output a BB that
7189 is already recorded for a line.
7190 (branch_prob): Use streamed_locations.
7192 2018-07-31 Martin Liska <mliska@suse.cz>
7194 PR gcov-profile/85370
7195 * coverage.c (coverage_begin_function): Do not mark target
7196 clones as artificial functions.
7198 2018-07-31 Martin Liska <mliska@suse.cz>
7200 PR gcov-profile/83813
7201 PR gcov-profile/84758
7202 PR gcov-profile/85217
7203 PR gcov-profile/85332
7204 * profile.c (branch_prob): Do not record GOTO expressions
7205 for GIMPLE statements which locations are already streamed.
7207 2018-07-31 Olivier Hainque <hainque@adacore.com>
7209 * gcc.c (handle_spec_function): Accept a soft_matched_part
7210 argument, as do_spec_1. Pass it down to ...
7211 (eval_spec_function): Accept a soft_matched_part argument,
7212 and pass it down to ...
7213 (do_spec_2): Accept a soft_matched_part argument, and pass
7214 it down to do_spec_1.
7215 (do_spec_1): Pass soft_matched_part to handle_spec_function.
7216 (handle_braces): Update call to handle_spec_function.
7217 (driver::set_up_specs): Update calls to do_spec_2.
7218 (compare_debug_dump_opt_spec_function): Likewise.
7219 (compare_debug_self_opt_spec_function): Likewise.
7221 2018-07-31 Olivier Hainque <hainque@adacore.com>
7223 * common.opt (nolibc): New option.
7224 * doc/invoke.texi (Link Options): Document it.
7225 * gcc.c (LINK_GCC_C_SEQUENCE_SPEC): Honor nolibc.
7226 * config/alpha/linux.h: Likewise.
7227 * config/arc/elf.h: Likewise.
7228 * config/arm/uclinux-elf.h: Likewise.
7229 * config/arm/unknown-elf.h: Likewise.
7230 * config/avr/avrlibc.h: Likewise.
7231 * config/bfin/bfin.h: Likewise.
7232 * config/bfin/linux.h: Likewise.
7233 * config/bfin/uclinux.h: Likewise.
7234 * config/darwin.h: Likewise.
7235 * config/darwin10.h: Likewise.
7236 * config/darwin12.h: Likewise.
7237 * config/gnu-user.h: Likewise.
7238 * config/lm32/uclinux-elf.h: Likewise.
7239 * config/pa/pa-hpux11.h: Likewise.
7240 * config/pa/pa64-hpux.h: Likewise.
7241 * config/sparc/sparc.h: Likewise.
7243 2018-07-31 Olivier Hainque <hainque@adacore.com>
7245 * gcc.c (getenv_spec_function): Prepend '/' to value for allowed
7246 undefined variables.
7248 2018-07-30 Segher Boessenkool <segher@kernel.crashing.org>
7251 * config/arm/arm.c (arm_block_set_aligned_vect): Use gen_int_mode
7254 2018-07-30 Bernd Edlinger <bernd.edlinger@hotmail.de>
7256 * tree-ssa-forwprop.c (simplify_builtin_call): Don't create a not NUL
7257 terminated string literal.
7259 2018-07-30 Segher Boessenkool <segher@kernel.crashing.org>
7261 PR rtl-optimization/85160
7262 * combine.c (is_just_move): New function.
7263 (try_combine): Allow combining two instructions into two if neither of
7264 the original instructions was a move.
7266 2018-07-30 Alexander Monakov <amonakov@ispras.ru>
7269 * doc/extend.texi (Global Register Variables): Discourage use of type
7271 (Local Register Variables): Likewise.
7273 2018-07-30 Richard Sandiford <richard.sandiford@arm.com>
7275 PR tree-optimization/86506
7276 * hwint.h (ceil_log2): Resync with hwint.c implementation.
7278 2018-07-30 Ilya Leoshkevich <iii@linux.ibm.com>
7281 * lra-constraints.c (spill_hard_reg_in_range): When selecting the
7282 hard_regno, make sure no insn between `from` and `to` clobbers it.
7284 2018-07-30 Cesar Philippidis <cesar@codesourcery.com>
7285 Tom de Vries <tdevries@suse.de>
7287 * config/nvptx/nvptx.c (PTX_GANG_DEFAULT): Rename to ...
7288 (PTX_DEFAULT_RUNTIME_DIM): ... this.
7289 (nvptx_goacc_validate_dims): Set default worker and gang dims to
7290 PTX_DEFAULT_RUNTIME_DIM.
7291 (nvptx_dim_limit): Ignore GOMP_DIM_WORKER.
7293 2018-07-29 John David Anglin <danglin@gcc.gnu.org>
7295 * config/pa/pa.c (pa_output_addr_vec): Align address table.
7296 * config/pa/pa.h (JUMP_TABLES_IN_TEXT_SECTION): Revise comment.
7297 * config/pa/pa32-linux.h (JUMP_TABLES_IN_TEXT_SECTION): Define.
7299 2018-07-27 Michael Meissner <meissner@linux.ibm.com>
7301 * config/rs6000/constraints.md (wG constraint): Delete, no longer
7303 * config/rs6000/predicates.md (p9_fusion_reg_operand): Rename
7304 predicate to reflect toc fusion has been deleted.
7305 (toc_fusion_mem_raw): Delete, no longer used.
7306 (toc_fusion_mem_wrapped): Likewise.
7307 * config/rs6000/rs6000-cpus.def (POWERPC_MASKS): Delete toc
7309 * config/rs6000/rs6000-protos.h (fusion_wrap_memory_address):
7310 Delete, no longer used.
7311 * config/rs6000/rs6000.c (struct rs6000_reg_addr): Delete fields
7312 meant to be used for toc fusion.
7313 (rs6000_debug_print_mode): Delete toc fusion debugging.
7314 (rs6000_debug_reg_global): Likewise.
7315 (rs6000_init_hard_regno_mode_ok): Delete setting up fields for toc
7316 fusion and secondary reload support that were never used.
7317 (rs6000_option_override_internal): Delete TOC fusion, that was only
7318 partially defined, and it did not work unless you also used the
7320 (rs6000_legitimate_address_p): Delete TOC fusion support.
7321 (rs6000_opt_masks): Likewise.
7322 (fusion_wrap_memory_address): Delete function, no longer used.
7323 (fusion_split_address); Delete TOC fusion support.
7324 * config/rs6000/rs6000.h (TARGET_TOC_FUSION_INT): Delete, no
7325 longer used with toc fusion being deleted.
7326 (TARGET_TOC_FUSION_FP): Likewise.
7327 * config/rs6000/rs6000.md (UNSPEC_FUSION_ADDIS): Delete TOC fusion
7329 (toc fusion spliter): Delete TOC fusion support.
7330 (toc_fusionload_<mode>): Likewise.
7331 (toc_fusionload_di): Likewise.
7332 (fusion_gpr_load_<mode>): Delete generator function, this insn no
7333 longer needs to be named. Rename predicate to delete TOC fusion.
7334 (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_load): Likewise.
7335 (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_store): Likewise.
7336 (fusion_vsx_<P:mode>_<GPR_FUSION:mode>_load): Likewise.
7337 (fusion_vsx_<P:mode>_<GPR_FUSION:mode>_store): Likewise.
7338 (p9 fusion peephole2s): Rename predicate to delete TOC fusion.
7340 2018-07-27 Kelvin Nilsen <kelvin@gcc.gnu.org>
7342 * doc/extend.texi (Basic PowerPC Built-in Functions Available on
7343 ISA 2.05): Replace __uint128_t with __uint128 and __int128_t with
7344 __int128 in built-in function prototypes.
7345 (PowerPC AltiVec Built-in Functions on ISA 2.07): Likewise.
7346 (PowerPC AltiVec Built-in Functions on ISA 3.0): Likewise.
7348 2018-07-27 Martin Sebor <msebor@redhat.com>
7350 PR tree-optimization/86696
7351 * tree-ssa-strlen.c (get_min_string_length): Handle all integer
7352 types, including enums.
7353 (handle_char_store): Be prepared for the above function to fail.
7355 2018-07-26 Qing Zhao <qing.zhao@oracle.com>
7357 * builtins.c (inline_expand_builtin_string_cmp): Disable inlining
7358 when optimization level is lower than 2 or optimize for size.
7360 2018-07-26 Martin Sebor <msebor@redhat.com>
7362 PR tree-optimization/86043
7363 PR tree-optimization/86042
7364 * tree-ssa-strlen.c (handle_builtin_memcpy): Handle strict overlaps.
7365 (get_string_cst_length): Rename...
7366 (get_min_string_length): ...to this. Add argument.
7367 (handle_char_store): Extend to handle multi-character stores by
7369 * tree.c (initializer_zerop): Use new argument. Handle MEM_REF.
7370 * tree.h (initializer_zerop): Add argument.
7372 2018-07-26 Jakub Jelinek <jakub@redhat.com>
7375 * omp-low.c (scan_sharing_clauses): Don't ignore map clauses for
7376 declare target to variables if they have always,{to,from,tofrom} map
7379 2018-07-26 Martin Liska <mliska@suse.cz>
7382 * lto-wrapper.c: Add linker_output as prefix
7383 for ltrans_output_file.
7385 2018-07-26 Segher Boessenkool <segher@kernel.crashing.org>
7387 PR rtl-optimization/85805
7388 * combine.c (reg_nonzero_bits_for_combine): Only use the last set
7389 value for hard registers if that was written in the same mode.
7391 2018-07-26 Martin Liska <mliska@suse.cz>
7393 PR gcov-profile/86536
7394 * gcov.c (format_gcov): Use printf format %.*f directly
7395 and do not handle special values.
7397 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
7399 * common/config/arc/arc-common.c (arc_option_optimization_table):
7400 Update default optimizations for size.
7402 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
7404 * config/arc/arc.md (movsf_insn): Add short instruction selection.
7405 * config/arc/constraints.md (CfZ): New constraint.
7406 * config/arc/fpu.md (addssf3_fpu): Use CfZ constraint.
7407 (subsf3_fpu): Likewise.
7408 (cmpsf_fpu): Likewise.
7409 (cmpsf_fpu_uneq): Likewise.
7411 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
7413 * config/arc/arc.c (compact_memory_operand_p): Check for uncached
7415 (arc_is_uncached_mem_p): uncached applies to both the variable and
7418 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
7420 * config/arc/arc.h (ADDITIONAL_REGISTER_NAMES): Add additional
7423 2018-07-25 David Malcolm <dmalcolm@redhat.com>
7425 * optinfo-emit-json.cc (class optrecord_json_writer): Convert
7426 field "m_scopes" from vec to auto_vec.
7428 2018-07-25 Martin Liska <mliska@suse.cz>
7430 * config/powerpcspe/powerpcspe-protos.h (rs6000_loop_align): Fix
7433 2018-07-25 Richard Biener <rguenther@suse.de>
7436 * dwarf2out.c (dwarf2out_decl): Do not handle nested functions
7437 special wrt context_die late.
7438 (gen_subprogram_die): Re-use DIEs in local scope.
7440 2018-07-25 Richard Sandiford <richard.sandiford@arm.com>
7442 PR tree-optimization/86644
7443 * hwint.c (ceil_log2): Fix comment. Return 0 for 0.
7445 2018-07-25 Martin Liska <mliska@suse.cz>
7448 * dumpfile.c: And excluded values with TDF_ALL_VALUES.
7449 * dumpfile.h (enum dump_flag): Defince TDF_ALL_VALUES.
7451 2018-07-25 Martin Liska <mliska@suse.cz>
7454 * params.def: Explain ASan abbreviation and provide
7455 a documentation link.
7457 2018-07-24 Martin Sebor <msebor@redhat.com>
7459 PR tree-optimization/86622
7460 PR tree-optimization/86532
7461 * builtins.h (string_length): Declare.
7462 * builtins.c (c_strlen): Correct handling of non-constant offsets.
7463 (check_access): Be prepared for non-constant length ranges.
7464 (string_length): Make extern.
7465 * expr.c (string_constant): Only handle the minor non-constant
7466 array index. Use string_constant to compute the length of
7467 a generic string constant.
7469 2018-07-24 Richard Sandiford <richard.sandiford@arm.com>
7471 PR tree-optimization/86618
7472 * tree-vect-stmts.c (vectorizable_call): Don't take the address
7473 of LOOP_VINFO_MASKS (loop_vinfo) when loop_vinfo is null.
7475 2018-07-24 David Malcolm <dmalcolm@redhat.com>
7477 PR tree-optimization/86636
7478 * json.cc (json::object::set): Fix comment. Add assertions.
7479 (json::array::append): Move here from json.h. Add comment and an
7481 (json::string::string): Likewise.
7482 * json.h (json::array::append): Move to json.cc.
7483 (json::string::string): Likewise.
7484 * optinfo-emit-json.cc
7485 (optrecord_json_writer::impl_location_to_json): Assert that we
7486 aren't attempting to write out UNKNOWN_LOCATION, or an ad-hoc
7487 wrapper around it. Expand the location once, rather than three
7489 (optrecord_json_writer::inlining_chain_to_json): Fix the check for
7490 UNKNOWN_LOCATION, to use LOCATION_LOCUS to look through ad-hoc
7492 (optrecord_json_writer::optinfo_to_json): Likewise, in four
7493 places. Fix some overlong lines.
7495 2018-07-24 Matthew Malcomson <matthew.malcomson@arm.com>
7497 * config/aarch64/aarch64-simd.md
7498 (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w<mode>): Split into...
7499 (aarch64_<ANY_EXTEND:su>subw<mode>): ... This...
7500 (aarch64_<ANY_EXTEND:su>addw<mode>): ... And this.
7501 (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w<mode>_internal): Split into...
7502 (aarch64_<ANY_EXTEND:su>subw<mode>_internal): ... This...
7503 (aarch64_<ANY_EXTEND:su>addw<mode>_internal): ... And this.
7504 (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w2<mode>_internal): Split into...
7505 (aarch64_<ANY_EXTEND:su>subw2<mode>_internal): ... This...
7506 (aarch64_<ANY_EXTEND:su>addw2<mode>_internal): ... And this.
7508 2018-07-24 Jakub Jelinek <jakub@redhat.com>
7511 * expmed.c (expand_divmod): Punt if d == HOST_WIDE_INT_MIN
7512 and size > HOST_BITS_PER_WIDE_INT. For size > HOST_BITS_PER_WIDE_INT
7513 and abs_d == d, do the power of two handling if profitable.
7515 2018-07-24 Richard Biener <rguenther@suse.de>
7517 * match.pd: Add BIT_FIELD_REF canonicalizations.
7519 2018-07-23 Bernd Edlinger <bernd.edlinger@hotmail.de>
7522 * genmatch.c (dt_operand::gen_match_op): Avoid folding volatile values.
7524 2018-07-23 Bernd Edlinger <bernd.edlinger@hotmail.de>
7526 * gimple-fold.c (gimple_fold_builtin_printf): Don't create a not NUL
7527 terminated STRING_CST object.
7529 2018-07-23 Bernd Edlinger <bernd.edlinger@hotmail.de>
7531 hsa-dump.c (dump_hsa_symbol): Avoid out of scope access to buf.
7533 2018-07-23 Segher Boessenkool <segher@kernel.crashing.org>
7535 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Adjust.
7536 * config/rs6000/rs6000-protos.h (rs6000_split_v4si_init): Delete.
7537 * config/rs6000/rs6000.c (rs6000_expand_vector_init): Always force
7538 the elements into a register.
7539 (rs6000_split_v4si_init_di_reg): Delete.
7540 (rs6000_split_v4si_init): Delete.
7541 * config/rs6000/vsx.md (unspec): Delete UNSPEC_VSX_VEC_INIT.
7542 (vsx_init_v4si): Rewrite as a define_expand.
7544 2018-07-23 Segher Boessenkool <segher@kernel.crashing.org>
7546 * config/rs6000/rs6000.md (splitters for rldimi and rlwimi with the
7547 zero_extend argument from memory): New.
7549 2018-07-22 Martin Sebor <msebor@redhat.com>
7552 * gimple-ssa-warn-alloca.c (alloca_call_type_by_arg): Avoid
7553 diagnosing calls with unknown arguments unless -Walloca-larger-than
7554 is restricted to less than PTRDIFF_MAX bytes.
7556 2018-07-22 Gerald Pfeifer <gerald@pfeifer.com>
7558 * doc/gcov.texi (Invoking Gcov): Editorial changes.
7560 2018-07-20 David Malcolm <dmalcolm@redhat.com>
7562 * pretty-print.c (text_info::set_location): Remove redundant
7563 "line_table" parameter from call to rich_location::set_range.
7565 2018-07-20 Martin Sebor <msebor@redhat.com>
7568 * builtins.c (expand_builtin_alloca): Adjust.
7569 * calls.c (alloc_max_size): Simplify.
7570 * cgraphunit.c (cgraph_node::expand): Adjust.
7571 * common.opt (larger_than_size, warn_frame_larger_than): Remove
7573 (frame_larger_than_size): Same.
7574 (-Wframe-larger-than, -Wlarger-than, -Wstack-usage): Change options
7575 to take a HOST_WIDE_INT argument and accept a byte-size suffix.
7577 * doc/invoke.texi (GCC Command Options): Document option arguments.
7578 Explain byte-size arguments and suffixes.
7579 (-Wvla-larger-than, -Wno-alloc-size-larger-than): Update.
7580 (-Wno-alloca-larger-than, -Wno-vla-larger-than): Same.
7581 (-Wframe-larger-than, -Wlarger-than, -Wstack-usage): Same.
7582 * doc/options.texi (UInteger): Expand.
7583 (Host_Wide_Int, ByteSize): Document new properties.
7584 * final.c (final_start_function_1): Include sizes in an error message.
7585 * function.c (frame_offset_overflow): Same.
7586 * gimple-ssa-warn-alloca.c (pass_walloca::gate): Adjust.
7587 (alloca_call_type_by_arg): Change function argument to HOST_WIDE_INT.
7588 Diagnose unbounded alloca calls only for limits of less than
7590 (alloca_call_type): Adjust. Diagnose possibly out-of-bounds alloca
7591 calls and VLA size only for limits of less than PTRDIFF_MAX. Same
7593 (pass_walloca::execute): Adjust. Diagnose alloca calls in loops
7594 only for limits of less than PTRDIFF_MAX.
7595 * langhooks-def.h (lhd_handle_option): Change function argument
7597 * langhooks.c (lhd_handle_option): Same.
7598 * langhooks.h (handle_option): Same.
7599 * opt-functions.awk (switch_bit_fields): Handle Host_Wide_Int and
7601 (var_type, var_type_struct): Same.
7602 (var_set): Handle ByteSize flag.
7603 * optc-gen.awk: Add comments to output to ease debugging. Make
7604 use of HOST_WIDE_INT where appropriate.
7605 * opts-gen-save.awk: Use %lx to format unsigned long.
7606 * opth-gen.awk: Change function argument to HOST_WIDE_INT.
7607 * opts-common.c (integral_argument): Return HOST_WIDE_INT and add
7608 arguments. Parse bytes-size suffixes.
7609 (enum_arg_to_value): Change function argument to HOST_WIDE_INT.
7610 (enum_value_to_arg): Same.
7611 (decode_cmdline_option): Handle cl_host_wide_int. Adjust.
7612 (handle_option): Adjust.
7613 (generate_option): Change function argument to HOST_WIDE_INT.
7614 (cmdline_handle_error): Adjust.
7615 (read_cmdline_option): Change function argument to HOST_WIDE_INT.
7616 (set_option): Change function argument to HOST_WIDE_INT.
7617 (option_enabled): Handle cl_host_wide_int.
7618 (get_option_state): Handle CLVC_SIZE.
7619 (control_warning_option): Same.
7620 * opts.c (common_handle_option): Change function argument to
7621 HOST_WIDE_INT. Remove handling of OPT_Walloca_larger_than_ and
7622 OPT_Wvla_larger_than_.
7623 * opts.h (enum cl_var_type): Add an enumerator.
7624 * stor-layout.c (layout_decl): Print a more meaningful warning.
7625 * toplev.c (output_stack_usage): Adjust.
7627 2018-07-20 Qing Zhao <qing.zhao@oracle.com>
7629 * builtins.c (expand_builtin_memcmp): Delete the last parameter for
7630 call to inline_expand_builtin_string_cmp.
7631 (expand_builtin_strcmp): Likewise.
7632 (expand_builtin_strncmp): Likewise.
7633 (inline_string_cmp): Delete the last parameter, change char_type_node
7634 to unsigned_char_type_node for strcmp/strncmp, add conversions to the
7636 (inline_expand_builtin_string_cmp): Delete the last parameter, give up
7637 the inlining expansion on target where the type of the call has same or
7638 narrower precision than unsigned char.
7640 2018-07-20 David Malcolm <dmalcolm@redhat.com>
7642 * Makefile.in (OBJS): Add json.o and optinfo-emit-json.o.
7643 (CFLAGS-optinfo-emit-json.o): Define TARGET_NAME.
7644 * common.opt (fsave-optimization-record): New option.
7645 * coretypes.h (struct kv_pair): Move here from dumpfile.c.
7646 * doc/invoke.texi (-fsave-optimization-record): New option.
7647 * dumpfile.c: Include "optinfo-emit-json.h".
7648 (struct kv_pair): Move to coretypes.h.
7649 (optgroup_options): Make non-static.
7650 (dump_context::end_scope): Call
7651 optimization_records_maybe_pop_dump_scope.
7652 * dumpfile.h (optgroup_options): New decl.
7653 * json.cc: New file.
7655 * optinfo-emit-json.cc: New file.
7656 * optinfo-emit-json.h: New file.
7657 * optinfo.cc: Include "optinfo-emit-json.h".
7658 (optinfo::emit): Call optimization_records_maybe_record_optinfo.
7659 (optinfo_enabled_p): Check optimization_records_enabled_p.
7660 (optinfo_wants_inlining_info_p): Likewise.
7661 * optinfo.h: Update comment.
7662 * profile-count.c (profile_quality_as_string): New function.
7663 * profile-count.h (profile_quality_as_string): New decl.
7664 (profile_count::quality): New accessor.
7665 * selftest-run-tests.c (selftest::run_tests): Call json_cc_tests
7666 and optinfo_emit_json_cc_tests.
7667 * selftest.h (selftest::json_cc_tests): New decl.
7668 (selftest::optinfo_emit_json_cc_tests): New decl.
7669 * toplev.c: Include "optinfo-emit-json.h".
7670 (compile_file): Call optimization_records_finish.
7671 (do_compile): Call optimization_records_start.
7672 * tree-ssa-live.c: Include optinfo.h.
7673 (remove_unused_scope_block_p): Retain inlining information if
7674 optinfo_wants_inlining_info_p returns true.
7676 2018-07-20 Richard Biener <rguenther@suse.de>
7679 * dwarf2out.c (dwarf2out_die_ref_for_decl): Test in_lto_p
7680 to cover -flto-partition=none.
7682 2018-07-20 Martin Liska <mliska@suse.cz>
7684 * tree.h (DECL_LOCATION_RANGE): Remove unused macro.
7685 (get_decl_source_range): Remove unused function.
7687 2018-07-20 Richard Biener <rguenther@suse.de>
7689 * tree-ssa-sccvn.h (struct vn_nary_op_s): Add next member.
7690 (struct vn_phi_s): Likewise.
7691 (struct vn_reference_s): Likewise.
7692 * tree-ssa-sccvn.c (vn_nary_op_hasher::equal): Add shortcut
7693 for searching the slot of an entry known to be in the hash itself.
7694 (vn_phi_hasher::equal): Likewise.
7695 (vn_reference_hasher::equal): Likewise.
7696 (last_inserted_ref, last_inserted_phi, last_inserted_nary): New
7698 (optimistic_info, current_info): Remove, keeping only valid_info.
7699 (vn_reference_lookup_1): Remove fallback lookup.
7700 (vn_reference_lookup_2): Likewise.
7701 (vn_nary_op_lookup_1): Likewise.
7702 (vn_phi_lookup): Likewise.
7703 (vn_nary_build_or_lookup_1): Make sure to not chain the built
7705 (vn_reference_insert): Adjust, chain the inserted hash element
7706 at last_inserted_ref.
7707 (vn_reference_insert_pieces): Likewise.
7708 (visit_reference_op_call): Likewise.
7709 (vn_nary_op_insert_into): Chain the inserted hash element at
7711 (vn_nary_op_insert_pieces): Adjust.
7712 (vn_nary_op_insert): Likewise.
7713 (vn_nary_op_insert_stmt): Likewise.
7714 (vn_phi_insert): Adjust, chain the inserted hash element at
7716 (process_scc): Remove clearing and copying the optimistic
7717 table. Instead remove elements inserted during an optimistic
7718 iteration from the single table we maintain.
7719 (init_scc_vn): Adjust.
7720 (free_scc_vn): Likewise.
7721 (sccvn_dom_walker::record_cond): Likewise.
7722 (sccvn_dom_walker::after_dom_children): Likewise.
7724 2018-07-19 Martin Sebor <msebor@redhat.com>
7726 PR tree-optimization/84047
7727 PR tree-optimization/83776
7728 * tree-vrp.c (vrp_prop::check_mem_ref): New function.
7729 (check_array_bounds): Call it.
7731 2018-07-19 Martin Sebor <msebor@redhat.com>
7733 * align.h (align_flags): Use member initialization.
7735 2018-07-19 David Malcolm <dmalcolm@redhat.com>
7737 * Makefile.in (OBJS): Add optinfo.o.
7738 * coretypes.h (class symtab_node): New forward decl.
7739 (struct cgraph_node): New forward decl.
7740 (class varpool_node): New forward decl.
7741 * dump-context.h: New file.
7742 * dumpfile.c: Include "optinfo.h", "dump-context.h", "cgraph.h",
7744 (refresh_dumps_are_enabled): Use optinfo_enabled_p.
7745 (set_dump_file): Call dumpfile_ensure_any_optinfo_are_flushed.
7746 (set_alt_dump_file): Likewise.
7747 (dump_context::~dump_context): New dtor.
7748 (dump_gimple_stmt): Move implementation to...
7749 (dump_context::dump_gimple_stmt): ...this new member function.
7750 Add the stmt to any pending optinfo, creating one if need be.
7751 (dump_gimple_stmt_loc): Move implementation to...
7752 (dump_context::dump_gimple_stmt_loc): ...this new member function.
7753 Start a new optinfo and add the stmt to it.
7754 (dump_gimple_expr): Move implementation to...
7755 (dump_context::dump_gimple_expr): ...this new member function.
7756 Add the stmt to any pending optinfo, creating one if need be.
7757 (dump_gimple_expr_loc): Move implementation to...
7758 (dump_context::dump_gimple_expr_loc): ...this new member function.
7759 Start a new optinfo and add the stmt to it.
7760 (dump_generic_expr): Move implementation to...
7761 (dump_context::dump_generic_expr): ...this new member function.
7762 Add the tree to any pending optinfo, creating one if need be.
7763 (dump_generic_expr_loc): Move implementation to...
7764 (dump_context::dump_generic_expr_loc): ...this new member
7765 function. Add the tree to any pending optinfo, creating one if
7767 (dump_printf): Move implementation to...
7768 (dump_context::dump_printf_va): ...this new member function. Add
7769 the text to any pending optinfo, creating one if need be.
7770 (dump_printf_loc): Move implementation to...
7771 (dump_context::dump_printf_loc_va): ...this new member function.
7772 Start a new optinfo and add the stmt to it.
7773 (dump_dec): Move implementation to...
7774 (dump_context::dump_dec): ...this new member function. Add the
7775 value to any pending optinfo, creating one if need be.
7776 (dump_context::dump_symtab_node): New member function.
7777 (dump_context::get_scope_depth): New member function.
7778 (dump_context::begin_scope): New member function.
7779 (dump_context::end_scope): New member function.
7780 (dump_context::ensure_pending_optinfo): New member function.
7781 (dump_context::begin_next_optinfo): New member function.
7782 (dump_context::end_any_optinfo): New member function.
7783 (dump_context::s_current): New global.
7784 (dump_context::s_default): New global.
7785 (dump_scope_depth): Delete global.
7786 (dumpfile_ensure_any_optinfo_are_flushed): New function.
7787 (dump_symtab_node): New function.
7788 (get_dump_scope_depth): Reimplement in terms of dump_context.
7789 (dump_begin_scope): Likewise.
7790 (dump_end_scope): Likewise.
7791 (selftest::temp_dump_context::temp_dump_context): New ctor.
7792 (selftest::temp_dump_context::~temp_dump_context): New dtor.
7793 (selftest::verify_item): New function.
7794 (ASSERT_IS_TEXT): New macro.
7795 (ASSERT_IS_TREE): New macro.
7796 (ASSERT_IS_GIMPLE): New macro.
7797 (selftest::test_capture_of_dump_calls): New test.
7798 (selftest::dumpfile_c_tests): Call it.
7799 * dumpfile.h (dump_printf, dump_printf_loc, dump_basic_block)
7800 (dump_generic_expr_loc, dump_generic_expr, dump_gimple_stmt_loc)
7801 (dump_gimple_stmt, dump_dec): Gather these related decls and add a
7802 descriptive comment.
7803 (dump_function, print_combine_total_stats, enable_rtl_dump_file)
7804 (dump_node, dump_bb): Move these unrelated decls.
7805 (class dump_manager): Add leading comment.
7806 * optinfo.cc: New file.
7807 * optinfo.h: New file.
7809 2018-07-19 Michael Collison <michael.collison@arm.com>
7810 Richard Henderson <rth@redhat.com>
7812 * config/aarch64/aarch64.md (subv<GPI>4, usubv<GPI>4): New patterns.
7813 (subti): Handle op1 zero.
7814 (subvti4, usub4ti4): New.
7815 (*sub<GPI>3_compare1_imm): New.
7816 (sub<GPI>3_carryinCV): New.
7817 (*sub<GPI>3_carryinCV_z1_z2, *sub<GPI>3_carryinCV_z1): New.
7818 (*sub<GPI>3_carryinCV_z2, *sub<GPI>3_carryinCV): New.
7820 2018-07-19 Michael Collison <michael.collison@arm.com>
7821 Richard Henderson <rth@redhat.com>
7823 * config/aarch64/aarch64.md: (addv<GPI>4, uaddv<GPI>4): New.
7824 (addti3): Create simpler code if low part is already known to be 0.
7825 (addvti4, uaddvti4): New.
7826 (*add<GPI>3_compareC_cconly_imm): New.
7827 (*add<GPI>3_compareC_cconly): New.
7828 (*add<GPI>3_compareC_imm): New.
7829 (*add<GPI>3_compareC): Rename from add<GPI>3_compare1; do not
7830 handle constants within this pattern..
7831 (*add<GPI>3_compareV_cconly_imm): New.
7832 (*add<GPI>3_compareV_cconly): New.
7833 (*add<GPI>3_compareV_imm): New.
7834 (add<GPI>3_compareV): New.
7835 (add<GPI>3_carryinC, add<GPI>3_carryinV): New.
7836 (*add<GPI>3_carryinC_zero, *add<GPI>3_carryinV_zero): New.
7837 (*add<GPI>3_carryinC, *add<GPI>3_carryinV): New.
7838 ((*add<GPI>3_compareC_cconly_imm): Replace 'ne' operator
7839 with 'comparison' operator.
7840 (*add<GPI>3_compareV_cconly_imm): Ditto.
7841 (*add<GPI>3_compareV_cconly): Ditto.
7842 (*add<GPI>3_compareV_imm): Ditto.
7843 (add<GPI>3_compareV): Ditto.
7844 (add<mode>3_carryinC): Ditto.
7845 (*add<mode>3_carryinC_zero): Ditto.
7846 (*add<mode>3_carryinC): Ditto.
7847 (add<mode>3_carryinV): Ditto.
7848 (*add<mode>3_carryinV_zero): Ditto.
7849 (*add<mode>3_carryinV): Ditto.
7851 2018-07-19 Michael Collison <michael.collison@arm.com>
7852 Richard Henderson <rth@redhat.com>
7854 * config/aarch64/aarch64-modes.def (CC_V): New.
7855 * config/aarch64/aarch64-protos.h
7856 (aarch64_addti_scratch_regs): Declare
7857 (aarch64_subvti_scratch_regs): Declare.
7858 (aarch64_expand_subvti): Declare.
7859 (aarch64_gen_unlikely_cbranch): Declare
7860 * config/aarch64/aarch64.c (aarch64_select_cc_mode): Test
7861 for signed overflow using CC_Vmode.
7862 (aarch64_get_condition_code_1): Handle CC_Vmode.
7863 (aarch64_gen_unlikely_cbranch): New function.
7864 (aarch64_addti_scratch_regs): New function.
7865 (aarch64_subvti_scratch_regs): New function.
7866 (aarch64_expand_subvti): New function.
7868 2018-07-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
7870 * config/aarch64/aarch64-option-extensions.def: New entry for profile
7872 * config/aarch64/aarch64.h (AARCH64_FL_PROFILE): New.
7873 * doc/invoke.texi (aarch64-feature-modifiers): New entry for profile
7876 2018-07-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
7879 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
7880 address check not strict.
7882 2018-07-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
7884 * config/aarch64/aarch64-simd.md (aarch64_simd_mov<VQ:mode>): Replace
7886 (store_pair_lanes<mode>): Likewise.
7887 * config/aarch64/aarch64-protos.h (aarch64_addr_query_type): Add new
7888 enum value 'ADDR_QUERY_LDP_STP_N'.
7889 * config/aarch64/aarch64.c (aarch64_addr_query_type): Likewise.
7890 (aarch64_print_address_internal): Add declaration.
7891 (aarch64_print_ldpstp_address): Remove.
7892 (aarch64_classify_address): Adapt mode for 'ADDR_QUERY_LDP_STP_N'.
7893 (aarch64_print_operand): Change printing of 'y'.
7894 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Use
7895 new enum value 'ADDR_QUERY_LDP_STP_N', don't hardcode mode and use
7896 'true' rather than '1'.
7897 * config/aarch64/constraints.md (Uml): Likewise.
7898 (Uml): Rename to Umn.
7901 2018-07-19 Richard Biener <rguenther@suse.de>
7903 * tree-ssa-sccvn.h (struct vn_phi_s): Make phiargs member
7905 * tree-ssa-sccvn.c: Remove alloc-pool.h use.
7906 (vn_phi_hasher): Derive from nofree_ptr_hash and remove remove method.
7907 (vn_reference_hasher): Likewise.
7908 (struct vn_tables_s): Remove obstack and alloc-pool members.
7909 (vn_tables_obstack, vn_tables_insert_obstack): New global obstacks.
7910 (vn_nary_build_or_lookup_1): Manually build in vn_tables_insert_obstack.
7911 (vn_reference_insert): Allocate from obstack instead of from alloc-pool.
7912 (vn_reference_insert_pieces): Likewise.
7913 (alloc_vn_nary_op_noinit): Adjust.
7914 (vn_nary_op_insert_stmt): Allocate phiargs in-place.
7915 (vn_phi_eq): Adjust.
7916 (shared_lookup_phiargs): Remove.
7917 (vn_phi_lookup): Allocate temporary vn_phi_s on the stack.
7918 (vn_phi_insert): Allocate from obstack instead of from alloc-pool.
7919 (visit_reference_op_call): Likewise.
7920 (copy_nary, copy_phi, copy_reference): Remove.
7921 (process_scc): Rewind the obstack when iterating. Do not
7922 copy the elements to valid_info but just move them from one
7923 hashtable to the other.
7924 (allocate_vn_table): Adjust.
7925 (free_vn_table): Likewise.
7926 (init_scc_vn): Likewise.
7927 (free_scc_vn): Likewise.
7929 2018-07-19 H.J. Lu <hongjiu.lu@intel.com>
7932 * config/i386/i386.c (rest_of_insert_endbranch): Lookup
7933 indirect_return as function type attribute.
7934 (ix86_attribute_table): Change indirect_return to function
7936 * doc/extend.texi: Update indirect_return attribute.
7938 2018-07-19 Aldy Hernandez <aldyh@redhat.com>
7940 * wide-int.h (widest2_int): New.
7941 * gimple-fold.c (arith_overflowed_p): Use it.
7942 * tree.h (widest2_int_cst): New.
7943 * tree-vrp.c (wide_int_binop_overflow): Rename from
7944 vrp_int_const_binop.
7945 Rewrite to work on trees.
7946 (extract_range_from_multiplicative_op_1): Abstract code to...
7947 (wide_int_range_min_max): ...here.
7948 (wide_int_range_cross_product): ...and here.
7949 (extract_range_from_binary_expr_1): Abstract overflow code to...
7950 (wide_int_range_mult_wrapping): ...here.
7951 * tree-vrp.h (wide_int_range_cross_product): New.
7952 (wide_int_range_mult_wrapping): New.
7954 2018-07-19 Andrew Senkevich <andrew.senkevich@intel.com>
7955 Julia Koval <julia.koval@intel.com>
7957 * config/i386/x86-tune-costs.h (skylake_memcpy,
7958 skylake_memset): Replace rep_prefix with unrolling for size 512.
7960 2018-07-18 Kugan Vivekanandarajah <kuganv@linaro.org>
7963 * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): Handle
7964 comparision with EQ_EXPR in last stmt.
7966 2018-07-18 Kelvin Nilsen <kelvin@gcc.gnu.org>
7968 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Rename
7969 this subsection to "PowerPC AltiVec/VSX Built-in Functions".
7970 (PowerPC AltiVec/VSX Built-in Functions): New name for subsection
7971 previously known as "PowerPC AltiVec Built-in Functions". Move
7972 some material to new subsubsections "PowerPC AltiVec Built-in
7973 Functions on ISA 2.06" and "PowerPC AltiVec Built-in Functions on
7975 (PowerPC Altivec Built-in Functions on ISA 2.05): New subsubsection.
7976 (PowerPC Altivec Built-in Functions on ISA 2.06): Likewise.
7977 (PowerPC Altivec Built-in Functions on ISA 2.07): Likewise.
7978 (PowerPC Altivec Built-in Functions on ISA 3.0): Likewise.
7980 2018-07-18 Richard Biener <rguenther@suse.de>
7982 PR tree-optimization/86557
7983 * tree-vect-patterns.c (vect_recog_divmod_pattern): Also handle
7986 2018-07-18 Ilya Leoshkevich <iii@linux.ibm.com>
7988 * config/s390/s390.c (s390_function_profiler): Generate CFI.
7990 2018-07-17 Jeff Law <law@redhat.com>
7992 * config/arm/arm.c (get_label_padding): Update for recent
7993 changes to label_to_alignment.
7995 PR tree-optimization/86010
7996 * tree-ssa-dse.c (compute_trims): Fix typo/thinko.
7998 * config/mips/mips.c (vr4130_align_insns): Update for recent
7999 changes to label_to_alignment.
8001 * config/frv/frv.c (frv_label_align): Update for recent changes
8002 to label_to_alignment.
8004 * config/nios2/nios2.c (nios2_label_align): Update for recent
8005 changes which dropped ALIGN_LABELS_LOG.
8007 2018-07-17 Andreas Schwab <schwab@linux-m68k.org>
8009 * config/m68k/m68k.md (umulsi3_highpart+1, const_umulsi3_highpart)
8010 (smulsi3_highpart+1, const_smulsi3_highpart): Add CC_STATUS_INIT.
8012 2018-07-17 Claudiu Zissulescu <claziss@synopsys.com>
8014 * config/arc/arc.c (arc_label_align): Use align_labels instead of
8015 deprecated align_labels_log.
8017 2018-07-17 Richard Biener <rguenther@suse.de>
8020 * dwarf2out.c (init_sections_and_labels): Always generate
8021 a debug_line_str_section for early LTO debug.
8022 (dwarf2out_finish): Reset debug_line_str_hash output early.
8023 Bump counter for extra dwarf5 .debug_loc labels to not conflict
8025 (dwarf2out_early_finish): Output debug_line_str.
8027 2018-07-17 Robin Dapp <rdapp@linux.ibm.com>
8029 * config/s390/s390.c (preferred_la_operand_p): Do not use LA with
8030 index register on z196 or later.
8032 2018-07-17 Robin Dapp <rdapp@linux.ibm.com>
8034 * config/s390/s390.c (s390_default_align): Set default function
8036 (s390_override_options_after_change): Call s390_default align.
8037 (s390_option_override_internal): Call s390_default_align.
8038 (TARGET_OVERRIDE_OPTIONS_AFTER_CHANGE): Define.
8040 2018-07-17 Jakub Jelinek <jakub@redhat.com>
8043 * omp-low.c (create_task_copyfn): Copy over also fields corresponding
8044 to _looptemp_ clauses, other than the first two.
8046 2018-07-17 Martin Liska <mliska@suse.cz>
8048 * opts.c: Do not enable OPT_falign_* for -Os.
8050 2018-07-17 Martin Liska <mliska@suse.cz>
8052 * align.h (MAX_CODE_ALIGN): New.
8053 (MAX_CODE_ALIGN_VALUE): New.
8054 * common/config/i386/i386-common.c (ix86_handle_option):
8055 (MAX_CODE_ALIGN): Moved to align.h.
8056 * final.c (MAX_CODE_ALIGN): Likewise.
8057 * opts.c (parse_and_check_align_values):
8058 (MAX_CODE_ALIGN): Likewise.
8059 (MAX_CODE_ALIGN_VALUE): Likewise.
8061 2018-07-17 Martin Liska <mliska@suse.cz>
8063 * config/i386/att.h (ASM_OUTPUT_ALIGN): Fix spacing
8064 in order to fulfil coding style.
8065 * config/i386/cygming.h (ASM_OUTPUT_ALIGN): Likewise.
8066 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
8067 * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
8068 * config/iq2000/iq2000.h (ASM_OUTPUT_ALIGN): Likewise.
8069 * config/pa/pa.h (ASM_OUTPUT_ALIGN): Likewise.
8070 * config/sparc/sol2.h (ASM_OUTPUT_ALIGN_WITH_NOP): Likewise.
8071 * config/sparc/sparc.h (ASM_OUTPUT_ALIGN): Likewise.
8072 * config/visium/visium.h (ASM_OUTPUT_ALIGN): Likewise.
8073 (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
8075 2018-07-17 Martin Liska <mliska@suse.cz>
8077 * align.h: New file.
8078 * config/alpha/alpha.c (alpha_align_insns_1): Use align_functions
8080 * config/i386/i386.c (ix86_avoid_jump_mispredicts): Use new return type
8081 align_flags of label_to_alignment.
8082 * config/m32r/m32r.h (LOOP_ALIGN): Wrap returned values into
8084 * config/m68k/m68k.c: Do not use removed align_labels_value and
8086 * config/nds32/nds32.h (JUMP_ALIGN): Wrap result into align_flags class.
8087 (LOOP_ALIGN): Likewise.
8088 (LABEL_ALIGN): Likewise.
8089 * config/powerpcspe/powerpcspe.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
8090 Remove not used macro.
8091 (rs6000_loop_align): Change return type to align_flags.
8092 (rs6000_loop_align_max_skip): Remove.
8093 * config/rs6000/rs6000-protos.h (rs6000_loop_align):
8094 Change return type to align_flags.
8095 * config/rs6000/rs6000.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
8096 Remove not used macro.
8097 (rs6000_loop_align): Change return type to align_flags.
8098 (rs6000_loop_align_max_skip): Remove.
8099 * config/rx/rx.h (JUMP_ALIGN): Wrap integer values
8100 * config/rx/rx-protos.h (rx_align_for_label): Make it
8102 * config/rx/rx.c (rx_align_for_label): Change return type
8104 (rx_max_skip_for_label): Remove TARGET_ASM_*_ALIGN_MAX_SKIP
8106 into align_flags class.
8107 (LABEL_ALIGN): Likewise.
8108 (LOOP_ALIGN): Likewise.
8109 * config/s390/s390.c (s390_label_align): Use align_flags
8111 (s390_asm_output_function_label): Likewise.
8112 * config/sh/sh.c (sh_override_options_after_change):
8113 Use align_flags class directly without macros.
8114 (find_barrier): Likewise.
8115 (barrier_align): Likewise.
8116 (sh_loop_align): Likewise.
8117 * config/spu/spu.c (spu_option_override):
8118 Use align_flags_tuple::get_value instead of removed macros.
8119 (spu_sched_init): Likewise.
8120 * config/spu/spu.h (GTY): Likewise.
8121 * config/visium/visium.c (visium_option_override):
8122 Set "8" as default secondary alignment.
8123 * config/visium/visium.h (SUBALIGN_LOG): Define to 3
8124 in order to guarantee secondary alignment of 8.
8125 * coretypes.h: Include align.h header file.
8126 * doc/tm.texi: Remove TARGET_ASM_JUMP_ALIGN_MAX_SKIP,
8127 TARGET_ASM_LOOP_ALIGN_MAX_SKIP, TARGET_ASM_LABEL_ALIGN_MAX_SKIP
8128 and TARGET_ASM_LABEL_ALIGN_AFTER_BARRIER_MAX_SKIP macros.
8129 * doc/tm.texi.in: Likewise.
8130 * final.c (struct label_alignment): Remove not used structure.
8131 (LABEL_ALIGN): Change type to align_flags.
8132 (LOOP_ALIGN): Likewise.
8133 (JUMP_ALIGN): Likewise.
8134 (default_loop_align_max_skip): Remove.
8135 (default_label_align_max_skip): Likewise.
8136 (default_jump_align_max_skip): Likewise.
8137 (default_label_align_after_barrier_max_skip):
8138 (LABEL_TO_ALIGNMENT): Change to access label_align vector.
8139 (LABEL_TO_MAX_SKIP): Remove.
8140 (label_to_alignment): Return align_flags type instead of integer.
8141 (label_to_max_skip): Remove.
8142 (align_fuzz): Use align_flags type.
8143 (compute_alignments): Use align_flags type and use align_flags::max
8144 to combine multiple alignments.
8145 (grow_label_align): Grow vec instead of C array.
8146 (update_alignments): Assign just LABEL_TO_ALIGNMENT.
8147 (shorten_branches): Use align_flags type and use align_flags::max
8148 to combine multiple alignments.
8149 (final_scan_insn_1): Remove usage of secondary alignment that comes
8150 from label alignment, but instead use proper secondary alignment
8151 which is computed in grow_label_align.
8152 * flags.h (struct align_flags_tuple): Move to align.h.
8153 (struct align_flags): Likewise.
8154 (state_align_loops): Rename to align_loops.
8155 (state_align_jumps): Rename to align_jumps.
8156 (state_align_labels): Rename to align_labels.
8157 (state_align_functions): Rename to align_functions.
8158 (align_loops_log): Remove.
8159 (align_jumps_log): Remove.
8160 (align_labels_log): Remove.
8161 (align_functions_log): Remove.
8162 (align_loops_max_skip): Remove.
8163 (align_jumps_max_skip): Remove.
8164 (align_labels_max_skip): Remove.
8165 (align_functions_max_skip): Remove.
8166 (align_loops_value): Remove.
8167 (align_jumps_value): Remove.
8168 (align_labels_value): Remove.
8169 (align_functions_value): Remove.
8170 * output.h (label_to_alignment): Change return type to align_flags.
8171 (label_to_max_skip): Remove.
8172 * target.def: Remove loop_align_max_skip, label_align_max_skip,
8173 jump_align_max_skip macros.
8174 * targhooks.h (default_loop_align_max_skip): Remove.
8175 (default_label_align_max_skip): Likewise.
8176 (default_jump_align_max_skip): Likewise.
8177 (default_label_align_after_barrier_max_skip): Remove.
8178 * toplev.c (read_log_maxskip): Use ::normalize function.
8179 (parse_N_M): Remove not used argument and also call ::normalize.
8180 (parse_alignment_opts): Do not pass unused arguments.
8181 * varasm.c (assemble_start_function): Use directly align_functions
8182 instead of removed macros.
8183 * system.h: Do not poison removed macros.
8185 2018-07-17 Jakub Jelinek <jakub@redhat.com>
8188 * gimplify.c (gimplify_omp_for): Ensure taskloop firstprivatized init
8189 and cond temporaries don't have reference type if iterator has
8190 pointer type. For init use &for_pre_body instead of pre_p if
8191 for_pre_body is non-empty.
8193 2018-07-16 Segher Boessenkool <segher@kernel.crashing.org>
8195 * config/rs6000/rs6000.md (trunc<mode>sf2): Expand truncates of
8196 double-double modes to SFmode directly directly.
8197 (trunc<mode>sf2_fprs): Delete.
8199 2018-07-16 Segher Boessenkool <segher@kernel.crashing.org>
8201 * config/rs6000/rs6000.c (init_float128_ibm): Use the correct names
8202 for conversions between IFmode and the decimal floating point modes.
8203 (init_float128_ieee): Use the correct names for conversions between
8204 KFmode and the decimal floating point modes.
8206 2018-07-16 Segher Boessenkool <segher@kernel.crashing.org>
8208 * config/rs6000/rs6000.c (init_float128_ibm): Use more correct names
8209 for the conversions between TDmode and IFmode.
8210 (init_float128_ieee): Use more correct names for the conversions
8211 between TDmode and KFmode.
8213 2018-07-16 Jakub Jelinek <jakub@redhat.com>
8215 PR tree-optimization/86526
8216 * builtins.c (expand_builtin_memcmp): Formatting fixes.
8217 (inline_expand_builtin_string_cmp): Likewise.
8218 (inline_string_cmp): Likewise. Use c_readstr instead of
8219 builtin_memcpy_read_str. Add unit_mode temporary.
8221 2018-07-16 Bernd Edlinger <bernd.edlinger@hotmail.de>
8224 * builtins.c (check_access): Bail out if range[0] is no INTEGER_CST.
8225 * expr.c (string_constant): Fix the element size of ARRAY_TYPE.
8227 2018-07-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
8229 * doc/extend.texi (PowerPC AltiVec Built-in Functions):
8230 Alphabetize prototypes of built-in functions, separating out
8231 built-in functions that are listed in this section but should be
8232 described elsewhere.
8234 2018-07-16 Uros Bizjak <ubizjak@gmail.com>
8237 * expmed.c (emit_store_flag): Do not emit setcc followed by a
8238 conditional move when trapping comparison was split to a
8239 non-trapping one (and vice versa).
8241 2018-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
8243 * config/s390/s390.c (s390_function_profiler): Generate nops
8244 instead of profiler call sequences.
8245 * config/s390/s390.opt: Add the new option.
8247 2018-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
8249 * config/s390/s390.c (s390_function_profiler): Generate
8250 __mcount_loc section.
8251 * config/s390/s390.opt: Add the new option.
8253 2018-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
8255 * common.opt: Add the new warning.
8256 * config/s390/s390.c (s390_function_profiler): Emit "brasl
8257 %r0,__fentry__" when -mfentry is specified.
8258 (s390_option_override_internal): Disallow -mfentry for 31-bit
8260 * config/s390/s390.opt: Add the new option.
8262 2018-07-16 Richard Biener <rguenther@suse.de>
8265 * dwarf2out.c (dwarf2out_register_external_die): Assign DIE parents
8266 for function-local FUNCTION_DECL and RESULT_DECL immediately.
8268 2018-07-16 Martin Liska <mliska@suse.cz>
8271 * ipa-pure-const.c (malloc_candidate_p): Revert ::get
8274 2017-07-16 Claudiu Zissulescu <claziss@synopsys.com>
8276 * config/arc/arcHS.md: Update ARCHS scheduling rules.
8278 2017-07-16 Claudiu Zissulescu <claziss@synopsys.com>
8280 * config/arc/arc-arch.h (arc_tune_attr): Add new tune parameters
8282 * config/arc/arc-cpus.def (hs4x): New cpu.
8284 * config/arc/arc-tables.opt: Regenerate.
8285 * config/arc/arc.c (arc_sched_issue_rate): New function.
8286 (TARGET_SCHED_ISSUE_RATE): Define.
8287 (TARGET_SCHED_EXPOSED_PIPELINE): Likewise.
8288 * config/arc/arc.md (attr type): Add fpu_fuse, fpu_sdiv, fpu_ddiv,
8290 (attr tune): Add ARCHS4x tune values.
8291 (attr tune_dspmpy): Define.
8292 (*tst): Correct instruction type.
8293 * config/arc/arcHS.md: Don't use this automaton for ARCHS4x cpus.
8294 * config/arc/arcHS4x.md: New file.
8295 * config/arc/fpu.md: Update instruction type attributes.
8296 * config/arc/t-multilib: Regenerate.
8298 2018-07-16 Tom de Vries <tdevries@suse.de>
8301 * var-tracking.c (vt_initialize): Fix pre_dec handling.
8303 2018-07-16 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
8305 * config/aarch64/atomics.md (aarch64_store_execlusive<mode>): Add
8308 2018-07-16 Eric Botcazou <ebotcazou@adacore.com>
8310 PR tree-optimization/86514
8311 * tree-ssa-reassoc.c (init_range_entry) <CASE_CONVERT>: Return for a
8312 conversion to a boolean type from a type with greater precision.
8314 2018-07-16 Tom de Vries <tdevries@suse.de>
8316 * var-tracking.c (vt_initialize): Print adjusted insn slim if
8317 dump_flags request TDF_SLIM.
8319 2018-07-16 Aldy Hernandez <aldyh@redhat.com>
8321 * fold-const.c (int_const_binop_1): Abstract...
8322 (wide_int_binop): ...wide int code here.
8323 (poly_int_binop): ...poly int code here.
8324 Abstract the rest of int_const_binop_1 into int_const_binop.
8325 * fold-const.h (wide_int_binop): New.
8326 * tree-vrp.c (vrp_int_const_binop): Call wide_int_binop.
8327 Remove useless PLUS/MINUS_EXPR case.
8328 (zero_nonzero_bits_from_vr): Move wide int code...
8329 (zero_nonzero_bits_from_bounds): ...here.
8330 (extract_range_from_binary_expr_1): Move mask optimization code...
8331 (range_easy_mask_min_max): ...here.
8332 * tree-vrp.h (zero_nonzero_bits_from_bounds): New.
8333 (range_easy_mask_min_max): New.
8335 2018-07-15 Jeff Law <law@redhat.com>
8338 * config/sh/sh.c (output_mi_thunk): Remove dead conditional
8341 2018-07-14 Jim Wilson <jimw@sifive.com>
8343 * config/riscv/linux.h (TARGET_ASM_FILE_END): New.
8345 2018-07-14 Paul Koning <ni1d@arrl.net>
8347 * config/pdp11/pdp11.c (pdp11_rtx_costs): Bugfixes.
8349 2018-07-13 Jan Hubicka <hubicka@ucw.cz>
8351 * lto-streamer-out.c (copy_function_or_variable): Dump info about
8354 2018-07-13 Bill Schmidt <wschmidt@linux.ibm.com>
8355 Steve Munroe <munroesj52@gmail.com>
8357 * config/rs6000/emmintrin.h (_mm_and_si128): New function.
8358 (_mm_andnot_si128): Likewise.
8359 (_mm_or_si128): Likewise.
8360 (_mm_xor_si128): Likewise.
8362 2018-07-13 Qing Zhao <qing.zhao@oracle.com>
8365 * builtins.c (expand_builtin_memcmp): Inline the calls first
8366 when result_eq is false.
8367 (expand_builtin_strcmp): Inline the calls first.
8368 (expand_builtin_strncmp): Likewise.
8369 (inline_string_cmp): New routine. Expand a string compare
8370 call by using a sequence of char comparison.
8371 (inline_expand_builtin_string_cmp): New routine. Inline expansion
8372 a call to str(n)cmp/memcmp.
8373 * doc/invoke.texi (--param builtin-string-cmp-inline-length):
8375 * params.def (BUILTIN_STRING_CMP_INLINE_LENGTH): New.
8377 2018-07-13 Richard Earnshaw <rearnsha@arm.com>
8379 * config/arm/driver-arm.c: Include arm-native.h.
8380 (host_detect_local_cpu): Use auto-generated data tables.
8381 (vendors, arm_cpu_table): Delete. Move part information to ...
8382 * config/arm/arm-cpus.in: ... here.
8383 * config/arm/parsecpu.awk (gen_native): New function.
8384 (vendor, part): New CPU fields.
8385 (END): Add support for building the native CPU detection tables.
8386 * config/arm/t-arm (arm-native.h): Add build rule.
8387 (driver-arm.o): Add dependency on arm-native.h.
8389 2018-07-13 Richard Biener <rguenther@suse.de>
8392 * match.pd (addr1 - addr2): Allow either of the operand to
8395 2018-07-13 Tom de Vries <tdevries@suse.de>
8397 * tree-inline.c (remap_ssa_name): Save and reuse debug exprs generated
8400 2018-07-13 Jackson Woodruff <jackson.woodruff@arm.com>
8402 * config/aarch64/aarch64.c (aarch64_operands_adjust_ok_for_ldpstp): Use
8403 arrays instead of numbered variables.
8405 2018-07-13 Eric Botcazou <ebotcazou@adacore.com>
8407 * config/sparc/sparc-protos.h (sparc_compute_frame_size): Delete.
8408 * config/sparc/sparc.c (sparc_compute_frame_size): Make static.
8410 2018-07-13 Richard Biener <rguenther@suse.de>
8413 * dwarf2out.c (gen_type_die_with_usage): Use scope_die_for
8414 instead of get_context_die.
8416 2018-07-13 Kugan Vivekanandarajah <kuganv@linaro.org>
8417 Richard Biener <rguenther@suse.de>
8420 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Check
8421 that the loop latch destination where phi is defined.
8423 2018-07-12 Kito Cheng <kito.cheng@gmail.com>
8425 * config/riscv/riscv.c (enum riscv_privilege_levels): Add UNKNOWN_MODE.
8426 (riscv_expand_epilogue): Add assertion to check interrupt mode.
8427 (riscv_set_current_function): Extract getting interrupt type to new
8429 (riscv_get_interrupt_type): New function.
8430 (riscv_merge_decl_attributes): New function, checking interrupt type is
8432 (TARGET_MERGE_DECL_ATTRIBUTES): Define.
8434 2018-07-12 Paul Koning <ni1d@arrl.net>
8436 * config/pdp11/pdp11.c (pdp11_output_def): Fix typo in .set
8439 2018-07-12 Paul Koning <ni1d@arrl.net>
8441 * doc/rtl.texi (REG_NONNEG): Remove decrement and branch until
8442 zero reference, add doloop_end instead.
8443 * doc/md.texi (decrement_and_branch_until_zero): Remove.
8444 (Looping patterns): Remove decrement_and_branch_until_zero. Add
8445 detail for doloop_end.
8447 2018-07-12 Martin Sebor <msebor@redhat.com>
8450 * attribs.c (decl_attributes): Reject conflicting attributes before
8451 calling attribute handlers.
8453 2018-07-12 Jan Hubicka <hubicka@ucw.cz>
8455 * dumpfile.c (gcc::dump_manager::get_dump_file_name): Add PART
8457 (gcc::dump_manager::get_dump_file_name): likewise.
8458 (dump_begin): Likewise.
8459 * dumpfile.h (dump_begin): Update prototype.
8460 (gcc::dump_manager::get_dump_file_name,
8461 gcc::dump_manager::get_dump_file_name): Update prototype.
8463 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
8465 * internal-fn.h (vectorizable_internal_fn_p): New function.
8466 * tree-vect-slp.c (compatible_calls_p): Likewise.
8467 (vect_build_slp_tree_1): Remove nops argument. Handle calls
8468 to internal functions.
8469 (vect_build_slp_tree_2): Update call to vect_build_slp_tree_1.
8471 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
8473 * fold-const.h (inverse_conditions_p): Declare.
8474 * fold-const.c (inverse_conditions_p): New function.
8475 * match.pd: Use inverse_conditions_p. Add folds of view_converts
8476 that test the inverse condition of a conditional internal function.
8477 * internal-fn.h (vectorized_internal_fn_supported_p): Declare.
8478 * internal-fn.c (internal_fn_mask_index): Handle conditional
8480 (vectorized_internal_fn_supported_p): New function.
8481 * tree-if-conv.c: Include internal-fn.h and fold-const.h.
8482 (any_pred_load_store): Replace with...
8483 (need_to_predicate): ...this new variable.
8484 (redundant_ssa_names): New variable.
8485 (ifcvt_can_use_mask_load_store): Move initial checks to...
8486 (ifcvt_can_predicate): ...this new function. Handle tree codes
8487 for which a conditional internal function exists.
8488 (if_convertible_gimple_assign_stmt_p): Use ifcvt_can_predicate
8489 instead of ifcvt_can_use_mask_load_store. Update after variable
8491 (predicate_load_or_store): New function, split out from
8492 predicate_mem_writes.
8493 (check_redundant_cond_expr): New function.
8494 (value_available_p): Likewise.
8495 (predicate_rhs_code): Likewise.
8496 (predicate_mem_writes): Rename to...
8497 (predicate_statements): ...this. Use predicate_load_or_store
8498 and predicate_rhs_code.
8499 (combine_blocks, tree_if_conversion): Update after above name changes.
8500 (ifcvt_local_dce): Handle redundant_ssa_names.
8501 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): Handle
8502 general conditional functions.
8503 * tree-vect-stmts.c (vectorizable_call): Likewise.
8505 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
8506 Alan Hayward <alan.hayward@arm.com>
8507 David Sherwood <david.sherwood@arm.com>
8509 * internal-fn.h (can_interpret_as_conditional_op_p): Declare.
8510 * internal-fn.c (can_interpret_as_conditional_op_p): New function.
8511 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Handle conditional
8512 plus and minus and convert them into IFN_COND_FMA-based sequences.
8513 (convert_mult_to_fma): Handle conditional plus and minus.
8515 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
8517 * doc/md.texi (cond_fma, cond_fms, cond_fnma, cond_fnms): Document.
8518 * optabs.def (cond_fma_optab, cond_fms_optab, cond_fnma_optab)
8519 (cond_fnms_optab): New optabs.
8520 * internal-fn.def (COND_FMA, COND_FMS, COND_FNMA, COND_FNMS): New
8522 (FMA): Use DEF_INTERNAL_FLT_FN rather than DEF_INTERNAL_FLT_FLOATN_FN.
8523 * internal-fn.h (get_conditional_internal_fn): Declare.
8524 (get_unconditional_internal_fn): Likewise.
8525 * internal-fn.c (cond_ternary_direct): New macro.
8526 (expand_cond_ternary_optab_fn): Likewise.
8527 (direct_cond_ternary_optab_supported_p): Likewise.
8528 (FOR_EACH_COND_FN_PAIR): Likewise.
8529 (get_conditional_internal_fn): New function.
8530 (get_unconditional_internal_fn): Likewise.
8531 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 5.
8532 (gimple_match_op::gimple_match_op): Add a new overload for 5
8534 (gimple_match_op::set_op): Likewise.
8535 (gimple_resimplify5): Declare.
8536 * genmatch.c (decision_tree::gen): Generate simplifications for
8538 * gimple-match-head.c (gimple_simplify): Define an overload for
8539 5 operands. Handle calls with 5 arguments in the top-level overload.
8540 (convert_conditional_op): Handle conversions from unconditional
8541 internal functions to conditional ones.
8542 (gimple_resimplify5): New function.
8543 (build_call_internal): Pass a fifth operand.
8544 (maybe_push_res_to_seq): Likewise.
8545 (try_conditional_simplification): Try converting conditional
8546 internal functions to unconditional internal functions.
8547 Handle 3-operand unconditional forms.
8548 * match.pd (UNCOND_TERNARY, COND_TERNARY): Operator lists.
8549 Define ternary equivalents of the current rules for binary conditional
8551 * config/aarch64/aarch64.c (aarch64_preferred_else_value): Handle
8553 * config/aarch64/iterators.md (UNSPEC_COND_FMLA, UNSPEC_COND_FMLS)
8554 (UNSPEC_COND_FNMLA, UNSPEC_COND_FNMLS): New unspecs.
8555 (optab): Handle them.
8556 (SVE_COND_FP_TERNARY): New int iterator.
8557 (sve_fmla_op, sve_fmad_op): New int attributes.
8558 * config/aarch64/aarch64-sve.md (cond_<optab><mode>)
8559 (*cond_<optab><mode>_2, *cond_<optab><mode_4)
8560 (*cond_<optab><mode>_any): New SVE_COND_FP_TERNARY patterns.
8562 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
8564 * target.def (preferred_else_value): New target hook.
8565 * doc/tm.texi.in (TARGET_PREFERRED_ELSE_VALUE): New hook.
8566 * doc/tm.texi: Regenerate.
8567 * targhooks.h (default_preferred_else_value): Declare.
8568 * targhooks.c (default_preferred_else_value): New function.
8569 * internal-fn.h (conditional_internal_fn_code): Declare.
8570 * internal-fn.c (FOR_EACH_CODE_MAPPING): New macro.
8571 (get_conditional_internal_fn): Use it.
8572 (conditional_internal_fn_code): New function.
8573 * gimple-match.h (gimple_match_cond): New struct.
8574 (gimple_match_op): Add a cond member function.
8575 (gimple_match_op::gimple_match_op): Update all forms to take a
8577 * genmatch.c (expr::gen_transform): Use the same condition as res_op
8578 for the suboperation, but don't specify a particular else_value.
8579 * tree-ssa-sccvn.c (vn_nary_simplify, vn_reference_lookup_3)
8580 (visit_nary_op, visit_reference_op_load): Pass
8581 gimple_match_cond::UNCOND to the gimple_match_op constructor.
8582 * gimple-match-head.c: Include tree-eh.h
8583 (convert_conditional_op): New function.
8584 (maybe_resimplify_conditional_op): Likewise.
8585 (gimple_resimplify1): Call maybe_resimplify_conditional_op.
8586 (gimple_resimplify2): Likewise.
8587 (gimple_resimplify3): Likewise.
8588 (gimple_resimplify4): Likewise.
8589 (maybe_push_res_to_seq): Return null for conditional operations.
8590 (try_conditional_simplification): New function.
8591 (gimple_simplify): Call it. Pass conditions to the gimple_match_op
8593 * match.pd: Fold VEC_COND_EXPRs of an IFN_COND_* call to a new
8595 * config/aarch64/aarch64.c (aarch64_preferred_else_value): New
8597 (TARGET_PREFERRED_ELSE_VALUE): Redefine.
8599 2018-07-12 Jan Hubicka <hubicka@ucw.cz>
8601 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
8603 (hash_tree): Do not hash DECL_FCONTEXT
8604 * tree-streamer-in.c (lto_input_ts_field_decl_tree_pointers):
8605 Do not stream DECL_FCONTEXT.
8606 * tree-streamer-out.c (write_ts_field_decl_tree_pointers): Likewise.
8607 * tree.c (free_lang_data_in_decl): Free DECL_FCONTEXT.
8609 2018-07-12 Richard Biener <rguenther@suse.de>
8612 * dwarf2out.c (gen_block_die): Only output blocks when they have
8613 at least one !DECL_IGNORED_P variable.
8615 2018-07-12 Richard Biener <rguenther@suse.de>
8618 * config/gnu-user.h (GNU_USER_TARGET_NO_PTHREADS_LIB_SPEC):
8619 Remove -mieee-fp handling.
8621 2018-07-12 Richard Biener <rguenther@suse.de>
8623 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Remove bogus
8624 left-over from last patch.
8626 2018-07-12 Jakub Jelinek <jakub@redhat.com>
8628 PR tree-optimization/86492
8629 * gimple-ssa-store-merging.c
8630 (imm_store_chain_info::coalesce_immediate_stores): Call
8631 check_no_overlap even for the merge_overlapping case. Formatting fix.
8633 2018-07-12 Richard Biener <rguenther@suse.de>
8636 * fold-const.c (fold_binary_op_with_conditional_arg): Do not
8637 move possibly trapping operations into the conditional.
8639 2018-07-12 Richard Biener <rguenther@suse.de>
8641 * tree-ssa-sccvn.c (mprts_hook_cnt): Remove.
8642 (vn_lookup_simplify_result): Remove recursion limit applied
8644 (vn_nary_build_or_lookup_1): Adjust.
8645 (try_to_simplify): Likewise.
8646 * gimple-match-head.c (gimple_resimplify1): Instead apply one
8648 (gimple_resimplify2): Likewise.
8649 (gimple_resimplify3): Likewise.
8650 (gimple_resimplify4): Likewise.
8652 2018-07-11 Jakub Jelinek <jakub@redhat.com>
8654 * config/i386/avx512bitalgintrin.h (_mm512_mask_bitshuffle_epi64_mask):
8655 Use __mmask64 type instead of __mmask8 for __M argument.
8656 * config/i386/avx512fintrin.h (_mm512_mask_xor_epi64,
8657 _mm512_maskz_xor_epi64): Use __mmask8 type instead of __mmask16 for
8659 (_mm512_mask_cmpneq_epi64_mask): Use __mmask8 type instead of
8660 __mmask16 for __M argument.
8661 (_mm512_maskz_insertf32x4, _mm512_maskz_inserti32x4,
8662 _mm512_mask_insertf32x4, _mm512_mask_inserti32x4): Cast last argument
8663 to __mmask16 instead of __mmask8.
8664 * config/i386/avx512vlintrin.h (_mm_mask_add_ps, _mm_maskz_add_ps,
8665 _mm256_mask_add_ps, _mm256_maskz_add_ps, _mm_mask_sub_ps,
8666 _mm_maskz_sub_ps, _mm256_mask_sub_ps, _mm256_maskz_sub_ps,
8667 _mm256_maskz_cvtepi32_ps, _mm_maskz_cvtepi32_ps): Use __mmask8 type
8668 instead of __mmask16 for __U argument.
8669 * config/i386/avx512vlbwintrin.h (_mm_mask_cmp_epi8_mask): Use
8670 __mmask16 instead of __mmask8 for __U argument.
8671 (_mm256_mask_cmp_epi8_mask): Use __mmask32 instead of __mmask16 for
8673 (_mm256_cmp_epi8_mask): Use __mmask32 return type instead of
8675 (_mm_mask_cmp_epu8_mask): Use __mmask16 instead of __mmask8 for __U
8677 (_mm256_mask_cmp_epu8_mask): Use __mmask32 instead of __mmask16 for
8679 (_mm256_cmp_epu8_mask): Use __mmask32 return type instead of
8681 (_mm_mask_cmp_epi16_mask): Cast last argument to __mmask8 instead
8683 (_mm256_mask_cvtepi8_epi16): Use __mmask16 instead of __mmask32 for
8685 (_mm_mask_cvtepi8_epi16): Use __mmask8 instead of __mmask32 for
8687 (_mm256_mask_cvtepu8_epi16): Use __mmask16 instead of __mmask32 for
8689 (_mm_mask_cvtepu8_epi16): Use __mmask8 instead of __mmask32 for
8691 (_mm256_mask_cmpneq_epu8_mask, _mm256_mask_cmplt_epu8_mask,
8692 _mm256_mask_cmpge_epu8_mask, _mm256_mask_cmple_epu8_mask): Change
8693 return type as well as __M argument type and all casts from __mmask8
8695 (_mm256_mask_cmpneq_epu16_mask, _mm256_mask_cmplt_epu16_mask,
8696 _mm256_mask_cmpge_epu16_mask, _mm256_mask_cmple_epu16_mask): Change
8697 return type as well as __M argument type and all casts from __mmask8
8699 (_mm256_mask_cmpneq_epi8_mask, _mm256_mask_cmplt_epi8_mask,
8700 _mm256_mask_cmpge_epi8_mask, _mm256_mask_cmple_epi8_mask): Change
8701 return type as well as __M argument type and all casts from __mmask8
8703 (_mm256_mask_cmpneq_epi16_mask, _mm256_mask_cmplt_epi16_mask,
8704 _mm256_mask_cmpge_epi16_mask, _mm256_mask_cmple_epi16_mask): Change
8705 return type as well as __M argument type and all casts from __mmask8
8707 * config/i386/avx512vbmi2vlintrin.h (_mm_mask_shrdi_epi32,
8708 _mm_mask_shldi_epi32): Cast last argument to __mmask8 instead of
8711 2018-07-11 Grazvydas Ignotas <notasas@gmail.com>
8713 * config/i386/avx512bwintrin.h: (_mm512_mask_cmp_epi8_mask,
8714 _mm512_mask_cmp_epu8_mask): Use __mmask64 type instead of __mmask32
8717 2018-07-11 Paul Koning <ni1d@arrl.net>
8719 * doc/md.texi (define_subst): Document how multiple occurrences of
8720 the same argument in the replacement pattern are handled.
8722 2018-07-11 Paul Koning <ni1d@arrl.net>
8724 * doc/extend.texi (Common Variable Attributes): Move "mode" into
8726 (Common Type Attributes): Add "mode" attribute.
8728 2018-07-11 Jan Hubicka <hubicka@ucw.cz>
8730 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not
8731 stream DECL_ORIGINAL_TYPE.
8732 (DFS::DFS_write_tree_body): Drop hack handling local external decls.
8733 (hash_tree): Do not walk DECL_ORIGINAL_TYPE.
8734 * tree-streamer-in.c (lto_input_ts_decl_non_common_tree_pointers):
8735 Do not walk original type.
8736 * tree-streamer-out.c (streamer_write_chain): Drop hack handling
8738 (write_ts_decl_non_common_tree_pointers): Do not stream
8740 * tree.c (free_lang_data_in_decl): Clear DECL_ORIGINAL_TYPE.
8741 (find_decls_types_r): Do not walk DEC_ORIGINAL_TYPE.
8743 2018-07-11 Aldy Hernandez <aldyh@redhat.com>
8745 * tree-ssa-threadupdate.c (thread_through_all_blocks): Do not jump
8746 thread twice from the same starting edge.
8748 2018-07-11 Aldy Hernandez <aldyh@redhat.com>
8750 * vr-values.c (gimple_stmt_nonzero_p): Abstract common code to...
8751 * gimple.c (gimple_call_nonnull_result_p): ...here...
8752 (gimple_call_nonnull_arg): ...and here.
8753 * gimple.h (gimple_call_nonnull_result_p): New.
8754 (gimple_call_nonnull_arg): New.
8756 2018-07-11 Richard Earnshaw <rearnsha@arm.com>
8758 * config/arm/arm-cpus.in: Move information from fpu field of each
8759 cpu definition to the isa field.
8760 * config/arm/parsecpu.awk (fpu): Delete match rule.
8761 (gen_comm_data): Don't add bits from the CPU's FPU entry.
8763 2018-07-11 Richard Biener <rguenther@suse.de>
8766 * dwarf2out.c (init_sections_and_labels): Use
8767 output_asm_line_debug_info consistently.
8768 (dwarf2out_early_finish): Likewise.
8769 (dwarf2out_finish): Remove DW_AT_stmt_list from early generated
8772 2018-07-11 Richard Biener <rguenther@suse.de>
8774 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1):
8775 Rework father_bb setting in a way to avoid propagating constants
8776 multiple times on a loop body.
8778 2018-07-10 Mark Wielaard <mark@klomp.org>
8781 * dwarf2out.c (output_macinfo_op): Fix dwarf_FORM typo in gcc_assert.
8783 2018-07-10 Richard Biener <rguenther@suse.de>
8785 * hash-map.h (hash_map::iterator::operator*): Return
8786 references to key and value.
8788 2018-07-10 Jakub Jelinek <jakub@redhat.com>
8791 * gimplify.c (find_combined_omp_for): Add DATA argument, in addition
8792 to finding the inner OMP_FOR/OMP_SIMD stmt find non-trivial wrappers,
8793 BLOCKs with BLOCK_VARs, OMP_PARALLEL in between, OMP_FOR in between.
8794 (gimplify_omp_for): For composite loops, move outer
8795 OMP_{DISTRIBUTE,TASKLOOP,FOR,PARALLEL} right around innermost
8796 OMP_FOR/OMP_SIMD if there are any non-trivial wrappers. For class
8797 iterators add any needed clauses. Allow OMP_FOR_ORIG_DECLS to contain
8798 TREE_LIST for both the original class iterator and the "last" helper
8799 var. Gimplify OMP_FOR_PRE_BODY before the outermost composite
8800 loop, remember has_decl_expr from outer composite loops for the
8801 innermost OMP_SIMD in TREE_PRIVATE bit on OMP_FOR_INIT.
8803 2018-07-09 Martin Sebor <msebor@redhat.com>
8807 * builtins.c (c_strlen): Avoid out-of-bounds warnings when
8808 accessing implicitly initialized array elements.
8809 * expr.c (string_constant): Handle string initializers of
8810 character arrays within aggregates.
8811 * gimple-fold.c (fold_array_ctor_reference): Add argument.
8812 Store element offset. As a special case, handle zero size.
8813 (fold_nonarray_ctor_reference): Same.
8814 (fold_ctor_reference): Add argument. Store subobject offset.
8815 * gimple-fold.h (fold_ctor_reference): Add argument.
8817 2018-07-09 Paul Koning <ni1d@arrl.net>
8819 * config/pdp11/pdp11.c (pdp11_addr_cost): New function.
8820 (pdp11_insn_cost): New function.
8821 (pdp11_md_asm_adjust): New function.
8822 (TARGET_INVALID_WITHIN_DOLOOP): Define.
8823 (pdp11_rtx_costs): Update to match machine better.
8824 (output_addr_const_pdp11): Correct format mismatch warnings.
8825 * config/pdp11/pdp11.h (SLOW_BYTE_ACCESS): Correct definition.
8826 * config/pdp11/pdp11.md: General change to add base_cost and/or
8827 length attributes for use by new pdp11_insn_cost function.
8828 (MIN_BRANCH): Correct definition.
8830 (doloop_end): Use standard pattern name for looping pattern.
8831 (doloop_end_nocc): New.
8832 (movsf): Add another constraint alternative.
8833 (zero_extendqihi2): Add constraint alternatives for not in place
8835 (zero_extendhisi2): Remove.
8836 (shift patterns): Add CC handling variants.
8840 (define_peephole2): New peephole to recognize mov that sets CC for
8843 2018-07-09 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8845 * config/sparc/sparc.c (sparc_fold_builtin) <SPARC_BUILTIN_PDIST,
8846 SPARC_BUILTIN_PDISTN>: Adapt for signature change of wi::neg,
8849 2018-07-09 Jakub Jelinek <jakub@redhat.com>
8852 * real.c (real_nextafter): Return true if result is denormal.
8854 2018-07-09 Martin Liska <mliska@suse.cz>
8856 * common.opt: Add back wrongly removed attribute.
8858 2018-07-09 Richard Biener <rguenther@suse.de>
8861 * dwarf2out.c (gen_block_die): For an early generated DIE
8862 always output high/low PC attributes.
8864 2018-07-09 Tom de Vries <tdevries@suse.de>
8866 * cfgexpand.c (expand_debug_source_expr): Handle VAR_DECL.
8867 * tree-inline.c (remap_ssa_name): Handle default def ssa_name mapping
8868 onto VAR_DECL with abstract origin.
8870 2018-07-07 Jim Wilson <jimw@sifive.com>
8872 * config/riscv/riscv.c (TARGET_CUSTOM_FUNCTION_DESCRIPTORS): New.
8874 2018-07-07 Tom de Vries <tdevries@suse.de>
8876 * tree-dfa.c (dump_enumerated_decls): Handle cfun->cfg == NULL.
8878 2018-07-07 Aldy Hernandez <aldyh@redhat.com>
8880 * tree-vrp.c (vrp_int_const_binop): Change overflow type to
8882 (combine_bound): Use wide-int overflow calculation instead of
8884 * calls.c (maybe_warn_alloc_args_overflow): Change overflow type to
8886 * fold-const.c (int_const_binop_2): Same.
8887 (extract_muldiv_1): Same.
8888 (fold_div_compare): Same.
8889 (fold_abs_const): Same.
8891 * poly-int.h (add): Same.
8895 * predict.c (predict_iv_comparison): Same.
8896 * profile-count.c (slow_safe_scale_64bit): Same.
8897 * simplify-rtx.c (simplify_const_binary_operation): Same.
8898 * tree-chrec.c (tree_fold_binomial): Same.
8899 * tree-data-ref.c (split_constant_offset_1): Same.
8900 * tree-if-conv.c (idx_within_array_bound): Same.
8901 * tree-scalar-evolution.c (iv_can_overflow_p): Same.
8902 * tree-ssa-phiopt.c (minmax_replacement): Same.
8903 * tree-vect-loop.c (is_nonwrapping_integer_induction): Same.
8904 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Same.
8905 * vr-values.c (vr_values::adjust_range_with_scev): Same.
8906 * wide-int.cc (wi::add_large): Same.
8907 (wi::mul_internal): Same.
8908 (wi::sub_large): Same.
8909 (wi::divmod_internal): Same.
8910 * wide-int.h: Change overflow type to overflow_type for neg, add,
8911 mul, smul, umul, div_trunc, div_floor, div_ceil, div_round,
8912 mod_trunc, mod_ceil, mod_round, add_large, sub_large,
8913 mul_internal, divmod_internal.
8914 (overflow_type): New enum.
8915 (accumulate_overflow): New.
8917 2018-07-06 Kugan Vivekanandarajah <kugan.vivekanandarajah@linaro.org>
8919 * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): New.
8920 (tree_ssa_phiopt_worker): Call cond_removal_in_popcount_pattern.
8922 2018-07-06 Kugan Vivekanandarajah <kuganv@linaro.org>
8924 * tree-ssa-loop-niter.c (number_of_iterations_popcount): If popcount
8925 argument is checked for zero before entering loop, avoid checking again.
8927 2018-07-06 Kugan Vivekanandarajah <kuganv@linaro.org>
8929 * gimplify.h (generic_expr_could_trap_p): Set as global function.
8930 * gimplify.h (generic_expr_could_trap_p): Likwise.
8931 * tree-scalar-evolution.c (expression_expensive_p): Handle COND_EXPR.
8933 2018-07-06 Jakub Jelinek <jakub@redhat.com>
8935 PR tree-optimization/86401
8936 * fold-const.c (fold_binary_loc) <case BIT_AND_EXPR>: Move the
8937 ((A & N) + B) & M -> (A + B) & M etc. optimization into ...
8938 (fold_bit_and_mask): ... here. New helper function for match.pd.
8939 * fold-const.h (fold_bit_and_mask): Declare.
8940 * match.pd (((A & N) + B) & M -> (A + B) & M): New optimization.
8942 2018-07-06 Peter Bergner <bergner@linux.ibm.com>
8945 * target.def (translate_mode_attribute): New hook.
8946 * targhooks.h (default_translate_mode_attribute): Declare.
8947 * targhooks.c (default_translate_mode_attribute): New function.
8948 * doc/tm.texi.in (TARGET_TRANSLATE_MODE_ATTRIBUTE): New hook.
8949 * doc/tm.texi: Regenerate.
8950 * config/rs6000/rs6000.c (TARGET_TRANSLATE_MODE_ATTRIBUTE): Define.
8951 (rs6000_translate_mode_attribute): New function.
8953 2018-07-06 Paul Koning <ni1d@arrl.net>
8955 * doc/md.texi (define_split): Document DONE and FAIL.
8956 (define_peephole2): Ditto.
8958 2018-07-05 Jeff Law <law@redhat.com>
8960 PR tree-optimization/86010
8961 * tree-ssa-dse.c (compute_trims): More aggressively trim at
8962 both the head and tail of mem* and str* calls.
8964 2018-07-05 Jim Wilson <jimw@sifive.com>
8966 * config.gcc (riscv*-*-*): When setting xlen, handle riscv-*.
8968 2018-07-05 Indu Bhagat <indu.bhagat@oracle.com>
8970 * config/aarch64/aarch64-simd.md: correct flags text for
8971 MIN_EXPR replacement.
8973 2018-07-05 James Clarke <jrtc27@jrtc27.com>
8975 * configure: Regenerated.
8977 2018-07-05 Carl Love <cel@us.ibm.com>
8979 * config/rs6000/rs6000-c.c: Map ALTIVEC_BUILTIN_VEC_UNPACKH for
8980 float argument to VSX_BUILTIN_DOUBLEH_V4SF.
8981 Map ALTIVEC_BUILTIN_VEC_UNPACKL for float argument to
8982 VSX_BUILTIN_DOUBLEL_V4SF.
8984 2018-07-05 Martin Sebor <msebor@redhat.com>
8987 * tree-ssa-strlen.c (maybe_set_strlen_range): Use type size rather
8988 than its domain to compute its the upper bound of a char array.
8990 2018-07-05 Nathan Sidwell <nathan@acm.org>
8992 Replace NO_IMPLICIT_EXTERN_C with SYSTEM_IMPLICIT_EXTERN_C.
8993 * doc/cpp.texi: Update comment.
8994 * doc/tm.texi: Rebuilt.
8995 * doc/tm.texi.in (NO_IMPLICIT_EXTERN_C): Replace with ...
8996 (SYSTEM_IMPLICIT_EXTERN_C): ... this, opposite sense.
8997 * doc/extend.texi (Backwards Compatibility): Clarify it is system
8998 headers affected by extern "C".
8999 * system.h: Poison NO_IMPLICIT_EXTERN_C.
9000 * config/alpha/alpha.h, config/arm/uclinux-elf.h,
9001 config/bfin/elf.h, config/cris/cris.h, config/darwin.h,
9002 config/dragonfly.h, config/freebsd.h, config/gnu-user.h,
9003 config/i386/cygming.h, config/i386/djgpp.h, config/i386/nto.h,
9004 config/ia64/hpux.h, config/lm32/lm32.h, config/lm32/uclinux-elf.h,
9005 config/lynx.h, config/mips/elf.h, config/mmix/mmix.h,
9006 config/netbsd.h, config/pa/pa-hpux.h, config/powerpcspe/sysv4.h,
9007 config/riscv/elf.h, config/rs6000/sysv4.h, config/rtems.h,
9008 config/s390/tpf.h, config/sh/newlib.h, config/sol2.h,
9009 config/sparc/openbsd64.h, config/sparc/sp-elf.h,
9010 config/sparc/sp64-elf.h, config/spu/spu.h,
9011 config/stormy16/stormy16.h, config/v850/v850.h,
9012 config/visium/visium.h, config/vx-common.h, config/xtensa/elf.h: Don't
9013 define NO_IMPLICIT_EXTERN_C.
9014 * config/rs6000/aix.h: Set SYSTEM_IMPLICIT_EXTERN_C.
9016 2018-07-05 Tamar Christina <tamar.christina@arm.com>
9019 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
9020 instead of GET_MODE_SIZE when comparing Units.
9022 2018-07-05 Tamar Christina <tamar.christina@arm.com>
9025 * rtlanal.c (set_noop_p): Constrain on mode change,
9026 include hard-reg-set.h
9028 2018-07-05 Tamar Christina <tamar.christina@arm.com>
9030 * config/aarch64/aarch64.c (aarch64_expand_movmem): Fix mode size.
9032 2018-07-05 Jakub Jelinek <jakub@redhat.com>
9035 2018-07-04 Maxim Ostapenko <m.ostapenko@samsung.com>
9038 * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
9040 * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
9042 2018-07-04 Maxim Ostapenko <m.ostapenko@samsung.com>
9045 * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
9047 * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
9049 2018-07-04 Eric Botcazou <ebotcazou@adacore.com>
9052 * expmed.c (choose_multiplier): Fix incorrect comparison with mask.
9054 2018-07-04 Aldy Hernandez <aldyh@redhat.com>
9056 * tree-vrp.c (extract_range_from_binary_expr_1): Initialize
9057 neg_*_op* variables.
9059 2018-07-04 Martin Liska <mliska@suse.cz>
9061 * tree-switch-conversion.c: Define
9062 max_ratio_for_speed and max_ratio_for_size constants.
9064 2018-07-04 Denys Vlasenko <dvlasenk@redhat.com>
9065 Martin Liska <mliska@suse.cz>
9070 * common.opt: Rename align options with 'str_' prefix.
9071 * common/config/i386/i386-common.c (set_malign_value): New
9073 (ix86_handle_option): Use it to set -falign-* options/
9074 * config/aarch64/aarch64-protos.h (struct tune_params): Change
9075 type from int to string.
9076 * config/aarch64/aarch64.c: Update default values from int
9078 * config/alpha/alpha.c (alpha_override_options_after_change):
9080 * config/arm/arm.c (arm_override_options_after_change_1): Likewise.
9081 * config/i386/dragonfly.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
9082 max skip conditionally.
9083 * config/i386/freebsd.h (SUBALIGN_LOG): New.
9084 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
9085 max skip conditionally.
9086 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
9087 max skip conditionally.
9088 * config/i386/gnu-user.h (SUBALIGN_LOG): New.
9089 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
9090 max skip conditionally.
9091 * config/i386/i386.c (struct ptt): Change type from int to
9093 (ix86_default_align): Set default values.
9094 * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Print
9095 max skip conditionally.
9096 * config/i386/iamcu.h (SUBALIGN_LOG): New.
9097 (ASM_OUTPUT_MAX_SKIP_ALIGN):
9098 * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN):
9099 * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
9100 max skip conditionally.
9101 * config/i386/openbsdelf.h (SUBALIGN_LOG): New.
9102 (ASM_OUTPUT_MAX_SKIP_ALIGN) Print max skip conditionally.:
9103 * config/i386/x86-64.h (SUBALIGN_LOG): New.
9104 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
9105 max skip conditionally.
9106 (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
9107 * config/ia64/ia64.c (ia64_option_override): Set default values
9108 for alignment options.
9109 * config/m68k/m68k.c: Handle new str_align_* options.
9110 * config/mips/mips.c (mips_set_compression_mode): Change
9112 (mips_option_override): Set default values for options.
9113 * config/powerpcspe/powerpcspe.c (rs6000_option_override_internal):
9115 * config/rs6000/rs6000.c (rs6000_option_override_internal):
9117 * config/rx/rx.c (rx_option_override): Likewise.
9118 * config/rx/rx.h (JUMP_ALIGN): Use align_jumps_log.
9119 (LABEL_ALIGN): Use align_labels_log.
9120 (LOOP_ALIGN): Use align_loops_align.
9121 * config/s390/s390.c (s390_asm_output_function_label): Use new
9123 * config/sh/sh.c (sh_override_options_after_change):
9124 Change type of constants.
9125 * config/spu/spu.c (spu_sched_init): Likewise.
9126 * config/sparc/sparc.c (sparc_option_override): Set default
9128 * config/visium/visium.c (visium_option_override): Likewise.
9129 * config/visium/visium.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Do not
9130 emit p2align format with last argument if it's not needed.
9131 * doc/invoke.texi: Document extended format of -falign-*.
9132 * final.c: Use align_labels alignment.
9133 * flags.h (struct target_flag_state): Change type to use
9135 (struct align_flags_tuple): New.
9136 (struct align_flags): Likewise.
9137 (align_loops_log): Redefine macro to use new types.
9138 (align_loops_max_skip): Redefine macro to use new types.
9139 (align_jumps_log): Redefine macro to use new types.
9140 (align_jumps_max_skip): Redefine macro to use new types.
9141 (align_labels_log): Redefine macro to use new types.
9142 (align_labels_max_skip): Redefine macro to use new types.
9143 (align_functions_log): Redefine macro to use new types.
9144 (align_loops): Redefine macro to use new types.
9145 (align_jumps): Redefine macro to use new types.
9146 (align_labels): Redefine macro to use new types.
9147 (align_functions): Redefine macro to use new types.
9148 (align_functions_max_skip): Redefine macro to use new types.
9149 (align_loops_value): New macro.
9150 (align_jumps_value): New macro.
9151 (align_labels_value): New macro.
9152 (align_functions_value): New macro.
9153 * function.c (invoke_set_current_function_hook): Propagate
9154 alignment values from flags to global variables default in
9156 * ipa-icf.c (sem_function::equals_wpa): Use
9157 cl_optimization_option_eq instead of memcmp.
9158 * lto-streamer.h (cl_optimization_stream_out): Support streaming
9160 (cl_optimization_stream_in): Likewise.
9161 * optc-save-gen.awk: Support strings in cl_optimization.
9162 * opth-gen.awk: Likewise.
9163 * opts.c (finish_options): Remove error checking of invalid
9165 (MAX_CODE_ALIGN): Remove.
9166 (MAX_CODE_ALIGN_VALUE): Likewise.
9167 (parse_and_check_align_values): New function.
9168 (check_alignment_argument): Likewise.
9169 (common_handle_option): Use check_alignment_argument.
9170 * opts.h (parse_and_check_align_values): Declare.
9171 * toplev.c (init_alignments): Remove.
9172 (read_log_maxskip): New.
9173 (parse_N_M): Likewise.
9174 (parse_alignment_opts): Likewise.
9175 (backend_init_target): Remove usage of init_alignments.
9176 * toplev.h (parse_alignment_opts): Declare.
9177 * tree-streamer-in.c (streamer_read_tree_bitfields): Add new
9179 * tree-streamer-out.c (streamer_write_tree_bitfields): Likewise.
9180 * tree.c (cl_option_hasher::equal): New.
9181 * varasm.c: Use new global macros.
9183 2018-07-04 Denys Vlasenko <dvlasenk@redhat.com>
9185 * config/i386/dragonfly.h: (ASM_OUTPUT_MAX_SKIP_ALIGN):
9186 Use a simpler align directive also if MAXSKIP = ALIGN-1.
9187 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
9188 * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
9189 * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
9190 * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
9191 * config/i386/freebsd.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Remove "If N
9192 is large, do at least 8 byte alignment" code. Add SUBALIGN_LOG
9193 define. Use a simpler align directive also if MAXSKIP = ALIGN-1.
9194 * config/i386/gnu-user.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
9195 * config/i386/iamcu.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
9196 * config/i386/openbsdelf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
9197 * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
9199 2018-07-04 Martin Liska <mliska@suse.cz>
9200 Jonathan Wakely <jwakely@redhat.com>
9202 * coverage.c: Use correct type.
9203 * doc/invoke.texi: Language correction.
9205 2018-07-03 H.J. Lu <hongjiu.lu@intel.com>
9208 * config/i386/i386.c (rest_of_insert_endbranch): Also generate
9209 ENDBRANCH for non-tail call which may return via indirect branch.
9210 * doc/extend.texi: Document indirect_return attribute.
9212 2018-07-03 Martin Sebor <msebor@redhat.com>
9214 PR tree-optimization/86274
9215 * gimple-ssa-sprintf.c (fmtresult::type_max_digits): Verify
9217 (format_floating): Correct handling of infinities and NaNs.
9219 2018-07-03 Martin Sebor <msebor@redhat.com>
9221 * print-tree.c (print_real_cst): New function.
9222 (print_node_brief): Call it.
9223 (print_node): Ditto.
9225 2018-07-03 Jeff Law <law@redhat.com>
9227 * config/h8300/h8300.md (logical<mode>3_sn, logical<mode>3): Merge
9228 into a single pattern.
9230 * config/h8300/h8300.md (ors code_iterator): New.
9231 (bsetqi_msx, bnotqi_msx patterns and splitters): Consolidate into
9232 a single pattern and single splitter.
9233 (bsethi_msx, bnothi_msx patterns): Consolidate into a single pattern.
9234 (iorqi3_1, xorqi3_1): Likewise.
9235 (iorqi3, xorqi3 expanders): Similarly.
9237 * config/h8300/h8300.md (movmd_internal_normal): Consolidated with
9238 (movmd_internal) into a single pattern using the P mode iterator.
9239 (movmd splitters): Similarly.
9240 (stpcpy_internal_normal, stpcpy_internal): Similarly for thes patterns.
9241 (movsd splitters): Similarly.
9243 * config/h8300/h8300.c (h8300_insn_length_from_table): Consolidate
9244 ADDB, ADDW and ADDL into a single ADD attribute which selects the
9245 right table based on the size of the operand.
9246 * config/h8300/h8300.md (length_table): Corresponding changes. All
9247 references to "addb", "addw" and "addl" changed to "add".
9248 (btst patterns): Merge two variants into a single pattern.
9249 (tstqi, tsthi): Likewise.
9250 (addhi3_incdec, addsi3_incdec): Likewise.
9251 (subhi3_h8300hs, subsi3_h8300hs): Likewise.
9252 (mulhi3, mulsi3): Likewise.
9253 (udivhi3, udivsi3): Likewise.
9254 (divhi3, divsi3): Likewise.
9255 (andorqi3, andorhi3, andorsi3): Likewise.
9257 2018-07-03 Uros Bizjak <ubizjak@gmail.com>
9260 * config/i386/sse.md (uavg<mode>3_ceil): New expander.
9261 (<sse2_avx2>_uavg<mode>3<mask_name>): Simplify expander.
9263 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
9265 PR tree-optimization/85694
9266 * config/aarch64/iterators.md (HADD, RHADD): New int iterators.
9267 (u): Handle UNSPEC_SHADD, UNSPEC_UHADD, UNSPEC_SRHADD and
9269 * config/aarch64/aarch64-simd.md (<u>avg<mode>3_floor)
9270 (<u>avg<mode>3_ceil): New patterns.
9272 2018-07-03 David Malcolm <dmalcolm@redhat.com>
9274 * gcc.dg/vect/slp-perm-1.c: Remove "note: " prefix from
9275 scan-tree-dump directive.
9276 * gcc.dg/vect/slp-perm-2.c: Likewise.
9277 * gcc.dg/vect/slp-perm-3.c: Likewise.
9278 * gcc.dg/vect/slp-perm-5.c: Likewise.
9279 * gcc.dg/vect/slp-perm-6.c: Likewise.
9280 * gcc.dg/vect/slp-perm-7.c: Likewise.
9281 * gcc.dg/vect/slp-perm-8.c: Likewise.
9283 2018-07-03 Marek Polacek <polacek@redhat.com>
9286 * gimple-fold.c (size_must_be_zero_p): Check the type of the size.
9288 2018-07-03 Richard Biener <rguenther@suse.de>
9291 * tree-ssa-structalias.c (find_func_clobbers): Properly
9292 handle indirect calls.
9294 2018-07-03 Jeff Law <law@redhat.com>
9296 * config/h8300/h8300.md (HSI, QHSI, QHSIF): New mode iterators.
9297 (shifts): New code iterator.
9298 (movqi, movhi, movsi, movsf expanders): Consolidate into a single
9299 expander. Fix HImode handling on H8/SX.
9300 (addqi3, addhi3, addsi3 expanders): Consolidate into a single expander.
9301 (subqi3, subhi3, subsi3 expanders): Likewise.
9302 (andqi3, andhi3, andsi3 expanders): Likewise.
9303 (iorqi3, iorhi3, iorsi3 expanders): Likewise.
9304 (xorqi3, xorhi3, xorsi3 expanders): Likewise.
9305 (negqi2, neghi2, negsi2, negsf2 expanders): Likewise.
9306 (one_cmplqi2, one_cmplhi2, one_cmplsi2): Likewise.
9307 (zero_extendqihi2, zero_extendqisi2): Likewise.
9308 (extendqihi2, extendqisi2): Likewise.
9309 (rotlqi3, rotlhi3, rotlsi3): Likewise.
9310 (neghi2_h8300, negsi2_h8300): Likewise for these patterns.
9311 (rotlqi3_1, rotlhi3_1): Likewise.
9312 (logicalhi3_sn, logicalsi3_sn): Likewise.
9313 (logicalhi3, logicalsi3): Likewise.
9315 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
9317 * tree-vect-patterns.c (vect_recog_rotate_pattern)
9318 (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
9319 (vect_recog_mixed_size_cond_pattern, adjust_bool_pattern_cast)
9320 (adjust_bool_pattern, vect_recog_bool_pattern): Pass the vector
9321 type to append_pattern_def_seq instead of creating a stmt_vec_info
9323 (build_mask_conversion): Likewise. Remove vinfo argument.
9324 (vect_add_conversion_to_patterm): Likewise, renaming to...
9325 (vect_add_conversion_to_pattern): ...this.
9326 (vect_recog_mask_conversion_pattern): Update call to
9327 build_mask_conversion. Pass the vector type to
9328 append_pattern_def_seq here too.
9329 (vect_recog_gather_scatter_pattern): Update call to
9330 vect_add_conversion_to_pattern.
9332 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
9334 * tree-vect-patterns.c (new_pattern_def_seq): Delete.
9335 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
9336 (vect_recog_widen_op_pattern, vect_recog_over_widening_pattern)
9337 (vect_recog_rotate_pattern, vect_synth_mult_by_constant): Don't set
9338 STMT_VINFO_PATTERN_DEF_SEQ to null here.
9339 (vect_recog_pow_pattern, vect_recog_vector_vector_shift_pattern)
9340 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern): Use
9341 append_pattern_def_seq instead of new_pattern_def_seq.
9342 (vect_recog_divmod_pattern): Do both of the above.
9343 (vect_pattern_recog_1): Assert that STMT_VINO_PATTERN_DEF_SEQ
9346 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
9348 * tree-vect-patterns.c (vect_recog_dot_prod_pattern):
9349 (vect_recog_sad_pattern, vect_recog_widen_op_pattern)
9350 (vect_recog_widen_mult_pattern, vect_recog_pow_pattern):
9351 (vect_recog_widen_sum_pattern, vect_recog_over_widening_pattern)
9352 (vect_recog_average_pattern, vect_recog_cast_forwprop_pattern)
9353 (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
9354 (vect_recog_vector_vector_shift_pattern, vect_synth_mult_by_constant)
9355 (vect_recog_mult_pattern, vect_recog_divmod_pattern)
9356 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
9357 (vect_recog_mask_conversion_pattern): Replace vec<gimple *>
9358 parameter with a single stmt_vec_info.
9359 (vect_recog_func_ptr): Likewise.
9360 (vect_recog_gather_scatter_pattern): Likewise, folding in...
9361 (vect_try_gather_scatter_pattern): ...this.
9362 (vect_pattern_recog_1): Remove stmts_to_replace and just pass
9363 the stmt_vec_info of the statement to be matched. Don't clear
9364 STMT_VINFO_RELATED_STMT.
9365 (vect_pattern_recog): Update call accordingly.
9367 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
9369 PR tree-optimization/85694
9370 * doc/md.texi (avgM3_floor, uavgM3_floor, avgM3_ceil)
9371 (uavgM3_ceil): Document new optabs.
9372 * doc/sourcebuild.texi (vect_avg_qi): Document new target selector.
9373 * internal-fn.def (IFN_AVG_FLOOR, IFN_AVG_CEIL): New internal
9375 * optabs.def (savg_floor_optab, uavg_floor_optab, savg_ceil_optab)
9376 (savg_ceil_optab): New optabs.
9377 * tree-vect-patterns.c (vect_recog_average_pattern): New function.
9378 (vect_vect_recog_func_ptrs): Add it.
9379 * tree-vect-stmts.c (vectorizable_call): Get the type of the zero
9380 constant directly from the associated lhs.
9382 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
9384 * tree-vect-patterns.c (vect_split_statement): New function.
9385 (vect_convert_input): Use it to try to split an existing cast.
9387 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
9389 * poly-int.h (print_hex): New function.
9390 * dumpfile.h (dump_dec, dump_hex): Declare.
9391 * dumpfile.c (dump_dec, dump_hex): New poly_wide_int functions.
9392 * tree-vectorizer.h (_stmt_vec_info): Add min_output_precision,
9393 min_input_precision, operation_precision and operation_sign.
9394 * tree-vect-patterns.c (vect_get_range_info): New function.
9395 (vect_same_loop_or_bb_p, vect_single_imm_use)
9396 (vect_operation_fits_smaller_type): Delete.
9397 (vect_look_through_possible_promotion): Add an optional
9398 single_use_p parameter.
9399 (vect_recog_over_widening_pattern): Rewrite to use new
9400 stmt_vec_info infomration. Handle one operation at a time.
9401 (vect_recog_cast_forwprop_pattern, vect_narrowable_type_p)
9402 (vect_truncatable_operation_p, vect_set_operation_type)
9403 (vect_set_min_input_precision): New functions.
9404 (vect_determine_min_output_precision_1): Likewise.
9405 (vect_determine_min_output_precision): Likewise.
9406 (vect_determine_precisions_from_range): Likewise.
9407 (vect_determine_precisions_from_users): Likewise.
9408 (vect_determine_stmt_precisions, vect_determine_precisions): Likewise.
9409 (vect_vect_recog_func_ptrs): Put over_widening first.
9411 (vect_pattern_recog): Call vect_determine_precisions.
9413 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
9415 * tree-vect-patterns.c (vect_mark_pattern_stmts): Remove pattern
9416 statements that have been replaced by further pattern statements.
9417 (vect_pattern_recog_1): Clear STMT_VINFO_PATTERN_DEF_SEQ on failure.
9419 2018-07-03 Richard Biener <rguenther@suse.de>
9421 * tree-vect-stmts.c (vect_is_simple_use): Consolidate dumping,
9422 always set *dt. Dump vectype in vectype overload.
9423 * dumpfile.h (dump_gimple_expr): New function.
9424 (dump_gimple_expr_loc): Likewise.
9425 * dumpfile.c (dump_gimple_expr): New function.
9426 (dump_gimple_expr_loc): Likewise.
9428 2018-07-02 Jeff Law <law@redhat.com>
9430 * config/h8300/h8300.md (movqi_h8300, movqi_h8300hs): Consolidate
9431 the H8/300, H8/300H and H8/S variants into a single pattern.
9432 (movhi_h8300, movqi_h8300hs): Similarly.
9433 (pushqi_h8300hs, pushhi_h8300hs): Consolidate into a single pattern.
9434 (QHI mode iterator): New.
9436 * config/h8300/h8300.md: Remove trailing whitespace.
9438 2018-07-02 Jim Wilson <jimw@sifive.com>
9440 * config/riscv/riscv.c (riscv_expand_epilogue): Use emit_jump_insn
9441 instead of emit_insn for interrupt returns.
9442 * config/riscv/riscv.md (riscv_met): Add (return) to rtl.
9443 (riscv_sret, riscv_uret): Likewise.
9445 2018-07-02 David Malcolm <dmalcolm@redhat.com>
9447 * pretty-print.c (selftest::test_pp_format): Move save and restore
9448 of quotes to class auto_fix_quotes, and add an instance.
9449 * selftest.c: Include "intl.h".
9450 (selftest::auto_fix_quotes::auto_fix_quotes): New ctor.
9451 (selftest::auto_fix_quotes::~auto_fix_quotes): New dtor.
9452 * selftest.h (selftest::auto_fix_quotes): New class.
9454 2018-07-02 Richard Henderson <richard.henderson@linaro.org>
9456 * config/aarch64/aarch64-protos.h, config/aarch64/aarch64.c
9457 (aarch64_sve_prepare_conditional_op): Remove.
9458 * config/aarch64/aarch64-sve.md (cond_<SVE_INT_BINARY><SVE_I>):
9459 Allow aarch64_simd_reg_or_zero as select operand; remove
9460 the aarch64_sve_prepare_conditional_op call.
9461 (cond_<SVE_INT_BINARY_SD><SVE_SDI>): Likewise.
9462 (cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
9463 (*cond_<SVE_INT_BINARY><SVE_I>_z): New pattern.
9464 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_z): New pattern.
9465 (*cond_<SVE_COND_FP_BINARY><SVE_F>_z): New pattern.
9466 (*cond_<SVE_INT_BINARY><SVE_I>_any): New pattern.
9467 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_any): New pattern.
9468 (*cond_<SVE_COND_FP_BINARY><SVE_F>_any): New pattern
9469 and a splitters to match all of the *_any patterns.
9470 * config/aarch64/predicates.md (aarch64_sve_any_binary_operator): New.
9472 * config/aarch64/iterators.md (SVE_INT_BINARY_REV): Remove.
9473 (SVE_COND_FP_BINARY_REV): Remove.
9474 (sve_int_op_rev, sve_fp_op_rev): New.
9475 * config/aarch64/aarch64-sve.md (*cond_<SVE_INT_BINARY><SVE_I>_0): New.
9476 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_0): New.
9477 (*cond_<SVE_COND_FP_BINARY><SVE_F>_0): New.
9478 (*cond_<SVE_INT_BINARY><SVE_I>_2): Rename, add movprfx alternative.
9479 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_2): Similarly.
9480 (*cond_<SVE_COND_FP_BINARY><SVE_F>_2): Similarly.
9481 (*cond_<SVE_INT_BINARY><SVE_I>_3): Similarly; use sve_int_op_rev.
9482 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_3): Similarly.
9483 (*cond_<SVE_COND_FP_BINARY><SVE_F>_3): Similarly; use sve_fp_op_rev.
9485 * config/aarch64/aarch64-sve.md (cond_<SVE_COND_FP_BINARY><SVE_F>):
9486 Remove match_dup 1 from the inner unspec.
9487 (*cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
9489 * config/aarch64/aarch64.md (movprfx): New attr.
9490 (length): Default movprfx to 8.
9491 * config/aarch64/aarch64-sve.md (*mul<SVE_I>3): Add movprfx alt.
9492 (*madd<SVE_I>, *msub<SVE_I): Likewise.
9493 (*<su>mul<SVE_I>3_highpart): Likewise.
9494 (*<SVE_INT_BINARY_SD><SVE_SDI>3): Likewise.
9495 (*v<ASHIFT><SVE_I>3): Likewise.
9496 (*<su><MAXMIN><SVE_I>3): Likewise.
9497 (*<su><MAXMIN><SVE_F>3): Likewise.
9498 (*fma<SVE_F>4, *fnma<SVE_F>4): Likewise.
9499 (*fms<SVE_F>4, *fnms<SVE_F>4): Likewise.
9500 (*div<SVE_F>4): Likewise.
9502 2018-07-02 Richard Sandiford <richard.sandiford@arm.com>
9504 * tree-vect-patterns.c (vect_recog_widen_shift_pattern): Fix typo
9507 2018-07-02 Richard Biener <rguenther@suse.de>
9509 PR tree-optimization/86363
9510 * tree-ssa-sccvn.c (vn_reference_lookup_3): Check the
9511 memset argument refers to a non-variable address.
9513 2018-07-02 Aldy Hernandez <aldyh@redhat.com>
9515 * tree-vrp.c (extract_range_from_binary_expr_1): Abstract a lot of the
9516 {PLUS,MINUS}_EXPR code to...
9517 (adjust_symbolic_bound): ...here,
9518 (combine_bound): ...here,
9519 (set_value_range_with_overflow): ...and here.
9521 2018-07-02 Aldy Hernandez <aldyh@redhat.com>
9523 * tree-vrp.c (extract_range_from_unary_expr): Abstract ABS_EXPR
9525 (extract_range_from_abs_expr): ...here.
9527 2018-07-02 Eric Botcazou <ebotcazou@adacore.com>
9529 * config/i386/i386.c (ix86_finalize_stack_frame_flags): Do not overrule
9530 -fno-omit-frame-pointer when not optimizing.
9532 2018-07-02 Martin Liska <mliska@suse.cz>
9535 * ipa-pure-const.c (malloc_candidate_p): Revert usage of ::get.
9536 (propagate_nothrow): Likewise.
9538 2018-07-02 Martin Liska <mliska@suse.cz>
9541 * ipa-inline.c (early_inliner): Revert wrongly added ::get call.
9543 2018-07-02 David Malcolm <dmalcolm@redhat.com>
9545 * dumpfile.c (dump_generic_expr_loc): Undo removal of this
9546 function in r262149, changing "loc" param from source_location to
9547 const dump_location_t &.
9548 * dumpfile.h (dump_generic_expr_loc): Undo removal of this
9549 declaration, as above.
9551 2018-07-01 Paul Koning <ni1d@arrl.net>
9553 * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
9554 -munit-asm, -mgnu-asm, -mdec-asm.
9555 * config/pdp11/pdp11-protos.h (pdp11_gen_int_label): New.
9556 (pdp11_output_labelref): New.
9557 (pdp11_output_def): New.
9558 (pdp11_output_addr_vec_elt): New.
9559 * config/pdp11/pdp11.c: Use tab between opcode and operands. Use
9560 %# and %@ format codes.
9561 (pdp11_option_override): New.
9562 (TARGET_ASM_FILE_START_FILE_DIRECTIVE): Define.
9563 (pdp11_output_ident): New.
9564 (pdp11_asm_named_section): New.
9565 (pdp11_asm_init_sections): New.
9566 (pdp11_file_start): New.
9567 (pdp11_file_end): New.
9568 (output_ascii): Use .ascii/.asciz for -mdec-asm.
9569 (pdp11_asm_print_operand): Update %# and %$ for -mdec-asm. Add
9570 %o, like %c but octal.
9571 (pdp11_option_override): New.
9572 * config/pdp11/pdp11.h (TEXT_SECTION_ASM_OP): Update for
9574 (DATA_SECTION_ASM_OP): Ditto.
9575 (READONLY_DATA_SECTION_ASM_OP): New.
9576 (IS_ASM_LOGICAL_LINE_SEPARATOR): New.
9577 (ASM_GENERATE_INTERNAL_LABEL): Use new function.
9578 (ASM_OUTPUT_LABELREF): Ditto.
9579 (ASM_OUTPUT_DEF): Ditto.
9580 (ASM_OUTPUT_EXTERNAL): New.
9581 (ASM_OUTPUT_SOURCE_FILENAME): New.
9582 (ASM_OUTPUT_ADDR_VEC_ELT): Use new function.
9583 (ASM_OUTPUT_SKIP): Update for -mdec-asm.
9584 * config/pdp11/pdp11.md: Use tab between opcode and operands. Use
9585 %# and %@ format codes.
9586 * config/pdp11/pdp11.opt (mgnu-asm): New.
9587 (mdec-asm): Conflicts with -mgnu-asm and -munix-asm.
9588 (munix-asm): Conflicts with -mdec-asm and -mgnu-asm.
9589 * doc/invoke.txt (PDP-11 Options): Add -mgnu-asm.
9591 2018-07-01 Aldy Hernandez <aldyh@redhat.com>
9593 * tree-ssa-threadupdate.c (mark_threaded_blocks): Avoid
9594 dereferencing path[] beyond its length.
9596 (debug_all_paths): New.
9597 (rewire_first_differing_edge): New.
9598 (adjust_paths_after_duplication): New.
9599 (duplicate_thread_path): Call adjust_paths_after_duplication.
9601 (thread_through_all_blocks): Add new argument to
9602 duplicate_thread_path.
9604 2018-06-30 Jim Wilson <jimw@sifive.com>
9606 * config/riscv/predicates.md (p2m1_shift_operand): New.
9607 (high_mask_shift_operand): New.
9608 * config/riscv/riscv.md (lshrsi3_zero_extend_3+1): New combiner
9609 pattern using p2m1_shift_operand.
9610 (lshsi3_zero_extend_3+2): New combiner pattern using
9611 high_mask_shift_operand.
9613 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
9615 * tree-vect-patterns.c (vect_get_external_def_edge): New function,
9617 (vect_recog_rotate_pattern): ...here.
9618 (vect_convert_input): Try to insert casts of invariants in the
9620 * tree-vect-loop-manip.c (vect_loop_versioning): Don't require the
9621 preheader to be empty.
9623 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
9625 * tree-vect-patterns.c (append_pattern_def_seq): Take an optional
9626 vector type. If given, install it in the new statement's
9628 (vect_element_precision): New function.
9629 (vect_unpromoted_value): New struct.
9630 (vect_unpromoted_value::vect_unpromoted_value): New function.
9631 (vect_unpromoted_value::set_op): Likewise.
9632 (vect_look_through_possible_promotion): Likewise.
9633 (vect_joust_widened_integer, vect_joust_widened_type): Likewise.
9634 (vect_widened_op_tree, vect_convert_input): Likewise.
9635 (vect_convert_inputs, vect_convert_output): Likewise.
9636 (vect_recog_dot_prod_pattern): Use vect_look_through_possible_promotion
9637 to handle the optional cast of the multiplication result and
9638 vect_widened_op_tree to detect the widened multiplication itself.
9639 Do not require the input and output of promotion casts to have
9640 the same sign, but base the signedness of the operation on the
9641 input rather than the result. If the pattern includes two
9642 promotions, check that those promotions have the same sign.
9643 Do not restrict the MULT_EXPR handling to a double-width result;
9644 handle quadruple-width results and wider. Use vect_convert_inputs
9645 to convert the inputs to the common type.
9646 (vect_recog_sad_pattern): Use vect_look_through_possible_promotion
9647 to handle the optional cast of the ABS result. Also allow a sign
9648 change or a sign extension between the ABS and MINUS.
9649 Use vect_widened_op_tree to detect the widened subtraction and use
9650 vect_convert_inputs to convert the inputs to the common type.
9651 (vect_handle_widen_op_by_const): Delete.
9652 (vect_recog_widen_op_pattern): New function.
9653 (vect_recog_widen_mult_pattern): Use it.
9654 (vect_recog_widen_shift_pattern): Likewise.
9655 (vect_recog_widen_sum_pattern): Use
9656 vect_look_through_possible_promotion to handle the promoted
9659 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
9661 * gimple-iterator.c (gsi_for_stmt): Add a new overload that takes
9662 the containing gimple_seq *.
9663 * gimple-iterator.h (gsi_for_stmt): Declare it.
9664 * tree-vect-patterns.c (vect_recog_dot_prod_pattern)
9665 (vect_recog_sad_pattern, vect_recog_widen_sum_pattern)
9666 (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
9667 (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
9668 (vect_recog_mask_conversion_pattern): Remove STMT_VINFO_IN_PATTERN_P
9670 (vect_init_pattern_stmt, vect_set_pattern_stmt): New functions,
9672 (vect_mark_pattern_stmts): ...here. Handle cases in which the
9673 statement being replaced is part of an existing pattern
9674 definition sequence, inserting the new pattern statements before
9676 (vect_pattern_recog_1): Don't return a bool. If the statement
9677 is already part of a pattern, instead apply pattern matching
9678 to the pattern definition statements. Don't clear the
9679 STMT_VINFO_RELATED_STMT if is_pattern_stmt_p.
9680 (vect_pattern_recog): Don't break after the first match;
9681 continue processing the pattern definition statements instead.
9682 Don't bail out for STMT_VINFO_IN_PATTERN_P here.
9684 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
9686 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
9687 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
9688 (vect_recog_widen_sum_pattern): Use it.
9690 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
9692 * tree-vect-loop.c (vectorizable_reduction): Assert that the
9693 phi is not a pattern statement and has not been replaced by
9694 a pattern statement.
9695 * tree-vect-patterns.c (type_conversion_p): Don't check
9696 STMT_VINFO_IN_PATTERN_P.
9697 (vect_recog_vector_vector_shift_pattern): Likewise.
9698 (vect_recog_dot_prod_pattern): Expect vect_is_simple_use to return
9699 the pattern statement rather than the original statement; check
9700 directly for a WIDEN_MULT_EXPR here.
9701 * tree-vect-slp.c (vect_get_and_check_slp_defs): Expect
9702 vect_is_simple_use to return the pattern statement rather
9703 than the original statement; use is_pattern_stmt_p to check
9704 for such a pattern statement.
9705 * tree-vect-stmts.c (process_use): Expect vect_is_simple_use
9706 to return the pattern statement rather than the original statement;
9707 don't do the same transformation here.
9708 (vect_is_simple_use): If the defining statement has been replaced
9709 by a pattern statement, return the pattern statement instead.
9710 Remove the corresponding (local) transformation from the vectype
9713 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
9715 * tree-vectorizer.h (vect_is_simple_use): Move the gimple ** to the
9716 end and default to null.
9717 * tree-vect-loop.c (vect_create_epilog_for_reduction)
9718 (vectorizable_reduction): Update calls accordingly, dropping the
9719 gimple ** argument if the passed-back statement isn't needed.
9720 * tree-vect-patterns.c (vect_get_internal_def, type_conversion_p)
9721 (vect_recog_rotate_pattern): Likewise.
9722 (vect_recog_mask_conversion_pattern): Likewise.
9723 * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
9724 (vect_mask_constant_operand_p): Likewise.
9725 * tree-vect-stmts.c (is_simple_and_all_uses_invariant, process_use):
9726 (vect_model_simple_cost, vect_get_vec_def_for_operand): Likewise.
9727 (get_group_load_store_type, get_load_store_type): Likewise.
9728 (vect_check_load_store_mask, vect_check_store_rhs): Likewise.
9729 (vectorizable_call, vectorizable_simd_clone_call): Likewise.
9730 (vectorizable_conversion, vectorizable_assignment): Likewise.
9731 (vectorizable_shift, vectorizable_operation): Likewise.
9732 (vectorizable_store, vect_is_simple_cond): Likewise.
9733 (vectorizable_condition, vectorizable_comparison): Likewise.
9734 (get_same_sized_vectype, vect_get_mask_type_for_stmt): Likewise.
9735 (vect_is_simple_use): Rename the def_stmt argument to def_stmt_out
9736 and move it to the end. Cope with null def_stmt_outs.
9738 2018-06-30 Bernd Edlinger <bernd.edlinger@hotmail.de>
9740 * Makefile.in (FLAGS_TO_PASS): Add RANLIB_FOR_TARGET.
9742 2018-06-29 Jeff Law <law@redhat.com>
9744 * config/v850/v850.c (v850_legitimate_address_p): Handle large
9745 displacements for TARGET_V850E2V3 and newer.
9746 (TARGET_LRA_P): Remove. Defaults to LRA now.
9747 * config/v850/v850.md (sign23byte_load): Remove.
9748 (unsign23byte_load, sign23hword_load, unsign23hword_load): Likewise.
9749 (23word_load, 23byte_store, 23hword_store, 23word_store): Likewise.
9751 2018-06-29 Martin Liska <mliska@suse.cz>
9754 * coverage.c (coverage_init): Mangle full path name.
9755 * doc/invoke.texi: Document the change.
9756 * gcov-io.c (mangle_path): New.
9757 * gcov-io.h (mangle_path): Likewise.
9758 * gcov.c (mangle_name): Use mangle_path for path mangling.
9760 2018-06-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9762 * config/arm/arm.c (output_move_double): Don't allow STRD instructions
9763 if starting source register is not even.
9765 2018-06-29 Martin Liska <mliska@suse.cz>
9767 PR tree-optimization/86263
9768 * tree-switch-conversion.c (switch_decision_tree::try_switch_expansion):
9769 Make edge redirection.
9771 2018-06-29 David Malcolm <dmalcolm@redhat.com>
9773 * dumpfile.c (dump_loc): Add indentation based on scope depth.
9774 (dump_scope_depth): New variable.
9775 (get_dump_scope_depth): New function.
9776 (dump_begin_scope): New function.
9777 (dump_end_scope): New function.
9778 * dumpfile.h (get_dump_scope_depth): New declaration.
9779 (dump_begin_scope): New declaration.
9780 (dump_end_scope): New declaration.
9781 (class auto_dump_scope): New class.
9782 (AUTO_DUMP_SCOPE): New macro.
9783 * tree-vectorizer.h (DUMP_VECT_SCOPE): Reimplement in terms of
9786 2018-06-29 Richard Biener <rguenther@suse.de>
9788 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences): Assert
9789 compute_all_dependences succeeds.
9790 * tree-vect-loop.c (vect_get_datarefs_in_loop): Fail early if we
9791 exceed --param loop-max-datarefs-for-datadeps.
9793 2018-06-29 Jakub Jelinek <jakub@redhat.com>
9795 * config/rs6000/t-rs6000: Append rs6000-modes.h to TM_H.
9797 2018-06-28 Uros Bizjak <ubizjak@gmail.com>
9800 * config/i386/sse.md (*vec_extractv4si_0_zext_sse4): Use
9801 alternative 0 in preferred_for_speed attribute.
9803 2018-06-28 Paul Koning <ni1d@arrl.net>
9805 * config/pdp11/pdp11-protos.h (pdp11_shift_length): New function.
9806 * config/pdp11/pdp11.c (pdp11_shift_length): New function.
9807 * config/pdp11/pdp11.h (ADJUST_INSN_LENGTH): Remove.
9808 * config/pdp11/pdp11.md: Correct "length" attribute calculation
9809 for shift insn patterns.
9811 2018-06-28 David Malcolm <dmalcolm@redhat.com>
9813 * cgraph.c (cgraph_node::get_body): Replace assignments to
9814 "dump_file" with calls to set_dump_file.
9815 * dumpfile.c (alt_dump_file): Make static, and group with...
9816 (alt_flags): ...this definition.
9817 (dumps_are_enabled): New variable.
9818 (refresh_dumps_are_enabled): New function.
9819 (set_dump_file): New function.
9820 (set_alt_dump_file): New function.
9821 (gcc::dump_manager::dump_start): Replace assignments to
9822 "dump_file" and "alt_dump_file" with calls to set_dump_file and
9824 (gcc::dump_manager::dump_finish): Likewise.
9825 * dumpfile.h (alt_dump_file): Delete decl.
9826 (dumps_are_enabled): New variable decl.
9827 (set_dump_file): New function decl.
9828 (dump_enabled_p): Rewrite in terms of new "dumps_are_enabled"
9830 * tree-nested.c (lower_nested_functions): Replace assignments to
9831 "dump_file" with calls to set_dump_file.
9833 2018-06-28 Eric Botcazou <ebotcazou@adacore.com>
9835 * tree-cfg.c (verify_gimple_in_cfg): Call verify_location on the
9836 goto_locus of each outgoing edge of each basic block.
9838 2018-06-28 Richard Biener <rguenther@suse.de>
9840 * dwarf2out.c (decl_scope_table): Remove.
9841 (push_decl_scope): Likewise.
9842 (pop_decl_scope): Likewise.
9843 (gen_type_die_for_member): Do not call push/pop_decl_scope.
9844 (gen_struct_or_union_type_die): Likewise.
9845 (gen_tagged_type_die): Likewise.
9846 (dwarf2out_init): Do not initialize decl_scope_table.
9847 (dwarf2out_c_finalize): Do not free it.
9849 2018-06-28 Richard Biener <rguenther@suse.de>
9851 * dwarf2out.c (gen_subprogram_die): Use is_unit_die when
9852 deciding whether to not re-use a DIE.
9854 2018-06-28 Richard Biener <rguenther@suse.de>
9856 * dwarf2out.c (gen_subprogram_die): Always re-use DIEs with an
9857 DW_AT_abstract_origin attribute.
9859 2018-06-28 Martin Liska <mliska@suse.cz>
9861 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
9862 Use newly introduced constants.
9863 * tree-switch-conversion.h (struct jump_table_cluster):
9864 Define max_ratio_for_size and max_ratio_for_speed.
9866 2018-06-28 Martin Liska <mliska@suse.cz>
9868 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
9869 Add new checking assert to catch invalid state.
9870 (jump_table_cluster::can_be_handled): Handle single case
9872 (jump_table_cluster::is_beneficial): Bail out for such case.
9873 (bit_test_cluster::find_bit_tests):
9874 Add new checking assert to catch invalid state.
9875 (bit_test_cluster::can_be_handled): Handle single case
9877 (bit_test_cluster::is_beneficial): Bail out for such case.
9878 (switch_decision_tree::analyze_switch_statement):
9881 2018-06-28 Martin Liska <mliska@suse.cz>
9883 * common.opt: Introduce -completion option.
9884 * gcc.c (driver_handle_option): Handle it.
9885 (driver::main): Print completions if completion
9887 * opt-suggestions.c (option_proposer::get_completions):
9889 (option_proposer::suggest_completion): Likewise.
9890 (option_proposer::find_param_completions): Likewise.
9891 (verify_autocompletions): Likewise.
9892 (test_completion_valid_options): Likewise.
9893 (test_completion_valid_params): Likewise.
9894 (in_completion_p): Likewise.
9895 (empty_completion_p): Likewise.
9896 (test_completion_partial_match): Likewise.
9897 (test_completion_garbage): Likewise.
9898 (opt_proposer_c_tests): Likewise.
9899 * opt-suggestions.h: Declare new functions.
9900 * opts.c (common_handle_option): Handle OPT__completion_.
9901 * selftest-run-tests.c (selftest::run_tests): Add
9902 opt_proposer_c_tests.
9903 * selftest.c (assert_str_startswith): New.
9904 * selftest.h (assert_str_startswith): Likewise.
9905 (opt_proposer_c_tests): New.
9906 (ASSERT_STR_STARTSWITH): Likewise.
9908 2018-06-28 Martin Liska <mliska@suse.cz>
9910 * Makefile.in: Add opt-suggestions.o.
9911 * gcc-main.c: Include opt-suggestions.h.
9912 * gcc.c (driver::driver): Likewise.
9913 (driver::~driver): Remove m_option_suggestions.
9914 (driver::build_option_suggestions): Moved to option_proposer.
9915 (driver::suggest_option): Likewise.
9916 (driver::handle_unrecognized_options): Use option_proposer.
9917 * gcc.h (class driver): Add new memver m_option_proposer.
9918 * opt-suggestions.c: New file.
9919 * opt-suggestions.h: New file.
9921 2018-06-28 Martin Liska <mliska@suse.cz>
9923 * vec.h (class auto_string_vec): New (moved from auto_argvec).
9924 (auto_string_vec::~auto_string_vec): Likewise.
9926 2018-06-28 Eric Botcazou <ebotcazou@adacore.com>
9928 * tree-inline.h (struct copy_body_data): Move remapping_type_depth and
9929 prevent_decl_creation_for_types fields up and add reset_location field.
9930 * tree-inline.c (remap_gimple_stmt): Force input_location on the new
9931 statement if id->reset_location is true.
9932 (copy_edges_for_bb): Do not set goto_locus on the new edges if
9933 id->reset_location is true.
9934 (copy_phis_for_bb): Force input_location on the arguments if
9935 id->reset_location is true.
9936 (expand_call_inline): Set id->reset_location if DECL_IGNORED_P
9937 is set on the function to be inlined.
9939 2018-06-27 Stephan Bergmann <sbergman@redhat.com>
9941 * doc/invoke.texi (Debugging Options): Add -gsplit-dwarf.
9943 2018-06-27 Dimitar Dimitrov <dimitar@dinux.eu>
9945 * lra-eliminations.c (update_reg_eliminate): Mark all spanning hard
9946 registers for Pmode.
9947 * lra-lives.c (check_pseudos_live_through_calls): Mark all spanning
9948 hard registers for the clobbered pseudo.
9950 2018-06-27 Paul Koning <ni1d@arrl.net>
9952 * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
9953 mutually exclusive options.
9954 * config/pdp11/constraints.md (h): New constraint.
9955 (O): Update definition to match shift code generation.
9956 (D): New constraint.
9957 * config/pdp11/pdp11-modes.def (CCNZ): Define mode.
9959 * config/pdp11/pdp11-protos.h (int_no_side_effect_operand): New
9961 (output_jump): Change arguments.
9962 (pdp11_fixed_cc_regs): New function.
9963 (pdp11_cc_mode): Ditto.
9964 (pdp11_expand_shift): Ditto.
9965 (pdp11_assemble_shift): Ditto.
9966 (pdp11_small_shift): Ditto.
9967 (pdp11_branch_cost): Remove.
9968 * config/pdp11/pdp11.c (pdp11_assemble_integer): Remove comments
9970 (pdp11_register_move_cost): Update for CC registers.
9971 (pdp11_rtx_costs): Add case for LSHIFTRT.
9972 (pdp11_output_jump): Add CCNZ mode conditional branches.
9973 (notice_update_cc_on_set): Remove.
9974 (pdp11_cc_mode): New function.
9975 (simple_memory_operand): Correct pre/post decrement case.
9976 (no_side_effect_operand): New function.
9977 (pdp11_regno_reg_class): Add CC_REGS class.
9978 (pdp11_fixed_cc_regs): New function.
9979 (pdp11_small_shift): New function.
9980 (pdp11_expand_shift): New function to expand shift insns.
9981 (pdp11_assemble_shift): New function to output shifts.
9982 (pdp11_branch_cost): Remove.
9983 (pdp11_modes_tieable_p): Make QI/HI modes tieable.
9984 * config/pdp11/pdp11.h (SIZE_TYPE): Ensure 16-bit type.
9985 (WCHAR_TYPE): Ditto.
9986 (PTRDIFF_TYPE): Ditto.
9987 (ADJUST_INSN_LENGTH): New macro.
9988 (FIXED_REGISTERS): Add CC registers.
9989 (CALL_USED_REGISTERS): Ditto.
9991 (REG_CLASS_NAMES): Ditto.
9992 (REG_CLASS_CONTENTS): Ditto.
9993 (SELECT_CC_MODE): Use new function.
9994 (TARGET_FLAGS_REGNUM): New macro.
9995 (TARGET_FIXED_CONDITION_CODE_REGS): Ditto.
9996 (cc0_reg_rtx): Remove.
9997 (CC_STATUS_MDEP): Remove.
9998 (CC_STATUS_MDEFP_INIT): Remove.
9999 (CC_IN_FPU): Remove.
10000 (NOTICE_UPDATE_CC): Remove.
10001 (REGISTER_NAMES): Add CC registers.
10002 (BRANCH_COST): Change to constant 1.
10003 * config/pdp11/pdp11.md: Rewrite for CCmode condition code
10005 * config/pdp11/pdp11.opt (mbcopy): Remove.
10006 (mbcopy-builtin): Remove.
10007 (mbranch-cheap): Remove.
10008 (mbranch-expensive): Remove.
10009 * config/pdp11/predicates.md (expand_shift_operand): Update to
10010 match shift code generation.
10011 (ccnz_operator): New predicate.
10012 * doc/invoke.texi (PDP-11 Options): Remove deleted options
10013 -mbcopy, -mbcopy-builtin, -mbranch-cheap, -mbranch-expensive.
10014 Remove non-existent option -mabshi, -mno-abshi. Document mutually
10016 * doc/md.texi (PDP-11): Document new D and h constraints. Update
10017 description of O constraint.
10019 2018-06-27 Jeff Law <law@redhat.com>
10020 Austin Law <austinklaw@gmail.com>
10022 * config/v850/v850.md (addsi3_set_flags): New pattern.
10023 (subsi3_set_flags, negsi2_set_flags, andsi3_set_flags): Likewise.
10024 (iorsi3_set_flags, xorsi3_set_flags, one_cmplsi2_set_flags): Likewise.
10025 (zero_extendhisi2_v850_set_flags): Likewise.
10026 (zero_extendqisi2_v850_set_flags): Likewise.
10027 (ashlsi3_set_flags, ashlsi3_v850e2_set_flags): Likewise.
10028 (lshrsi3_set_flags, lshrsi3_v850e2_set_flags): Likewise.
10029 (ashrsi3_set_flags, ashrsi3_v850e2_set_flags): Likewise.
10031 * config/v850/v850-protos.h (notice_update_cc): Remove.
10032 * config/v850/v850.c (v850_compare_op0, v850_compare_op1): Remove.
10033 (v850_print_operand): Handle 'D' and "d".
10034 (v850_select_cc_mode): Remove ATTRIBUTE_UNUSED for last argument.
10035 Add handling of arithmetic/logical operations compared against zero.
10036 (v850_gen_float_compare): Remove ATTRIBUTE_UNUSED for last argument.
10037 Do not look at v850_compare_op, instead get mode from last argument.
10038 (v850_gen_compare): Remove
10039 (increment_stack): Use addsi3_clobber_flags to avoid splitting failure
10040 after reload for prologue insns.
10041 (expand_prologue): Account for CLOBBER of CC_REGNUM in various
10043 (construct_save_jarl): Likewise.
10044 (TARGET_FLAGS_REGNUM): Define.
10045 * config/v850/v850.h (v850_compare_op0, v850_compare_op1): Remove.
10046 (NOTICE_UPDATE_CC): Remove.
10047 * config/v850/v850.md (v850_tst1): Use (reg:CCZ CC_REGNUM) rather
10048 than cc0. Conditionalize on reload_completed.
10049 (cmpsi_insn, setfcc_insn): Likewise.
10050 (tst1 splitter): Turn into define_and_split which sets the flags
10052 (cstoresi4, cbranchsf4, cbranchdf4, cbranchsi4_insn): Likewise.
10053 (cbranchsi4, branch_normal, branch_invert): Do not expose cc0 here.
10054 (cstoresf4, cstoredf4): Clobber the flags.
10055 (cmpsi, cmpsf, cmpdf): Remove expanders.
10056 (setf_insn): Remove pattern.
10057 (addsi3): Turn into define_and_split which clobbers the flags after
10058 reload and a suitable pattern (addsi3_clobber_flags) for use after
10060 (subsi3, negsi2, andsi3, iorsi3, xorsi3, one_cmplsi2) Likewise.
10061 (ashlsi3, ashlsi3_v850e2, lshrsi3, lsh4si3_v850e2): Likewise.
10062 (ashrsi3, ashrsi3_v850e2): Likewise.
10063 (bins): Clobber the flags.
10064 (movsicc_normal_cc, movsicc_normal, movsicc_tst1): Likewise.
10065 (movsicc_tst1_revesed, sasf, swap and rotate patterns): Likewise.
10066 (fix_loop_counter, call_internal_short, call_internal_long): Likewise.
10067 (call_value_internal_short, call_value_internal_long): Likewise.
10068 (callt_save_interrupt, callt_return_interrupt): Likewise.
10069 (save_interrupt, return_interrupt): Likewise.
10070 (callt_save_all_interrupt, save_all_interrupt): Likewise.
10071 (_save_all_interrupt, callt_restore_all_interrupt): Likewise.
10072 (restore_all_interrupt, _restore_all_interrupt): Likewise.
10073 (All FP comparisons): Only allow after reload has completed.
10075 (divh, divhu): Tweak output template.
10076 (branch_z_normal, branch_z_invert): Remove
10077 (branch_nz_normal, branch_nz_invert): Likewise.
10078 (extendhisi_insn, extendqisi_insn): Do not clobber flags.
10080 * config/v850/v850-modes.def (CCZ, CCNZ): Add new modes.
10081 * config/v850/v850.c (notice_update_cc): Remove.
10082 * config/v850/v850.h (CC_OVERFLOW_UNUSABLE): Remove
10083 (CC_NO_CARRY): Likewise.
10084 (NOTICE_UPDATE_CC): Define to nothing.
10085 * config/v850/v850.md: Remove block comment on cc0 handling
10086 Remove "cc" attribute from all patterns. Remove cc_status handling
10087 from all patterns. Minor formatting fixes.
10089 2018-06-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10091 * config/aarch64/aarch64-cores.def (cortex-a76): New entry.
10092 (cortex-a76.cortex-a55): Likewise.
10093 * config/aarch64/aarch64-tune.md: Regenerate.
10094 * doc/invoke.texi (AArch64 Options): Document cortex-a76 and
10095 cortex-a76.cortex-a55.
10097 2018-06-27 Jeff Law <law@redhat.com>
10099 * config/v850/t-v850 (MULTILIB_OPTIONS): Remove 8byte-align.
10100 (MULTILIB_DIRNAMES): Similarly.
10102 2018-06-27 Eric Botcazou <ebotcazou@adacore.com>
10104 * gimple.h (gimple_return_retbnd): Delete.
10105 (gimple_return_set_retbnd): Likewise.
10106 * cgraphunit.c (cgraph_node::expand_thunk): Remove call to
10107 gimple_return_set_retbnd.
10108 * gimple-pretty-print.c (dump_gimple_return): Remove call to
10109 gimple_return_retbnd and adjust.
10110 * tree-inline.h (struct copy_body_data): Remove retbnd field.
10111 * tree-inline.c (remap_gimple_stmt): Remove handling of retbnd.
10112 Explicitly return NULL in a couple more cases. Move assertion
10113 on debug statements and remove unreachable code.
10114 (reset_debug_binding): Do not test id->retbnd.
10115 (expand_call_inline): Do not set it.
10117 2018-06-27 Rasmus Villemoes <rv@rasmusvillemoes.dk>
10119 * configure.ac: Add --disable-gcov option.
10120 * configure: Regenerate.
10121 * Makefile.in: Honour @enable_gcov@.
10122 * doc/install.texi: Document --disable-gcov.
10124 2018-06-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10126 * config/arm/arm-cpus.in (cortex-a76): New entry.
10127 (cortex-a76.cortex-a55): Likewise.
10128 * config/arm/arm-tables.opt: Regenerate.
10129 * config/arm/arm-tune.md: Likewise.
10130 * config/arm/driver-arm.c (arm_cpu_table): Add Cortex-A76 entry.
10131 * doc/invoke.texi (ARM Options): Document cortex-a76 and
10132 cortex-a76.cortex-a55.
10134 2018-06-27 Tamar Christina <tamar.christina@arm.com>
10137 * config/aarch64/aarch64.md (*movhf_aarch64): Add dup v0.4h pattern.
10139 2018-06-27 Siddhesh Poyarekar <siddhesh@sourceware.org>
10141 * config/aarch64/aarch64.h (CALL_USE_REGISTERS): Fix obsolete
10143 (EPILOGUE_USES): Likewise.
10145 2018-06-26 Eric Botcazou <ebotcazou@adacore.com>
10147 * tree-inline.c (remap_location): New function extracted from...
10148 (copy_edges_for_bb): Add ID parameter. Remap goto_locus.
10149 (copy_phis_for_bb): ...here. Call remap_location.
10150 (copy_cfg_body): Adjust call to copy_edges_for_bb.
10152 2018-06-26 Aaron Sawdey <acsawdey@linux.ibm.com>
10154 * config/rs6000/rs6000-string.c (expand_block_clear): Don't use
10155 unaligned vsx for 16B memset.
10157 2018-06-26 Segher Boessenkool <segher@kernel.crashing.org>
10160 * config/rs6000/rs6000.c (rs6000_init_builtins): Do not set
10161 ieee128_float_type_node to long_double_type_node unless
10162 TARGET_LONG_DOUBLE_128 is set.
10164 2018-06-26 David Malcolm <dmalcolm@redhat.com>
10166 * cfgloop.c (get_loop_location): Convert return type from
10167 location_t to dump_user_location_t, replacing INSN_LOCATION lookups
10168 by implicit construction from rtx_insn *, and using
10169 dump_user_location_t::from_function_decl for the fallback case.
10170 * cfgloop.h (get_loop_location): Convert return type from
10171 location_t to dump_user_location_t.
10172 * cgraphunit.c (walk_polymorphic_call_targets): Update call to
10173 dump_printf_loc to pass in a dump_location_t rather than a
10174 location_t, via the gimple stmt.
10175 * coverage.c (get_coverage_counts): Update calls to
10176 dump_printf_loc to pass in dump_location_t rather than a
10178 * doc/optinfo.texi (Dump types): Convert example of
10179 dump_printf_loc from taking "locus" to taking "insn". Update
10180 description of the "_loc" calls to cover dump_location_t.
10181 * dumpfile.c: Include "backend.h", "gimple.h", "rtl.h", and
10183 (dump_user_location_t::dump_user_location_t): New constructors,
10184 from gimple *stmt and rtx_insn *.
10185 (dump_user_location_t::from_function_decl): New function.
10186 (dump_loc): Make static.
10187 (dump_gimple_stmt_loc): Convert param "loc" from location_t to
10188 const dump_location_t &.
10189 (dump_generic_expr_loc): Delete.
10190 (dump_printf_loc): Convert param "loc" from location_t to
10191 const dump_location_t &.
10192 (selftest::test_impl_location): New function.
10193 (selftest::dumpfile_c_tests): New function.
10194 * dumpfile.h: Include "profile-count.h".
10195 (class dump_user_location_t): New class.
10196 (struct dump_impl_location_t): New struct.
10197 (class dump_location_t): New class.
10198 (dump_printf_loc): Convert 2nd param from source_location to
10199 const dump_location_t &.
10200 (dump_generic_expr_loc): Delete.
10201 (dump_gimple_stmt_loc): Convert 2nd param from source_location to
10202 const dump_location_t &.
10203 * gimple-fold.c (fold_gimple_assign): Update call to
10204 dump_printf_loc to pass in a dump_location_t rather than a
10205 location_t, via the gimple stmt.
10206 (gimple_fold_call): Likewise.
10207 * gimple-loop-interchange.cc
10208 (loop_cand::analyze_iloop_reduction_var): Update for change to
10209 check_reduction_path.
10210 (tree_loop_interchange::interchange): Update for change to
10211 find_loop_location.
10212 * graphite-isl-ast-to-gimple.c (scop_to_isl_ast): Update for
10213 change in return-type of find_loop_location.
10214 (graphite_regenerate_ast_isl): Likewise.
10215 * graphite-optimize-isl.c (optimize_isl): Likewise.
10216 * graphite.c (graphite_transform_loops): Likewise.
10217 * ipa-devirt.c (ipa_devirt): Update call to dump_printf_loc to
10218 pass in a dump_location_t rather than a location_t, via the
10220 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
10221 * ipa.c (walk_polymorphic_call_targets): Likewise.
10222 * loop-unroll.c (report_unroll): Convert "locus" param from
10223 location_t to dump_location_t.
10224 (decide_unrolling): Update for change to get_loop_location's
10226 * omp-grid.c (struct grid_prop): Convert field "target_loc" from
10227 location_t to dump_user_location_t.
10228 (grid_find_single_omp_among_assignments_1): Updates calls to
10229 dump_printf_loc to pass in a dump_location_t rather than a
10230 location_t, via the gimple stmt.
10231 (grid_parallel_clauses_gridifiable): Convert "tloc" from
10232 location_t to dump_location_t. Updates calls to dump_printf_loc
10233 to pass in a dump_location_t rather than a location_t, via the
10235 (grid_inner_loop_gridifiable_p): Likewise.
10236 (grid_dist_follows_simple_pattern): Likewise.
10237 (grid_gfor_follows_tiling_pattern): Likewise.
10238 (grid_target_follows_gridifiable_pattern): Likewise.
10239 (grid_attempt_target_gridification): Convert initialization
10240 of local "grid" from memset to zero-initialization; FIXME: does
10241 this require C++11? Update call to dump_printf_loc to pass in a
10242 optinfo_location rather than a location_t, via the gimple stmt.
10243 * profile.c (read_profile_edge_counts): Updates call to
10244 dump_printf_loc to pass in a dump_location_t rather than a
10246 (compute_branch_probabilities): Likewise.
10247 * selftest-run-tests.c (selftest::run_tests): Call
10249 * selftest.h (dumpfile_c_tests): New decl.
10250 * tree-loop-distribution.c (pass_loop_distribution::execute):
10251 Update for change in return type of find_loop_location.
10252 * tree-parloops.c (parallelize_loops): Likewise.
10253 * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Convert
10254 "locus" from location_t to dump_user_location_t.
10255 (canonicalize_loop_induction_variables): Likewise.
10256 * tree-ssa-loop-ivopts.c (tree_ssa_iv_optimize_loop): Update
10257 for change in return type of find_loop_location.
10258 * tree-ssa-loop-niter.c (number_of_iterations_exit): Update call
10259 to dump_printf_loc to pass in a dump_location_t rather than a
10260 location_t, via the stmt.
10261 * tree-ssa-sccvn.c (eliminate_dom_walker::before_dom_children):
10263 * tree-vect-loop-manip.c (find_loop_location): Convert return
10264 type from source_location to dump_user_location_t.
10265 (vect_do_peeling): Update for above change.
10266 (vect_loop_versioning): Update for change in type of
10268 * tree-vect-loop.c (check_reduction_path): Convert "loc" param
10269 from location_t to dump_user_location_t.
10270 (vect_estimate_min_profitable_iters): Update for change in type
10272 * tree-vect-slp.c (vect_print_slp_tree): Convert param "loc" from
10273 location_t to dump_location_t.
10274 (vect_slp_bb): Update for change in type of vect_location.
10275 * tree-vectorizer.c (vect_location): Convert from source_location
10276 to dump_user_location_t.
10277 (try_vectorize_loop_1): Update for change in vect_location's type.
10278 (vectorize_loops): Likewise.
10279 (increase_alignment): Likewise.
10280 * tree-vectorizer.h (vect_location): Convert from source_location
10281 to dump_user_location_t.
10282 (find_loop_location): Convert return type from source_location to
10283 dump_user_location_t.
10284 (check_reduction_path): Convert 1st param from location_t to
10285 dump_user_location_t.
10286 * value-prof.c (check_counter): Update call to dump_printf_loc to
10287 pass in a dump_user_location_t rather than a location_t; update
10288 call to error_at for change in type of "locus".
10289 (check_ic_target): Update call to dump_printf_loc to
10290 pass in a dump_user_location_t rather than a location_t, via the
10293 2018-06-26 Robin Dapp <rdapp@linux.vnet.ibm.com>
10295 * config/s390/s390.h (enum processor_flags): Do not use
10296 default tune parameter when -march was specified.
10298 2018-06-26 Jakub Jelinek <jakub@redhat.com>
10301 * config/i386/i386.md (setcc + movzbl to xor + setcc peephole2s):
10302 Check reg_overlap_mentioned_p in addition to reg_set_p with the same
10305 2018-06-26 Richard Biener <rguenther@suse.de>
10307 PR tree-optimization/86287
10309 * tree-vect-loop.c (vect_transform_loop_stmt): Fix read-after-free.
10310 (vect_analyze_loop): Initialize n_stmts.
10312 2018-06-26 Richard Biener <rguenther@suse.de>
10314 PR middle-end/86271
10315 * fold-const.c (fold_convertible_p): Pointer extension
10318 2018-06-26 Alexandre Oliva <aoliva@redhat.com>
10321 * dwarf2out.c (loc_list_has_views): Adjust comments.
10322 (dw_loc_list): Split single cross-partition range with
10325 2018-06-25 Jeff Law <law@redhat.com>
10327 * common/config/v850/v850-common.c (TARGET_DEFAULT_TARGET_FLAGS): Turn
10328 on -mbig-switch by default.
10330 * config/v850/predicates.md (const_float_1_operand): Fix match_code
10332 (const_float_0_operand): Remove unused predicate.
10333 * config/v850/v850.md (define_constants): Remove UNSPEC_LOOP.
10334 (define_c_enum unspec): Add LOOP, RCP and RSQRT constants.
10335 (recipsf2): New expander. Original pattern now called
10337 (recipdf2, recipdf2_insn): Similarly.
10338 (rsqrtsf2, rsqrtsf2_insn): Similarly
10339 (rsqrtdf2, rsqrtdf2_insn): Similarly
10341 2018-06-26 Gerald Pfeifer <gerald@pfeifer.com>
10343 * ginclude/stddef.h: Remove an obsolete comment on FreeBSD 5.
10344 Simplify logic for FreeBSD (twice).
10346 2018-06-25 Martin Sebor <msebor@redhat.com>
10348 PR tree-optimization/86204
10349 * tree-ssa-strlen.c (handle_builtin_strlen): Avoid storing
10350 a strnlen result if it's less than the length of the string.
10352 2018-06-25 Martin Sebor <msebor@redhat.com>
10354 PR tree-optimization/85700
10355 * gimple-fold.c (gimple_fold_builtin_strncat): Adjust comment.
10356 * tree-ssa-strlen.c (is_strlen_related_p): Handle integer subtraction.
10357 (maybe_diag_stxncpy_trunc): Distinguish strncat from strncpy.
10359 2018-06-25 Martin Sebor <msebor@redhat.com>
10361 * doc/extend.texi (Zero-length arrays): Update and clarify.
10363 2018-06-25 Michael Meissner <meissner@linux.ibm.com>
10365 * config.gcc (powerpc64le*): Revert January 16th, 2018 patch that
10366 added IEEE/IBM long double multilib support on PowerPC little
10367 endian Linux systems.
10368 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Likewise.
10369 (MULTILIB_DEFAULTS): Likewise.
10370 * config/rs6000/rs6000.c (rs6000_option_override_internal):
10372 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Likewise.
10373 * config/rs6000/t-ldouble-linux64le-ibm: Delete, no longer used.
10374 * config/rs6000/t-ldouble-linux64le-ieee: Delete, no longer used.
10376 2018-06-25 Alexander Monakov <amonakov@ispras.ru>
10378 PR middle-end/86311
10379 * sort.cc (REORDER_23): Avoid memcpy with same destination and source.
10380 (REORDER_45): Likewise.
10382 2018-06-25 Jeff Law <law@redhat.com>
10384 * config/v850/v850.md (divmodhi4): Make sure to sign extend the
10385 dividend to 32 bits. Adjust length.
10386 (udivmodhi4): Cleanup output template. Fix length.
10388 2018-06-25 Carl Love <cel@us.ibm.com>
10390 * config/rs6000/vsx.md: Change word selector to prefered location.
10392 2018-06-25 Richard Biener <rguenther@suse.de>
10394 PR tree-optimization/86304
10395 * tree-vectorizer.c (vectorize_loops): Walk over new possibly
10396 epilogue-if-converted loops as well.
10398 2018-06-25 Jan Hubicka <hubicka@ucw.cz>
10400 * lto-section-out.c (lto_begin_section): Do not print section
10401 name for noaddr and unnumbered dumps.
10403 2018-06-25 Richard Biener <rguenther@suse.de>
10405 * tree-vectorizer.h (struct vec_info_shared): New structure
10406 with parts split out from struct vec_info and loop_nest from
10407 struct _loop_vec_info.
10408 (struct vec_info): Adjust accordingly.
10409 (struct _loop_vec_info): Likewise.
10410 (LOOP_VINFO_LOOP_NEST): Adjust.
10411 (LOOP_VINFO_DATAREFS): Likewise.
10412 (LOOP_VINFO_DDRS): Likewise.
10413 (struct _bb_vec_info): Likewise.
10414 (BB_VINFO_DATAREFS): Likewise.
10415 (BB_VINFO_DDRS): Likewise.
10416 (struct _stmt_vec_info): Add dr_aux member.
10417 (DR_VECT_AUX): Adjust to refer to member of DR_STMTs vinfo.
10418 (DR_MISALIGNMENT_UNINITIALIZED): New.
10419 (set_dr_misalignment): Adjust.
10420 (dr_misalignment): Assert misalign isn't DR_MISALIGNMENT_UNINITIALIZED.
10421 (vect_analyze_loop): Adjust prototype.
10422 (vect_analyze_loop_form): Likewise.
10423 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
10424 Compute dependences lazily.
10425 (vect_record_base_alignments): Use shared datarefs/ddrs.
10426 (vect_verify_datarefs_alignment): Likewise.
10427 (vect_analyze_data_refs_alignment): Likewise.
10428 (vect_analyze_data_ref_accesses): Likewise.
10429 (vect_analyze_data_refs): Likewise.
10430 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Add
10431 constructor parameter for shared part.
10432 (vect_analyze_loop_form): Pass in shared part and adjust.
10433 (vect_analyze_loop_2): Pass in storage for the number of
10434 stmts. Move loop nest finding to the caller. Compute
10436 (vect_analyze_loop): Pass in shared part.
10437 (vect_transform_loop): Verify shared datarefs are unchanged.
10438 * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Add
10439 constructor parameter for shared part.
10440 (vect_slp_analyze_bb_1): Pass in shared part and adjust.
10441 (vect_slp_bb): Verify shared datarefs are unchanged before
10443 * tree-vect-stmts.c (ensure_base_align): Adjust for DR_AUX
10445 (new_stmt_vec_info): Initialize DR_AUX misalignment to
10446 DR_MISALIGNMENT_UNINITIALIZED.
10447 * tree-vectorizer.c (vec_info::vec_info): Add constructor
10448 parameter for shared part.
10449 (vec_info::~vec_info): Adjust.
10450 (vec_info_shared::vec_info_shared): New.
10451 (vec_info_shared::~vec_info_shared): Likewise.
10452 (vec_info_shared::save_datarefs): Likewise.
10453 (vec_info_shared::check_datarefs): Likewise.
10454 (try_vectorize_loop_1): Construct shared part live for analyses
10455 of a single loop for multiple vector sizes.
10456 * tree-parloops.c (gather_scalar_reductions): Adjust.
10458 2018-06-25 Richard Biener <rguenther@suse.de>
10460 * tree-vect-data-refs.c (vect_find_stmt_data_reference): Modify
10461 DR for SIMD lane accesses here and mark DR with (void *)-1 aux.
10462 (vect_analyze_data_refs): Remove similar code from here and
10463 simplify accordingly.
10465 2018-06-25 Richard Biener <rguenther@suse.de>
10467 * tree-vect-data-refs.c (vect_check_gather_scatter): Fail
10468 for reverse storage order accesses rather than asserting
10469 they cannot happen here.
10471 2018-06-25 Tom de Vries <tdevries@suse.de>
10474 * config/i386/i386.md (define_insn "*tls_global_dynamic_64_<mode>"):
10475 Use data16 instead of .byte for insn prefix.
10477 2018-06-25 Andreas Krebbel <krebbel@linux.ibm.com>
10480 * parser.c (make_char_string_pack): Pass this literal chars
10481 through cpp_interpret_string.
10482 (cp_parser_userdef_numeric_literal): Check the result of
10483 make_char_string_pack.
10485 2018-06-24 Maya Rashish <coypu@sdf.org>
10487 * ginclude/stddef.h: Simplify conditions around avoiding
10488 re-definition of __size_t.
10490 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
10492 * lto-streamer-out.c (tree_is_indexable): Make LABEL_DECL nonindexable
10493 unless it is forced or nonlocal; assert that we stream no IMPORTED_DECL.
10495 2018-06-22 Maya Rashish <coypu@sdf.org>
10497 * doc/invoke.texi (mno-fancy-math-387): Update for changes
10498 made to OpenBSD and NetBSD through the years.
10500 2018-06-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
10502 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
10503 behavior of vec_pack (vector double, vector double) to match
10504 behavior of vec_float2 (vector double, vector double).
10506 2018-06-22 Olivier Hainque <hainque@adacore.com>
10508 * gimplify.c (gimplify_function_tree): Prevent creation
10509 of a trampoline for the address of the current function
10510 passed to entry/exit instrumentation hooks.
10512 2018-06-22 Aaron Sawdey <acsawdey@linux.ibm.com>
10515 * config/rs6000/rs6000-string.c (expand_strn_compare): Handle -m32
10518 2018-06-22 Martin Liska <mliska@suse.cz>
10520 PR tree-optimization/86263
10521 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
10522 Bail out if is_enabled is false.
10523 * tree-switch-conversion.h (jump_table_cluster::is_enabled):
10525 (jump_table_cluster::is_enabled): New function.
10527 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
10529 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
10530 BINFO_BASE_ACCESSES and BINFO_VPTR_FIELD.
10531 * tree-streamer-in.c (streamer_read_tree_bitfields): Likewise.
10532 (lto_input_ts_binfo_tree_pointers): Likewise.
10533 * tree-streamer-out.c (streamer_write_tree_bitfields,
10534 write_ts_binfo_tree_pointers): Likewise.
10535 * tree.c (free_lang_data_in_binfo): Clear BINFO_VPTR_FIELD.
10537 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
10539 * tree.c (free_lang_data_in_type): Free all TYPE_VFIELDs.
10541 2018-06-22 Martin Liska <mliska@suse.cz>
10543 * symbol-summary.h (get): Make it pure and inline move
10544 functionality from ::get function.
10545 (get): Remove and inline into ::get and ::get_create.
10546 (get_create): Move code from ::get function.
10548 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
10551 * config/i386/sol2.h (CPP_SPEC): Don't pass -P for
10552 -x assembler-with-cpp.
10554 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
10556 * config/sol2.h (TARGET_OS_CPP_BUILTINS): Define
10557 _FILE_OFFSET_BITS=64 for C++.
10559 2018-06-21 Michael Meissner <meissner@linux.ibm.com>
10561 * config/rs6000/rs6000.md (extendtfif2): Add missing 128-bit
10562 conversion insn that shows up when pr85657-3.c is compiled using
10563 IEEE 128-bit long double.
10564 (neg<mode>2_internal): Use the correct mode to check whether the
10565 mode is IBM extended.
10566 * config/rs6000/rs6000.c (init_float128_ieee): Prevent complex
10567 multiply and divide external functions from being created more
10570 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
10572 * cfgrtl.c (fixup_reorder_chain): Do not emit NOPs in DECL_IGNORED_P
10574 (rtl_merge_blocks): Likewise. Do not emit a NOP if the location of
10575 the edge can be forwarded.
10576 (cfg_layout_merge_blocks): Likewise.
10578 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
10580 * except.c (finish_eh_generation): Commit edge insertions only after
10581 the EH edges have been redirected from post-landing to landing pads.
10583 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
10585 * tree-nested.c (get_frame_type): Use create_tmp_var_raw instead of
10586 create_tmp_var_for to create the FRAME decl.
10587 (finalize_nesting_tree_1): Do not unchain the FRAME decl.
10589 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
10591 * tree-inline.c (copy_edges_for_bb): Minor tweak.
10592 (maybe_move_debug_stmts_to_successors): Also reset the locus of the
10593 debug statement when resetting its value.
10594 (expand_call_inline): Copy the locus of the call onto the assignment
10595 of the return value, if any. Use local variable in more cases.
10597 2018-06-21 Martin Liska <mliska@suse.cz>
10599 * ipa-pure-const.c (propagate_nothrow): Use
10600 funct_state_summaries->get.
10601 (dump_malloc_lattice): Likewise.
10602 (propagate_malloc): Likewise.
10604 2018-06-21 Richard Biener <rguenther@suse.de>
10606 * lto-streamer-out.c (DFS::DFS_write_tree_body): Update outdated
10607 comment. Follow BLOCK_ABSTRACT_ORIGIN unconditionally.
10608 * tree-streamer-in.c (lto_input_ts_block_tree_pointers): Update
10610 * tree-streamer-out.c (write_ts_block_tree_pointers): Stream
10611 BLOCK_ABSTRACT_ORIGIN unconditionally.
10613 2018-06-21 David Malcolm <dmalcolm@redhat.com>
10615 * ipa-cp.c (ipcp_driver): Set edge_clone_summaries to NULL after
10617 * ipa-reference.c (ipa_reference_c_finalize): Delete
10618 ipa_ref_opt_sum_summaries and set it to NULL.
10620 2018-06-21 Tom de Vries <tdevries@suse.de>
10622 PR tree-optimization/85859
10623 * tree-ssa-tail-merge.c (stmt_local_def): Copy gimple_is_call
10624 test with comment from bb_no_side_effects_p.
10626 2018-06-21 Richard Biener <rguenther@suse.de>
10628 PR tree-optimization/86232
10629 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Adjust
10630 max for constant niter.
10632 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
10634 * config/aarch64/aarch64-simd.md
10635 (*aarch64_crypto_aes<aes_op>v16qi_xor_combine): New.
10637 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
10639 * config/aarch64/aarch64-simd.md (aarch64_crypto_aes<aes_op>v16qi):
10640 Make opernads of the unspec commutative.
10642 2018-06-21 Richard Biener <rguenther@suse.de>
10644 * tree-data-ref.c (dr_step_indicator): Handle NULL DR_STEP.
10645 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr):
10646 Avoid calling vect_mark_for_runtime_alias_test with gathers or scatters.
10647 (vect_analyze_data_ref_dependence): Re-order checks to deal with
10649 (vect_record_base_alignments): Do not record base alignment
10650 for gathers or scatters.
10651 (vect_compute_data_ref_alignment): Drop return value that is always
10652 true. Bail out early for gathers or scatters.
10653 (vect_enhance_data_refs_alignment): Bail out early for gathers
10655 (vect_find_same_alignment_drs): Likewise.
10656 (vect_analyze_data_refs_alignment): Remove dead code.
10657 (vect_slp_analyze_and_verify_node_alignment): Likewise.
10658 (vect_analyze_data_refs): For possible gathers or scatters do
10659 not create an alternate DR, just check their possible validity
10660 and mark them. Adjust DECL_NONALIASED handling to not rely
10661 on DR_BASE_ADDRESS.
10662 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Do not
10663 update inits of gathers or scatters.
10664 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern):
10665 Also copy gather/scatter flag to pattern vinfo.
10667 2018-06-20 Kelvin Nilsen <kelvin@gcc.gnu.org>
10669 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
10670 behavior of vec_packsu (vector unsigned long long, vector unsigned
10671 long long) to match behavior of vec_packs with same signature.
10673 2018-06-20 Chung-Lin Tang <cltang@codesourcery.com>
10674 Thomas Schwinge <thomas@codesourcery.com>
10675 Cesar Philippidis <cesar@codesourcery.com>
10677 * gimplify.c (gimplify_scan_omp_clauses): Add support for
10678 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
10679 (gimplify_adjust_omp_clauses): Likewise.
10680 (gimplify_oacc_declare_1): Add support for GOMP_MAP_RELEASE, remove
10681 support for GOMP_MAP_FORCE_{ALLOC,TO,FROM,TOFROM}.
10682 (gimplify_omp_target_update): Update handling of acc update and
10684 * omp-low.c (install_var_field): Remove unused parameter
10685 base_pointers_restrict.
10686 (scan_sharing_clauses): Remove base_pointers_restrict parameter.
10687 Update call to install_var_field. Handle OMP_CLAUSE_{IF_PRESENT,
10689 (omp_target_base_pointers_restrict_p): Delete.
10690 (scan_omp_target): Update call to scan_sharing_clauses.
10691 * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_{IF_PRESENT,
10693 * tree-nested.c (convert_nonlocal_omp_clauses): Handle
10694 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
10695 (convert_local_omp_clauses): Likewise.
10696 * tree-pretty-print.c (dump_omp_clause): Likewise.
10697 * tree.c (omp_clause_num_ops): Add entries for OMP_CLAUSE_{IF_PRESENT,
10699 (omp_clause_code_name): Likewise.
10701 2018-06-20 Jakub Jelinek <jakub@redhat.com>
10704 * var-tracking.c (use_narrower_mode_test): Check if shift amount can
10707 PR tree-optimization/86231
10708 * tree-vrp.c (union_ranges): For ( [ ) ] or ( )[ ] range and
10709 anti-range don't overwrite *vr0min before using it to compute *vr0max.
10711 2018-06-20 Tom de Vries <tdevries@suse.de>
10713 PR tree-optimization/86097
10714 * tree-ssa-loop-manip.c (canonicalize_loop_ivs): Also convert *nit to
10715 iv type if signedness of iv type is not the same as that of *nit.
10717 2018-06-20 Jakub Jelinek <jakub@redhat.com>
10719 * cfgrtl.c (rtl_verify_edges): Formatting fix. If bb->preds has any
10720 EDGE_EH edges, verify they are all EDGE_EH.
10722 2018-06-20 Maya Rashish <coypu@sdf.org>
10724 * ginclude/stddef.h: Limit #include <machine/ansi.h> to NetBSD.
10726 2018-06-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10728 * config/aarch64/aarch64-tuning-flags.def (no_ldp_stp_qregs): New.
10729 * config/aarch64/aarch64.c (xgene1_tunings): Add
10730 AARCH64_EXTRA_TUNE_NO_LDP_STP_QREGS to tune_flags.
10731 (aarch64_mode_valid_for_sched_fusion_p):
10732 Allow 16-byte modes.
10733 (aarch64_classify_address): Allow 16-byte modes for load_store_pair_p.
10734 * config/aarch64/aarch64-ldpstp.md: Add peepholes for LDP STP of
10736 * config/aarch64/aarch64-simd.md (load_pair<VQ:mode><VQ2:mode>):
10738 (vec_store_pair<VQ:mode><VQ2:mode>): Likewise.
10739 * config/aarch64/iterators.md (VQ2): New mode iterator.
10741 2018-06-20 Martin Liska <mliska@suse.cz>
10743 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
10744 Change default ratio from 10 to 8.
10746 2018-06-20 Martin Liska <mliska@suse.cz>
10748 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
10750 (bit_test_cluster::find_bit_tests): Likewise.
10751 (switch_decision_tree::analyze_switch_statement): Find clusters.
10752 * tree-switch-conversion.h (struct jump_table_cluster): Document
10755 2018-06-20 Martin Liska <mliska@suse.cz>
10757 * tree-switch-conversion.c (switch_conversion::collect):
10758 Record m_uniq property.
10759 (switch_conversion::expand): Bail out for special conditions.
10760 (group_cluster::~group_cluster): New.
10761 (group_cluster::group_cluster): Likewise.
10762 (group_cluster::dump): Likewise.
10763 (jump_table_cluster::emit): New.
10764 (switch_decision_tree::fix_phi_operands_for_edges): New.
10765 (struct case_node): Remove struct.
10766 (jump_table_cluster::can_be_handled): New.
10767 (case_values_threshold): Moved to header.
10768 (reset_out_edges_aux): Likewise.
10769 (jump_table_cluster::is_beneficial): New.
10770 (bit_test_cluster::can_be_handled): Likewise.
10771 (add_case_node): Remove.
10772 (bit_test_cluster::is_beneficial): New.
10773 (case_bit_test::cmp): New.
10774 (bit_test_cluster::emit): New.
10775 (expand_switch_as_decision_tree_p): Remove.
10776 (bit_test_cluster::hoist_edge_and_branch_if_true): New.
10777 (fix_phi_operands_for_edge): Likewise.
10778 (switch_decision_tree::analyze_switch_statement): New.
10779 (compute_cases_per_edge): Move ...
10780 (switch_decision_tree::compute_cases_per_edge): ... here.
10781 (try_switch_expansion): Likewise.
10782 (switch_decision_tree::try_switch_expansion): Likewise.
10783 (record_phi_operand_mapping): Likewise.
10784 (switch_decision_tree::record_phi_operand_mapping): Likewise.
10785 (emit_case_decision_tree): Likewise.
10786 (switch_decision_tree::emit): Likewise.
10787 (balance_case_nodes): Likewise.
10788 (switch_decision_tree::balance_case_nodes): Likewise.
10789 (dump_case_nodes): Likewise.
10790 (switch_decision_tree::dump_case_nodes): Likewise.
10791 (emit_jump): Likewise.
10792 (switch_decision_tree::emit_jump): Likewise.
10793 (emit_cmp_and_jump_insns): Likewise.
10794 (switch_decision_tree::emit_cmp_and_jump_insns): Likewise.
10795 (emit_case_nodes): Likewise.
10796 (switch_decision_tree::emit_case_nodes): Likewise.
10797 (conditional_probability): Remove.
10798 * tree-switch-conversion.h (enum cluster_type): New.
10800 (struct cluster): Likewise.
10801 (cluster::cluster): Likewise.
10802 (struct simple_cluster): Likewise.
10803 (simple_cluster::simple_cluster): Likewise.
10804 (struct group_cluster): Likewise.
10805 (struct jump_table_cluster): Likewise.
10806 (struct bit_test_cluster): Likewise.
10807 (struct min_cluster_item): Likewise.
10808 (struct case_tree_node): Likewise.
10809 (case_tree_node::case_tree_node): Likewise.
10810 (jump_table_cluster::case_values_threshold): Likewise.
10811 (struct case_bit_test): Likewise.
10812 (struct switch_decision_tree): Likewise.
10813 (struct switch_conversion): Likewise.
10814 (switch_decision_tree::reset_out_edges_aux): Likewise.
10816 2018-06-20 Martin Liska <mliska@suse.cz>
10818 * tree-switch-conversion.c (MAX_CASE_BIT_TESTS): Remove.
10819 (hoist_edge_and_branch_if_true): Likewise.
10820 (expand_switch_using_bit_tests_p): Likewise.
10821 (struct case_bit_test): Likewise.
10822 (case_bit_test_cmp): Likewise.
10823 (emit_case_bit_tests): Likewise.
10824 (switch_conversion::switch_conversion): New class.
10825 (struct switch_conv_info): Remove old struct.
10826 (collect_switch_conv_info): More to ...
10827 (switch_conversion::collect): ... this.
10828 (check_range): Likewise.
10829 (switch_conversion::check_range): Likewise.
10830 (check_all_empty_except_final): Likewise.
10831 (switch_conversion::check_all_empty_except_final): Likewise.
10832 (check_final_bb): Likewise.
10833 (switch_conversion::check_final_bb): Likewise.
10834 (create_temp_arrays): Likewise.
10835 (switch_conversion::create_temp_arrays): Likewise.
10836 (free_temp_arrays): Likewise.
10837 (gather_default_values): Likewise.
10838 (switch_conversion::gather_default_values): Likewise.
10839 (build_constructors): Likewise.
10840 (switch_conversion::build_constructors): Likewise.
10841 (constructor_contains_same_values_p): Likewise.
10842 (switch_conversion::contains_same_values_p): Likewise.
10843 (array_value_type): Likewise.
10844 (switch_conversion::array_value_type): Likewise.
10845 (build_one_array): Likewise.
10846 (switch_conversion::build_one_array): Likewise.
10847 (build_arrays): Likewise.
10848 (switch_conversion::build_arrays): Likewise.
10849 (gen_def_assigns): Likewise.
10850 (switch_conversion::gen_def_assigns): Likewise.
10851 (prune_bbs): Likewise.
10852 (switch_conversion::prune_bbs): Likewise.
10853 (fix_phi_nodes): Likewise.
10854 (switch_conversion::fix_phi_nodes): Likewise.
10855 (gen_inbound_check): Likewise.
10856 (switch_conversion::gen_inbound_check): Likewise.
10857 (process_switch): Use the newly created class.
10858 (switch_conversion::expand): New.
10859 (switch_conversion::~switch_conversion): New.
10860 * tree-switch-conversion.h: New file.
10862 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
10864 * tree-vectorizer.h (NUM_PATTERNS, vect_recog_func_ptr): Move to
10865 tree-vect-patterns.c.
10866 * tree-vect-patterns.c (vect_supportable_direct_optab_p): New function.
10867 (vect_recog_dot_prod_pattern): Use it. Remove the type_in argument.
10868 (vect_recog_sad_pattern): Likewise.
10869 (vect_recog_widen_sum_pattern): Likewise.
10870 (vect_recog_pow_pattern): Likewise. Check for a null vectype.
10871 (vect_recog_widen_shift_pattern): Remove the type_in argument.
10872 (vect_recog_rotate_pattern): Likewise.
10873 (vect_recog_mult_pattern): Likewise.
10874 (vect_recog_vector_vector_shift_pattern): Likewise.
10875 (vect_recog_divmod_pattern): Likewise.
10876 (vect_recog_mixed_size_cond_pattern): Likewise.
10877 (vect_recog_bool_pattern): Likewise.
10878 (vect_recog_mask_conversion_pattern): Likewise.
10879 (vect_try_gather_scatter_pattern): Likewise.
10880 (vect_recog_widen_mult_pattern): Likewise. Check for a null vectype.
10881 (vect_recog_over_widening_pattern): Likewise.
10882 (vect_recog_gather_scatter_pattern): Likewise.
10883 (vect_recog_func_ptr): Move from tree-vectorizer.h
10884 (vect_vect_recog_func_ptrs): Move further down the file.
10885 (vect_recog_func): Likewise. Remove the third argument.
10886 (NUM_PATTERNS): Define based on vect_vect_recog_func_ptrs.
10887 (vect_pattern_recog_1): Expect the pattern function to do any
10888 necessary target tests. Also expect it to provide a vector type.
10889 Remove the type_in handling.
10891 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
10893 * tree-vect-patterns.c (vect_pattern_detected): New function.
10894 (vect_recog_dot_prod_patternm, vect_recog_sad_pattern)
10895 (vect_recog_widen_mult_pattern, vect_recog_widen_sum_pattern)
10896 (vect_recog_over_widening_pattern, vect_recog_widen_shift_pattern
10897 (vect_recog_rotate_pattern, vect_recog_vector_vector_shift_pattern)
10898 (vect_recog_mult_pattern, vect_recog_divmod_pattern)
10899 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
10900 (vect_recog_mask_conversion_pattern)
10901 (vect_try_gather_scatter_pattern): Likewise.
10903 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
10905 * tree-vect-patterns.c (vect_get_internal_def): New function.
10906 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
10907 (vect_recog_vector_vector_shift_pattern, check_bool_pattern)
10908 (search_type_for_mask_1): Use it.
10910 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
10912 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
10913 redundant WIDEN_SUM_EXPR handling.
10914 (vect_recog_sad_pattern): Likewise.
10916 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
10918 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
10919 redundant check that the types of a PLUS_EXPR or MULT_EXPR agree.
10920 (vect_recog_sad_pattern): Likewise PLUS_EXPR, ABS_EXPR and MINUS_EXPR.
10921 (vect_recog_widen_mult_pattern): Likewise MULT_EXPR.
10922 (vect_recog_widen_sum_pattern): Likewise PLUS_EXPR.
10924 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
10926 * tree-vect-stmts.c (vectorizable_call): Make sure that we
10927 use the stmt_vec_info of the original bb statement for the
10928 new zero assignment, even if the call is part of a pattern.
10930 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
10932 * tree-vectorizer.h (_stmt_vec_info): Note above pattern_def_seq
10933 that the sequence is attached to the original statement rather
10934 than the pattern statement.
10935 * tree-vect-loop.c (vect_determine_vf_for_stmt): Take the
10936 PATTERN_DEF_SEQ from the original statement rather than
10937 the main pattern statement.
10938 * tree-vect-stmts.c (free_stmt_vec_info): Likewise.
10939 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Likewise.
10940 (vect_mark_pattern_stmts): Don't copy the PATTERN_DEF_SEQ.
10942 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
10944 * tree-vect-stmts.c (vect_analyze_stmt): Move the handling of pattern
10945 definition statements before the early exit for statements that aren't
10947 * tree-vect-loop.c (vect_transform_loop_stmt): New function,
10949 (vect_transform_loop): ...here. Process pattern definition
10950 statements without first checking whether the main pattern
10951 statement is live or relevant.
10953 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
10955 * tree-cfgcleanup.c (tree_forwarder_block_p): Do not return false at
10956 -O0 if the locus represent UNKNOWN_LOCATION but have different values.
10958 2018-06-19 Aaron Sawdey <acsawdey@linux.ibm.com>
10960 * config/rs6000/rs6000-string.c (select_block_compare_mode): Check
10961 TARGET_EFFICIENT_OVERLAPPING_UNALIGNED here instead of in caller.
10962 (do_and3, do_and3_mask, do_compb3, do_rotl3): New functions.
10963 (expand_block_compare): Change select_block_compare_mode call.
10964 (expand_strncmp_align_check): Use new functions, fix comment.
10965 (emit_final_str_compare_gpr): New function.
10966 (expand_strn_compare): Refactor and clean up code.
10967 * config/rs6000/vsx.md (vsx_mov<mode>_64bit): Remove *.
10969 2018-06-19 Tony Reix <tony.reix@atos.com>
10970 Damien Bergamini <damien.bergamini@atos.com>
10971 David Edelsohn <dje.gcc@gmail.com>
10973 * collect2.c (static_obj): New variable.
10974 (static_libs): New variable.
10975 (is_in_list): Uncomment declaration.
10976 (main): Track AIX libraries linked statically.
10977 (is_in_list): Uncomment definition.
10978 (scan_prog_file): Don't add AIX shared libraries initializer
10979 to constructor list if linking statically.
10981 2018-06-19 Max Filippov <jcmvbkbc@gmail.com>
10983 * config/xtensa/xtensa.md (UNSPEC_FRAME_BLOCKAGE): New unspec
10985 (allocate_stack, frame_blockage, *frame_blockage): New patterns.
10987 2018-06-19 Jan Hubicka <hubicka@ucw.cz>
10989 * tree.c (find_decls_types_r): Remove all non-VAR_DECLs from
10992 2018-06-19 Martin Liska <mliska@suse.cz>
10994 * config/i386/i386.c (ix86_can_inline_p): Do not use
10995 ipa_fn_summaries::get_create.
10996 * ipa-cp.c (ipcp_cloning_candidate_p): Replace get_create with
10998 (devirtualization_time_bonus): Likewise.
10999 (ipcp_propagate_stage): Likewise.
11000 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
11001 (edge_set_predicate): Likewise.
11002 (evaluate_conditions_for_known_args): Likewise.
11003 (evaluate_properties_for_edge): Likewise.
11004 (ipa_call_summary::reset): Tranform to ...
11005 (ipa_call_summary::~ipa_call_summary): ... this.
11006 (ipa_fn_summary::reset): Transform to ...
11007 (ipa_fn_summary::~ipa_fn_summary): ... this.
11008 (ipa_fn_summary_t::remove): Rename to ...
11009 (ipa_fn_summary_t::remove_callees): ... this.
11010 (ipa_fn_summary_t::duplicate): Use placement new
11011 instead of memory copy.
11012 (ipa_call_summary_t::duplicate): Likewise.
11013 (ipa_call_summary_t::remove): Remove.
11014 (dump_ipa_call_summary): Change get_create to get.
11015 (ipa_dump_fn_summary): Dump only when summary exists.
11016 (analyze_function_body): Use symbol_summary::get instead
11018 (compute_fn_summary): Likewise.
11019 (estimate_edge_devirt_benefit): Likewise.
11020 (estimate_edge_size_and_time): Likewise.
11021 (inline_update_callee_summaries): Likewise.
11022 (remap_edge_change_prob): Likewise.
11023 (remap_edge_summaries): Likewise.
11024 (ipa_merge_fn_summary_after_inlining): Likewise.
11025 (write_ipa_call_summary): Likewise.
11026 (ipa_fn_summary_write): Likewise.
11027 (ipa_free_fn_summary): Likewise.
11028 * ipa-fnsummary.h (struct GTY): Add new ctor and copy ctor.
11029 (struct ipa_call_summary): Likewise.
11030 * ipa-icf.c (sem_function::merge): Use symbol_summary::get instead
11032 * ipa-inline-analysis.c (do_estimate_edge_time): Likewise.
11033 (estimate_size_after_inlining): Likewise.
11034 (estimate_growth): Likewise.
11035 (growth_likely_positive): Likewise.
11036 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
11037 (inline_call): Likewise.
11038 * ipa-inline.c (caller_growth_limits): Likewise.
11039 (can_inline_edge_p): Likewise.
11040 (can_inline_edge_by_limits_p): Likewise.
11041 (compute_uninlined_call_time): Likewise.
11042 (compute_inlined_call_time): Likewise.
11043 (want_inline_small_function_p): Likewise.
11044 (edge_badness): Likewise.
11045 (update_caller_keys): Likewise.
11046 (update_callee_keys): Likewise.
11047 (inline_small_functions): Likewise.
11048 (inline_to_all_callers_1): Likewise.
11049 (dump_overall_stats): Likewise.
11050 (early_inline_small_functions): Likewise.
11051 (early_inliner): Likewise.
11052 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
11053 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
11054 * ipa-pure-const.c (malloc_candidate_p): Likewise.
11055 * ipa-split.c (execute_split_functions): Likewise.
11056 * symbol-summary.h: Likewise.
11057 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
11059 2018-06-19 Richard Biener <rguenther@suse.de>
11061 * tree-vectorizer.c (try_vectorize_loop_1): Split out of ...
11062 (vectorize_loops): ... here. Fix dbgcnt handling.
11063 (try_vectorize_loop): Wrap try_vectorize_loop_1.
11065 2018-06-19 Segher Boessenkool <segher@kernel.crashing.org>
11068 * config/rs6000/rs6000.md (rs6000_discover_homogeneous_aggregate): An
11069 ieee128 argument takes up only one (vector) register, not two (floating
11072 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
11074 * gimplify.c (gimplify_init_constructor): Really never clear for an
11075 incomplete constructor if CONSTRUCTOR_NO_CLEARING is set.
11077 2018-06-19 Richard Biener <rguenther@suse.de>
11079 PR tree-optimization/86179
11080 * tree-vect-patterns.c (vect_pattern_recog_1): Clean up
11081 after failed recognition.
11083 2018-06-18 Martin Sebor <msebor@redhat.com>
11085 PR middle-end/85602
11086 * calls.c (maybe_warn_nonstring_arg): Handle strncat.
11087 * tree-ssa-strlen.c (is_strlen_related_p): Make extern.
11088 Handle integer subtraction.
11089 (maybe_diag_stxncpy_trunc): Handle nonstring source arguments.
11090 * tree-ssa-strlen.h (is_strlen_related_p): Declare.
11092 2018-06-18 David Malcolm <dmalcolm@redhat.com>
11094 * config/frv/frv-protos.h (frv_ifcvt_modify_insn): Strengthen 3rd
11095 param from rtx to rtx_insn *.
11096 * config/frv/frv.c (frv_ifcvt_add_insn): Likewise for "insn"
11098 (frv_ifcvt_modify_insn): Likwise.
11099 (frv_ifcvt_modify_final): Likwise for local "existing_insn",
11100 adding an as_a <rtx_insn *> cast. Likewise for local "insn".
11101 * config/mips/mips.c (r10k_insert_cache_barriers): Add an
11102 as_a <rtx_insn *> cast to local "unprotected_region" once
11103 it's been established that it's not NULL or pc_rtx.
11104 * config/nds32/nds32-relax-opt.c (nds32_group_insns): Strengthen
11105 param "sethi" from rtx to rtx_insn *.
11106 (nds32_group_float_insns): Likewise for param "insn".
11107 * config/vax/vax-protos.h (vax_output_int_add): Likewise for 1st
11109 (vax_output_int_subtract): Likewise.
11110 * config/vax/vax.c (vax_output_int_add): Likewise for param
11112 (vax_output_int_subtract): Likewise.
11113 * emit-rtl.c (set_insn_deleted): Likewise, removing cast.
11114 (emit_pattern_after): Likewise for param "after".
11115 (emit_insn_after): Likewise.
11116 (emit_jump_insn_after): Likewise.
11117 (emit_call_insn_after): Likewise.
11118 (emit_debug_insn_after): Likewise.
11119 (emit_pattern_before): Likewise for param "before".
11120 (emit_insn_before): Likewise.
11121 (emit_jump_insn_before): Likewise.
11122 * final.c (get_insn_template): Likewise for param "insn", removing
11124 * output.h (get_insn_template): Likewise for 2nd param.
11125 * rtl.h (emit_insn_before): Likewise.
11126 (emit_jump_insn_before): Likewise.
11127 (emit_debug_insn_before_noloc): Likewise.
11128 (emit_insn_after): Likewise.
11129 (emit_jump_insn_after): Likewise.
11130 (emit_call_insn_after): Likewise.
11131 (emit_debug_insn_after): Likewise.
11132 (set_insn_deleted): Likewise for param.
11134 2018-06-18 Michael Meissner <meissner@linux.ibm.com>
11137 * config/rs6000/rs6000-modes.def (toplevel): Rework the 128-bit
11138 floating point modes, so that IFmode is numerically greater than
11139 TFmode, which is greater than KFmode using FRACTIONAL_FLOAT_MODE
11140 to declare the ordering. This prevents IFmode from being
11141 converted to TFmode when long double is IEEE 128-bit on an ISA 3.0
11142 machine. Include rs6000-modes.h to share the fractional values
11143 between genmodes* and the rest of the compiler.
11144 (IFmode): Likewise.
11145 (KFmode): Likewise.
11146 (TFmode): Likewise.
11147 * config/rs6000/rs6000-modes.h: New file.
11148 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Change the
11149 meaning of rs6000_long_double_size so that 126..128 selects an
11150 appropriate 128-bit floating point type.
11151 (rs6000_option_override_internal): Likewise.
11152 * config/rs6000/rs6000.h (toplevel): Include rs6000-modes.h.
11153 (TARGET_LONG_DOUBLE_128): Change the meaning of
11154 rs6000_long_double_size so that 126..128 selects an appropriate
11155 128-bit floating point type.
11156 (LONG_DOUBLE_TYPE_SIZE): Update comment.
11157 * config/rs6000/rs6000.md (trunciftf2): Correct the modes of the
11158 source and destination to match the standard usage.
11159 (truncifkf2): Likewise.
11160 (copysign<mode>3, IEEE iterator): Rework copysign of float128 on
11161 ISA 2.07 to use an explicit clobber, instead of passing in a
11163 (copysign<mode>3_soft): Likewise.
11165 2018-06-18 David Malcolm <dmalcolm@redhat.com>
11167 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
11168 Replace dump_printf_loc call with DUMP_VECT_SCOPE.
11169 (vect_slp_analyze_instance_dependence): Likewise.
11170 (vect_enhance_data_refs_alignment): Likewise.
11171 (vect_analyze_data_refs_alignment): Likewise.
11172 (vect_slp_analyze_and_verify_instance_alignment
11173 (vect_analyze_data_ref_accesses): Likewise.
11174 (vect_prune_runtime_alias_test_list): Likewise.
11175 (vect_analyze_data_refs): Likewise.
11176 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Likewise.
11177 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
11178 (vect_analyze_scalar_cycles_1): Likewise.
11179 (vect_get_loop_niters): Likewise.
11180 (vect_analyze_loop_form_1): Likewise.
11181 (vect_update_vf_for_slp): Likewise.
11182 (vect_analyze_loop_operations): Likewise.
11183 (vect_analyze_loop): Likewise.
11184 (vectorizable_induction): Likewise.
11185 (vect_transform_loop): Likewise.
11186 * tree-vect-patterns.c (vect_pattern_recog): Likewise.
11187 * tree-vect-slp.c (vect_analyze_slp): Likewise.
11188 (vect_make_slp_decision): Likewise.
11189 (vect_detect_hybrid_slp): Likewise.
11190 (vect_slp_analyze_operations): Likewise.
11191 (vect_slp_bb): Likewise.
11192 * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Likewise.
11193 (vectorizable_bswap): Likewise.
11194 (vectorizable_call): Likewise.
11195 (vectorizable_simd_clone_call): Likewise.
11196 (vectorizable_conversion): Likewise.
11197 (vectorizable_assignment): Likewise.
11198 (vectorizable_shift): Likewise.
11199 (vectorizable_operation): Likewise.
11200 * tree-vectorizer.h (DUMP_VECT_SCOPE): New macro.
11202 2018-06-18 Martin Sebor <msebor@redhat.com>
11204 PR tree-optimization/81384
11205 * builtin-types.def (BT_FN_SIZE_CONST_STRING_SIZE): New.
11206 * builtins.c (expand_builtin_strnlen): New function.
11207 (expand_builtin): Call it.
11208 (fold_builtin_n): Avoid setting TREE_NO_WARNING.
11209 * builtins.def (BUILT_IN_STRNLEN): New.
11210 * calls.c (maybe_warn_nonstring_arg): Handle BUILT_IN_STRNLEN.
11211 Warn for bounds in excess of maximum object size.
11212 * tree-ssa-strlen.c (maybe_set_strlen_range): Return tree representing
11213 single-value ranges. Handle strnlen.
11214 (handle_builtin_strlen): Handle strnlen.
11215 (strlen_check_and_optimize_stmt): Same.
11216 * doc/extend.texi (Other Builtins): Document strnlen.
11218 2018-06-18 Maya Rashish <coypu@sdf.org>
11220 * config/alpha/openbsd.h (TARGET_DEFAULT): Define.
11221 (LINK_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): Likewise.
11222 (INTMAX_TYPE, UINTMAX_TYPE, WINT_TYPE): Likewise.
11224 * config/alpha/elf.h (STARTFILE_SPEC, ENDFILE_SPEC): Move from
11226 * config/alpha/linux.h (STARTFILE_SPEC, ENDFILE_SPEC): Here.
11228 2018-06-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
11230 * tree.c (escaped_string::escape): Replace cast to char * by
11231 const_cast<char *> (unescaped).
11233 2018-06-18 Nick Clifton <nickc@redhat.com>
11236 * tree.c (escaped_string): New class. Converts an unescaped
11237 string into its escaped equivalent.
11238 (warn_deprecated_use): Use the new class to convert the
11239 deprecation message, if present.
11240 (test_escaped_strings): New self test.
11241 (test_c_tests): Add test_escaped_strings.
11242 * doc/extend.texi (deprecated): Add a note that the
11243 deprecation message is affected by the -fmessage-length
11244 option, and that control characters will be escaped.
11245 (#pragma GCC error): Document this pragma.
11246 (#pragma GCC warning): Likewise.
11247 * doc/invoke.texi (-fmessage-length): Document this option's
11248 effect on the #warning and #error preprocessor directives and
11249 the deprecated attribute.
11251 2018-06-18 Eric Botcazou <ebotcazou@adacore.com>
11253 * tree.c (decl_value_expr_lookup): Revert latest change.
11254 (decl_value_expr_insert): Likewise.
11256 2018-06-17 Eric Botcazou <ebotcazou@adacore.com>
11258 * gimplify.c (nonlocal_vlas): Delete.
11259 (nonlocal_vla_vars): Likewise.
11260 (gimplify_var_or_parm_decl): Do not add debug VAR_DECLs for non-local
11262 (gimplify_body): Do not create and destroy nonlocal_vlas.
11263 * tree-nested.c: Include diagnostic.h.
11264 (use_pointer_in_frame): Tweak.
11265 (lookup_field_for_decl): Add assertion and declare the transformation.
11266 (convert_nonlocal_reference_op) <PARM_DECL>: Rework and issue an
11267 internal error when the reference is in a wrong context. Do not
11268 create a debug decl by default.
11269 (note_nonlocal_block_vlas): Delete.
11270 (convert_nonlocal_reference_stmt) <GIMPLE_BIND>: Do not call it.
11271 (convert_local_reference_op) <PARM_DECL>: Skip the frame decl. Do not
11272 create a debug decl by default.
11273 (convert_gimple_call) <GIMPLE_CALL>: Issue an internal error when the
11274 call is in a wrong context.
11275 (fixup_vla_decls): New function.
11276 (finalize_nesting_tree_1): Adjust comment. Call fixup_vla_decls if no
11277 debug variables were created.
11278 * tree.c (decl_value_expr_lookup): Add checking assertion.
11279 (decl_value_expr_insert): Likewise.
11281 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
11283 PR middle-end/82479
11284 * ipa-fnsummary.c (will_be_nonconstant_expr_predicate): Handle CALL_EXPR.
11285 * tree-scalar-evolution.c (interpret_expr): Likewise.
11286 (expression_expensive_p): Likewise.
11287 * tree-ssa-loop-ivopts.c (contains_abnormal_ssa_name_p): Likewise.
11288 * tree-ssa-loop-niter.c (number_of_iterations_popcount): New.
11289 (number_of_iterations_exit_assumptions): Use number_of_iterations_popcount.
11290 (ssa_defined_by_minus_one_stmt_p): New.
11292 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
11294 PR middle-end/64946
11295 * cfgexpand.c (expand_debug_expr): Hande ABSU_EXPR.
11296 * config/i386/i386.c (ix86_add_stmt_cost): Likewise.
11297 * dojump.c (do_jump): Likewise.
11298 * expr.c (expand_expr_real_2): Check operand type's sign.
11299 * fold-const.c (const_unop): Handle ABSU_EXPR.
11300 (fold_abs_const): Likewise.
11301 * gimple-pretty-print.c (dump_unary_rhs): Likewise.
11302 * gimple-ssa-backprop.c (backprop::process_assign_use): Likesie.
11303 (strip_sign_op_1): Likesise.
11304 * match.pd: Add new pattern to generate ABSU_EXPR.
11305 * optabs-tree.c (optab_for_tree_code): Handle ABSU_EXPR.
11306 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
11307 * tree-eh.c (operation_could_trap_helper_p): Likewise.
11308 * tree-inline.c (estimate_operator_cost): Likewise.
11309 * tree-pretty-print.c (dump_generic_node): Likewise.
11310 * tree-vect-patterns.c (vect_recog_sad_pattern): Likewise.
11311 * tree.def (ABSU_EXPR): New.
11313 2018-06-16 Jakub Jelinek <jakub@redhat.com>
11315 PR middle-end/86095
11316 * common.opt (Wunsafe-loop-optimizations): Add Ignore, remove Var,
11317 documented as preserved for backward compatibility only.
11318 * doc/invoke.texi: Remove -Wunsafe-loop-optimizations documentation.
11320 PR rtl-optimization/86108
11321 * bb-reorder.c (create_forwarder_block): Renamed to ...
11322 (create_eh_forwarder_block): ... this. Split OLD_BB after labels and
11323 jump from new landing pad to the second part.
11324 (sjlj_fix_up_crossing_landing_pad, dw2_fix_up_crossing_landing_pad):
11327 2018-06-15 Jakub Jelinek <jakub@redhat.com>
11329 PR middle-end/85878
11330 * expr.c (expand_assignment): Remove now redundant COMPLEX_MODE_P
11331 check from first store_expr, use to_mode instead of GET_MODE (to_rtx).
11332 Only call store_expr for halves if the mode is the same.
11334 PR middle-end/86123
11335 * match.pd ((X / Y) == 0 -> X < Y): Don't transform complex divisions.
11336 Fix up comment formatting.
11338 2018-06-15 Bernd Edlinger <bernd.edlinger@hotmail.de>
11340 * typed-splay-tree.h (typed_splay_tree::remove): New function.
11341 (typed_splay_tree::closure,
11342 typed_splay_tree::inner_foreach_fn, typed_splay_tree::m_inner): Deleted.
11343 (typed_splay_tree::typed_splay_tree,
11344 typed_splay_tree::operator =): Declared private.
11345 (typed_splay_tree::splay_tree_key, typed_splay_tree::splay_tree_value,
11346 typed_splay_tree::splay_tree_node_s, typed_splay_tree::KDEL,
11347 typed_splay_tree::VDEL, typed_splay_tree::splay_tree_delete_helper,
11348 typed_splay_tree::rotate_left, typed_splay_tree::rotate_right,
11349 typed_splay_tree::splay_tree_splay,
11350 typed_splay_tree::splay_tree_foreach_helper,
11351 typed_splay_tree::splay_tree_insert,
11352 typed_splay_tree::splay_tree_remove,
11353 typed_splay_tree::splay_tree_lookup,
11354 typed_splay_tree::splay_tree_predecessor,
11355 typed_splay_tree::splay_tree_successor,
11356 typed_splay_tree::splay_tree_min,
11357 typed_splay_tree::splay_tree_max): Took over from splay-tree.c/.h.
11358 (typed_splay_tree::root, typed_splay_tree::comp,
11359 typed_splay_tree::delete_key,
11360 typed_splay_tree::delete_value): New data members.
11361 * typed-splay-tree.c (selftest::test_str_to_int): Add a test for
11362 typed_splay_tree::remove.
11364 2018-06-15 Matthew Fortune <matthew.fortune@mips.com>
11366 * config/mips/mips.h (ASM_SPEC): Pass through -mcrc, -mno-crc,
11367 -mginv and -mno-ginv to the assembler.
11368 * config/mips/mips.opt (-mcrc): New option.
11369 (-mginv): Likewise.
11370 * doc/invoke.text (-mcrc): Document.
11371 (-mginv): Likewise.
11373 2018-06-15 Nick Clifton <nickc@redhat.com>
11376 * tree.c (escaped_string): New class. Converts an unescaped
11377 string into its escaped equivalent.
11378 (warn_deprecated_use): Use the new class to convert the
11379 deprecation message, if present.
11380 (test_escaped_strings): New self test.
11381 (test_c_tests): Add test_escaped_strings.
11382 * doc/extend.texi (deprecated): Add a note that the
11383 deprecation message is affected by the -fmessage-length
11384 option, and that control characters will be escaped.
11385 (#pragma GCC error): Document this pragma.
11386 (#pragma GCC warning): Likewise.
11387 * doc/invoke.texi (-fmessage-length): Document this option's
11388 effect on the #warning and #error preprocessor directives and
11389 the deprecated attribute.
11391 2018-06-15 Richard Biener <rguenther@suse.de>
11393 * tree-vect-slp.c (vect_slp_bb): Dump MSG_OPTIMIZED_LOCATIONS
11394 here, also noting vector size used.
11395 * tree-vectorizer.c (vectorize_loops): Adjust. Note vector
11396 size used in MSG_OPTIMIZED_LOCATIONS dump.
11397 (pass_slp_vectorize::execute): Adjust.
11399 2018-06-15 Claudiu Zissulescu <claziss@synopsys.com>
11402 * config/arc/arc.c (arc_return_address_register): Fix
11405 2018-06-15 Richard Biener <rguenther@suse.de>
11407 PR middle-end/86159
11408 * tree-cfg.c (gimplify_build3): Do not strip sign conversions,
11409 leave useless conversion stripping to force_gimple_operand_gsi.
11410 (gimplify_build2): Likewise.
11411 (gimplify_build1): Likewise.
11413 2018-06-15 Richard Biener <rguenther@suse.de>
11415 PR middle-end/86076
11416 * tree-cfg.c (move_stmt_op): unshare invariant addresses
11417 before adjusting their block.
11419 2018-06-15 Sebastian Huber <sebastian.huber@embedded-brains.de>
11421 * config.gcc (riscv*-*-elf* | riscv*-*-rtems*): Use custom
11422 multilibs for *-*-rtems*.
11423 * config/riscv/t-rtems: New file.
11425 2018-06-14 Jakub Jelinek <jakub@redhat.com>
11427 PR middle-end/86122
11428 * match.pd ((A +- CST1) +- CST2): Punt if last resort
11429 unsigned_type_for returns NULL.
11432 * lower-subreg.c (find_decomposable_subregs): Don't decompose float
11433 subregs of multi-word pseudos unless the float mode has word size.
11435 2018-06-14 Richard Biener <rguenther@suse.de>
11437 PR middle-end/86139
11438 * tree-vect-generic.c (build_word_mode_vector_type): Remove
11439 duplicate and harmful type_hash_canon.
11440 * tree.c (type_hash_canon): Assert we didn't find ourselves.
11442 2018-06-14 Richard Biener <rguenther@suse.de>
11445 * tree-ssa-struct-alias.c (create_variable_info_for): Handle
11448 2018-06-14 Sebastian Huber <sebastian.huber@embedded-brains.de>
11450 * config/rtems.h (STDINT_LONG32): Define.
11452 2018-06-13 Matthew Fortune <matthew.fortune@mips.com>
11453 Prachi Godbole <prachi.godbole@imgtec.com>
11455 * config/mips/mips-cpus.def: Define P6600.
11456 * config/mips/mips-tables.opt: Regenerate.
11457 * config/mips/mips.c (mips_ucbranch_type): New enum.
11458 (mips_rtx_cost_data): Add support for P6600.
11459 (mips_issue_rate): Likewise.
11460 (mips_multipass_dfa_lookahead): Likewise.
11461 (mips_avoid_hazard): Likewise.
11462 (mips_reorg_process_insns): Likewise.
11463 (mips_classify_branch_p6600): New function.
11464 * config/mips/mips.h (TUNE_P6600): New define.
11465 (MIPS_ISA_LEVEL_SPEC): Infer mips64r6 from p6600.
11466 (ENABLE_LD_ST_PAIRS): Enable load/store bonding for p6600.
11467 * config/mips/mips.md: Include p6600.md.
11468 (processor): Add p6600.
11469 * config/mips/p6600.md: New file.
11470 * doc/invoke.texi: Add p6600 to supported architectures.
11472 2018-06-13 Martin Sebor <msebor@redhat.com>
11474 PR tree-optimization/86114
11475 * gimple-fold.c (gimple_fold_builtin_strlen): Only handle LHS
11477 * tree-ssa-strlen.c (maybe_set_strlen_range): Same.
11479 2018-06-13 Richard Biener <rguenther@suse.de>
11481 * tree-vect-patterns.c (vect_recog_vector_vector_shift_pattern):
11482 Properly set vector type of the intermediate stmt.
11483 * tree-vect-stmts.c (vectorizable_operation): The destination
11484 var always has vectype_out type.
11486 2018-06-13 Jeff Law <law@redhat.com>
11488 * config/rl78/rl78.c (move_elim_pass): Use TDF_NONE rather than
11489 integer 0 for argument to print_rtl_with_bb.
11490 (rl78_reorg): Likewise.
11492 2018-06-13 David Malcolm <dmalcolm@redhat.com>
11494 * config/arc/arc.c (hwloop_optimize): Strengthen local "end_label"
11495 from rtx to rtx_insn *.
11496 * config/bfin/bfin.c (hwloop_optimize): Likewise for local
11498 (add_sched_insns_for_speculation): Likewise for local "target",
11499 converting usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
11500 * config/c6x/c6x.c (reorg_split_calls): Strengthen param "call_labels"
11501 from rtx_insn ** to rtx_code_label **.
11502 (reorg_emit_nops): Likewise.
11503 (c6x_reorg): Likewise for local "call_labels".
11504 * config/sh/sh-protos.h (get_dest_uid): Strengthen 1st param from
11506 * config/sh/sh.c (dump_table): Strengthen local "lab" from rtx to
11507 rtx_code_label *, adding safe_as_a <rtx_code_label *> casts to
11508 the loops over LABEL_REFS.
11509 (fixup_addr_diff_vecs): Add as_a <rtx_insn *> to usage of
11511 (barrier_align): Convert usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
11512 (get_dest_uid): Strengthen param "label" from rtx to rtx_insn *.
11513 (split_branches): Strengthen local "olabel" from rtx to
11514 rtx_insn *, adding a safe_as_a cast.
11515 * emit-rtl.c (next_real_insn): Strengthen param from "rtx"
11517 (add_insn_after): Likewise for first two params.
11518 (add_insn_before): Likewise.
11519 (remove_insn): Likewise for param.
11520 (emit_pattern_before_noloc): Likewise for second and third params.
11521 (emit_jump_insn_before_noloc): Convert NULL_RTX to NULL.
11522 (emit_call_insn_before_noloc): Likewise.
11523 (emit_debug_insn_before_noloc): Strengthen "before" param from "rtx"
11525 (emit_barrier_before): Likewise.
11526 (emit_label_before): Strengthen "label" param from "rtx" to
11527 "rtx_code_label *". Strengthen "before" param from "rtx" to
11529 (emit_insn_after_1): Strengthen "after" param from "rtx" to
11531 (emit_pattern_after_noloc): Likewise.
11532 (emit_insn_after_noloc): Likewise.
11533 (emit_jump_insn_after_noloc): Likewise.
11534 (emit_call_insn_after_noloc): Likewise.
11535 (emit_debug_insn_after_noloc): Likewise.
11536 (emit_barrier_after): Likewise.
11537 (emit_label_after): Likewise for both params.
11538 (emit_pattern_after_setloc): Likewise for "after" param. Convert
11539 "loc" param from "int" to "location_t".
11540 (emit_insn_after_setloc): Likewise.
11541 (emit_jump_insn_after_setloc): Likewise.
11542 (emit_call_insn_after_setloc): Likewise.
11543 (emit_debug_insn_after_setloc): Likewise.
11544 (emit_pattern_before_setloc): Likewise for "before" param. Convert
11545 "loc" param from "int" to "location_t".
11546 (emit_pattern_before): Convert NULL_RTX to NULL.
11547 (emit_insn_before_setloc): Convert "loc" param from "int" to
11549 (emit_jump_insn_before_setloc): Likewise.
11550 (emit_call_insn_before_setloc): Likewise.
11551 (emit_debug_insn_before_setloc): Strengthen "before" param from rtx to
11552 rtx_insn *. Convert "loc" param from "int" to "location_t".
11553 * rtl.h (emit_insn_before_setloc, emit_jump_insn_before_setloc,
11554 emit_call_insn_before_setloc, emit_debug_insn_before_setloc):
11555 Convert 3rd param from "int" to "location_t".
11556 (emit_barrier_before, emit_barrier_after, next_real_insn):
11557 Strengthen param from rtx to rtx_insn *.
11558 (emit_label_before): Strengthen 1st param from "rtx" to
11559 "rtx_code_label *". Strengthen 2nd param from "rtx" to
11561 (emit_insn_after_noloc, emit_jump_insn_after_noloc,
11562 emit_call_insn_after_noloc, emit_debug_insn_after_noloc):
11563 Strengthen 2nd param from "rtx" to "rtx_insn *".
11564 (emit_insn_after_setloc, emit_jump_insn_after_setloc)
11565 emit_call_insn_after_setloc, emit_debug_insn_after_setloc):
11566 Likewise. Convert 3rd param from "int" to "location_t".
11567 (emit_label_after): Strengthen 1st param from "rtx" to
11568 "rtx_code_label *".
11569 (next_real_insn, remove_insn): Strengthen param from "rtx" to
11571 (add_insn_before, add_insn_after): Strengthen 1st and 2nd params
11572 from "rtx" to "rtx_insn *".
11574 2018-06-13 Jan Hubicka <hubicka@gcc.gnu.org>
11576 * cgraph.c (cgraph_node::get_untransformed_body): Dump function
11577 bodies streamed in with -Q.
11578 * dumpfile.c (dump_files): Add lto-stream-out dump file.
11579 * dumpfile.h (tree_dump_index): Add lto_stream_out.
11580 * gimple-streamer-out.c: Include gimple-pretty-print.h
11581 (output_bb): Dump stmts streamed.
11582 * lto-section-out.c: Include print-tree.h
11583 (lto_begin_section): Dump sections created.
11584 (lto_output_decl_index): Dump decl encoded.
11585 * lto-streamer-out.c: Include print-tree.h
11586 (create_output_block): Dump output block created.
11587 (DFS::DFS_write_tree_body): Dump DFS SCCs streamed.
11588 (output_function): Dump function output.
11589 (output_constructor): Dump constructor streamed.
11590 (write_global_stream): Output indexes encoded.
11591 (produce_asm_for_decls): Dump streams encoded.
11592 * lto-streamer.c (streamer_dump_file): New global var.
11593 * lto-streamer.h (streamer_dump_file): Declare.
11594 * passes.c (ipa_write_summaries): Initialize streamer dump.
11595 * varpool.c (varpool_node::get_constructor): Dump constructors streamed
11598 2018-06-13 Eric Botcazou <ebotcazou@adacore.com>
11601 * config/i386/winnt.c (i386_pe_seh_cold_init): Do not emit negative
11602 offsets for register save directives. Emit a second batch of save
11603 directives, if need be, when the function accesses prior frames.
11605 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
11607 * config/arc/fpu.md (fmasf4): Force operand to register.
11608 (fnmasf4): Likewise.
11610 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
11612 * config/arc/arc-protos.h (arc_pad_return): Remove.
11613 * config/arc/arc.c (machine_function): Remove force_short_suffix
11615 (arc_print_operand): Adjust printing of '&'.
11616 (arc_verify_short): Remove conditional printing of short suffix.
11617 (arc_final_prescan_insn): Remove reference to size_reason.
11618 (pad_return): New function.
11619 (arc_reorg): Call pad_return.
11620 (arc_pad_return): Remove.
11621 (arc_init_machine_status): Remove reference to force_short_suffix.
11622 * config/arc/arc.md (vunspec): Add VUNSPEC_ARC_BLOCKAGE.
11623 (attr length): When attribute iscompact is true force to 2
11624 regardless; in the case of maybe check if we want to force the
11625 instruction to have 4 bytes length.
11626 (nopv): Change it to generate 4 byte long nop as well.
11627 (blockage): New pattern.
11628 (simple_return): Remove call to arc_pad_return.
11629 (p_return_i): Likewise.
11631 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
11633 * config/arc/elf.h (LINK_GCC_C_SEQUENCE_SPEC): Define.
11635 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
11637 * config/arc/builtins.def (SYNC): SYNC instruction is valid on all
11640 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
11642 * config/arc/arc.c (atomic_exchangesi): EX instruction is default
11643 for ARC700 and ARCv2.
11645 2018-06-13 Chenghua Xu <paul.hua.gm@gmail.com>
11648 * config/mips/loongson.md (vec_setv4hi): Gen_lowpart for
11649 operands[2] instead of operands[1].
11652 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
11654 * lra-constraints.c (simplify_operand_subreg): In the paradoxical
11655 case, check whether the outer register overlaps an unallocatable
11656 register, not just whether it fits the required class.
11658 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
11660 * poly-int.h (can_div_trunc_p): Add new overload in which all values
11662 * alias.c (get_addr): Extend CONST_INT handling to poly_int_rtx_p.
11663 (memrefs_conflict_p): Likewise.
11664 (init_alias_analysis): Likewise.
11665 * cfgexpand.c (expand_debug_expr): Likewise.
11666 * combine.c (combine_simplify_rtx, force_int_to_mode): Likewise.
11667 * cse.c (fold_rtx): Likewise.
11668 * explow.c (adjust_stack, anti_adjust_stack): Likewise.
11669 * expr.c (emit_block_move_hints): Likewise.
11670 (clear_storage_hints, push_block, emit_push_insn): Likewise.
11671 (store_expr_with_bounds, reduce_to_bit_field_precision): Likewise.
11672 (emit_group_load_1): Use rtx_to_poly_int64 for group offsets.
11673 (emit_group_store): Likewise.
11674 (find_args_size_adjust): Use strip_offset. Use rtx_to_poly_int64
11675 to read the PRE/POST_MODIFY increment.
11676 * calls.c (store_one_arg): Use strip_offset.
11677 * rtlanal.c (rtx_addr_can_trap_p_1): Extend CONST_INT handling to
11679 (set_noop_p): Use rtx_to_poly_int64 for the elements selected
11681 * simplify-rtx.c (avoid_constant_pool_reference): Use strip_offset.
11682 (simplify_binary_operation_1): Extend CONST_INT handling to
11684 * var-tracking.c (compute_cfa_pointer): Take a poly_int64 rather
11685 than a HOST_WIDE_INT.
11686 (hard_frame_pointer_adjustment): Change from HOST_WIDE_INT to
11688 (adjust_mems, add_stores): Update accodingly.
11689 (vt_canonicalize_addr): Track polynomial offsets.
11690 (emit_note_insn_var_location): Likewise.
11691 (vt_add_function_parameter): Likewise.
11692 (vt_initialize): Likewise.
11694 2018-06-12 Jeff Law <law@redhat.com>
11696 * config.gcc (alpha*-*-freebsd*): Remove.
11697 * config/alpha/freebsd.h: Remove.
11699 2018-06-12 David Malcolm <dmalcolm@redhat.com>
11702 * spellcheck-tree.c (levenshtein_distance): Rename to...
11703 (get_edit_distance): ...this, and update for underlying renaming.
11704 * spellcheck-tree.h (levenshtein_distance): Rename to...
11705 (get_edit_distance): ...this.
11706 * spellcheck.c (levenshtein_distance): Rename to...
11707 (get_edit_distance): ...this. Convert from Levenshtein distance
11708 to Damerau-Levenshtein distance by supporting transpositions of
11709 adjacent characters. Rename "v1" to "v_next" and "v0" to
11711 (selftest::levenshtein_distance_unit_test_oneway): Rename to...
11712 (selftest::test_edit_distance_unit_test_oneway): ...this, and
11713 update for underlying renaming.
11714 (selftest::levenshtein_distance_unit_test): Rename to...
11715 (selftest::test_get_edit_distance_unit): ...this, and update for
11716 underlying renaming.
11717 (selftest::test_find_closest_string): Add example from PR 69968
11718 where transposition helps
11719 (selftest::test_metric_conditions): Update for renaming.
11720 (selftest::test_metric_conditions): Likewise.
11721 (selftest::spellcheck_c_tests): Likewise.
11722 * spellcheck.h (levenshtein_distance): Rename both overloads to...
11723 (get_edit_distance): ...this.
11724 (best_match::consider): Update for renaming.
11726 2018-06-12 Martin Sebor <msebor@redhat.com>
11728 PR tree-optimization/85259
11729 * builtins.c (compute_objsize): Handle constant offsets.
11730 * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Return
11731 true iff a warning has been issued.
11732 * gimple.h (gimple_nonartificial_location): New function.
11733 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Call
11734 gimple_nonartificial_location and handle -Wno-system-headers.
11735 (handle_builtin_stxncpy): Same.
11737 2018-06-12 Martin Sebor <msebor@redhat.com>
11740 * fold-const.c (operand_equal_p): Handle SAVE_EXPR.
11742 2018-06-12 Will Schmidt <will_schmidt@vnet.ibm.com>
11744 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
11745 BUILTIN_VEC_XST entries for pointer to double and long long.
11747 2018-06-12 H.J. Lu <hongjiu.lu@intel.com>
11750 * config/i386/gnu-user.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
11752 * config/i386/gnu-user64.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
11755 2018-06-12 Martin Liska <mliska@suse.cz>
11757 * doc/options.texi: Document IntegerRange.
11759 2018-06-12 Martin Liska <mliska@suse.cz>
11761 * config/i386/i386.opt: Make MPX-related options as Deprecated.
11762 * opt-functions.awk: Handle Deprecated flag.
11763 * opts-common.c (decode_cmdline_option): Handle cl_deprecated
11765 (read_cmdline_option): Report warning for a deprecated option.
11766 * opts.h (struct cl_option): Add new field cl_deprecated.
11767 (CL_ERR_DEPRECATED): New.
11769 2018-06-12 Martin Liska <mliska@suse.cz>
11771 * doc/options.texi: Document Deprecated option flag.
11773 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
11775 * config/arc/arc-arch.h (arc_extras): New enum.
11776 (arc_cpu_t):Add field extra.
11777 (arc_cpu_types): Consider the extras.
11778 * config/arc/arc-cpus.def: Add extras info.
11779 * config/arc/arc-opts.h (processor_type): Consider extra field.
11780 * config/arc/arc.c (arc_override_options): Handle extra field.
11782 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
11784 * config/arc/arc-arch.h: Update ARC_OPTX macro.
11785 * config/arc/arc-options.def (ARC_OPTX): Introduce a new doc
11787 * config/arc/arc.c (arc_init): Update pic warning.
11788 (irq_range): Update irq range parsing warnings.
11789 (arc_override_options): Update various warning messages.
11790 (arc_handle_aux_attribute): Likewise.
11792 2018-06-12 Robert Suchanek <robert.suchanek@mips.com>
11794 * config/mips/i6400.md (i6400_fpu_fadd): Remove frint.
11796 2018-06-12 Jozef Lawrynowicz <jozef.l@mittosystems.com>
11798 * doc/sourcebuild.texi: Document usage of line number 0 in verify
11799 compiler messages directives.
11801 2018-06-12 Matthew Fortune <mfortune@gmail.com>
11803 * config/mips/mips-cpus.def: New MIPS_CPU for i6500.
11804 * config/mips/mips-tables.opt: Regenerate.
11805 * config/mips/mips.h (MIPS_ISA_LEVEL_SPEC): Mark i6500 as
11807 * doc/invoke.texi: Document -march=i6500.
11809 2018-06-12 Prachi Godbole <prachi.godbole@imgtec.com>
11811 * config/mips/i6400.md (i6400_gpmuldiv): Remove cpu_unit.
11812 (i6400_gpmul): Add cpu_unit.
11813 (i6400_gpdiv): Likewise.
11814 (i6400_msa_add_d): Update reservations.
11815 (i6400_msa_int_add) Likewise.
11816 (i6400_msa_short_logic3) Likewise.
11817 (i6400_msa_short_logic2) Likewise.
11818 (i6400_msa_short_logic) Likewise.
11819 (i6400_msa_move) Likewise.
11820 (i6400_msa_cmp) Likewise.
11821 (i6400_msa_short_float2) Likewise.
11822 (i6400_msa_div_d) Likewise.
11823 (i6400_msa_long_logic1) Likewise.
11824 (i6400_msa_long_logic2) Likewise.
11825 (i6400_msa_mult) Likewise.
11826 (i6400_msa_long_float2) Likewise.
11827 (i6400_msa_long_float4) Likewise.
11828 (i6400_msa_long_float5) Likewise.
11829 (i6400_msa_long_float8) Likewise.
11830 (i6400_fpu_fadd): Include frint type.
11831 (i6400_fpu_store): New define_insn_reservation.
11832 (i6400_fpu_load): Likewise.
11833 (i6400_fpu_move): Likewise.
11834 (i6400_fpu_fcmp): Likewise.
11835 (i6400_fpu_fmadd): Likewise.
11836 (i6400_int_mult): Include imul3nc type and update reservation.
11837 (i6400_int_div): Include idiv3 type and update reservation.
11838 (i6400_int_load): Update to check type not move_type.
11839 (i6400_int_store): Likewise.
11840 (i6400_int_prefetch): Set zero latency.
11842 2018-06-12 Eric Botcazou <ebotcazou@adacore.com>
11844 * gcc.c: Document new %@{...} sequence.
11845 (LINK_COMMAND_SPEC): Use it for the -L switches.
11846 (cpp_unique_options): Use it for the -I switches.
11847 (at_file_argbuf): New global variable.
11848 (in_at_file): Likewise.
11849 (alloc_args): Create at_file_argbuf.
11850 (clear_args): Truncate at_file_argbuf.
11851 (store_arg): If in_at_file, push the argument onto at_file_argbuf.
11852 (open_at_file): New function.
11853 (close_at_file): Likewise.
11854 (create_at_file): Delete.
11855 (do_spec_1) <'i'>: Use open_at_file/close_at_file.
11858 (validate_switches_from_spec): Deal with %@{...} sequence.
11859 (validate_switches): Likewise.
11860 (driver::finalize): Call clear_args.
11862 2018-06-11 Rasmus Villemoes <rasmus.villemoes@prevas.dk>
11864 * config/vx-common.h (USE_TM_CLONE_REGISTRY): #define to 0.
11866 2018-06-11 Martin Sebor <msebor@redhat.com>
11868 * doc/invoke.texi (-Wall): List -Wc++17-compat.
11869 (Wno-class-memaccess): Add @opindex.
11870 (Wno-templates, Wno-multiple-inheritance): Same.
11871 (Wno-virtual-inheritance, Wno-namespaces): Same.
11872 (Wno-pedantic, Wno-chkp, Wcoverage-mismatch): Same.
11873 (Wno-format-overflow, Wno-format-truncation): Same.
11874 (Wno-unused-local-typedefs, Walloc-size-larger-than): Same.
11875 (Wno-alloc-size-larger-than, Wframe-larger-than): Same
11876 (Wno-stack-usage, Wno-c++-compat, Wno-c++11-compat): Same.
11877 (Wno-c++14-compat, Wno-c++17-compat, Wno-openmp-simd): Same.
11878 (Wno-unsuffixed-float-constants, Wno-addr-space-convert): Same.
11879 (Wno-misspelled-isr): Same.
11881 2018-06-11 Martin Sebor <msebor@redhat.com>
11883 * PR tree-optimization/86083
11884 * tree-ssa-strlen.c (handle_char_store): Use tree_expr_nonzero_p.
11886 2018-06-11 Zhouyi Zhou <zhouzhouyi@gmail.com>
11888 * tree-eh.c (lower_eh_constructs_2): Add a comma to comment.
11890 2018-06-11 Segher Boessenkool <segher@kernel.crashing.org>
11893 * config/rs6000/rs6000.md (*movdi_internal32): Put constraint modifiers
11894 on the correct operand.
11895 (*movdi_internal64): Ditto.
11897 2018-06-11 Martin Liska <mliska@suse.cz>
11899 PR tree-optimization/86089
11900 * tree-ssa-strlen.c (get_string_length): Move back removed hunk.
11902 2018-06-11 Julia Koval <julia.koval@intel.com>
11904 * config/i386/vaesintrin.h (_mm_aesdec_epi128, _mm_aesdeclast_epi128)
11905 _mm_aesenc_epi128, _mm_aesenclast_epi128: Remove.
11906 * config/i386/vpclmulqdqintrin.h (_mm_clmulepi64_epi128): Remove.
11908 2018-06-11 Olivier Hainque <hainque@adacore.com>
11910 * dwarf2out.c (gen_compile_unit_die): Fallout to DW_LANG_Ada83
11911 for Ada with strict dwarf2.
11913 2018-06-08 Peter Bergner <bergner@vnet.ibm.com>
11916 * config/rs6000/rs6000.c (mem_operand_gpr): Enable PRE_INC and PRE_DEC
11919 2018-06-08 Jan Hubicka <hubicka@ucw.cz>
11921 * dumpfile.c (FIRST_ME_AUTO_NUMBERED_DUMP): Bump to 4.
11923 2018-06-08 David Edelsohn <dje.gcc@gmail.com>
11925 * config/rs6000/rs6000.c (rs6000_passes_ieee128): Protect with #if
11928 2018-06-08 Martin Liska <mliska@suse.cz>
11930 * tree-cfg.h (debug_function): Fix argument type to match
11933 2018-06-08 Martin Liska <mliska@suse.cz>
11935 * config/powerpcspe/powerpcspe.c (rs6000_xcoff_visibility):
11936 Remove usage of MPX-related (and removed) fields.
11937 * config/rs6000/rs6000.c (rs6000_xcoff_visibility): Likewise.
11939 2018-06-08 David Malcolm <dmalcolm@redhat.com>
11941 * cfg.c (debug): Use TDF_NONE rather than 0.
11942 * cfghooks.c (debug): Likewise.
11943 * dumpfile.c (DUMP_FILE_INFO): Likewise; also for OPTGROUP.
11944 (struct dump_option_value_info): Convert to...
11945 (struct kv_pair): ...this template type.
11946 (dump_options): Convert to kv_pair<dump_flags_t>; use TDF_NONE
11948 (optinfo_verbosity_options): Likewise.
11949 (optgroup_options): Convert to kv_pair<optgroup_flags_t>; use
11951 (gcc::dump_manager::dump_register): Use optgroup_flags_t rather
11952 than int for "optgroup_flags" param.
11953 (dump_generic_expr_loc): Use dump_flags_t rather than int for
11955 (dump_dec): Likewise.
11956 (dump_finish): Use TDF_NONE rather than 0.
11957 (gcc::dump_manager::opt_info_enable_passes): Use optgroup_flags_t
11958 rather than int for "optgroup_flags" param. Use TDF_NONE rather
11959 than 0. Update for change to option_ptr.
11960 (opt_info_switch_p_1): Convert "optgroup_flags" param from int *
11961 to optgroup_flags_t *. Use TDF_NONE and OPTGROUP_NONE rather than
11962 0. Update for changes to optinfo_verbosity_options and
11964 (opt_info_switch_p): Convert optgroup_flags from int to
11966 (dump_basic_block): Use dump_flags_t rather than int
11967 for "dump_kind" param.
11968 * dumpfile.h (TDF_ADDRESS, TDF_SLIM, TDF_RAW, TDF_DETAILS,
11969 TDF_STATS, TDF_BLOCKS, TDF_VOPS, TDF_LINENO, TDF_UID)
11970 TDF_STMTADDR, TDF_GRAPH, TDF_MEMSYMS, TDF_RHS_ONLY, TDF_ASMNAME,
11971 TDF_EH, TDF_NOUID, TDF_ALIAS, TDF_ENUMERATE_LOCALS, TDF_CSELIB,
11972 TDF_SCEV, TDF_GIMPLE, TDF_FOLDING, MSG_OPTIMIZED_LOCATIONS,
11973 MSG_MISSED_OPTIMIZATION, MSG_NOTE, MSG_ALL, TDF_COMPARE_DEBUG,
11974 TDF_NONE): Convert from macros to...
11975 (enum dump_flag): ...this new enum.
11976 (dump_flags_t): Update to use enum.
11977 (operator|, operator&, operator~, operator|=, operator&=):
11978 Implement for dump_flags_t.
11979 (OPTGROUP_NONE, OPTGROUP_IPA, OPTGROUP_LOOP, OPTGROUP_INLINE,
11980 OPTGROUP_OMP, OPTGROUP_VEC, OPTGROUP_OTHER, OPTGROUP_ALL):
11981 Convert from macros to...
11982 (enum optgroup_flag): ...this new enum.
11983 (optgroup_flags_t): New typedef.
11984 (operator|, operator|=): Implement for optgroup_flags_t.
11985 (struct dump_file_info): Convert field "alt_flags" to
11986 dump_flags_t. Convert field "optgroup_flags" to
11988 (dump_basic_block): Use dump_flags_t rather than int for param.
11989 (dump_generic_expr_loc): Likewise.
11990 (dump_dec): Likewise.
11991 (dump_register): Convert param "optgroup_flags" to
11993 (opt_info_enable_passes): Likewise.
11994 * early-remat.c (early_remat::dump_edge_list): Use TDF_NONE rather
11996 * gimple-pretty-print.c (debug): Likewise.
11997 * gimple-ssa-store-merging.c (bswap_replace): Likewise.
11998 (merged_store_group::apply_stores): Likewise.
11999 * gimple-ssa-strength-reduction.c (insert_initializers): Likewise.
12000 * gimple.c (verify_gimple_pp): Likewise.
12001 * graphite-poly.c (print_pbb_body): Likewise.
12002 * passes.c (pass_manager::register_one_dump_file): Convert
12003 local "optgroup_flags" to optgroup_flags_t.
12004 * print-tree.c (print_node): Use TDF_NONE rather than 0.
12006 (debug_body): Likewise.
12007 * tree-pass.h (struct pass_data): Convert field "optgroup_flags"
12008 to optgroup_flags_t.
12009 * tree-pretty-print.c (print_struct_decl): Use TDF_NONE rather
12011 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Likewise.
12012 (convert_mult_to_fma): Likewise.
12013 * tree-ssa-reassoc.c (undistribute_ops_list): Likewise.
12014 * tree-ssa-sccvn.c (vn_eliminate): Likewise.
12015 * tree-vect-data-refs.c (dump_lower_bound): Convert param
12016 "dump_kind" to dump_flags_t.
12018 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
12020 * config/rs6000/rs6000.c (min, max): Delete.
12022 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
12024 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete mention of
12025 -mabi=spe and -mabi=no-spe.
12027 2018-06-08 Martin Liska <mliska@suse.cz>
12029 * ipa-pure-const.c (propagate_pure_const): Use ::get at places
12030 where we expect an existing summary.
12032 2018-06-08 Martin Liska <mliska@suse.cz>
12034 * ipa-inline-analysis.c (simple_edge_hints): Use ::get method.
12035 * ipa-inline.h (estimate_edge_growth): Likewise.
12037 2018-06-08 Martin Liska <mliska@suse.cz>
12039 * cgraph.c (function_version_hasher::hash): Use
12040 cgraph_node::get_uid ().
12041 (function_version_hasher::equal):
12042 * cgraph.h (cgraph_node::get_uid): New method.
12043 * ipa-inline.c (update_caller_keys): Use
12044 cgraph_node::get_uid ().
12045 (update_callee_keys): Likewise.
12046 * ipa-utils.c (searchc): Likewise.
12047 (ipa_reduced_postorder): Likewise.
12048 * lto-cgraph.c (input_node): Likewise.
12049 * passes.c (is_pass_explicitly_enabled_or_disabled): Likewise.
12050 * symbol-summary.h (symtab_insertion): Likewise.
12051 (symtab_removal): Likewise.
12052 (symtab_duplication): Likewise.
12053 * tree-pretty-print.c (dump_function_header): Likewise.
12054 * tree-sra.c (convert_callers_for_node): Likewise.
12056 2018-06-08 Martin Liska <mliska@suse.cz>
12058 * cgraph.c (symbol_table::create_edge): Always assign a new
12060 (symbol_table::free_edge): Do not recycle numbers.
12061 * cgraph.h (cgraph_edge::get): New method.
12062 * symbol-summary.h (symtab_removal): Use it.
12063 (symtab_duplication): Likewise.
12064 (call_summary::hashable_uid): Remove.
12066 2018-06-08 Martin Liska <mliska@suse.cz>
12068 * ipa-inline-analysis.c (inline_edge_removal_hook): Remove.
12069 (initialize_growth_caches): Remove.
12070 (free_growth_caches): Likewise.
12071 (do_estimate_edge_time): Use edge_growth_cache.
12072 (do_estimate_edge_size): Likewise.
12073 (do_estimate_edge_hints): Likewise.
12074 * ipa-inline.c (reset_edge_caches): Likewise.
12075 (recursive_inlining): Likewise.
12076 (inline_small_functions): Likewise.
12077 * ipa-inline.h (initialize_growth_caches): Remove.
12078 (estimate_edge_size): Likewise.
12079 (estimate_edge_time): Likewise.
12080 (estimate_edge_hints): Likewise.
12081 (reset_edge_growth_cache): Likewise.
12082 * symbol-summary.h (call_summary::remove): New method.
12084 2018-06-08 Martin Liska <mliska@suse.cz>
12086 * ipa-cp.c (class edge_clone_summary): New summary.
12087 (grow_edge_clone_vectors): Remove.
12088 (ipcp_edge_duplication_hook): Remove.
12089 (class edge_clone_summary_t): New call_summary class.
12090 (ipcp_edge_removal_hook): Remove.
12091 (edge_clone_summary_t::duplicate): New function.
12092 (get_next_cgraph_edge_clone): Use edge_clone_summaries.
12093 (create_specialized_node): Likewise.
12094 (ipcp_driver): Initialize edge_clone_summaries and do not
12097 2018-06-08 Martin Liska <mliska@suse.cz>
12099 * symbol-summary.h (get): New function.
12100 (call_summary::m_initialize_when_cloning): New class member.
12102 2018-06-08 Martin Liska <mliska@suse.cz>
12104 * cgraph.c (cgraph_node::remove): Do not recycle uid.
12105 * cgraph.h (symbol_table::release_symbol): Do not pass uid.
12106 (symbol_table::allocate_cgraph_symbol): Do not set uid.
12107 * passes.c (uid_hash_t): Record removed_nodes by their uids.
12108 (remove_cgraph_node_from_order): Use the removed_nodes set.
12109 (do_per_function_toporder): Likwise.
12110 * symbol-summary.h (symtab_insertion): Use cgraph_node::uid
12111 instead of summary_uid.
12112 (symtab_removal): Likewise.
12113 (symtab_duplication): Likewise.
12115 2018-06-08 Martin Liska <mliska@suse.cz>
12117 * ipa-cp.c (ipcp_store_bits_results): Use
12118 ipcp_transformation_sum.
12119 (ipcp_store_vr_results): Likewise.
12120 * ipa-prop.c (ipcp_grow_transformations_if_necessary): Renamed
12122 (ipcp_transformation_initialize): ... this.
12123 (ipa_set_node_agg_value_chain):
12124 (ipa_node_params_t::duplicate): Use ipcp_transformation_sum.
12125 (write_ipcp_transformation_info): Likewise.
12126 (read_ipcp_transformation_info): Likewise.
12127 (ipcp_update_bits): Likewise.
12128 (ipcp_update_vr): Likewise.
12129 (ipcp_transform_function): Likewise.
12130 * ipa-prop.h: Rename ipcp_transformation_summary to
12131 ipcp_transformation.
12132 (class ipcp_transformation_t): New function summary.
12133 (ipcp_get_transformation_summary): Use ipcp_transformation_sum.
12134 (ipa_get_agg_replacements_for_node): Likewise.
12136 2018-06-08 Martin Liska <mliska@suse.cz>
12138 * ipa-pure-const.c (struct funct_state_d): Do it class instead
12140 (class funct_state_summary_t): New function_summary class.
12141 (has_function_state): Remove.
12142 (get_function_state): Likewise.
12143 (set_function_state): Likewise.
12144 (add_new_function): Likewise.
12145 (funct_state_summary_t::insert): New function.
12146 (duplicate_node_data): Remove.
12147 (remove_node_data): Remove.
12148 (funct_state_summary_t::duplicate): New function.
12149 (register_hooks): Create new funct_state_summaries.
12150 (pure_const_generate_summary): Use it.
12151 (pure_const_write_summary): Likewise.
12152 (pure_const_read_summary): Likewise.
12153 (propagate_pure_const): Likewise.
12154 (propagate_nothrow): Likewise.
12155 (dump_malloc_lattice): Likewise.
12156 (propagate_malloc): Likewise.
12157 (execute): Do not register hooks, just remove summary
12159 (pass_ipa_pure_const::pass_ipa_pure_const): Simplify
12162 2018-06-08 Martin Liska <mliska@suse.cz>
12164 * ipa-reference.c (remove_node_data): Remove.
12165 (duplicate_node_data): Likewise.
12166 (class ipa_ref_var_info_summary_t): New class.
12167 (class ipa_ref_opt_summary_t): Likewise.
12168 (get_reference_vars_info): Use ipa_ref_var_info_summaries.
12169 (get_reference_optimization_summary): Use
12170 ipa_ref_opt_sum_summaries.
12171 (set_reference_vars_info): Remove.
12172 (set_reference_optimization_summary): Likewise.
12173 (ipa_init): Create summaries.
12174 (init_function_info): Use function summary.
12175 (ipa_ref_opt_summary_t::duplicate): New function.
12176 (ipa_ref_opt_summary_t::remove): New function.
12177 (get_read_write_all_from_node): Fix GNU coding style.
12178 (propagate): Use function summary.
12179 (write_node_summary_p): Fix GNU coding style.
12180 (stream_out_bitmap): Likewise.
12181 (ipa_reference_read_optimization_summary): Use function summary.
12182 (ipa_reference_c_finalize): Do not release hooks.
12184 2018-06-08 Martin Liska <mliska@suse.cz>
12186 * ipa-fnsummary.c (dump_ipa_call_summary): Use ::get method.
12187 (analyze_function_body): Extract multiple calls of get_create.
12188 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
12189 * ipa-inline.c (recursive_inlining): Use ::get method.
12190 * ipa-inline.h (estimate_edge_growth): Likewise.
12192 2018-06-08 Martin Liska <mliska@suse.cz>
12194 * hsa-common.h (enum hsa_function_kind): Rename HSA_NONE to
12196 (hsa_function_summary::hsa_function_summary): Use the new enum
12198 (hsa_gpu_implementation_p): Use hsa_summaries::get.
12199 * hsa-gen.c (hsa_get_host_function): Likewise.
12200 (get_brig_function_name): Likewise.
12201 * ipa-hsa.c (process_hsa_functions): Likewise.
12202 (ipa_hsa_write_summary): Likewise.
12203 * symbol-summary.h (symtab_duplication): Use ::get function/
12204 (get): New function.
12206 2018-06-08 Martin Liska <mliska@suse.cz>
12208 * config/i386/i386.c (ix86_can_inline_p): Use get_create instead
12210 * hsa-common.c (hsa_summary_t::link_functions): Likewise.
12211 (hsa_register_kernel): Likewise.
12212 * hsa-common.h (hsa_gpu_implementation_p): Likewise.
12213 * hsa-gen.c (hsa_get_host_function): Likewise.
12214 (get_brig_function_name): Likewise.
12215 (generate_hsa): Likewise.
12216 (pass_gen_hsail::execute): Likewise.
12217 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
12218 (devirtualization_time_bonus): Likewise.
12219 (ipcp_propagate_stage): Likewise.
12220 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
12221 (edge_set_predicate): Likewise.
12222 (evaluate_conditions_for_known_args): Likewise.
12223 (evaluate_properties_for_edge): Likewise.
12224 (ipa_fn_summary::reset): Likewise.
12225 (ipa_fn_summary_t::duplicate): Likewise.
12226 (dump_ipa_call_summary): Likewise.
12227 (ipa_dump_fn_summary): Likewise.
12228 (analyze_function_body): Likewise.
12229 (compute_fn_summary): Likewise.
12230 (estimate_edge_devirt_benefit): Likewise.
12231 (estimate_edge_size_and_time): Likewise.
12232 (estimate_calls_size_and_time): Likewise.
12233 (estimate_node_size_and_time): Likewise.
12234 (inline_update_callee_summaries): Likewise.
12235 (remap_edge_change_prob): Likewise.
12236 (remap_edge_summaries): Likewise.
12237 (ipa_merge_fn_summary_after_inlining): Likewise.
12238 (ipa_update_overall_fn_summary): Likewise.
12239 (read_ipa_call_summary): Likewise.
12240 (inline_read_section): Likewise.
12241 (write_ipa_call_summary): Likewise.
12242 (ipa_fn_summary_write): Likewise.
12243 (ipa_free_fn_summary): Likewise.
12244 * ipa-hsa.c (process_hsa_functions): Likewise.
12245 (ipa_hsa_write_summary): Likewise.
12246 (ipa_hsa_read_section): Likewise.
12247 * ipa-icf.c (sem_function::merge): Likewise.
12248 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
12249 (do_estimate_edge_time): Likewise.
12250 (estimate_size_after_inlining): Likewise.
12251 (estimate_growth): Likewise.
12252 (growth_likely_positive): Likewise.
12253 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
12254 (inline_call): Likewise.
12255 * ipa-inline.c (caller_growth_limits): Likewise.
12256 (can_inline_edge_p): Likewise.
12257 (can_inline_edge_by_limits_p): Likewise.
12258 (compute_uninlined_call_time): Likewise.
12259 (compute_inlined_call_time): Likewise.
12260 (want_inline_small_function_p): Likewise.
12261 (edge_badness): Likewise.
12262 (update_caller_keys): Likewise.
12263 (update_callee_keys): Likewise.
12264 (recursive_inlining): Likewise.
12265 (inline_small_functions): Likewise.
12266 (inline_to_all_callers_1): Likewise.
12267 (dump_overall_stats): Likewise.
12268 (early_inline_small_functions): Likewise.
12269 (early_inliner): Likewise.
12270 * ipa-inline.h (estimate_edge_growth): Likewise.
12271 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
12272 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
12273 * ipa-prop.h (IPA_NODE_REF): Likewise.
12274 (IPA_EDGE_REF): Likewise.
12275 * ipa-pure-const.c (malloc_candidate_p): Likewise.
12276 (propagate_malloc): Likewise.
12277 * ipa-split.c (execute_split_functions): Likewise.
12278 * symbol-summary.h: Rename get to get_create.
12280 (get_create): Likewise.
12281 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
12283 2018-06-08 Martin Liska <mliska@suse.cz>
12285 * symbol-summary.h (release): Move definition out of class
12287 (symtab_removal): Likewise.
12288 (symtab_duplication): Likewise.
12290 2018-06-08 Martin Liska <mliska@suse.cz>
12292 * symbol-summary.h (function_summary): Move constructor
12293 implementation out of class declaration.
12294 (release): Likewise.
12295 (symtab_insertion): Likewise.
12296 (symtab_removal): Likewise.
12297 (symtab_duplication): Likewise.
12300 2018-06-08 Martin Liska <mliska@suse.cz>
12302 * Makefile.in: Remove support for MPX (macros, related functions,
12303 fields in cgraph_node, ...).
12304 * builtin-types.def (BT_BND): Likewise.
12305 (BT_FN_BND_CONST_PTR): Likewise.
12306 (BT_FN_CONST_PTR_BND): Likewise.
12307 (BT_FN_VOID_PTR_BND): Likewise.
12308 (BT_FN_BND_CONST_PTR_SIZE): Likewise.
12309 (BT_FN_VOID_CONST_PTR_BND_CONST_PTR): Likewise.
12310 * builtins.c (expand_builtin_memcpy_with_bounds): Likewise.
12311 (expand_builtin_mempcpy_with_bounds): Likewise.
12312 (expand_builtin_memset_with_bounds): Likewise.
12313 (expand_builtin_memset_args): Likewise.
12314 (std_expand_builtin_va_start): Likewise.
12315 (expand_builtin): Likewise.
12316 (expand_builtin_with_bounds): Likewise.
12317 * builtins.def (DEF_BUILTIN_CHKP): Likewise.
12318 (DEF_LIB_BUILTIN_CHKP): Likewise.
12319 (DEF_EXT_LIB_BUILTIN_CHKP): Likewise.
12320 (DEF_CHKP_BUILTIN): Likewise.
12321 (BUILT_IN_MEMCPY): Likewise.
12322 (BUILT_IN_MEMMOVE): Likewise.
12323 (BUILT_IN_MEMPCPY): Likewise.
12324 (BUILT_IN_MEMSET): Likewise.
12325 (BUILT_IN_STPCPY): Likewise.
12326 (BUILT_IN_STRCAT): Likewise.
12327 (BUILT_IN_STRCHR): Likewise.
12328 (BUILT_IN_STRCPY): Likewise.
12329 (BUILT_IN_STRLEN): Likewise.
12330 (BUILT_IN_MEMCPY_CHK): Likewise.
12331 (BUILT_IN_MEMMOVE_CHK): Likewise.
12332 (BUILT_IN_MEMPCPY_CHK): Likewise.
12333 (BUILT_IN_MEMSET_CHK): Likewise.
12334 (BUILT_IN_STPCPY_CHK): Likewise.
12335 (BUILT_IN_STRCAT_CHK): Likewise.
12336 (BUILT_IN_STRCPY_CHK): Likewise.
12337 * calls.c (store_bounds): Likewise.
12338 (emit_call_1): Likewise.
12339 (special_function_p): Likewise.
12340 (maybe_warn_nonstring_arg): Likewise.
12341 (initialize_argument_information): Likewise.
12342 (finalize_must_preallocate): Likewise.
12343 (compute_argument_addresses): Likewise.
12344 (expand_call): Likewise.
12345 * cfgexpand.c (expand_call_stmt): Likewise.
12346 (expand_return): Likewise.
12347 (expand_gimple_stmt_1): Likewise.
12348 (pass_expand::execute): Likewise.
12349 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
12350 (cgraph_node::remove): Likewise.
12351 (cgraph_node::dump): Likewise.
12352 (cgraph_node::verify_node): Likewise.
12353 * cgraph.h (chkp_function_instrumented_p): Likewise.
12354 (symtab_node::get_alias_target): Likewise.
12355 (cgraph_node::can_remove_if_no_direct_calls_and_refs_p): Likewise.
12356 (cgraph_local_p): Likewise.
12357 * cgraphbuild.c (cgraph_edge::rebuild_edges): Likewise.
12358 (cgraph_edge::rebuild_references): Likewise.
12359 * cgraphunit.c (varpool_node::finalize_decl): Likewise.
12360 (walk_polymorphic_call_targets): Likewise.
12361 (cgraph_node::expand_thunk): Likewise.
12362 (symbol_table::output_weakrefs): Likewise.
12363 * common/config/i386/i386-common.c (OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET): Likewise.
12364 (ix86_handle_option): Likewise.
12365 * config/i386/constraints.md: Likewise.
12366 * config/i386/i386-builtin-types.def (BND): Likewise.
12370 * config/i386/i386-builtin.def (BDESC_END): Likewise.
12371 (BDESC_FIRST): Likewise.
12373 * config/i386/i386-c.c (ix86_target_macros_internal): Likewise.
12374 * config/i386/i386-protos.h (ix86_bnd_prefixed_insn_p): Likewise.
12375 * config/i386/i386.c (enum reg_class): Likewise.
12376 (ix86_target_string): Likewise.
12377 (ix86_option_override_internal): Likewise.
12378 (ix86_conditional_register_usage): Likewise.
12379 (ix86_valid_target_attribute_inner_p): Likewise.
12380 (ix86_set_indirect_branch_type): Likewise.
12381 (ix86_set_current_function): Likewise.
12382 (ix86_function_arg_regno_p): Likewise.
12383 (init_cumulative_args): Likewise.
12384 (ix86_function_arg_advance): Likewise.
12385 (ix86_function_arg): Likewise.
12386 (ix86_pass_by_reference): Likewise.
12387 (ix86_function_value_regno_p): Likewise.
12388 (ix86_function_value_1): Likewise.
12389 (ix86_function_value_bounds): Likewise.
12390 (ix86_return_in_memory): Likewise.
12391 (ix86_setup_incoming_vararg_bounds): Likewise.
12392 (ix86_va_start): Likewise.
12393 (indirect_thunk_need_prefix): Likewise.
12394 (print_reg): Likewise.
12395 (ix86_print_operand): Likewise.
12396 (ix86_expand_call): Likewise.
12397 (ix86_output_function_return): Likewise.
12398 (reg_encoded_number): Likewise.
12399 (BDESC_VERIFYS): Likewise.
12400 (ix86_init_mpx_builtins): Likewise.
12401 (ix86_init_builtins): Likewise.
12402 (ix86_emit_cmove): Likewise.
12403 (ix86_emit_move_max): Likewise.
12404 (ix86_expand_builtin): Likewise.
12405 (ix86_builtin_mpx_function): Likewise.
12406 (ix86_get_arg_address_for_bt): Likewise.
12407 (ix86_load_bounds): Likewise.
12408 (ix86_store_bounds): Likewise.
12409 (ix86_load_returned_bounds): Likewise.
12410 (ix86_store_returned_bounds): Likewise.
12411 (ix86_class_likely_spilled_p): Likewise.
12412 (ix86_hard_regno_mode_ok): Likewise.
12413 (x86_order_regs_for_local_alloc): Likewise.
12414 (ix86_mitigate_rop): Likewise.
12415 (ix86_bnd_prefixed_insn_p): Likewise.
12416 (ix86_mpx_bound_mode): Likewise.
12417 (ix86_make_bounds_constant): Likewise.
12418 (ix86_initialize_bounds): Likewise.
12419 (TARGET_LOAD_BOUNDS_FOR_ARG): Likewise.
12420 (TARGET_STORE_BOUNDS_FOR_ARG): Likewise.
12421 (TARGET_LOAD_RETURNED_BOUNDS): Likewise.
12422 (TARGET_STORE_RETURNED_BOUNDS): Likewise.
12423 (TARGET_CHKP_BOUND_MODE): Likewise.
12424 (TARGET_BUILTIN_CHKP_FUNCTION): Likewise.
12425 (TARGET_CHKP_FUNCTION_VALUE_BOUNDS): Likewise.
12426 (TARGET_CHKP_MAKE_BOUNDS_CONSTANT): Likewise.
12427 (TARGET_CHKP_INITIALIZE_BOUNDS): Likewise.
12428 * config/i386/i386.h (TARGET_MPX): Likewise.
12429 (TARGET_MPX_P): Likewise.
12430 (VALID_BND_REG_MODE): Likewise.
12431 (FIRST_BND_REG): Likewise.
12432 (LAST_BND_REG): Likewise.
12433 (enum reg_class): Likewise.
12434 (BND_REG_P): Likewise.
12435 (BND_REGNO_P): Likewise.
12436 (BNDmode): Likewise.
12437 (ADJUST_INSN_LENGTH): Likewise.
12438 * config/i386/i386.md: Likewise.
12439 * config/i386/i386.opt: Likewise.
12440 * config/i386/linux-common.h (LIBMPX_LIBS): Likewise.
12441 (defined): Likewise.
12442 (LINK_MPX): Likewise.
12443 (MPX_SPEC): Likewise.
12444 (LIBMPX_SPEC): Likewise.
12445 (LIBMPXWRAPPERS_SPEC): Likewise.
12446 (CHKP_SPEC): Likewise.
12447 * config/i386/predicates.md: Likewise.
12448 * dbxout.c (dbxout_type): Likewise.
12449 * doc/extend.texi: Likewise.
12450 * doc/invoke.texi: Likewise.
12451 * doc/md.texi: Likewise.
12452 * doc/tm.texi: Likewise.
12453 * doc/tm.texi.in: Likewise.
12454 * dwarf2out.c (is_base_type): Likewise.
12455 (gen_formal_types_die): Likewise.
12456 (gen_subprogram_die): Likewise.
12457 (gen_type_die_with_usage): Likewise.
12458 (gen_decl_die): Likewise.
12459 (dwarf2out_late_global_decl): Likewise.
12460 * expr.c (expand_assignment): Likewise.
12461 (emit_storent_insn): Likewise.
12462 (store_expr_with_bounds): Likewise.
12463 (store_expr): Likewise.
12464 (expand_expr_real_1): Likewise.
12465 * expr.h (store_expr_with_bounds): Likewise.
12466 * function.c (use_register_for_decl): Likewise.
12467 (struct bounds_parm_data): Likewise.
12468 (assign_parms_augmented_arg_list): Likewise.
12469 (assign_parm_find_entry_rtl): Likewise.
12470 (assign_parm_is_stack_parm): Likewise.
12471 (assign_parm_load_bounds): Likewise.
12472 (assign_bounds): Likewise.
12473 (assign_parms): Likewise.
12474 (expand_function_start): Likewise.
12475 * gcc.c (CHKP_SPEC): Likewise.
12476 * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
12477 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Likewise.
12478 (wrestrict_dom_walker::check_call): Likewise.
12479 * gimple.c (gimple_build_call_from_tree): Likewise.
12480 * gimple.h (enum gf_mask): Likewise.
12481 (gimple_call_with_bounds_p): Likewise.
12482 (gimple_call_set_with_bounds): Likewise.
12483 * gimplify.c (gimplify_init_constructor): Likewise.
12484 * ipa-cp.c (initialize_node_lattices): Likewise.
12485 (propagate_constants_across_call): Likewise.
12486 (find_more_scalar_values_for_callers_subset): Likewise.
12487 * ipa-hsa.c (process_hsa_functions): Likewise.
12488 * ipa-icf-gimple.c (func_checker::compare_gimple_call): Likewise.
12489 * ipa-icf.c (sem_function::merge): Likewise.
12490 * ipa-inline.c (early_inliner): Likewise.
12491 * ipa-pure-const.c (warn_function_noreturn): Likewise.
12492 (warn_function_cold): Likewise.
12493 (propagate_pure_const): Likewise.
12494 * ipa-ref.h (enum GTY): Likewise.
12495 * ipa-split.c (find_retbnd): Likewise.
12496 (consider_split): Likewise.
12497 (split_function): Likewise.
12498 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
12499 * ipa.c (walk_polymorphic_call_targets): Likewise.
12500 (symbol_table::remove_unreachable_nodes): Likewise.
12501 (process_references): Likewise.
12502 (cgraph_build_static_cdtor_1): Likewise.
12503 * lto-cgraph.c (lto_output_node): Likewise.
12504 (output_refs): Likewise.
12505 (compute_ltrans_boundary): Likewise.
12506 (input_overwrite_node): Likewise.
12507 (input_node): Likewise.
12508 (input_cgraph_1): Likewise.
12509 * params.def (PARAM_CHKP_MAX_CTOR_SIZE): Likewise.
12510 * passes.c (pass_manager::execute_early_local_passes): Likewise.
12511 (class pass_chkp_instrumentation_passes): Likewise.
12512 (make_pass_chkp_instrumentation_passes): Likewise.
12513 * passes.def: Likewise.
12514 * rtl.h (struct GTY): Likewise.
12515 (CALL_EXPR_WITH_BOUNDS_P): Likewise.
12516 * stor-layout.c (layout_type): Likewise.
12517 * symtab.c: Likewise.
12518 * target.def: Likewise.
12519 * targhooks.c (default_chkp_bound_type): Likewise.
12520 (default_chkp_bound_mode): Likewise.
12521 (default_builtin_chkp_function): Likewise.
12522 (default_chkp_function_value_bounds): Likewise.
12523 (default_chkp_make_bounds_constant): Likewise.
12524 (default_chkp_initialize_bounds): Likewise.
12525 * targhooks.h (default_chkp_bound_type): Likewise.
12526 (default_chkp_bound_mode): Likewise.
12527 (default_builtin_chkp_function): Likewise.
12528 (default_chkp_function_value_bounds): Likewise.
12529 (default_chkp_make_bounds_constant): Likewise.
12530 (default_chkp_initialize_bounds): Likewise.
12531 * toplev.c (compile_file): Likewise.
12532 (process_options): Likewise.
12533 * tree-core.h (DEF_BUILTIN): Likewise.
12534 (DEF_BUILTIN_CHKP): Likewise.
12535 * tree-inline.c (declare_return_variable): Likewise.
12536 (remap_gimple_stmt): Likewise.
12537 (copy_bb): Likewise.
12538 (initialize_inlined_parameters): Likewise.
12539 (expand_call_inline): Likewise.
12540 * tree-pass.h (make_pass_ipa_chkp_versioning): Likewise.
12541 (make_pass_ipa_chkp_early_produce_thunks): Likewise.
12542 (make_pass_ipa_chkp_produce_thunks): Likewise.
12543 (make_pass_chkp): Likewise.
12544 (make_pass_chkp_opt): Likewise.
12545 (make_pass_chkp_instrumentation_passes): Likewise.
12546 * tree-pretty-print.c (dump_generic_node): Likewise.
12547 * tree-ssa-ccp.c (insert_clobber_before_stack_restore): Likewise.
12548 * tree-ssa-dce.c (propagate_necessity): Likewise.
12549 (eliminate_unnecessary_stmts): Likewise.
12550 * tree-ssa-pre.c (create_expression_by_pieces): Likewise.
12551 * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
12552 * tree-ssa-sccvn.h: Likewise.
12553 * tree-ssa-strlen.c (get_string_length): Likewise.
12554 (valid_builtin_call): Likewise.
12555 (adjust_last_stmt): Likewise.
12556 (handle_builtin_strchr): Likewise.
12557 (handle_builtin_strcpy): Likewise.
12558 (handle_builtin_stxncpy): Likewise.
12559 (handle_builtin_memcpy): Likewise.
12560 (handle_builtin_strcat): Likewise.
12561 (strlen_check_and_optimize_stmt): Likewise.
12562 * tree-stdarg.c (expand_ifn_va_arg_1): Likewise.
12563 * tree-streamer-in.c: Likewise.
12564 * tree-streamer.c (record_common_node): Likewise.
12565 * tree.c (tree_code_size): Likewise.
12566 (wide_int_to_tree_1): Likewise.
12567 (type_contains_placeholder_1): Likewise.
12568 (build_common_tree_nodes): Likewise.
12569 * tree.def (POINTER_BOUNDS_TYPE): Likewise.
12570 * tree.h (POINTER_BOUNDS_TYPE_P): Likewise.
12571 (POINTER_BOUNDS_P): Likewise.
12572 (BOUNDED_TYPE_P): Likewise.
12573 (BOUNDED_P): Likewise.
12574 (CALL_WITH_BOUNDS_P): Likewise.
12575 (pointer_bounds_type_node): Likewise.
12576 * value-prof.c (gimple_ic): Likewise.
12577 * var-tracking.c (vt_add_function_parameters): Likewise.
12578 * varasm.c (make_decl_rtl): Likewise.
12579 (assemble_start_function): Likewise.
12580 (output_constant): Likewise.
12581 (maybe_assemble_visibility): Likewise.
12582 * varpool.c (ctor_for_folding): Likewise.
12583 * chkp-builtins.def: Remove.
12584 * ipa-chkp.c: Remove.
12585 * ipa-chkp.h: Remove.
12586 * rtl-chkp.c: Remove.
12587 * rtl-chkp.h: Remove.
12588 * tree-chkp-opt.c: Remove.
12589 * tree-chkp.c: Remove.
12590 * tree-chkp.h: Remove.
12592 2018-06-07 Carl Love <cel@us.ibm.com>
12594 * config/rs6000/vsx.md (vextract_fp_from_shorth,
12595 vextract_fp_from_shortl): Add BE support.
12597 2018-06-07 Paul Koning <ni1d@arrl.net>
12599 * compare-elim.c (try_merge_compare): Don't merge compare if
12600 address contains a side effect.
12601 (try_eliminate_compare): Likewise.
12603 2018-06-07 Olga Makhotina <olga.makhotina@intel.com>
12605 * config.gcc: Support "tremont".
12606 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "tremont".
12607 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
12609 * config/i386/i386.c (m_TREMONT): Define.
12610 (processor_target_table): Add "tremont".
12611 (PTA_TREMONT): Define.
12612 (ix86_lea_outperforms): Add TARGET_TREMONT.
12613 (get_builtin_code_for_version): Handle PROCESSOR_TREMONT.
12614 (fold_builtin_cpu): Add M_INTEL_TREMONT, replace M_INTEL_GOLDMONT
12615 and M_INTEL_GOLDMONT_PLUS.
12616 (fold_builtin_cpu): Add "tremont".
12617 (ix86_add_stmt_cost): Add TARGET_TREMONT.
12618 (ix86_option_override_internal): Add "tremont".
12619 * config/i386/i386.h (processor_costs): Define TARGET_TREMONT.
12620 (processor_type): Add PROCESSOR_TREMONT.
12621 * config/i386/x86-tune.def: Add m_TREMONT.
12622 * doc/invoke.texi: Add tremont as x86 -march=/-mtune= CPU type.
12624 2018-06-07 Jozef Lawrynowicz <jozef.l@mittosystems.com>
12626 * config/msp430/msp430.c (msp430_mcu_name): Set the "i" in the
12627 symbol defined for msp430i* devices to be lower case.
12629 2018-06-07 Richard Biener <rguenther@suse.de>
12631 * graphite-sese-to-poly.c (extract_affine): Avoid unneded
12632 wrapping. Properly wrap the result of a BIT_NOT_EXPR.
12633 Properly wrap signed arithmetic if overflow wraps.
12635 2018-06-07 Jakub Jelinek <jakub@redhat.com>
12637 PR tree-optimization/69615
12638 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): If rhs2 is lhs
12639 of a cast from a same precision integral SSA_NAME in a bb dominated
12640 by first_bb, retry with rhs2 set to the rhs1 of the cast. Don't emit
12641 cast to utype if rhs2 has already a compatible type.
12643 2018-06-07 Richard Biener <rguenther@suse.de>
12645 PR tree-optimization/85935
12646 * graphite-scop-detection.c (find_params_in_bb): Analyze
12647 condition operands with respect to the correct loop. Assert
12648 the analysis doesn't fail.
12650 2018-06-04 Carl Love <cel@us.ibm.com>
12652 * config/rs6000/vsx.md (first_match_index_<mode>): Calculate index
12653 using natural element order. Use gen_lshrsi3 instead of gen_ashrsi3
12654 as it is slightly cheaper.
12655 (first_match_or_eos_index_<mode>):
12656 Calculate index using natural element order.
12657 (first_match_index_<mode>):
12658 Calculate index using natural element order.
12659 (first_match_or_eos_index_<mode>):
12660 Calculate index using natural order.
12661 (define_insn vclzlsbb): Change to define_insn vclzlsbb_<mode>.
12662 for BE and LE modes.
12663 * config/rs6000/rs6000-c.c: Rename P9V_BUILTIN_VCLZLSBB,
12664 P9V_BUILTIN_VCLZLSBB_V16QI.
12665 * config/rs6000/rs6000-builtin.def: Make VCLZLSBB mode
12668 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
12670 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Adjust
12671 indentation and line wrap for many prototypes. Add missing
12672 @smallexample directives around block of prototypes for vec_xl and
12675 2018-06-05 Michael Meissner <meissner@linux.ibm.com>
12677 * config/rs6000/rs6000.c (rs6000_passes_ieee128): New boolean to
12678 track if we pass or return IEEE 128-bit floating point.
12679 (ieee128_mangling_gcc_8_1): New boolean to say whether to generate
12680 C++ mangling that is compatible with GCC 8.1.
12681 (TARGET_ASM_GLOBALIZE_DECL_NAME): Override target hook.
12682 (init_cumulative_args): Note if we pass or return IEEE 128-bit
12683 floating point types.
12684 (rs6000_function_arg_advance_1): Likewise.
12685 (rs6000_mangle_type): Optionally generate mangled names that match
12686 what GCC 8.1 generated for IEEE 128-bit floating point types.
12687 (rs6000_globalize_decl_name): If we have an external function that
12688 passes or returns IEEE 128-bit types, generate a weak reference
12689 from the mangled name used in GCC 8.1 to the current mangled
12691 (rs6000_init_builtins): Make __ibm128 use the long double type if
12692 long double is IBM extended double. Make __float128 use the long
12693 double type if long double is IEEE 128-bit.
12696 * config/rs6000/rs6000-builtin.def (BU_IBM128_2): New helper
12697 macro for __ibm128 built-in functions.
12698 (PACK_IF): Add __ibm128 pack/unpack functions.
12699 (UNPACK_IF): Likewise.
12700 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Do not
12701 enable long double built-in functions if long double is IEEE
12702 128-bit floating point.
12703 (rs6000_invalid_builtin): Update long double built-in function
12705 (rs6000_expand_builtin): For PACK_IF and UNPACK_IF built-in
12706 functions, adjust the built-in function to use the long double
12707 built-in function if __ibm128 and long double are the same type.
12708 * doc/extend.texi (PowerPC builtins): Update documention for
12709 __builtin_{,un}pack_longdouble. Add documentation for
12710 __builtin_{,un}pack_ibm128.
12712 2018-06-06 Jim Wilson <jimw@sifive.com>
12714 * config/riscv/riscv.c (enum riscv_privilege_levels): New.
12715 (struct machine_function): New field interrupt_mode.
12716 (riscv_handle_type_attribute): New function. Add forward declaration.
12717 (riscv_attribute_table) <interrupt>: Use riscv_handle_type_attribute.
12718 (riscv_expand_epilogue): Check interrupt_mode field.
12719 (riscv_set_current_function): Check interrupt attribute args and
12720 set interrupt_mode field.
12721 * config/riscv/riscv.md (UNSPECV_SRET, UNSPECV_URET): New.
12722 (riscv_sret, riscv_uret): New.
12723 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: Document
12724 new arguments to interrupt attribute.
12726 2018-06-06 Peter Bergner <bergner@vnet.ibm.com>
12729 * /config/rs6000/rs6000.h (ASM_CPU_SPEC): Add support for -mpower9.
12730 Don't handle -mcpu=power8 if -mpower9-vector is also used.
12732 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
12734 * config/rs6000/rs6000-builtin.def (VSX_BUILTIN_VEC_LD,
12735 VSX_BUILTIN_VEC_ST): Add comment to explain non-traditional uses.
12736 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
12737 several redundant entries.
12739 2018-06-06 David Malcolm <dmalcolm@redhat.com>
12741 * config/i386/i386-protos.h (ix86_expand_call): Strengthen return
12742 type from "rtx" to "rtx_insn *".
12743 * config/i386/i386.c (ix86_expand_split_stack_prologue): Likewise
12744 for local "call_insn", removing cast.
12745 (ix86_expand_call): Likewise, introducing a "call_insn" local.
12747 2018-06-06 Eric Botcazou <ebotcazou@adacore.com>
12749 PR tree-optimization/86066
12750 * gimple-ssa-store-merging.c (process_store): Do not bypass BIT_NOT_EXPR
12751 for BIT_INSERT_EXPR stores.
12753 2018-06-06 Richard Biener <rguenther@suse.de>
12755 PR tree-optimization/86062
12756 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
12757 component refs ontop
12758 of to be offsetted base.
12760 2018-06-06 Jozef Lawrynowicz <jozef.l@mittosystems.com>
12762 * config/msp430/msp430.c (msp430_attr): Allow interrupt handlers
12763 to be static and remove check on interrupt attribute name.
12765 2018-06-05 Kelvin Nilsen <kelvin@gcc.gnu.org>
12767 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
12768 volatile qualifier from vec_lvsl and vec_lvsr argument prototypes.
12770 2018-06-05 Steve Ellcey <sellcey@cavium.com>
12773 * config/aarch64/aarch64-protos.h (aarch64_err_no_fpadvsimd): Remove
12775 * config/aarch64/aarch64-protos..c (aarch64_err_no_fpadvsimd):
12776 Remove second argument, change how error is called.
12777 (aarch64_layout_arg): Remove second argument from
12778 aarch64_err_no_fpadvsimd call.
12779 (aarch64_init_cumulative_args): Ditto.
12780 (aarch64_gimplify_va_arg_expr): Ditto.
12781 * config/aarch64/aarch64.md (mov<mode>): Ditto.
12783 2018-06-05 Uros Bizjak <ubizjak@gmail.com>
12785 * config/i386/i386.md (simple_return_indirect_internal): New expander.
12786 (*simple_return_indirect_internal<mode>): Rename from
12787 simple_return_indirect_internal. Use W mode iterator.
12788 (rstorssp): New expander.
12789 (*rstorssp<mode>): Rename from rstorssp. Use P mode iterator.
12790 (clrssbsy): New expander.
12791 (*clrssbsy<mode>): Rename from clrssbsy. Use P mode iterator.
12793 2018-06-05 Andre Vieira <andre.simoesdiasvieira@arm.com>
12795 * config/arm/arm_cmse.h (cmse_nsfptr_create): Change typeof to
12797 (cmse_check_pointed_object): Likewise.
12799 2018-06-05 Martin Liska <mliska@suse.cz>
12801 PR gcov-profile/47618
12802 * doc/invoke.texi: Document how -fprofile-dir format
12805 2018-06-05 Richard Biener <rguenther@suse.de>
12807 * tree-cfgcleanup.c (cleanup_control_flow_pre): For edge
12808 removal pretend DOM info isn't available so we do not update
12809 it and only remove edges, not dominated blocks. Actually free
12810 DOM info in case we removed something. Remove unreachable blocks.
12811 (mfb_keep_latches): Work with either DOM info or marked backedges.
12812 (cleanup_tree_cfg_noloop): Do not remove unreachable blocks
12813 first. Mark backedges if DOM info isn't available.
12814 (Re-)compute DOM info after cleanup_control_flow_pre.
12816 2018-06-05 Richard Biener <rguenther@suse.de>
12818 * tree-cfg.c (struct locus_discrim_map): Store line, not location.
12819 (locus_discrim_hasher::hash): Adjust.
12820 (locus_discrim_hasher::equal): Likewise.
12821 (next_discriminator_for_locus): Work on line directly.
12822 (same_line_p): Pass in expanded locus1 as well.
12823 (assign_discriminators): Avoid redundant location expansions.
12825 2018-06-05 Richard Biener <rguenther@suse.de>
12827 PR tree-optimization/86046
12828 * tree-ssa.c (maybe_optimize_var): Clear DECL_GIMPLE_REG_P
12829 if required after clearing TREE_ADDRESSABLE.
12831 2018-06-05 Richard Biener <rguenther@suse.de>
12833 PR tree-optimization/86047
12834 * tree-ssa-loop.c (for_each_index): Glob handling of all
12835 decls and constants and really handle all of them.
12837 2018-06-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12840 * config/arm/arm-builtins.c (arm_type_qualifiers): Add
12841 qualifier_void_pointer and qualifier_const_void_pointer.
12842 (arm_ldc_qualifiers, arm_stc_qualifiers): Use the above.
12843 (arm_init_builtins): Handle the above.
12844 * config/arm/arm_acle.h (__arm_cdp, __arm_ldc, __arm_ldcl, __arm_stc,
12845 __arm_stcl, __arm_mcr, __arm_cdp2, __arm_ldc2, __arm_ldcl2, __arm_stc2,
12846 __arm_stcl2,__arm_mcr2, __arm_mcrr, __arm_mcrr2): Remove return for
12849 2018-06-05 Martin Liska <mliska@suse.cz>
12851 * auto-profile.c (read_autofdo_file): Do not use
12852 gcov_ctr_summary struct.
12853 (afdo_callsite_hot_enough_for_early_inline): Likewise.
12854 * coverage.c (struct counts_entry): Likewise.
12855 (read_counts_file): Read just single summary entry.
12856 (get_coverage_counts): Use gcov_summary struct.
12857 * coverage.h (get_coverage_counts): Likewise.
12858 * gcov-dump.c (dump_working_sets): Likewise.
12859 (tag_summary): Dump just single summary.
12860 * gcov-io.c (gcov_write_summary): Write just histogram
12862 (gcov_read_summary): Read just single summary.
12863 (compute_working_sets): Use gcov_summary struct.
12864 * gcov-io.h (GCOV_TAG_SUMMARY_LENGTH): Remove usage
12865 of GCOV_COUNTERS_SUMMABLE.
12866 (GCOV_COUNTERS_SUMMABLE): Remove.
12867 (GCOV_FIRST_VALUE_COUNTER): Replace with
12868 GCOV_COUNTER_V_INTERVAL.
12869 (struct gcov_ctr_summary): Remove.
12870 (struct gcov_summary): Directly use fields of former
12872 (compute_working_sets): Use gcov_summary struct.
12873 * gcov.c (read_count_file): Do not use ctrs fields.
12874 * lto-cgraph.c (merge_profile_summaries): Use gcov_summary
12876 * lto-streamer.h (struct GTY): Make profile_info gcov_summary
12878 * profile.c: Likewise.
12879 * profile.h: Likewise.
12881 2018-06-05 Martin Liska <mliska@suse.cz>
12883 PR gcov-profile/84846
12884 * gcov.c (output_lines): Print working directory only
12885 in intermediate format.
12887 2018-06-05 Andreas Krebbel <krebbel@linux.ibm.com>
12889 * config/s390/s390-builtin-types.def: Add void function type.
12890 * config/s390/s390-builtins.def: Use the function type for the
12893 2018-06-04 Jim Wilson <jimw@sifive.com>
12895 * config/riscv/riscv-protos.h (riscv_expand_epilogue): Change bool arg
12897 * config/riscv/riscv.c (riscv_for_each_saved_reg): New args epilogue
12898 and maybe_eh_return. Change regno to unsigned int. Use new args to
12899 handle EH_RETURN_DATA_REGNO registers properly.
12900 (riscv_expand_prologue): Pass new args to riscv_for_each_saved_reg.
12901 (riscv_expand_epilogue): Update comment. Change argument name and
12902 type. Update code to use new name and type. Pass new args to
12903 riscv_for_each_saved_reg. Only use EH_RETURN_STACKADJ_RTX when
12905 * config/riscv/riscv.md (NORMAL_RETURN): New.
12906 (SIBCALL_RETURN, EXCEPTION_RETURN): New.
12907 (epilogue, sibcall_epilogue): Update riscv_expand_epilogue arg.
12908 (eh_return): Call gen_eh_return_internal and emit barrier.
12909 (eh_return_internal): Call riscv_expand_epilogue.
12911 2018-06-04 Eric Botcazou <ebotcazou@adacore.com>
12913 * gimple-ssa-store-merging.c (struct merged_store_group): Move up
12914 bit_insertion field and declare can_be_merged_into method.
12915 (merged_store_group::can_be_merged_into): New method.
12916 (imm_store_chain_info::coalesce_immediate): Call it to decide whether
12917 consecutive non-overlapping stores can be merged. Turn MEM_REF stores
12918 into BIT_INSERT_EXPR stores if the group contains a non-MEM_REF store.
12920 2018-06-04 Richard Biener <rguenther@suse.de>
12922 PR tree-optimization/85955
12923 * builtins.c (fold_builtin_sincos): Convert pointers to
12924 destination to appropriate type before dereferencing.
12926 2018-06-04 Segher Boessenkool <segher@kernel.crashing.org>
12928 * config/rs6000/rs6000.md (abs<mode>2 for FLOAT128): Handle IFmode.
12930 2018-06-04 Richard Sandiford <richard.sandiford@linaro.org>
12932 * expr.c (expand_expr_real_1): Force the operand into memory if
12933 its TYPE_MODE is BLKmode and if there is no integer mode for
12934 the number of bits being extracted.
12936 2018-06-04 Jakub Jelinek <jakub@redhat.com>
12940 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
12941 Use vptestnm rather than vptestm in (=Yc,v,C) variant.
12943 2018-06-04 Richard Biener <rguenther@suse.de>
12945 * tree-cfgcleanup.c (cleanup_tree_cfg_1): Fold into...
12946 (cleanup_tree_cfg_noloop): ... single caller. Do
12947 start_recording_case_labels later.
12949 2018-06-04 Sebastian Peryt <sebastian.peryt@intel.com>
12951 * config/i386/cldemoteintrin.h: Change define from _X86INTRIN_H_INCLUDED
12952 to _IMMINTRIN_H_INCLUDED.
12953 * config/i386/pconfigintrin.h: Ditto.
12954 * config/i386/waitpkgintrin.h: Ditto.
12955 * config/i386/immintrin.h: Add includes for sgxintrin.h,
12956 pconfigintrin.h, waitpkgintrin.h and cldemoteintrin.h.
12957 * config/i386/x86intrin.h: Remove includes for mintrin.h, xmmintrin.h,
12958 emmintrin.h, pmmintrin.h, tmmintrin.h, smmintrin.h, wmmintrin.h,
12959 bmiintrin.h, bmi2intrin.h, lzcntintrin.h, sgxintrin.h, pconfigintrin.h,
12960 waitpkgintrin.h and cldemoteintrin.h.
12962 2018-06-04 Richard Biener <rguenther@suse.de>
12964 PR tree-optimization/86038
12965 * tracer.c (find_best_successor): Check probability for
12966 being initialized, bail out if not.
12968 2018-06-04 Richard Earnshaw <rearnsha@arm.com>
12971 * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
12972 of bits to ignore when comparing architectures.
12974 2018-06-04 Jakub Jelinek <jakub@redhat.com>
12976 PR tree-optimization/69615
12977 * fold-const.c (merge_ranges): If range1 is - [x, x] and x is the
12978 maximum or minimum of the type, try to merge it also as if
12979 range1 is + [-, x - 1] or + [x + 1, -].
12982 * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
12984 2018-06-03 Eric Botcazou <ebotcazou@adacore.com>
12986 PR tree-optimization/86034
12987 * gimple-ssa-store-merging.c (output_merged_store): Convert the RHS to
12988 the unsigned bitfield type in a bit insertion sequence if it does not
12989 have a larger precision than the bitfield size.
12990 (process_store): Also bypass widening conversions for BIT_INSERT_EXPR.
12992 2018-06-03 Kito Cheng <kito.cheng@gmail.com>
12994 * config/nds32/nds32-peephole2.md: Add new patterns for code size.
12996 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
12998 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3J.
12999 * config/nds32/nds32.c (nds32_option_override): Consider ARCH_V3J.
13000 * config/nds32/nds32.h (TARGET_ISA_V3): Add ARCH_V3J checking.
13001 * config/nds32/nds32.opt (march): Add enum value Value(ARCH_V3J).
13003 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
13005 * common/config/nds32/nds32-common.c (nds32_option_optimization_table):
13006 Disable -fdelete-null-pointer-checks for ELF toolchain.
13008 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
13009 Kito Cheng <kito.cheng@gmail.com>
13011 * config.gcc (nds32*): Use nds32-linux.opt and nds32-elf.opt.
13012 (nds32le-*-*, nds32be-*-*): Integrate checking process.
13013 (nds32*-*-*): Add glibc and uclibc conditions.
13014 * common/config/nds32/nds32-common.c (nds32_except_unwind_info): New.
13015 (TARGET_EXCEPT_UNWIND_INFO): Define.
13016 * config/nds32/elf.h: New file.
13017 * config/nds32/linux.h: New file.
13018 * config/nds32/nds32-elf.opt: New file.
13019 * config/nds32/nds32-linux.opt: New file.
13020 * config/nds32/nds32-fp-as-gp.c
13021 (pass_nds32_fp_as_gp::gate): Consider TARGET_LINUX_ABI.
13022 * config/nds32/nds32.c (nds32_conditional_register_usage): Consider
13024 (nds32_asm_file_end): Ditto.
13025 (nds32_print_operand): Ditto.
13026 (nds32_insert_attributes): Ditto.
13027 (nds32_init_libfuncs): New function.
13028 (TARGET_HAVE_TLS): Define.
13029 (TARGET_INIT_LIBFUNCS): Define.
13030 * config/nds32/nds32.h (TARGET_DEFAULT_RELAX): Apply different relax
13032 (TARGET_ELF): Apply different mcmodel setting.
13033 (LINK_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): The content has
13034 been migrated into elf.h and linux.h files.
13035 * config/nds32/nds32.md (add_pc): Consider TARGET_LINUX_ABI.
13036 * config/nds32/nds32.opt (mvh): Consider TARGET_LINUX_ABI.
13037 (mcmodel): The content has been migrated into nds32-elf.opt and
13038 nds32-linux.opt files.
13039 * config/nds32/t-elf: New file.
13040 * config/nds32/t-linux: New file.
13042 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
13043 Shiva Chen <shiva0217@gmail.com>
13045 * config/nds32/constants.md (unspec_volatile_element): Add
13046 UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END.
13047 * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp
13049 * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare.
13050 (make_pass_nds32_fp_as_gp): Declare.
13051 * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one
13053 (nds32_asm_function_end_prologue): Remove unused asm output.
13054 (nds32_asm_function_begin_epilogue): Remove unused asm output.
13055 (nds32_asm_file_start): Output necessary fp_as_gp information.
13056 (nds32_option_override): Adjust register usage.
13057 (nds32_expand_prologue): Consider fp_as_gp situation.
13058 (nds32_expand_prologue_v3push): Consider fp_as_gp situation.
13059 * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p.
13062 (simple_return): Ditto.
13063 (omit_fp_begin): Output special directive for fp_as_gp.
13064 (omit_fp_end): Output special directive for fp_as_gp.
13065 * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp,
13066 mforbid-fp-as-gp): New options.
13068 2018-06-01 Mark Wielaard <mark@klomp.org>
13070 * dwarf2out.c (dwarf2out_finish): Remove generation of
13071 DW_AT_loclists_base.
13073 2018-06-01 Eric Botcazou <ebotcazou@adacore.com>
13075 * gimple-ssa-store-merging.c: Include gimple-fold.h.
13076 (struct store_immediate_info): Document BIT_INSERT_EXPR stores.
13077 (struct merged_store_group): Add bit_insertion field.
13078 (dump_char_array): Use standard hexadecimal format.
13079 (merged_store_group::merged_store_group): Set bit_insertion to false.
13080 (merged_store_group::apply_stores): Use optimal buffer size. Deal
13081 with BIT_INSERT_EXPR stores. Move up code updating the mask and
13082 also print the mask in the dump file.
13083 (pass_store_merging::gate): Minor tweak.
13084 (imm_store_chain_info::coalesce_immediate): Fix wrong association
13085 of stores with groups in dump. Allow coalescing of BIT_INSERT_EXPR
13086 stores with INTEGER_CST stores.
13087 (count_multiple_uses) <BIT_INSERT_EXPR>: New case.
13088 (imm_store_chain_info::output_merged_store): Add try_bitpos variable
13089 and use it throughout. Generate bit insertion sequences if need be.
13090 (pass_store_merging::process_store): Remove redundant condition.
13091 Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR.
13093 2018-06-01 Segher Boessenkool <segher@kernel.crashing.org>
13095 * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
13096 the 128-bit floating point types. Fix function comment.
13098 2018-06-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13100 * config/aarch64/aarch64-simd.md
13101 (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler
13103 (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
13106 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
13108 PR tree-optimization/85989
13109 * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
13111 (backprop::intersect_uses): Check it when deciding whether this
13112 is a backedge reference.
13113 (backprop::process_block): Add each phi to m_visited_phis
13114 after visiting it, then clear it at the end.
13116 2018-06-01 Richard Biener <rguenther@suse.de>
13118 * tree-vectorizer.h (vect_dr_stmt): New function.
13119 (vect_get_load_cost): Adjust.
13120 (vect_get_store_cost): Likewise.
13121 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
13122 Use vect_dr_stmt instead of DR_SMTT.
13123 (vect_record_base_alignments): Likewise.
13124 (vect_calculate_target_alignment): Likewise.
13125 (vect_compute_data_ref_alignment): Likewise and make static.
13126 (vect_update_misalignment_for_peel): Likewise.
13127 (vect_verify_datarefs_alignment): Likewise.
13128 (vector_alignment_reachable_p): Likewise.
13129 (vect_get_data_access_cost): Likewise. Pass down
13130 vinfo to vect_get_load_cost/vect_get_store_cost instead of DR.
13131 (vect_get_peeling_costs_all_drs): Likewise.
13132 (vect_peeling_hash_get_lowest_cost): Likewise.
13133 (vect_enhance_data_refs_alignment): Likewise.
13134 (vect_find_same_alignment_drs): Likewise.
13135 (vect_analyze_data_refs_alignment): Likewise.
13136 (vect_analyze_group_access_1): Likewise.
13137 (vect_analyze_group_access): Likewise.
13138 (vect_analyze_data_ref_access): Likewise.
13139 (vect_analyze_data_ref_accesses): Likewise.
13140 (vect_vfa_segment_size): Likewise.
13141 (vect_small_gap_p): Likewise.
13142 (vectorizable_with_step_bound_p): Likewise.
13143 (vect_prune_runtime_alias_test_list): Likewise.
13144 (vect_analyze_data_refs): Likewise.
13145 (vect_supportable_dr_alignment): Likewise.
13146 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
13147 (vect_gen_prolog_loop_niters): Likewise.
13148 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
13149 * tree-vect-patterns.c (vect_recog_bool_pattern): Do not
13151 (vect_recog_mask_conversion_pattern): Likewise.
13152 (vect_try_gather_scatter_pattern): Likewise.
13153 * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info
13154 to vect_get_store_cost.
13155 (vect_get_store_cost): Get stmt_info instead of DR.
13156 (vect_model_load_cost): Pass stmt_info to vect_get_load_cost.
13157 (vect_get_load_cost): Get stmt_info instead of DR.
13159 2018-06-01 Richard Biener <rguenther@suse.de>
13161 PR middle-end/86017
13162 * gimple-fold.c (var_decl_component_p): Also allow offsetted
13163 vars wrapped in MEM_REFs.
13165 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
13167 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
13168 Fix subreg tests so that we only return a choice between
13169 GENERAL_REGS and FP_REGS if the original classes included both.
13171 2018-06-01 Richard Biener <rguenther@suse.de>
13174 * tree-ssa-structalias.c (get_function_part_constraint):
13175 Handle NULL fi->decl.
13176 (find_func_aliases_for_call): Properly handle indirect
13177 fi from direct call.
13178 (find_func_clobbers): Likewise.
13179 (ipa_pta_execute): Likewise.
13180 (create_variable_info_for): For functions that are ifunc_resolver
13181 resolve to a varinfo that contains the result of the resolver call.
13182 (associate_varinfo_to_alias): Do not treat ifunc resolvers as
13185 2018-05-31 Michael Collison <michael.collison@arm.com>
13187 * config/aarch64/aarch64.md:
13188 (*fix_to_zero_extenddfdi2): New pattern.
13189 * gcc.target/aarch64/fix_extend1.c: New testcase.
13191 2018-05-31 Qing Zhao <qing.zhao@oracle.com>
13193 PR middle-end/78809
13194 PR middle-end/83026
13195 * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
13196 and BUILT_IN_STRNCMP_EQ.
13197 * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
13198 BUILT_IN_STRNCMP_EQ.
13199 * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
13200 handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
13201 (gimple_fold_builtin): Likewise.
13202 * tree-ssa-strlen.c (compute_string_length): New function.
13203 (determine_min_obsize): New function.
13204 (handle_builtin_string_cmp): New function to handle calls to
13205 string compare functions.
13206 (strlen_optimize_stmt): Add handling to builtin string compare
13208 * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
13209 Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
13210 * tree.c (build_common_builtin_nodes): Add new defines of
13211 BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
13213 2018-05-31 Jakub Jelinek <jakub@redhat.com>
13216 * bb-reorder.c (pass_partition_blocks::gate): Return false for
13217 functions with naked attribute.
13219 2018-05-31 Uros Bizjak <ubizjak@gmail.com>
13221 * config/i386/sse.md (avx_vec_concat<mode>):
13222 Substitute concat_tg_mode mode attribute with xtg_mode.
13223 (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
13224 (concat_tg_mode): Remove mode attribute.
13226 2018-05-31 Martin Sebor <msebor@redhat.com>
13229 * calls.c (alloc_max_size): Correct a logic error/typo.
13230 Treat excessive arguments as infinite. Warn for invalid arguments.
13231 * doc/invoke.texi (-Walloc-size-larger-than): Update.
13233 2018-05-31 H.J. Lu <hongjiu.lu@intel.com>
13236 * config/i386/x86-tune.def: Re-enable partial_reg_dependency
13237 and movx for Haswell.
13239 2018-05-31 Chung-Lin Tang <cltang@codesourcery.com>
13240 Cesar Philippidis <cesar@codesourcery.com>
13242 PR middle-end/85879
13243 * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
13244 when emitting error on private/firstprivate reductions.
13245 * omp-low.c (lower_omp_target): Avoid reference-type processing
13246 on pointers for firstprivate clause.
13248 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
13250 * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
13253 * config/aarch64/aarch64-simd.md
13254 (aarch64_ld1x3<VALLDIF:mode>): New pattern.
13255 (aarch64_ld1_x3_<mode>): Likewise
13256 (aarch64_st1x2<VALLDIF:mode>): Likewise
13257 (aarch64_st1_x2_<mode>): Likewise
13258 (aarch64_st1x3<VALLDIF:mode>): Likewise
13259 (aarch64_st1_x3_<mode>): Likewise
13260 * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
13261 (vld1_s8_x3): Likewise.
13262 (vld1_u16_x3): Likewise.
13263 (vld1_s16_x3): Likewise.
13264 (vld1_u32_x3): Likewise.
13265 (vld1_s32_x3): Likewise.
13266 (vld1_u64_x3): Likewise.
13267 (vld1_s64_x3): Likewise.
13268 (vld1_f16_x3): Likewise.
13269 (vld1_f32_x3): Likewise.
13270 (vld1_f64_x3): Likewise.
13271 (vld1_p8_x3): Likewise.
13272 (vld1_p16_x3): Likewise.
13273 (vld1_p64_x3): Likewise.
13274 (vld1q_u8_x3): Likewise.
13275 (vld1q_s8_x3): Likewise.
13276 (vld1q_u16_x3): Likewise.
13277 (vld1q_s16_x3): Likewise.
13278 (vld1q_u32_x3): Likewise.
13279 (vld1q_s32_x3): Likewise.
13280 (vld1q_u64_x3): Likewise.
13281 (vld1q_s64_x3): Likewise.
13282 (vld1q_f16_x3): Likewise.
13283 (vld1q_f32_x3): Likewise.
13284 (vld1q_f64_x3): Likewise.
13285 (vld1q_p8_x3): Likewise.
13286 (vld1q_p16_x3): Likewise.
13287 (vld1q_p64_x3): Likewise.
13288 (vst1_s64_x2): Likewise.
13289 (vst1_u64_x2): Likewise.
13290 (vst1_f64_x2): Likewise.
13291 (vst1_s8_x2): Likewise.
13292 (vst1_p8_x2): Likewise.
13293 (vst1_s16_x2): Likewise.
13294 (vst1_p16_x2): Likewise.
13295 (vst1_s32_x2): Likewise.
13296 (vst1_u8_x2): Likewise.
13297 (vst1_u16_x2): Likewise.
13298 (vst1_u32_x2): Likewise.
13299 (vst1_f16_x2): Likewise.
13300 (vst1_f32_x2): Likewise.
13301 (vst1_p64_x2): Likewise.
13302 (vst1q_s8_x2): Likewise.
13303 (vst1q_p8_x2): Likewise.
13304 (vst1q_s16_x2): Likewise.
13305 (vst1q_p16_x2): Likewise.
13306 (vst1q_s32_x2): Likewise.
13307 (vst1q_s64_x2): Likewise.
13308 (vst1q_u8_x2): Likewise.
13309 (vst1q_u16_x2): Likewise.
13310 (vst1q_u32_x2): Likewise.
13311 (vst1q_u64_x2): Likewise.
13312 (vst1q_f16_x2): Likewise.
13313 (vst1q_f32_x2): Likewise.
13314 (vst1q_f64_x2): Likewise.
13315 (vst1q_p64_x2): Likewise.
13316 (vst1_s64_x3): Likewise.
13317 (vst1_u64_x3): Likewise.
13318 (vst1_f64_x3): Likewise.
13319 (vst1_s8_x3): Likewise.
13320 (vst1_p8_x3): Likewise.
13321 (vst1_s16_x3): Likewise.
13322 (vst1_p16_x3): Likewise.
13323 (vst1_s32_x3): Likewise.
13324 (vst1_u8_x3): Likewise.
13325 (vst1_u16_x3): Likewise.
13326 (vst1_u32_x3): Likewise.
13327 (vst1_f16_x3): Likewise.
13328 (vst1_f32_x3): Likewise.
13329 (vst1_p64_x3): Likewise.
13330 (vst1q_s8_x3): Likewise.
13331 (vst1q_p8_x3): Likewise.
13332 (vst1q_s16_x3): Likewise.
13333 (vst1q_p16_x3): Likewise.
13334 (vst1q_s32_x3): Likewise.
13335 (vst1q_s64_x3): Likewise.
13336 (vst1q_u8_x3): Likewise.
13337 (vst1q_u16_x3): Likewise.
13338 (vst1q_u32_x3): Likewise.
13339 (vst1q_u64_x3): Likewise.
13340 (vst1q_f16_x3): Likewise.
13341 (vst1q_f32_x3): Likewise.
13342 (vst1q_f64_x3): Likewise.
13343 (vst1q_p64_x3): Likewise.
13345 2018-05-30 Jozef Lawrynowicz <jozef.l@mittosystems.com>
13347 * config/msp430/msp430.c (msp430_output_labelref): Prepend
13348 user_label_prefix to name.
13350 * tree-core.h: Update comment about the format of NAME string
13351 passed to handler in attribute_spec.
13353 * config/msp430/msp430.md: Remove erroneous subreg expression from
13354 zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
13355 zero_extend{q,h}isi2.
13357 2018-05-30 Borislav Petkov <bp@suse.de>
13359 * doc/extend.texi: Document some architecture specific
13360 constraints and sort entries.
13362 2018-05-30 Martin Sebor <msebor@redhat.com>
13364 PR middle-end/85369
13365 * builtins.c (expand_builtin_stpcpy_1): New function.
13366 (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
13367 only if the former succeeds.
13369 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
13371 * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
13374 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
13376 * doc/invoke.texi (-flinker-output): Document
13378 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
13380 * passes.c (ipa_write_summaries): Only modify statements if body
13382 * cgraphunit.c (ipa_passes): Also produce intermeidate code when
13383 incrementally linking.
13384 (ipa_passes): Likewise.
13385 * lto-cgraph.c (lto_output_node): When incrementally linking do not
13386 pass down resolution info.
13387 * common.opt (flag_incremental_link): Update info.
13388 * gcc.c (plugin specs): Turn flinker-output=* to
13389 -plugin-opt=-linker-output-known
13390 * toplev.c (compile_file): Also cut compilation when doing incremental
13392 * flag-types. (enum lto_partition_model): Add
13393 LTO_LINKER_OUTPUT_NOLTOREL.
13394 (invoke.texi): Add -flinker-output docs.
13395 * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
13396 link same way as WPA; do not stream in dead initializers.
13398 * dwarf2out.c (dwarf2out_die_ref_for_decl,
13399 darf2out_register_external_decl): Support incremental link.
13401 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
13403 * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
13405 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
13407 * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
13408 it down to simple_object_copy_lto_debug_sections.
13409 (run_gcc): Determine incremental LTO link time and configure
13410 lto1 into non-wpa mode, disable renaming of debug sections.
13412 2018-05-30 Kelvin Nilsen <kelvin@gcc.gnu.org>
13414 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
13415 descriptions of various incorrectly documented functions.
13417 2018-05-30 Andre Vieira <andre.simoesdiasvieira@arm.com>
13420 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
13421 address check not strict.
13423 2018-05-30 Richard Biener <rguenther@suse.de>
13425 PR tree-optimization/85964
13426 * tracer.c (better_p): Drop initialized count check, we only
13427 call the function with initialized counts now.
13428 (find_best_successor): Do find a best edge if one
13429 has uninitialized count.
13430 (find_best_predecessor): Likewise. Do BB frequency check only
13431 if count is initialized.
13433 2017-05-30 Jackson Woodruff <jackson.woodruff@arm.com>
13435 * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
13436 (aarch64_ldrstr_offset_compare): New.
13437 (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
13438 load/store orderings.
13439 (aarch64_gen_adjusted_ldpstp): Likewise.
13441 2018-05-30 Wilco Dijkstra <wdijkstr@arm.com>
13443 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
13444 Check for subset of GENERAL_REGS and FP_REGS.
13445 * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
13448 2018-05-30 Richard Sandiford <richard.sandiford@linaro.org>
13450 * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
13451 and wi::to_poly_offset. Add the current offset and then check
13452 whether the sum fits, rather than using an unchecked addition of
13453 a checked term. Check for a shwi rather than a uhwi.
13454 * expr.c (get_bit_range): Use tree_to_poly_uint64.
13455 (store_constructor): Use poly_int_tree_p.
13456 (expand_expr_real_1): Likewise.
13457 * function.c (assign_temp): Likewise.
13458 * fold-const.c (const_binop): Use poly_int_tree_p and
13459 wi::to_poly_offset.
13460 (fold_indirect_ref_1): Likewise. Use multiple_p to attempt an exact
13462 * ipa-icf-gimple.c (func_checker::compare_operand): Use
13463 to_poly_offset for MEM offsets.
13464 * ipa-icf.c (sem_variable::equals): Likewise.
13465 * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
13466 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
13467 wi::to_poly_offset for BIT_FIELD_REF offsets.
13468 (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
13469 wi::to_poly_offset.
13470 * var-tracking.c (emit_note_insn_var_location): Use
13471 tree_to_poly_uint64.
13473 2018-05-29 Jim Wilson <jimw@sifive.com>
13475 * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
13477 2018-05-29 Uros Bizjak <ubizjak@gmail.com>
13480 * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
13481 Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
13483 (sse4_1_round<mode>2): Use nonimmediate_operand
13484 for operand 1 predicate.
13486 2018-05-29 Martin Sebor <msebor@redhat.com>
13487 Richard Biener <rguenther@suse.de>
13490 * calls.c (get_size_range): Call determine_value_range instead
13491 of get_value_range..
13492 * tree-vrp.h (determine_value_range): Declared new function.
13493 * tree-vrp.c (determine_value_range_1, determine_value_range): New.
13495 2018-05-29 Richard Biener <rguenther@suse.de>
13497 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
13498 sure to use non-pattern stmts for get_earlier_stmt arguments.
13499 * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
13500 called on pattern stmts.
13501 (get_later_stmt): Likewise.
13503 2018-05-29 Martin Liska <mliska@suse.cz>
13505 PR gcov-profile/85759
13506 * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
13509 2018-05-29 Jakub Jelinek <jakub@redhat.com>
13511 * tree-cfg.c (verify_gimple_assign_unary): Add checking for
13513 (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
13517 * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
13518 VEC_PACK_FLOAT_EXPR): New tree codes.
13519 * tree-pretty-print.c (op_code_prio): Handle
13520 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
13521 (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
13522 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
13523 * tree-inline.c (estimate_operator_cost): Likewise.
13524 * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
13525 * fold-const.c (const_binop): Likewise.
13526 (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
13527 VEC_UNPACK_FIX_TRUNC_LO_EXPR.
13528 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
13529 (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
13530 * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
13531 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
13532 * expr.c (expand_expr_real_2): Likewise.
13533 * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
13534 vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
13535 vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
13537 * optabs.c (expand_widen_pattern_expr): For
13538 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
13539 sign from result type rather than operand's type.
13540 (expand_binop_directly): For vec_packu_float_optab and
13541 vec_packs_float_optab allow result type to be different from operand's
13543 * optabs-tree.c (optab_for_tree_code): Handle
13544 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
13545 VEC_PACK_FLOAT_EXPR. Formatting fixes.
13546 * tree-vect-generic.c (expand_vector_operations_1): Handle
13547 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
13548 VEC_PACK_FLOAT_EXPR.
13549 * tree-vect-stmts.c (supportable_widening_operation): Handle
13551 (supportable_narrowing_operation): Handle FLOAT_EXPR.
13552 * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
13553 * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
13554 (float<floatunssuffix>v2div2sf2): ... this. Formatting fix.
13555 (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
13557 (vec_pack<floatprefix>_float_<mode>): New expander.
13558 (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
13560 (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
13561 vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
13562 * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
13563 vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
13564 vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
13566 * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
13567 VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
13568 (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
13569 VEC_PACK_FLOAT_EXPR): Document.
13571 2018-05-29 Richard Biener <rguenther@suse.de>
13573 * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
13575 (stmt_vec_info_vec): Make pointer.
13576 (init_stmt_vec_info_vec): Remove.
13577 (free_stmt_vec_info_vec): Likewise.
13578 (set_stmt_vec_info_vec): New function.
13579 (free_stmt_vec_infos): Likewise.
13580 (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
13581 (set_vinfo_for_stmt): Likewise.
13582 (get_earlier_stmt): Likewise.
13583 (get_later_stmt): Likewise.
13584 * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
13585 (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
13586 (vec_info::~vec_info): Free stmt_vec_infos.
13587 (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
13588 Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
13589 (pass_slp_vectorize::execute): Likewise.
13590 * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
13591 (free_stmt_vec_info_vec): Likewise.
13592 (set_stmt_vec_info_vec): New function.
13593 (free_stmt_vec_infos): Likewise.
13594 * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
13595 the global stmt_vec_info_vec.
13596 * tree-parloops.c (gather_scalar_reductions): Use
13597 set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
13600 2018-05-29 Richard Biener <rguenther@suse.de>
13602 * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
13604 2018-05-29 Martin Liska <mliska@suse.cz>
13605 David Malcolm <dmalcolm@redhat.com>
13607 * vec.c (test_reverse): New.
13608 (vec_c_tests): Add new test.
13609 * vec.h (vl_ptr>::reverse): New function.
13611 2018-05-29 Gerald Pfeifer <gerald@pfeifer.com>
13613 * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
13615 * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
13618 2018-05-28 Bernd Edlinger <bernd.edlinger@hotmail.de>
13620 * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
13622 2018-05-28 Richard Biener <rguenther@suse.de>
13624 PR tree-optimization/85933
13625 * tree-vect-data-refs.c (vect_record_base_alignments): Only
13626 look at stmts marked as vectorizable.
13628 2018-05-28 Richard Biener <rguenther@suse.de>
13630 PR tree-optimization/85934
13631 * tree-vect-generic.c (expand_vector_operations_1): Hoist
13632 vector boolean check before scalar optimization.
13634 2018-05-28 Jakub Jelinek <jakub@redhat.com>
13636 * doc/invoke.texi (ARM Options): Use @item instead of @itemx
13639 2018-05-28 Mark Wielaard <mark@klomp.org>
13641 * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
13642 if it is an expression containing a minus sign.
13644 2018-05-27 John David Anglin <danglin@gcc.gnu.org>
13646 * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
13648 2018-05-27 Paul Koning <ni1d@arrl.net>
13650 * config/pdp11/pdp11.md (truncsihi2): Remove.
13652 2018-05-27 Monk Chiang <sh.chiang04@gmail.com>
13653 Chung-Ju Wu <jasonwucj@gmail.com>
13655 * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
13657 (unaligned_store_dw): Ditto.
13658 * config/nds32/nds32-memory-manipulation.c
13659 (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
13660 (nds32_gen_dup_4_byte_to_word_value): Rename to ...
13661 (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
13662 (emit_setmem_word_loop): Rename to ...
13663 (emit_setmem_doubleword_loop): ... this.
13664 (nds32_gen_dup_4_byte_to_word_value): New function.
13665 (nds32_gen_dup_8_byte_to_double_word_value): New function.
13666 (nds32_expand_setmem_loop): Refine implementation.
13667 (nds32_expand_setmem_loop_v3m): Ditto.
13668 * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
13671 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
13673 * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
13675 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
13677 * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
13678 (nds32_init_machine_status): Initialize machine->attr_naked_p and
13679 machine->attr_no_prologue_p.
13680 (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
13681 (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
13682 (nds32_expand_epilogue): Consider attr_naked_p.
13683 (nds32_expand_epilogue_v3pop): Likewise.
13684 (nds32_can_use_return_insn): Likewise.
13685 * config/nds32/nds32.h (machine_function): Add attr_naked_p and
13686 attr_no_prologue_p fields.
13687 * config/nds32/nds32.opt (mret-in-naked-func): New option.
13689 2018-05-27 Jakub Jelinek <jakub@redhat.com>
13692 * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
13694 * config/i386/sse.md
13695 (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
13697 (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
13699 (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
13701 (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
13703 (*<floatsuffix>floatv2div2sf2): Rename to ...
13704 (*float<floatunssuffix>v2div2sf2): ... this.
13705 (<floatsuffix>floatv2div2sf2_mask): Rename to ...
13706 (float<floatunssuffix>v2div2sf2_mask): ... this.
13707 (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
13708 (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
13709 (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
13711 (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
13713 (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
13715 (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
13717 (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
13719 (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
13721 (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
13722 (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
13723 (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
13724 gen_ufix_truncv8dfv8si2.
13725 * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
13726 __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
13727 __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
13728 __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
13729 __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
13730 __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
13731 __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
13732 Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
13734 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
13738 * varasm.c (assemble_alias): Lookup ifunc attribute on error.
13740 2018-05-25 Jim Wilson <jimw@sifive.com>
13742 * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
13743 * config/riscv/riscv.c (struct machine_function): Add
13744 interrupt_handler_p and attribute_checked_p fields.
13745 (riscv_attribute_table): Add interrupt.
13746 (riscv_interrupt_type_p): New.
13747 (riscv_save_reg_p): Save extra regs for interrupt handler.
13748 (riscv_use_save_libcall): Return false for interrupt handler.
13749 (riscv_first_stack_step): Add forward declaration.
13750 (riscv_compute_frame_info): New local interrupt_save_t1. Set it
13751 for interrupt handler with large frame. Use it for saved reg list.
13752 (riscv_expand_prologue): Move flag_stack_usage_info support to
13753 eliminate duplication.
13754 (riscv_expand_epilogue): Generate mret for interrupt handler.
13755 (riscv_epilogue_uses): New.
13756 (riscv_can_use_return_insn): Return false for interrupt handler.
13757 (riscv_function_ok_for_sibcall): Likewise.
13758 (riscv_set_current_function): Add interrupt handler support.
13759 * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
13760 * config/riscv/riscv.md (UNSPECV_MRET): New.
13762 (riscv_frflags, riscv_fsflags): Use tab after opcode.
13764 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
13766 2018-05-25 Bill Schmidt <wschmidt@linux.ibm.com>
13768 PR tree-optimization/85712
13769 * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
13770 this candidate has already been replaced in-situ by a copy.
13772 2018-05-25 Jason Merrill <jason@redhat.com>
13774 PR c++/80485 - inline function non-zero address.
13775 * symtab.c (nonzero_address): Check DECL_COMDAT.
13777 2018-05-25 Uros Bizjak <ubizjak@gmail.com>
13780 * config/alpha/alpha.md (ashlsi3): New insn pattern.
13781 (*ashlsi_se): Rename from *ashldi_se. Define as sign
13782 extension of SImode operation. Use const123_operand predicate.
13783 (*saddsi_1): Remove.
13784 (*saddl_se_1): Ditto.
13785 (*ssubsi_1): Ditto.
13786 (*ssubl_se_1): Ditto.
13787 * config/alpha/predicates.md (const123_operand): New predicate.
13788 * config/alpha/constraints.md (P): Use IN_RANGE.
13790 2018-05-25 Richard Biener <rguenther@suse.de>
13792 * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
13794 (ref_maybe_used_by_stmt_p): Likewise.
13795 (stmt_may_clobber_ref_p): Likewise.
13796 (stmt_may_clobber_ref_p_1): Likewise.
13797 * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
13799 (ref_maybe_used_by_stmt_p): Likewise.
13800 (stmt_may_clobber_ref_p): Likewise.
13801 (stmt_may_clobber_ref_p_1): Likewise.
13802 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
13803 the alias oracle to disambiguate DRs with stmts DR analysis
13805 (vect_analyze_data_refs): Do not give up on not analyzable
13806 DRs for BB vectorization. Remove code truncating the dataref
13809 2018-05-25 Jakub Jelinek <jakub@redhat.com>
13812 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
13813 Add (=Yk,v,C) variant using vptestm insn. Use TARGET_AVX512BW
13814 in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
13816 2018-05-25 Richard Biener <rguenther@suse.de>
13818 * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
13819 function, combining stmt data ref gathering and fatal analysis
13821 (vect_analyze_data_refs): Remove now redudnant code and simplify.
13822 * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
13823 vect_analyze_loop_2 and use vect_find_stmt_data_reference.
13824 * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
13825 * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
13827 2018-05-25 Bin Cheng <bin.cheng@arm.com>
13829 PR tree-optimization/85720
13830 * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
13831 SCC if all partitions are builtins.
13832 (version_loop_by_alias_check): New parameter. Generate cancelable
13833 runtime alias check if all partitions are builtins.
13834 (distribute_loop): Update call to above function.
13836 2018-05-25 Bin Cheng <bin.cheng@arm.com>
13838 * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
13839 (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
13840 (parm_default_def_partition_arg): Ditto.
13841 (set_parm_default_def_partition): Ditto.
13842 (get_parm_default_def_partitions): Ditto and make it static.
13843 (get_undefined_value_partitions): Ditto and make it static.
13844 (remove_ssa_form): Refactor call to init_var_map here.
13845 * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
13846 computation for loop region.
13847 (coalesce_partitions, compute_optimized_partition_bases): Ditto.
13848 (register_default_def): Delete.
13849 (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
13850 (parm_default_def_partition_arg): Ditto.
13851 (set_parm_default_def_partition): Ditto.
13852 (get_parm_default_def_partitions): Ditto and make it static.
13853 (get_undefined_value_partitions): Ditto and make it static.
13854 (coalesce_with_default, coalesce_with_default): Update comment.
13855 (create_coalesce_list_for_region): New func factored out from
13856 create_outofssa_var_map.
13857 (populate_coalesce_list_for_outofssa): New func factored out from
13858 create_outofssa_var_map and coalesce_ssa_name.
13859 (create_outofssa_var_map): Delete.
13860 (coalesce_ssa_name): Refactor to support live range computation.
13861 * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
13862 (get_parm_default_def_partitions): Delete.
13863 (get_undefined_value_partitions): Ditto.
13864 * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
13865 computation for loop region.
13866 (new_tree_live_info, loe_visit_block): Ditto.
13867 (live_worklist, set_var_live_on_entry): Ditto.
13868 (calculate_live_on_exit, verify_live_on_entry): Ditto.
13869 * tree-ssa-live.h (struct _var_map): New fields.
13870 (init_var_map): Change decl.
13871 (region_contains_p): New.
13873 2018-05-25 Bin Cheng <bin.cheng@arm.com>
13875 * tree-ssa-live.h (live_merge_and_clear): Delete.
13877 2018-05-25 Richard Biener <rguenther@suse.de>
13880 * tree-dump.c (dequeue_and_dump): Remove access to removed
13881 operand 2 of a SWITCH_EXPR.
13883 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
13885 * doc/sourcebuild.texi (vect_double_cond_arith): Include
13886 multiplication and division.
13887 * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
13888 (cond_udiv@var{m}, cond_umod@var{m}): Document.
13889 * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
13890 (cond_udiv_optab, cond_umod_optab): New optabs.
13891 * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
13892 (IFN_COND_RDIV): New internal functions.
13893 * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
13894 TRUNC_MOD_EXPR and RDIV_EXPR.
13895 * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
13896 * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
13898 (SVE_INT_BINARY): Include mult.
13899 (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
13900 (optab, sve_int_op): Handle mult.
13901 (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
13903 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
13904 for SVE_INT_BINARY_SD.
13906 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
13908 * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
13909 (optab, sve_int_op): Handle div and udiv.
13910 * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
13911 for SVE_INT_BINARY_SD.
13912 (*<optab><mode>3): New insn for the same.
13914 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
13916 * tree-vect-patterns.c: Include predict.h.
13917 (vect_recog_divmod_pattern): Restrict check for division support
13918 to when optimizing for size.
13920 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
13922 * doc/sourcebuild.texi (vect_double_cond_arith: Document.
13923 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
13924 (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
13925 (gimple_match_op::set_op): Likewise.
13926 (gimple_resimplify4): Declare.
13927 * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
13928 (expr::gen_transform): Likewise.
13929 (decision_tree::gen): Generate a simplification routine for 4 operands.
13930 * gimple-match-head.c (gimple_simplify): Add an overload for
13931 4 operands. In the top-level function, handle up to 4 call
13932 arguments and call gimple_resimplify4.
13933 (gimple_resimplify4): New function.
13934 (build_call_internal): Pass a fourth operand.
13935 (maybe_push_to_seq): Likewise.
13936 * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
13937 Fold VEC_COND_EXPRs of an operation and a default value into
13938 an IFN_COND_* function if possible.
13939 * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
13941 (SVE_COND_FP_BINARY): Include them.
13942 (optab, sve_fp_op): Handle them.
13943 (SVE_INT_BINARY_REV): New code iterator.
13944 (SVE_COND_FP_BINARY_REV): New int iterator.
13945 (commutative): New int attribute.
13946 * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
13948 * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
13950 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
13951 (*cond_<optab><mode>): New patterns for reversed operands.
13953 2018-05-25 Richard Biener <rguenther@suse.de>
13955 * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
13956 (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
13957 (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
13958 (STMT_VINFO_GROUPED_ACCESS): Adjust.
13959 * tree-vect-data-refs.c (everywhere): Adjust users.
13960 * tree-vect-loop.c (everywhere): Likewise.
13961 * tree-vect-slp.c (everywhere): Likewise.
13962 * tree-vect-stmts.c (everywhere): Likewise.
13963 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
13965 2018-05-25 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
13967 * configure.ac (gcc_cv_as_section_has_e): Move to common section.
13969 (gcc_cv_as_section_exclude): ... this.
13970 Try Solaris as #exclude syntax.
13971 * configure: Regenerate.
13972 * config.in: Regenerate.
13973 * config/i386/i386.c (i386_solaris_elf_named_section): Handle
13975 * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
13976 [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
13978 * varasm.c (default_elf_asm_named_section): Don't check if
13979 HAVE_GAS_SECTION_EXCLUDE is defined.
13981 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
13983 * doc/md.texi: Update the documentation of the cond_* optabs
13984 to mention the new final operand. Fix GET_MODE_NUNITS call.
13985 Describe the scalar case too.
13986 * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
13987 * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
13989 (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
13990 (get_conditional_internal_fn): Update comment.
13991 * tree-vect-loop.c (vectorizable_reduction): Pass the original
13992 accumulator value as a final argument to conditional functions.
13993 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
13994 a define_expand and add an "else" operand. Assert for now that
13995 the else operand is equal to operand 2. Use SVE_INT_BINARY and
13996 SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
13997 (*cond_<optab><mode>): New patterns.
13998 * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
13999 (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
14000 (UNSPEC_COND_EOR): Delete.
14001 (optab): Remove associated mappings.
14002 (SVE_INT_BINARY): New code iterator.
14003 (sve_int_op): Remove int attribute and add "minus" to the code
14005 (SVE_COND_INT_OP): Delete.
14006 (SVE_COND_FP_OP): Rename to...
14007 (SVE_COND_FP_BINARY): ...this.
14009 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
14011 * optabs.c (can_reuse_operands_p): New function.
14012 (maybe_legitimize_operands): Try to reuse the results for
14015 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
14017 * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
14018 Add {q} suffix to insn mnemonic.
14020 2018-05-23 Jozef Lawrynowicz <jozef.l@mittosystems.com>
14022 * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
14023 (msp430_warn_func_return): New.
14025 2018-05-24 Roger Sayle <roger@nextmovesoftware.com>
14027 * fold-const.c (tree_nonzero_bits): New function.
14028 * fold-const.h (tree_nonzero_bits): Likewise.
14029 * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
14030 friends. POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
14032 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
14036 * varasm.c (assemble_alias): Check ifunc_resolver only on
14039 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
14042 * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
14043 when memory input operand is handled.
14045 2018-05-24 Luis Machado <luis.machado@linaro.org>
14047 * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
14049 (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
14051 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
14053 * match.pd: Delay FMA folds until after vectorization.
14055 2018-05-24 Andre Vieira <andre.simoesdiasvieira@arm.com>
14058 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
14059 address check not strict.
14061 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
14063 * gimple-match.h (gimple_match_op): New class.
14064 (mprts_hook): Replace parameters with a gimple_match_op *.
14065 (maybe_build_generic_op): Likewise.
14066 (gimple_simplified_result_is_gimple_val): Replace parameters with
14067 a const gimple_match_op *.
14068 (gimple_simplify): Replace code_helper * and tree * parameters with
14069 a gimple_match_op * parameter.
14070 (gimple_resimplify1): Replace code_helper *, tree and tree *
14071 parameters with a gimple_match_op * parameter.
14072 (gimple_resimplify2): Likewise.
14073 (gimple_resimplify3): Likewise.
14074 (maybe_push_res_to_seq): Replace code_helper, tree and tree *
14075 parameters with a gimple_match_op * parameter.
14076 * gimple-match-head.c (gimple_simplify): Change prototypes of
14077 auto-generated functions to take a gimple_match_op * instead of
14078 separate code_helper * and tree * parameters. Make the same
14079 change in the top-level overload and update calls to the
14080 gimple_resimplify routines. Update calls to the auto-generated
14081 functions and to maybe_push_res_to_seq in the publicly-facing
14082 operation-specific gimple_simplify overloads.
14083 (gimple_match_op::MAX_NUM_OPS): Define.
14084 (gimple_resimplify1): Replace rcode and ops with a single res_op
14085 parameter. Update call to gimple_simplify.
14086 (gimple_resimplify2): Likewise.
14087 (gimple_resimplify3): Likewise.
14088 (mprts_hook): Replace parameters with a gimple_match_op *.
14089 (maybe_build_generic_op): Likewise.
14090 (build_call_internal): Replace type, nargs and ops with
14091 a gimple_match_op *.
14092 (maybe_push_res_to_seq): Replace res_code, type and ops parameters
14093 with a single gimple_match_op *. Update calls to mprts_hook,
14094 build_call_internal and gimple_simplified_result_is_gimple_val.
14095 Factor out code that is common to the tree_code and combined_fn cases.
14096 * genmatch.c (expr::gen_transform): Replace tem_code and
14097 tem_ops with a gimple_match_op called tem_op. Update calls
14098 to the gimple_resimplify functions and maybe_push_res_to_seq.
14099 (dt_simplify::gen_1): Manipulate res_op instead of res_code and
14100 res_ops. Update call to the gimple_resimplify functions.
14101 (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
14102 (decision_tree::gen): Make the functions take a gimple_match_op *
14103 called res_op instead of separate res_code and res_ops parameters.
14104 Update call accordingly.
14105 * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
14106 and ops with a single res_op parameter. Update calls to
14107 maybe_build_generic_op and maybe_push_res_to_seq.
14108 (fold_stmt_1): Update calls to gimple_simplify and
14109 replace_stmt_with_simplification.
14110 (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
14111 and gimple_simplified_result_is_gimple_val.
14112 * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
14114 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
14115 with a gimple_match_op *.
14116 (vn_nary_build_or_lookup): Likewise. Update call to
14117 vn_nary_build_or_lookup_1.
14118 (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
14119 gimple_match_op *. Update calls to the gimple_resimplify routines
14120 and to gimple_simplified_result_is_gimple_val.
14121 (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
14122 Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
14123 (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
14124 (visit_nary_op): Likewise.
14125 (visit_reference_op_load): Likewise.
14127 2018-05-23 Luis Machado <luis.machado@linaro.org>
14129 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
14130 modifier for printing the step amount.
14132 2018-05-23 Jozef Lawrynowicz <jozef.l@somniumtech.com>
14135 * tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
14138 2018-05-23 Segher Boessenkool <segher@kernel.crashing.org>
14140 * doc/sourcebuild.texi (Endianness): New subsubsection.
14142 2018-05-23 Luis Machado <luis.machado@linaro.org>
14144 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
14145 <prefetch_dynamic_strides>: New const bool field.
14146 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
14147 prefetch_dynamic_strides.
14148 (exynosm1_prefetch_tune): Likewise.
14149 (thunderxt88_prefetch_tune): Likewise.
14150 (thunderx_prefetch_tune): Likewise.
14151 (thunderx2t99_prefetch_tune): Likewise.
14152 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
14154 (aarch64_override_options_internal): Update to set
14155 PARAM_PREFETCH_DYNAMIC_STRIDES.
14156 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
14157 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
14158 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
14159 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
14160 prefetch-dynamic-strides setting.
14162 2018-05-23 Luis Machado <luis.machado@linaro.org>
14164 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
14165 <minimum_stride>: New const int field.
14166 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
14167 minimum_stride field defaulting to -1.
14168 (exynosm1_prefetch_tune): Likewise.
14169 (thunderxt88_prefetch_tune): Likewise.
14170 (thunderx_prefetch_tune): Likewise.
14171 (thunderx2t99_prefetch_tune): Likewise.
14172 (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
14173 <default_opt_level>: Set to 3.
14174 (aarch64_override_options_internal): Update to set
14175 PARAM_PREFETCH_MINIMUM_STRIDE.
14176 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
14177 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
14178 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
14179 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
14180 stride is constant and is below the minimum stride threshold.
14182 2018-05-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14184 * config/arm/arm-cpus.in (mode26): Delete.
14185 (armv4): Delete mode26 reference.
14186 * config/arm/arm.c (arm_configure_build_target): Delete use of
14189 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
14191 * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
14193 (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
14194 Rewrite expander pattern. Emit gen_floatunssi<mode>2_i387_with_xmm
14196 (floatunsdisf2): Rewrite expander pattern. Hanlde TARGET_AVX512F.
14197 (floatunsdidf2): Ditto.
14199 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
14201 * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
14202 (fixuns_trunc<mode>si2_avx512f): Ditto.
14203 (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
14204 (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
14205 Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
14207 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
14209 PR rtl-optimization/79985
14210 * df-scan.c (df_insn_refs_collect): Remove special case for
14211 global registers and asm statements.
14213 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
14215 * extend.texi (Global Register Variables): Rewrite the bullet list.
14216 Note that the register is available for allocation. Note that access
14217 via inline asm must use constraints. Add note about async-signal
14218 handlers. Remove paragraph about automagic register selection.
14220 2018-05-23 Richard Biener <rguenther@suse.de>
14222 * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
14223 of fixed offset from memset VN.
14225 2018-05-23 Bill Schmidt <wschmidt@linux.ibm.com>
14227 * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
14228 first_interp field.
14229 (alloc_cand_and_find_basis): Initialize first_interp field.
14230 (slsr_process_mul): Modify first_interp field.
14231 (slsr_process_add): Likewise.
14232 (slsr_process_cast): Modify first_interp field for each new
14234 (slsr_process_copy): Likewise.
14235 (dump_candidate): Dump first_interp field.
14236 (replace_mult_candidate): Process all interpretations, not just
14238 (replace_rhs_if_not_dup): Likewise.
14239 (replace_one_candidate): Likewise.
14241 2018-05-23 Wilco Dijkstra <wdijkstr@arm.com>
14243 * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
14245 (aarch64_needs_frame_chain): New function.
14246 (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
14248 2018-05-23 Sudakshina Das <sudi.das@arm.com>
14251 * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
14252 Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
14253 * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
14254 * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
14255 as true for strict-align.
14256 (aarch64_can_inline_p): Perform checks even when callee has no
14257 attributes to check for strict alignment.
14258 * doc/extend.texi (AArch64 Function Attributes): Document
14260 * doc/invoke.texi: (AArch64 Options): Likewise.
14262 2018-05-23 Richard Sandiford <richard.sandiford@linaro.org>
14264 PR tree-optimization/85853
14265 * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
14266 the handling of the root of the node to...
14267 (vect_slp_analyze_node_operations_1): ...this new function,
14268 and run the whole thing with the child nodes' def types
14269 set according to their SLP node's def type.
14271 2018-05-23 Richard Biener <rguenther@suse.de>
14273 PR middle-end/85874
14274 * tree-data-ref.c (create_runtime_alias_checks): Defer
14275 and ignore overflow warnings.
14277 2018-05-23 Yury Gribov <tetra2005@gmail.com>
14279 PR tree-optimization/85822
14280 * tree-vrp.c (is_masked_range_test): Fix handling of negative
14283 2018-05-23 Richard Biener <rguenther@suse.de>
14285 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
14286 memset constants via native_interpret_expr.
14288 2018-05-22 H.J. Lu <hongjiu.lu@intel.com>
14291 * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
14293 (cgraph_node::create_alias): Likewise.
14294 (cgraph_node::get_availability): Check ifunc_resolver instead
14295 of looking up ifunc attribute.
14296 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
14297 * varasm.c (do_assemble_alias): Likewise.
14298 (assemble_alias): Likewise.
14299 (default_binds_local_p_3): Likewise.
14300 * cgraph.h (cgraph_node): Add ifunc_resolver.
14301 (cgraph_node::only_called_directly_or_aliased_p): Return false
14302 for IFUNC resolver.
14303 * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
14305 * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
14306 is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
14307 (symtab_node::binds_to_current_def_p): Check ifunc_resolver
14308 instead of looking up ifunc attribute.
14310 2018-05-22 Luis Machado <luis.machado@linaro.org>
14312 * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
14314 2018-05-22 Martin Sebor <msebor@redhat.com>
14316 PR middle-end/85359
14317 * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
14318 only when expasion succeeds.
14319 (expand_builtin_strcmp): Same.
14320 (expand_builtin_strncmp): Same.
14322 2018-05-22 Martin Sebor <msebor@redhat.com>
14324 * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
14326 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
14327 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14329 * config/aarch64/aarch64-ldpstp.md: Replace uses of
14330 aarch64_mem_pair_operand with memory_operand and delete operand swapping
14332 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
14333 Add check for legitimate_address.
14334 (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
14335 (aarch64_swap_ldrstr_operands): New.
14336 * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
14339 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
14340 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14342 * config/aarch64/aarch64.md: New patterns to generate stp
14344 (store_pair_sw, store_pair_dw): New patterns to generate stp for
14345 single words and double words.
14346 (load_pair_sw, load_pair_dw): Likewise.
14347 (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
14349 (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
14351 * config/aarch64/aarch64-ldpstp.md: Modify peephole
14352 for different mode ldpstp and add peephole for merged zero stores.
14353 Likewise for loads.
14354 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
14356 (aarch64_gen_store_pair): Rename calls to match new patterns.
14357 (aarch64_gen_load_pair): Rename calls to match new patterns.
14358 * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
14359 (load_pair<DREG:mode><DREG2:mode>): ... This.
14360 (store_pair<mode>): Rename to...
14361 (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
14362 * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
14363 New mode iterators.
14364 (V_INT_EQUIV): Handle SImode.
14365 * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
14368 2018-05-22 Martin Sebor <msebor@redhat.com>
14371 * calls.c (maybe_warn_nonstring_arg): Use string length to set
14372 or ajust the presumed bound on an operation to avoid unnecessary
14375 2018-05-22 Martin Sebor <msebor@redhat.com>
14377 PR tree-optimization/85826
14378 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
14379 assuming that a DECL necesarily has a constant size.
14381 2018-05-22 Richard Sandiford <richard.sandiford@linaro.org>
14383 PR middle-end/85862
14384 * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
14386 2018-05-22 Richard Biener <rguenther@suse.de>
14388 PR tree-optimization/85834
14389 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
14390 non-constant and non-zero memset arguments.
14392 2018-05-22 Martin Liska <mliska@suse.cz>
14395 * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
14397 2018-05-22 Richard Biener <rguenther@suse.de>
14399 PR tree-optimization/85863
14400 * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
14401 comparisons when vectype is specified.
14402 (vectorizable_condition): Do not specify vectype for
14403 vect_is_simple_cond when SLP vectorizing.
14405 2018-05-21 Michael Meissner <meissner@linux.ibm.com>
14408 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
14409 define __ibm128 as long double.
14410 * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
14411 as a distinct type when IEEE 128-bit support is enabled.
14412 (init_float128_ieee): Fix up conversions between IFmode and IEEE
14413 128-bit types to use the correct functions.
14414 (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
14415 convert between 128-bit floating point types that have different
14416 modes but the same representation, instead of using gen_lowpart to
14418 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
14420 (IFKF_reg): New attributes to give the register constraints for
14422 (extend<mode>tf2_internal): New insns to mark an explicit
14423 conversion between 128-bit floating point types that have a
14424 different mode but share the same representation.
14426 2018-05-21 Richard Sandiford <richard.sandiford@linaro.org>
14428 PR tree-optimization/85814
14429 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
14430 a null return from get_strinfo when unsharing the next
14431 strinfo in the chain.
14433 2018-05-21 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
14436 * varasm.c (weak_finish): Clean up weak_decls.
14438 2018-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14440 * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
14441 UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
14442 UNSPEC_UADALP values.
14443 * config/aarch64/iterators.md (ABAL): New int iterator.
14446 (sur): Add mappings for the above.
14447 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
14449 (aarch64_<sur>abal<mode>_4): Likewise.
14450 (aarch64_<sur>adalp<mode>_3): Likewise.
14451 (<sur>sadv16qi): New define_expand.
14453 2018-05-21 Alexander Nesterovskiy <alexander.nesterovskiy@intel.com>
14455 * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
14456 (*movdf_internal): Ditto.
14457 (*rcpsf2_sse): Ditto.
14458 (*rsqrtsf2_sse): Ditto.
14459 (*sqrt<mode>2_sse): Ditto.
14461 2018-05-21 Tamar Christina <tamar.christina@arm.com>
14463 * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
14465 (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
14466 * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
14467 veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
14468 vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
14470 * config/aarch64/arm_neon.h: Likewise.
14471 * config/aarch64/iterators.md (VQ_I): New.
14473 2018-05-21 Alexey Brodkin <abrodkin@synopsys.com>
14475 * config.gcc: Add arc/t-multilib-linux to tmake_file for
14477 * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
14480 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
14482 * config/nds32/constraints.md (S): New constraint.
14483 * config/nds32/nds32.md (call_internal): Use constraint S.
14484 (call_value_internal): Likewise.
14485 (sibcall_internal): Likewise.
14486 (sibcall_value_internal): Likewise.
14488 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
14489 Chung-Ju Wu <jasonwucj@gmail.com>
14491 * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
14492 into consideration.
14494 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
14495 Chung-Ju Wu <jasonwucj@gmail.com>
14497 * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
14498 (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
14499 (nds32_rtx_costs_impl): Simplify.
14500 (nds32_address_cost_impl): Simplify.
14501 (nds32_init_rtx_costs): New function.
14502 (nds32_rtx_costs_speed_prefer): Likewise.
14503 (nds32_rtx_costs_size_prefer): Likewise.
14504 (nds32_address_cost_speed_prefer): Likewise.
14505 (nds32_address_cost_speed_fwprop): Likewise.
14506 (nds32_address_cost_size_prefer): Likewise.
14507 * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
14508 * config/nds32/nds32.c (nds32_option_override): Use
14509 nds32_init_rtx_costs function.
14511 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
14513 * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
14514 * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
14515 (TARGET_PIPELINE_N8): Likewise.
14516 (TARGET_PIPELINE_N10): Likewise.
14517 (TARGET_PIPELINE_N13): Likewise.
14518 (TARGET_PIPELINE_GRAYWOLF): Likewise.
14520 2018-05-19 Monk Chiang <sh.chiang04@gmail.com>
14522 * config/nds32/nds32-fpu.md: Update copyright year.
14524 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
14526 * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
14528 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
14530 * config/nds32/nds32.c
14531 (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
14532 * config/nds32/nds32.opt (minline-asm-r15): New option.
14534 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
14536 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
14538 * config/nds32/nds32.md (abssi2): New pattern.
14540 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
14542 * config/i386/i386.md (rex64namesuffix): New mode attribute.
14543 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
14544 Merge insn pattern from sse_cvtsi2ss<round_name> and
14545 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
14546 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
14547 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
14548 using SWI48 mode iterator.
14549 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
14550 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
14551 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
14552 pattern from sse_cvttss2si<round_saeonly_name>
14553 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
14554 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
14555 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
14556 using SWI48 mode iterator.
14557 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
14558 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
14559 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
14560 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
14561 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
14562 using SWI48 mode iterator.
14563 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
14564 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
14565 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
14566 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
14567 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
14568 SWI48 mode iterator.
14569 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
14570 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
14571 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
14572 pattern from sse_cvttsd2si<round_saeonly_name>
14573 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
14575 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
14577 * config/nds32/nds32-md-auxiliary.c
14578 (nds32_valid_smw_lwm_base_p): Refine.
14579 (nds32_output_smw_single_word): Refine.
14580 (nds32_output_smw_double_word): New.
14581 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
14583 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
14585 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
14586 (nds32_output_stack_pop): Refine.
14587 (nds32_expand_unaligned_load): Refine.
14588 (nds32_expand_unaligned_store): Refine.
14590 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
14591 Chung-Ju Wu <jasonwucj@gmail.com>
14593 * config/nds32/constants.md: Add TP_REGNUM constant.
14594 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
14595 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
14597 * config/nds32/nds32-doubleword.md: Consider flag_pic.
14598 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
14599 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
14600 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
14601 and PIC code generation.
14602 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
14604 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
14606 * config/nds32/nds32.md: Support TLS and PIC.
14607 * config/nds32/nds32.c: Support TLS and PIC.
14608 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
14609 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
14612 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
14614 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
14615 mode with E_ prefix.
14617 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
14618 Chung-Ju Wu <jasonwucj@gmail.com>
14620 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
14621 * config/nds32/nds32-md-auxiliary.c
14622 (symbolic_reference_mentioned_p): New.
14623 (nds32_legitimize_ict_address): New.
14624 (nds32_expand_ict_move): New.
14625 (nds32_indirect_call_referenced_p): New.
14626 (nds32_symbol_binds_local_p): Delete.
14627 (nds32_long_call_p): Modify.
14628 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
14629 * config/nds32/nds32-protos.h
14630 (symbolic_reference_mentioned_p): Declare.
14631 (nds32_legitimize_ict_address): Declare.
14632 (nds32_expand_ict_move): Declare.
14633 (nds32_indirect_call_referenced_p): Declare.
14634 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
14635 (nds32_relax_group): Use nds32_ict_const_p as condition.
14636 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
14637 (nds32_asm_file_start): Output ict_model directive in asm code.
14638 (nds32_legitimate_address_p): Consider indirect call.
14639 (nds32_print_operand): Consider indirect call.
14640 (nds32_print_operand_address): Consider indirect call.
14641 (nds32_insert_attributes): Handle "indirect_call" attribute.
14642 (TARGET_LEGITIMATE_ADDRESS_P): Define.
14643 (TARGET_LEGITIMATE_CONSTANT_P): Define.
14644 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
14645 (TARGET_DELEGITIMIZE_ADDRESS): Define.
14646 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
14647 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
14648 (TARGET_ICT_MODEL_SMALL): Define.
14649 (TARGET_ICT_MODEL_LARGE): Define.
14650 * config/nds32/nds32.md (movsi): Consider ict model.
14651 (call, call_value): Consider ict model.
14652 (sibcall, sibcall_value): Consider ict model.
14653 * config/nds32/nds32.opt (mict-model): New option.
14654 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
14657 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
14658 Monk Chiang <sh.chiang04@gmail.com>
14659 Jim Wilson <jimw@sifive.com>
14661 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
14662 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
14663 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
14664 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
14665 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
14666 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
14667 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
14668 compute save_libcall_adjustment properly.
14669 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
14670 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
14671 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
14672 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
14673 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
14674 (ABI_SPEC): Handle mabi=ilp32e.
14675 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
14676 (RVE): Add RVE mask.
14677 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
14678 <-march>: Add rv32e as an example.
14680 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
14683 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
14684 (intra_create_variable_infos): Handle C++ constructors.
14686 2018-05-18 Martin Liska <mliska@suse.cz>
14688 * passes.def: Remove a redundant pass.
14690 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
14693 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
14695 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14697 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
14699 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
14701 (armv2, armv2a, armv3, armv3m): Delete architectures.
14702 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
14703 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
14704 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
14706 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
14707 (*mulsidi3adddi): Likewise.
14708 (mulsidi3): Likewise.
14709 (*mulsidi3_nov6): Likewise.
14710 (umulsidi3): Likewise.
14711 (umulsidi3_nov6): Likewise.
14712 (umaddsidi4): Likewise.
14713 (*umulsidi3adddi): Likewise.
14714 (smulsi3_highpart): Likewise.
14715 (*smulsi3_highpart_nov6): Likewise.
14716 (umulsi3_highpart): Likewise.
14717 (*umulsi3_highpart_nov6): Likewise.
14718 * config/arm/arm.h (arm_arch3m): Delete.
14719 * config/arm/arm.c (arm_arch3m): Delete.
14720 (arm_option_override_internal): Update armv3-related comment.
14721 (arm_configure_build_target): Delete use of isa_bit_mode32.
14722 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
14723 (arm_rtx_costs_internal): Delete check of arm_arch3m.
14724 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
14725 (mulsa3): Likewise.
14726 (mulusa3): Likewise.
14727 * config/arm/arm-protos.h (arm_arch3m): Delete.
14728 * config/arm/arm-tables.opt: Regenerate.
14729 * config/arm/arm-tune.md: Likewise.
14730 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
14731 deleted architectures.
14733 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14735 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
14736 (armv5t, armv5te): New features.
14737 (ARMv5, ARMv5e): Delete fgroups.
14738 (ARMv5t, ARMv5te): Adjust for above changes.
14739 (ARMv6m): Likewise.
14740 (armv5, armv5e): Delete arches.
14741 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
14743 (*call_reg_arm): Likewise.
14744 (*call_value_reg_armv5): Likewise.
14745 (*call_value_reg_arm): Likewise.
14746 (*call_symbol): Likewise.
14747 (*call_value_symbol): Likewise.
14748 (*sibcall_insn): Likewise.
14749 (*sibcall_value_insn): Likewise.
14750 (clzsi2): Likewise.
14751 (prefetch): Likewise.
14752 (define_split and define_peephole2 dependent on arm_arch5):
14754 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
14756 (TARGET_ARM_QBIT): Likewise.
14757 (TARGET_DSP_MULTIPLY): Likewise.
14758 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
14759 (arm_arch5, arm_arch5e): Delete.
14760 (arm_arch5t, arm_arch5te): Declare.
14761 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
14762 (arm_arch5t): Declare.
14763 (arm_option_reconfigure_globals): Update for the above.
14764 (arm_options_perform_arch_sanity_checks): Update comment, replace
14765 use of arm_arch5 with arm_arch5t.
14766 (use_return_insn): Likewise.
14767 (arm_emit_call_insn): Likewise.
14768 (output_return_instruction): Likewise.
14769 (arm_final_prescan_insn): Likewise.
14770 (arm_coproc_builtin_available): Likewise.
14771 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
14772 arm_arch5e with arm_arch5t and arm_arch5te.
14773 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
14774 (arm_arch5t, arm_arch5te): Declare.
14775 * config/arm/arm-tables.opt: Regenerate.
14776 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
14777 * config/arm/t-multilib: Likewise.
14778 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
14779 instead of arm_arch5.
14780 (*call_reg_thumb1): Likewise.
14781 (*call_value_reg_thumb1_v5): Likewise.
14782 (*call_value_reg_thumb1): Likewise.
14783 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
14785 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
14787 2018-05-18 Martin Liska <mliska@suse.cz>
14789 PR gcov-profile/84846
14790 * doc/gcov.texi: Document -t option of gcov tool.
14792 2018-05-18 Martin Liska <mliska@suse.cz>
14794 PR gcov-profile/84846
14795 * gcov.c (print_usage): Add new -t option.
14796 (process_args): Handle the option.
14797 (generate_results): Use stdout as output when requested by
14800 2018-05-18 Martin Liska <mliska@suse.cz>
14802 PR gcov-profile/84846
14803 * coverage.c (coverage_init): Write PWD to .gcno file.
14804 * doc/gcov.texi: Document how working directory is printed.
14805 * gcov-dump.c (dump_gcov_file): Print PWD.
14806 * gcov.c (output_intermediate_file): Likewise.
14807 (read_graph_file): Read PWD string.
14808 (output_lines): Print PWD.
14810 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
14812 PR middle-end/85817
14813 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
14814 for retval and return false if all args to phi are zero.
14816 2018-05-18 Richard Biener <rguenther@suse.de>
14818 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
14820 (evrp_dom_walker::before_dom_children): Call it.
14822 2018-05-18 Richard Biener <rguenther@suse.de>
14824 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
14825 results when processing array refs with variable index.
14827 2018-05-18 Toon Moene <toon@moene.org>
14829 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
14830 directly after that of -floop-interchange. Indicate that both
14831 options are enabled by default when specifying -O3.
14833 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14835 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
14836 iterator. Delete separate integer-mode vec_set<mode> expander.
14837 (aarch64_simd_vec_setv2di): Delete.
14838 (vec_setv2di): Delete.
14839 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
14840 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
14841 the "w, r" alternative.
14843 2018-05-18 Martin Liska <mliska@suse.cz>
14845 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
14846 * tree-pass.h (make_pass_lower_switch_O0): New function.
14847 * tree-switch-conversion.c (node_has_low_bound): Remove.
14848 (node_has_high_bound): Likewise.
14849 (node_is_bounded): Likewise.
14850 (class pass_lower_switch): Make it a template type and create
14852 (pass_lower_switch::execute): Add template argument.
14853 (make_pass_lower_switch): New function.
14854 (make_pass_lower_switch_O0): New function.
14855 (do_jump_if_equal): Remove.
14856 (emit_case_nodes): Simplify to just handle all 3 cases and leave
14857 all the hard work to tree optimization passes.
14859 2018-05-18 Martin Liska <mliska@suse.cz>
14861 * dbgcnt.c (limit_low): Renamed from limit.
14862 (limit_high): New variable.
14863 (dbg_cnt_is_enabled): Check for upper limit.
14864 (dbg_cnt): Adjust dumping.
14865 (dbg_cnt_set_limit_by_index): Add new argument for high
14867 (dbg_cnt_set_limit_by_name): Likewise.
14868 (dbg_cnt_process_single_pair): Parse new format.
14869 (dbg_cnt_process_opt): Use strtok.
14870 (dbg_cnt_list_all_counters): Remove 'value' and add
14872 * doc/invoke.texi: Document changes.
14874 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
14876 * doc/sourcebuild.texi (scalar_all_fma): Document.
14877 * tree.def (FMA_EXPR): Delete.
14878 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
14879 * internal-fn.c (ternary_direct): New macro.
14880 (expand_ternary_optab_fn): Likewise.
14881 (direct_ternary_optab_supported_p): Likewise.
14882 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
14883 * builtins.c (fold_builtin_fma): Delete.
14884 (fold_builtin_3): Don't call it.
14885 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
14886 * expr.c (expand_expr_real_2): Likewise.
14887 * fold-const.c (operand_equal_p): Likewise.
14888 (fold_ternary_loc): Likewise.
14889 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
14890 * gimple.c (DEFTREECODE): Likewise.
14891 * gimplify.c (gimplify_expr): Likewise.
14892 * optabs-tree.c (optab_for_tree_code): Likewise.
14893 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
14894 * tree-eh.c (operation_could_trap_p): Likewise.
14895 (stmt_could_throw_1_p): Likewise.
14896 * tree-inline.c (estimate_operator_cost): Likewise.
14897 * tree-pretty-print.c (dump_generic_node): Likewise.
14898 (op_code_prio): Likewise.
14899 * tree-ssa-loop-im.c (stmt_cost): Likewise.
14900 * tree-ssa-operands.c (get_expr_operands): Likewise.
14901 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
14902 * fold-const-call.h (fold_fma): Delete.
14903 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
14904 CFN_FNMA and CFN_FNMS.
14905 (fold_fma): Delete.
14906 * genmatch.c (combined_fn): New enum.
14907 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
14908 (commutative_op): New function.
14909 (commutate): Use it. Handle more than 2 operands.
14910 (dt_operand::gen_gimple_expr): Use commutative_op.
14911 (parser::parse_expr): Allow :c to be used with non-binary
14912 operators if the commutative operand is known.
14913 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
14914 CFN_FMS, CFN_FNMA and CFN_FNMS.
14915 (backprop::process_assign_use): Remove FMA_EXPR handling.
14916 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
14917 (gen_hsa_fma): New function.
14918 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
14919 IFN_FNMA and IFN_FNMS.
14920 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
14921 * gimple-fold.h (follow_all_ssa_edges): Declare.
14922 * gimple-fold.c (follow_all_ssa_edges): New function.
14923 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
14924 gimple_build interface and use follow_all_ssa_edges to fold the result.
14925 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
14926 instead of checking for optabs directly.
14927 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
14928 rather than FMA_EXPRs.
14929 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
14930 call to IFN_FMA instead of an FMA_EXPR.
14932 2018-05-17 Jim Wilson <jimw@sifive.com>
14934 * expr.c (do_tablejump): When converting index to Pmode, if we have a
14935 sign extended promoted subreg, and the range does not have the sign bit
14936 set, then do a sign extend.
14938 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
14939 test, check for sign extended subreg and/or constant operands, and
14940 do a sign extend in that case.
14942 2018-05-17 Steve Ellcey <sellcey@cavium.com>
14944 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
14945 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
14947 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
14948 Change logics_shift_reg to logics_shift_imm.
14949 (thunderx2t99_fp_loadpair_basic): Delete.
14950 (thunderx2t99_fp_storepair_basic): Delete.
14951 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
14952 (thunderx2t99_asimd_polynomial): Delete.
14953 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
14954 and neon_fp_mul_d_scalar_q.
14955 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
14956 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
14957 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
14958 (thunderx2t99_asimd_lut): Add missing tbl types.
14959 (thunderx2t99_asimd_ext): Delete.
14960 (thunderx2t99_asimd_load1_1_mult): Delete.
14961 (thunderx2t99_asimd_load1_2_mult): Delete.
14962 (thunderx2t99_asimd_load1_ldp): New.
14963 (thunderx2t99_asimd_load1): New.
14964 (thunderx2t99_asimd_load2): Add missing *load2* types.
14965 (thunderx2t99_asimd_load3): New.
14966 (thunderx2t99_asimd_load4): New.
14967 (thunderx2t99_asimd_store1_1_mult): Delete.
14968 (thunderx2t99_asimd_store1_2_mult): Delete.
14969 (thunderx2t99_asimd_store2_mult): Delete.
14970 (thunderx2t99_asimd_store2_onelane): Delete.
14971 (thunderx2t99_asimd_store_stp): New.
14972 (thunderx2t99_asimd_store1): New.
14973 (thunderx2t99_asimd_store2): New.
14974 (thunderx2t99_asimd_store3): New.
14975 (thunderx2t99_asimd_store4): New.
14977 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
14979 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
14980 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
14982 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
14983 Segher Boessenkool <segher@kernel.crashing.org>
14986 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
14989 2018-05-17 Richard Biener <rguenther@suse.de>
14991 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
14992 for pruning loop and prune defs feeding only already visited PHIs.
14994 2018-05-17 Richard Biener <rguenther@suse.de>
14996 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
14998 2018-05-17 Bin Cheng <bin.cheng@arm.com>
14999 Richard Biener <rguenther@suse.de>
15001 PR tree-optimization/85793
15002 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
15003 for VMAT_ELEMENTWISE.
15005 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
15007 * internal-fn.h (lookup_internal_fn): Declare
15008 * internal-fn.c (lookup_internal_fn): New function.
15009 * gimple.c (gimple_build_call_from_tree): Handle calls to
15010 internal functions.
15011 * gimple-pretty-print.c (dump_gimple_call): Print "." before
15012 internal function names.
15013 * tree-pretty-print.c (dump_generic_node): Likewise.
15014 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
15016 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
15018 * gimple-fold.h (gimple_build): Make the function forms take
15019 combined_fn rather than built_in_function.
15020 (gimple_simplify): Likewise.
15021 * gimple-match-head.c (gimple_simplify): Likewise.
15022 * gimple-fold.c (gimple_build): Likewise.
15023 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
15024 rather than gimple_build_call_internal.
15025 (get_initial_defs_for_reduction): Likewise.
15026 (vect_create_epilog_for_reduction): Likewise.
15027 (vectorizable_live_operation): Likewise.
15029 2018-05-17 Martin Liska <mliska@suse.cz>
15031 * gimple-ssa-sprintf.c (format_directive): Do not use
15032 space in between 'G_' and '('.
15034 2018-05-17 Jakub Jelinek <jakub@redhat.com>
15037 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
15038 even if the mask is not all ones.
15041 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
15043 (ix86_gimple_fold_builtin): Likewise.
15046 * config/i386/i386.c: Include tree-vector-builder.h.
15047 (ix86_vector_shift_count): New function.
15048 (ix86_fold_builtin): Fold shift builtins by scalar count.
15049 (ix86_gimple_fold_builtin): Likewise.
15051 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
15052 _mm512_setzero): New intrinsics.
15054 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
15055 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15057 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
15058 code generation for cases where splatting a value is not useful.
15059 * simplify-rtx.c (simplify_ternary_operation): Simplify
15060 vec_merge across a vec_duplicate and a paradoxical subreg forming
15061 a vector mode to a vec_concat.
15063 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
15065 * config.gcc: Support "goldmont-plus".
15066 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
15068 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
15069 PROCESSOR_GOLDMONT_PLUS.
15070 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
15071 (processor_target_table): Add "goldmont-plus".
15072 (PTA_GOLDMONT_PLUS): Define.
15073 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
15074 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
15075 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
15076 (fold_builtin_cpu): Add "goldmont-plus".
15077 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
15078 (ix86_option_override_internal): Add "goldmont-plus".
15079 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
15080 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
15081 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
15082 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
15084 2018-05-17 Richard Biener <rguenther@suse.de>
15086 PR tree-optimization/85757
15087 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
15088 remove defs that only feed that PHI from further processing.
15090 2018-05-16 Jim Wilson <jimw@sifive.com>
15092 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
15094 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
15096 2018-05-16 Mark Wielaard <mark@klomp.org>
15098 * dwarf2out.c (count_index_strings): New function.
15099 (output_indirect_strings): Call count_index_strings and generate
15100 header for dwarf_version >= 5.
15102 2018-05-16 Mark Wielaard <mark@klomp.org>
15104 * dwarf2out.c (dwarf_FORM): New function.
15105 (set_indirect_string): Use dwarf_FORM.
15106 (reset_indirect_string): Likewise.
15107 (size_of_die): Likewise.
15108 (value_format): Likewise.
15109 (output_die): Likewise.
15110 (add_skeleton_AT_string): Likewise.
15111 (output_macinfo_op): Likewise.
15112 (index_string): Likewise.
15113 (output_index_string_offset): Likewise.
15114 (output_index_string): Likewise.
15115 (count_index_strings): Likewise.
15117 2018-05-16 Carl Love <cel@us.ibm.com>
15119 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
15120 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
15122 2018-05-16 Martin Jambor <mjambor@suse.cz>
15124 * ipa-prop.c (ipa_free_all_edge_args): Remove.
15125 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
15127 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
15129 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
15130 (fnma<mode>4): Likewise.
15131 (fms<mode>4): Likewise.
15132 (fnms<mode>4): Likewise.
15133 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
15134 (aarch64_fnma<mode>4): Likewise.
15135 (aarch64_fms<mode>4): Likewise.
15136 (aarch64_fnms<mode>4): Likewise.
15137 (aarch64_fnmadd<mode>4): Likewise.
15139 2018-05-16 Jason Merrill <jason@redhat.com>
15141 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
15143 2018-05-16 Richard Biener <rguenther@suse.de>
15145 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
15146 (dump_stmt_cost): Declare.
15147 (add_stmt_cost): Dump cost we add.
15148 (add_stmt_costs): New function.
15149 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
15150 No longer exported.
15151 (vect_analyze_stmt): Adjust prototype.
15152 (vectorizable_condition): Likewise.
15153 (vectorizable_live_operation): Likewise.
15154 (vectorizable_reduction): Likewise.
15155 (vectorizable_induction): Likewise.
15156 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
15157 cost vector to pass to vectorizable_ and record afterwards.
15158 (vect_model_reduction_cost): Take cost vector argument and adjust.
15159 (vect_model_induction_cost): Likewise.
15160 (vectorizable_reduction): Likewise.
15161 (vectorizable_induction): Likewise.
15162 (vectorizable_live_operation): Likewise.
15163 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
15164 SLP_TREE_NUMBER_OF_VEC_STMTS.
15165 (vect_analyze_slp_cost_1): Remove.
15166 (vect_analyze_slp_cost): Likewise.
15167 (vect_slp_analyze_node_operations): Take visited args and
15168 a target cost vector. Avoid processing already visited stmt sets.
15169 (vect_slp_analyze_operations): Use a local cost vector to gather
15170 costs and register those of non-discarded instances.
15171 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
15172 (vect_schedule_slp_instance): Remove copying of
15173 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
15175 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
15176 adding cost. Record cost entry location.
15177 (vect_prologue_cost_for_slp_op): Function to compute cost of
15178 a constant or invariant generated for SLP vect in the prologue,
15179 split out from vect_analyze_slp_cost_1.
15180 (vect_model_simple_cost): Make static. Adjust for SLP costing.
15181 (vect_model_promotion_demotion_cost): Likewise.
15182 (vect_model_store_cost): Likewise, make static.
15183 (vect_model_load_cost): Likewise.
15184 (vectorizable_bswap): Add cost vector arg and adjust.
15185 (vectorizable_call): Likewise.
15186 (vectorizable_simd_clone_call): Likewise.
15187 (vectorizable_conversion): Likewise.
15188 (vectorizable_assignment): Likewise.
15189 (vectorizable_shift): Likewise.
15190 (vectorizable_operation): Likewise.
15191 (vectorizable_store): Likewise.
15192 (vectorizable_load): Likewise.
15193 (vectorizable_condition): Likewise.
15194 (vectorizable_comparison): Likewise.
15195 (can_vectorize_live_stmts): Likewise.
15196 (vect_analyze_stmt): Likewise.
15197 (vect_transform_stmt): Adjust calls to vectorizable_*.
15198 * tree-vectorizer.c: Include gimple-pretty-print.h.
15199 (dump_stmt_cost): New function.
15201 2018-05-16 Richard Biener <rguenther@suse.de>
15203 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
15204 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
15205 * tree-ssa-dse.c: Include tree-ssa-loop.h.
15206 (check_name): New callback.
15207 (dse_classify_store): Track cycles via a visited bitmap of PHI
15208 defs and simplify handling of in-loop and across loop dead stores
15209 and properly fail for loop-variant refs. Handle byte-tracking with
15210 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
15213 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
15215 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
15216 (vect_get_mask_type_for_stmt): Likewise.
15217 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
15219 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
15220 to determine the statement's vector type and the vector type that
15221 should be used for calculating nunits. Deal with cases in which
15222 the type has to be deferred.
15223 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
15224 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
15225 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
15226 (vect_determine_vf_for_stmt): New functions, split out from...
15227 (vect_determine_vectorization_factor): ...here.
15228 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
15229 (vect_get_mask_type_for_stmt): New functions, split out from
15230 vect_determine_vectorization_factor.
15232 2018-05-16 Richard Biener <rguenther@suse.de>
15234 * tree-cfg.c (verify_gimple_assign_ternary): Properly
15235 verify the [VEC_]COND_EXPR embedded comparison.
15237 2018-05-15 Martin Sebor <msebor@redhat.com>
15239 PR tree-optimization/85753
15240 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
15241 RECORD_TYPE in addition to ARRAY_TYPE.
15243 2018-05-15 Martin Sebor <msebor@redhat.com>
15245 PR middle-end/85643
15246 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
15248 2018-05-15 Richard Biener <rguenther@suse.de>
15250 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
15251 add by_clobber_p one. Change algorithm to collect all defs
15252 representing uses we need to walk and try reducing them to
15253 a single one before failing.
15254 (dse_dom_walker::dse_optimize_stmt): Adjust.
15256 2018-05-13 Mark Wielaard <mark@klomp.org>
15258 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
15259 (size_of_loc_descr): Likewise.
15260 (output_loc_operands): Likewise.
15261 (output_loc_operands_raw): Likewise.
15262 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
15263 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
15264 (hash_loc_operands): Likewise.
15265 (compare_loc_operands): Likewise.
15267 2018-05-14 Mark Wielaard <mark@klomp.org>
15269 * dwarf2out.c (count_index_addrs): New function.
15270 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
15272 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
15274 PR tree-optimization/83648
15275 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
15276 return value as malloc candidate.
15278 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
15281 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite
15282 param as true in call to suggest_attribute.
15284 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
15286 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
15287 -mreadonly-in-sdata.
15289 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15291 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
15293 (aarch64_crypto_aesd_fused): Likewise.
15295 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
15297 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
15298 (movsi_aarch64): Likewise.
15299 (load_pairsi): Likewise.
15300 (load_pairdi): Likewise.
15301 (store_pairsi): Likewise.
15302 (store_pairdi): Likewise.
15303 (load_pairsf): Likewise.
15304 (load_pairdf): Likewise.
15305 (store_pairsf): Likewise.
15306 (store_pairdf): Likewise.
15307 (zero_extend): Likewise.
15308 (trunc): Swap alternatives.
15309 (fcvt_target): Add '?' to prefer w over r.
15311 2018-05-14 Jakub Jelinek <jakub@redhat.com>
15314 * config/i386/i386.md: Disallow non-commutative arithmetics in
15315 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
15316 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
15317 in the peephole2 before it.
15319 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
15321 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
15322 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
15323 (ix86_handle_option): Handle -mcldemote.
15324 * config.gcc: New header.
15325 * config/i386/cldemoteintrin.h: New file.
15326 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
15327 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
15329 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
15330 OPTION_MASK_ISA_CLDEMOTE.
15331 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
15332 (ix86_valid_target_attribute_inner_p): Ditto.
15333 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
15334 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
15335 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
15336 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
15337 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
15339 * config/i386/i386.opt: Add -mcldemote.
15340 * config/i386/x86intrin.h: New header.
15341 * doc/invoke.texi: Add -mcldemote.
15343 2018-05-14 Richard Biener <rguenther@suse.de>
15345 * doc/match-and-simplify.texi: Adjust :s documentation.
15347 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
15349 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
15350 intended memcpy size.
15351 (REORDER_45): Likewise.
15353 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
15355 * sort.cc: New file.
15356 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
15357 * vec.c (qsort_chk): Use gcc_qsort.
15358 * Makefile.in (OBJS-libcommon): Add sort.o.
15359 (build/sort.o): New target. Use it...
15360 (BUILD_RTL): ... here, and...
15361 (build/gencfn-macros): ... here, and...
15362 (build/genmatch): ... here.
15364 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
15365 Chung-Ju Wu <jasonwucj@gmail.com>
15367 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
15368 * config/nds32/nds32-graywolf.md: New file.
15369 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
15370 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
15372 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
15373 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
15374 * config/nds32/nds32.md (pipeline_model): Add graywolf.
15375 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
15376 * config/nds32/pipelines.md: Include n15 settings.
15378 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
15379 Chung-Ju Wu <jasonwucj@gmail.com>
15381 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
15382 * config/nds32/nds32-n13.md: New file.
15383 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
15384 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
15386 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
15387 * config/nds32/nds32.md (pipeline_model): Add n13.
15388 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
15389 * config/nds32/pipelines.md: Include n13 settings.
15391 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
15392 Chung-Ju Wu <jasonwucj@gmail.com>
15394 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
15395 * config/nds32/nds32-n10.md: New file.
15396 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
15397 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
15399 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
15400 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
15401 * config/nds32/nds32.md (pipeline_model): Add n10.
15402 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
15403 * config/nds32/pipelines.md: Include n10 settings.
15405 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
15406 Kito Cheng <kito.cheng@gmail.com>
15407 Chung-Ju Wu <jasonwucj@gmail.com>
15409 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
15410 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
15411 Add enum values for DSP extension instructions.
15412 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
15414 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
15415 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
15416 New code iterators.
15417 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
15418 * config/nds32/nds32-dspext.md: New file for DSP implementation.
15419 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
15420 * config/nds32/nds32-intrinsic.md: Likewise.
15421 * config/nds32/nds32_intrinsic.h: Likewise.
15422 * config/nds32/nds32-md-auxiliary.c: Likewise.
15423 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
15424 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
15425 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
15426 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
15427 * config/nds32/nds32-protos.h: New declarations for DSP extension.
15428 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
15429 TYPE_DMAC in switch statement.
15430 * config/nds32/nds32.c: New checking and implementation for DSP
15431 extension instructions.
15432 * config/nds32/nds32.h: Likewise.
15433 * config/nds32/nds32.md: Likewise.
15434 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
15435 * config/nds32/predicates.md: Implement new predicates for DSP
15438 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
15440 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
15441 Reformat alternatives and attributes so it is easier to identify
15442 which constraints/attributes go with which instruction.
15443 (mov<mode>_hardfloat32, FMOVE64): Likewise.
15444 (mov<mode>_softfloat32, FMOVE64): Likewise.
15445 (mov<mode>_hardfloat64, FMOVE64): Likewise.
15446 (mov<mode>_softfloat64, FMOVE64): Likewise.
15448 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
15450 * doc/extend.texi (PowerPC Built-in Functions): Rename this
15452 (Basic PowerPC Built-in Functions): The new name of the
15453 subsection previously known as "PowerPC Built-in Functions".
15454 (Basic PowerPC Built-in Functions Available on all Configurations):
15456 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
15457 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
15458 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
15459 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
15461 2018-05-11 Martin Jambor <mjambor@suse.cz>
15464 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
15467 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
15470 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
15472 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
15474 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
15475 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
15476 (ix86_handle_option): Handle -mwaitpkg.
15477 * config.gcc: New header.
15478 * config/i386/cpuid.h (bit_WAITPKG): New bit.
15479 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
15480 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
15482 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
15483 OPTION_MASK_ISA_WAITPKG.
15484 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
15485 (ix86_option_override_internal): Add PTA_WAITPKG.
15486 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
15487 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
15488 IX86_BUILTIN_TPAUSE.
15489 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
15490 __builtin_ia32_umwait and __builtin_ia32_tpause.
15491 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
15492 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
15493 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
15494 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
15495 UNSPECV_TPAUSE): New.
15496 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
15497 * config/i386/i386.opt: Add -mwaitpkg.
15498 * config/i386/waitpkgintrin.h: New file.
15499 * config/i386/x86intrin.h: New header.
15500 * doc/invoke.texi: Add -mwaitpkg.
15502 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
15505 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
15507 (cortex-m0): Use armv6s-m isa.
15508 (cortex-m0plus): Likewise.
15509 (cortex-m1): Likewise.
15510 (cortex-m0.small-multiply): Likewise.
15511 (cortex-m0plus.small-multiply): Likewise.
15512 (cortex-m1.small-multiply): Likewise.
15514 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
15515 Jakub Jelinek <jakub@redhat.com>
15517 PR tree-optimization/85692
15518 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
15519 source permute as well.
15521 2018-05-11 Martin Liska <mliska@suse.cz>
15524 * doc/extend.texi: Document LLVM style format for no_sanitize
15527 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
15529 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
15530 mode_supports_vsx_dform_quad to mode_supports_dq_form.
15531 (mode_supports_vsx_dform_quad): Likewise.
15532 (mode_supports_vmx_dform): Move these functions to be next to the
15533 other mode_supports functions.
15534 (mode_supports_dq_form): Likewise.
15535 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
15536 mode_supports_dq_form.
15537 (reg_offset_addressing_ok_p): Likewise.
15538 (offsettable_ok_by_alignment): Likewise.
15539 (rs6000_legitimate_offset_address_p): Likewise.
15540 (legitimate_lo_sum_address_p): Likewise.
15541 (rs6000_legitimize_address): Likewise.
15542 (rs6000_legitimize_reload_address): Likewise.
15543 (rs6000_secondary_reload_inner): Likewise.
15544 (rs6000_preferred_reload_class): Likewise.
15545 (rs6000_output_move_128bit): Likewise.
15547 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
15549 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
15550 Generate SImode target register for null target.
15551 <case IX86_BUILTIN_XGETBV>: Ditto.
15552 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
15553 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
15555 2018-05-10 Carl Love <cel@us.ibm.com>
15557 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
15558 dcbtt and dcbtstt if operands[2] is 0.
15560 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
15563 * config/i386/sse.md (usadv64qi): New expander.
15565 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
15567 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
15568 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
15569 -maltivec=be support.
15570 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
15571 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
15572 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
15573 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
15574 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
15575 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
15576 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
15577 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
15578 altivec_vsumsws): Adjust.
15579 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
15580 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
15581 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
15583 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
15584 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
15585 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
15586 (altivec_lve<VI_char>x): Delete expand.
15587 (*altivec_lve<VI_char>x_internal): Rename to...
15588 (altivec_lve<VI_char>x): ... this.
15589 (altivec_lvxl_<mode>): Delete expand.
15590 (*altivec_lvxl_<mode>_internal): Rename to ...
15591 (altivec_lvxl_<mode>): ... this.
15592 (altivec_stvxl_<mode>): Delete expand.
15593 (*altivec_stvxl_<mode>_internal): Rename to ...
15594 (altivec_stvxl_<mode>): ... this.
15595 (altivec_stve<VI_char>x): Delete expand.
15596 (*altivec_stve<VI_char>x_internal): Rename to ...
15597 (altivec_stve<VI_char>x): ... this.
15598 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
15599 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
15600 reduc_plus_scal_<mode>): Adjust.
15601 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
15603 (rs6000_cpu_cpp_builtins): Adjust.
15604 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
15605 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
15606 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
15607 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
15608 -maltivec=be support.
15609 (rs6000_split_vec_extract_var): Adjust.
15610 (rs6000_split_v4si_init): Adjust.
15611 (swap_selector_for_mode): Delete.
15612 (altivec_expand_lvx_be, altivec_expand_stvx_be,
15613 altivec_expand_stvex_be): Delete.
15614 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
15615 -maltivec=be support.
15616 (rs6000_gimple_fold_builtin): Ditto.
15617 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
15619 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
15620 (TARGET_DIRECT_MOVE_64BIT): Adjust.
15621 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
15622 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
15623 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
15624 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
15625 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
15626 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
15627 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
15628 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
15629 anonymous split): Adjust.
15630 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
15631 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
15633 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
15635 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
15636 when --with-gxx-include-dir is also specified.
15637 * configure: Regenerate.
15639 2018-05-09 Jim Wilson <jimw@sifive.com>
15642 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
15643 * config/riscv/t-withmultilib: New.
15644 * config/riscv/withmultilib.h: New.
15645 * doc/install.texi: Document RISC-V --with-multilib-list support.
15647 2018-05-09 Richard Biener <rguenther@suse.de>
15649 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
15651 (vect_bb_vectorization_profitable_p): Adjust. Compute
15652 actual scalar cost using the cost vector and the add_stmt_cost
15655 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
15657 PR rtl-optimization/85645
15658 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
15659 in the REG_CFA_REGISTER note for LR, don't leave it empty.
15661 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
15663 PR rtl-optimization/85645
15664 * shrink-wrap.c (spread_components): Return a boolean saying if
15665 anything was changed.
15666 (try_shrink_wrapping_separate): Iterate spread_components until
15667 nothing changes anymore.
15669 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
15671 PR rtl-optimization/85645
15672 * regrename.c (build_def_use): Also kill the chains that include the
15673 destination of a REG_CFA_REGISTER note.
15675 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
15677 PR rtl-optimization/85645
15678 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
15679 insn that has a REG_CFA_REGISTER note.
15681 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
15683 * cfgexpand.c (expand_clobber): New function.
15684 (expand_gimple_stmt_1): Use it.
15685 * tree-vect-stmts.c (vect_clobber_variable): New function,
15687 (vectorizable_simd_clone_call): ...here.
15688 (vectorizable_store): Emit a clobber either side of an
15689 IFN_STORE_LANES sequence.
15690 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
15692 2018-05-09 Tom de Vries <tom@codesourcery.com>
15695 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
15696 (define_insn "trap_if_false"): Add exit after trap.
15698 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
15700 PR rtl-optimization/85638
15701 * bb-reorder.c: Include common/common-target.h.
15702 (create_forwarder_block): New function extracted from...
15703 (fix_up_crossing_landing_pad): ...here. Rename into...
15704 (dw2_fix_up_crossing_landing_pad): ...this.
15705 (sjlj_fix_up_crossing_landing_pad): New function.
15706 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
15707 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
15708 from both partitions and exit the loop after one iteration.
15710 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
15713 * doc/extend.texi (PowerPC Built-in Functions): Rename this
15715 (Basic PowerPC Built-in Functions): The new name of the
15716 subsection previously known as "PowerPC Built-in Functions".
15717 (Basic PowerPC Built-in Functions Available on all Configurations):
15719 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
15721 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
15723 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
15725 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
15728 2018-05-08 Jim Wilson <jimw@sifive.com>
15730 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
15731 (LD_EMUL_SUFFIX): New.
15732 (LINK_SPEC): Use it.
15734 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
15736 * doc/extend.texi (PowerPC Built-in Functions): Rename this
15738 (Basic PowerPC Built-in Functions): The new name of the
15739 subsection previously known as "PowerPC Built-in Functions".
15740 (Basic PowerPC Built-in Functions Available on all Configurations):
15742 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
15744 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
15746 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
15748 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
15751 2018-05-08 Jakub Jelinek <jakub@redhat.com>
15754 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
15755 after cmpelim optimization.
15757 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
15759 * config.gcc: Support "goldmont".
15760 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
15761 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
15762 PROCESSOR_GOLDMONT.
15763 * config/i386/i386.c (m_GOLDMONT): Define.
15764 (processor_target_table): Add "goldmont".
15765 (PTA_GOLDMONT): Define.
15766 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
15767 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
15768 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
15769 (fold_builtin_cpu): Add "goldmont".
15770 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
15771 (ix86_option_override_internal): Add "goldmont".
15772 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
15773 (processor_type): Add PROCESSOR_GOLDMONT.
15774 * config/i386/i386.md: Add CPU "glm".
15775 * config/i386/glm.md: New file.
15776 * config/i386/x86-tune.def: Add m_GOLDMONT.
15777 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
15779 2018-05-08 Jakub Jelinek <jakub@redhat.com>
15782 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
15784 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
15785 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
15786 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
15789 * config/i386/i386.c (ix86_fold_builtin): Handle
15790 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
15793 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
15794 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
15796 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
15799 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
15800 (check_arch): Likewise.
15801 (check_fpu): Return the result rather than printing it.
15802 (end arch): Fix operator precedence.
15803 (end cpu): Likewise.
15804 (END): Print the result from check_fpu.
15806 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
15807 Alan Hayward <alan.hayward@arm.com>
15808 David Sherwood <david.sherwood@arm.com>
15810 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
15811 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
15812 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
15813 (*fcmuo<mode>_and): New patterns.
15815 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
15817 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
15818 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
15819 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
15820 (cmp_op, sve_imm_con): New code attributes.
15821 (SVE_COND_INT_CMP, imm_con): Delete.
15822 (cmp_op): Remove above unspecs from int attribute.
15823 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
15825 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
15826 comparison-specific unspecs.
15827 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
15828 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
15829 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
15830 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
15831 (*vec_fcm<cmp_op><mode>): Rename to...
15832 (*fcm<cmp_op><mode>): ...this and adjust likewise.
15833 (*vec_fcmuo<mode>): Rename to...
15834 (*fcmuo<mode>): ...this and adjust likewise.
15835 (*pred_fcm<cmp_op><mode>): New pattern.
15836 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
15837 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
15839 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
15841 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
15842 (aarch64_emit_sve_predicated_cond): New function.
15843 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
15844 (aarch64_emit_unspec_cond_or): Replace with...
15845 (aarch64_emit_sve_or_conds): ...this new function. Use
15846 aarch64_emit_sve_ptrue_op for the individual comparisons and
15847 aarch64_emit_binop to OR them together.
15848 (aarch64_emit_inverted_unspec_cond): Replace with...
15849 (aarch64_emit_sve_inverted_cond): ...this new function. Use
15850 aarch64_emit_sve_ptrue_op for the comparison and
15851 aarch64_emit_unop to invert the result.
15852 (aarch64_expand_sve_vec_cmp_float): Update after the above
15853 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
15855 2018-05-07 Nathan Sidwell <nathan@acm.org>
15857 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
15858 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
15859 (Backwards Compatibility): Likewise.
15861 2018-05-07 Luis Machado <luis.machado@linaro.org>
15865 2018-05-07 Luis Machado <luis.machado@linaro.org>
15867 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
15868 <prefetch_dynamic_strides>: New const bool field.
15869 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
15870 prefetch_dynamic_strides.
15871 (exynosm1_prefetch_tune): Likewise.
15872 (thunderxt88_prefetch_tune): Likewise.
15873 (thunderx_prefetch_tune): Likewise.
15874 (thunderx2t99_prefetch_tune): Likewise.
15875 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
15877 (aarch64_override_options_internal): Update to set
15878 PARAM_PREFETCH_DYNAMIC_STRIDES.
15879 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
15880 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
15881 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
15882 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
15883 prefetch-dynamic-strides setting.
15885 2018-05-07 Luis Machado <luis.machado@linaro.org>
15887 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
15888 <minimum_stride>: New const int field.
15889 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
15890 minimum_stride field.
15891 (exynosm1_prefetch_tune): Likewise.
15892 (thunderxt88_prefetch_tune): Likewise.
15893 (thunderx_prefetch_tune): Likewise.
15894 (thunderx2t99_prefetch_tune): Likewise.
15895 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
15896 (aarch64_override_options_internal): Update to set
15897 PARAM_PREFETCH_MINIMUM_STRIDE.
15898 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
15899 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
15900 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
15901 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
15902 stride is constant and is below the minimum stride threshold.
15904 2018-05-07 Luis Machado <luis.machado@linaro.org>
15906 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
15909 2018-05-07 Luis Machado <luis.machado@linaro.org>
15911 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
15912 <prefetch_dynamic_strides>: New const bool field.
15913 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
15914 prefetch_dynamic_strides.
15915 (exynosm1_prefetch_tune): Likewise.
15916 (thunderxt88_prefetch_tune): Likewise.
15917 (thunderx_prefetch_tune): Likewise.
15918 (thunderx2t99_prefetch_tune): Likewise.
15919 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
15921 (aarch64_override_options_internal): Update to set
15922 PARAM_PREFETCH_DYNAMIC_STRIDES.
15923 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
15924 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
15925 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
15926 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
15927 prefetch-dynamic-strides setting.
15929 2018-05-07 Luis Machado <luis.machado@linaro.org>
15931 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
15932 <minimum_stride>: New const int field.
15933 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
15934 minimum_stride field.
15935 (exynosm1_prefetch_tune): Likewise.
15936 (thunderxt88_prefetch_tune): Likewise.
15937 (thunderx_prefetch_tune): Likewise.
15938 (thunderx2t99_prefetch_tune): Likewise.
15939 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
15940 (aarch64_override_options_internal): Update to set
15941 PARAM_PREFETCH_MINIMUM_STRIDE.
15942 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
15943 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
15944 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
15945 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
15946 stride is constant and is below the minimum stride threshold.
15948 2018-05-06 Jakub Jelinek <jakub@redhat.com>
15951 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
15952 the type is addressable. Don't force op into register if it has
15955 2018-05-05 Roland McGrath <mcgrathr@google.com>
15958 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
15959 any section for which we don't know a specific type it should have,
15960 regardless of name. Previously this was done only for the exact
15961 names ".init_array", ".fini_array", and ".preinit_array".
15962 (default_elf_asm_named_section): Add comment about
15963 relationship with default_section_type_flags and SECTION_NOTYPE.
15964 (get_section): Don't consider it a type conflict if one side has
15965 SECTION_NOTYPE and the other doesn't, as long as neither has the
15966 SECTION_BSS et al used in the default_section_type_flags logic.
15968 2018-05-05 Tom de Vries <tom@codesourcery.com>
15971 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
15972 (workaround_barsyncs): New function.
15973 (nvptx_reorg): Use workaround_barsyncs.
15974 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
15975 (define_expand "nvptx_membar_cta"): New define_expand.
15976 (define_insn "*nvptx_membar_cta"): New insn.
15978 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
15980 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
15981 To improve optimization opportunities.
15982 * builtin-types.def: The new needed builtin types for the above.
15984 2018-05-04 Richard Biener <rguenther@suse.de>
15986 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
15987 * gimple-ssa-store-merging.c
15988 (imm_store_chain_info::output_merged_store): Remove redundant create,
15989 release split_store vector contents on failure.
15990 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
15991 scalar stmt vector on cache hit.
15993 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
15995 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
15997 * config.gcc (powerpc-xilinx-eabi*): Remove.
15998 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
16000 (fusion_addis_mem_combo_load): Ditto.
16001 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
16003 (rs6000_cpu_cpp_builtins): Ditto.
16004 * config/rs6000/rs6000-linux.c
16005 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
16006 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
16007 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
16009 (rs6000_setup_reg_addr_masks): Ditto.
16010 (rs6000_init_hard_regno_mode_ok): Ditto.
16011 (rs6000_option_override_internal): Ditto.
16012 (legitimate_lo_sum_address_p): Ditto.
16013 (rs6000_legitimize_address): Ditto.
16014 (rs6000_legitimize_reload_address): Ditto.
16015 (rs6000_legitimate_address_p): Ditto.
16016 (abi_v4_pass_in_fpr): Ditto.
16017 (setup_incoming_varargs): Ditto.
16018 (rs6000_gimplify_va_arg): Ditto.
16019 (rs6000_split_multireg_move): Ditto.
16020 (rs6000_savres_strategy): Ditto.
16021 (rs6000_emit_prologue_components): Ditto.
16022 (rs6000_emit_epilogue_components): Ditto.
16023 (rs6000_emit_prologue): Ditto.
16024 (rs6000_emit_epilogue): Ditto.
16025 (rs6000_elf_file_end): Ditto.
16026 (rs6000_function_value): Ditto.
16027 (rs6000_libcall_value): Ditto.
16028 * config/rs6000/rs6000.h: Ditto.
16029 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
16030 (TARGET_MINMAX): ... this. New.
16031 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
16032 * config/rs6000/rs6000.md: Remove Xilinx FP support.
16033 (*movsi_internal1_single): Delete.
16034 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
16035 mfpu=, mxilinx-fpu): Delete.
16036 * config/rs6000/singlefp.h: Delete.
16037 * config/rs6000/sysv4.h: Remove Xilinx FP support.
16038 * config/rs6000/t-rs6000: Ditto.
16039 * config/rs6000/t-xilinx: Delete.
16040 * config/rs6000/titan.md: Adjust for fp_type removal.
16041 * config/rs6000/vsx.md: Remove Xilinx FP support.
16042 (VStype_simple): Delete.
16043 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
16044 * config/rs6000/xfpu.h: Delete.
16045 * config/rs6000/xfpu.md: Delete.
16046 * config/rs6000/xilinx.h: Delete.
16047 * config/rs6000/xilinx.opt: Delete.
16048 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove
16049 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
16051 2018-05-04 Tom de Vries <tom@codesourcery.com>
16054 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
16057 2018-05-04 Richard Biener <rguenther@suse.de>
16059 PR middle-end/85627
16060 * tree-complex.c (update_complex_assignment): We are always in SSA form.
16061 (expand_complex_div_wide): Likewise.
16062 (expand_complex_operations_1): Likewise.
16063 (expand_complex_libcall): Preserve EH info of the original stmt.
16064 (tree_lower_complex): Handle removed blocks.
16065 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
16066 on complex multiplication and division libcall builtins.
16068 2018-05-04 Richard Biener <rguenther@suse.de>
16070 PR middle-end/85574
16071 * fold-const.c (negate_expr_p): Restrict negation of operand
16072 zero of a division to when we know that can happen without
16074 (fold_negate_expr_1): Likewise.
16076 2018-05-04 Jakub Jelinek <jakub@redhat.com>
16079 * real.h (real_nextafter): Declare.
16080 * real.c (real_nextafter): New function.
16081 * fold-const-call.c (fold_const_nextafter): New function.
16082 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
16083 CASE_CFN_NEXTTOWARD.
16084 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
16085 even when arg1_mode is different from arg0_mode.
16087 2018-05-03 Nathan Sidwell <nathan@acm.org>
16089 * doc/extend.texi (Deprecated Features): Remove
16090 -ffriend-injection.
16091 (Backwards Compatibility): Likewise.
16092 * doc/invoke.texi (C++ Language Options): Likewise.
16093 (C++ Dialect Options): Likewise.
16095 2018-05-03 Jakub Jelinek <jakub@redhat.com>
16098 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
16099 _mm512_mask_mullox_epi64): New intrinsics.
16101 2018-05-03 Tom de Vries <tom@codesourcery.com>
16104 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
16105 dump files): Add offload-tree.
16107 2018-05-03 Richard Biener <rguenther@suse.de>
16109 PR tree-optimization/85615
16110 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
16111 to loops not nested in BBs loop father to avoid creating multi-entry
16114 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
16116 PR tree-optimization/70291
16117 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
16118 arguments. Change return type to tree. Emit libcall as a new
16119 statement rather than replacing existing one when inplace_p is true.
16120 (expand_complex_multiplication_components): New function.
16121 (expand_complex_multiplication): Expand floating-point complex
16122 multiplication using the above.
16123 (expand_complex_division): Rename inner_type parameter to type.
16124 Update expand_complex_libcall call-site.
16125 (expand_complex_operations_1): Update expand_complex_multiplication
16126 and expand_complex_division call-sites.
16128 2018-05-02 Jakub Jelinek <jakub@redhat.com>
16131 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
16132 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
16133 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
16134 the highest significant bit of the shift count mask is clear. In
16135 check whether and[sq]i3 is needed verify that all significant bits
16136 of the shift count other than the highest are set.
16138 2018-05-02 Tom de Vries <tom@codesourcery.com>
16141 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
16142 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
16143 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
16144 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
16145 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
16146 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
16147 __builtin_goacc_parlevel_size.
16149 2018-05-02 Richard Biener <rguenther@suse.de>
16151 PR tree-optimization/85597
16152 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
16153 do not use split vect_get_vec_defs call but call vect_get_slp_defs
16156 2018-05-02 Tom de Vries <tom@codesourcery.com>
16159 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
16160 dump files): Add ltrans-tree.
16162 2018-05-02 Tom de Vries <tom@codesourcery.com>
16165 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
16166 dump files): Add wpa-ipa.
16168 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
16170 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
16171 powerpc*-*-linux*paired* target.
16172 * config/rs6000/750cl.h: Delete.
16173 * config/rs6000/paired.h: Delete.
16174 * config/rs6000/paired.md: Delete.
16175 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
16177 * config/rs6000/rs6000-builtin.def: Remove paired float support.
16178 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
16179 comment. Remove paired float support.
16180 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
16181 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
16183 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
16184 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
16186 * config/rs6000/rs6000.c: Remove paired float support.
16187 (paired_expand_vector_init, paired_expand_vector_move,
16188 paired_emit_vector_compare, paired_emit_vector_cond_expr,
16189 (paired_expand_lv_builtin, paired_expand_stv_builtin,
16190 paired_expand_builtin, paired_expand_predicate_builtin,
16191 paired_init_builtins): Delete.
16192 * config/rs6000/rs6000.h: Remove paired float support.
16193 * config/rs6000/rs6000.md: Remove paired float support.
16194 (move_from_CR_ov_bit): Delete.
16195 * config/rs6000/rs6000.opt (mpaired): Delete.
16196 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
16197 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
16199 2018-05-02 Richard Biener <rguenther@suse.de>
16201 PR middle-end/85567
16202 * gimplify.c (gimplify_save_expr): When in SSA form allow
16203 SAVE_EXPRs to compute to SSA vars.
16205 2018-05-02 Jakub Jelinek <jakub@redhat.com>
16208 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
16209 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
16210 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
16211 clobber operands[2], instead use a new pseudo. Formatting fixes.
16213 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
16215 PR tree-optimization/85586
16216 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
16217 exit early for statements in the same group if the accesses are
16220 2018-05-02 Tom de Vries <tom@codesourcery.com>
16223 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
16226 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
16228 PR tree-optimization/85143
16229 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
16231 2018-05-01 Tom de Vries <tom@codesourcery.com>
16234 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
16235 not found" error message.
16237 2018-05-01 Tom de Vries <tom@codesourcery.com>
16240 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
16241 * vec.c (test_ordered_remove_if): New function.
16242 (vec_c_tests): Call test_ordered_remove_if.
16243 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
16244 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
16245 * tree-vect-patterns.c (vect_pattern_recog_1): Use
16246 VEC_ORDERED_REMOVE_IF.
16248 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
16250 PR tree-optimization/82665
16251 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
16252 pointer subtraction where arguments come from a memchr call.
16254 2018-05-01 Jakub Jelinek <jakub@redhat.com>
16256 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
16257 --push-state --as-needed and --pop-state instead of --as-needed and
16258 --no-as-needed if ld supports it.
16259 * configure: Regenerated.
16262 * doc/install.texi2html: Replace _002d with - and _002a with * in
16263 generated html files using sed.
16265 2018-04-30 David Malcolm <dmalcolm@redhat.com>
16268 * gcc-rich-location.c (blank_line_before_p): New function.
16269 (use_new_line): New function.
16270 (gcc_rich_location::add_fixit_insert_formatted): New function.
16271 * gcc-rich-location.h
16272 (gcc_rich_location::add_fixit_insert_formatted): New function.
16274 2018-04-30 David Malcolm <dmalcolm@redhat.com>
16276 * selftest.c (assert_streq): Rename "expected" and "actual" to
16277 "val1" and "val2". Extend NULL-handling to cover both inputs
16278 symmetrically, while still requiring both to be non-NULL for a pass.
16279 * selftest.h (assert_streq): Rename "expected" and "actual" to
16281 (ASSERT_EQ): Likewise.
16282 (ASSERT_EQ_AT): Likewise.
16283 (ASSERT_KNOWN_EQ): Likewise.
16284 (ASSERT_KNOWN_EQ_AT): Likewise.
16285 (ASSERT_NE): Likewise.
16286 (ASSERT_MAYBE_NE): Likewise.
16287 (ASSERT_MAYBE_NE_AT): Likewise.
16288 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
16289 the assertion to pass.
16290 (ASSERT_STREQ_AT): Likewise.
16292 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
16294 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
16295 interaction with -pie.
16297 2018-04-30 David Malcolm <dmalcolm@redhat.com>
16299 * selftest.h: Fix alphabetization of per-source-file selftest
16302 2018-04-30 Jason Merrill <jason@redhat.com>
16304 PR c++/61982 - dead stores to destroyed objects.
16305 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
16308 2018-04-30 Jason Merrill <jason@redhat.com>
16310 * tree.c (build_clobber): New.
16311 * tree.h: Declare it.
16312 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
16314 2018-04-30 David Malcolm <dmalcolm@redhat.com>
16316 * diagnostic-show-locus.c (layout::layout): Update for
16317 location_get_source_line returning a char_span.
16318 (struct char_span): Move to input.h.
16319 (struct correction): Update for fields in char_span becoming
16321 (struct source_line): Update for location_get_source_line
16322 returning a char_span.
16323 (layout::print_line): Likewise.
16324 * edit-context.c (edited_file::print_content): Likewise.
16325 (edited_file::print_diff_hunk): Likewise.
16326 (edited_file::print_run_of_changed_lines): Likewise.
16327 (edited_file::get_num_lines): Likewise.
16328 (edited_line::edited_line): Likewise.
16329 * final.c (asm_show_source): Likewise.
16330 * input.c (location_get_source_line): Convert return type
16331 from const char * to char_span, losing the final "line_len"
16333 (dump_location_info): Update for the above.
16334 (get_substring_ranges_for_loc): Likewise. Use a char_span
16335 when handling the literal within the line.
16336 (test_reading_source_line): Update for location_get_source_line
16337 returning a char_span.
16338 * input.h (class char_span): Move here from
16339 diagnostic-show-locus.c, converting from a struct to a class.
16340 Make data members private.
16341 (char_span::operator bool): New.
16342 (char_span::length): New.
16343 (char_span::get_buffer): New.
16344 (char_span::operator[]): New.
16345 (char_span::subspan): Make const.
16346 (char_span::xstrdup): New.
16347 (location_get_source_line): Convert return type from const char *
16348 to char_span, losing the final "line_size" param.
16350 2018-04-30 Jan Hubicka <jh@suse.cz>
16352 * lto-wrapper.c (ltrans_priorities): New static var.
16353 (cmp_priority): New.
16354 (run_gcc): Read priorities and if doing parallel build order
16355 the Makefile by them.
16357 2018-04-30 David Malcolm <dmalcolm@redhat.com>
16359 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
16361 2018-04-30 Richard Biener <rguenther@suse.de>
16363 * tree-cfg.c (verify_address): Remove base argument, add
16364 flag whether to check TREE_ADDRESSABLE and do that.
16365 (verify_expr): Remove.
16366 (verify_types_in_gimple_reference): Add pieces from verify_expr.
16367 (verify_gimple_assign_single): Likewise.
16368 (verify_gimple_switch): Likewise.
16369 (verify_expr_location_1): Dereference tp once. Add (disabled)
16370 piece from verify_expr.
16371 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
16373 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
16375 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
16377 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
16379 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
16380 (small_data_pattern): Likewise.
16381 (arc_rewrite_small_data): Likewise.
16382 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
16383 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
16384 (get_symbol_alignment): New function.
16385 (legitimate_small_data_address_p): Likewise.
16386 (legitimate_scaled_address): Update, call
16387 legitimate_small_data_address_p.
16388 (output_sdata): New static variable.
16389 (arc_print_operand): Update how we handle small data operands.
16390 (arc_print_operand_address): Likewise.
16391 (arc_legitimate_address_p): Update, use
16392 legitimate_small_data_address_p.
16393 (arc_rewrite_small_data_p): Remove.
16394 (arc_rewrite_small_data_1): Likewise.
16395 (arc_rewrite_small_data): Likewise.
16396 (small_data_pattern): Likewise.
16397 (compact_sda_memory_operand): Update to use
16398 legitimate_small_data_address_p and get_symbol_alignment.
16399 (prepare_move_operands): Don't rewite sdata pattern.
16400 (prepare_extend_operands): Remove.
16401 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
16403 (zero_extendqisi2): Likewise.
16404 (zero_extendhisi2): Likewise.
16405 (extendqihi2): Likewise.
16406 (extendqisi2): Likewise.
16407 (extendhisi2): Likewise.
16408 (addsi3): Likewise.
16409 (subsi3): Likewise.
16410 (andsi3): Likewise.
16411 * config/arc/constraints.md (Usd): Change it to memory constraint.
16413 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
16415 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
16416 as source of std instructions.
16417 * config/arc/arc.md (movsi_insn): Update pattern predicate to
16418 allow 6-bit constants as source for store instructions.
16419 (movdi_insn): Update instruction pattern to allow 6-bit constants
16420 as source for store instructions.
16422 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
16424 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
16426 2018-04-30 Nathan Sidwell <nathan@acm.org>
16427 Sandra Loosemore <sandra@codesourcery.com>
16429 * dumpfile.c (dump_open): Allow '-' for stdout.
16430 * doc/invoke.texi (Developer Options): Document dump filename
16431 determination early. Document stdin/stdout selection.
16433 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
16435 Microblaze Target: PIC data text relative
16437 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
16438 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
16440 * config/microblaze/microblaze.h (microblaze_constant_address_p):
16441 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
16442 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
16443 New addressing mode for data-text relative position indepenedent code.
16444 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
16445 'ADDRESS_SYMBOLIC_TXT_REL'.
16446 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
16447 (microblaze_legitimate_pic_operand): Exclude function calls from
16448 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
16449 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
16451 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
16452 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
16453 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
16454 for 'address + offset'.
16455 (microblaze_expand_prologue): Add new function prologue call for
16457 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
16458 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
16459 table in case of TARGET_PIC_DATA_TEXT_REL.
16460 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
16461 * config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
16462 Add new macros 'UNSPEC_TEXT',
16463 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
16464 + exclude function calls from 'UNSPEC_PLT' in case of data text
16466 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
16467 new target hook for generating address diff vector tables in case of
16469 * doc/tm.texi : Regenerate.
16470 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
16471 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
16472 of addr diff vector generation.
16473 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
16474 target hook definition.
16475 * targhooks.h, targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
16476 Add default function for generate_pic_addr_diff_vec -> flag_pic.
16477 * doc/invoke.texi (Add new pic option): Add new microblaze pic
16478 option for data text relative.
16480 2018-04-30 Richard Biener <rguenther@suse.de>
16482 * tree-chrec.h (evolution_function_is_constant_p): Remove
16484 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
16486 2018-04-30 Richard Biener <rguenther@suse.de>
16489 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
16491 2018-04-30 Richard Biener <rguenther@suse.de>
16493 PR tree-optimization/28364
16494 PR tree-optimization/85275
16495 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
16496 copying first exit test.
16498 2018-04-28 Mark Wielaard <mark@klomp.org>
16500 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
16501 dwarf_version >= 5.
16502 (dwarf_AT): Handle DW_AT_addr_base.
16503 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
16505 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
16508 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
16509 (*ashl<dwi>3_doubleword_mask_1): Ditto.
16510 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
16511 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
16513 2018-04-28 Richard Biener <rguenther@suse.de>
16515 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
16516 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
16517 to reflect use. Only add interesting stmts.
16519 2018-04-27 Martin Jambor <mjambor@suse.cz>
16522 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
16523 the jump function allows for passing through aggregate values.
16525 2018-04-27 David Malcolm <dmalcolm@redhat.com>
16527 * input.h (in_system_header_at): Convert from macro to inline
16529 (from_macro_expansion_at): Likewise.
16530 (from_macro_definition_at): Likewise.
16532 2018-04-27 Jeff Law <law@redhat.com>
16534 * config.gcc: Mark tile* targets as deprecated/obsolete.
16536 2018-04-27 Richard Biener <rguenther@suse.de>
16538 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
16541 2018-04-27 Richard Biener <rguenther@suse.de>
16543 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
16545 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
16547 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
16548 with Yd constraint. Set "preferred_for_speed" attribute from
16549 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
16550 with Yd constraint.
16551 (*movdi_internal): Ditto.
16552 (movti_interunit splitters): Remove
16553 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
16554 (movdi_interunit splitters): Ditto.
16555 * config/i386/constraints.md (Ye): Remove.
16556 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
16558 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
16561 * config/aarch64/constraints.md (Usg): Limit to 31.
16562 (Usj): Limit to 63.
16564 2018-04-27 Jakub Jelinek <jakub@redhat.com>
16566 PR tree-optimization/85529
16567 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
16568 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
16569 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
16570 zero extension or masking of the MSB bit.
16571 (optimize_range_tests): Add FIRST_BB argument, pass it through
16572 to optimize_range_tests_var_bound.
16573 (maybe_optimize_range_tests, reassociate_bb): Adjust
16574 optimize_range_tests callers.
16576 2018-04-26 Richard Biener <rguenther@suse.de>
16577 Jakub Jelinek <jakub@redhat.com>
16579 * cgraph.h (symbol_table): Just declare debug method here.
16580 * symtab.c (symbol_table::debug): Define.
16582 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
16584 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
16586 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
16588 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
16589 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
16590 (*movdi_internal): Substitute Yi and Yj constraint with x
16591 and Ym and Yn constraint with y constraint. Update "isa"
16592 attribute and set "preferred_for_speed" attribute from
16593 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
16594 (*movsi_internal): Ditto.
16595 (*movdf_internal): Ditto.
16596 (*movsf_internal): Ditto.
16597 (*zero_extendsidi2): Ditto.
16598 * config/i386/sse.md (vec_set<mode>_0): Ditto.
16599 (sse2_loadld): Ditto.
16600 (*vec_extract<ssevecmodelower>_0): Ditto.
16601 (*vec_extractv4si_0_zext_sse4): Ditto.
16602 (vec_concatv2di): Ditto.
16603 (*vec_dup<mode>): Ditto.
16604 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
16605 * config/i386/constraints.md (Yi): Remove.
16610 2018-04-26 Nathan Sidwell <nathan@acm.org>
16612 * dumpfile.c (dump_open): New.
16613 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
16614 (dump_finish): Detect stdio/stderr by value not name.
16616 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
16618 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
16620 2018-04-26 Tom de Vries <tom@codesourcery.com>
16623 * config/nvptx/nvptx.c (verify_neutering_jumps)
16624 (verify_neutering_labels): New function
16625 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
16627 2018-04-26 Tom de Vries <tom@codesourcery.com>
16630 * config/nvptx/nvptx.c (needs_neutering_p): New function.
16631 (nvptx_single): Use needs_neutering_p to skip over insns that do not
16634 2018-04-26 Richard Biener <rguenther@suse.de>
16635 Tom de Vries <tom@codesourcery.com>
16638 * lto-streamer-out.c (output_function): Fixup loops if required to match
16639 discovery done in the reader.
16641 2018-04-26 Richard Biener <rguenther@suse.de>
16643 PR tree-optimization/85116
16644 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
16645 have a loop exit from the single latch predecessor. Remove
16646 case of header with just condition.
16647 (ch_base::copy_headers): Exclude infinite loops from any
16649 (pass_ch::execute): Record exits.
16651 2018-04-26 Richard Biener <rguenther@suse.de>
16653 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
16654 prologue cost vector and pass it to vect_get_load_cost.
16655 (vect_get_peeling_costs_all_drs): Likewise.
16656 (vect_peeling_hash_get_lowest_cost): Likewise.
16657 (vect_enhance_data_refs_alignment): Likewise.
16659 2018-04-26 Richard Biener <rguenther@suse.de>
16661 PR middle-end/85450
16662 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
16663 checking of integer<->pointer conversions.
16664 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
16665 sign-/zero-extending pointer types.
16666 (expand_omp_for_static_chunk): Likewise.
16668 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
16669 Jean Lee <xiaoyur347@gmail.com>
16671 * config/mips/mips.c (mips_asan_shadow_offset): New function.
16672 (TARGET_ASAN_SHADOW_OFFSET): Define.
16673 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
16674 true for -fsanitize=address.
16676 2018-04-25 Mark Wielaard <mark@klomp.org>
16678 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
16681 2018-04-25 Jakub Jelinek <jakub@redhat.com>
16683 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
16684 than "alu", remove explicit "memory" and "imm_disp" attributes.
16685 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
16687 PR middle-end/85414
16688 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
16689 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
16690 gen_lowpart_no_emit.
16692 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
16695 * config/i386/i386.c (ix86_expand_builtin): Change memory
16696 operand to XI, extend p0 to Pmode.
16697 * config/i386/i386.md: Change unspec volatile and operand
16698 1 mode to XI, change operand 0 mode to P.
16700 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
16702 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
16703 GET_MODE_MASK before any checking.
16704 (nds32_can_use_bset_p): Likewise.
16705 (nds32_can_use_btgl_p): Likewise.
16707 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
16709 * config/nds32/nds32-doubleword.md: New define_split pattern for
16710 illegal register number.
16712 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
16714 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
16716 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
16718 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
16720 2018-04-25 Richard Biener <rguenther@suse.de>
16722 * lto-streamer.h (LTO_major_version): Bump to 8.
16724 2018-04-25 Jakub Jelinek <jakub@redhat.com>
16726 * BASE-VER: Set to 9.0.0.
16728 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
16730 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
16731 in __abskf2 and __powikf2.
16733 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
16736 * config/aarch64/constraints.md (Usg, Usj): New constraints.
16737 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
16738 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
16739 Use the above on operand 2. Reindent.
16740 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
16742 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
16745 * common/config/i386/i386-common.c (ix86_handle_option): Don't
16747 * config/i386/i386.opt (mcet): Removed.
16748 * doc/install.texi: Remove -mcet documentation.
16749 * doc/invoke.texi: Likewise.
16751 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
16754 * doc/install.texi: Remove -mcet from bootstrap-cet.
16756 2018-04-24 Jakub Jelinek <jakub@redhat.com>
16759 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
16760 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
16764 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
16765 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
16766 containing a CONST_VECTOR.
16768 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
16770 * doc/install.texi: Update newlib dependency for nvptx.
16772 2018-04-24 Jakub Jelinek <jakub@redhat.com>
16775 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
16776 instead of INTVAL when shifting x left.
16778 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
16780 PR tree-optimization/85478
16781 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
16782 vect_grouped_store_supported for single element vectors.
16784 2018-04-24 Richard Biener <rguenther@suse.de>
16787 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
16788 load cost increase to the case of non-constant step.
16790 2018-04-24 Jakub Jelinek <jakub@redhat.com>
16793 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
16794 destination if any_malformed_asm.
16796 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
16798 PR middle-end/85496
16799 * expr.c (store_field): In the bitfield case, if the value comes from
16800 a function call and is returned in registers by means of a PARALLEL,
16801 do not change the mode of the temporary unless BLKmode and VOIDmode.
16803 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
16805 PR rtl-optimization/85423
16806 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
16807 dependencies to debug insns when the previous insn is non-debug.
16809 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
16811 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
16812 enums into a single definition.
16813 (fls): Fix predicates and printing.
16816 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
16818 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
16819 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
16820 and short u6 immediate.
16821 (check_if_valid_sleep_operand): Remove.
16822 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
16824 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
16826 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
16827 flag_always_save_lp condition.
16828 * config/nds32/nds32.opt (malways-save-lp): New option.
16830 2018-04-22 Shiva Chen <shiva0217@gmail.com>
16832 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
16833 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
16834 * config/nds32/nds32.h
16835 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
16836 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
16838 2018-04-22 Shiva Chen <shiva0217@gmail.com>
16840 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
16841 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
16843 2018-04-22 Shiva Chen <shiva0217@gmail.com>
16844 Chung-Ju Wu <jasonwucj@gmail.com>
16846 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
16848 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
16849 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
16851 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
16853 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
16855 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
16857 * config/nds32/nds32-protos.h (nds32_data_alignment,
16858 nds32_local_alignment): Declare.
16859 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
16860 nds32_local_alignment): New functions.
16861 (TARGET_CONSTANT_ALIGNMENT): Define.
16862 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
16864 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
16866 * config/nds32/nds32.c
16867 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
16868 (TARGET_MODES_TIEABLE_P): Likewise.
16870 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
16872 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
16873 level Ofast and Og.
16875 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
16876 Chung-Ju Wu <jasonwucj@gmail.com>
16878 * config/nds32/constants.md (unspec_volatile_element): Add enum values
16879 for unaligned access.
16880 * config/nds32/nds32-intrinsic.c: Implementation of expanding
16882 * config/nds32/nds32-intrinsic.md: Likewise.
16883 * config/nds32/nds32_intrinsic.h: Likewise.
16884 * config/nds32/nds32.h (nds32_builtins): Likewise.
16885 * config/nds32/nds32.opt (munaligned-access): New option.
16886 * config/nds32/nds32.c (nds32_asm_file_start): Display
16887 flag_unaligned_access status.
16889 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
16891 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
16892 -mno-relax is present.
16893 * config/riscv/linux.h (LINK_SPEC): Ditto.
16895 2018-04-20 Martin Sebor <msebor@redhat.com>
16898 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
16900 (gimple_fold_builtin_stxcpy_chk): Same.
16901 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
16903 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
16906 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
16907 __powikf2 when long double is IEEE 128-bit.
16909 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
16911 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
16912 step to make sure stack always aligned.
16914 2018-04-20 Carl Love <cel@us.ibm.com>
16917 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
16918 size check for arg0.
16920 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
16921 Tom de Vries <tom@codesourcery.com>
16924 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
16925 Emit insns for calls too.
16926 (nvptx_find_par): Always look for worker-level predecessor insn.
16927 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
16929 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
16930 (nvptx_process_pars): Propagate frames for calls.
16932 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
16935 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
16937 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
16938 (ix86_handle_option): Don't handle OPT_mibt.
16939 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
16941 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
16943 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
16944 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
16945 (ix86_target_macros): Define __CET__ with flag_cf_protection
16946 for -fcf-protection.
16947 * config/i386/i386.c (isa2_opts): Remove -mibt.
16948 * config/i386/i386.h (TARGET_IBT): Removed.
16949 (TARGET_IBT_P): Likewise.
16950 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
16951 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
16952 * config/i386/i386.opt (mcet): Update help message.
16953 (mshstk): Likewise.
16955 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
16956 -mcet as an alias for -mshstk.
16958 2018-04-20 Richard Biener <rguenther@suse.de>
16960 PR middle-end/85475
16961 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
16962 complexity by forcing a single use of the multiply operand.
16964 2018-04-20 Martin Jambor <mjambor@suse.cz>
16967 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
16968 recursion dependency to only apply to non-clones.
16970 2018-04-20 Martin Jambor <mjambor@suse.cz>
16973 * ipa-cp.c (create_specialized_node): Check that clones of
16974 self-recursive edges exist during IPA-CP.
16976 2018-04-19 Toon Moene <toon@moene.org>
16978 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
16981 2018-04-19 Jakub Jelinek <jakub@redhat.com>
16983 PR tree-optimization/85467
16984 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
16985 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
16986 VECTOR_CST element to type.
16988 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
16991 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
16992 * config/i386/i386.md (builtin_setjmp_setup): Removed.
16993 (builtin_longjmp): Likewise.
16994 (save_stack_nonlocal): New pattern.
16995 (restore_stack_nonlocal): Likewise.
16997 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
17000 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
17001 Replace ASM_OUTPUT_LABEL with fprintf.
17003 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
17006 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
17007 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
17008 * config/i386/i386-c.c (ix86_target_macros_internal): Also
17009 define __IBT__ and __SHSTK__ for -fcf-protection.
17010 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
17012 (ix86_trampoline_init): Likewise.
17013 (x86_output_mi_thunk): Likewise.
17014 (ix86_notrack_prefixed_insn_p): Likewise.
17015 (ix86_option_override_internal): Don't disallow -fcf-protection.
17016 * config/i386/i386.md (rdssp<mode>): Also enable for
17018 (incssp<mode>): Likewise.
17019 (nop_endbr): Likewise.
17020 * config/i386/i386.opt (mcet): Change help message to built-in
17023 (mshstk): Likewise.
17024 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
17025 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
17026 enable CET built-in functions.
17028 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
17030 * common/config/i386/i386-common.c
17031 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
17032 OPTION_MASK_ISA_MOVDIRI_UNSET,
17033 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
17034 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
17035 * config.gcc (movdirintrin.h): New header.
17036 * config/i386/cpuid.h (bit_MOVDIRI,
17037 bit_MOVDIR64B): New bits.
17038 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
17040 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
17041 (VOID, PVOID, PCVOID)): New function types.
17042 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
17043 __builtin_ia32_directstoreu_u64,
17044 __builtin_ia32_movdir64b): New builtins.
17045 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
17046 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
17048 (ix86_valid_target_attribute_inner_p): Ditto.
17049 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
17050 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
17051 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
17052 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
17053 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
17054 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
17055 (movdiri<mode>, movdir64b_<mode>): New.
17056 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
17057 * config/i386/immintrin.h: Include movdirintrin.h.
17058 * config/i386/movdirintrin.h: New file.
17059 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
17061 2018-04-19 Richard Biener <rguenther@suse.de>
17063 PR middle-end/85455
17064 * cfg.c (clear_bb_flags): When loop state says we have
17065 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
17067 2018-04-19 Richard Biener <rguenther@suse.de>
17069 PR tree-optimization/84737
17070 * tree-vect-data-refs.c (vect_copy_ref_info): New function
17071 copying restrict info.
17072 (vect_setup_realignment): Use it.
17073 * tree-vectorizer.h (vect_copy_ref_info): Declare.
17074 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
17075 the first DR to all generated stores.
17076 (vectorizable_load): Likewise for loads.
17078 2018-04-19 Jakub Jelinek <jakub@redhat.com>
17080 PR tree-optimization/85446
17081 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
17082 the integral and pointer types to have the same precision.
17084 * doc/install.texi: Document --disable-cet being the default and
17087 2018-04-18 Martin Liska <mliska@suse.cz>
17089 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
17092 2018-04-18 Martin Liska <mliska@suse.cz>
17095 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
17098 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
17099 arguments if they are comparable.
17101 2018-04-18 Martin Liska <mliska@suse.cz>
17104 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
17107 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
17110 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
17113 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
17114 ENDBR after calling __morestack.
17116 2018-04-18 David Malcolm <dmalcolm@redhat.com>
17119 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
17120 by using gcc_base_ver to generate a gcc_driver_version, and use
17121 it when generating GCC_DRIVER_NAME.
17122 * configure: Regenerate.
17124 2018-04-18 Jakub Jelinek <jakub@redhat.com>
17127 * config.gcc: Obsolete powerpc*-*-*spe*.
17129 2018-04-17 Jakub Jelinek <jakub@redhat.com>
17132 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
17133 (stabstr_D): Change type of unum from unsigned int to
17134 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
17137 2018-04-17 Jim Wilson <jimw@sifive.com>
17140 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
17141 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
17142 Set arg_pointer_offset after using pretend_args_size.
17144 2018-04-17 Jakub Jelinek <jakub@redhat.com>
17146 PR rtl-optimization/85431
17147 * dse.c (record_store): Ignore zero width stores.
17150 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
17151 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
17152 __builtin_stack_restore rather than after it.
17153 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
17154 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
17155 argument instead of virtual_dynamic_stack_rtx.
17157 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
17159 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
17161 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
17162 Add note to error message to explain internal mapping of overloaded
17163 built-in function name to non-overloaded built-in function name.
17164 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
17167 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
17170 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
17171 where the inputs overlap with the output.
17173 2018-04-17 Jakub Jelinek <jakub@redhat.com>
17175 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
17176 (=v, v) alternative and explicit "memory" attribute.
17177 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
17178 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
17180 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
17181 "sselog1" type instead of "sselog".
17182 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
17183 "sselog". Remove explicit "memory" attribute.
17184 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
17185 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
17187 (vec_extract_hi_v32hi): Merge all alternatives into one, use
17188 "sselog1" type instead of "sselog". Remove explicit "memory"
17190 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
17191 use "sselog1" type instead of "sselog". Remove explicit "memory"
17193 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
17194 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
17196 (vec_extract_hi_v64qi): Merge all alternatives into one, use
17197 "sselog1" type instead of "sselog". Remove explicit "memory"
17199 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
17200 use "sselog1" type instead of "sselog". Remove explicit "memory"
17204 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
17206 PR middle-end/85414
17207 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
17210 2018-04-17 Martin Jambor <mjambor@suse.cz>
17213 * ipa-cp.c (create_specialized_node): Call
17214 expand_all_artificial_thunks if necessary.
17216 2018-04-17 Martin Liska <mliska@suse.cz>
17219 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
17220 in message, remote space in between '_G' and '('.
17222 2018-04-17 Jakub Jelinek <jakub@redhat.com>
17225 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
17226 avx512f_vmcmp<mode>3<round_saeonly_name>,
17227 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
17228 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
17229 avx512f_rndscale<mode><round_saeonly_name>,
17230 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
17231 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
17232 Use %<iptr>2 instead of %2 for -masm=intel.
17233 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
17234 avx512f_vcvttss2usi<round_saeonly_name>,
17235 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
17237 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
17238 avx512f_vcvttsd2usi<round_saeonly_name>,
17239 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
17240 Use %q1 instead of %1 for -masm=intel.
17241 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
17242 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
17243 of %3 for -masm=intel.
17244 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
17246 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
17248 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
17250 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
17251 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
17253 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
17255 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
17256 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
17257 %g1 and one with %0 and %1.
17258 (avx512er_vmrcp28<mode><round_saeonly_name>,
17259 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
17260 %1 for -masm=intel.
17261 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
17262 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
17263 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
17264 of %0 and %{%4%} for -masm=intel.
17265 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
17266 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
17267 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
17268 order of %0 and %{%5%}%{z%} for -masm=intel.
17270 2018-04-17 Jan Hubicka <jh@suse.cz>
17273 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
17275 2018-04-17 Martin Liska <mliska@suse.cz>
17278 * multiple_target.c (create_dispatcher_calls): Set apostrophes
17279 for target_clone error message. Make default implementation
17280 clone to be a local declaration.
17281 (separate_attrs): Add new argument and check for an empty
17283 (expand_target_clones): Handle it.
17284 (ipa_target_clone): Make redirection just for target_clones
17287 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
17288 Tom de Vries <tom@codesourcery.com>
17290 PR middle-end/84955
17291 * omp-expand.c (expand_oacc_for): Add dummy false branch for
17292 tiled basic blocks without omp continue statements.
17294 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
17297 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
17298 vec_extract expression as having side effects to make sure it gets
17301 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
17304 * config/i386/i386.c (get_builtin_code_for_version): Check
17307 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
17310 * config.gcc: Support "skylake".
17311 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
17313 * config/i386/i386.c (m_SKYLAKE): Define.
17314 (processor_target_table): Add "skylake".
17315 (ix86_option_override_internal): Add "skylake".
17316 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
17317 PROCESSOR_CANNONLAKE.
17318 (get_builtin_code_for_version): Fix priority for
17319 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
17320 PROCESSOR_SKYLAKE-AVX512.
17321 * config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
17322 (processor_type): Add PROCESSOR_SKYLAKE.
17324 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
17325 Jason Merrill <jason@redhat.com>
17328 * convert.c (convert_to_integer_1): Use direct recursion for
17329 enumeral types and types with a precision less than the number
17330 of bits in their mode.
17332 2018-04-16 Julia Koval <julia.koval@intel.com>
17335 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
17336 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
17338 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
17341 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
17342 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
17343 and -mno-direct-move.
17345 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
17348 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
17349 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
17350 Ensure negative shifts result in {0}.
17352 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
17354 PR rtl-optimization/79916
17355 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
17356 regs (if any) to define how to gnerate SD moves when LRA is in
17359 2018-04-13 Jakub Jelinek <jakub@redhat.com>
17361 PR rtl-optimization/85393
17362 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
17363 * except.c (expand_dw2_landing_pad_for_region): Make static.
17364 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
17365 a label and unconditional jump to old_bb, rather than
17366 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
17369 PR rtl-optimization/85376
17370 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
17371 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
17372 instead of a specific value.
17374 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
17375 Bin Cheng <bin.cheng@arm.com>
17377 PR tree-optimization/82965
17378 PR tree-optimization/83991
17379 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
17380 by_profile_only parameter.
17381 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
17382 information if the loop was predicted to iterate too many times.
17383 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
17385 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
17388 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
17391 2018-04-13 Martin Liska <mliska@suse.cz>
17392 Jakub Jelinek <jakub@redhat.com>
17394 PR middle-end/81657
17395 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
17396 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
17397 * builtins.c (expand_builtin_memory_copy_args): Use
17398 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
17399 handle dest_addr == pc_rtx.
17401 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
17404 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
17405 asked to not generate direct moves.
17406 (fix_trunc<mode>si2_stfiwx): Similar.
17407 (fix_trunc<mode>si2_internal): Similar.
17409 2018-04-12 Jakub Jelinek <jakub@redhat.com>
17412 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
17413 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
17414 lookup if dest in some wider mode is known to be const0_rtx and
17415 if so, record permanent equivalence for it to be ZERO_EXTEND of
17416 the narrower mode destination.
17418 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
17420 * lto-streamer-out.c (output_function): Revert 259346.
17421 * omp-expand.c (expand_oacc_for): Likewise.
17423 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
17425 PR rtl-optimization/85354
17426 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
17427 * sel-sched.c (sel_global_init): ... here.
17429 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
17432 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
17433 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
17434 mode for PE-COFF targets.
17435 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
17436 (i386_pe_asm_lto_end): Likewise.
17437 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
17438 (TARGET_ASM_LTO_END): Likewise.
17439 * config/i386/winnt.c (saved_debug_info_level): New static variable.
17440 (i386_pe_asm_lto_start): New function.
17441 (i386_pe_asm_lto_end): Likewise.
17443 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
17444 Richard Biener <rguenther@suse.de>
17446 PR middle-end/84955
17447 * lto-streamer-out.c (output_function): Fix CFG loop state before
17449 * omp-expand.c (expand_oacc_for): Handle calls to internal
17450 functions like regular functions.
17452 2018-04-12 Richard Biener <rguenther@suse.de>
17455 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
17456 for the early LTO debug to properly generate references to it
17457 during DIE emission. Do not re-use that for the skeleton for
17459 (dwarf2out_early_finish): Likewise.
17461 2018-04-12 Jakub Jelinek <jakub@redhat.com>
17464 * config/i386/sse.md
17465 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
17466 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
17467 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
17468 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
17469 and output is a reg, avoid creating invalid lowpart subreg, but
17470 instead split into a 512-bit move. Don't split if not AVX512VL,
17471 input is xmm16+ reg and output is a mem.
17472 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
17473 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
17474 xmm16+ reg and output is a mem.
17476 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
17478 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
17479 also for flag_dwarf2_cfi_asm.
17481 2018-04-12 Jakub Jelinek <jakub@redhat.com>
17483 PR rtl-optimization/85342
17484 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
17485 a bool scalar var inside of the loop instead. Don't try to update
17486 recog_data.operand after failed apply_change_group.
17488 2018-04-12 Tom de Vries <tom@codesourcery.com>
17491 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
17492 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
17493 array with flexible array member as array without given dimension.
17494 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
17495 argument for undefined param to true.
17497 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
17500 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
17501 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
17502 from PowerPC section.
17503 * config/rs6000/sysv4.opt (mcall-): Improve help text.
17504 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
17505 help text that is too long.
17506 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
17507 help text that is too long.
17508 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
17509 help text that is too long.
17511 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
17513 * config/alpha/alpha.md (stack_probe_internal): Rename
17514 from "probe_stack". Update all callers.
17516 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
17518 PR rtl-optimization/84566
17519 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
17520 sched_macro_fuse_insns.
17522 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
17525 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
17526 (compute_block_dependences): ... from here.
17528 2018-04-11 Jakub Jelinek <jakub@redhat.com>
17530 PR tree-optimization/85331
17531 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
17532 from int to HOST_WIDE_INT.
17534 2018-04-11 Martin Jambor <mjambor@suse.cz>
17537 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
17538 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
17539 not the same as the source val.
17540 (cgraph_edge_brings_value_p): New parameter.
17541 (gather_edges_for_value): Pass destination value to
17542 cgraph_edge_brings_value_p.
17543 (perhaps_add_new_callers): Likewise.
17544 (get_info_about_necessary_edges): Likewise and exclude values brought
17545 only by self-recursive edges.
17546 (create_specialized_node): Redirect only clones of self-calling edges.
17547 (+self_recursive_pass_through_p): New function.
17548 (find_more_scalar_values_for_callers_subset): Use it.
17549 (find_aggregate_values_for_callers_subset): Likewise.
17550 (known_aggs_to_agg_replacement_list): Removed.
17551 (decide_whether_version_node): Re-calculate known constants for all
17552 remaining context clones.
17554 2018-04-11 Richard Biener <rguenther@suse.de>
17557 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
17558 from early DWARF output.
17559 (dwarf2out_early_finish): Output line info unconditionally into
17560 early DWARF and add reference to it.
17562 2018-04-11 Jakub Jelinek <jakub@redhat.com>
17565 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
17566 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
17567 other than V2DFmode using iptr mode attribute.
17568 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
17570 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
17572 PR rtl-optimization/84659
17573 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
17575 2018-04-11 Jakub Jelinek <jakub@redhat.com>
17578 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
17580 (output_loc_list): Pass address of a dummy size variable even in the
17581 locview handling loop.
17582 (index_location_lists): Add comment on why skip_loc_list_entry can't
17585 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
17588 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
17591 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
17594 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
17595 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
17596 and -mstring-compare-inline-limit.
17598 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
17601 * config/rs6000/rs6000.md (allocate_stack): Put the residual size
17602 for stack clash protection in a register whenever we need it to be in
17605 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
17607 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
17608 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
17610 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
17613 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
17615 (mlong-double-): Ditto.
17616 * config/rs6000/sysv4.opt (msdata=): Ditto.
17617 (mtls-size=): Ditto.
17619 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
17621 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
17622 erroneous entries for
17623 "vector int vec_ldl (int, long int *)", and
17624 "vector unsigned int vec_ldl (int, unsigned long int *)".
17625 Add comments and entries for
17626 "vector bool char vec_ldl (int, bool char *)",
17627 "vector bool short vec_ldl (int, bool short *)",
17628 "vector bool int vec_ldl (int, bool int *)",
17629 "vector bool long long vec_ldl (int, bool long long *)",
17630 "vector pixel vec_ldl (int, pixel *)",
17631 "vector long long vec_ldl (int, long long *)",
17632 "vector unsigned long long vec_ldl (int, unsigned long long *)".
17633 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
17634 type tree bool_long_long_type_node and correct definition of
17635 bool_V2DI_type_node to make reference to this new type tree.
17636 (rs6000_mangle_type): Replace erroneous reference to
17637 bool_long_type_node with bool_long_long_type_node.
17638 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
17639 comments to emphasize sign distinctions for char and int types and
17640 replace RS6000_BTI_bool_long constant with
17641 RS6000_BTI_bool_long_long constant. Also add comment to restrict
17642 use of RS6000_BTI_pixel.
17643 (bool_long_type_node): Remove this macro definition.
17644 (bool_long_long_type_node): New macro definition
17646 2018-04-10 Jakub Jelinek <jakub@redhat.com>
17648 PR rtl-optimization/85300
17649 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
17650 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
17651 simplify_unary_operation fails.
17653 2018-04-10 Martin Liska <mliska@suse.cz>
17655 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
17656 cgraph_edge and ipa_ref.
17658 2018-04-10 Jakub Jelinek <jakub@redhat.com>
17662 * config/i386/sse.md
17663 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
17664 computation of the VEC_MERGE selector from mask.
17665 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
17666 Fix decoding of the VEC_MERGE selector into mask.
17668 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
17670 PR tree-optimization/85286
17671 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
17673 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
17675 * final.c (final_1): Set insn_last_address as well as
17676 insn_current_address.
17678 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
17681 * explow.c (emit_stack_probe): Call validize_mem on memory location
17682 before passing it to gen_probe_stack. Create address operand and
17683 legitimize it for the probe_stack_address case.
17685 2018-04-09 Jan Hubicka <jh@suse.cz>
17688 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
17689 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
17690 * tree.c (free_lang_data_in_type): Fix handling of binfos;
17692 (free_lang_data): Rebuild type inheritance graph.
17694 2018-04-09 Martin Sebor <msebor@redhat.com>
17696 * invoke.texi (-finline-small-functions): Mention other optimization
17698 (-findirect-inlining, -fpartial-inlining): Same.
17699 (-finline-functions-called-once): Same.
17700 (-freorder-blocks-and-partition): Same.
17702 2018-04-09 Jan Hubicka <jh@suse.cz>
17705 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
17706 jumps; choose last target that matches the criteria (i.e.
17707 no partition changes for non-crossing jumps).
17708 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
17709 support for redirecting crossing jumps to non-crossing.
17711 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
17713 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
17714 also for naked functions.
17716 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
17718 * config/arc/arc.md (add_shift): New pattern.
17719 (add_shift2): Likewise.
17720 (sub_shift): Likewise.
17721 (sub_shift_cmp0_noout): Likewise.
17722 (compare_si_ashiftsi): Likewise.
17723 (xbfu_cmp0_noout): New combine pattern.
17724 (xbfu_cmp0"): Likewise.
17725 (movsi_set_cc_insn): Place the predicable variant first.
17726 (commutative_binary_cmp0_noout): Remove clobber.
17727 (commutative_binary_cmp0): New pattern.
17728 (noncommutative_binary_cmp0): Likewise.
17729 (noncommutative_binary_cmp0_noout): Likewise.
17730 (noncommutative_binary_comparison_result_used): Removed.
17731 (rsub_cmp0): New pattern.
17732 (rsub_cmp0_noout): Likewise.
17733 (extzvsi): Changed, keep only meaningful variants.
17734 (SQH, SEZ): New iterators.
17735 (SQH_postfix): New mode attribute.
17736 (SEZ_prefix): New code attribute.
17737 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
17738 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
17739 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
17740 of numerical value.
17741 (noncommutative_operator): Check the availability of barrel
17744 2018-04-09 Richard Biener <rguenther@suse.de>
17746 PR tree-optimization/85284
17747 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
17748 Only use the niter constraining form of simple_iv when the exit
17749 is always executed.
17751 2018-04-09 Tom de Vries <tom@codesourcery.com>
17754 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
17755 (define_expand "*memory_barrier"): New define_expand.
17756 (define_insn "memory_barrier"): New insn.
17758 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
17760 PR rtl-optimization/80463
17761 PR rtl-optimization/83972
17762 PR rtl-optimization/83480
17764 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
17765 correct producer for the insn.
17766 (tidy_control_flow): Fixup seqnos in case of debug insns.
17768 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
17770 PR rtl-optimization/83913
17772 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
17773 different sched-times when merging exprs.
17775 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
17777 PR rtl-optimization/83962
17779 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
17780 tidy_fallthru_edge and tidy_control_flow.
17782 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
17784 PR rtl-optimization/83530
17786 * sel-sched.c (force_next_insn): New global variable.
17787 (remove_insn_for_debug): When force_next_insn is true, also leave only
17788 next insn in the ready list.
17789 (sel_sched_region): When the region wasn't scheduled, make another pass
17790 over it with force_next_insn set to 1.
17792 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
17794 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
17796 * config/nds32/constants.md (unspec_volatile_element): Add enum values
17797 for interrupt control.
17798 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
17799 functions for interrupt control.
17800 * config/nds32/nds32-intrinsic.md: Likewise.
17801 * config/nds32/nds32_intrinsic.h: Likewise.
17802 * config/nds32/nds32.h (nds32_builtins): Likewise.
17804 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
17806 * config/nds32/nds32.c (nds32_init_machine_status,
17807 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
17808 strict_aligned_p field.
17809 (nds32_expand_to_rtl_hook): New function.
17810 (TARGET_EXPAND_TO_RTL_HOOK): Define.
17811 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
17813 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
17814 Chung-Ju Wu <jasonwucj@gmail.com>
17816 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
17817 * config/nds32/nds32-n7.md: New file.
17818 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
17819 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
17821 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
17822 * config/nds32/nds32.md (pipeline_model): Add n7.
17823 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
17824 * config/nds32/pipelines.md: Include n7 settings.
17826 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
17827 Chung-Ju Wu <jasonwucj@gmail.com>
17829 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
17830 * config/nds32/nds32-e8.md: New file.
17831 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
17832 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
17834 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
17835 * config/nds32/nds32.md (pipeline_model): Add e8.
17836 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
17837 * config/nds32/pipelines.md: Include e8 settings.
17839 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
17840 Chung-Ju Wu <jasonwucj@gmail.com>
17842 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
17843 * config/nds32/nds32-n8.md: New file.
17844 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
17845 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
17847 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
17848 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
17849 * config/nds32/nds32.md (pipeline_model): Add n8.
17850 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
17851 * config/nds32/pipelines.md: Include n8 settings.
17853 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
17854 Chung-Ju Wu <jasonwucj@gmail.com>
17856 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
17857 * config/nds32/nds32-n9-2r1w.md: New file.
17858 * config/nds32/nds32-n9-3r2w.md: New file.
17859 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
17860 nds32_register_ports): New or modify for cpu n9.
17861 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
17863 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
17864 * config/nds32/nds32-utils.c: New file.
17865 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
17866 TARGET_MUL_SLOW): Define.
17867 * config/nds32/nds32.md (pipeline_model): New attribute.
17868 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
17869 New options that support cpu n9.
17870 * config/nds32/pipelines.md: Include n9 settings.
17871 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
17873 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
17875 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
17876 information if necessary.
17877 (output_cond_branch_compare_zero): Likewise.
17878 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
17879 (nds32_target_alignment): Refine for alignment.
17880 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
17881 (FUNCTION_BOUNDARY): Modify.
17882 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
17884 * config/nds32/nds32.opt (malways-align, malign-functions): New.
17886 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
17888 * config/nds32/constants.md (unspec_volatile_element): Add values for
17889 TLB operation and data prefetch.
17890 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
17891 functions for TLB operation and data prefetch.
17892 * config/nds32/nds32-intrinsic.md: Likewise.
17893 * config/nds32/nds32_intrinsic.h: Likewise.
17894 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
17895 (nds32_print_operand): Likewise.
17896 * config/nds32/nds32.h (nds32_builtins): Likewise.
17898 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
17899 Andrew Pinski <pinsika@gcc.gnu.org>
17901 PR middle-end/82976
17902 * match.pd: Use constant_boolean_node of correct type instead of
17903 boolean_true_node or boolean_false_node for simplifying
17904 pointer comparisons to zero.
17906 2018-04-07 Jakub Jelinek <jakub@redhat.com>
17908 PR tree-optimization/80021
17909 * tree.c (verify_type_variant): Make error call in verify_variant_match
17910 translatable and remove final full stop.
17912 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
17914 * config/nds32/constants.md (unspec_volatile_element): Add
17915 UNSPEC_VOLATILE_EH_RETURN.
17916 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
17917 nds32_output_stack_pop): Support dwarf exception handling process.
17918 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
17919 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
17920 exception handling process.
17921 (nds32_compute_stack_frame): Likewise.
17922 (nds32_return_addr_rtx): Likewise.
17923 (nds32_initial_elimination_offset): Likewise.
17924 (nds32_expand_prologue): Likewise.
17925 (nds32_expand_epilogue): Likewise.
17926 (nds32_dynamic_chain_address): New function.
17927 * config/nds32/nds32.h (machine_function): Add fields for dwarf
17928 exception handling.
17929 (DYNAMIC_CHAIN_ADDRESS): Define.
17930 (EH_RETURN_DATA_REGNO): Define.
17931 (EH_RETURN_STACKADJ_RTX): Define.
17932 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
17933 patterns for dwarf exception handling.
17935 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
17937 * config/nds32/nds32.h: Clean up obsolete macros.
17939 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
17941 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
17942 Add enum values for particular instructions.
17943 * config/nds32/nds32-intrinsic.c: Implementation of expanding
17944 particular intrinsic functions.
17945 * config/nds32/nds32-intrinsic.md: Likewise.
17946 * config/nds32/nds32_intrinsic.h: Likewise.
17947 * config/nds32/nds32.h (nds32_builtins): Likewise.
17948 * config/nds32/nds32.md (type): Add pbsad and pbsada.
17949 (btst, ave): New patterns for particular instructions.
17951 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
17953 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
17954 Add enum values for atomic load/store and memory sync.
17955 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
17957 * config/nds32/nds32-intrinsic.md: Likewise.
17958 * config/nds32/nds32_intrinsic.h: Likewise.
17959 * config/nds32/nds32.h (nds32_builtins): Likewise.
17961 2018-04-07 Jakub Jelinek <jakub@redhat.com>
17963 PR tree-optimization/85257
17964 * fold-const.c (native_encode_vector): If not all elts could fit
17965 and off is -1, return 0 rather than offset.
17966 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
17967 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
17968 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
17969 adjust buffer in native_interpret_expr call.
17971 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
17973 * config/nds32/constants.md (unspec_volatile_element): Add cache
17974 control enum values.
17975 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
17976 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
17977 * config/nds32/nds32.c (nds32_cctl_names): New.
17978 (nds32_print_operand): Handle cache control register names.
17979 * config/nds32/nds32.h (nds32_builtins): New enum values.
17980 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
17982 * config/nds32/nds32.md (type): Add mmu.
17983 * config/nds32/pipelines.md (simple_insn): Add mmu.
17985 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
17987 * config/nds32/nds32.md (type): Remove call.
17988 * config/nds32/pipelines.md (simple_insn): Likewise.
17990 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
17992 * config/nds32/constants.md (unspec_volatile_element): Add
17993 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
17994 UNSPEC_VOLATILE_FMFCFG.
17995 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
17996 description for fmfcfg and fmfcsr.
17997 (bdesc_1arg): Add fmtcsr.
17998 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
17999 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
18000 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
18001 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
18002 unspec_fmfcfg): New patterns.
18003 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
18004 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
18005 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
18006 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
18007 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
18008 __nds32__fmfcfg): Define.
18010 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
18012 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
18013 intrinsic register names.
18014 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
18015 intrinsic register enum values and macros.
18017 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
18019 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
18020 for load/store addressing form.
18021 (nds32_print_operand_address): Likewise.
18023 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
18026 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
18027 based on LABEL_REF. Remove useless assertion.
18028 (pic_address_needs_scratch): Fix formatting.
18029 (sparc_legitimize_pic_address): Minor tweaks.
18030 (sparc_delegitimize_address): Adjust assertion accordingly.
18031 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
18032 into symbolic_operand.
18033 (movsi_high_pic_label_ref): Likewise.
18034 (movsi_lo_sum_pic_label_ref): Likewise.
18035 (movdi_pic_label_ref): Likewise.
18036 (movdi_high_pic_label_ref): Likewise.
18037 (movdi_lo_sum_pic_label_ref): Likewise.
18039 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
18041 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
18042 custom LIB_SPEC setup.
18044 2018-04-06 Ruslan Bukin <br@bsdpad.com>
18045 Kito Cheng <kito.cheng@gmail.com>
18047 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
18048 * config/riscv/freebsd.h: New.
18050 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
18052 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
18053 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
18056 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
18057 Kito Cheng <kito.cheng@gmail.com>
18059 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
18060 nds32_output_call, nds32_symbol_binds_local_p): New functions.
18061 * config/nds32/nds32-protos.h (nds32_output_call,
18062 nds32_output_return): Declare.
18063 * config/nds32/nds32.md: Refine all the call and return patterns.
18065 2018-04-06 Jakub Jelinek <jakub@redhat.com>
18068 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
18069 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
18071 PR rtl-optimization/84872
18072 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
18073 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
18074 EDGE_CROSSING edge.
18076 2018-04-06 Tamar Christina <tamar.christina@arm.com>
18078 * expr.c (copy_blkmode_to_reg): Revert 254862.
18079 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
18081 2018-04-06 Richard Biener <rguenther@suse.de>
18083 PR middle-end/85244
18084 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
18085 after seeing a component reference with an adjacent field. Treat
18086 refs to arrays at struct end of external decls similar to
18087 refs to unconstrained commons.
18089 2018-04-06 Jakub Jelinek <jakub@redhat.com>
18092 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
18093 look through SAVE_EXPRs with non-side-effects argument. Adjust
18095 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
18098 2018-04-06 Richard Biener <rguenther@suse.de>
18100 PR middle-end/85180
18101 * alias.c (find_base_term): New wrapper around find_base_term
18102 unwinding CSELIB_VAL_PTR changes.
18103 (find_base_term): Do not restore CSELIB_VAL_PTR during the
18106 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
18108 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
18110 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
18111 constant definitions.
18112 ("nop"): lr 0,0 -> nopr r0
18113 ("nop_lr0", "nop_lr1"): New insn definitions.
18115 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
18117 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
18118 NDS32_V3PUSH_AVAILABLE_P macro.
18120 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
18121 Chung-Ju Wu <jasonwucj@gmail.com>
18123 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
18124 (nds32*-*-*): Add float and fpu_config into supported_defaults.
18125 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
18126 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
18127 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
18128 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
18129 * config/nds32/constraints.md: New constraints and checking for hard
18130 float configuration.
18131 * config/nds32/iterators.md: New mode iterator and attribute for hard
18132 float configuration.
18133 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
18135 * config/nds32/nds32-fpu.md: New file.
18136 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
18137 deal with hard float code generation.
18138 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
18140 (abi_type, float_reg_number): New enum type.
18141 * config/nds32/nds32-predicates.c: New predicates for hard float.
18142 * config/nds32/nds32-protos.h: Declare functions for hard float.
18143 * config/nds32/nds32.c: Implementation for hard float configuration.
18144 * config/nds32/nds32.h: Definitions for hard float configuration.
18145 * config/nds32/nds32.md: Include hard float machine description and
18146 modify patterns for hard float configuration.
18147 * config/nds32/nds32.opt: New options for hard float configuration.
18148 * config/nds32/predicates.md: New predicates for hard float
18151 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
18153 * common/config/nds32/nds32-common.c
18154 (nds32_option_optimization_table): Enable -mreleax-hint by default.
18156 2018-04-05 Jakub Jelinek <jakub@redhat.com>
18158 PR middle-end/85195
18159 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
18160 CONSTRUCTOR_ELT (ctor, ...)->value.
18162 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
18165 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
18167 2018-04-05 Tom de Vries <tom@codesourcery.com>
18170 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
18173 2018-04-05 Shiva Chen <shiva0217@gmail.com>
18174 Kito Cheng <kito.cheng@gmail.com>
18176 * config/nds32/constraints.md (U33): Fine-tune checking condition.
18177 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
18178 * config/nds32/nds32.h (nds32_16bit_address_type): Add
18179 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
18181 2018-04-05 Shiva Chen <shiva0217@gmail.com>
18182 Kito Cheng <kito.cheng@gmail.com>
18184 * config/nds32/constraints.md (Ufe): New memory constraint.
18185 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
18186 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
18187 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
18189 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
18190 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
18192 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
18194 * config/nds32/nds32.md: Use optimize_size in the condition for
18195 alu-shift instructions.
18197 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
18199 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
18201 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
18203 * config/nds32/nds32.md (negsi2): Refine pattern.
18205 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
18206 Chung-Ju Wu <jasonwucj@gmail.com>
18208 * config/nds32/iterators.md (shift_rotate): New code iterator.
18209 (shift): New code attribute.
18210 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
18211 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
18212 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
18213 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
18214 bit-wise operations.
18215 (andsi3, *andsi3): Ditto.
18216 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
18217 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
18218 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
18219 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
18220 nds32_ior_operand, nds32_xor_operand): New predicates.
18222 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
18224 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
18225 (addsi3, subsi3): ... this.
18227 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
18229 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
18231 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
18233 * config/nds32/nds32.md: Adjust indention.
18235 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
18237 * config/nds32/nds32.md (feature): New attribute.
18239 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
18241 * config/nds32/nds32.md (subtype): New attribute.
18243 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
18246 * config/arm/arm-builtins.c (arm_expand_builtin): Change
18247 expansion to perform a bitwise AND of the argument followed by a
18248 boolean negation of the result.
18250 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
18252 PR rtl-optimization/84878
18253 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
18254 the basic block. Assert the use reference is not artificial and that
18255 it has an associated insn.
18257 2018-04-04 Michael Matz <matz@suse.de>
18259 * builtins.c (compute_objsize): Pass correct operand
18260 to array_at_struct_end_p.
18262 2018-04-04 Richard Biener <rguenther@suse.de>
18265 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
18266 from contexts for DINFO_LEVEL_TERSE and below.
18268 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
18270 * config/nds32/nds32-doubleword.md (move_<mode>): Require
18271 resiter_operand condition.
18272 * config/nds32/nds32.md (*move<mode>): Ditto.
18274 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
18275 Monk Chiang <sh.chiang04@gmail.com>
18277 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
18279 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
18281 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
18283 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
18284 Kito Cheng <kito.cheng@gmail.com>
18286 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
18287 nds32_cond_code_str, output_cond_branch,
18288 output_cond_branch_compare_zero, nds32_expand_cbranch,
18289 nds32_expand_cstore, nds32_expand_movcc,
18290 nds32_output_cbranchsi4_equality_zero,
18291 nds32_output_cbranchsi4_equality_reg,
18292 nds32_output_cbranchsi4_equality_reg_or_const_int,
18293 nds32_output_cbranchsi4_greater_less_zero: New functions.
18294 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
18295 nds32_expand_cstore, nds32_expand_movcc,
18296 nds32_output_cbranchsi4_equality_zero,
18297 nds32_output_cbranchsi4_equality_reg,
18298 nds32_output_cbranchsi4_equality_reg_or_const_int,
18299 nds32_output_cbranchsi4_greater_less_zero): Declare.
18300 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
18301 nds32_rimm11s_operand): New predicates.
18302 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
18303 * config/nds32/nds32.md: Rewrite all the branch and conditional move
18306 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
18308 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
18309 * config/nds32/nds32.md: Ditto.
18310 * config/nds32/pipelines.md: Ditto.
18312 2018-04-04 Richard Biener <rguenther@suse.de>
18314 PR tree-optimization/85168
18315 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
18316 propagating abnormals.
18318 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
18320 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
18322 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
18323 Kito Cheng <kito.cheng@gmail.com>
18325 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
18326 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
18327 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
18328 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
18329 * config/nds32/nds32.md (sibcall_internal): New.
18330 (sibcall_register): Remove.
18331 (sibcall_immediate): Remove.
18332 (sibcall_value_internal): New.
18333 (sibcall_value_register): Remove.
18334 (sibcall_value_immediate): Remove.
18335 * config/nds32/predicates.md (nds32_general_register_operand): New.
18336 (nds32_call_address_operand): New.
18338 2018-04-03 Jakub Jelinek <jakub@redhat.com>
18340 PR rtl-optimization/85167
18341 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
18342 bb_defs if *split_p, instead preinitialize it to NULL.
18344 PR tree-optimization/85156
18345 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
18346 evaluating the argument multiple times.
18348 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
18350 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
18352 (_mm_cvtpd_ps): Likewise.
18353 (_mm_cvttpd_epi32): Likewise.
18354 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
18355 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
18356 vector, pixel, and bool following altivec.h include.
18358 2018-04-03 Martin Sebor <msebor@redhat.com>
18360 * doc/extend.texi (Common Function Attributes): Clarify.
18361 (const attribute): Likewise.
18362 (pure attribute): Likewise.
18364 2018-04-03 Jakub Jelinek <jakub@redhat.com>
18367 * config/i386/i386.c (ix86_expand_vector_set): Use
18368 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
18370 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
18372 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
18373 instructions when changing rounding bits to preserve precision bits
18374 in the x87 control word.
18376 2018-04-03 Martin Liska <mliska@suse.cz>
18378 PR tree-optimization/82491
18379 * rtl.h (strip_offset_and_add): Replace += suboffset with
18380 poly_uint64 () + suboffset.
18382 2018-03-29 Martin Liska <mliska@suse.cz>
18383 Martin Jambor <mjambor@suse.cz>
18386 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
18387 param_type is not an integral or pointer type.
18389 2018-04-03 Richard Biener <rguenther@suse.de>
18391 * sese.h (recompute_all_dominators): Remove.
18393 2018-04-02 Martin Sebor <msebor@redhat.com>
18395 * doc/invoke.texi (-Wrestrict): Fix typos.
18397 2018-04-02 Jim Wilson <jimw@sifive.com>
18399 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
18400 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
18401 (<optab>di3, <optab>si3_extend): Likewise.
18402 (<optab>si3_mask, <optab>si3_mask_1): New.
18403 (<optab>di3_mask, <optab>di3_mask_1): New.
18404 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
18405 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
18406 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
18408 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
18410 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
18413 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
18415 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
18416 (nds32_canonicalize_comparison): New function.
18418 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
18419 Kito Cheng <kito.cheng@gmail.com>
18420 Kuan-Lin Chen <kuanlinchentw@gmail.com>
18422 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
18423 * config/nds32/constants.md (unspec_volatile_element): Add
18424 UNSPEC_VOLATILE_RELAX_GROUP.
18425 * config/nds32/nds32-relax-opt.c: New file.
18426 * config/nds32/nds32-predicates.c
18427 (nds32_symbol_load_store_p): New function.
18428 * config/nds32/nds32-protos.h
18429 (nds32_symbol_load_store_p): Declare function.
18430 (make_pass_nds32_relax_opt): Declare new rtl pass function.
18431 * config/nds32/nds32.c
18432 (nds32_register_pass): New function to register pass.
18433 (nds32_register_passes): New function to register passes.
18434 * config/nds32/nds32.md (relax_group): New pattern.
18435 * config/nds32/nds32.opt (mrelax-hint): New option.
18436 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
18438 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
18440 * config/nds32/t-nds32: Modify files dependency.
18442 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
18444 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
18445 (PROFILE_HOOK): Define its implementation.
18447 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
18449 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
18450 type and 32-bit size.
18452 2018-04-01 Jakub Jelinek <jakub@redhat.com>
18454 PR middle-end/85090
18455 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
18456 (V_128_256): New mode iterator.
18457 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
18458 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
18459 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
18461 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
18462 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
18464 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
18467 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
18468 NaN inputs correctly.
18470 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
18473 * config/rs6000/vsx.md (??r): New mode attribute.
18474 (*vsx_mov<mode>_64bit): Use it.
18475 (*vsx_mov<mode>_32bit): Likewise.
18477 2018-03-30 Martin Sebor <msebor@redhat.com>
18479 PR tree-optimization/84818
18480 * builtins.c (check_access): Use warning_n.
18482 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
18485 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
18487 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
18490 2018-03-30 Julia Koval <julia.koval@intel.com>
18493 * x86-tune.def (movx, partial_reg_dependency): Enable for
18496 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
18498 PR inline-asm/84985
18499 * lra-constraints.c (process_alt_operands): Move setting
18500 this_alternative_matches below.
18502 2018-03-29 Martin Liska <mliska@suse.cz>
18505 * doc/invoke.texi: Document how LTO works with debug info.
18506 Describe auto-load support of binutils. Mention 'x86-64'
18507 as valid option value of -march option.
18509 2018-03-29 Jakub Jelinek <jakub@redhat.com>
18511 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
18514 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
18515 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
18516 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
18519 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
18522 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
18523 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
18524 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
18525 for RS6000_BTM_POWERPC64.
18526 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
18527 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
18528 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
18531 (DIVDEU): Likewise.
18533 2018-03-28 Carl Love <cel@us.ibm.com>
18536 2017-09-27 Carl Love <cel@us.ibm.com>
18538 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
18539 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
18540 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
18543 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
18545 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
18546 instead of __vector bool.
18547 (_mm_max_pu8): Likewise.
18548 (_mm_min_pi16): Likewise.
18550 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
18553 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
18554 (DIVWEUO): Likewise.
18555 (DIVDEO): Likewise.
18556 (DIVDEUO): Likewise.
18557 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
18558 DIVWEUO and DIVDEUO.
18559 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
18560 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
18561 (div_extend): Likewise.
18562 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
18564 (__builtin_divweuo): Likewise.
18565 (__builtin_divdeo): Likewise.
18566 (__builtin_divdeuo): Likewise.
18568 2018-03-28 Jakub Jelinek <jakub@redhat.com>
18571 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
18572 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
18574 PR tree-optimization/82004
18575 * gimple-match-head.c (optimize_pow_to_exp): New function.
18576 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
18577 Don't fold to exp if optimize_pow_to_exp is false.
18579 2018-03-28 Martin Liska <mliska@suse.cz>
18582 * calls.c (initialize_argument_information): Fix trailing space.
18583 * common.opt: Fix typo and provide better explanation for
18584 -fsanitize-coverage option.
18585 * config/i386/i386.opt: Fix typo.
18587 2018-03-28 Jakub Jelinek <jakub@redhat.com>
18588 Martin Liska <mliska@suse.cz>
18591 * gimplify.c (asan_poison_variable): Don't do the check for
18592 gimplify_omp_ctxp here.
18593 (gimplify_decl_expr): Do it here.
18594 (gimplify_target_expr): Likewise.
18596 2018-03-28 Martin Liska <mliska@suse.cz>
18599 * config/i386/i386.c (ix86_function_arg_advance): Do not call
18600 chkp_type_bounds_count if MPX is not enabled.
18602 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
18604 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
18606 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
18609 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
18610 function to create the function decl for complex long double
18611 multiply and divide for -mabi=ieeelongdouble.
18612 (init_float128_ieee): Call it.
18614 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
18617 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
18618 -fcf-protection=branch -mibt.
18619 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
18621 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
18624 * config/arm/arm.c (arm_valid_symbolic_address): Handle
18625 arm_word_relocations.
18627 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
18630 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
18631 extern array declarations.
18633 2018-03-27 Richard Biener <rguenther@suse.de>
18635 PR middle-end/84067
18636 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
18637 explicit single_use checks.
18639 2018-03-27 Richard Biener <rguenther@suse.de>
18641 PR tree-optimization/85082
18642 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
18645 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
18647 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
18648 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
18649 Turn on fasynchronous-unwind-tables and funwind-tables.
18651 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
18654 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
18655 (*bmi_blsr_<mode>_ccz): Ditto.
18657 2018-03-26 Tom de Vries <tom@codesourcery.com>
18659 PR tree-optimization/85063
18660 * omp-general.c (offloading_function_p): New function. Factor out
18662 * omp-offload.c (pass_omp_target_link::gate): ... here.
18663 * omp-general.h (offloading_function_p): Declare.
18664 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
18665 with attribute omp declare target for offloading functions.
18667 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
18669 PR tree-optimization/84005
18670 * tree-data-ref.h (get_base_for_alignment): Declare.
18671 * tree-data-ref.c (get_base_for_alignment_1): New function.
18672 (get_base_for_alignment): Likewise.
18673 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
18674 get_base_for_alignment to find a suitable base object, instead
18675 of always using drb->base_address.
18677 2018-03-23 Jakub Jelinek <jakub@redhat.com>
18679 PR inline-asm/85022
18680 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
18681 known size by default.
18683 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
18685 PR inline-asm/85030
18686 * lra-constraints.c (process_alt_operands): Don't match BLKmode
18687 and non BLKmode operands.
18689 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
18692 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
18693 Clean up attributes.
18695 2018-03-23 Richard Biener <rguenther@suse.de>
18698 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
18699 we are going to emit early debug for LTO.
18701 2018-03-23 Jakub Jelinek <jakub@redhat.com>
18703 PR inline-asm/85034
18704 * function.c (match_asm_constraints_1): Don't optimize if input
18705 doesn't satisfy general_operand predicate for output's mode.
18707 PR inline-asm/85022
18708 * alias.c (write_dependence_p): Don't require for x_canonicalized
18709 non-VOIDmode if x has VOIDmode.
18712 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
18713 just don't try to optimize it rather than assert it never happens.
18715 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
18717 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
18718 macro expansions for definition of ST_INTERNAL_<mode> and
18719 LD_INTERNAL_<mode> builtins.
18720 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
18722 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
18724 (altivec_expand_st_builtin): Likewise.
18725 (altivec_expand_builtin): Remove calls to deleted functions.
18726 (rs6000_address_for_altivec): Delete this function.
18727 * config/rs6000/vector.md: Remove expands for
18728 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
18730 2018-03-22 Sudakshina Das <sudi.das@arm.com>
18733 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
18734 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
18735 re-computing once computed.
18736 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
18737 (arm_init_machine_status): Initialize
18738 machine->static_chain_stack_bytes.
18740 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
18743 * doc/extend.texi: Add four new prototypes for vec_ld.
18744 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
18745 definitions for more logical presentation.
18746 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
18747 entries for V1TI variants of __builtin_altivec_ld builtin.
18748 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
18749 handling of V1TI variant of LVX icode pattern.
18750 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
18751 (rs6000_gimple_fold_builtin): Likewise.
18752 (altivec_init_builtins): Add code to define
18753 __builtin_altivec_lvx_v1ti function.
18755 2018-03-22 Jakub Jelinek <jakub@redhat.com>
18757 PR inline-asm/84941
18758 * function.c (match_asm_constraints_1): Don't do the optimization
18759 if input isn't a REG, SUBREG, MEM or constant.
18761 2018-03-22 Tom de Vries <tom@codesourcery.com>
18763 PR tree-optimization/84956
18764 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
18765 bb_has_abnormal_pred.
18767 2018-03-22 Jakub Jelinek <jakub@redhat.com>
18770 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
18771 DECL_INITIAL (decl) to decl at the end.
18772 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
18773 adjust the comment.
18775 2018-03-21 Joseph Myers <joseph@codesourcery.com>
18777 * doc/extend.texi (__builtin_tgmath): Document when complex
18778 integer types are treated as _Complex _Float64.
18780 2018-03-21 Tom de Vries <tom@codesourcery.com>
18782 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
18784 2018-03-21 Jakub Jelinek <jakub@redhat.com>
18786 PR tree-optimization/84960
18787 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
18788 if it is ENTRY block, move them into single succ of ENTRY in that case.
18790 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
18792 PR tree-optimization/84811
18793 * poly-int.h (poly_span_traits): Remove the T3 parameter and
18794 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
18795 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
18796 (known_subrange_p): Update accordingly. Cast each value involved
18797 in the size comparison, rather than casting the result of the
18800 2018-03-21 Jakub Jelinek <jakub@redhat.com>
18802 PR tree-optimization/84982
18803 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
18804 by flipping the least significant bit rather than all bits from
18805 bitpos to bitpos + bitsize - 1.
18807 2018-03-21 Nathan Sidwell <nathan@acm.org>
18809 * doc/extend.texi (Deprecated Features): Remove mention of
18810 long-deleted deprecations.
18812 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
18815 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
18816 * configure: Regenerate.
18818 2018-03-21 Tom de Vries <tom@codesourcery.com>
18820 PR tree-optimization/83126
18821 * tree-parloops.c (num_phis): New function.
18822 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
18824 2018-03-21 Nathan Sidwell <nathan@acm.org>
18826 * doc/extend.texi (Deprecated Features): Update deprecated flags,
18827 mention anon-struct/union members and trailing attributes.
18829 2018-03-21 Bin Cheng <bin.cheng@arm.com>
18831 PR tree-optimization/84969
18832 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
18833 builtin memset partitions if they set different rhs values.
18835 2018-03-21 Jakub Jelinek <jakub@redhat.com>
18837 PR rtl-optimization/84989
18838 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
18839 VEC_DUPLICATE with scalar result mode.
18841 2018-03-21 Martin Liska <mliska@suse.cz>
18844 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
18845 not intended return statement.
18847 2018-03-21 Martin Liska <mliska@suse.cz>
18850 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
18851 (chkp_find_bound_slots_1): Limit number of iterations.
18853 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
18856 * Minor grammar fixes for x86 options.
18858 2018-03-20 Jakub Jelinek <jakub@redhat.com>
18861 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
18862 holding REG_CFA_RESTORE notes, instead turn them into a USE.
18864 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
18867 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
18868 (altivec_lvx_<mode>_1op): Likewise.
18869 (altivec_stvx_<mode>_2op): Likewise.
18870 (altivec_stvx_<mode>_1op): Likewise.
18871 (altivec_lvx_<VM2:mode>): New define_expand.
18872 (altivec_stvx_<VM2:mode>): Likewise.
18873 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
18874 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
18875 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
18876 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
18877 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
18878 (rs6000_gen_lvx): Likewise.
18879 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
18880 (altivec_expand_stv_builtin): Likewise.
18881 (altivec_expand_builtin): Likewise.
18882 * config/rs6000/vector.md: Likewise.
18884 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
18887 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
18890 2018-03-20 Richard Biener <rguenther@suse.de>
18893 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
18894 sign-conversions as zero, fall back to standard scalar_stmt
18897 2018-03-20 Martin Liska <mliska@suse.cz>
18900 * predict.c (rebuild_frequencies): Handle case when we have
18901 PROFILE_ABSENT, but flag_guess_branch_prob is false.
18903 2018-03-20 Jakub Jelinek <jakub@redhat.com>
18906 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
18907 flag_section_anchors.
18908 * varasm.c (use_blocks_for_decl_p): Remove hack for
18909 dw2_force_const_mem.
18912 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
18914 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
18915 be created, use lowpart_subreg of operands[0] rather than operands[0]
18917 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
18918 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
18919 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
18920 and n constraint instead of aarch64_shift_imm_di and Usd.
18921 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
18922 (*aarch64_<optab>_reg_minus<mode>3): ... this.
18924 2018-03-20 Sudakshina Das <sudi.das@arm.com>
18927 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
18928 to favor GPR over NEON registers.
18929 (<shift>di3_neon): Likewise.
18931 2018-03-20 Tom de Vries <tom@codesourcery.com>
18934 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
18935 (nvptx_process_pars): Emit bar.sync asap and alap.
18937 2018-03-20 Tom de Vries <tom@codesourcery.com>
18940 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
18941 seen_label if seen_label is already set.
18943 2018-03-20 Jakub Jelinek <jakub@redhat.com>
18946 * config/i386/i386.c (fold_builtin_cpu): For features above 31
18947 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
18948 Use 1U instead of 1. Formatting fixes.
18951 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
18952 instead of TREE_TYPE (s1) for the return value.
18954 2018-03-19 Jakub Jelinek <jakub@redhat.com>
18956 PR tree-optimization/84946
18957 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
18958 bitsize + bitsize in poly_uint64 rather than poly_int64.
18961 * dwarf2asm.c: Include fold-const.c.
18962 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
18963 of decl rather than decl itself.
18965 PR rtl-optimization/84643
18966 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
18968 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
18971 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
18972 calling assemble_variable.
18974 2018-03-19 Sudakshina Das <sudi.das@arm.com>
18977 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
18978 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
18980 2018-03-19 Jim Wilson <jimw@sifive.com>
18983 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
18984 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
18985 (riscv_first_stack_step): Likewise.
18986 (riscv_option_override): Use STACK_BOUNDARY instead of
18987 MIN_STACK_BOUNDARY.
18988 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
18989 MIN_STACK_BOUNDARY.
18990 (BIGGEST_ALIGNMENT): Set to 128.
18991 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
18992 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
18995 2018-03-19 Richard Biener <rguenther@suse.de>
18997 PR tree-optimization/84933
18998 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
18999 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
19001 2018-03-19 Richard Biener <rguenther@suse.de>
19003 PR tree-optimization/84859
19004 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
19005 (cond_if_else_store_replacement): Perform sinking operation on
19006 single-store BBs regardless of MAX_STORES_TO_SINK setting.
19007 Generalize what a BB with a single eligible store is.
19009 2018-03-19 Richard Biener <rguenther@suse.de>
19011 PR tree-optimization/84929
19012 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
19013 chrec_is_positive against non-chrec arg.
19015 2018-03-19 Tamar Christina <tamar.christina@arm.com>
19018 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
19020 2018-03-18 Martin Liska <mliska@suse.cz>
19022 PR rtl-optimization/84635
19023 * regrename.c (build_def_use): Use matches_mode only when
19026 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
19028 PR tree-optimization/84913
19029 * tree-vect-loop.c (vectorizable_reduction): Don't try to
19030 vectorize chains of COND_EXPRs.
19032 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
19034 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
19036 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
19038 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
19040 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
19042 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
19044 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
19045 Kito Cheng <kito.cheng@gmail.com>
19047 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
19048 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
19049 (nds32_adjust_reg_alloc_order): New function.
19050 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
19052 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
19054 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
19055 nds32_print_operand, nds32_print_operand_address): Use
19056 HOST_WIDE_INT_PRINT_DEC instead.
19058 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
19060 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
19062 2018-03-17 Jakub Jelinek <jakub@redhat.com>
19065 * config/i386/i386.c (initial_ix86_tune_features,
19066 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
19067 unsigned long long.
19068 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
19069 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
19070 rather than 1u << ix86_tune. Formatting fix.
19071 (ix86_option_override_internal): Change ix86_arch_mask from
19072 unsigned int to unsigned HOST_WIDE_INT, initialize to
19073 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
19074 (ix86_function_specific_restore): Likewise.
19076 2018-03-16 Jakub Jelinek <jakub@redhat.com>
19079 * postreload.c (reload_combine_recognize_pattern): Perform
19080 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
19081 truncate_int_for_mode the result for the destination's mode.
19084 * hsa-gen.c (mem_type_for_type): Fix comment typo.
19085 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
19087 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
19090 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
19093 * lra-assigns.c (lra_split_hard_reg_for): Don't use
19094 regno_allocno_class_array and sorted_pseudos.
19095 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
19096 insns where regno is used.
19098 2018-03-16 Martin Liska <mliska@suse.cz>
19101 * multiple_target.c (create_dispatcher_calls): Redirect
19102 reference in the symbol table.
19104 2018-03-16 Martin Liska <mliska@suse.cz>
19107 * multiple_target.c (create_dispatcher_calls): Redirect also
19110 2018-03-16 Jakub Jelinek <jakub@redhat.com>
19114 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
19115 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
19116 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
19118 2018-03-16 Julia Koval <julia.koval@intel.com>
19120 * doc/invoke.texi (Skylake Server): Add CLWB.
19121 Cannonlake): Remove CLWB.
19123 2018-03-16 Jakub Jelinek <jakub@redhat.com>
19125 PR tree-optimization/84841
19126 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
19128 (FLOAT_ONE_CONST_TYPE): Define.
19129 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
19130 (sort_by_operand_rank): Put entries with higher constant_type last
19131 rather than first to match comments.
19133 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
19135 * config/nios2/nios2.md (movsi_internal): Fix thinko in
19138 2018-03-15 Jakub Jelinek <jakub@redhat.com>
19141 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
19142 check and use address of target always.
19144 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
19147 * config/i386/i386.c (indirect_thunk_needed): Update comments.
19148 (indirect_thunk_bnd_needed): Likewise.
19149 (indirect_thunks_used): Likewise.
19150 (indirect_thunks_bnd_used): Likewise.
19151 (indirect_return_needed): New.
19152 (indirect_return_bnd_needed): Likewise.
19153 (output_indirect_thunk_function): Add a bool argument for
19155 (output_indirect_thunk_function): Don't generate alias for
19156 function return thunk.
19157 (ix86_code_end): Call output_indirect_thunk_function to generate
19158 function return thunks.
19159 (ix86_output_function_return): Set indirect_return_bnd_needed
19160 and indirect_return_needed instead of indirect_thunk_bnd_needed
19161 and indirect_thunk_needed.
19163 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
19165 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
19166 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
19167 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
19169 2018-03-15 David Malcolm <dmalcolm@redhat.com>
19170 Paul Hua <paul.hua.gm@gmail.com>
19173 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
19175 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
19177 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
19178 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
19179 resp. SFmode cases.
19181 2018-03-15 Tamar Christina <tamar.christina@arm.com>
19184 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
19185 instead of GET_MODE_SIZE when comparing Units.
19187 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
19190 * varasm.c (hash_section): Return an unchangeble hash value
19191 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
19192 Return !aarch64_can_use_per_function_literal_pools_p ().
19194 2018-03-15 Jakub Jelinek <jakub@redhat.com>
19197 * optabs.c (emit_conditional_move): Pass address of cmode's copy
19198 rather than address of cmode as last argument to prepare_cmp_insn.
19200 2018-03-15 Julia Koval <julia.koval@intel.com>
19202 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
19203 F_AVX512VNNI, F_AVX512BITALG): New.
19205 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
19208 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
19209 insn for floating-point loads and stores.
19211 2018-03-14 Carl Love <cel@us.ibm.com>
19213 * config/rs6000/rs6000-c.c: Add macro definitions for
19214 ALTIVEC_BUILTIN_VEC_PERMXOR.
19215 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
19216 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
19217 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
19218 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
19220 * config/doc/extend.texi: Add prototypes for vec_permxor.
19222 2018-03-14 David Malcolm <dmalcolm@redhat.com>
19225 * diagnostic-show-locus.c (class layout_point): Convert m_line
19226 from int to linenum_type.
19227 (line_span::comparator): Use linenum "compare" function when
19228 comparing line numbers.
19229 (test_line_span): New function.
19230 (layout_range::contains_point): Convert param "row" from int to
19232 (layout_range::intersects_line_p): Likewise.
19233 (layout::will_show_line_p): Likewise.
19234 (layout::print_source_line): Likewise.
19235 (layout::should_print_annotation_line_p): Likewise.
19236 (layout::print_annotation_line): Likewise.
19237 (layout::print_leading_fixits): Likewise.
19238 (layout::annotation_line_showed_range_p): Likewise.
19239 (struct line_corrections): Likewise for field m_row.
19240 (line_corrections::line_corrections): Likewise for param "row".
19241 (layout::print_trailing_fixits): Likewise.
19242 (layout::get_state_at_point): Likewise.
19243 (layout::get_x_bound_for_row): Likewise.
19244 (layout::print_line): Likewise.
19245 (diagnostic_show_locus): Likewise for locals "last_line" and "row".
19246 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
19247 * input.c (selftest::test_linenum_comparisons): New function.
19248 (selftest::input_c_tests): Call it.
19249 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
19250 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
19251 * selftest.h (ASSERT_GT): New macro.
19252 (ASSERT_GT_AT): New macro.
19253 (ASSERT_LT): New macro.
19254 (ASSERT_LT_AT): New macro.
19256 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
19258 PR rtl-optimization/84780
19259 * combine.c (distribute_links): Don't make a link based on pc_rtx.
19261 2018-03-14 Martin Liska <mliska@suse.cz>
19263 * tree.c (record_node_allocation_statistics): Use
19264 get_stats_node_kind.
19265 (get_stats_node_kind): New function extracted from
19266 record_node_allocation_statistics.
19267 (free_node): Use get_stats_node_kind.
19269 2018-03-14 Richard Biener <rguenther@suse.de>
19271 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
19272 that the value-set of ANTIC_IN doesn't grow.
19275 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
19277 (BB_VISITED_WITH_VISITED_SUCCS): New define.
19278 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
19280 2018-03-14 Julia Koval <julia.koval@intel.com>
19282 * config.gcc (icelake-client, icelake-server): New.
19284 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
19285 (initial_ix86_arch_features): Ditto.
19286 (PTA_SKYLAKE): Add SGX.
19287 (PTA_ICELAKE): Remove.
19288 (PTA_ICELAKE_CLIENT): New.
19289 (PTA_ICELAKE_SERVER): New.
19290 (ix86_option_override_internal): Split up icelake on icelake client and
19292 (get_builtin_code_for_version): Ditto.
19293 (fold_builtin_cpu): Ditto.
19294 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
19295 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
19296 * config/i386/i386.h (processor_type): Ditto.
19297 * doc/invoke.texi: Ditto.
19299 2018-03-14 Jakub Jelinek <jakub@redhat.com>
19302 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
19303 INTEGER_CST offset, add it together with bitpos / 8 and
19304 sign extend based on POINTER_SIZE.
19308 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
19311 * config/i386/constraints.md (Yc): New register constraint.
19312 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
19313 Use Yc constraint for alternative 2 of operand 0. Remove
19314 preferred_for_speed attribute.
19316 2018-03-14 Richard Biener <rguenther@suse.de>
19318 PR tree-optimization/84830
19319 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
19320 with the old one to avoid oscillations.
19322 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
19325 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
19327 (assign_by_spills): Return a flag of reload assignment failure.
19328 Do not process the reload assignment failures. Do not spill other
19329 reload pseudos if they has the same reg class. Update n if
19331 (lra_assign): Add a return arg. Set up from the result of
19332 assign_by_spills call.
19333 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
19334 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
19335 usage_insns if it is not NULL.
19336 (spill_hard_reg_in_range): New function.
19337 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
19338 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
19339 function prototypes.
19340 (lra_assign): Change prototype.
19341 * lra.c (lra): Add code to deal with fails by splitting hard reg
19344 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
19346 * config/riscv/riscv.opt (mrelax): New option.
19347 * config/riscv/riscv.c (riscv_file_start): Emit ".option
19348 "norelax" when riscv_mrelax is disabled.
19349 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
19351 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
19354 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
19355 reassociation for int modes.
19357 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
19359 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
19360 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
19362 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
19363 * config/aarch64/aarch64-sve.md
19364 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
19365 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
19366 (*extend<mode><Vwide>2): Rename to...
19367 (aarch64_sve_extend<mode><Vwide>2): ...this.
19368 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
19369 renaming the old pattern to...
19370 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
19372 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
19373 define_expand, renaming the old pattern to...
19374 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
19375 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
19376 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
19377 account when deciding which SVE instruction the optab should use.
19378 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
19380 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
19382 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
19383 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
19384 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
19385 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
19386 (tlsdesc_small_<mode>): Turn a define_expand and use
19387 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
19388 (tlsdesc_small_advsimd_<mode>): ...this.
19389 (tlsdesc_small_sve_<mode>): New pattern.
19391 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
19393 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
19394 (UNSPEC_UMUL_HIGHPART): New constants.
19395 (MUL_HIGHPART): New int iteraor.
19396 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
19397 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
19399 (*<su>mul<mode>3_highpart): New define_insn.
19401 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
19404 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
19407 2018-03-13 Martin Liska <mliska@suse.cz>
19410 * (sem_item_optimizer::sem_item_optimizer): Initialize new
19412 (sem_item_optimizer::~sem_item_optimizer): Release it.
19413 (sem_item_optimizer::merge_classes): Register variable aliases.
19414 (sem_item_optimizer::fixup_pt_set): New function.
19415 (sem_item_optimizer::fixup_points_to_sets): Likewise.
19416 * ipa-icf.h: Declare new variables and functions.
19418 2018-03-13 Jakub Jelinek <jakub@redhat.com>
19420 PR middle-end/84834
19421 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
19422 integer_pow2p@2 and test integer_pow2p in condition.
19423 (A < 0 ? C : 0): Similarly for @1.
19425 PR middle-end/84831
19426 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
19427 characters starting at p contain '\0' character, don't look beyond
19431 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
19432 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
19435 * reg-stack.c (change_stack): Change update_end var from int to
19436 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
19437 also call set_block_for_insn on the newly added insns and rescan.
19440 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
19441 on the last operand.
19444 * tree.c (stabilize_reference_1): Return save_expr (e) for
19445 STATEMENT_LIST even if it doesn't have side-effects.
19447 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
19449 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
19451 2018-03-12 Renlin Li <renlin.li@arm.com>
19453 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
19454 aarch64_output_scalar_simd_mov_immediate.
19456 2018-03-12 Martin Sebor <msebor@redhat.com>
19458 PR tree-optimization/83456
19459 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
19460 for perfectly overlapping calls to memcpy.
19461 (gimple_fold_builtin_memory_chk): Same.
19462 (gimple_fold_builtin_strcpy): Handle no-warning.
19463 (gimple_fold_builtin_stxcpy_chk): Same.
19464 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
19466 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
19468 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
19469 parameter. Use it for SFmode.
19470 (rs6000_function_arg_advance_1): Adjust.
19471 (rs6000_function_arg): Adjust.
19472 (rs6000_gimplify_va_arg): Pass false for that new parameter.
19474 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
19476 PR rtl-optimization/84169
19477 PR rtl-optimization/84780
19478 * combine.c (can_combine_p): Check for a 2-insn combination whether
19479 the destination register is used between the two insns, too.
19481 2018-03-12 Richard Biener <rguenther@suse.de>
19483 PR tree-optimization/84803
19484 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
19485 for refs DR analysis didn't process.
19487 2018-03-12 Richard Biener <rguenther@suse.de>
19489 PR tree-optimization/84777
19490 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
19491 force-vectorize loops ignore whether we are optimizing for size.
19493 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
19495 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
19496 (TARGET_MD_ASM_ADJUST): Define.
19498 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
19499 Kito Cheng <kito.cheng@gmail.com>
19500 Chung-Ju Wu <jasonwucj@gmail.com>
19502 * config/nds32/nds32.c (nds32_compute_stack_frame,
19503 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
19504 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
19505 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
19506 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
19507 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
19508 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
19509 * config/nds32/nds32.md (prologue, epilogue): Use macro
19510 NDS32_V3PUSH_AVAILABLE_P to do checking.
19512 2018-03-11 Jakub Jelinek <jakub@redhat.com>
19515 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
19516 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
19517 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
19518 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
19519 addition of most attributes on !orig_type_die or the attribute not
19520 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
19522 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
19523 Chung-Ju Wu <jasonwucj@gmail.com>
19525 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
19526 __NDS32_VH__ macro.
19527 * config/nds32/nds32.opt (mvh): New option.
19529 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
19530 Chung-Ju Wu <jasonwucj@gmail.com>
19532 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
19534 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
19535 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
19538 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
19539 Chung-Ju Wu <jasonwucj@gmail.com>
19541 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
19543 * config/nds32/nds32-multiple.md (strlensi): New pattern.
19544 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
19546 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
19547 Kito Cheng <kito.cheng@gmail.com>
19548 Chung-Ju Wu <jasonwucj@gmail.com>
19550 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
19551 UNSPEC_FFMISM and UNSPEC_FLMISM.
19552 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
19553 for ffb, ffmism and flmism.
19554 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
19555 (unspec_ffmism): Ditto.
19556 (unspec_flmism): Ditto.
19557 (nds32_expand_builtin_impl): Check if string extension is available.
19558 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
19559 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
19561 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
19564 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
19567 * lra-assigns.c (assign_by_spills): Return a flag of reload
19568 assignment failure. Do not process the reload assignment
19569 failures. Do not spill other reload pseudos if they has the same
19571 (lra_assign): Add a return arg. Set up from the result of
19572 assign_by_spills call.
19573 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
19574 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
19575 usage_insns if it is not NULL.
19576 (spill_hard_reg_in_range): New function.
19577 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
19578 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
19579 function prototypes.
19580 (lra_assign): Change prototype.
19581 * lra.c (lra): Add code to deal with fails by splitting hard reg
19584 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
19587 * config/i386/i386.opt: Replace Enforcment with Enforcement.
19589 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
19592 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
19593 (dw_val_node): Add val_symbolic_view.
19594 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
19595 (symview_upper_bound): New.
19596 (new_line_info_table): Initialize symviews_since_reset.
19597 (dwarf2out_source_line): Count symviews_since_reset and set
19598 symview_upper_bound.
19599 (dw_val_equal_p): Handle symview.
19600 (add_AT_symview): New.
19601 (print_dw_val): Handle symview.
19602 (attr_checksum, attr_checksum_ordered): Likewise.
19603 (same_dw_val_p, size_of_die): Likewise.
19604 (value_format, output_die): Likewise.
19605 (add_high_low_attributes): Use add_AT_symview for entry_view.
19606 (dwarf2out_finish): Reset symview_upper_bound, clear
19609 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
19612 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
19613 Add strict argument and use it.
19614 (rs6000_split_multireg_move): Update for new strict argument.
19615 (mem_operand_gpr): Disallow all non-offsettable addresses.
19616 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
19618 2018-03-09 Jakub Jelinek <jakub@redhat.com>
19621 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
19622 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
19623 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
19626 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
19627 decl, use remap_type if we want to use the type.
19629 2018-03-09 Martin Sebor <msebor@redhat.com>
19631 PR tree-optimization/84526
19632 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
19634 (builtin_access::generic_overlap): Be prepared to handle non-array
19637 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
19639 PR rtl-optimization/84682
19640 * lra-constraints.c (process_address_1): Check is_address flag
19641 for address constraints.
19642 (process_alt_operands): Likewise.
19643 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
19644 preprocess_constraints.
19645 * recog.h (preprocess_constraints): Add oploc parameter.
19647 * recog.c (preprocess_constraints): Test address_operand for
19648 CT_ADDRESS constraints.
19650 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
19653 * lra-assigns.c (assign_by_spills): Return a flag of reload
19654 assignment failure. Do not process the reload assignment
19655 failures. Do not spill other reload pseudos if they has the same
19657 (lra_assign): Add a return arg. Set up from the result of
19658 assign_by_spills call.
19659 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
19660 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
19661 usage_insns if it is not NULL.
19662 (spill_hard_reg_in_range): New function.
19663 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
19664 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
19665 function prototypes.
19666 (lra_assign): Change prototype.
19667 * lra.c (lra): Add code to deal with fails by splitting hard reg
19670 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
19673 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
19674 Accept complain bool parameter. Only emit errors if it is true.
19675 (arm_parse_cpu_option_name): Likewise.
19676 (arm_target_thumb_only): Adjust callers of the above.
19677 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
19678 prototype to take a default true bool parameter.
19679 (arm_parse_arch_option_name): Likewise.
19681 2018-03-09 David Malcolm <dmalcolm@redhat.com>
19682 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
19686 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
19687 * configure: Regenerate.
19688 * configure.ac ("linker --version-script option"): New.
19689 ("linker soname option"): New.
19691 2018-03-09 Richard Biener <rguenther@suse.de>
19693 PR tree-optimization/84775
19694 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
19695 immediate uses of predicate stmts and mark them modified.
19698 PR tree-optimization/84178
19699 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
19701 (version_loop_for_if_conversion): Delay update_ssa call.
19702 (tree_if_conversion): Delay update_ssa until after predicate
19705 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
19708 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
19709 when the function accesses prior frames.
19711 2018-03-08 Jakub Jelinek <jakub@redhat.com>
19714 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
19715 gen_llsym, otherwise call maybe_gen_llsym.
19717 PR inline-asm/84742
19718 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
19719 has ',' character inside of it.
19721 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
19724 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
19725 as clobbering CC_REGNUM.
19727 2018-03-08 Richard Biener <rguenther@suse.de>
19729 PR middle-end/84552
19730 * tree-scalar-evolution.c: Include tree-into-ssa.h.
19731 (follow_copies_to_constant): Do not follow SSA names registered
19734 2018-03-08 Richard Biener <rguenther@suse.de>
19736 PR tree-optimization/84178
19737 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
19739 (version_loop_for_if_conversion): Delay update_ssa call.
19740 (tree_if_conversion): Delay update_ssa until after predicate
19743 2018-03-08 David Malcolm <dmalcolm@redhat.com>
19745 PR tree-optimization/84178
19746 * tree-if-conv.c (release_bb_predicate): Remove the
19747 the assertion that the stmts have NULL use_ops.
19748 Discard the statements, asserting that they haven't
19749 yet been added to a BB.
19751 2018-03-08 Richard Biener <rguenther@suse.de>
19753 PR tree-optimization/84746
19754 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
19755 (phi_translate): Pass in destination ANTIC_OUT set.
19756 (phi_translate_1): Likewise. For a simplified result lookup
19757 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
19758 (phi_translate_set): Adjust.
19759 (do_pre_regular_insertion): Likewise.
19760 (do_pre_partial_partial_insertion): Likewise.
19762 2018-03-08 Martin Liska <mliska@suse.cz>
19764 PR gcov-profile/84735
19765 * doc/gcov.texi: Document usage of profile files.
19766 * gcov-io.h: Document changes in the format.
19768 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
19772 * dwarf2out.c (struct dw_line_info_table): Update comments for
19774 (FORCE_RESET_NEXT_VIEW): New.
19775 (FORCE_RESETTING_VIEW_P): New.
19776 (RESETTING_VIEW_P): Check for -1 too.
19777 (ZERO_VIEW_P): Likewise.
19778 (new_line_info_table): Force-reset next view.
19779 (dwarf2out_begin_function): Likewise.
19780 (dwarf2out_source_line): Simplify zero_view_p initialization.
19781 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
19782 view directly. Omit view when omitting .loc at line 0.
19784 2018-03-08 Jakub Jelinek <jakub@redhat.com>
19786 PR tree-optimization/84740
19787 * tree-switch-conversion.c (process_switch): Call build_constructors
19788 only if info.phi_count is non-zero.
19790 PR tree-optimization/84739
19791 * tree-tailcall.c (find_tail_calls): Check call arguments against
19792 DECL_ARGUMENTS (current_function_decl) rather than
19793 DECL_ARGUMENTS (func) when checking for tail recursion.
19795 2018-03-07 Jakub Jelinek <jakub@redhat.com>
19797 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
19798 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
19799 Volker Reichelt's entry and add entries for people that perform
19800 GCC fuzzy testing and report numerous bugs.
19802 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
19805 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
19806 readonly data in sdata, if that is disabled.
19807 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
19808 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
19809 -mreadonly-in-sdata option.
19811 2018-03-07 Martin Sebor <msebor@redhat.com>
19813 PR tree-optimization/84468
19814 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
19815 basic block when looking for nul assignment.
19817 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
19820 * except.h (output_function_exception_table): Adjust prototype.
19821 * except.c (output_function_exception_table): Remove FNNAME parameter
19822 and add SECTION parameter. Ouput one part of the table at a time.
19823 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
19824 the first part of the exception table and emit unwind directives.
19825 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
19826 (i386_pe_seh_cold_init): Likewise.
19827 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
19828 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
19829 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
19830 (ix86_output_call_insn): Emit a nop in one more case for SEH.
19831 * config/i386/winnt.c: Include except.h.
19832 (struct seh_frame_state): Add reg_offset, after_prologue and
19833 in_cold_section fields.
19834 (i386_pe_seh_end_prologue): Set seh->after_prologue.
19835 (i386_pe_seh_cold_init): New function.
19836 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
19837 to seh->in_cold_section.
19838 (seh_emit_push): Record the offset of the push.
19839 (seh_emit_save): Record the offet of the save.
19840 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
19841 Test seh->after_prologue to disregard the epilogue.
19842 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
19843 (i386_pe_end_cold_function): New function.
19845 2018-03-07 Jakub Jelinek <jakub@redhat.com>
19848 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
19849 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
19852 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
19854 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
19855 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
19857 PR middle-end/84723
19858 * multiple_target.c: Include tree-inline.h and intl.h.
19859 (expand_target_clones): Diagnose and fail if node->definition and
19860 !tree_versionable_function_p (node->decl).
19862 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
19864 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
19866 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
19867 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
19868 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
19870 2018-03-06 Jakub Jelinek <jakub@redhat.com>
19873 * combine.c (try_combine): Use reg_or_subregno instead of handling
19874 just paradoxical SUBREGs and REGs.
19876 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
19878 * config/arc/arc.c (arc_finalize_pic): Remove function.
19879 (arc_must_save_register): We use single base PIC register, remove
19880 checks to save/restore the PIC register.
19881 (arc_expand_prologue): Likewise.
19882 * config/arc/arc-protos.h (arc_set_default_type_attributes):
19884 (arc_verify_short): Likewise.
19885 (arc_attr_type): Likewise.
19886 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
19887 (walk_stores): Likewise.
19888 (arc_address_cost): Make it static.
19889 (arc_verify_short): Likewise.
19890 (branch_dest): Likewise.
19891 (arc_attr_type): Likewise.
19892 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
19893 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
19894 (arc_final_prescan_insn): Remove inserting the nops due to
19895 hardware hazards. It is done in reorg step.
19896 (insn_length_variant_t): Remove.
19897 (insn_length_parameters_t): Likewise.
19898 (arc_insn_length_parameters): Likewise.
19899 (arc_get_insn_variants): Likewise.
19900 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
19902 2018-03-06 Jakub Jelinek <jakub@redhat.com>
19904 PR inline-asm/84683
19905 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
19908 PR tree-optimization/84687
19909 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
19911 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
19913 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
19915 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
19916 Rename to ppc_speculation_barrier.
19917 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
19918 __builtin_ppc_speculation_barrier.
19920 2018-03-05 Jakub Jelinek <jakub@redhat.com>
19923 * combine.c (combine_simplify_rtx): Don't try to simplify if
19924 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
19927 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
19929 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
19930 to 32 bytes when compiling for POWER9.
19932 2018-03-05 Jakub Jelinek <jakub@redhat.com>
19935 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
19936 regparm >= 3 with no arg reg available also for calls with
19937 flag_force_indirect_call. Pass decl to ix86_function_regparm.
19940 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
19942 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
19944 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
19947 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
19949 2018-03-05 Richard Biener <rguenther@suse.de>
19951 PR tree-optimization/84486
19952 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
19953 When inserting a __builtin_assume_aligned call set the LHS
19954 SSA name alignment info accordingly.
19956 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
19958 PR tree-optimization/84114
19959 * config/aarch64/aarch64.c (aarch64_reassociation_width)
19960 Avoid reassociation of FLOAT_MODE addition.
19962 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
19964 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
19965 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
19966 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
19967 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
19968 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
19969 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
19970 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
19972 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
19973 __builtin_ia32_wbinvd): New builtins.
19974 (SPECIAL_ARGS2): New.
19975 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
19976 (SPECIAL_ARGS2): New.
19977 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
19978 (ix86_valid_target_attribute_inner_p): Ditto.
19979 (ix86_init_mmx_sse_builtins): Add special_args2.
19980 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
19981 TARGET_WBNOINVD_P): New.
19982 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
19983 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
19984 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
19985 * config/i386/immintrin.h (_wbinvd): New intrinsic.
19986 * config/i386/pconfigintrin.h: New file.
19987 * config/i386/wbnoinvdintrin.h: Ditto.
19988 * config/i386/x86intrin.h: Add headers pconfigintrin.h and
19990 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
19992 2018-03-05 Richard Biener <rguenther@suse.de>
19994 PR tree-optimization/84670
19995 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
19997 (BB_VISITED_WITH_VISITED_SUCCS): New define.
19998 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
19999 (compute_antic_aux): Only assert the number of values in ANTIC_IN
20000 doesn't grow if all successors (recursively) were visited at least
20003 2018-03-05 Richard Biener <rguenther@suse.de>
20005 PR tree-optimization/84650
20006 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
20007 if executed in the loop pipeline.
20009 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
20011 * doc/configfiles.texi (Configuration Files): Move info about
20012 conditionalizing $target-protos.h to...
20013 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
20014 differs from $target-protos.h.
20016 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
20017 Chung-Ju Wu <jasonwucj@gmail.com>
20019 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
20020 * config/nds32/nds32-multiple.md (setmemsi): Define.
20021 * config/nds32/nds32-memory-manipulation.c
20022 (nds32_gen_dup_4_byte_to_word_value): New.
20023 (emit_setmem_word_loop): New.
20024 (emit_setmem_byte_loop): New.
20025 (nds32_expand_setmem_loop): New.
20026 (nds32_expand_setmem_loop_v3m): New.
20027 (nds32_expand_setmem_unroll): New.
20028 (nds32_expand_setmem): New.
20030 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
20031 Chung-Ju Wu <jasonwucj@gmail.com>
20033 * config/nds32/nds32-memory-manipulation.c
20034 (nds32_emit_load_store): New.
20035 (nds32_emit_post_inc_load_store): New.
20036 (nds32_emit_mem_move): New.
20037 (nds32_emit_mem_move_block): New.
20038 (nds32_expand_movmemsi_loop_unknown_size): New.
20039 (nds32_expand_movmemsi_loop_known_size): New.
20040 (nds32_expand_movmemsi_loop): New.
20041 (nds32_expand_movmemsi_unroll): New.
20042 (nds32_expand_movmemqi): Rename ...
20043 (nds32_expand_movmemsi): ... to this.
20044 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
20045 (movmemsi): ... to this.
20046 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
20047 (nds32_expand_movmemsi): ... to this.
20049 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
20050 Monk Chiang <sh.chiang04@gmail.com>
20051 Chung-Ju Wu <jasonwucj@gmail.com>
20053 * config/nds32/nds32-protos.h
20054 (nds32_expand_load_multiple): New arguments.
20055 (nds32_expand_store_multiple): Ditto.
20056 (nds32_valid_multiple_load_store): Rename ...
20057 (nds32_valid_multiple_load_store_p): ... to this.
20058 * config/nds32/nds32-memory-manipulation.c
20059 (nds32_expand_load_multiple): Refine implementation.
20060 (nds32_expand_store_multiple): Ditto.
20061 * config/nds32/nds32-multiple.md
20062 (load_multiple): Update nds32_expand_load_multiple interface.
20063 (store_multiple): Update nds32_expand_store_multiple interface.
20064 * config/nds32/nds32-predicates.c
20065 (nds32_valid_multiple_load_store): Rename ...
20066 (nds32_valid_multiple_load_store_p): ... to this and refine
20068 * config/nds32/predicates.md
20069 (nds32_load_multiple_and_update_address_operation): New predicate.
20070 (nds32_store_multiple_and_update_address_operation): New predicate.
20072 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
20073 Chung-Ju Wu <jasonwucj@gmail.com>
20075 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
20076 (combo): New attribute.
20077 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
20079 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
20081 * config/nds32/nds32.opt: Change -mcmodel= default value.
20083 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
20084 Monk Chiang <sh.chiang04@gmail.com>
20085 Chung-Ju Wu <jasonwucj@gmail.com>
20087 * config/nds32/constants.md (unspec_element): New enum.
20088 * config/nds32/constraints.md (Umw): New constraint.
20089 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
20090 * config/nds32/nds32-intrinsic.md: Likewise.
20091 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
20092 (nds32_valid_smw_lwm_base_p): New.
20093 (nds32_output_smw_single_word): New.
20094 (nds32_output_lmw_single_word): New.
20095 (nds32_expand_unaligned_load): New.
20096 (nds32_expand_unaligned_store): New.
20097 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
20098 (nds32_output_smw_single_word): Declare.
20099 (nds32_output_lmw_single_word): Declare.
20100 (nds32_expand_unaligned_load): Declare.
20101 (nds32_expand_unaligned_store): Declare.
20102 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
20103 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
20104 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
20105 NDS32_BUILTIN_UASTORE_DW.
20106 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
20109 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
20110 Kito Cheng <kito.cheng@gmail.com>
20111 Chung-Ju Wu <jasonwucj@gmail.com>
20113 * config/nds32/nds32-intrinsic.c
20114 (nds32_expand_builtin_null_ftype_reg): Delete.
20115 (nds32_expand_builtin_reg_ftype_imm): Ditto.
20116 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
20117 (nds32_read_argument): New.
20118 (nds32_legitimize_target): Ditto.
20119 (nds32_legitimize_argument): Ditto.
20120 (nds32_check_constant_argument): Ditto.
20121 (nds32_expand_unop_builtin): Ditto.
20122 (nds32_expand_unopimm_builtin): Ditto.
20123 (nds32_expand_binop_builtin): Ditto.
20124 (nds32_builtin_decl_impl): Ditto.
20125 (builtin_description): Ditto.
20126 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
20127 (nds32_init_builtins_impl): Ditto.
20128 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
20129 (nds32_builtin_decl): New.
20130 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
20131 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
20133 2018-03-02 Jeff Law <law@redhat.com>
20135 * reorg.c (stop_search_p): Handle DEBUG_INSN.
20136 (redundant_insn, fill_simple_delay_slots): Likewise.
20137 (fill_slots_from_thread): Likewise.
20138 * resource.c (mark_referenced_resources): Likewise.
20139 (mark_set_resources, find_dead_or_set_registers): Likewise.
20141 2018-03-02 Jakub Jelinek <jakub@redhat.com>
20143 * substring-locations.h (format_warning_va): Formatting fix for
20144 ATTRIBUTE_GCC_DIAG.
20145 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
20147 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
20148 * substring-locations.c: Include intl.h.
20149 (format_warning_va): Turned into small wrapper around
20150 format_warning_n_va, renamed to ...
20151 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
20152 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
20154 (format_warning_at_substring_n): New function.
20155 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
20156 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
20157 format_warning_at_substring with just a shorter name instead of
20158 const function pointer.
20159 (fmtwarn_n): New function.
20160 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
20161 appropriate, get rid of all the fmtstr temporaries, move conditionals
20162 with G_() wrapped string literals directly into fmtwarn arguments,
20163 cast dir.len to (int), formatting fixes.
20165 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
20167 * doc/invoke.texi: Remove "Cilk Plus" references.
20169 2018-03-02 Jakub Jelinek <jakub@redhat.com>
20170 Richard Biener <rguenther@suse.de>
20173 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
20174 for error or warning attributes if CALL_FROM_THUNK_P is set.
20177 2018-03-02 Jakub Jelinek <jakub@redhat.com>
20180 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
20181 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
20184 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
20185 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
20187 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
20188 instead of -1U in last predictors element's probability member.
20190 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
20193 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
20194 arguments if they are comparable.
20196 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
20198 PR tree-optimization/84634
20199 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
20200 masks and masked_loop_p with a single loop_masks, making sure it's
20201 null for bb vectorization.
20203 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
20205 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
20206 (vect_analyze_data_ref_access): Use loop->safe_len rather than
20207 loop->force_vectorize to check whether there is no alias.
20209 2018-03-02 Jakub Jelinek <jakub@redhat.com>
20212 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
20214 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
20216 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
20217 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
20218 instead of a loop around prev_real_insn.
20219 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
20222 PR inline-asm/84625
20223 * config/i386/i386.c (ix86_print_operand): Use conditional
20224 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
20227 2018-03-02 Richard Biener <rguenther@suse.de>
20229 PR tree-optimization/84427
20230 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
20231 (bitmap_set_subtract_values): Rewrite to handle multiple
20234 (prune_clobbered_mems): Likewise.
20235 (phi_translate): Take edge instead of pred/phiblock.
20236 (phi_translate_1): Likewise.
20237 (phi_translate_set): Likewise. Insert all translated
20238 exprs for a value into the set, keeping possibly multiple
20239 expressions per value.
20240 (compute_antic_aux): Adjust for phi_translate changes.
20241 When intersecting union the expressions and prune those
20242 not in the final value set, keeping possibly multiple
20243 expressions per value. Do not use value-insertion
20244 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
20245 all expressions. Add verification that the value-sets
20246 only shrink during iteration.
20247 (compute_partial_antic_aux): Adjust for the phi_translate changes.
20248 (do_pre_regular_insertion): Likewise.
20249 (do_pre_partial_partial_insertion): Likewise.
20251 2018-03-02 Richard Biener <rguenther@suse.de>
20254 * config/darwin.c (saved_debug_info_level): New static global.
20255 (darwin_asm_lto_start): Disable debug info generation for LTO out.
20256 (darwin_asm_lto_end): Restore debug info generation settings.
20258 2018-03-01 Martin Liska <mliska@suse.cz>
20261 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
20262 volatile arguments.
20264 2018-03-01 Richard Biener <rguenther@suse.de>
20267 * dwarf2out.c (gen_variable_die): Properly handle late VLA
20268 type annotation with LTO when debug was disabled at compile-time.
20270 2018-03-01 Matthew Fortune <mfortune@gmail.com>
20272 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
20274 (mips_final_postscan_insn): Likewise.
20276 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
20278 PR rtl-optimization/84528
20279 * alias.c (init_alias_target): Add commentary.
20280 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
20281 a unique base value if the frame pointer is not eliminated
20282 to the stack pointer.
20284 2018-03-01 Tom de Vries <tom@codesourcery.com>
20286 PR rtl-optimization/83327
20287 * lra-int.h (hard_regs_spilled_into): Declare.
20288 * lra.c (hard_regs_spilled_into): Define.
20289 (init_reg_info): Init hard_regs_spilled_into.
20290 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
20291 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
20292 (process_bb_lives): Handle hard_regs_spilled_into.
20293 (lra_create_live_ranges_1): Before doing liveness propagation, clear
20294 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
20296 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
20298 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
20299 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
20300 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
20301 * config/rs6000/aix72.h: New file.
20303 2018-02-28 Jakub Jelinek <jakub@redhat.com>
20305 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
20306 instead of warning_at with conditional singular and plural messages
20310 * stor-layout.c (update_alignment_for_field): For
20311 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
20312 && !DECL_PACKED (field), do the alignment update, just use
20313 only desired_align instead of MAX (type_align, desired_align)
20315 (place_field): Don't do known_align < desired_align handling
20316 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
20317 is non-NULL, instead do it after rli->prev_field handling and
20318 only if not within a bitfield word. For DECL_PACKED (field)
20319 use type_align of BITS_PER_UNIT.
20321 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
20323 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
20324 superfluous parentheses and trailing spaces.
20326 2018-02-28 Richard Biener <rguenther@suse.de>
20328 PR tree-optimization/84584
20329 * graphite-scop-detection.c (scop_detection::add_scop): Discard
20330 SCoPs with fake exit edge.
20332 2018-02-28 Martin Liska <mliska@suse.cz>
20335 * timevar.c (timer::print): Fix format to properly print 100%
20338 2018-02-28 Richard Biener <rguenther@suse.de>
20340 PR middle-end/84607
20341 * genmatch.c (capture_info::walk_match): Do not mark
20342 captured expressions without operands as expr_p given
20343 they act more like predicates and should be subject to
20344 "lost tail" side-effect preserving.
20346 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
20348 PR rtl-optimization/81611
20349 * auto-inc-dec.c (attempt_change): Move dead note from
20350 mem_insn if it's the next use of regno
20351 (find_address): Take address use of reg holding
20352 non-incremented value. Add parm to limit search to the named
20354 (merge_in_block): Attempt to use a mem insn that is the next
20355 use of the original regno.
20357 2018-02-27 Martin Sebor <msebor@redhat.com>
20360 * doc/invoke.texi (-Wmissing-attributes): New option.
20361 * print-tree.c (print_node): Handle DECL_UNINLINABLE.
20363 2018-02-27 Martin Sebor <msebor@redhat.com>
20365 PR translation/84207
20366 * diagnostic-core.h (warning_n, error_n, inform_n): Change
20367 n argument to unsigned HOST_WIDE_INT.
20368 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
20369 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
20370 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
20371 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
20373 2018-02-27 Richard Biener <rguenther@suse.de>
20375 PR tree-optimization/84512
20376 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
20377 Do not use the estimate returned from record_stmt_cost for
20378 the scalar iteration cost but sum properly using add_stmt_cost.
20380 2018-02-27 Richard Biener <rguenther@suse.de>
20382 PR tree-optimization/84466
20383 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
20384 Adjust last change to less strictly validate use operands.
20386 2018-02-27 Martin Liska <mliska@suse.cz>
20388 PR gcov-profile/84548
20389 * gcov.c (process_file): Allow partial overlap and consider it
20390 also as group functions.
20391 (output_lines): Properly calculate range of lines for a group.
20393 2018-02-27 Martin Liska <mliska@suse.cz>
20395 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
20396 'ggc' suffixes. Change first column width.
20397 (timer::print): Fix formatting of the column.
20399 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
20401 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
20402 preserve inline entry blocks for the sake of debug inline
20403 entry point markers alone.
20404 (remove_unused_locals): Suggest in comments a better place to
20405 force the preservation of inline entry blocks that are
20406 otherwise unused, but do not preserve them.
20408 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
20410 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
20412 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
20415 * config/i386/constraints.md (Bs): Replace
20416 ix86_indirect_branch_register with
20417 TARGET_INDIRECT_BRANCH_REGISTER.
20419 * config/i386/i386.md (indirect_jump): Likewise.
20420 (tablejump): Likewise.
20421 (*sibcall_memory): Likewise.
20422 (*sibcall_value_memory): Likewise.
20423 Peepholes of indirect call and jump via memory: Likewise.
20424 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
20425 (*sibcall_value_GOT_32): Likewise.
20426 * config/i386/predicates.md (indirect_branch_operand): Likewise.
20427 (GOT_memory_operand): Likewise.
20428 (call_insn_operand): Likewise.
20429 (sibcall_insn_operand): Likewise.
20430 (GOT32_symbol_operand): Likewise.
20431 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
20433 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
20435 PR rtl-optimization/83496
20436 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
20437 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
20438 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
20439 redundant insn, if any.
20440 (relax_delay_slots): Likewise.
20441 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
20443 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
20445 PR tree-optimization/83965
20446 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
20447 that grouped statements are part of a reduction chain. Return
20448 true if the statement is not marked as a reduction itself but
20449 is part of a group.
20450 (vect_recog_dot_prod_pattern): Don't check whether the statement
20451 is part of a group here.
20452 (vect_recog_sad_pattern): Likewise.
20453 (vect_recog_widen_sum_pattern): Likewise.
20455 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
20458 * final.c (rest_of_clean_state): Also look for calls inside sequences.
20460 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
20463 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
20465 (ix86_output_indirect_function_return): New prototype.
20466 (ix86_split_simple_return_pop_internal): Likewise.
20467 * config/i386/i386.c (indirect_return_via_cx): New.
20468 (indirect_return_via_cx_bnd): Likewise.
20469 (indirect_thunk_name): Handle return va CX_REG.
20470 (output_indirect_thunk_function): Create alias for
20471 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
20472 (ix86_output_indirect_jmp): Remove the bool argument.
20473 (ix86_output_indirect_function_return): New function.
20474 (ix86_split_simple_return_pop_internal): Likewise.
20475 * config/i386/i386.md (*indirect_jump): Don't pass false
20476 to ix86_output_indirect_jmp.
20477 (*tablejump_1): Likewise.
20478 (simple_return_pop_internal): Change it to define_insn_and_split.
20479 Call ix86_split_simple_return_pop_internal to split it for
20480 -mfunction-return=.
20481 (simple_return_indirect_internal): Call
20482 ix86_output_indirect_function_return instead of
20483 ix86_output_indirect_jmp.
20485 2018-02-26 Jakub Jelinek <jakub@redhat.com>
20488 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
20489 memset and value initialization afterwards.
20491 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
20493 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
20495 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
20498 * common/config/aarch64/aarch64-common.c
20499 (aarch_option_optimization_table[]): Switch
20500 off fomit-frame-pointer
20502 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
20503 Chung-Ju Wu <jasonwucj@gmail.com>
20505 * config/nds32/nds32-multiple.md (load_multiple): Disallow
20507 (store_multiple): Ditto.
20509 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
20511 * config.gcc: Add --with-cpu support for nds32 target.
20512 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
20513 * config/nds32/nds32.opt: Add -mcpu= option.
20515 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
20517 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
20518 isel=yes): Warn for these deprecated options.
20520 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
20522 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
20523 ISA_2_5_MASKS_EMBEDDED.
20525 2018-02-23 Jakub Jelinek <jakub@redhat.com>
20527 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
20528 p->max as pointers rather than using iterative_hash_expr.
20530 2018-02-23 Carl Love <cel@us.ibm.com>
20532 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
20533 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
20535 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
20536 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
20537 P8V_BUILTIN_VEC_VUNSIGNED2.
20539 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
20542 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
20543 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
20545 * lra-constraints.c (curr_insn_transform): Set up
20546 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
20548 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
20549 * lra-eliminations.c (spill_pseudos): Ditto.
20550 (process_insn_for_elimination): Ditto.
20551 * lra-lives.c (reg_early_clobber_p): Use the new macros.
20552 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
20553 LRA_NON_CLOBBERED_ALT.
20555 2018-02-22 Martin Sebor <msebor@redhat.com>
20557 PR tree-optimization/84480
20558 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
20559 to maybe_diag_stxncpy_trunc. Call it.
20560 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
20561 from gimple_fold_builtin_strcpy. Print inlining stack.
20562 (handle_builtin_stxncpy): Print inlining stack.
20563 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
20565 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
20568 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
20569 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
20570 and -fcheck-pointer-bounds are used together.
20571 (indirect_thunk_prefix): New enum.
20572 (indirect_thunk_need_prefix): New function.
20573 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
20574 "_nt" instead of "_bnd" for NOTRACK prefix.
20575 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
20576 (output_indirect_thunk_function): Likewise.
20578 (ix86_code_end): Update output_indirect_thunk_function calls.
20579 (ix86_output_indirect_branch_via_reg): Replace
20580 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
20581 (ix86_output_indirect_branch_via_push): Likewise.
20582 (ix86_output_function_return): Likewise.
20583 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
20584 incompatible with -fcf-protection=branch and
20585 -fcheck-pointer-bounds.
20587 2018-02-22 Steve Ellcey <sellcey@cavium.com>
20590 * config/aarch64/aarch64.c (aarch64_print_address_internal):
20591 Change gcc_assert call to output_operand_lossage.
20593 2018-02-22 Steve Ellcey <sellcey@cavium.com>
20595 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
20597 2018-02-22 DJ Delorie <dj@redhat.com>
20598 Sebastian Perta <sebastian.perta@renesas.com>
20599 Oleg Endo <olegendo@gcc.gnu.org>
20601 * config/rx/rx.c (rx_rtx_costs): New function.
20602 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
20604 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
20606 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
20608 2018-02-22 Martin Liska <mliska@suse.cz>
20611 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
20612 Add "native" as a possible value.
20614 2018-02-22 Martin Liska <mliska@suse.cz>
20617 * config/i386/i386.c (ix86_option_override_internal):
20618 Add "native" as a possible value for -march and -mtune.
20620 2018-02-22 Jakub Jelinek <jakub@redhat.com>
20623 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
20624 to all type variants.
20626 PR tree-optimization/84503
20627 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
20628 width as info->bitpos + info->bitsize - start.
20629 (merged_store_group::merge_overlapping): Simplify width computation.
20630 (check_no_overlap): New function.
20631 (imm_store_chain_info::try_coalesce_bswap): Compute expected
20632 start + width and last_order of the group, fail if check_no_overlap
20634 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
20635 to group if check_no_overlap fails.
20637 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
20639 * config/rs6000/altivec.md: Delete contraint arguments to
20640 define_expand, define_split, and define_peephole2, and in
20641 define_insn_and_split if always unused.
20642 * config/rs6000/darwin.md: Ditto.
20643 * config/rs6000/dfp.md: Ditto.
20644 * config/rs6000/rs6000.md: Ditto.
20645 * config/rs6000/sync.md: Ditto.
20646 * config/rs6000/vector.md: Ditto.
20647 * config/rs6000/vsx.md: Ditto.
20649 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
20651 * config/rs6000/altivec.md: Write output control strings as braced
20652 blocks instead of double-quoted strings.
20653 * config/rs6000/darwin.md: Ditto.
20654 * config/rs6000/rs6000.md: Ditto.
20655 * config/rs6000/vector.md: Ditto.
20656 * config/rs6000/vsx.md: Ditto.
20658 2018-02-21 Jason Merrill <jason@redhat.com>
20660 PR c++/84314 - ICE with templates and fastcall attribute.
20661 * attribs.c (build_type_attribute_qual_variant): Remove assert.
20663 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
20665 * ipa-cp.c (determine_versionability): Fix comment typos.
20667 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
20670 * ipa-cp.c (determine_versionability): Do not version functions caling
20673 2018-02-21 Martin Liska <mliska@suse.cz>
20676 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
20677 Add "native" as a possible value.
20678 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
20679 the macro when native cpu detection is available.
20681 2018-02-21 Martin Liska <mliska@suse.cz>
20684 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
20685 Add "native" as a possible value.
20686 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
20687 when native cpu detection is available.
20689 2018-02-21 Jakub Jelinek <jakub@redhat.com>
20690 Martin Sebor <msebor@redhat.com>
20692 PR tree-optimization/84478
20693 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
20695 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
20696 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
20697 support which is conservatively correct, for 2 only stay conservative
20698 for maxlen. Formatting and comment capitalization fixes. Add STRICT
20699 argument to the 2 argument get_range_strlen, adjust 6 arg
20700 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
20702 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
20703 (gimple_fold_builtin_strlen): Pass true as last argument to
20706 2018-02-20 Martin Sebor <msebor@redhat.com>
20708 PR middle-end/84095
20709 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
20710 (builtin_memref::set_base_and_offset): Same. Handle inner references.
20711 (builtin_memref::builtin_memref): Factor out parts into
20712 set_base_and_offset and call it.
20714 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
20716 PR middle-end/84406
20717 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
20718 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
20719 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
20720 search at the associated MODE_INT.
20722 2018-02-20 Jeff Law <law@redhat.com>
20724 PR middle-end/82123
20725 PR tree-optimization/81592
20726 PR middle-end/79257
20727 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
20728 for range data rather than using global data.
20729 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
20730 range data rather than using global data.
20731 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
20732 pass it to children as needed.
20733 (struct directive::fmtresult): Similarly.
20734 (struct directive::set_width): Similarly.
20735 (struct directive::set_precision): Similarly.
20736 (format_integer, format_directive, parse_directive): Similarly.
20737 (format_none): Accept unnamed vr_values parameter.
20738 (format_percent, format_floating, format_character): Similarly.
20739 (format_string, format_plain): Similarly.
20740 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
20741 the EVRP range analyzer for range data rather than using global data.
20742 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
20743 gimple-ssa-evrp-analyze.h
20744 (class sprintf_dom_walker): Add after_dom_children member function.
20745 Add evrp_range_analyzer member.
20746 (sprintf_dom_walker::before_dom_children): Call into the EVRP
20747 range analyzer as needed.
20748 (sprintf_dom_walker::after_dom_children): New member function.
20749 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
20751 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
20752 (evrp_range_analyzer::pop_to_marker): Likewise.
20754 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
20756 PR tree-optimization/84419
20757 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
20758 with the required type if its current type is compatible but
20761 2018-02-20 Jakub Jelinek <jakub@redhat.com>
20763 PR middle-end/82004
20764 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
20765 after vectorization.
20767 2018-02-20 Martin Liska <mliska@suse.cz>
20770 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
20771 possible values if we don't have a hint.
20773 2018-02-20 Martin Liska <mliska@suse.cz>
20777 * final.c (shorten_branches): Build align_tab array with one
20779 * opts.c (finish_options): Add alignment option limit check.
20780 (MAX_CODE_ALIGN): Likewise.
20781 (MAX_CODE_ALIGN_VALUE): Likewise.
20782 * doc/invoke.texi: Document maximum allowed option value for
20783 all -falign-* options.
20785 2018-02-19 Jakub Jelinek <jakub@redhat.com>
20788 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
20789 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
20790 * var-tracking.c (emit_note_insn_var_location): Remove all references
20791 to NOTE_INSN_CALL_ARG_LOCATION.
20792 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
20793 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
20794 Use copy_rtx_if_shared.
20795 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
20796 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
20797 (dwarf2out_var_location): Remove handling of
20798 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
20800 * final.c (final_scan_insn): Remove all references to
20801 NOTE_INSN_CALL_ARG_LOCATION.
20802 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
20803 before dumping final insns.
20804 * except.c (emit_note_eh_region_end): Remove all references to
20805 NOTE_INSN_CALL_ARG_LOCATION.
20806 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
20807 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
20808 * config/arc/arc.c (hwloop_optimize): Likewise.
20809 * config/arm/arm.c (create_fix_barrier): Likewise.
20810 * config/s390/s390.c (s390_chunkify_start): Likewise.
20811 * config/sh/sh.c (find_barrier): Likewise.
20812 * config/i386/i386.c (rest_of_insert_endbranch,
20813 ix86_seh_fixup_eh_fallthru): Likewise.
20814 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
20815 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
20816 * config/frv/frv.c (frv_function_prologue): Likewise.
20817 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
20819 (note_outside_basic_block_p): Remove all references to
20820 NOTE_INSN_CALL_ARG_LOCATION.
20821 * gengtype.c (adjust_field_rtx_def): Likewise.
20822 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
20824 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
20825 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
20828 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
20831 PR tree-optimization/84452
20832 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
20833 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
20836 2018-02-19 Martin Liska <mliska@suse.cz>
20839 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
20841 2018-02-19 Martin Liska <mliska@suse.cz>
20842 Richard Sandiford <richard.sandiford@linaro.org>
20844 PR tree-optimization/82491
20845 * gimple-fold.c (get_base_constructor): Make earlier bail out
20848 2018-02-19 Carl Love <cel@us.ibm.com>
20850 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
20851 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
20853 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
20854 P8V_BUILTIN_VEC_NEG.
20856 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
20858 * config/rl78/rl78.md (movdf): New define expand.
20860 2018-02-19 Martin Liska <mliska@suse.cz>
20863 * doc/invoke.texi: Fix typo.
20864 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
20866 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
20868 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
20869 handle rs6000_single_float and rs6000_double_float specially for
20872 2018-02-16 Jeff Law <law@redhat.com>
20874 * config/rx/rx.c (add_pop_cfi_notes): New function.;
20875 (pop_regs): Use it.
20877 2018-02-16 Jakub Jelinek <jakub@redhat.com>
20880 * ipa-inline.c (inline_small_functions): Fix a typo.
20882 2018-02-16 Nathan Sidwell <nathan@acm.org>
20884 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
20886 2018-02-16 Carl Love <cel@us.ibm.com>
20888 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
20889 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
20890 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
20891 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
20892 expansion to P8V_BUILTIN_VEC_FLOAT2.
20894 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
20896 PR rtl-optimization/70023
20897 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
20898 src_regno into account.
20900 2018-02-16 Carl Love <cel@us.ibm.com>
20902 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
20903 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
20904 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
20905 * config/rs6000/rs6000.c: Remove case statements for
20906 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
20907 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
20908 and P9V_BUILTIN_VEC_VINSERT4B.
20909 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
20910 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
20911 * config/rs6000/vsx.md:
20912 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
20915 2018-02-16 Carl Love <cel@us.ibm.com>
20917 * config/rs6000/altivec.h: Add builtin names vec_extract4b
20919 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
20921 * config/rs6000/rs6000-c.c: Add the definitions for
20922 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
20923 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
20924 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
20925 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
20926 definition for insert4b and define insn *insert3b_internal.
20927 * doc/extend.texi: Add documentation for vec_extract4b.
20929 2018-02-16 Nathan Sidwell <nathan@acm.org>
20931 * doc/extend.texi (Backwards Compatibility): Mention friend
20932 injection. Note for-scope is deprecated.
20933 * doc/invoke.texi (-ffriend-injection): Deprecate.
20935 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
20937 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
20938 that moved to I2, also allow destinations that are a paradoxical
20939 subreg (instead of a normal reg).
20941 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
20944 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
20947 2018-02-16 Richard Biener <rguenther@suse.de>
20949 PR tree-optimization/84037
20950 PR tree-optimization/84016
20952 * config/i386/i386.c (ix86_builtin_vectorization_cost):
20953 Adjust vec_construct for the fact we need additional higher latency
20954 128bit inserts for AVX256 and AVX512 vector builds.
20955 (ix86_add_stmt_cost): Scale vector construction cost for
20958 2018-02-16 Richard Biener <rguenther@suse.de>
20960 PR tree-optimization/84417
20961 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
20962 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
20963 (non_rewritable_lvalue_p): Likewise, use poly-ints.
20965 2018-02-16 Martin Liska <mliska@suse.cz>
20968 * internal-fn.def (ASAN_CHECK): Set proper flags.
20969 (ASAN_MARK): Likewise.
20971 2018-02-16 Julia Koval <julia.koval@intel.com>
20973 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
20974 from PTA_CANNONLAKE.
20976 2018-02-16 Jakub Jelinek <jakub@redhat.com>
20979 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
20980 Use ++iter rather than iter++ for std::list iterators.
20981 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
20982 defer deleting them until all nodes in the forest are processed. Do
20983 free even leaf nodes. Change to_process into auto_vec.
20986 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
20987 * vec.h (vec_default_construct): Use memset instead of placement new
20988 if BROKEN_VALUE_INITIALIZATION is defined.
20989 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
20990 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
20993 PR rtl-optimization/83723
20994 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
20995 * lra.c (lra_substitute_pseudo): Likewise. If true, use
20996 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
20998 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
21000 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
21002 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
21004 PR rtl-optimization/81443
21005 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
21006 from inner REGs to paradoxical SUBREGs.
21008 2018-02-16 Richard Biener <rguenther@suse.de>
21010 PR tree-optimization/84399
21011 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
21012 For operands we can analyze at their definition make sure we can
21013 analyze them at each use as well.
21015 2018-02-16 Richard Biener <rguenther@suse.de>
21017 PR tree-optimization/84190
21018 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
21019 volatile accesses if the decl isn't volatile.
21021 2018-02-15 Jason Merrill <jason@redhat.com>
21023 PR c++/84314 - ICE with templates and fastcall attribute.
21024 * attribs.c (build_type_attribute_qual_variant): Don't clobber
21025 TYPE_CANONICAL on an existing type.
21027 2018-02-15 Jakub Jelinek <jakub@redhat.com>
21029 PR tree-optimization/84383
21030 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
21031 dstoff nor call operand_equal_p if dstbase is NULL.
21033 PR tree-optimization/84334
21034 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
21035 also a CONSTANT_CLASS_P, punt.
21037 2018-02-14 Jim Wilson <jimw@sifive.com>
21039 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
21040 first SMALL_OPERAND check. New local min_second_step. Move assert
21041 to where locals are set. Add TARGET_RVC support.
21042 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
21044 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
21046 * doc/invoke.texi: Correct -Wformat-overflow code sample.
21048 2018-02-14 Martin Sebor <msebor@redhat.com>
21050 PR tree-optimization/83698
21051 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
21052 arrays constrain the offset range to their bounds.
21053 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
21054 (builtin_access::overlap): Avoid setting the size of overlap if it's
21056 (maybe_diag_overlap): Also consider arrays when deciding what values
21057 of offsets to include in diagnostics.
21059 2018-02-14 Martin Sebor <msebor@redhat.com>
21062 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
21063 that correspond to the kind of a declaration.
21065 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
21068 * config/pa/pa.md: Load address of PIC label using the linkage table
21069 if the label is nonlocal.
21071 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
21073 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
21074 warning message if user requests -maltivec=be.
21075 * doc/invoke.texi: Document deprecation of -maltivec=be.
21077 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
21080 * config/rs6000/rs6000-c.c: Update definitions for
21081 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
21082 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
21084 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
21087 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
21088 add _get_ssp intrinsics. Remove argument from
21089 __builtin_ia32_rdssp[d|q].
21090 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
21091 * config/i386/i386-builtin.def: Remove argument from
21092 __builtin_ia32_rdssp[d|q].
21093 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
21094 ix86_expand_special_args_builtin for _rdssp[d|q].
21095 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
21096 Clear register before usage.
21097 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
21098 Add documentation for new _get_ssp and _inc_ssp intrinsics.
21100 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
21102 PR tree-optimization/84357
21103 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
21104 operand 1 of an ARRAY_REF too.
21106 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
21109 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
21110 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
21112 (set_of_reg): New struct.
21113 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
21114 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
21115 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
21117 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
21118 Split into bitclr, bitset, bitinvert patterns if appropriate.
21119 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
21120 use rx_fuse_in_memory_bitop.
21121 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
21122 to named insn, correct maximum insn length.
21124 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
21127 * machmode.def: Define a complex mode for PARTIAL_INT.
21128 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
21130 * doc/rtl.texi: Document CSPImode.
21131 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
21133 (msp430_hard_regno_nregs_with_padding): Likewise.
21135 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
21138 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
21140 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
21142 PR rtl-optimization/84169
21143 * combine.c (try_combine): New variable split_i2i3. Set it to true if
21144 we generated a parallel as new i3 and we split that to new i2 and i3
21145 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
21146 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
21147 those to i2, not i1. Partially rewrite this scan code.
21149 2018-02-13 Jakub Jelinek <jakub@redhat.com>
21152 * stor-layout.c (place_field): For variable length fields, adjust
21153 offset_align afterwards not just based on the field's alignment,
21154 but also on the size.
21156 PR middle-end/84309
21157 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
21158 of exps and logs in the use_exp2 case.
21160 2018-02-13 Jeff Law <law@redhat.com>
21162 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
21163 entry for "vector".
21165 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
21168 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
21172 * common.opt (gas-loc-support, gas-locview-support): New.
21173 (ginline-points, ginternal-reset-location-views): New.
21174 * doc/invoke.texi: Document them. Use @itemx where intended.
21175 (gvariable-location-views): Adjust.
21176 * target.def (reset_location_view): New.
21177 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
21178 (TARGET_RESET_LOCATION_VIEW): New.
21179 * doc/tm.texi: Rebuilt.
21180 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
21181 (dwarf2out_default_as_locview_support): New.
21182 (output_asm_line_debug_info): Use option variables.
21183 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
21184 (output_loc_list): Likewise.
21185 (add_high_low_attributes): Check option variables.
21186 Don't output entry view attribute in strict mode.
21187 (gen_inlined_subroutine_die): Check option variables.
21188 (dwarf2out_inline_entry): Likewise.
21189 (init_sections_and_labels): Likewise.
21190 (dwarf2out_early_finish): Likewise.
21191 (maybe_reset_location_view): New, from...
21192 (dwarf2out_var_location): ... here. Call it.
21193 * debug.h (dwarf2out_default_as_loc_support): Declare.
21194 (dwarf2out_default_as_locview_support): Declare.
21195 * hooks.c (hook_int_rtx_insn_0): New.
21196 * hooks.h (hook_int_rtx_insn_0): Declare.
21197 * toplev.c (process_options): Take -gas-loc-support and
21198 -gas-locview-support from dwarf2out. Enable
21199 -gvariable-location-views by default only with locview
21200 assembler support. Enable -ginternal-reset-location-views by
21201 default only if the target defines the corresponding hook.
21202 Enable -ginline-points by default if location views are
21203 enabled; force it disabled if statement frontiers are
21205 * tree-inline.c (expand_call_inline): Check option variables.
21206 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
21208 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
21210 PR tree-optimization/84321
21211 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
21212 handling. Also check whether the anti-range contains any values
21213 that satisfy the mask; switch to a VR_RANGE if not.
21215 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
21218 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
21220 2018-02-13 Martin Jambor <mjambor@suse.cz>
21223 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
21224 of call statements, also set location of a load to a temporary.
21226 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
21228 * config/rl78/rl78.c (add_vector_labels): New function.
21229 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
21230 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
21231 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
21232 which checks that no arguments are passed.
21233 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
21234 * doc/extend.texi: Documentation for the new attribute.
21236 2018-02-13 Andreas Schwab <schwab@suse.de>
21238 * config/riscv/linux.h (CPP_SPEC): Define.
21240 2018-02-13 Jakub Jelinek <jakub@redhat.com>
21243 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
21244 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
21245 OPTION_MASK_ISA_AES as first argument to def_builtin_const
21246 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
21247 instead of OPTION_MASK_ISA_PCLMUL as first argument to
21248 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
21249 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
21250 temporarily for AES and PCLMUL builtins.
21252 PR tree-optimization/84339
21253 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
21254 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
21257 PR middle-end/84309
21258 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
21259 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
21260 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
21262 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
21264 * omp-simd-clone.h: New file.
21265 * omp-simd-clone.c: Include omp-simd-clone.h.
21266 (expand_simd_clones): No longer static.
21267 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
21268 cgraph.h and omp-simd-clone.h.
21269 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
21270 (vect_recog_widen_shift_pattern): Formatting fix.
21271 (vect_pattern_recog_1): Don't check optab for calls.
21274 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
21275 operands[2] into a REG before using gen_lowpart on it.
21277 2018-02-12 Jeff Law <law@redhat.com>
21280 * config/sh/sh.c (find_barrier): Consider a sibling call
21283 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
21284 successfully back substituting a reg.
21286 2018-02-12 Richard Biener <rguenther@suse.de>
21288 PR tree-optimization/84037
21289 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
21290 parameter, move visited init to caller.
21291 (vect_slp_analyze_operations): Separate cost from validity
21292 check, initialize visited once for all instances.
21293 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
21295 * tree-vect-stmts.c (vect_model_simple_cost): Make early
21297 (vect_model_promotion_demotion_cost): Likewise.
21298 (vectorizable_bswap): Guard cost modeling with !slp_node
21299 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
21301 (vectorizable_call): Likewise.
21302 (vectorizable_conversion): Likewise.
21303 (vectorizable_assignment): Likewise.
21304 (vectorizable_shift): Likewise.
21305 (vectorizable_operation): Likewise.
21306 (vectorizable_store): Likewise.
21307 (vectorizable_load): Likewise.
21308 (vectorizable_condition): Likewise.
21309 (vectorizable_comparison): Likewise.
21311 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
21314 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
21315 (ASAN_MARK): Fix fnspec to account for return value, change pointer
21316 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
21318 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
21320 PR middle-end/83665
21321 * params.def (inline-min-speedup): Increase from 8 to 15.
21322 (max-inline-insns-auto): Decrease from 40 to 30.
21323 * ipa-split.c (consider_split): Add some buffer for function to
21324 be considered inlining candidate.
21325 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
21328 2018-02-12 Richard Biener <rguenther@suse.de>
21330 PR tree-optimization/84037
21331 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
21332 matched stmts if we cannot swap the non-matched ones.
21334 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
21336 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
21337 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
21338 _mm_maskz_scalef_round_ss): New intrinsics.
21339 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
21340 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
21341 __builtin_ia32_scalefss_round): Remove.
21342 (__builtin_ia32_scalefsd_mask_round,
21343 __builtin_ia32_scalefss_mask_round): New intrinsics.
21344 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
21345 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
21346 ((match_operand:VF_128 2 "<round_nimm_predicate>"
21347 "<round_constraint>")): Changed to ...
21348 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
21349 "<round_scalar_constraint>")): ... this.
21350 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
21351 %0, %1, %2<round_op3>}"): Changed to ...
21352 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
21353 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
21354 %2<round_scalar_mask_op3>}"): ... this.
21355 * config/i386/subst.md (round_scalar_nimm_predicate): New.
21357 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
21359 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
21360 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
21361 (_mm_maskz_sqrt_round_ss): New intrinsics.
21362 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
21363 (__builtin_ia32_sqrtsd_mask_round)
21364 (__builtin_ia32_sqrtss_mask_round): New builtins.
21365 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
21366 (__builtin_ia32_sqrtss_round): Remove.
21367 (__builtin_ia32_sqrtsd_mask_round)
21368 (__builtin_ia32_sqrtss_mask_round): New builtins.
21369 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
21370 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
21371 ((match_operand:VF_128 1 "vector_operand"
21372 "xBm,<round_constraint>")): Changed to ...
21373 ((match_operand:VF_128 1 "vector_operand"
21374 "xBm,<round_scalar_constraint>")): ... this.
21375 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
21376 %0, %2, %<iptr>1<round_op3>}): Changed to ...
21377 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
21378 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
21379 %<iptr>1<round_scalar_mask_op3>}): ... this.
21380 ((set_attr "prefix" "<round_prefix>")): Changed to ...
21381 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
21383 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
21386 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
21387 Cast vec_cmpeq result to correct type.
21388 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
21389 Cast vec_cmpgt result to correct type.
21391 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
21393 * final.c (final_scan_insn_1): Renamed from...
21394 (final_scan_insn): ... this. New wrapper, to recover
21395 seen from the outermost call in recursive ones.
21396 * config/sparc/sparc.c (output_return): Drop seen from call.
21397 (output_sibcall): Likewise.
21398 * config/visium/visium.c (output_branch): Likewise.
21400 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
21402 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
21405 2018-02-10 Alan Modra <amodra@gmail.com>
21408 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
21409 Specify LR as an input.
21411 2018-02-10 Jakub Jelinek <jakub@redhat.com>
21414 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
21415 remove_member_access_dummy_vars): New functions.
21416 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
21417 lower_omp_1, execute_lower_omp): Use them.
21419 PR rtl-optimization/84308
21420 * shrink-wrap.c (spread_components): Release todo vector.
21422 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
21424 PR rtl-optimization/57193
21425 * ira-color.c (struct allocno_color_data): Add member
21426 conflict_allocno_hard_prefs.
21427 (update_conflict_allocno_hard_prefs): New.
21428 (bucket_allocno_compare_func): Add a preference based on
21429 conflict_allocno_hard_prefs.
21430 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
21431 (color_allocnos): Remove a dead code. Initiate
21432 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
21434 2018-02-09 Jakub Jelinek <jakub@redhat.com>
21437 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
21438 constraint from =wa to wa. Avoid a subreg on the output operand,
21439 instead use a pseudo and subreg it in a move.
21440 (p9_xxbrd_<mode>): Changed to ...
21441 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
21442 (p9_xxbrd_v2df): New expander.
21443 (p9_xxbrw_<mode>): Changed to ...
21444 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
21445 (p9_xxbrw_v4sf): New expander.
21447 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
21449 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
21451 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
21454 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
21455 multiply in 32-bit mode.
21456 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
21457 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
21460 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
21462 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
21463 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
21464 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
21465 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
21467 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
21470 * dwarf2out.c (is_trivial_indirect_ref): New function.
21471 (dwarf2out_late_global_decl): Do not generate a location
21472 attribute for variables that have a non-trivial DECL_VALUE_EXPR
21473 and that are not defined in the current unit.
21475 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
21477 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
21478 instead of a libcall for UNORDERED.
21480 2018-02-09 Tamar Christina <tamar.christina@arm.com>
21483 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
21484 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
21486 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
21489 * config/s390/s390.c (s390_set_current_function): Invoke
21490 s390_indirect_branch_settings also if fndecl didn't change.
21492 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
21494 * config/rs6000/rs6000.md (blockage): Set length to zero.
21496 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
21498 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
21500 2018-02-09 Jakub Jelinek <jakub@redhat.com>
21503 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
21504 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
21508 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
21509 PARALLEL incoming that failed vt_get_decl_and_offset check.
21511 PR middle-end/84237
21512 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
21513 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
21515 (get_variable_section): For decls in named .bss* sections pass true as
21516 second argument to bss_initializer_p.
21518 2018-02-09 Marek Polacek <polacek@redhat.com>
21519 Jakub Jelinek <jakub@redhat.com>
21522 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
21523 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
21524 Sync some changes from cxx_fold_indirect_ref.
21526 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
21528 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
21530 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
21531 (BLOCK_INLINE_ENTRY_LABEL): New.
21532 (dwarf2out_var_location): Disregard inline entry markers.
21533 (inline_entry_data): New struct.
21534 (inline_entry_data_hasher): New hashtable type.
21535 (inline_entry_data_hasher::hash): New.
21536 (inline_entry_data_hasher::equal): New.
21537 (inline_entry_data_table): New variable.
21538 (add_high_low_attributes): Add DW_AT_entry_pc and
21539 DW_AT_GNU_entry_view attributes if a pending entry is found
21540 in inline_entry_data_table. Add old entry_pc attribute only
21541 if debug nonbinding markers are disabled.
21542 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
21543 markers are enabled.
21544 (block_within_block_p, dwarf2out_inline_entry): New.
21545 (dwarf2out_finish): Check that no entries remained in
21546 inline_entry_data_table.
21547 * final.c (reemit_insn_block_notes): Handle inline entry notes.
21548 (final_scan_insn, notice_source_line): Likewise.
21549 (rest_of_clean_state): Skip inline entry markers.
21550 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
21552 * gimple.c (gimple_build_debug_inline_entry): New.
21553 * gimple.h (enum gimple_debug_subcode): Add
21554 GIMPLE_DEBUG_INLINE_ENTRY.
21555 (gimple_build_debug_inline_entry): Declare.
21556 (gimple_debug_inline_entry_p): New.
21557 (gimple_debug_nonbind_marker_p): Adjust.
21558 * insn-notes.def (INLINE_ENTRY): New.
21559 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
21560 inline entry marker notes.
21561 (print_insn): Likewise.
21562 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
21563 (INSN_DEBUG_MARKER_KIND): Likewise.
21564 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
21565 * tree-inline.c (expand_call_inline): Build and insert
21566 debug_inline_entry stmt.
21567 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
21568 inline entry blocks early, if nonbind markers are enabled.
21569 (dump_scope_block): Dump fragment info.
21570 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
21571 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
21572 (gimple_build_debug_inline_entry): New.
21573 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
21574 Enable/disable inline entry points too.
21575 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
21576 (DEBUG_INSN): Describe inline entry markers.
21578 * common.opt (gvariable-location-views): New.
21579 (gvariable-location-views=incompat5): New.
21580 * config.in: Rebuilt.
21581 * configure: Rebuilt.
21582 * configure.ac: Test assembler for view support.
21583 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
21584 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
21585 * dwarf2out.c (var_loc_view): New typedef.
21586 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
21587 (dwarf2out_locviews_in_attribute): New.
21588 (dwarf2out_locviews_in_loclist): New.
21589 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
21590 (enum dw_line_info_opcode): Add LI_adv_address.
21591 (struct dw_line_info_table): Add view.
21592 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
21593 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
21594 (zero_view_p): New variable.
21595 (ZERO_VIEW_P): New macro.
21596 (output_asm_line_debug_info): New.
21597 (struct var_loc_node): Add view.
21598 (add_AT_view_list, AT_loc_list): New.
21599 (add_var_loc_to_decl): Add view param. Test it against last.
21600 (new_loc_list): Add view params. Record them.
21601 (AT_loc_list_ptr): Handle loc and view lists.
21602 (view_list_to_loc_list_val_node): New.
21603 (print_dw_val): Handle dw_val_class_view_list.
21604 (size_of_die): Likewise.
21605 (value_format): Likewise.
21606 (loc_list_has_views): New.
21607 (gen_llsym): Set vl_symbol too.
21608 (maybe_gen_llsym, skip_loc_list_entry): New.
21609 (dwarf2out_maybe_output_loclist_view_pair): New.
21610 (output_loc_list): Output view list or entries too.
21611 (output_view_list_offset): New.
21612 (output_die): Handle dw_val_class_view_list.
21613 (output_dwarf_version): New.
21614 (output_compilation_unit_header): Use it.
21615 (output_skeleton_debug_sections): Likewise.
21616 (output_rnglists, output_line_info): Likewise.
21617 (output_pubnames, output_aranges): Update version comments.
21618 (output_one_line_info_table): Output view numbers in asm comments.
21619 (dw_loc_list): Determine current endview, pass it to new_loc_list.
21620 Call maybe_gen_llsym.
21621 (loc_list_from_tree_1): Adjust.
21622 (add_AT_location_description): Create view list attribute if
21623 needed, check it's absent otherwise.
21624 (convert_cfa_to_fb_loc_list): Adjust.
21625 (maybe_emit_file): Call output_asm_line_debug_info for test.
21626 (dwarf2out_var_location): Reset views as needed. Precompute
21627 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
21628 attribute. Set view.
21629 (new_line_info_table): Reset next view.
21630 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
21631 (dwarf2out_source_line): Likewise. Output view resets and labels to
21632 the assembler, or select appropriate line info opcodes.
21633 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
21634 (optimize_string_length): Catch it. Adjust.
21635 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
21636 dw_val_class_view_list, and remove it if no longer needed.
21637 (hash_loc_list): Hash view numbers.
21638 (loc_list_hasher::equal): Compare them.
21639 (optimize_location_lists): Check whether a view list symbol is
21640 needed, and whether the locview attribute is present, and
21641 whether they match. Remove the locview attribute if no longer
21643 (index_location_lists): Call skip_loc_list_entry for test.
21644 (dwarf2out_finish): Call output_asm_line_debug_info for test.
21645 Use output_dwarf_version.
21646 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
21647 (struct dw_val_node): Add val_view_list.
21648 * final.c (SEEN_NEXT_VIEW): New.
21649 (set_next_view_needed): New.
21650 (clear_next_view_needed): New.
21651 (maybe_output_next_view): New.
21652 (final_start_function): Rename to...
21653 (final_start_function_1): ... this. Take pointer to FIRST,
21654 add SEEN parameter. Emit param bindings in the initial view.
21655 (final_start_function): Reintroduce SEEN-less interface.
21656 (final): Rename to...
21657 (final_1): ... this. Take SEEN parameter. Output final pending
21658 next view at the end.
21659 (final): Reintroduce seen-less interface.
21660 (final_scan_insn): Output pending next view before switching
21661 sections or ending a block. Mark the next view as needed when
21662 outputting variable locations. Notify debug backend of section
21663 changes, and of location view changes.
21664 (rest_of_handle_final): Adjust.
21665 * toplev.c (process_options): Autodetect value for debug variable
21666 location views option. Warn on incompat5 without -gdwarf-5.
21667 * doc/invoke.texi (gvariable-location-views): New.
21668 (gvariable-location-views=incompat5): New.
21669 (gno-variable-location-views): New.
21671 2018-02-08 David Malcolm <dmalcolm@redhat.com>
21673 PR tree-optimization/84136
21674 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
21675 that the result of find_edge is non-NULL.
21677 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
21680 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
21681 storing integer register in SImode. Fix cost of 256 and 512
21682 byte aligned SSE register store.
21684 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
21686 * config/i386/i386.c (ix86_multiplication_cost): Fix
21687 multiplication cost for TARGET_AVX512DQ.
21689 2018-02-08 Marek Polacek <polacek@redhat.com>
21691 PR tree-optimization/84238
21692 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
21695 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
21697 PR tree-optimization/84265
21698 * tree-vect-stmts.c (vectorizable_store): Don't treat
21699 VMAT_CONTIGUOUS accesses as grouped.
21700 (vectorizable_load): Likewise.
21702 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
21704 PR tree-optimization/81635
21705 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
21706 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
21707 (test_round_for_mask): New functions.
21708 (wide_int_cc_tests): Call test_round_for_mask.
21709 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
21710 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
21711 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
21712 range returned by get_range_info.
21714 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
21717 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
21718 * symtab.c: Include builtins.h
21719 (symtab_node::output_to_lto_symbol_table_p): Move here
21720 from lto-streamer-out.c:output_symbol_p.
21721 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
21722 (output_symbol_p): Move all logic to symtab.c
21723 (produce_symtab): Update.
21725 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
21727 * config/s390/s390-opts.h (enum indirect_branch): Define.
21728 * config/s390/s390-protos.h (s390_return_addr_from_memory)
21729 (s390_indirect_branch_via_thunk)
21730 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
21731 (enum s390_indirect_branch_type): Define.
21732 * config/s390/s390.c (struct s390_frame_layout, struct
21733 machine_function): Remove.
21734 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
21735 (indirect_branch_table_label_no, indirect_branch_table_name):
21737 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
21738 (enum s390_indirect_branch_option): Define.
21739 (s390_return_addr_from_memory): New function.
21740 (s390_handle_string_attribute): New function.
21741 (s390_attribute_table): Add new attribute handler.
21742 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
21743 (s390_indirect_branch_via_thunk): New function.
21744 (s390_indirect_branch_via_inline_thunk): New function.
21745 (s390_function_ok_for_sibcall): When jumping via thunk disallow
21746 sibling call optimization for non z10 compiles.
21747 (s390_emit_call): Force indirect branch target to be a single
21748 register. Add r1 clobber for non-z10 compiles.
21749 (s390_emit_epilogue): Emit return jump via return_use expander.
21750 (s390_reorg): Handle JUMP_INSNs as execute targets.
21751 (s390_option_override_internal): Perform validity checks for the
21752 new command line options.
21753 (s390_indirect_branch_attrvalue): New function.
21754 (s390_indirect_branch_settings): New function.
21755 (s390_set_current_function): Invoke s390_indirect_branch_settings.
21756 (s390_output_indirect_thunk_function): New function.
21757 (s390_code_end): Implement target hook.
21758 (s390_case_values_threshold): Implement target hook.
21759 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
21761 * config/s390/s390.h (struct s390_frame_layout)
21762 (struct machine_function): Move here from s390.c.
21763 (TARGET_INDIRECT_BRANCH_NOBP_RET)
21764 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
21765 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
21766 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
21767 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
21768 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
21769 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
21770 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
21771 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
21772 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
21773 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
21774 (mnemonic attribute): Add values which aren't recognized
21776 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
21777 pattern for branch conversion. Fix mnemonic attribute.
21778 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
21779 indirect branch via thunk if requested.
21780 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
21781 ("*indirect_jump"): Disable for branch conversion using out of
21783 ("indirect_jump_via_thunk<mode>_z10")
21784 ("indirect_jump_via_thunk<mode>")
21785 ("indirect_jump_via_inlinethunk<mode>_z10")
21786 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
21787 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
21788 ("casesi_jump_via_inlinethunk<mode>_z10")
21789 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
21790 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
21791 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
21792 ("*indirect2_jump"): Disable for branch conversion.
21793 ("casesi_jump"): Turn into expander and expand patterns for branch
21795 ("return_use"): New expander.
21796 ("*return"): Emit return via thunk and rename it to ...
21797 ("*return<mode>"): ... this one.
21798 * config/s390/s390.opt: Add new options and and enum for the
21801 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
21803 * lra-constraints.c (match_reload): Unconditionally use
21804 gen_lowpart_SUBREG, rather than selecting between that
21805 and equivalent gen_rtx_SUBREG code.
21807 2018-02-08 Richard Biener <rguenther@suse.de>
21809 PR tree-optimization/84233
21810 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
21811 changed flag instead of boguously re-using phi_inserted.
21813 2018-02-08 Martin Jambor <mjambor@suse.cz>
21815 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
21816 static local variables.
21818 2018-02-08 Richard Biener <rguenther@suse.de>
21820 PR tree-optimization/84278
21821 * tree-vect-stmts.c (vectorizable_store): When looking for
21822 smaller vector types to perform grouped strided loads/stores
21823 make sure the mode is supported by the target.
21824 (vectorizable_load): Likewise.
21826 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
21828 * config/aarch64/aarch64.c (aarch64_components_for_bb):
21829 Increase LDP/STP opportunities by adding adjacent callee-saves.
21831 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
21833 PR rtl-optimization/84068
21834 PR rtl-optimization/83459
21835 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
21837 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
21839 PR tree-optimization/84224
21840 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
21841 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
21842 non-zero arguments.
21844 2018-02-07 Iain Sandoe <iain@codesourcery.com>
21847 * config/rs6000/altivec.md (*restore_world): Remove LR use.
21848 * config/rs6000/predicates.md (restore_world_operation): Adjust op
21849 count, remove one USE.
21851 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
21853 * doc/install.texi (Configuration): Document the
21854 --with-long-double-format={ibm,ieee} PowerPC configuration
21858 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
21859 Convert from define_expand to be define_insn_and_split. Rework
21860 float/double/_Float128 conversions to QI/HI/SImode to work with
21861 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
21862 conversions to QI/HImode types did a store and then a load to
21863 truncate the value. For conversions to VSX registers, don't split
21864 the insn, instead emit the code directly. Use the code iterator
21865 any_fix to combine signed and unsigned conversions.
21866 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
21867 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
21868 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
21869 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
21870 (fix_<mode>di2_hw): Likewise.
21871 (fixuns_<mode>di2_hw): Likewise.
21872 (fix_<mode>si2_hw): Likewise.
21873 (fixuns_<mode>si2_hw): Likewise.
21874 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
21875 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
21876 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
21877 fix<uns>_trunc<SFDF:mode>si2_p8.
21878 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer used.
21879 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
21880 (fix<uns>_<mode>_mem): Likewise.
21881 (fctiw<u>z_<mode>_mem): Likewise.
21882 (fix<uns>_<mode>_mem): Likewise.
21883 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
21884 the register allocator from doing a direct move to the GPRs to do
21885 a store, and instead use the ISA 3.0 store byte/half-word from
21886 vector register instruction. For IEEE 128-bit floating point,
21887 also optimize stores of 32-bit ints.
21888 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
21890 2018-02-07 Alan Hayward <alan.hayward@arm.com>
21892 * genextract.c (push_pathstr_operand): New function to support
21894 (walk_rtx): Call push_pathstr_operand.
21895 (print_path): Support [a-zA-Z].
21897 2018-02-07 Richard Biener <rguenther@suse.de>
21899 PR tree-optimization/84037
21900 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
21901 (cse_and_gimplify_to_preheader): Declare.
21902 (vect_get_place_in_interleaving_chain): Likewise.
21903 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
21905 (_loop_vec_info::~_loop_vec_info): Delete it.
21906 (cse_and_gimplify_to_preheader): New function.
21907 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
21908 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
21909 (vectorizable_load): Likewise. For grouped stores always base
21910 the IV on the first element.
21911 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
21912 condition before gimplifying.
21914 2018-02-07 Jakub Jelinek <jakub@redhat.com>
21916 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
21917 *DIV_EXPR and *MOD_EXPR.
21919 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
21922 * config/i386/i386.c (ix86_option_override_internal): Mask out
21923 the CF_SET bit when checking -fcf-protection.
21925 2018-02-07 Tom de Vries <tom@codesourcery.com>
21928 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
21931 2018-02-07 Richard Biener <rguenther@suse.de>
21933 PR tree-optimization/84204
21934 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
21937 PR tree-optimization/84205
21938 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
21939 special-case isl_ast_op_zdiv_r.
21941 PR tree-optimization/84223
21942 * graphite-scop-detection.c (gather_bbs::before_dom_children):
21943 Only add conditions from within the region.
21944 (gather_bbs::after_dom_children): Adjust.
21946 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
21949 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
21950 * config/avr/avr.md: Only post-reload split REG-REG moves if
21951 either register is GENERAL_REG_P.
21953 2018-02-07 Jakub Jelinek <jakub@redhat.com>
21955 PR tree-optimization/84235
21956 * tree-ssa-scopedtables.c
21957 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
21958 if the subtraction is performed in floating point type where NaNs are
21959 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
21960 build 1. Formatting fix.
21962 2018-02-06 Jakub Jelinek <jakub@redhat.com>
21965 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
21966 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
21967 and skip it regardless of bb boundaries. Use CALL_P macro,
21968 don't test INSN_P (insn) together with CALL_P or JUMP_P check
21969 unnecessarily, formatting fix.
21971 2018-02-06 Michael Collison <michael.collison@arm.com>
21973 * config/arm/thumb2.md:
21974 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
21975 (*thumb_mov_notscc): Ditto.
21977 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
21980 * config/rs6000/rs6000.md (su code attribute): Use "u" for
21981 unsigned_fix, not "s".
21983 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
21985 * configure.ac (gcc_fn_eh_frame_ro): New function.
21986 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
21987 correct .eh_frame permissions.
21988 * configure: Regenerate.
21990 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
21992 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
21993 irrelevant options.
21995 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
21997 * config/rs6000/rs6000.c (rs6000_option_override_internal):
21998 Display warning message for -mno-speculate-indirect-jumps.
22000 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
22002 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
22004 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
22006 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
22008 PR tree-optimization/84225
22009 * tree-eh.c (find_trapping_overflow): Only call
22010 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
22012 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
22015 * config/i386/i386.c: Reimplement the check of possible options
22016 -mibt/-mshstk conbination. Change error messages.
22017 * doc/invoke.texi: Fix a typo: remove extra '='.
22019 2018-02-06 Marek Polacek <polacek@redhat.com>
22021 PR tree-optimization/84228
22022 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
22024 2018-02-06 Tamar Christina <tamar.christina@arm.com>
22027 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
22028 emitted arch directives.
22029 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
22030 __ARM_FEATURE_COPROC before changing architectures.
22032 2018-02-06 Richard Biener <rguenther@suse.de>
22034 * config/i386/i386.c (print_reg): Fix typo.
22035 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
22037 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
22039 * configure: Regenerate.
22041 2018-02-05 Martin Sebor <msebor@redhat.com>
22043 PR tree-optimization/83369
22044 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
22047 2018-02-05 Martin Liska <mliska@suse.cz>
22049 * doc/invoke.texi: Cherry-pick upstream r323995.
22051 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
22053 * ira.c (ira_init_register_move_cost): Adjust comment.
22055 2018-02-05 Martin Liska <mliska@suse.cz>
22057 PR gcov-profile/84137
22058 * doc/gcov.texi: Fix typo in documentation.
22060 2018-02-05 Martin Liska <mliska@suse.cz>
22062 PR gcov-profile/83879
22063 * doc/gcov.texi: Document necessity of --dynamic-list-data when
22064 using dlopen functionality.
22066 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
22068 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
22069 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
22070 _mm_maskz_range_ss, _mm_mask_range_round_ss,
22071 _mm_maskz_range_round_ss): New intrinsics.
22072 (__builtin_ia32_rangesd128_round)
22073 (__builtin_ia32_rangess128_round): Remove.
22074 (__builtin_ia32_rangesd128_mask_round,
22075 __builtin_ia32_rangess128_mask_round): New builtins.
22076 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
22077 __builtin_ia32_rangess128_round): Remove.
22078 (__builtin_ia32_rangesd128_mask_round,
22079 __builtin_ia32_rangess128_mask_round): New builtins.
22080 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
22081 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
22082 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
22083 "<round_saeonly_constraint>")): Changed to ...
22084 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
22085 "<round_saeonly_scalar_constraint>")): ... this.
22086 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
22087 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
22088 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
22089 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
22090 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
22092 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
22094 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
22096 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
22097 Remove all values except native, 8540 and 8548.
22099 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
22101 * config/i386/i386.c (ix86_output_function_return): Pass
22102 INVALID_REGNUM, instead of -1, as invalid register number to
22103 indirect_thunk_name and output_indirect_thunk.
22105 2018-02-02 Julia Koval <julia.koval@intel.com>
22107 * config.gcc: Add -march=icelake.
22108 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
22109 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
22110 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
22111 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
22112 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
22113 (processor_target_table): Add icelake.
22114 (ix86_option_override_internal): Handle new PTAs.
22115 (get_builtin_code_for_version): Handle icelake.
22116 (M_INTEL_COREI7_ICELAKE): New.
22117 (fold_builtin_cpu): Handle icelake.
22118 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
22119 * doc/invoke.texi: Add -march=icelake.
22121 2018-02-02 Julia Koval <julia.koval@intel.com>
22123 * config/i386/i386.c (ix86_option_override_internal): Change flags type
22124 to wide_int_bitmask.
22125 * wide-int-bitmask.h: New.
22127 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
22130 * config/i386/i386.md: Replace Pmode with word_mode in
22131 builtin_setjmp_setup and builtin_longjmp to support x32.
22133 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
22137 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
22139 (rs6000_supported_cpu_names): New static variable.
22140 (linux_cpu_translation_table): Likewise.
22141 (elf_platform) <cpu>: Define new static variable and use it.
22142 Translate kernel AT_PLATFORM name to canonical name if needed.
22143 Error if platform name is unknown.
22145 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
22148 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
22150 2018-02-01 Jeff Law <law@redhat.com>
22153 * config/i386/i386.c (release_scratch_register_on_entry): Add new
22154 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
22155 the scratch if RELEASE_VIA_POP is false.
22156 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
22157 If we have to save a temporary register, decrement SIZE appropriately.
22158 Pass new arguments to release_scratch_register_on_entry.
22159 (ix86_adjust_stack_and_probe): Likewise.
22160 (ix86_emit_probe_stack_range): Pass new arguments to
22161 release_scratch_register_on_entry.
22163 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
22165 PR rtl-optimization/84157
22166 * combine.c (change_zero_ext): Use REG_P predicate in
22167 front of HARD_REGISTER_P predicate.
22169 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
22171 * config/avr/avr.c (avr_option_override): Move disabling of
22172 -fdelete-null-pointer-checks to...
22173 * common/config/avr/avr-common.c (avr_option_optimization_table):
22176 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
22178 PR tree-optimization/81635
22179 * tree-data-ref.c (split_constant_offset_1): For types that
22180 wrap on overflow, try to use range info to prove that wrapping
22183 2018-02-01 Renlin Li <renlin.li@arm.com>
22186 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
22187 TAILCALL_ADDR_REGS.
22188 (aarch64_register_move_cost): Likewise.
22189 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
22190 TAILCALL_ADDR_REGS.
22191 (REG_CLASS_NAMES): Likewise.
22192 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
22193 TAILCALL_ADDR_REGS. Remove IP registers.
22194 * config/aarch64/aarch64.md (Ucs): Update register constraint.
22196 2018-02-01 Richard Biener <rguenther@suse.de>
22198 * domwalk.h (dom_walker::dom_walker): Add additional constructor
22199 for specifying RPO order and allow NULL for that.
22200 * domwalk.c (dom_walker::dom_walker): Likewise.
22201 (dom_walker::walk): Handle NULL RPO order.
22202 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
22204 (rewrite_update_dom_walker): Likewise.
22205 (mark_def_dom_walker): Likewise.
22207 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
22209 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
22210 (aarch64_maybe_expand_sve_subreg_move): Declare.
22211 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
22212 * config/aarch64/predicates.md (aarch64_any_register_operand): New
22214 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
22215 that are semantically a reverse operation.
22216 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
22217 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
22218 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
22220 (aarch64_can_change_mode_class): For big-endian, forbid changes
22221 between two SVE modes if they have different element sizes.
22223 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
22225 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
22226 the TImode handling for big-endian targets.
22228 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
22230 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
22231 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
22233 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
22234 Remove BSWAP handing for big-endian targets and use the form of
22235 LD1RQ appropariate for the mode.
22237 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
22239 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
22240 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
22241 duplicated element.
22243 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
22246 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
22247 check for operands that need to go through aarch64_sve_reload_be.
22249 2018-02-01 Jakub Jelinek <jakub@redhat.com>
22251 PR tree-optimization/81661
22252 PR tree-optimization/84117
22253 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
22254 * tree-eh.c: Include gimplify.h.
22255 (find_trapping_overflow, replace_trapping_overflow,
22256 rewrite_to_non_trapping_overflow): New functions.
22257 * tree-vect-loop.c: Include tree-eh.h.
22258 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
22259 * tree-data-ref.c: Include tree-eh.h.
22260 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
22262 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
22264 PR rtl-optimization/84123
22265 * combine.c (change_zero_ext): Check if hard register satisfies
22266 can_change_dest_mode before calling gen_lowpart_SUBREG.
22268 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
22271 * ira.c (ira_init_register_move_cost): Remove assert.
22273 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
22275 PR rtl-optimization/84071
22276 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
22277 * doc/tm.texi: Regenerate.
22279 2018-01-31 Richard Biener <rguenther@suse.de>
22281 PR tree-optimization/84132
22282 * tree-data-ref.c (analyze_miv_subscript): Properly
22283 check whether evolution_function_is_affine_multivariate_p
22284 before calling gcd_of_steps_may_divide_p.
22286 2018-01-31 Julia Koval <julia.koval@intel.com>
22289 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
22290 * config/i386/i386.md (rdpid_rex64) New.
22291 (rdpid): Make 32bit only.
22293 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
22296 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
22297 an IDENTIFIER_NODE for FUNCTION_TYPE's.
22299 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
22302 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
22304 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
22306 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
22308 PR rtl-optimization/84071
22309 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
22310 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
22312 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
22314 * config/arc/arc.c (arc_handle_aux_attribute): New function.
22315 (arc_attribute_table): Add 'aux' attribute.
22316 (arc_in_small_data_p): Consider aux like variables.
22317 (arc_is_aux_reg_p): New function.
22318 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
22319 (arc_get_aux_arg): New function.
22320 (prepare_move_operands): Handle aux-register access.
22321 (arc_handle_aux_attribute): New function.
22322 * doc/extend.texi (ARC Variable attributes): Add subsection.
22324 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
22326 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
22327 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
22328 (arc_attribute_table): Add 'uncached' attribute.
22329 (arc_print_operand): Print '.di' flag for uncached memory
22331 (arc_in_small_data_p): Do not consider for small data the uncached
22333 (arc_is_uncached_mem_p): New function.
22334 * config/arc/predicates.md (compact_store_memory_operand): Check
22335 for uncached memory accesses.
22336 (nonvol_nonimm_operand): Likewise.
22337 * doc/extend.texi (ARC Type Attribute): New subsection.
22339 2018-01-31 Jakub Jelinek <jakub@redhat.com>
22342 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
22343 falign-loops=): Add Optimization flag.
22345 2018-01-30 Jeff Law <law@redhat.com>
22348 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
22349 INT_REGISTERS_SAVED. Check it prior to calling
22350 get_scratch_register_on_entry.
22351 (ix86_adjust_stack_and_probe): Similarly.
22352 (ix86_emit_probe_stack_range): Similarly.
22353 (ix86_expand_prologue): Corresponding changes.
22355 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
22358 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
22359 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
22361 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
22364 * lra-constraints.c (curr_insn_transform): Process AND in the
22367 2018-01-30 Jakub Jelinek <jakub@redhat.com>
22369 PR rtl-optimization/83986
22370 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
22371 dependence against last_pending_memory_flush in addition to
22372 pending_jump_insns.
22374 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
22376 PR tree-optimization/81611
22377 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
22380 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
22383 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
22386 2018-01-30 Richard Biener <rguenther@suse.de>
22387 Jakub Jelinek <jakub@redhat.com>
22389 PR tree-optimization/84111
22390 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
22391 inner loops added during recursion, as they don't have up-to-date
22394 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
22397 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
22398 (can_inline_edge_by_limits_p): ... here.
22399 (can_early_inline_edge_p, check_callers,
22400 update_caller_keys, update_callee_keys, recursive_inlining,
22401 add_new_edges_to_heap, speculation_useful_p,
22402 inline_small_functions,
22403 inline_small_functions, flatten_function,
22404 inline_to_all_callers_1): Update.
22406 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
22408 * profile-count.c (profile_count::combine_with_ipa_count): Handle
22411 2018-01-30 Richard Biener <rguenther@suse.de>
22413 PR tree-optimization/83008
22414 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
22415 invariant and constant vector uses in stmts when they need
22416 more than one stmt.
22418 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
22421 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
22422 * configure: Regenerate.
22424 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
22426 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
22428 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
22429 Use gen_rtx_REG rather than gen_lowpart.
22431 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
22433 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
22434 rather than 0 when creating partial subregs.
22436 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
22438 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
22441 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
22444 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
22445 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
22446 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
22447 flags. This restores the settings used before the 2017-07-24.
22448 Turning off pre increment/decrement/modify allows IVOPTS to
22449 optimize DF/SF loops where the index is an int.
22451 2018-01-29 Richard Biener <rguenther@suse.de>
22452 Kelvin Nilsen <kelvin@gcc.gnu.org>
22455 * tree-vect-stmts.c (vectorizable_call): Don't call
22456 targetm.vectorize_builtin_md_vectorized_function if callee is
22459 2018-01-22 Carl Love <cel@us.ibm.com>
22461 * doc/extend.tex: Fix typo in second arg in
22462 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
22464 2018-01-29 Richard Biener <rguenther@suse.de>
22466 PR tree-optimization/84086
22467 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
22468 (flush_ssaname_freelist): When SSA names were released reset
22469 the SCEV hash table.
22471 2018-01-29 Richard Biener <rguenther@suse.de>
22473 PR tree-optimization/84057
22474 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
22475 removed paths when removing edges.
22477 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
22479 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
22480 -mfunction-return=@var{choice}.
22482 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
22484 PR diagnostic/84034
22485 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
22486 Handle CR like TAB.
22487 (layout::print_source_line): Likewise.
22488 (test_get_line_width_without_trailing_whitespace): Add test cases.
22490 2018-01-27 Jakub Jelinek <jakub@redhat.com>
22492 PR middle-end/84040
22493 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
22496 2018-01-26 Jim Wilson <jimw@sifive.com>
22498 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
22500 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
22503 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22505 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
22506 and CMP + SUB-immediate -> SUBS.
22508 2018-01-26 Martin Sebor <msebor@redhat.com>
22510 PR tree-optimization/83896
22511 * tree-ssa-strlen.c (get_string_len): Rename...
22512 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
22513 Avoid assuming length is constant.
22514 (handle_char_store): Use HOST_WIDE_INT for string length.
22516 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
22519 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
22520 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
22522 2018-01-26 Richard Biener <rguenther@suse.de>
22524 PR rtl-optimization/84003
22525 * dse.c (record_store): Only record redundant stores when
22526 the earlier store aliases at least all accesses the later one does.
22528 2018-01-26 Jakub Jelinek <jakub@redhat.com>
22530 PR rtl-optimization/83985
22531 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
22532 REG_CFA_RESTORE insns.
22533 (delete_unmarked_insns): Don't ignore separate shrink wrapping
22534 REG_CFA_RESTORE insns here.
22537 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
22538 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
22540 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
22542 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
22543 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
22544 (arc_init): Likewise.
22545 (arc_override_options): Likewise.
22546 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
22548 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
22550 * config/arc/arc.h (TARGET_DBNZ): Define.
22551 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
22552 properly set the tune attribute.
22553 (dbnz): Use TARGET_DBNZ guard.
22554 * config/arc/arc.opt (mtune): Add core3 option.
22556 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
22558 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
22559 recognize new pic like addresses.
22560 (arc_delegitimize_address): Clean up.
22562 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
22564 * config/arc/arc-arches.def: Option mrf16 valid for all
22566 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
22567 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
22568 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
22569 * config/arc/arc-tables.opt: Regenerate.
22570 * config/arc/arc.c (arc_conditional_register_usage): Handle
22571 reduced register file case.
22572 (arc_file_start): Set must have build attributes.
22573 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
22574 mrf16 option value.
22575 * config/arc/arc.opt (mrf16): Add new option.
22576 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
22577 * config/arc/genmultilib.awk: Handle new mrf16 option.
22578 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
22579 * config/arc/t-multilib: Regenerate.
22580 * doc/invoke.texi (ARC Options): Document mrf16 option.
22582 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
22584 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
22585 * config/arc/arc.c (arc_handle_secure_attribute): New function.
22586 (arc_attribute_table): Add 'secure_call' attribute.
22587 (arc_print_operand): Print secure call operand.
22588 (arc_function_ok_for_sibcall): Don't optimize tail calls when
22590 (arc_is_secure_call_p): New function. * config/arc/arc.md
22591 (call_i): Add support for sjli instruction.
22592 (call_value_i): Likewise.
22593 * config/arc/constraints.md (Csc): New constraint.
22595 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
22596 John Eric Martin <John.Martin@emmicro-us.com>
22598 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
22599 * config/arc/arc.c (_arc_jli_section): New struct.
22600 (arc_jli_section): New type.
22601 (rc_jli_sections): New static variable.
22602 (arc_handle_jli_attribute): New function.
22603 (arc_attribute_table): Add jli_always and jli_fixed attribute.
22604 (arc_file_end): New function.
22605 (TARGET_ASM_FILE_END): Define.
22606 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
22607 (arc_add_jli_section): New function.
22608 (jli_call_scan): Likewise.
22609 (arc_reorg): Call jli_call_scan.
22610 (arc_output_addsi): Remove 'S' from printing asm operand.
22611 (arc_is_jli_call_p): New function.
22612 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
22614 (movhi_insn): Likewise.
22615 (movsi_insn): Likewise.
22616 (movsi_set_cc_insn): Likewise.
22617 (loadqi_update): Likewise.
22618 (load_zeroextendqisi_update): Likewise.
22619 (load_signextendqisi_update): Likewise.
22620 (loadhi_update): Likewise.
22621 (load_zeroextendhisi_update): Likewise.
22622 (load_signextendhisi_update): Likewise.
22623 (loadsi_update): Likewise.
22624 (loadsf_update): Likewise.
22625 (movsicc_insn): Likewise.
22626 (bset_insn): Likewise.
22627 (bxor_insn): Likewise.
22628 (bclr_insn): Likewise.
22629 (bmsk_insn): Likewise.
22630 (bicsi3_insn): Likewise.
22631 (cmpsi_cc_c_insn): Likewise.
22632 (movsi_ne): Likewise.
22633 (movsi_cond_exec): Likewise.
22634 (clrsbsi2): Likewise.
22635 (norm_f): Likewise.
22645 (call_i): Remove 'S' asm letter, add jli instruction.
22646 (call_value_i): Likewise.
22647 * config/arc/arc.op (mjli-always): New option.
22648 * config/arc/constraints.md (Cji): New constraint.
22649 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
22651 (subsf3_fpx): Likewise.
22652 (mulsf3_fpx): Likewise.
22653 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
22655 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
22656 function attrbutes.
22657 * doc/invoke.texi (ARC): Document mjli-always option.
22659 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
22661 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
22662 avoid addition with 0 and use incw and decw where possible.
22664 2018-01-26 Richard Biener <rguenther@suse.de>
22666 PR tree-optimization/81082
22667 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
22668 association if it requires casting to unsigned.
22669 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
22670 from fold_plusminus_mult_expr to catch important cases late when
22671 range info is available.
22673 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
22675 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
22676 * configure.ac (hidden_linkonce): New test.
22677 * configure: Regenerate.
22678 * config.in: Regenerate.
22680 2018-01-26 Julia Koval <julia.koval@intel.com>
22682 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
22683 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
22684 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
22685 _mm_mask_bitshuffle_epi64_mask): Fix type.
22686 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
22687 USI_FTYPE_V4DI_V4DI_USI): Remove.
22688 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
22689 __builtin_ia32_vpshufbitqmb256_mask,
22690 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
22691 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
22692 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
22694 2018-01-26 Alan Modra <amodra@gmail.com>
22697 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
22698 UNSPEC_VBPERMQ. Sort other unspecs.
22700 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
22702 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
22704 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
22706 PR middle-end/83055
22707 * predict.c (drop_profile): Do not push/pop cfun; update also
22709 (handle_missing_profiles): Fix logic looking for zero profiles.
22711 2018-01-25 Jakub Jelinek <jakub@redhat.com>
22713 PR middle-end/83977
22714 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
22715 on functions with #pragma omp declare simd or functions with simd
22717 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
22718 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
22719 Remove trailing \n from warning_at calls.
22721 2018-01-25 Tom de Vries <tom@codesourcery.com>
22724 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
22725 for neutered workers.
22727 2018-01-24 Joseph Myers <joseph@codesourcery.com>
22730 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
22731 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
22733 2018-01-24 Jeff Law <law@redhat.com>
22736 * i386.c (get_probe_interval): Move to earlier point.
22737 (ix86_compute_frame_layout): If -fstack-clash-protection and
22738 the frame is larger than the probe interval, then use pushes
22739 to save registers rather than reg->mem moves.
22740 (ix86_expand_prologue): Remove conditional for int_registers_saved
22743 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
22746 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
22747 min/max for never referenced object.
22749 2018-01-24 Jakub Jelinek <jakub@redhat.com>
22751 PR middle-end/83977
22752 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
22754 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
22755 attributes from DECL_ATTRIBUTES (decl) without affecting
22756 DECL_ATTRIBUTES (current_function_decl).
22757 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
22758 functions with non-NULL DECL_ABSTRACT_ORIGIN.
22760 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
22762 PR tree-optimization/83979
22763 * fold-const.c (fold_comparison): Use constant_boolean_node
22764 instead of boolean_{true,false}_node.
22766 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
22768 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
22771 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
22773 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
22774 Simplify the clause that sets the length attribute.
22775 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
22776 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
22777 clause that sets the length attribute.
22778 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
22780 2018-01-24 Tom de Vries <tom@codesourcery.com>
22783 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
22784 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
22785 Add strict parameter.
22786 (prevent_branch_around_nothing): Insert dummy insn between branch to
22787 label and label with no ptx insn inbetween.
22788 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
22790 2018-01-24 Tom de Vries <tom@codesourcery.com>
22793 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
22794 for neutered threads in warp.
22795 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
22797 2018-01-24 Richard Biener <rguenther@suse.de>
22799 PR tree-optimization/83176
22800 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
22803 2018-01-24 Richard Biener <rguenther@suse.de>
22805 PR tree-optimization/82819
22806 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
22807 code generating pluses that are no-ops in the target precision.
22809 2018-01-24 Richard Biener <rguenther@suse.de>
22811 PR middle-end/84000
22812 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
22814 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
22816 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
22817 to merge probabilities.
22818 * predict.c (probably_never_executed): Also mark as cold functions
22819 with global 0 profile and guessed local profile.
22820 * profile-count.c (profile_probability::combine_with_count): New
22822 * profile-count.h (profile_probability::operator*,
22823 profile_probability::operator*=, profile_probability::operator/,
22824 profile_probability::operator/=): Reduce precision to adjusted
22825 and set value to guessed on contradictory divisions.
22826 (profile_probability::combine_with_freq): Remove.
22827 (profile_probability::combine_wiht_count): Declare.
22828 (profile_count::force_nonzero):: Set to adjusted.
22829 (profile_count::probability_in):: Set quality to adjusted.
22830 * tree-ssa-tail-merge.c (replace_block_by): Use
22831 combine_with_count.
22833 2018-01-23 Andrew Waterman <andrew@sifive.com>
22834 Jim Wilson <jimw@sifive.com>
22836 * config/riscv/riscv.c (riscv_stack_boundary): New.
22837 (riscv_option_override): Set riscv_stack_boundary. Handle
22838 riscv_preferred_stack_boundary_arg.
22839 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
22840 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
22841 (STACK_BOUNDARY): Set to riscv_stack_boundary.
22842 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
22843 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
22844 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
22846 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
22849 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
22850 of struct ix86_frame.
22851 (ix86_expand_epilogue): Likewise. Add a local variable for
22852 the reg_save_offset field in struct ix86_frame.
22854 2018-01-23 Bin Cheng <bin.cheng@arm.com>
22856 PR tree-optimization/82604
22857 * tree-loop-distribution.c (enum partition_kind): New enum item
22858 PKIND_PARTIAL_MEMSET.
22859 (partition_builtin_p): Support above new enum item.
22860 (generate_code_for_partition): Ditto.
22861 (compute_access_range): Differentiate cases that equality can be
22862 proven at all loops, the innermost loops or no loops.
22863 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
22864 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
22865 (finalize_partitions, distribute_loop): Don't fuse partition of
22866 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
22867 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
22868 parloop is enabled.
22870 2018-01-23 Martin Liska <mliska@suse.cz>
22872 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
22873 order to ignore the predictor.
22874 (PRED_POLYMORPHIC_CALL): Likewise.
22875 (PRED_RECURSIVE_CALL): Likewise.
22877 2018-01-23 Martin Liska <mliska@suse.cz>
22879 * tree-profile.c (tree_profiling): Print function header to
22880 aware reader which function we are working on.
22881 * value-prof.c (gimple_find_values_to_profile): Do not print
22882 not interesting value histograms.
22884 2018-01-23 Martin Liska <mliska@suse.cz>
22886 * profile-count.h (enum profile_quality): Add
22887 profile_uninitialized as the first value. Do not number values
22888 as they are zero based.
22889 (profile_count::verify): Update sanity check.
22890 (profile_probability::verify): Likewise.
22892 2018-01-23 Nathan Sidwell <nathan@acm.org>
22894 * doc/invoke.texi (ffor-scope): Deprecate.
22896 2018-01-23 David Malcolm <dmalcolm@redhat.com>
22898 PR tree-optimization/83510
22899 * domwalk.c (set_all_edges_as_executable): New function.
22900 (dom_walker::dom_walker): Convert bool param
22901 "skip_unreachable_blocks" to enum reachability. Move setup of
22902 edge flags to set_all_edges_as_executable and only do it when
22903 reachability is REACHABLE_BLOCKS.
22904 * domwalk.h (enum dom_walker::reachability): New enum.
22905 (dom_walker::dom_walker): Convert bool param
22906 "skip_unreachable_blocks" to enum reachability.
22907 (set_all_edges_as_executable): New decl.
22908 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
22909 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
22911 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
22912 but converting true to REACHABLE_BLOCKS.
22913 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
22915 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
22916 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
22917 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
22919 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
22920 if check_all_array_refs will be called.
22922 2018-01-23 David Malcolm <dmalcolm@redhat.com>
22924 * tree.c (selftest::test_location_wrappers): Add more test
22927 2018-01-23 David Malcolm <dmalcolm@redhat.com>
22929 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
22930 (selftest::test_bit_in_range): Likewise.
22932 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
22935 * doc/sourcebuild.texi (vect_float): Say that the selector
22936 only describes the situation when -funsafe-math-optimizations is on.
22937 (vect_float_strict): Document.
22939 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
22941 PR tree-optimization/83965
22942 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
22943 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
22944 instead of checking only for a reduction.
22945 (vect_recog_widen_sum_pattern): Likewise.
22947 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
22949 * predict.c (probably_never_executed): Only use precise profile info.
22950 (compute_function_frequency): Skip after inlining hack since we now
22951 have quality checking.
22953 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
22955 * profile-count.h (profile_probability::very_unlikely,
22956 profile_probability::unlikely, profile_probability::even): Set
22957 precision to guessed.
22959 2018-01-23 Richard Biener <rguenther@suse.de>
22961 PR tree-optimization/83963
22962 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
22963 Properly terminate dominator walk when crossing the exit edge not
22964 when visiting its source block.
22966 2018-01-23 Jakub Jelinek <jakub@redhat.com>
22969 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
22970 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
22972 2018-01-22 Jakub Jelinek <jakub@redhat.com>
22974 PR tree-optimization/83957
22975 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
22976 semicolon after for body surrounded by braces.
22978 PR tree-optimization/83081
22979 * profile-count.h (profile_probability::split): New method.
22980 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
22981 Use profile_probability::split.
22982 (do_compare_rtx_and_jump): Fix adjustment of probabilities
22983 when splitting a single conditional jump into 2.
22985 2018-01-22 David Malcolm <dmalcolm@redhat.com>
22987 PR tree-optimization/69452
22988 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
22991 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
22993 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
22994 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
22995 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
22997 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
22999 * config/rl78/rl78-protos.h (rl78_split_movdi): New function
23001 * config/rl78/rl78.md (movdi): New define_expand.
23002 * config/rl78/rl78.c (rl78_split_movdi): New function.
23004 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
23007 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
23009 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
23010 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
23011 128-bit to produce an UNSPEC move to get the double word with the
23012 signbit and then a shift directly to do signbit.
23013 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
23014 implementation with a new version that just does either a direct
23015 move or a regular move. Move memory interface to separate insns.
23016 Move insns so they are next to the expander.
23017 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
23018 with signbit move. Split big and little endian case.
23019 (signbit<mode>2_dm_mem_le): Likewise.
23020 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
23021 (signbit<mode>2_dm2): Likewise.
23023 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
23025 * config/rl78/rl78.md (anddi3): New define_expand.
23027 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
23029 * config/rl78/rl78.md (umindi3): New define_expand.
23031 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
23033 * config/rl78/rl78.md (smindi3): New define_expand.
23035 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
23037 * config/rl78/rl78.md (smaxdi3): New define_expand.
23039 2018-01-22 Carl Love <cel@us.ibm.com>
23041 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
23042 LVX_V1TI): Add macro expansion.
23043 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
23044 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
23045 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
23046 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
23047 Change check to determine if the instruction is a byte reversing
23048 entry. Fix typo in comment.
23049 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
23050 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
23051 Add def_builtin calls for new builtins.
23052 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
23053 Add define_insn expansion.
23055 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
23057 * config/rl78/rl78.md (umaxdi3): New define_expand.
23059 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
23061 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
23062 for non-QImode registers.
23064 2018-01-22 Richard Biener <rguenther@suse.de>
23066 PR tree-optimization/83963
23067 * graphite-scop-detection.c (scop_detection::get_sese): Delay
23068 including the loop exit block.
23069 (scop_detection::merge_sese): Likewise.
23070 (scop_detection::add_scop): Do it here instead.
23072 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23074 * doc/sourcebuild.texi (arm_softfloat): Document.
23076 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
23079 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
23080 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
23081 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
23083 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
23084 David Edelsohn <dje.gcc@gmail.com>
23087 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
23088 Change "crset eq" to "crset 2".
23089 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
23090 (*call_indirect_aix<mode>_nospec): Likewise.
23091 (*call_value_indirect_aix<mode>_nospec): Likewise.
23092 (*call_indirect_elfv2<mode>_nospec): Likewise.
23093 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
23094 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
23095 change assembly output from . to $.
23096 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
23097 (indirect_jump<mode>_nospec): Change assembly output from . to $.
23098 (*tablejump<mode>_internal1_nospec): Likewise.
23100 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
23103 * config/sh/sh_optimize_sett_clrt.cc:
23104 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
23106 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
23108 PR tree-optimization/83940
23109 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
23110 offset_dt to vect_constant_def rather than vect_unknown_def_type.
23111 (vect_check_load_store_mask): Add a mask_dt_out parameter and
23112 use it to pass back the definition type.
23113 (vect_check_store_rhs): Likewise rhs_dt_out.
23114 (vect_build_gather_load_calls): Add a mask_dt argument and use
23115 it instead of a call to vect_is_simple_use.
23116 (vectorizable_store): Update calls to vect_check_load_store_mask
23117 and vect_check_store_rhs. Use the dt returned by the latter instead
23118 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
23119 instead of calls to vect_is_simple_use. Pass the scalar rather
23120 than the vector operand to vect_is_simple_use when handling
23121 second and subsequent copies of an rhs value.
23122 (vectorizable_load): Update calls to vect_check_load_store_mask
23123 and vect_build_gather_load_calls. Use the cached mask_dt and
23124 gs_info.offset_dt instead of calls to vect_is_simple_use.
23126 2018-01-20 Jakub Jelinek <jakub@redhat.com>
23128 PR middle-end/83945
23129 * tree-emutls.c: Include gimplify.h.
23130 (lower_emutls_2): New function.
23131 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
23132 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
23133 it before further processing.
23136 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
23137 UINTVAL (trueop1) instead of INTVAL (op1).
23139 2018-01-19 Jakub Jelinek <jakub@redhat.com>
23143 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
23144 INCOMING_FRAME_SP_OFFSET if not defined.
23145 (scan_trace): Add ENTRY argument. If true and
23146 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
23147 emit a note to adjust the CFA offset.
23148 (create_cfi_notes): Adjust scan_trace callers.
23149 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
23150 INCOMING_FRAME_SP_OFFSET in the CIE.
23151 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
23152 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
23154 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
23155 * doc/tm.texi: Regenerated.
23157 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
23159 PR rtl-optimization/83147
23160 * lra-constraints.c (remove_inheritance_pseudos): Use
23161 lra_substitute_pseudo_within_insn.
23163 2018-01-19 Tom de Vries <tom@codesourcery.com>
23164 Cesar Philippidis <cesar@codesourcery.com>
23167 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
23169 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
23172 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
23173 spaces for function labels.
23175 2018-01-19 Martin Liska <mliska@suse.cz>
23177 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
23178 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
23179 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
23180 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
23181 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
23182 (PRED_CONST_RETURN): Change from 69 to 65.
23183 (PRED_NULL_RETURN): Change from 91 to 71.
23184 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
23185 (PRED_LOOP_GUARD): Change from 66 to 73.
23187 2018-01-19 Martin Liska <mliska@suse.cz>
23189 * predict.c (predict_insn_def): Add new assert.
23190 (struct branch_predictor): Change type to signed integer.
23191 (test_prediction_value_range): Amend test to cover
23192 PROB_UNINITIALIZED.
23193 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
23194 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
23195 (PRED_LOOP_ITERATIONS_MAX): Likewise.
23196 (PRED_LOOP_IV_COMPARE): Likewise.
23197 * predict.h (PROB_UNINITIALIZED): Define new constant.
23199 2018-01-19 Martin Liska <mliska@suse.cz>
23201 * predict.c (dump_prediction): Add new format for
23202 analyze_brprob.py script which is enabled with -details
23204 * profile-count.h (precise_p): New function.
23206 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
23208 PR tree-optimization/83922
23209 * tree-vect-loop.c (vect_verify_full_masking): Return false if
23210 there are no statements that need masking.
23211 (vect_active_double_reduction_p): New function.
23212 (vect_analyze_loop_operations): Use it when handling phis that
23213 are not in the loop header.
23215 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
23217 PR tree-optimization/83914
23218 * tree-vect-loop.c (vectorizable_induction): Don't convert
23219 init_expr or apply the peeling adjustment for inductions
23220 that are nested within the vectorized loop.
23222 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23224 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
23227 2018-01-18 Jakub Jelinek <jakub@redhat.com>
23231 * function.h (gimplify_parameters): Add gimple_seq * argument.
23232 * function.c: Include gimple.h and options.h.
23233 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
23234 for the added local temporaries if needed.
23235 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
23236 if there are any parameter cleanups, wrap whole body into a
23237 try/finally with the cleanups.
23239 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
23242 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
23243 Use GET_MODE_CLASS for scalar floating point.
23245 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
23249 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
23250 Fix call of call_cgraph_insertion_hooks.
23252 2018-01-18 Martin Sebor <msebor@redhat.com>
23254 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
23256 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
23259 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
23262 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
23265 * common.opt: (-ffile-prefix-map): New option.
23266 * opts.c (common_handle_option): Defer it.
23267 * opts-global.c (handle_common_deferred_options): Handle it.
23268 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
23269 * file-prefix-map.h: New file.
23270 (remap_debug_filename, add_debug_prefix_map): ...here.
23271 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
23272 * final.c (debug_prefix_map, add_debug_prefix_map
23273 remap_debug_filename): Move to...
23274 * file-prefix-map.c: New file.
23275 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
23276 generalize, get rid of alloca(), use strrchr() instead of strchr().
23277 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
23278 Implement in terms of add_prefix_map().
23279 (remap_macro_filename, remap_debug_filename): Implement in term of
23281 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
23282 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
23283 * dbxout.c: Include file-prefix-map.h.
23284 * varasm.c: Likewise.
23285 * vmsdbgout.c: Likewise.
23286 * xcoffout.c: Likewise.
23287 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
23288 * doc/cppopts.texi (-fmacro-prefix-map): Document.
23289 * doc/invoke.texi (-ffile-prefix-map): Document.
23290 (-fdebug-prefix-map): Update description.
23292 2018-01-18 Martin Liska <mliska@suse.cz>
23294 * config/i386/i386.c (indirect_thunk_name): Document that also
23296 (output_indirect_thunk): Document why both instructions
23297 (pause and lfence) are generated.
23299 2018-01-18 Richard Biener <rguenther@suse.de>
23301 PR tree-optimization/83887
23302 * graphite-scop-detection.c
23303 (scop_detection::get_nearest_dom_with_single_entry): Remove.
23304 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
23305 (scop_detection::merge_sese): Re-implement with a flood-fill
23306 algorithm that properly finds a SESE region if it exists.
23308 2018-01-18 Jakub Jelinek <jakub@redhat.com>
23311 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
23312 pointer_diff optimizations use view_convert instead of convert.
23314 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
23316 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
23317 Generate different code for -mno-speculate-indirect-jumps.
23318 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
23319 (*call_indirect_aix<mode>): Disable for
23320 -mno-speculate-indirect-jumps.
23321 (*call_indirect_aix<mode>_nospec): New define_insn.
23322 (*call_value_indirect_aix<mode>): Disable for
23323 -mno-speculate-indirect-jumps.
23324 (*call_value_indirect_aix<mode>_nospec): New define_insn.
23325 (*sibcall_nonlocal_sysv<mode>): Generate different code for
23326 -mno-speculate-indirect-jumps.
23327 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
23329 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
23331 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
23332 long double type, set the flags for noting the default long double
23333 type, even if we don't pass or return a long double type.
23335 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
23338 * ipa-inline.c (flatten_function): Do not overwrite final inlining
23341 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
23343 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
23344 support for merge[hl].
23345 (fold_mergehl_helper): New helper function.
23346 (tree-vector-builder.h): New #include for tree_vector_builder usage.
23347 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
23348 (altivec_vmrglw_direct): Add xxmrglw insn.
23350 2018-01-17 Andrew Waterman <andrew@sifive.com>
23352 * config/riscv/riscv.c (riscv_conditional_register_usage): If
23353 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
23355 2018-01-17 David Malcolm <dmalcolm@redhat.com>
23358 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
23359 call the lto_location_cache before reading the
23360 DECL_SOURCE_LOCATION of the types.
23362 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
23363 Richard Sandiford <richard.sandiford@linaro.org>
23365 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
23366 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
23367 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
23368 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
23369 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
23371 * config/aarch64/constraints.md (aarch64_movti_operand):
23373 * config/aarch64/predicates.md (Uti): Add new constraint.
23375 2018-01-17 Carl Love <cel@us.ibm.com>
23377 * config/rs6000/vsx.md (define_expand xl_len_r,
23378 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
23379 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
23381 (define_expand, define_insn): Move the shift left from the
23382 define_insn to the define_expand for lxvl and stxvl instructions.
23383 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
23384 and XL_LEN_R definitions to PURE.
23386 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
23388 * config/i386/i386.c (indirect_thunk_name): Declare regno
23389 as unsigned int. Compare regno with INVALID_REGNUM.
23390 (output_indirect_thunk): Ditto.
23391 (output_indirect_thunk_function): Ditto.
23392 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
23393 in the call to output_indirect_thunk_function.
23395 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
23397 PR middle-end/83884
23398 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
23399 rather than the size of inner_type to determine the stack slot size
23400 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
23402 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
23405 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
23408 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
23410 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
23411 endian Linux systems to optionally enable multilibs for selecting
23412 the long double type if the user configured an explicit type.
23413 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
23414 have no long double multilibs if not defined.
23415 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
23416 warn if the user used -mabi={ieee,ibm}longdouble and we built
23417 multilibs for long double.
23418 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
23419 appropriate multilib option.
23420 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
23422 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
23423 for building long double multilibs.
23424 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
23426 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
23428 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
23431 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
23433 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
23436 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
23439 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
23442 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
23444 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
23445 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
23447 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
23449 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
23450 different rtl trees depending on TARGET_64BIT.
23451 (rs6000_gen_lvx): Likewise.
23453 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
23455 * config/visium/visium.md (nop): Tweak comment.
23456 (hazard_nop): Likewise.
23458 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
23460 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
23461 -mspeculate-indirect-jumps.
23462 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
23463 for -mno-speculate-indirect-jumps.
23464 (*call_indirect_elfv2<mode>_nospec): New define_insn.
23465 (*call_value_indirect_elfv2<mode>): Disable for
23466 -mno-speculate-indirect-jumps.
23467 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
23468 (indirect_jump): Emit different RTL for
23469 -mno-speculate-indirect-jumps.
23470 (*indirect_jump<mode>): Disable for
23471 -mno-speculate-indirect-jumps.
23472 (*indirect_jump<mode>_nospec): New define_insn.
23473 (tablejump): Emit different RTL for
23474 -mno-speculate-indirect-jumps.
23475 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
23476 (tablejumpsi_nospec): New define_expand.
23477 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
23478 (tablejumpdi_nospec): New define_expand.
23479 (*tablejump<mode>_internal1): Disable for
23480 -mno-speculate-indirect-jumps.
23481 (*tablejump<mode>_internal1_nospec): New define_insn.
23482 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
23485 2018-01-16 Artyom Skrobov tyomitch@gmail.com
23487 * caller-save.c (insert_save): Drop unnecessary parameter. All
23490 2018-01-16 Jakub Jelinek <jakub@redhat.com>
23491 Richard Biener <rguenth@suse.de>
23494 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
23495 return early, inline manually is_gimple_sizepos. Make sure if we
23496 call gimplify_expr we don't end up with a gimple constant.
23497 * tree.c (variably_modified_type_p): Don't return true for
23498 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
23499 * gimplify.h (is_gimple_sizepos): Remove.
23501 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
23503 PR tree-optimization/83857
23504 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
23505 vectorizable_live_operation for pure SLP statements.
23506 (vectorizable_live_operation): Handle PHIs.
23508 2018-01-16 Richard Biener <rguenther@suse.de>
23510 PR tree-optimization/83867
23511 * tree-vect-stmts.c (vect_transform_stmt): Precompute
23512 nested_in_vect_loop_p since the scalar stmt may get invalidated.
23514 2018-01-16 Jakub Jelinek <jakub@redhat.com>
23517 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
23518 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
23519 If off is not INTEGER_CST, issue a may not be aligned warning
23520 rather than isn't aligned. Use isn%'t rather than isn't.
23521 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
23523 <case MULT_EXPR>: Improve the case when bottom and one of the
23524 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
23525 operand, in that case check if the other operand is multiple of
23526 bottom divided by the INTEGER_CST operand.
23528 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
23531 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
23532 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
23533 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
23534 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
23535 * config/pa/pa.c (pa_function_arg_advance): Likewise.
23536 (pa_function_arg, pa_arg_partial_bytes): Likewise.
23537 (pa_function_arg_size): New function.
23539 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
23541 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
23542 in a separate statement.
23544 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
23546 PR tree-optimization/83847
23547 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
23548 group gathers and scatters.
23550 2018-01-16 Jakub Jelinek <jakub@redhat.com>
23552 PR rtl-optimization/86620
23553 * params.def (max-sched-ready-insns): Bump minimum value to 1.
23555 PR rtl-optimization/83213
23556 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
23557 to last if both are JUMP_INSNs.
23559 PR tree-optimization/83843
23560 * gimple-ssa-store-merging.c
23561 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
23562 store_immediate_info for bswap/nop orig_stores.
23564 2018-01-15 Andrew Waterman <andrew@sifive.com>
23566 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
23568 <UDIV>: Increase cost if !TARGET_DIV.
23570 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
23572 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
23573 (define_attr "cr_logical_3op"): New.
23574 (cceq_ior_compare): Adjust.
23575 (cceq_ior_compare_complement): Adjust.
23576 (*cceq_rev_compare): Adjust.
23577 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
23578 (is_cracked_insn): Adjust.
23579 (insn_must_be_first_in_group): Adjust.
23580 * config/rs6000/40x.md: Adjust.
23581 * config/rs6000/440.md: Adjust.
23582 * config/rs6000/476.md: Adjust.
23583 * config/rs6000/601.md: Adjust.
23584 * config/rs6000/603.md: Adjust.
23585 * config/rs6000/6xx.md: Adjust.
23586 * config/rs6000/7450.md: Adjust.
23587 * config/rs6000/7xx.md: Adjust.
23588 * config/rs6000/8540.md: Adjust.
23589 * config/rs6000/cell.md: Adjust.
23590 * config/rs6000/e300c2c3.md: Adjust.
23591 * config/rs6000/e500mc.md: Adjust.
23592 * config/rs6000/e500mc64.md: Adjust.
23593 * config/rs6000/e5500.md: Adjust.
23594 * config/rs6000/e6500.md: Adjust.
23595 * config/rs6000/mpc.md: Adjust.
23596 * config/rs6000/power4.md: Adjust.
23597 * config/rs6000/power5.md: Adjust.
23598 * config/rs6000/power6.md: Adjust.
23599 * config/rs6000/power7.md: Adjust.
23600 * config/rs6000/power8.md: Adjust.
23601 * config/rs6000/power9.md: Adjust.
23602 * config/rs6000/rs64.md: Adjust.
23603 * config/rs6000/titan.md: Adjust.
23605 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
23607 * config/i386/predicates.md (indirect_branch_operand): Rewrite
23608 ix86_indirect_branch_register logic.
23610 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
23612 * config/i386/constraints.md (Bs): Update
23613 ix86_indirect_branch_register check. Don't check
23614 ix86_indirect_branch_register with GOT_memory_operand.
23616 * config/i386/predicates.md (GOT_memory_operand): Don't check
23617 ix86_indirect_branch_register here.
23618 (GOT32_symbol_operand): Likewise.
23620 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
23622 * config/i386/predicates.md (constant_call_address_operand):
23623 Rewrite ix86_indirect_branch_register logic.
23624 (sibcall_insn_operand): Likewise.
23626 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
23628 * config/i386/constraints.md (Bs): Replace
23629 ix86_indirect_branch_thunk_register with
23630 ix86_indirect_branch_register.
23632 * config/i386/i386.md (indirect_jump): Likewise.
23633 (tablejump): Likewise.
23634 (*sibcall_memory): Likewise.
23635 (*sibcall_value_memory): Likewise.
23636 Peepholes of indirect call and jump via memory: Likewise.
23637 * config/i386/i386.opt: Likewise.
23638 * config/i386/predicates.md (indirect_branch_operand): Likewise.
23639 (GOT_memory_operand): Likewise.
23640 (call_insn_operand): Likewise.
23641 (sibcall_insn_operand): Likewise.
23642 (GOT32_symbol_operand): Likewise.
23644 2018-01-15 Jakub Jelinek <jakub@redhat.com>
23646 PR middle-end/83837
23647 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
23648 type rather than type addr's type points to.
23649 (expand_omp_atomic_mutex): Likewise.
23650 (expand_omp_atomic): Likewise.
23652 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
23655 * config/i386/i386.c (output_indirect_thunk_function): Use
23656 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
23657 for __x86_return_thunk.
23659 2018-01-15 Richard Biener <rguenther@suse.de>
23661 PR middle-end/83850
23662 * expmed.c (extract_bit_field_1): Fix typo.
23664 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23667 * config/arm/iterators.md (VF): New mode iterator.
23668 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
23669 Remove integer-related logic from pattern.
23670 (neon_vabd<mode>_3): Likewise.
23672 2018-01-15 Jakub Jelinek <jakub@redhat.com>
23674 PR middle-end/82694
23675 * common.opt (fstrict-overflow): No longer an alias.
23676 (fwrapv-pointer): New option.
23677 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
23678 also for pointer types based on flag_wrapv_pointer.
23679 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
23680 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
23681 opts->x_flag_wrapv got set.
23682 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
23683 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
23684 POINTER_TYPE_OVERFLOW_UNDEFINED.
23685 * match.pd: Likewise in address comparison pattern.
23686 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
23688 2018-01-15 Richard Biener <rguenther@suse.de>
23691 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
23692 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
23693 Reset type names to their identifier if their TYPE_DECL doesn't
23694 have linkage (and thus is used for ODR and devirt).
23695 (save_debug_info_for_decl): Remove.
23696 (save_debug_info_for_type): Likewise.
23697 (add_tree_to_fld_list): Adjust.
23698 * tree-pretty-print.c (dump_generic_node): Make dumping of
23699 type names more robust.
23701 2018-01-15 Richard Biener <rguenther@suse.de>
23703 * BASE-VER: Bump to 8.0.1.
23705 2018-01-14 Martin Sebor <msebor@redhat.com>
23708 * builtins.c (check_access): Avoid warning when the no-warning bit
23711 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
23713 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
23714 * ira-color (allocno_hard_regs_compare): Likewise.
23716 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
23719 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
23720 Use .pushsection/.popsection.
23722 2018-01-14 Martin Sebor <msebor@redhat.com>
23725 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
23727 2018-01-14 Jakub Jelinek <jakub@redhat.com>
23729 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
23730 entry from extra_headers.
23731 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
23732 extra_headers, make the list bitwise identical to the i?86-*-* one.
23734 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
23736 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
23737 -mcmodel=large with -mindirect-branch=thunk,
23738 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
23739 -mfunction-return=thunk-extern.
23740 * doc/invoke.texi: Document -mcmodel=large is incompatible with
23741 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
23742 -mfunction-return=thunk and -mfunction-return=thunk-extern.
23744 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
23746 * config/i386/i386.c (print_reg): Print the name of the full
23747 integer register without '%'.
23748 (ix86_print_operand): Handle 'V'.
23749 * doc/extend.texi: Document 'V' modifier.
23751 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
23753 * config/i386/constraints.md (Bs): Disallow memory operand for
23754 -mindirect-branch-register.
23756 * config/i386/predicates.md (indirect_branch_operand): Likewise.
23757 (GOT_memory_operand): Likewise.
23758 (call_insn_operand): Likewise.
23759 (sibcall_insn_operand): Likewise.
23760 (GOT32_symbol_operand): Likewise.
23761 * config/i386/i386.md (indirect_jump): Call convert_memory_address
23762 for -mindirect-branch-register.
23763 (tablejump): Likewise.
23764 (*sibcall_memory): Likewise.
23765 (*sibcall_value_memory): Likewise.
23766 Disallow peepholes of indirect call and jump via memory for
23767 -mindirect-branch-register.
23768 (*call_pop): Replace m with Bw.
23769 (*call_value_pop): Likewise.
23770 (*sibcall_pop_memory): Replace m with Bs.
23771 * config/i386/i386.opt (mindirect-branch-register): New option.
23772 * doc/invoke.texi: Document -mindirect-branch-register option.
23774 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
23776 * config/i386/i386-protos.h (ix86_output_function_return): New.
23777 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
23778 set function_return_type.
23779 (indirect_thunk_name): Add ret_p to indicate thunk for function
23781 (output_indirect_thunk_function): Pass false to
23782 indirect_thunk_name.
23783 (ix86_output_indirect_branch_via_reg): Likewise.
23784 (ix86_output_indirect_branch_via_push): Likewise.
23785 (output_indirect_thunk_function): Create alias for function
23786 return thunk if regno < 0.
23787 (ix86_output_function_return): New function.
23788 (ix86_handle_fndecl_attribute): Handle function_return.
23789 (ix86_attribute_table): Add function_return.
23790 * config/i386/i386.h (machine_function): Add
23791 function_return_type.
23792 * config/i386/i386.md (simple_return_internal): Use
23793 ix86_output_function_return.
23794 (simple_return_internal_long): Likewise.
23795 * config/i386/i386.opt (mfunction-return=): New option.
23796 (indirect_branch): Mention -mfunction-return=.
23797 * doc/extend.texi: Document function_return function attribute.
23798 * doc/invoke.texi: Document -mfunction-return= option.
23800 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
23802 * config/i386/i386-opts.h (indirect_branch): New.
23803 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
23804 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
23805 with local indirect jump when converting indirect call and jump.
23806 (ix86_set_indirect_branch_type): New.
23807 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
23808 (indirectlabelno): New.
23809 (indirect_thunk_needed): Likewise.
23810 (indirect_thunk_bnd_needed): Likewise.
23811 (indirect_thunks_used): Likewise.
23812 (indirect_thunks_bnd_used): Likewise.
23813 (INDIRECT_LABEL): Likewise.
23814 (indirect_thunk_name): Likewise.
23815 (output_indirect_thunk): Likewise.
23816 (output_indirect_thunk_function): Likewise.
23817 (ix86_output_indirect_branch_via_reg): Likewise.
23818 (ix86_output_indirect_branch_via_push): Likewise.
23819 (ix86_output_indirect_branch): Likewise.
23820 (ix86_output_indirect_jmp): Likewise.
23821 (ix86_code_end): Call output_indirect_thunk_function if needed.
23822 (ix86_output_call_insn): Call ix86_output_indirect_branch if
23824 (ix86_handle_fndecl_attribute): Handle indirect_branch.
23825 (ix86_attribute_table): Add indirect_branch.
23826 * config/i386/i386.h (machine_function): Add indirect_branch_type
23827 and has_local_indirect_jump.
23828 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
23830 (tablejump): Likewise.
23831 (*indirect_jump): Use ix86_output_indirect_jmp.
23832 (*tablejump_1): Likewise.
23833 (simple_return_indirect_internal): Likewise.
23834 * config/i386/i386.opt (mindirect-branch=): New option.
23835 (indirect_branch): New.
23838 (thunk-inline): Likewise.
23839 (thunk-extern): Likewise.
23840 * doc/extend.texi: Document indirect_branch function attribute.
23841 * doc/invoke.texi: Document -mindirect-branch= option.
23843 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
23846 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
23848 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
23850 * ipa-inline.c (want_inline_small_function_p): Return false if
23851 inlining has already failed with CIF_FINAL_ERROR.
23852 (update_caller_keys): Call want_inline_small_function_p before
23854 (update_callee_keys): Likewise.
23856 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
23858 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
23860 (rs6000_quadword_masked_address_p): Likewise.
23861 (quad_aligned_load_p): Likewise.
23862 (quad_aligned_store_p): Likewise.
23863 (const_load_sequence_p): Add comment to describe the outer-most loop.
23864 (mimic_memory_attributes_and_flags): New function.
23865 (rs6000_gen_stvx): Likewise.
23866 (replace_swapped_aligned_store): Likewise.
23867 (rs6000_gen_lvx): Likewise.
23868 (replace_swapped_aligned_load): Likewise.
23869 (replace_swapped_load_constant): Capitalize argument name in
23870 comment describing this function.
23871 (rs6000_analyze_swaps): Add a third pass to search for vector loads
23872 and stores that access quad-word aligned addresses and replace
23873 with stvx or lvx instructions when appropriate.
23874 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
23875 New function prototype.
23876 (rs6000_quadword_masked_address_p): Likewise.
23877 (rs6000_gen_lvx): Likewise.
23878 (rs6000_gen_stvx): Likewise.
23879 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
23880 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
23881 when memory address is aligned.
23882 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
23883 this split to select lvx instruction when memory address is aligned.
23884 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
23885 instruction when memory address is aligned.
23886 (*vsx_le_perm_load_v16qi): Likewise.
23887 (four unnamed splitters): Modify to select the stvx instruction
23888 when memory is aligned.
23890 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
23892 * predict.c (determine_unlikely_bbs): Handle correctly BBs
23893 which appears in the queue multiple times.
23895 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
23896 Alan Hayward <alan.hayward@arm.com>
23897 David Sherwood <david.sherwood@arm.com>
23899 * tree-vectorizer.h (vec_lower_bound): New structure.
23900 (_loop_vec_info): Add check_nonzero and lower_bounds.
23901 (LOOP_VINFO_CHECK_NONZERO): New macro.
23902 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
23903 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
23904 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
23905 fields. Make seg_len the distance travelled, not including the
23907 (dr_direction_indicator): Declare.
23908 (dr_zero_step_indicator): Likewise.
23909 (dr_known_forward_stride_p): Likewise.
23910 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
23912 (runtime_alias_check_p): Allow runtime alias checks with
23914 (operator ==): Compare access_size and align.
23915 (prune_runtime_alias_test_list): Rework for new distinction between
23916 the access_size and seg_len.
23917 (create_intersect_range_checks_index): Likewise. Cope with polynomial
23919 (get_segment_min_max): New function.
23920 (create_intersect_range_checks): Use it.
23921 (dr_step_indicator): New function.
23922 (dr_direction_indicator): Likewise.
23923 (dr_zero_step_indicator): Likewise.
23924 (dr_known_forward_stride_p): Likewise.
23925 * tree-loop-distribution.c (data_ref_segment_size): Return
23926 DR_STEP * (niters - 1).
23927 (compute_alias_check_pairs): Update call to the dr_with_seg_len
23929 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
23930 (vect_preserves_scalar_order_p): New function, split out from...
23931 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
23932 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
23933 (vect_vfa_access_size): New function.
23934 (vect_vfa_align): Likewise.
23935 (vect_compile_time_alias): Take access_size_a and access_b arguments.
23936 (dump_lower_bound): New function.
23937 (vect_check_lower_bound): Likewise.
23938 (vect_small_gap_p): Likewise.
23939 (vectorizable_with_step_bound_p): Likewise.
23940 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
23941 depencies if the vectorization factor is 1. Convert the checks
23942 for nonzero steps into checks on the bounds of DR_STEP. Try using
23943 a bunds check for variable steps if the minimum required step is
23944 relatively small. Update calls to the dr_with_seg_len
23945 constructor and to vect_compile_time_alias.
23946 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
23948 (vect_loop_versioning): Call it.
23949 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
23951 (vect_estimate_min_profitable_iters): Account for any bounds checks.
23953 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
23954 Alan Hayward <alan.hayward@arm.com>
23955 David Sherwood <david.sherwood@arm.com>
23957 * doc/sourcebuild.texi (vect_scatter_store): Document.
23958 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
23960 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
23962 * genopinit.c (main): Add supports_vec_scatter_store and
23963 supports_vec_scatter_store_cached to target_optabs.
23964 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
23965 IFN_MASK_SCATTER_STORE.
23966 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
23968 * internal-fn.h (internal_store_fn_p): Declare.
23969 (internal_fn_stored_value_index): Likewise.
23970 * internal-fn.c (scatter_store_direct): New macro.
23971 (expand_scatter_store_optab_fn): New function.
23972 (direct_scatter_store_optab_supported_p): New macro.
23973 (internal_store_fn_p): New function.
23974 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
23975 IFN_MASK_SCATTER_STORE.
23976 (internal_fn_mask_index): Likewise.
23977 (internal_fn_stored_value_index): New function.
23978 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
23979 for scatter stores.
23980 * optabs-query.h (supports_vec_scatter_store_p): Declare.
23981 * optabs-query.c (supports_vec_scatter_store_p): New function.
23982 * tree-vectorizer.h (vect_get_store_rhs): Declare.
23983 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
23984 true for scatter stores.
23985 (vect_gather_scatter_fn_p): Handle scatter stores too.
23986 (vect_check_gather_scatter): Consider using scatter stores if
23987 supports_vec_scatter_store_p.
23988 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
23989 scatter stores too.
23990 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
23991 internal_fn_stored_value_index.
23992 (check_load_store_masking): Handle scatter stores too.
23993 (vect_get_store_rhs): Make public.
23994 (vectorizable_call): Use internal_store_fn_p.
23995 (vectorizable_store): Handle scatter store internal functions.
23996 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
23997 when deciding whether the end of the group has been reached.
23998 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
23999 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
24000 (mask_scatter_store<mode>): New insns.
24002 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
24003 Alan Hayward <alan.hayward@arm.com>
24004 David Sherwood <david.sherwood@arm.com>
24006 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
24007 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
24008 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
24010 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
24011 Use vect_truncate_gather_scatter_offset if we can't treat the
24012 operation as a normal gather load or scatter store.
24013 (get_group_load_store_type): Take the gather_scatter_info
24014 as argument. Try using a gather load or scatter store for
24015 single-element groups.
24016 (get_load_store_type): Update calls to get_group_load_store_type
24017 and vect_use_strided_gather_scatters_p.
24019 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
24020 Alan Hayward <alan.hayward@arm.com>
24021 David Sherwood <david.sherwood@arm.com>
24023 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
24024 optional tree argument.
24025 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
24027 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
24028 but continue to use the current value as a fallback.
24029 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
24030 to compare the updates.
24031 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
24032 (get_load_store_type): Use it when handling a strided access.
24033 (vect_get_strided_load_store_ops): New function.
24034 (vect_get_data_ptr_increment): Likewise.
24035 (vectorizable_load): Handle strided gather loads. Always pass
24036 a step to vect_create_data_ref_ptr and bump_vector_ptr.
24038 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
24039 Alan Hayward <alan.hayward@arm.com>
24040 David Sherwood <david.sherwood@arm.com>
24042 * doc/md.texi (gather_load@var{m}): Document.
24043 (mask_gather_load@var{m}): Likewise.
24044 * genopinit.c (main): Add supports_vec_gather_load and
24045 supports_vec_gather_load_cached to target_optabs.
24046 * optabs-tree.c (init_tree_optimization_optabs): Use
24047 ggc_cleared_alloc to allocate target_optabs.
24048 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
24049 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
24051 * internal-fn.h (internal_load_fn_p): Declare.
24052 (internal_gather_scatter_fn_p): Likewise.
24053 (internal_fn_mask_index): Likewise.
24054 (internal_gather_scatter_fn_supported_p): Likewise.
24055 * internal-fn.c (gather_load_direct): New macro.
24056 (expand_gather_load_optab_fn): New function.
24057 (direct_gather_load_optab_supported_p): New macro.
24058 (direct_internal_fn_optab): New function.
24059 (internal_load_fn_p): Likewise.
24060 (internal_gather_scatter_fn_p): Likewise.
24061 (internal_fn_mask_index): Likewise.
24062 (internal_gather_scatter_fn_supported_p): Likewise.
24063 * optabs-query.c (supports_at_least_one_mode_p): New function.
24064 (supports_vec_gather_load_p): Likewise.
24065 * optabs-query.h (supports_vec_gather_load_p): Declare.
24066 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
24067 and memory_type field.
24068 (NUM_PATTERNS): Bump to 15.
24069 * tree-vect-data-refs.c: Include internal-fn.h.
24070 (vect_gather_scatter_fn_p): New function.
24071 (vect_describe_gather_scatter_call): Likewise.
24072 (vect_check_gather_scatter): Try using internal functions for
24073 gather loads. Recognize existing calls to a gather load function.
24074 (vect_analyze_data_refs): Consider using gather loads if
24075 supports_vec_gather_load_p.
24076 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
24077 (vect_get_gather_scatter_offset_type): Likewise.
24078 (vect_convert_mask_for_vectype): Likewise.
24079 (vect_add_conversion_to_patterm): Likewise.
24080 (vect_try_gather_scatter_pattern): Likewise.
24081 (vect_recog_gather_scatter_pattern): New pattern recognizer.
24082 (vect_vect_recog_func_ptrs): Add it.
24083 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
24084 internal_fn_mask_index and internal_gather_scatter_fn_p.
24085 (check_load_store_masking): Take the gather_scatter_info as an
24086 argument and handle gather loads.
24087 (vect_get_gather_scatter_ops): New function.
24088 (vectorizable_call): Check internal_load_fn_p.
24089 (vectorizable_load): Likewise. Handle gather load internal
24091 (vectorizable_store): Update call to check_load_store_masking.
24092 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
24093 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
24094 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
24095 (aarch64_gather_scale_operand_d): New predicates.
24096 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
24097 (mask_gather_load<mode>): New insns.
24099 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
24100 Alan Hayward <alan.hayward@arm.com>
24101 David Sherwood <david.sherwood@arm.com>
24103 * optabs.def (fold_left_plus_optab): New optab.
24104 * doc/md.texi (fold_left_plus_@var{m}): Document.
24105 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
24106 * internal-fn.c (fold_left_direct): Define.
24107 (expand_fold_left_optab_fn): Likewise.
24108 (direct_fold_left_optab_supported_p): Likewise.
24109 * fold-const-call.c (fold_const_fold_left): New function.
24110 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
24111 * tree-parloops.c (valid_reduction_p): New function.
24112 (gather_scalar_reductions): Use it.
24113 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
24114 (vect_finish_replace_stmt): Declare.
24115 * tree-vect-loop.c (fold_left_reduction_fn): New function.
24116 (needs_fold_left_reduction_p): New function, split out from...
24117 (vect_is_simple_reduction): ...here. Accept reductions that
24118 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
24119 (vect_force_simple_reduction): Also store the reduction type in
24120 the assignment's STMT_VINFO_REDUC_TYPE.
24121 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
24122 (merge_with_identity): New function.
24123 (vect_expand_fold_left): Likewise.
24124 (vectorize_fold_left_reduction): Likewise.
24125 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
24126 scalar phi in place for it. Check for target support and reject
24127 cases that would reassociate the operation. Defer the transform
24128 phase to vectorize_fold_left_reduction.
24129 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
24130 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
24131 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
24133 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
24135 * tree-if-conv.c (predicate_mem_writes): Remove redundant
24136 call to ifc_temp_var.
24138 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
24139 Alan Hayward <alan.hayward@arm.com>
24140 David Sherwood <david.sherwood@arm.com>
24142 * target.def (legitimize_address_displacement): Take the original
24143 offset as a poly_int.
24144 * targhooks.h (default_legitimize_address_displacement): Update
24146 * targhooks.c (default_legitimize_address_displacement): Likewise.
24147 * doc/tm.texi: Regenerate.
24148 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
24149 as an argument, moving assert of ad->disp == ad->disp_term to...
24150 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
24151 Try calling targetm.legitimize_address_displacement before expanding
24152 the address rather than afterwards, and adjust for the new interface.
24153 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
24154 Match the new hook interface. Handle SVE addresses.
24155 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
24156 new hook interface.
24158 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
24160 * Makefile.in (OBJS): Add early-remat.o.
24161 * target.def (select_early_remat_modes): New hook.
24162 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
24163 * doc/tm.texi: Regenerate.
24164 * targhooks.h (default_select_early_remat_modes): Declare.
24165 * targhooks.c (default_select_early_remat_modes): New function.
24166 * timevar.def (TV_EARLY_REMAT): New timevar.
24167 * passes.def (pass_early_remat): New pass.
24168 * tree-pass.h (make_pass_early_remat): Declare.
24169 * early-remat.c: New file.
24170 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
24172 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
24174 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
24175 Alan Hayward <alan.hayward@arm.com>
24176 David Sherwood <david.sherwood@arm.com>
24178 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
24179 vfm1 with a bound_epilog parameter.
24180 (vect_do_peeling): Update calls accordingly, and move the prologue
24181 call earlier in the function. Treat the base bound_epilog as 0 for
24182 fully-masked loops and retain vf - 1 for other loops. Add 1 to
24183 this base when peeling for gaps.
24184 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
24185 with fully-masked loops.
24186 (vect_estimate_min_profitable_iters): Handle the single peeled
24187 iteration in that case.
24189 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
24190 Alan Hayward <alan.hayward@arm.com>
24191 David Sherwood <david.sherwood@arm.com>
24193 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
24194 single-element interleaving even if the size is not a power of 2.
24195 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
24196 accesses for single-element interleaving if the group size is
24199 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
24200 Alan Hayward <alan.hayward@arm.com>
24201 David Sherwood <david.sherwood@arm.com>
24203 * doc/md.texi (fold_extract_last_@var{m}): Document.
24204 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
24205 * optabs.def (fold_extract_last_optab): New optab.
24206 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
24207 * internal-fn.c (fold_extract_direct): New macro.
24208 (expand_fold_extract_optab_fn): Likewise.
24209 (direct_fold_extract_optab_supported_p): Likewise.
24210 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
24211 * tree-vect-loop.c (vect_model_reduction_cost): Handle
24212 EXTRACT_LAST_REDUCTION.
24213 (get_initial_def_for_reduction): Do not create an initial vector
24214 for EXTRACT_LAST_REDUCTION reductions.
24215 (vectorizable_reduction): Leave the scalar phi in place for
24216 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
24217 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
24218 epilogue code for EXTRACT_LAST_REDUCTION and defer the
24219 transform phase to vectorizable_condition.
24220 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
24222 (vect_finish_stmt_generation): ...here.
24223 (vect_finish_replace_stmt): New function.
24224 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
24225 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
24227 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
24229 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
24230 Alan Hayward <alan.hayward@arm.com>
24231 David Sherwood <david.sherwood@arm.com>
24233 * doc/md.texi (extract_last_@var{m}): Document.
24234 * optabs.def (extract_last_optab): New optab.
24235 * internal-fn.def (EXTRACT_LAST): New internal function.
24236 * internal-fn.c (cond_unary_direct): New macro.
24237 (expand_cond_unary_optab_fn): Likewise.
24238 (direct_cond_unary_optab_supported_p): Likewise.
24239 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
24240 loops using EXTRACT_LAST.
24241 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
24242 (extract_last_<mode>): ...this optab.
24243 (vec_extract<mode><Vel>): Update accordingly.
24245 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
24246 Alan Hayward <alan.hayward@arm.com>
24247 David Sherwood <david.sherwood@arm.com>
24249 * target.def (empty_mask_is_expensive): New hook.
24250 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
24251 * doc/tm.texi: Regenerate.
24252 * targhooks.h (default_empty_mask_is_expensive): Declare.
24253 * targhooks.c (default_empty_mask_is_expensive): New function.
24254 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
24255 if the target says that empty masks are expensive.
24256 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
24258 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
24260 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
24261 Alan Hayward <alan.hayward@arm.com>
24262 David Sherwood <david.sherwood@arm.com>
24264 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
24265 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
24266 (vect_use_loop_mask_for_alignment_p): New function.
24267 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
24268 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
24269 niters_skip argument. Make sure that the first niters_skip elements
24270 of the first iteration are inactive.
24271 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
24272 Update call to vect_set_loop_masks_directly.
24273 (get_misalign_in_elems): New function, split out from...
24274 (vect_gen_prolog_loop_niters): ...here.
24275 (vect_update_init_of_dr): Take a code argument that specifies whether
24276 the adjustment should be added or subtracted.
24277 (vect_update_init_of_drs): Likewise.
24278 (vect_prepare_for_masked_peels): New function.
24279 (vect_do_peeling): Skip prologue peeling if we're using a mask
24280 instead. Update call to vect_update_inits_of_drs.
24281 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
24283 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
24284 alignment. Do not include the number of peeled iterations in
24285 the minimum threshold in that case.
24286 (vectorizable_induction): Adjust the start value down by
24287 LOOP_VINFO_MASK_SKIP_NITERS iterations.
24288 (vect_transform_loop): Call vect_prepare_for_masked_peels.
24289 Take the number of skipped iterations into account when calculating
24291 * tree-vect-stmts.c (vect_gen_while_not): New function.
24293 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
24294 Alan Hayward <alan.hayward@arm.com>
24295 David Sherwood <david.sherwood@arm.com>
24297 * doc/sourcebuild.texi (vect_fully_masked): Document.
24298 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
24299 default value to 0.
24300 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
24302 (vect_analyze_loop_2): ...here. Don't check the vectorization
24303 factor against the number of loop iterations if the loop is
24306 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
24307 Alan Hayward <alan.hayward@arm.com>
24308 David Sherwood <david.sherwood@arm.com>
24310 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
24311 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
24312 (dump_groups): Update accordingly.
24313 (iv_use::mem_type): New member variable.
24314 (address_p): New function.
24315 (record_use): Add a mem_type argument and initialize the new
24317 (record_group_use): Add a mem_type argument. Use address_p.
24318 Remove obsolete null checks of base_object. Update call to record_use.
24319 (find_interesting_uses_op): Update call to record_group_use.
24320 (find_interesting_uses_cond): Likewise.
24321 (find_interesting_uses_address): Likewise.
24322 (get_mem_type_for_internal_fn): New function.
24323 (find_address_like_use): Likewise.
24324 (find_interesting_uses_stmt): Try find_address_like_use before
24325 calling find_interesting_uses_op.
24326 (addr_offset_valid_p): Use the iv mem_type field as the type
24327 of the addressed memory.
24328 (add_autoinc_candidates): Likewise.
24329 (get_address_cost): Likewise.
24330 (split_small_address_groups_p): Use address_p.
24331 (split_address_groups): Likewise.
24332 (add_iv_candidate_for_use): Likewise.
24333 (autoinc_possible_for_pair): Likewise.
24334 (rewrite_groups): Likewise.
24335 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
24336 (determine_group_iv_cost): Update after split of USE_ADDRESS.
24337 (get_alias_ptr_type_for_ptr_address): New function.
24338 (rewrite_use_address): Rewrite address uses in calls that were
24339 identified by find_address_like_use.
24341 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
24342 Alan Hayward <alan.hayward@arm.com>
24343 David Sherwood <david.sherwood@arm.com>
24345 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
24347 * gimple-expr.h (is_gimple_addressable: Likewise.
24348 * gimple-expr.c (is_gimple_address): Likewise.
24349 * internal-fn.c (expand_call_mem_ref): New function.
24350 (expand_mask_load_optab_fn): Use it.
24351 (expand_mask_store_optab_fn): Likewise.
24353 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
24354 Alan Hayward <alan.hayward@arm.com>
24355 David Sherwood <david.sherwood@arm.com>
24357 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
24358 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
24359 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
24360 (cond_umax@var{mode}): Document.
24361 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
24362 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
24363 (cond_umin_optab, cond_umax_optab): New optabs.
24364 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
24365 (COND_IOR, COND_XOR): New internal functions.
24366 * internal-fn.h (get_conditional_internal_fn): Declare.
24367 * internal-fn.c (cond_binary_direct): New macro.
24368 (expand_cond_binary_optab_fn): Likewise.
24369 (direct_cond_binary_optab_supported_p): Likewise.
24370 (get_conditional_internal_fn): New function.
24371 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
24372 Cope with reduction statements that are vectorized as calls rather
24374 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
24375 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
24376 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
24377 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
24378 (UNSPEC_COND_EOR): New unspecs.
24379 (optab): Add mappings for them.
24380 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
24381 (sve_int_op, sve_fp_op): New int attributes.
24383 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
24384 Alan Hayward <alan.hayward@arm.com>
24385 David Sherwood <david.sherwood@arm.com>
24387 * optabs.def (while_ult_optab): New optab.
24388 * doc/md.texi (while_ult@var{m}@var{n}): Document.
24389 * internal-fn.def (WHILE_ULT): New internal function.
24390 * internal-fn.h (direct_internal_fn_supported_p): New override
24391 that takes two types as argument.
24392 * internal-fn.c (while_direct): New macro.
24393 (expand_while_optab_fn): New function.
24394 (convert_optab_supported_p): Likewise.
24395 (direct_while_optab_supported_p): New macro.
24396 * wide-int.h (wi::udiv_ceil): New function.
24397 * tree-vectorizer.h (rgroup_masks): New structure.
24398 (vec_loop_masks): New typedef.
24399 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
24400 and fully_masked_p.
24401 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
24402 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
24403 (vect_max_vf): New function.
24404 (slpeel_make_loop_iterate_ntimes): Delete.
24405 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
24406 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
24407 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
24408 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
24409 internal-fn.h, stor-layout.h and optabs-query.h.
24410 (vect_set_loop_mask): New function.
24411 (add_preheader_seq): Likewise.
24412 (add_header_seq): Likewise.
24413 (interleave_supported_p): Likewise.
24414 (vect_maybe_permute_loop_masks): Likewise.
24415 (vect_set_loop_masks_directly): Likewise.
24416 (vect_set_loop_condition_masked): Likewise.
24417 (vect_set_loop_condition_unmasked): New function, split out from
24418 slpeel_make_loop_iterate_ntimes.
24419 (slpeel_make_loop_iterate_ntimes): Rename to..
24420 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
24421 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
24422 (vect_do_peeling): Update call accordingly.
24423 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
24425 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
24426 mask_compare_type, can_fully_mask_p and fully_masked_p.
24427 (release_vec_loop_masks): New function.
24428 (_loop_vec_info): Use it to free the loop masks.
24429 (can_produce_all_loop_masks_p): New function.
24430 (vect_get_max_nscalars_per_iter): Likewise.
24431 (vect_verify_full_masking): Likewise.
24432 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
24433 retries, and free the mask rgroups before retrying. Check loop-wide
24434 reasons for disallowing fully-masked loops. Make the final decision
24435 about whether use a fully-masked loop or not.
24436 (vect_estimate_min_profitable_iters): Do not assume that peeling
24437 for the number of iterations will be needed for fully-masked loops.
24438 (vectorizable_reduction): Disable fully-masked loops.
24439 (vectorizable_live_operation): Likewise.
24440 (vect_halve_mask_nunits): New function.
24441 (vect_double_mask_nunits): Likewise.
24442 (vect_record_loop_mask): Likewise.
24443 (vect_get_loop_mask): Likewise.
24444 (vect_transform_loop): Handle the case in which the final loop
24445 iteration might handle a partial vector. Call vect_set_loop_condition
24446 instead of slpeel_make_loop_iterate_ntimes.
24447 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
24448 (check_load_store_masking): New function.
24449 (prepare_load_store_mask): Likewise.
24450 (vectorizable_store): Handle fully-masked loops.
24451 (vectorizable_load): Likewise.
24452 (supportable_widening_operation): Use vect_halve_mask_nunits for
24454 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
24455 (vect_gen_while): New function.
24456 * config/aarch64/aarch64.md (umax<mode>3): New expander.
24457 (aarch64_uqdec<mode>): New insn.
24459 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
24460 Alan Hayward <alan.hayward@arm.com>
24461 David Sherwood <david.sherwood@arm.com>
24463 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
24464 (reduc_xor_scal_optab): New optabs.
24465 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
24466 (reduc_xor_scal_@var{m}): Document.
24467 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
24468 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
24469 internal functions.
24470 * fold-const-call.c (fold_const_call): Handle them.
24471 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
24472 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
24473 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
24474 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
24475 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
24476 (UNSPEC_XORV): New unspecs.
24477 (optab): Add entries for them.
24478 (BITWISEV): New int iterator.
24479 (bit_reduc_op): New int attributes.
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/md.texi (vec_shl_insert_@var{m}): New optab.
24486 * internal-fn.def (VEC_SHL_INSERT): New internal function.
24487 * optabs.def (vec_shl_insert_optab): New optab.
24488 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
24489 (duplicate_and_interleave): Likewise.
24490 * tree-vect-loop.c: Include internal-fn.h.
24491 (neutral_op_for_slp_reduction): New function, split out from
24492 get_initial_defs_for_reduction.
24493 (get_initial_def_for_reduction): Handle option 2 for variable-length
24494 vectors by loading the neutral value into a vector and then shifting
24495 the initial value into element 0.
24496 (get_initial_defs_for_reduction): Replace the code argument with
24497 the neutral value calculated by neutral_op_for_slp_reduction.
24498 Use gimple_build_vector for constant-length vectors.
24499 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
24500 but the first group_size elements have a neutral value.
24501 Use duplicate_and_interleave otherwise.
24502 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
24503 Update call to get_initial_defs_for_reduction. Handle SLP
24504 reductions for variable-length vectors by creating one vector
24505 result for each scalar result, with the elements associated
24506 with other scalar results stubbed out with the neutral value.
24507 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
24508 Require IFN_VEC_SHL_INSERT for double reductions on
24509 variable-length vectors, or SLP reductions that have
24510 a neutral value. Require can_duplicate_and_interleave_p
24511 support for variable-length unchained SLP reductions if there
24512 is no neutral value, such as for MIN/MAX reductions. Also require
24513 the number of vector elements to be a multiple of the number of
24514 SLP statements when doing variable-length unchained SLP reductions.
24515 Update call to vect_create_epilog_for_reduction.
24516 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
24517 and remove initial values.
24518 (duplicate_and_interleave): Make public.
24519 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
24520 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
24522 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
24523 Alan Hayward <alan.hayward@arm.com>
24524 David Sherwood <david.sherwood@arm.com>
24526 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
24527 (can_duplicate_and_interleave_p): New function.
24528 (vect_get_and_check_slp_defs): Take the vector of statements
24529 rather than just the current one. Remove excess parentheses.
24530 Restriction rejectinon of vect_constant_def and vect_external_def
24531 for variable-length vectors to boolean types, or types for which
24532 can_duplicate_and_interleave_p is false.
24533 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
24534 (duplicate_and_interleave): New function.
24535 (vect_get_constant_vectors): Use gimple_build_vector for
24536 constant-length vectors and suitable variable-length constant
24537 vectors. Use duplicate_and_interleave for other variable-length
24538 vectors. Don't defer the update when inserting new statements.
24540 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
24541 Alan Hayward <alan.hayward@arm.com>
24542 David Sherwood <david.sherwood@arm.com>
24544 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
24545 min_profitable_iters doesn't go negative.
24547 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
24548 Alan Hayward <alan.hayward@arm.com>
24549 David Sherwood <david.sherwood@arm.com>
24551 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
24552 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
24553 * optabs.def (vec_mask_load_lanes_optab): New optab.
24554 (vec_mask_store_lanes_optab): Likewise.
24555 * internal-fn.def (MASK_LOAD_LANES): New internal function.
24556 (MASK_STORE_LANES): Likewise.
24557 * internal-fn.c (mask_load_lanes_direct): New macro.
24558 (mask_store_lanes_direct): Likewise.
24559 (expand_mask_load_optab_fn): Handle masked operations.
24560 (expand_mask_load_lanes_optab_fn): New macro.
24561 (expand_mask_store_optab_fn): Handle masked operations.
24562 (expand_mask_store_lanes_optab_fn): New macro.
24563 (direct_mask_load_lanes_optab_supported_p): Likewise.
24564 (direct_mask_store_lanes_optab_supported_p): Likewise.
24565 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
24567 (vect_load_lanes_supported): Likewise.
24568 * tree-vect-data-refs.c (strip_conversion): New function.
24569 (can_group_stmts_p): Likewise.
24570 (vect_analyze_data_ref_accesses): Use it instead of checking
24571 for a pair of assignments.
24572 (vect_store_lanes_supported): Take a masked_p parameter.
24573 (vect_load_lanes_supported): Likewise.
24574 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
24575 vect_store_lanes_supported and vect_load_lanes_supported.
24576 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
24577 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
24578 parameter. Don't allow gaps for masked accesses.
24579 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
24580 and vect_load_lanes_supported.
24581 (get_load_store_type): Take a masked_p parameter and update
24582 call to get_group_load_store_type.
24583 (vectorizable_store): Update call to get_load_store_type.
24584 Handle IFN_MASK_STORE_LANES.
24585 (vectorizable_load): Update call to get_load_store_type.
24586 Handle IFN_MASK_LOAD_LANES.
24588 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
24589 Alan Hayward <alan.hayward@arm.com>
24590 David Sherwood <david.sherwood@arm.com>
24592 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
24594 * config/aarch64/aarch64-protos.h
24595 (aarch64_sve_struct_memory_operand_p): Declare.
24596 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
24597 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
24598 (VPRED, vpred): Handle SVE structure modes.
24599 * config/aarch64/constraints.md (Utx): New constraint.
24600 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
24601 (aarch64_sve_struct_nonimmediate_operand): New predicates.
24602 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
24603 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
24604 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
24605 structure modes. Split into pieces after RA.
24606 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
24607 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
24609 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
24610 SVE structure modes.
24611 (aarch64_classify_address): Likewise.
24612 (sizetochar): Move earlier in file.
24613 (aarch64_print_operand): Handle SVE register lists.
24614 (aarch64_array_mode): New function.
24615 (aarch64_sve_struct_memory_operand_p): Likewise.
24616 (TARGET_ARRAY_MODE): Redefine.
24618 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
24619 Alan Hayward <alan.hayward@arm.com>
24620 David Sherwood <david.sherwood@arm.com>
24622 * target.def (array_mode): New target hook.
24623 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
24624 * doc/tm.texi: Regenerate.
24625 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
24626 * hooks.c (hook_optmode_mode_uhwi_none): New function.
24627 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
24628 targetm.array_mode.
24629 * stor-layout.c (mode_for_array): Likewise. Support polynomial
24632 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
24633 Alan Hayward <alan.hayward@arm.com>
24634 David Sherwood <david.sherwood@arm.com>
24636 * fold-const.c (fold_binary_loc): Check the argument types
24637 rather than the result type when testing for a vector operation.
24639 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
24641 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
24642 * doc/tm.texi: Regenerate.
24644 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
24645 Alan Hayward <alan.hayward@arm.com>
24646 David Sherwood <david.sherwood@arm.com>
24648 * doc/invoke.texi (-msve-vector-bits=): Document new option.
24649 (sve): Document new AArch64 extension.
24650 * doc/md.texi (w): Extend the description of the AArch64
24651 constraint to include SVE vectors.
24652 (Upl, Upa): Document new AArch64 predicate constraints.
24653 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
24655 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
24656 (msve-vector-bits=): New option.
24657 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
24658 SVE when these are disabled.
24659 (sve): New extension.
24660 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
24661 modes. Adjust their number of units based on aarch64_sve_vg.
24662 (MAX_BITSIZE_MODE_ANY_MODE): Define.
24663 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
24664 aarch64_addr_query_type.
24665 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
24666 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
24667 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
24668 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
24669 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
24670 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
24671 (aarch64_simd_imm_zero_p): Delete.
24672 (aarch64_check_zero_based_sve_index_immediate): Declare.
24673 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
24674 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
24675 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
24676 (aarch64_sve_float_mul_immediate_p): Likewise.
24677 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
24678 rather than an rtx.
24679 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
24680 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
24681 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
24682 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
24683 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
24684 (aarch64_regmode_natural_size): Likewise.
24685 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
24686 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
24688 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
24689 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
24690 for VG and the SVE predicate registers.
24691 (V_ALIASES): Add a "z"-prefixed alias.
24692 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
24693 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
24694 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
24695 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
24696 (REG_CLASS_NAMES): Add entries for them.
24697 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
24698 and the predicate registers.
24699 (aarch64_sve_vg): Declare.
24700 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
24701 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
24702 (REGMODE_NATURAL_SIZE): Define.
24703 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
24705 * config/aarch64/aarch64.c: Include cfgrtl.h.
24706 (simd_immediate_info): Add a constructor for series vectors,
24707 and an associated step field.
24708 (aarch64_sve_vg): New variable.
24709 (aarch64_dbx_register_number): Handle VG and the predicate registers.
24710 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
24711 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
24712 (VEC_ANY_DATA, VEC_STRUCT): New constants.
24713 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
24714 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
24715 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
24716 (aarch64_get_mask_mode): New functions.
24717 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
24718 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
24719 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
24720 predicate modes and predicate registers. Explicitly restrict
24721 GPRs to modes of 16 bytes or smaller. Only allow FP registers
24722 to store a vector mode if it is recognized by
24723 aarch64_classify_vector_mode.
24724 (aarch64_regmode_natural_size): New function.
24725 (aarch64_hard_regno_caller_save_mode): Return the original mode
24727 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
24728 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
24729 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
24730 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
24732 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
24733 does not overlap dest if the function is frame-related. Handle
24735 (aarch64_split_add_offset): New function.
24736 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
24737 them aarch64_add_offset.
24738 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
24739 and update call to aarch64_sub_sp.
24740 (aarch64_add_cfa_expression): New function.
24741 (aarch64_expand_prologue): Pass extra temporary registers to the
24742 functions above. Handle the case in which we need to emit new
24743 DW_CFA_expressions for registers that were originally saved
24744 relative to the stack pointer, but now have to be expressed
24745 relative to the frame pointer.
24746 (aarch64_output_mi_thunk): Pass extra temporary registers to the
24748 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
24749 IP0 and IP1 values for SVE frames.
24750 (aarch64_expand_vec_series): New function.
24751 (aarch64_expand_sve_widened_duplicate): Likewise.
24752 (aarch64_expand_sve_const_vector): Likewise.
24753 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
24754 Handle SVE constants. Use emit_move_insn to move a force_const_mem
24755 into the register, rather than emitting a SET directly.
24756 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
24757 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
24758 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
24759 (offset_9bit_signed_scaled_p): New functions.
24760 (aarch64_replicate_bitmask_imm): New function.
24761 (aarch64_bitmask_imm): Use it.
24762 (aarch64_cannot_force_const_mem): Reject expressions involving
24763 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
24764 (aarch64_classify_index): Handle SVE indices, by requiring
24765 a plain register index with a scale that matches the element size.
24766 (aarch64_classify_address): Handle SVE addresses. Assert that
24767 the mode of the address is VOIDmode or an integer mode.
24768 Update call to aarch64_classify_symbol.
24769 (aarch64_classify_symbolic_expression): Update call to
24770 aarch64_classify_symbol.
24771 (aarch64_const_vec_all_in_range_p): New function.
24772 (aarch64_print_vector_float_operand): Likewise.
24773 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
24774 "vN" for FP registers with SVE modes. Handle (const ...) vectors
24775 and the FP immediates 1.0 and 0.5.
24776 (aarch64_print_address_internal): Handle SVE addresses.
24777 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
24778 (aarch64_regno_regclass): Handle predicate registers.
24779 (aarch64_secondary_reload): Handle big-endian reloads of SVE
24781 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
24782 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
24783 (aarch64_convert_sve_vector_bits): New function.
24784 (aarch64_override_options): Use it to handle -msve-vector-bits=.
24785 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
24786 rather than an rtx.
24787 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
24788 Handle SVE vector and predicate modes. Accept VL-based constants
24789 that need only one temporary register, and VL offsets that require
24790 no temporary registers.
24791 (aarch64_conditional_register_usage): Mark the predicate registers
24792 as fixed if SVE isn't available.
24793 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
24794 Return true for SVE vector and predicate modes.
24795 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
24796 rather than an unsigned int. Handle SVE modes.
24797 (aarch64_preferred_simd_mode): Update call accordingly. Handle
24799 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
24801 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
24802 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
24803 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
24804 (aarch64_sve_float_mul_immediate_p): New functions.
24805 (aarch64_sve_valid_immediate): New function.
24806 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
24807 Explicitly reject structure modes. Check for INDEX constants.
24808 Handle PTRUE and PFALSE constants.
24809 (aarch64_check_zero_based_sve_index_immediate): New function.
24810 (aarch64_simd_imm_zero_p): Delete.
24811 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
24812 vector modes. Accept constants in the range of CNT[BHWD].
24813 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
24814 ask for an Advanced SIMD mode.
24815 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
24816 (aarch64_simd_vector_alignment): Handle SVE predicates.
24817 (aarch64_vectorize_preferred_vector_alignment): New function.
24818 (aarch64_simd_vector_alignment_reachable): Use it instead of
24820 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
24821 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
24823 (MAX_VECT_LEN): Delete.
24824 (expand_vec_perm_d): Add a vec_flags field.
24825 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
24826 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
24827 (aarch64_evpc_ext): Don't apply a big-endian lane correction
24829 (aarch64_evpc_rev): Rename to...
24830 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
24831 (aarch64_evpc_rev_global): New function.
24832 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
24833 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
24835 (aarch64_evpc_sve_tbl): New function.
24836 (aarch64_expand_vec_perm_const_1): Update after rename of
24837 aarch64_evpc_rev. Handle SVE permutes too, trying
24838 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
24839 than aarch64_evpc_tbl.
24840 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
24841 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
24842 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
24843 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
24844 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
24845 (aarch64_expand_sve_vcond): New functions.
24846 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
24847 of aarch64_vector_mode_p.
24848 (aarch64_dwarf_poly_indeterminate_value): New function.
24849 (aarch64_compute_pressure_classes): Likewise.
24850 (aarch64_can_change_mode_class): Likewise.
24851 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
24852 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
24853 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
24854 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
24855 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
24856 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
24857 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
24858 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
24860 (Dn, Dl, Dr): Accept const as well as const_vector.
24861 (Dz): Likewise. Compare against CONST0_RTX.
24862 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
24863 of "vector" where appropriate.
24864 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
24865 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
24866 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
24867 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
24868 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
24869 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
24870 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
24871 (v_int_equiv): Extend to SVE modes.
24872 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
24874 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
24875 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
24876 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
24877 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
24878 (SVE_COND_FP_CMP): New int iterators.
24879 (perm_hilo): Handle the new unpack unspecs.
24880 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
24882 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
24883 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
24884 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
24885 (aarch64_equality_operator, aarch64_constant_vector_operand)
24886 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
24887 (aarch64_sve_nonimmediate_operand): Likewise.
24888 (aarch64_sve_general_operand): Likewise.
24889 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
24890 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
24891 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
24892 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
24893 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
24894 (aarch64_sve_float_arith_immediate): Likewise.
24895 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
24896 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
24897 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
24898 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
24899 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
24900 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
24901 (aarch64_sve_float_arith_operand): Likewise.
24902 (aarch64_sve_float_arith_with_sub_operand): Likewise.
24903 (aarch64_sve_float_mul_operand): Likewise.
24904 (aarch64_sve_vec_perm_operand): Likewise.
24905 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
24906 (aarch64_mov_operand): Accept const_poly_int and const_vector.
24907 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
24908 as well as const_vector.
24909 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
24910 in file. Use CONST0_RTX and CONSTM1_RTX.
24911 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
24912 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
24913 Use aarch64_simd_imm_zero.
24914 * config/aarch64/aarch64-sve.md: New file.
24915 * config/aarch64/aarch64.md: Include it.
24916 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
24917 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
24918 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
24919 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
24920 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
24921 (sve): New attribute.
24922 (enabled): Disable instructions with the sve attribute unless
24924 (movqi, movhi): Pass CONST_POLY_INT operaneds through
24925 aarch64_expand_mov_immediate.
24926 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
24927 CNT[BHSD] immediates.
24928 (movti): Split CONST_POLY_INT moves into two halves.
24929 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
24930 Split additions that need a temporary here if the destination
24931 is the stack pointer.
24932 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
24933 (*add<mode>3_poly_1): New instruction.
24934 (set_clobber_cc): New expander.
24936 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
24938 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
24939 parameter and use it instead of GET_MODE_SIZE (innermode). Use
24940 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
24941 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
24942 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
24943 Change innermode from fixed_mode_size to machine_mode.
24944 (simplify_subreg): Update call accordingly. Handle a constant-sized
24945 subreg of a variable-length CONST_VECTOR.
24947 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
24948 Alan Hayward <alan.hayward@arm.com>
24949 David Sherwood <david.sherwood@arm.com>
24951 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
24952 (add_offset_to_base): New function, split out from...
24953 (create_mem_ref): ...here. When handling a scale other than 1,
24954 check first whether the address is valid without the offset.
24955 Add it into the base if so, leaving the index and scale as-is.
24957 2018-01-12 Jakub Jelinek <jakub@redhat.com>
24960 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
24961 fold_for_warn before checking if arg2 is INTEGER_CST.
24963 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
24965 * config/rs6000/predicates.md (load_multiple_operation): Delete.
24966 (store_multiple_operation): Delete.
24967 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
24968 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
24969 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
24970 guarded by TARGET_STRING.
24971 (rs6000_output_load_multiple): Delete.
24972 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
24973 OPTION_MASK_STRING / TARGET_STRING handling.
24974 (print_operand) <'N', 'O'>: Add comment that these are unused now.
24975 (const rs6000_opt_masks) <"string">: Change mask to 0.
24976 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
24977 (MASK_STRING): Delete.
24978 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
24980 (load_multiple): Delete.
24987 (store_multiple): Delete.
24994 (movmemsi_8reg): Delete.
24995 (corresponding unnamed define_insn): Delete.
24996 (movmemsi_6reg): Delete.
24997 (corresponding unnamed define_insn): Delete.
24998 (movmemsi_4reg): Delete.
24999 (corresponding unnamed define_insn): Delete.
25000 (movmemsi_2reg): Delete.
25001 (corresponding unnamed define_insn): Delete.
25002 (movmemsi_1reg): Delete.
25003 (corresponding unnamed define_insn): Delete.
25004 * config/rs6000/rs6000.opt (mno-string): New.
25005 (mstring): Replace by deprecation warning stub.
25006 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
25008 2018-01-12 Jakub Jelinek <jakub@redhat.com>
25010 * regrename.c (regrename_do_replace): If replacing the same
25011 reg multiple times, try to reuse last created gen_raw_REG.
25014 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
25015 main to workaround a bug in GDB.
25017 2018-01-12 Tom de Vries <tom@codesourcery.com>
25020 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
25022 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
25024 PR rtl-optimization/80481
25025 * ira-color.c (get_cap_member): New function.
25026 (allocnos_conflict_by_live_ranges_p): Use it.
25027 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
25028 (setup_slot_coalesced_allocno_live_ranges): Ditto.
25030 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
25033 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
25034 (*saddl_se_1): Ditto.
25035 (*ssubsi_1): Ditto.
25036 (*ssubl_se_1): Ditto.
25038 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
25040 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
25041 rather than wi::to_widest for DR_INITs.
25042 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
25043 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
25044 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
25046 (vect_analyze_group_access_1): Note that here.
25048 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
25050 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
25051 polynomial type sizes.
25053 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
25055 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
25056 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
25057 (gimple_add_tmp_var): Likewise.
25059 2018-01-12 Martin Liska <mliska@suse.cz>
25061 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
25062 (gimple_alloc_sizes): Likewise.
25063 (dump_gimple_statistics): Use PRIu64 in printf format.
25064 * gimple.h: Change uint64_t to int.
25066 2018-01-12 Martin Liska <mliska@suse.cz>
25068 * tree-core.h: Use uint64_t instead of int.
25069 * tree.c (tree_node_counts): Likewise.
25070 (tree_node_sizes): Likewise.
25071 (dump_tree_statistics): Use PRIu64 in printf format.
25073 2018-01-12 Martin Liska <mliska@suse.cz>
25075 * Makefile.in: As qsort_chk is implemented in vec.c, add
25076 vec.o to linkage of gencfn-macros.
25077 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
25078 passing the info to record_node_allocation_statistics.
25079 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
25081 * ggc-common.c (struct ggc_usage): Add operator== and use
25082 it in operator< and compare function.
25083 * mem-stats.h (struct mem_usage): Likewise.
25084 * vec.c (struct vec_usage): Remove operator< and compare
25085 function. Can be simply inherited.
25087 2018-01-12 Martin Jambor <mjambor@suse.cz>
25090 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
25091 * tree-ssa-math-opts.c: Include domwalk.h.
25092 (convert_mult_to_fma_1): New function.
25093 (fma_transformation_info): New type.
25094 (fma_deferring_state): Likewise.
25095 (cancel_fma_deferring): New function.
25096 (result_of_phi): Likewise.
25097 (last_fma_candidate_feeds_initial_phi): Likewise.
25098 (convert_mult_to_fma): Added deferring logic, split actual
25099 transformation to convert_mult_to_fma_1.
25100 (math_opts_dom_walker): New type.
25101 (math_opts_dom_walker::after_dom_children): New method, body moved
25102 here from pass_optimize_widening_mul::execute, added deferring logic
25104 (pass_optimize_widening_mul::execute): Moved most of code to
25105 math_opts_dom_walker::after_dom_children.
25106 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
25107 * config/i386/i386.c (ix86_option_override_internal): Added
25108 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
25110 2018-01-12 Richard Biener <rguenther@suse.de>
25113 * dwarf2out.c (gen_variable_die): Do not reset old_die for
25114 inline instance vars.
25116 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
25119 * config/rx/rx.c (rx_is_restricted_memory_address):
25120 Handle SUBREG case.
25122 2018-01-12 Richard Biener <rguenther@suse.de>
25124 PR tree-optimization/80846
25125 * target.def (split_reduction): New target hook.
25126 * targhooks.c (default_split_reduction): New function.
25127 * targhooks.h (default_split_reduction): Declare.
25128 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
25129 target requests first reduce vectors by combining low and high
25131 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
25132 (get_vectype_for_scalar_type_and_size): Export.
25133 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
25134 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
25135 * doc/tm.texi: Regenerate.
25136 * config/i386/i386.c (ix86_split_reduction): Implement
25137 TARGET_VECTORIZE_SPLIT_REDUCTION.
25139 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
25142 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
25143 in PIC mode except for TARGET_VXWORKS_RTP.
25144 * config/sparc/sparc.c: Include cfgrtl.h.
25145 (TARGET_INIT_PIC_REG): Define.
25146 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
25147 (sparc_pic_register_p): New predicate.
25148 (sparc_legitimate_address_p): Use it.
25149 (sparc_legitimize_pic_address): Likewise.
25150 (sparc_delegitimize_address): Likewise.
25151 (sparc_mode_dependent_address_p): Likewise.
25152 (gen_load_pcrel_sym): Remove 4th parameter.
25153 (load_got_register): Adjust call to above. Remove obsolete stuff.
25154 (sparc_expand_prologue): Do not call load_got_register here.
25155 (sparc_flat_expand_prologue): Likewise.
25156 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
25157 (sparc_use_pseudo_pic_reg): New function.
25158 (sparc_init_pic_reg): Likewise.
25159 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
25160 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
25162 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
25164 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
25165 Add item for branch_cost.
25167 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
25169 PR rtl-optimization/83565
25170 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
25171 not extend the result to a larger mode for rotate operations.
25172 (num_sign_bit_copies1): Likewise.
25174 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
25177 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
25179 Use values-Xc.o for -pedantic.
25180 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
25182 2018-01-12 Martin Liska <mliska@suse.cz>
25185 * ipa-devirt.c (final_warning_record::grow_type_warnings):
25187 (possible_polymorphic_call_targets): Use it.
25188 (ipa_devirt): Likewise.
25190 2018-01-12 Martin Liska <mliska@suse.cz>
25192 * profile-count.h (enum profile_quality): Use 0 as invalid
25193 enum value of profile_quality.
25195 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
25197 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
25198 -mext-string options.
25200 2018-01-12 Richard Biener <rguenther@suse.de>
25202 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
25203 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
25204 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
25206 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
25208 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
25210 * configure.ac (--with-long-double-format): Add support for the
25211 configuration option to change the default long double format on
25213 * config.gcc (powerpc*-linux*-*): Likewise.
25214 * configure: Regenerate.
25215 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
25216 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
25217 used without modification.
25219 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
25221 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
25222 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
25223 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
25224 MISC_BUILTIN_SPEC_BARRIER.
25225 (rs6000_init_builtins): Likewise.
25226 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
25228 (speculation_barrier): New define_insn.
25229 * doc/extend.texi: Document __builtin_speculation_barrier.
25231 2018-01-11 Jakub Jelinek <jakub@redhat.com>
25234 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
25235 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
25236 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
25238 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
25239 integral modes instead of "ss" and "sd".
25240 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
25241 vectors with 32-bit and 64-bit elements.
25242 (vecdupssescalarmodesuffix): New mode attribute.
25243 (vec_dup<mode>): Use it.
25245 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
25248 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
25249 frame if argument is passed on stack.
25251 2018-01-11 Jakub Jelinek <jakub@redhat.com>
25254 * ree.c (combine_reaching_defs): Optimize also
25255 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
25256 reg2=any_extend(exp); reg1=reg2;, formatting fix.
25258 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
25260 PR middle-end/83189
25261 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
25263 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
25265 PR middle-end/83718
25266 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
25267 after they are computed.
25269 2018-01-11 Bin Cheng <bin.cheng@arm.com>
25271 PR tree-optimization/83695
25272 * gimple-loop-linterchange.cc
25273 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
25274 reset cached scev information after interchange.
25275 (pass_linterchange::execute): Remove call to scev_reset_htab.
25277 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
25279 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
25280 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
25281 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
25282 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
25283 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
25284 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
25285 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
25286 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
25287 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
25288 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
25289 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
25290 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
25291 (V_lane_reg): Likewise.
25292 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
25294 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
25295 (vfmal_lane_low<mode>_intrinsic,
25296 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
25297 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
25298 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
25299 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
25300 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
25301 vfmsl_lane_high<mode>_intrinsic): New define_insns.
25303 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
25305 * config/arm/arm-cpus.in (fp16fml): New feature.
25306 (ALL_SIMD): Add fp16fml.
25307 (armv8.2-a): Add fp16fml as an option.
25308 (armv8.3-a): Likewise.
25309 (armv8.4-a): Add fp16fml as part of fp16.
25310 * config/arm/arm.h (TARGET_FP16FML): Define.
25311 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
25313 * config/arm/arm-modes.def (V2HF): Define.
25314 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
25315 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
25316 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
25317 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
25318 vfmsl_low, vfmsl_high): New set of builtins.
25319 * config/arm/iterators.md (PLUSMINUS): New code iterator.
25320 (vfml_op): New code attribute.
25321 (VFMLHALVES): New int iterator.
25322 (VFML, VFMLSEL): New mode attributes.
25323 (V_reg): Define mapping for V2HF.
25324 (V_hi, V_lo): New mode attributes.
25325 (VF_constraint): Likewise.
25326 (vfml_half, vfml_half_selector): New int attributes.
25327 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
25329 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
25330 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
25332 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
25333 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
25334 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
25335 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
25337 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
25338 Document new effective target and option set.
25340 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
25342 * config/arm/arm-cpus.in (armv8_4): New feature.
25343 (ARMv8_4a): New fgroup.
25344 (armv8.4-a): New arch.
25345 * config/arm/arm-tables.opt: Regenerate.
25346 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
25347 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
25348 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
25349 Add matching rules for -march=armv8.4-a and extensions.
25350 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
25352 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
25355 * config/rx/rx.md (BW): New mode attribute.
25356 (sync_lock_test_and_setsi): Add mode suffix to insn output.
25358 2018-01-11 Richard Biener <rguenther@suse.de>
25360 PR tree-optimization/83435
25361 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
25362 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
25363 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
25365 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
25366 Alan Hayward <alan.hayward@arm.com>
25367 David Sherwood <david.sherwood@arm.com>
25369 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
25371 (aarch64_classify_address): Initialize it. Track polynomial offsets.
25372 (aarch64_print_address_internal): Use it to check for a zero offset.
25374 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
25375 Alan Hayward <alan.hayward@arm.com>
25376 David Sherwood <david.sherwood@arm.com>
25378 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
25379 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
25380 Return a poly_int64 rather than a HOST_WIDE_INT.
25381 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
25382 rather than a HOST_WIDE_INT.
25383 * config/aarch64/aarch64.h (aarch64_frame): Protect with
25384 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
25385 hard_fp_offset, frame_size, initial_adjust, callee_offset and
25386 final_offset from HOST_WIDE_INT to poly_int64.
25387 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
25388 to_constant when getting the number of units in an Advanced SIMD
25390 (aarch64_builtin_vectorized_function): Check for a constant number
25392 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
25394 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
25395 attribute instead of GET_MODE_NUNITS.
25396 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
25397 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
25398 GET_MODE_SIZE for fixed-size registers.
25399 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
25400 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
25401 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
25402 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
25403 (aarch64_print_operand, aarch64_print_address_internal)
25404 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
25405 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
25406 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
25407 Handle polynomial GET_MODE_SIZE.
25408 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
25409 wider than SImode without modification.
25410 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
25411 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
25412 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
25413 passing and returning SVE modes.
25414 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
25415 rather than GEN_INT.
25416 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
25417 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
25418 (aarch64_allocate_and_probe_stack_space): Likewise.
25419 (aarch64_layout_frame): Cope with polynomial offsets.
25420 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
25421 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
25422 polynomial offsets.
25423 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
25424 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
25425 poly_int64 rather than a HOST_WIDE_INT.
25426 (aarch64_get_separate_components, aarch64_process_components)
25427 (aarch64_expand_prologue, aarch64_expand_epilogue)
25428 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
25429 (aarch64_anchor_offset): New function, split out from...
25430 (aarch64_legitimize_address): ...here.
25431 (aarch64_builtin_vectorization_cost): Handle polynomial
25432 TYPE_VECTOR_SUBPARTS.
25433 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
25435 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
25436 number of elements from the PARALLEL rather than the mode.
25437 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
25438 rather than GET_MODE_BITSIZE.
25439 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
25440 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
25441 (aarch64_expand_vec_perm_const_1): Handle polynomial
25442 d->perm.length () and d->perm elements.
25443 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
25444 Apply to_constant to d->perm elements.
25445 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
25446 polynomial CONST_VECTOR_NUNITS.
25447 (aarch64_move_pointer): Take amount as a poly_int64 rather
25449 (aarch64_progress_pointer): Avoid temporary variable.
25450 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
25451 the mode attribute instead of GET_MODE.
25453 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
25454 Alan Hayward <alan.hayward@arm.com>
25455 David Sherwood <david.sherwood@arm.com>
25457 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
25458 x exists before using it.
25459 (aarch64_add_constant_internal): Rename to...
25460 (aarch64_add_offset_1): ...this. Replace regnum with separate
25461 src and dest rtxes. Handle the case in which they're different,
25462 including when the offset is zero. Replace scratchreg with an rtx.
25463 Use 2 additions if there is no spare register into which we can
25464 move a 16-bit constant.
25465 (aarch64_add_constant): Delete.
25466 (aarch64_add_offset): Replace reg with separate src and dest
25467 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
25468 Use aarch64_add_offset_1.
25469 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
25470 an rtx rather than an int. Take the delta as a poly_int64
25471 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
25472 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
25473 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
25474 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
25475 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
25476 and aarch64_add_sp.
25477 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
25478 aarch64_add_constant.
25480 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
25482 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
25483 Use scalar_float_mode.
25485 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
25487 * config/aarch64/aarch64-simd.md
25488 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
25489 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
25490 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
25491 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
25492 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
25493 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
25494 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
25495 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
25496 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
25497 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
25499 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
25502 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
25503 targ_options->x_arm_arch_string is non NULL.
25505 2018-01-11 Tamar Christina <tamar.christina@arm.com>
25507 * config/aarch64/aarch64.h
25508 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
25510 2018-01-11 Sudakshina Das <sudi.das@arm.com>
25513 * expmed.c (emit_store_flag_force): Swap if const op0
25514 and change VOIDmode to mode of op0.
25516 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
25518 PR rtl-optimization/83761
25519 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
25520 than bytes to mode_for_size.
25522 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
25524 PR middle-end/83189
25525 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
25526 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
25529 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
25531 PR middle-end/83575
25532 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
25533 when in layout mode.
25534 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
25535 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
25538 2018-01-10 Michael Collison <michael.collison@arm.com>
25540 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
25541 * config/aarch64/aarch64-option-extension.def: Add
25542 AARCH64_OPT_EXTENSION of 'fp16fml'.
25543 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
25544 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
25545 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
25546 * config/aarch64/constraints.md (Ui7): New constraint.
25547 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
25548 (VFMLA_SEL_W): Ditto.
25551 (VFMLA16_LOW): New int iterator.
25552 (VFMLA16_HIGH): Ditto.
25553 (UNSPEC_FMLAL): New unspec.
25554 (UNSPEC_FMLSL): Ditto.
25555 (UNSPEC_FMLAL2): Ditto.
25556 (UNSPEC_FMLSL2): Ditto.
25557 (f16mac): New code attribute.
25558 * config/aarch64/aarch64-simd-builtins.def
25559 (aarch64_fmlal_lowv2sf): Ditto.
25560 (aarch64_fmlsl_lowv2sf): Ditto.
25561 (aarch64_fmlalq_lowv4sf): Ditto.
25562 (aarch64_fmlslq_lowv4sf): Ditto.
25563 (aarch64_fmlal_highv2sf): Ditto.
25564 (aarch64_fmlsl_highv2sf): Ditto.
25565 (aarch64_fmlalq_highv4sf): Ditto.
25566 (aarch64_fmlslq_highv4sf): Ditto.
25567 (aarch64_fmlal_lane_lowv2sf): Ditto.
25568 (aarch64_fmlsl_lane_lowv2sf): Ditto.
25569 (aarch64_fmlal_laneq_lowv2sf): Ditto.
25570 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
25571 (aarch64_fmlalq_lane_lowv4sf): Ditto.
25572 (aarch64_fmlsl_lane_lowv4sf): Ditto.
25573 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
25574 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
25575 (aarch64_fmlal_lane_highv2sf): Ditto.
25576 (aarch64_fmlsl_lane_highv2sf): Ditto.
25577 (aarch64_fmlal_laneq_highv2sf): Ditto.
25578 (aarch64_fmlsl_laneq_highv2sf): Ditto.
25579 (aarch64_fmlalq_lane_highv4sf): Ditto.
25580 (aarch64_fmlsl_lane_highv4sf): Ditto.
25581 (aarch64_fmlalq_laneq_highv4sf): Ditto.
25582 (aarch64_fmlsl_laneq_highv4sf): Ditto.
25583 * config/aarch64/aarch64-simd.md:
25584 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
25585 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
25586 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
25587 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
25588 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
25589 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
25590 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
25591 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
25592 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
25593 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
25594 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
25595 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
25596 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
25597 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
25598 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
25599 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
25600 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
25601 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
25602 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
25603 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
25604 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
25605 (vfmlsl_low_u32): Ditto.
25606 (vfmlalq_low_u32): Ditto.
25607 (vfmlslq_low_u32): Ditto.
25608 (vfmlal_high_u32): Ditto.
25609 (vfmlsl_high_u32): Ditto.
25610 (vfmlalq_high_u32): Ditto.
25611 (vfmlslq_high_u32): Ditto.
25612 (vfmlal_lane_low_u32): Ditto.
25613 (vfmlsl_lane_low_u32): Ditto.
25614 (vfmlal_laneq_low_u32): Ditto.
25615 (vfmlsl_laneq_low_u32): Ditto.
25616 (vfmlalq_lane_low_u32): Ditto.
25617 (vfmlslq_lane_low_u32): Ditto.
25618 (vfmlalq_laneq_low_u32): Ditto.
25619 (vfmlslq_laneq_low_u32): Ditto.
25620 (vfmlal_lane_high_u32): Ditto.
25621 (vfmlsl_lane_high_u32): Ditto.
25622 (vfmlal_laneq_high_u32): Ditto.
25623 (vfmlsl_laneq_high_u32): Ditto.
25624 (vfmlalq_lane_high_u32): Ditto.
25625 (vfmlslq_lane_high_u32): Ditto.
25626 (vfmlalq_laneq_high_u32): Ditto.
25627 (vfmlslq_laneq_high_u32): Ditto.
25628 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
25629 (AARCH64_FL_FOR_ARCH8_4): New.
25630 (AARCH64_ISA_F16FML): New ISA flag.
25631 (TARGET_F16FML): New feature flag for fp16fml.
25632 (doc/invoke.texi): Document new fp16fml option.
25634 2018-01-10 Michael Collison <michael.collison@arm.com>
25636 * config/aarch64/aarch64-builtins.c:
25637 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
25638 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
25639 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
25640 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
25641 (AARCH64_ISA_SHA3): New ISA flag.
25642 (TARGET_SHA3): New feature flag for sha3.
25643 * config/aarch64/iterators.md (sha512_op): New int attribute.
25644 (CRYPTO_SHA512): New int iterator.
25645 (UNSPEC_SHA512H): New unspec.
25646 (UNSPEC_SHA512H2): Ditto.
25647 (UNSPEC_SHA512SU0): Ditto.
25648 (UNSPEC_SHA512SU1): Ditto.
25649 * config/aarch64/aarch64-simd-builtins.def
25650 (aarch64_crypto_sha512hqv2di): New builtin.
25651 (aarch64_crypto_sha512h2qv2di): Ditto.
25652 (aarch64_crypto_sha512su0qv2di): Ditto.
25653 (aarch64_crypto_sha512su1qv2di): Ditto.
25654 (aarch64_eor3qv8hi): Ditto.
25655 (aarch64_rax1qv2di): Ditto.
25656 (aarch64_xarqv2di): Ditto.
25657 (aarch64_bcaxqv8hi): Ditto.
25658 * config/aarch64/aarch64-simd.md:
25659 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
25660 (aarch64_crypto_sha512su0qv2di): Ditto.
25661 (aarch64_crypto_sha512su1qv2di): Ditto.
25662 (aarch64_eor3qv8hi): Ditto.
25663 (aarch64_rax1qv2di): Ditto.
25664 (aarch64_xarqv2di): Ditto.
25665 (aarch64_bcaxqv8hi): Ditto.
25666 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
25667 (vsha512h2q_u64): Ditto.
25668 (vsha512su0q_u64): Ditto.
25669 (vsha512su1q_u64): Ditto.
25670 (veor3q_u16): Ditto.
25671 (vrax1q_u64): Ditto.
25672 (vxarq_u64): Ditto.
25673 (vbcaxq_u16): Ditto.
25674 * config/arm/types.md (crypto_sha512): New type attribute.
25675 (crypto_sha3): Ditto.
25676 (doc/invoke.texi): Document new sha3 option.
25678 2018-01-10 Michael Collison <michael.collison@arm.com>
25680 * config/aarch64/aarch64-builtins.c:
25681 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
25682 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
25683 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
25684 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
25685 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
25686 (AARCH64_ISA_SM4): New ISA flag.
25687 (TARGET_SM4): New feature flag for sm4.
25688 * config/aarch64/aarch64-simd-builtins.def
25689 (aarch64_sm3ss1qv4si): Ditto.
25690 (aarch64_sm3tt1aq4si): Ditto.
25691 (aarch64_sm3tt1bq4si): Ditto.
25692 (aarch64_sm3tt2aq4si): Ditto.
25693 (aarch64_sm3tt2bq4si): Ditto.
25694 (aarch64_sm3partw1qv4si): Ditto.
25695 (aarch64_sm3partw2qv4si): Ditto.
25696 (aarch64_sm4eqv4si): Ditto.
25697 (aarch64_sm4ekeyqv4si): Ditto.
25698 * config/aarch64/aarch64-simd.md:
25699 (aarch64_sm3ss1qv4si): Ditto.
25700 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
25701 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
25702 (aarch64_sm4eqv4si): Ditto.
25703 (aarch64_sm4ekeyqv4si): Ditto.
25704 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
25705 (sm3part_op): Ditto.
25706 (CRYPTO_SM3TT): Ditto.
25707 (CRYPTO_SM3PART): Ditto.
25708 (UNSPEC_SM3SS1): New unspec.
25709 (UNSPEC_SM3TT1A): Ditto.
25710 (UNSPEC_SM3TT1B): Ditto.
25711 (UNSPEC_SM3TT2A): Ditto.
25712 (UNSPEC_SM3TT2B): Ditto.
25713 (UNSPEC_SM3PARTW1): Ditto.
25714 (UNSPEC_SM3PARTW2): Ditto.
25715 (UNSPEC_SM4E): Ditto.
25716 (UNSPEC_SM4EKEY): Ditto.
25717 * config/aarch64/constraints.md (Ui2): New constraint.
25718 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
25719 * config/arm/types.md (crypto_sm3): New type attribute.
25720 (crypto_sm4): Ditto.
25721 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
25722 (vsm3tt1aq_u32): Ditto.
25723 (vsm3tt1bq_u32): Ditto.
25724 (vsm3tt2aq_u32): Ditto.
25725 (vsm3tt2bq_u32): Ditto.
25726 (vsm3partw1q_u32): Ditto.
25727 (vsm3partw2q_u32): Ditto.
25728 (vsm4eq_u32): Ditto.
25729 (vsm4ekeyq_u32): Ditto.
25730 (doc/invoke.texi): Document new sm4 option.
25732 2018-01-10 Michael Collison <michael.collison@arm.com>
25734 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
25735 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
25736 (AARCH64_FL_FOR_ARCH8_4): New.
25737 (AARCH64_FL_V8_4): New flag.
25738 (doc/invoke.texi): Document new armv8.4-a option.
25740 2018-01-10 Michael Collison <michael.collison@arm.com>
25742 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
25743 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
25744 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
25745 * config/aarch64/aarch64-option-extension.def: Add
25746 AARCH64_OPT_EXTENSION of 'sha2'.
25747 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
25748 (crypto): Disable sha2 and aes if crypto disabled.
25749 (crypto): Enable aes and sha2 if enabled.
25750 (simd): Disable sha2 and aes if simd disabled.
25751 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
25753 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
25754 (TARGET_SHA2): New feature flag for sha2.
25755 (TARGET_AES): New feature flag for aes.
25756 * config/aarch64/aarch64-simd.md:
25757 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
25758 conditional on TARGET_AES.
25759 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
25760 (aarch64_crypto_sha1hsi): Make pattern conditional
25762 (aarch64_crypto_sha1hv4si): Ditto.
25763 (aarch64_be_crypto_sha1hv4si): Ditto.
25764 (aarch64_crypto_sha1su1v4si): Ditto.
25765 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
25766 (aarch64_crypto_sha1su0v4si): Ditto.
25767 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
25768 (aarch64_crypto_sha256su0v4si): Ditto.
25769 (aarch64_crypto_sha256su1v4si): Ditto.
25770 (doc/invoke.texi): Document new aes and sha2 options.
25772 2018-01-10 Martin Sebor <msebor@redhat.com>
25774 PR tree-optimization/83781
25775 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
25778 2018-01-11 Martin Sebor <msebor@gmail.com>
25779 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
25781 PR tree-optimization/83501
25782 PR tree-optimization/81703
25784 * tree-ssa-strlen.c (get_string_cst): Rename...
25785 (get_string_len): ...to this. Handle global constants.
25786 (handle_char_store): Adjust.
25788 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
25789 Jim Wilson <jimw@sifive.com>
25791 * config/riscv/riscv-protos.h (riscv_output_return): New.
25792 * config/riscv/riscv.c (struct machine_function): New naked_p field.
25793 (riscv_attribute_table, riscv_output_return),
25794 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
25795 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
25796 (riscv_compute_frame_info): Only compute frame->mask if not a naked
25798 (riscv_expand_prologue): Add early return for naked function.
25799 (riscv_expand_epilogue): Likewise.
25800 (riscv_function_ok_for_sibcall): Return false for naked function.
25801 (riscv_set_current_function): New.
25802 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
25803 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
25804 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
25805 * doc/extend.texi (RISC-V Function Attributes): New.
25807 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
25809 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
25810 check for 128-bit long double before checking TCmode.
25811 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
25812 128-bit long doubles before checking TFmode or TCmode.
25813 (FLOAT128_IBM_P): Likewise.
25815 2018-01-10 Martin Sebor <msebor@redhat.com>
25817 PR tree-optimization/83671
25818 * builtins.c (c_strlen): Unconditionally return zero for the empty
25820 Use -Warray-bounds for warnings.
25821 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
25822 for non-constant array indices with COMPONENT_REF, arrays of
25823 arrays, and pointers to arrays.
25824 (gimple_fold_builtin_strlen): Determine and set length range for
25825 non-constant character arrays.
25827 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
25829 PR middle-end/81897
25830 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
25833 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
25835 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
25837 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
25840 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
25841 VECTOR_MEM_ALTIVEC_OR_VSX_P.
25842 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
25843 indexed_or_indirect_operand predicate.
25844 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
25845 (*vsx_le_perm_load_v8hi): Likewise.
25846 (*vsx_le_perm_load_v16qi): Likewise.
25847 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
25848 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
25849 (*vsx_le_perm_store_v8hi): Likewise.
25850 (*vsx_le_perm_store_v16qi): Likewise.
25851 (eight unnamed splitters): Likewise.
25853 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
25855 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
25856 * config/rs6000/emmintrin.h: Likewise.
25857 * config/rs6000/mmintrin.h: Likewise.
25858 * config/rs6000/xmmintrin.h: Likewise.
25860 2018-01-10 David Malcolm <dmalcolm@redhat.com>
25863 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
25865 * tree.c (tree_nop_conversion): Return true for location wrapper
25867 (maybe_wrap_with_location): New function.
25868 (selftest::check_strip_nops): New function.
25869 (selftest::test_location_wrappers): New function.
25870 (selftest::tree_c_tests): Call it.
25871 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
25872 (maybe_wrap_with_location): New decl.
25873 (EXPR_LOCATION_WRAPPER_P): New macro.
25874 (location_wrapper_p): New inline function.
25875 (tree_strip_any_location_wrapper): New inline function.
25877 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
25880 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
25881 stack_realign_offset for the largest alignment of stack slot
25883 (ix86_find_max_used_stack_alignment): New function.
25884 (ix86_finalize_stack_frame_flags): Use it. Set
25885 max_used_stack_alignment if we don't realign stack.
25886 * config/i386/i386.h (machine_function): Add
25887 max_used_stack_alignment.
25889 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
25891 * config/arm/arm.opt (-mbranch-cost): New option.
25892 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
25895 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
25898 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
25899 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
25901 2018-01-10 Richard Biener <rguenther@suse.de>
25904 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
25905 early out so it also covers the case where we have a non-NULL
25908 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
25910 PR tree-optimization/83753
25911 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
25912 for non-strided grouped accesses if the number of elements is 1.
25914 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
25917 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
25918 * i386.h (TARGET_USE_GATHER): Define.
25919 * x86-tune.def (X86_TUNE_USE_GATHER): New.
25921 2018-01-10 Martin Liska <mliska@suse.cz>
25924 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
25925 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
25927 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
25928 CLEANUP_NO_PARTITIONING is not set.
25930 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
25932 * doc/rtl.texi: Remove documentation of (const ...) wrappers
25933 for vectors, as a partial revert of r254296.
25934 * rtl.h (const_vec_p): Delete.
25935 (const_vec_duplicate_p): Don't test for vector CONSTs.
25936 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
25937 * expmed.c (make_tree): Likewise.
25940 * common.md (E, F): Use CONSTANT_P instead of checking for
25942 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
25943 checking for CONST_VECTOR.
25945 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
25947 PR middle-end/83575
25948 * predict.c (force_edge_cold): Handle in more sane way edges
25949 with no prediction.
25951 2018-01-09 Carl Love <cel@us.ibm.com>
25953 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
25955 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
25956 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
25957 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
25958 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
25959 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
25960 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
25961 * config/rs6000/rs6000-protos.h: Add extern defition for
25962 rs6000_generate_float2_double_code.
25963 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
25965 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
25966 (float2_v2df): Add define_expand.
25968 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
25971 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
25972 op_mode in the force_to_mode call.
25974 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
25976 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
25977 instead of checking each element individually.
25978 (aarch64_evpc_uzp): Likewise.
25979 (aarch64_evpc_zip): Likewise.
25980 (aarch64_evpc_ext): Likewise.
25981 (aarch64_evpc_rev): Likewise.
25982 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
25983 instead of checking each element individually. Return true without
25985 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
25986 whether all selected elements come from the same input, instead of
25987 checking each element individually. Remove calls to gen_rtx_REG,
25988 start_sequence and end_sequence and instead assert that no rtl is
25991 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
25993 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
25994 order of HIGH and CONST checks.
25996 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
25998 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
25999 if the destination isn't an SSA_NAME.
26001 2018-01-09 Richard Biener <rguenther@suse.de>
26003 PR tree-optimization/83668
26004 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
26006 (canonicalize_loop_form): ... here, renamed from ...
26007 (canonicalize_loop_closed_ssa_form): ... this and amended to
26008 swap successor edges for loop exit blocks to make us use
26009 the RPO order we need for initial schedule generation.
26011 2018-01-09 Joseph Myers <joseph@codesourcery.com>
26013 PR tree-optimization/64811
26014 * match.pd: When optimizing comparisons with Inf, avoid
26015 introducing or losing exceptions from comparisons with NaN.
26017 2018-01-09 Martin Liska <mliska@suse.cz>
26020 * asan.c (shadow_mem_size): Add gcc_assert.
26022 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
26024 Don't save registers in main().
26027 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
26028 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
26029 * config/avr/avr.c (avr_set_current_function): Don't error if
26030 naked, OS_task or OS_main are specified at the same time.
26031 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
26033 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
26035 * common/config/avr/avr-common.c (avr_option_optimization_table):
26036 Switch on -mmain-is-OS_task for optimizing compilations.
26038 2018-01-09 Richard Biener <rguenther@suse.de>
26040 PR tree-optimization/83572
26041 * graphite.c: Include cfganal.h.
26042 (graphite_transform_loops): Connect infinite loops to exit
26043 and remove fake edges at the end.
26045 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
26047 * ipa-inline.c (edge_badness): Revert accidental checkin.
26049 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
26052 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
26053 symbols; not inline clones.
26055 2018-01-09 Jakub Jelinek <jakub@redhat.com>
26058 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
26059 hard registers. Formatting fixes.
26061 PR preprocessor/83722
26062 * gcc.c (try_generate_repro): Pass
26063 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
26064 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
26067 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
26068 Kito Cheng <kito.cheng@gmail.com>
26070 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
26071 (riscv_leaf_function_p): Delete.
26072 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
26074 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
26076 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
26078 (do_ifelse): New function.
26079 (do_isel): New function.
26080 (do_sub3): New function.
26081 (do_add3): New function.
26082 (do_load_mask_compare): New function.
26083 (do_overlap_load_compare): New function.
26084 (expand_compare_loop): New function.
26085 (expand_block_compare): Call expand_compare_loop() when appropriate.
26086 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
26087 option description.
26088 (-mblock-compare-inline-loop-limit): New option.
26090 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
26093 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
26094 Reverse order of second and third operands in first alternative.
26095 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
26096 of first and second elements in UNSPEC_VPERMR vector.
26097 (altivec_expand_vec_perm_le): Likewise.
26099 2018-01-08 Jeff Law <law@redhat.com>
26101 PR rtl-optimizatin/81308
26102 * tree-switch-conversion.c (cfg_altered): New file scoped static.
26103 (process_switch): If group_case_labels makes a change, then set
26105 (pass_convert_switch::execute): If a switch is converted, then
26106 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
26108 PR rtl-optimization/81308
26109 * recog.c (split_all_insns): Conditionally cleanup the CFG after
26112 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
26114 PR target/83663 - Revert r255946
26115 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
26116 generation for cases where splatting a value is not useful.
26117 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
26118 across a vec_duplicate and a paradoxical subreg forming a vector
26119 mode to a vec_concat.
26121 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
26123 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
26124 -march=armv8.3-a variants.
26125 * config/arm/t-multilib: Likewise.
26126 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
26128 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
26130 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
26132 (cceq_ior_compare_complement): Give it a name so I can use it, and
26133 change boolean_or_operator predicate to boolean_operator so it can
26134 be used to generate a crand.
26135 (eqne): New code iterator.
26136 (bd/bd_neg): New code_attrs.
26137 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
26138 a single define_insn.
26139 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
26140 decrement (bdnzt/bdnzf/bdzt/bdzf).
26141 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
26142 with the new names of the branch decrement patterns, and added the
26143 names of the branch decrement conditional patterns.
26145 2018-01-08 Richard Biener <rguenther@suse.de>
26147 PR tree-optimization/83563
26148 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
26151 2018-01-08 Richard Biener <rguenther@suse.de>
26153 PR middle-end/83713
26154 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
26156 2018-01-08 Richard Biener <rguenther@suse.de>
26158 PR tree-optimization/83685
26159 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
26160 references to abnormals.
26162 2018-01-08 Richard Biener <rguenther@suse.de>
26165 * dwarf2out.c (output_indirect_strings): Handle empty
26166 skeleton_debug_str_hash.
26167 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
26169 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
26171 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
26172 (emit_store_direct): Likewise.
26173 (arc_trampoline_adjust_address): Likewise.
26174 (arc_asm_trampoline_template): New function.
26175 (arc_initialize_trampoline): Use asm_trampoline_template.
26176 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
26177 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
26178 * config/arc/arc.md (flush_icache): Delete pattern.
26180 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
26182 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
26183 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
26186 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
26189 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
26190 by not USED_FOR_TARGET.
26191 (make_pass_resolve_sw_modes): Likewise.
26193 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
26195 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
26198 2018-01-08 Richard Biener <rguenther@suse.de>
26200 PR middle-end/83580
26201 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
26203 2018-01-08 Richard Biener <rguenther@suse.de>
26205 PR middle-end/83517
26206 * match.pd ((t * 2) / 2) -> t): Add missing :c.
26208 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
26210 PR middle-end/81897
26211 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
26212 basic blocks with a small number of successors.
26213 (convert_control_dep_chain_into_preds): Improve handling of
26215 (dump_predicates): Split apart into...
26216 (dump_pred_chain): ...here...
26217 (dump_pred_info): ...and here.
26218 (can_one_predicate_be_invalidated_p): Add debugging printfs.
26219 (can_chain_union_be_invalidated_p): Improve check for invalidation
26221 (uninit_uses_cannot_happen): Avoid unnecessary if
26222 convert_control_dep_chain_into_preds yielded nothing.
26224 2018-01-06 Martin Sebor <msebor@redhat.com>
26226 PR tree-optimization/83640
26227 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
26228 subtracting negative offset from size.
26229 (builtin_access::overlap): Adjust offset bounds of the access to fall
26230 within the size of the object if possible.
26232 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
26234 PR rtl-optimization/83699
26235 * expmed.c (extract_bit_field_1): Restrict the vector usage of
26236 extract_bit_field_as_subreg to cases in which the extracted
26237 value is also a vector.
26239 * lra-constraints.c (process_alt_operands): Test for the equivalence
26240 substitutions when detecting a possible reload cycle.
26242 2018-01-06 Jakub Jelinek <jakub@redhat.com>
26245 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
26246 by default if flag_selective_schedling{,2}. Formatting fixes.
26248 PR rtl-optimization/83682
26249 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
26250 if it has non-VECTOR_MODE element mode.
26251 (vec_duplicate_p): Likewise.
26253 PR middle-end/83694
26254 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
26255 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
26257 2018-01-05 Jakub Jelinek <jakub@redhat.com>
26260 * config/i386/i386-builtin.def
26261 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
26262 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
26263 Require also OPTION_MASK_ISA_AVX512F in addition to
26264 OPTION_MASK_ISA_GFNI.
26265 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
26266 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
26267 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
26268 to OPTION_MASK_ISA_GFNI.
26269 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
26270 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
26271 OPTION_MASK_ISA_AVX512BW.
26272 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
26273 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
26274 addition to OPTION_MASK_ISA_GFNI.
26275 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
26276 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
26277 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
26278 to OPTION_MASK_ISA_GFNI.
26279 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
26280 a requirement for all ISAs rather than any of them with a few
26282 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
26284 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
26285 bitmasks to be enabled with 3 exceptions, instead of requiring any
26286 enabled ISA with lots of exceptions.
26287 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
26288 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
26289 Change avx512bw in isa attribute to avx512f.
26290 * config/i386/sgxintrin.h: Add license boilerplate.
26291 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
26292 to __AVX512F__ and __AVX512VL to __AVX512VL__.
26293 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
26294 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
26296 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
26297 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
26298 temporarily sse2 rather than sse if not enabled already.
26301 * config/i386/sse.md (VI248_VLBW): Rename to ...
26302 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
26303 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
26304 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
26305 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
26306 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
26307 mode iterator instead of VI248_VLBW.
26309 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
26311 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
26312 (record_modified): Skip clobbers; add debug output.
26313 (param_change_prob): Use sreal frequencies.
26315 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
26317 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
26318 punt for user-aligned variables.
26320 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
26322 * tree-chrec.c (chrec_contains_symbols): Return true for
26325 2018-01-05 Sudakshina Das <sudi.das@arm.com>
26328 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
26329 of (x|y) == x for BICS pattern.
26331 2018-01-05 Jakub Jelinek <jakub@redhat.com>
26333 PR tree-optimization/83605
26334 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
26335 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
26338 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
26340 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
26341 * config/epiphany/rtems.h: New file.
26343 2018-01-04 Jakub Jelinek <jakub@redhat.com>
26344 Uros Bizjak <ubizjak@gmail.com>
26347 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
26348 QIreg_operand instead of register_operand predicate.
26349 * config/i386/i386.c (ix86_rop_should_change_byte_p,
26350 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
26351 comments instead of -fmitigate[-_]rop.
26353 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
26356 * cgraphunit.c (symbol_table::compile): Switch to text_section
26357 before calling assembly_start debug hook.
26358 * run-rtl-passes.c (run_rtl_passes): Likewise.
26361 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
26363 * tree-vrp.c (extract_range_from_binary_expr_1): Check
26364 range_int_cst_p rather than !symbolic_range_p before calling
26365 extract_range_from_multiplicative_op_1.
26367 2018-01-04 Jeff Law <law@redhat.com>
26369 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
26370 redundant test in assertion.
26372 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
26374 * doc/rtl.texi: Document machine_mode wrapper classes.
26376 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
26378 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
26379 using tree_to_uhwi.
26381 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
26383 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
26384 the VEC_PERM_EXPR fold to fail.
26386 2018-01-04 Jakub Jelinek <jakub@redhat.com>
26389 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
26390 to switched_sections.
26392 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
26395 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
26396 test for d.testing.
26398 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
26401 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
26402 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
26404 2018-01-04 Jakub Jelinek <jakub@redhat.com>
26407 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
26408 is BLKmode and bitpos not zero or mode change is needed.
26410 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
26413 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
26416 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
26419 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
26420 instead of MULT rtx. Update all corresponding splitters.
26421 (*saddl_se): Ditto.
26422 (*ssub<modesuffix>): Ditto.
26423 (*ssubl_se): Ditto.
26424 (*cmp_sadd_di): Update split patterns.
26425 (*cmp_sadd_si): Ditto.
26426 (*cmp_sadd_sidi): Ditto.
26427 (*cmp_ssub_di): Ditto.
26428 (*cmp_ssub_si): Ditto.
26429 (*cmp_ssub_sidi): Ditto.
26430 * config/alpha/predicates.md (const23_operand): New predicate.
26431 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
26432 Look for ASHIFT, not MULT inner operand.
26433 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
26435 2018-01-04 Martin Liska <mliska@suse.cz>
26437 PR gcov-profile/83669
26438 * gcov.c (output_intermediate_file): Add version to intermediate
26440 * doc/gcov.texi: Document new field 'version' in intermediate
26441 file format. Fix location of '-k' option of gcov command.
26443 2018-01-04 Martin Liska <mliska@suse.cz>
26446 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
26448 2018-01-04 Jakub Jelinek <jakub@redhat.com>
26450 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
26452 2018-01-03 Martin Sebor <msebor@redhat.com>
26454 PR tree-optimization/83655
26455 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
26456 checking calls with invalid arguments.
26458 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26460 * tree-vect-stmts.c (vect_get_store_rhs): New function.
26461 (vectorizable_mask_load_store): Delete.
26462 (vectorizable_call): Return false for masked loads and stores.
26463 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
26464 instead of gimple_assign_rhs1.
26465 (vectorizable_load): Handle IFN_MASK_LOAD.
26466 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
26468 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26470 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
26472 (vectorizable_mask_load_store): ...here.
26473 (vectorizable_load): ...and here.
26475 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26477 * tree-vect-stmts.c (vect_build_all_ones_mask)
26478 (vect_build_zero_merge_argument): New functions, split out from...
26479 (vectorizable_load): ...here.
26481 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26483 * tree-vect-stmts.c (vect_check_store_rhs): New function,
26485 (vectorizable_mask_load_store): ...here.
26486 (vectorizable_store): ...and here.
26488 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26490 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
26492 (vectorizable_mask_load_store): ...here.
26494 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26496 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
26497 (vect_model_store_cost): Take a vec_load_store_type instead of a
26499 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
26500 (vect_model_store_cost): Take a vec_load_store_type instead of a
26502 (vectorizable_mask_load_store): Update accordingly.
26503 (vectorizable_store): Likewise.
26504 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
26506 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26508 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
26509 IFN_MASK_LOAD calls here rather than...
26510 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
26512 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26513 Alan Hayward <alan.hayward@arm.com>
26514 David Sherwood <david.sherwood@arm.com>
26516 * expmed.c (extract_bit_field_1): For vector extracts,
26517 fall back to extract_bit_field_as_subreg if vec_extract
26520 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26521 Alan Hayward <alan.hayward@arm.com>
26522 David Sherwood <david.sherwood@arm.com>
26524 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
26525 they are variable or constant sized.
26526 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
26527 slots for constant-sized data.
26529 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26530 Alan Hayward <alan.hayward@arm.com>
26531 David Sherwood <david.sherwood@arm.com>
26533 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
26534 handling COND_EXPRs with boolean comparisons, try to find a better
26535 basis for the mask type than the boolean itself.
26537 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26539 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
26540 is calculated and how it can be overridden.
26541 * genmodes.c (max_bitsize_mode_any_mode): New variable.
26542 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
26544 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
26547 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26548 Alan Hayward <alan.hayward@arm.com>
26549 David Sherwood <david.sherwood@arm.com>
26551 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
26552 Remove the mode argument.
26553 (aarch64_simd_valid_immediate): Remove the mode and inverse
26555 * config/aarch64/iterators.md (bitsize): New iterator.
26556 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
26557 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
26558 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
26559 aarch64_simd_valid_immediate.
26560 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
26561 (aarch64_reg_or_bic_imm): Likewise.
26562 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
26563 with an insn_type enum and msl with a modifier_type enum.
26564 Replace element_width with a scalar_mode. Change the shift
26565 to unsigned int. Add constructors for scalar_float_mode and
26566 scalar_int_mode elements.
26567 (aarch64_vect_float_const_representable_p): Delete.
26568 (aarch64_can_const_movi_rtx_p)
26569 (aarch64_simd_scalar_immediate_valid_for_move)
26570 (aarch64_simd_make_constant): Update call to
26571 aarch64_simd_valid_immediate.
26572 (aarch64_advsimd_valid_immediate_hs): New function.
26573 (aarch64_advsimd_valid_immediate): Likewise.
26574 (aarch64_simd_valid_immediate): Remove mode and inverse
26575 arguments. Rewrite to use the above. Use const_vec_duplicate_p
26576 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
26577 and aarch64_float_const_representable_p on the result.
26578 (aarch64_output_simd_mov_immediate): Remove mode argument.
26579 Update call to aarch64_simd_valid_immediate and use of
26580 simd_immediate_info.
26581 (aarch64_output_scalar_simd_mov_immediate): Update call
26584 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26585 Alan Hayward <alan.hayward@arm.com>
26586 David Sherwood <david.sherwood@arm.com>
26588 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
26589 (mode_nunits): Likewise CONST_MODE_NUNITS.
26590 * machmode.def (ADJUST_NUNITS): Document.
26591 * genmodes.c (mode_data::need_nunits_adj): New field.
26592 (blank_mode): Update accordingly.
26593 (adj_nunits): New variable.
26594 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
26596 (emit_mode_size_inline): Set need_bytesize_adj for all modes
26597 listed in adj_nunits.
26598 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
26599 listed in adj_nunits. Don't emit case statements for such modes.
26600 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
26601 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
26602 nothing if adj_nunits is nonnull.
26603 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
26604 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
26605 (emit_mode_fbit): Update use of print_maybe_const_decl.
26606 (emit_move_size): Likewise. Treat the array as non-const
26608 (emit_mode_adjustments): Handle adj_nunits.
26610 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26612 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
26613 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
26614 (VECTOR_MODES): Use it.
26615 (make_vector_modes): Take the prefix as an argument.
26617 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26618 Alan Hayward <alan.hayward@arm.com>
26619 David Sherwood <david.sherwood@arm.com>
26621 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
26622 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
26623 for MODE_VECTOR_BOOL.
26624 * machmode.def (VECTOR_BOOL_MODE): Document.
26625 * genmodes.c (VECTOR_BOOL_MODE): New macro.
26626 (make_vector_bool_mode): New function.
26627 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
26629 * lto-streamer-in.c (lto_input_mode_table): Likewise.
26630 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
26632 * stor-layout.c (int_mode_for_mode): Likewise.
26633 * tree.c (build_vector_type_for_mode): Likewise.
26634 * varasm.c (output_constant_pool_2): Likewise.
26635 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
26636 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
26637 for MODE_VECTOR_BOOL.
26638 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
26639 of mode class checks.
26640 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
26641 instead of a list of mode class checks.
26642 (expand_vector_scalar_condition): Likewise.
26643 (type_for_widest_vector_mode): Handle BImode as an inner mode.
26645 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26646 Alan Hayward <alan.hayward@arm.com>
26647 David Sherwood <david.sherwood@arm.com>
26649 * machmode.h (mode_size): Change from unsigned short to
26651 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
26652 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
26653 or if measurement_type is not polynomial.
26654 (fixed_size_mode::includes_p): Check for constant-sized modes.
26655 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
26656 return a poly_uint16 rather than an unsigned short.
26657 (emit_mode_size): Change the type of mode_size from unsigned short
26658 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
26659 (emit_mode_adjustments): Cope with polynomial vector sizes.
26660 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
26662 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
26664 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
26665 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
26666 * caller-save.c (setup_save_areas): Likewise.
26667 (replace_reg_with_saved_mem): Likewise.
26668 * calls.c (emit_library_call_value_1): Likewise.
26669 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
26670 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
26671 (gen_lowpart_for_combine): Likewise.
26672 * convert.c (convert_to_integer_1): Likewise.
26673 * cse.c (equiv_constant, cse_insn): Likewise.
26674 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
26675 (cselib_subst_to_values): Likewise.
26676 * dce.c (word_dce_process_block): Likewise.
26677 * df-problems.c (df_word_lr_mark_ref): Likewise.
26678 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
26679 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
26680 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
26681 (rtl_for_decl_location): Likewise.
26682 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
26683 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
26684 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
26685 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
26686 (expand_expr_real_1): Likewise.
26687 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
26688 (pad_below): Likewise.
26689 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
26690 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
26691 * ira.c (get_subreg_tracking_sizes): Likewise.
26692 * ira-build.c (ira_create_allocno_objects): Likewise.
26693 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
26694 (ira_sort_regnos_for_alter_reg): Likewise.
26695 * ira-costs.c (record_operand_costs): Likewise.
26696 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
26697 (resolve_simple_move): Likewise.
26698 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
26699 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
26700 (lra_constraints): Likewise.
26701 (CONST_POOL_OK_P): Reject variable-sized modes.
26702 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
26703 (add_pseudo_to_slot, lra_spill): Likewise.
26704 * omp-low.c (omp_clause_aligned_alignment): Likewise.
26705 * optabs-query.c (get_best_extraction_insn): Likewise.
26706 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
26707 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
26708 (expand_mult_highpart, valid_multiword_target_p): Likewise.
26709 * recog.c (offsettable_address_addr_space_p): Likewise.
26710 * regcprop.c (maybe_mode_change): Likewise.
26711 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
26712 * regrename.c (build_def_use): Likewise.
26713 * regstat.c (dump_reg_info): Likewise.
26714 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
26715 (find_reloads, find_reloads_subreg_address): Likewise.
26716 * reload1.c (eliminate_regs_1): Likewise.
26717 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
26718 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
26719 (simplify_binary_operation_1, simplify_subreg): Likewise.
26720 * targhooks.c (default_function_arg_padding): Likewise.
26721 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
26722 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
26723 (verify_gimple_assign_ternary): Likewise.
26724 * tree-inline.c (estimate_move_cost): Likewise.
26725 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
26726 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
26727 (get_address_cost_ainc): Likewise.
26728 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
26729 (vect_supportable_dr_alignment): Likewise.
26730 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
26731 (vectorizable_reduction): Likewise.
26732 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
26733 (vectorizable_operation, vectorizable_load): Likewise.
26734 * tree.c (build_same_sized_truth_vector_type): Likewise.
26735 * valtrack.c (cleanup_auto_inc_dec): Likewise.
26736 * var-tracking.c (emit_note_insn_var_location): Likewise.
26737 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
26738 (ADDR_VEC_ALIGN): Likewise.
26740 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26741 Alan Hayward <alan.hayward@arm.com>
26742 David Sherwood <david.sherwood@arm.com>
26744 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
26746 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
26747 or if measurement_type is polynomial.
26748 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
26749 * combine.c (make_extraction): Likewise.
26750 * dse.c (find_shift_sequence): Likewise.
26751 * dwarf2out.c (mem_loc_descriptor): Likewise.
26752 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
26753 (extract_bit_field, extract_low_bits): Likewise.
26754 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
26755 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
26756 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
26757 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
26758 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
26759 * reload.c (find_reloads): Likewise.
26760 * reload1.c (alter_reg): Likewise.
26761 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
26762 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
26763 * tree-if-conv.c (predicate_mem_writes): Likewise.
26764 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
26765 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
26766 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
26767 * valtrack.c (dead_debug_insert_temp): Likewise.
26768 * varasm.c (mergeable_constant_section): Likewise.
26769 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
26771 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26772 Alan Hayward <alan.hayward@arm.com>
26773 David Sherwood <david.sherwood@arm.com>
26775 * expr.c (expand_assignment): Cope with polynomial mode sizes
26776 when assigning to a CONCAT.
26778 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26779 Alan Hayward <alan.hayward@arm.com>
26780 David Sherwood <david.sherwood@arm.com>
26782 * machmode.h (mode_precision): Change from unsigned short to
26784 (mode_to_precision): Return a poly_uint16 rather than an unsigned
26786 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
26787 or if measurement_type is not polynomial.
26788 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
26789 in which the mode is already known to be a scalar_int_mode.
26790 * genmodes.c (emit_mode_precision): Change the type of mode_precision
26791 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
26793 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
26794 for GET_MODE_PRECISION.
26795 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
26796 for GET_MODE_PRECISION.
26797 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
26799 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
26800 (expand_field_assignment, make_extraction): Likewise.
26801 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
26802 (get_last_value): Likewise.
26803 * convert.c (convert_to_integer_1): Likewise.
26804 * cse.c (cse_insn): Likewise.
26805 * expr.c (expand_expr_real_1): Likewise.
26806 * lra-constraints.c (simplify_operand_subreg): Likewise.
26807 * optabs-query.c (can_atomic_load_p): Likewise.
26808 * optabs.c (expand_atomic_load): Likewise.
26809 (expand_atomic_store): Likewise.
26810 * ree.c (combine_reaching_defs): Likewise.
26811 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
26812 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
26813 * tree.h (type_has_mode_precision_p): Likewise.
26814 * ubsan.c (instrument_si_overflow): Likewise.
26816 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26817 Alan Hayward <alan.hayward@arm.com>
26818 David Sherwood <david.sherwood@arm.com>
26820 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
26821 polynomial numbers of units.
26822 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
26823 (valid_vector_subparts_p): New function.
26824 (build_vector_type): Remove temporary shim and take the number
26825 of units as a poly_uint64 rather than an int.
26826 (build_opaque_vector_type): Take the number of units as a
26827 poly_uint64 rather than an int.
26828 * tree.c (build_vector_from_ctor): Handle polynomial
26829 TYPE_VECTOR_SUBPARTS.
26830 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
26831 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
26832 (build_vector_from_val): If the number of units is variable,
26833 use build_vec_duplicate_cst for constant operands and
26834 VEC_DUPLICATE_EXPR otherwise.
26835 (make_vector_type): Remove temporary is_constant ().
26836 (build_vector_type, build_opaque_vector_type): Take the number of
26837 units as a poly_uint64 rather than an int.
26838 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
26840 * cfgexpand.c (expand_debug_expr): Likewise.
26841 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
26842 (store_constructor, expand_expr_real_1): Likewise.
26843 (const_scalar_mask_from_tree): Likewise.
26844 * fold-const-call.c (fold_const_reduction): Likewise.
26845 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
26846 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
26847 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
26848 (fold_relational_const): Likewise.
26849 (native_interpret_vector): Likewise. Change the size from an
26850 int to an unsigned int.
26851 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
26852 TYPE_VECTOR_SUBPARTS.
26853 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
26854 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
26855 duplicating a non-constant operand into a variable-length vector.
26856 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
26857 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
26858 * ipa-icf.c (sem_variable::equals): Likewise.
26859 * match.pd: Likewise.
26860 * omp-simd-clone.c (simd_clone_subparts): Likewise.
26861 * print-tree.c (print_node): Likewise.
26862 * stor-layout.c (layout_type): Likewise.
26863 * targhooks.c (default_builtin_vectorization_cost): Likewise.
26864 * tree-cfg.c (verify_gimple_comparison): Likewise.
26865 (verify_gimple_assign_binary): Likewise.
26866 (verify_gimple_assign_ternary): Likewise.
26867 (verify_gimple_assign_single): Likewise.
26868 * tree-pretty-print.c (dump_generic_node): Likewise.
26869 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
26870 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
26871 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
26872 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
26873 (vect_shift_permute_load_chain): Likewise.
26874 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
26875 (expand_vector_condition, optimize_vector_constructor): Likewise.
26876 (lower_vec_perm, get_compute_type): Likewise.
26877 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
26878 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
26879 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
26880 (vect_recog_mask_conversion_pattern): Likewise.
26881 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
26882 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
26883 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
26884 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
26885 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
26886 (vectorizable_shift, vectorizable_operation, vectorizable_store)
26887 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
26888 (supportable_widening_operation): Likewise.
26889 (supportable_narrowing_operation): Likewise.
26890 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
26892 * varasm.c (output_constant): Likewise.
26894 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26895 Alan Hayward <alan.hayward@arm.com>
26896 David Sherwood <david.sherwood@arm.com>
26898 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
26899 so that both the length == 3 and length != 3 cases set up their
26900 own permute vectors. Add comments explaining why we know the
26901 number of elements is constant.
26902 (vect_permute_load_chain): Likewise.
26904 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26905 Alan Hayward <alan.hayward@arm.com>
26906 David Sherwood <david.sherwood@arm.com>
26908 * machmode.h (mode_nunits): Change from unsigned char to
26910 (ONLY_FIXED_SIZE_MODES): New macro.
26911 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
26912 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
26913 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
26915 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
26916 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
26917 or if measurement_type is not polynomial.
26918 * genmodes.c (ZERO_COEFFS): New macro.
26919 (emit_mode_nunits_inline): Make mode_nunits_inline return a
26921 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
26922 Use ZERO_COEFFS when emitting initializers.
26923 * data-streamer.h (bp_pack_poly_value): New function.
26924 (bp_unpack_poly_value): Likewise.
26925 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
26926 for GET_MODE_NUNITS.
26927 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
26928 for GET_MODE_NUNITS.
26929 * tree.c (make_vector_type): Remove temporary shim and make
26930 the real function take the number of units as a poly_uint64
26931 rather than an int.
26932 (build_vector_type_for_mode): Handle polynomial nunits.
26933 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
26934 * emit-rtl.c (const_vec_series_p_1): Likewise.
26935 (gen_rtx_CONST_VECTOR): Likewise.
26936 * fold-const.c (test_vec_duplicate_folding): Likewise.
26937 * genrecog.c (validate_pattern): Likewise.
26938 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
26939 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
26940 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
26941 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
26942 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
26943 * rtlanal.c (subreg_get_info): Likewise.
26944 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
26945 (vect_grouped_load_supported): Likewise.
26946 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
26947 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
26948 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
26949 (simplify_const_unary_operation, simplify_binary_operation_1)
26950 (simplify_const_binary_operation, simplify_ternary_operation)
26951 (test_vector_ops_duplicate, test_vector_ops): Likewise.
26952 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
26953 instead of CONST_VECTOR_NUNITS.
26954 * varasm.c (output_constant_pool_2): Likewise.
26955 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
26956 explicit-encoded elements in the XVEC for variable-length vectors.
26958 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26960 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
26962 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26963 Alan Hayward <alan.hayward@arm.com>
26964 David Sherwood <david.sherwood@arm.com>
26966 * coretypes.h (fixed_size_mode): Declare.
26967 (fixed_size_mode_pod): New typedef.
26968 * builtins.h (target_builtins::x_apply_args_mode)
26969 (target_builtins::x_apply_result_mode): Change type to
26970 fixed_size_mode_pod.
26971 * builtins.c (apply_args_size, apply_result_size, result_vector)
26972 (expand_builtin_apply_args_1, expand_builtin_apply)
26973 (expand_builtin_return): Update accordingly.
26975 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26977 * cse.c (hash_rtx_cb): Hash only the encoded elements.
26978 * cselib.c (cselib_hash_rtx): Likewise.
26979 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
26980 CONST_VECTOR encoding.
26982 2018-01-03 Jakub Jelinek <jakub@redhat.com>
26983 Jeff Law <law@redhat.com>
26986 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
26987 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
26988 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
26989 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
26992 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
26993 explicitly probe *sp in a noreturn function if there were any callee
26994 register saves or frame pointer is needed.
26996 2018-01-03 Jakub Jelinek <jakub@redhat.com>
26999 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
27000 BLKmode for ternary, binary or unary expressions.
27003 * var-tracking.c (delete_vta_debug_insn): New inline function.
27004 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
27005 insns from get_insns () to NULL instead of each bb separately.
27006 Use delete_vta_debug_insn. No longer static.
27007 (vt_debug_insns_local, variable_tracking_main_1): Adjust
27008 delete_vta_debug_insns callers.
27009 * rtl.h (delete_vta_debug_insns): Declare.
27010 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
27011 instead of variable_tracking_main.
27013 2018-01-03 Martin Sebor <msebor@redhat.com>
27015 PR tree-optimization/83603
27016 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
27017 arguments past the endof the argument list in functions declared
27018 without a prototype.
27019 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
27020 Avoid checking when arguments are null.
27022 2018-01-03 Martin Sebor <msebor@redhat.com>
27025 * doc/extend.texi (attribute const): Fix a typo.
27026 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
27027 issuing -Wsuggest-attribute for void functions.
27029 2018-01-03 Martin Sebor <msebor@redhat.com>
27031 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
27032 offset_int::from instead of wide_int::to_shwi.
27033 (maybe_diag_overlap): Remove assertion.
27034 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
27035 * gimple-ssa-sprintf.c (format_directive): Same.
27036 (parse_directive): Same.
27037 (sprintf_dom_walker::compute_format_length): Same.
27038 (try_substitute_return_value): Same.
27040 2018-01-03 Jeff Law <law@redhat.com>
27042 PR middle-end/83654
27043 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
27044 non-constant residual for zero at runtime and avoid probing in
27045 that case. Reorganize code for trailing problem to mirror handling
27048 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
27050 PR tree-optimization/83501
27051 * tree-ssa-strlen.c (get_string_cst): New.
27052 (handle_char_store): Call get_string_cst.
27054 2018-01-03 Martin Liska <mliska@suse.cz>
27056 PR tree-optimization/83593
27057 * tree-ssa-strlen.c: Include tree-cfg.h.
27058 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
27059 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
27060 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
27062 (strlen_dom_walker::before_dom_children): Call
27063 gimple_purge_dead_eh_edges. Dump tranformation with details
27065 (strlen_dom_walker::before_dom_children): Update call by adding
27066 new argument cleanup_eh.
27067 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
27069 2018-01-03 Martin Liska <mliska@suse.cz>
27072 * cif-code.def (VARIADIC_THUNK): New enum value.
27073 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
27076 2018-01-03 Jan Beulich <jbeulich@suse.com>
27078 * sse.md (mov<mode>_internal): Tighten condition for when to use
27079 vmovdqu<ssescalarsize> for TI and OI modes.
27081 2018-01-03 Jakub Jelinek <jakub@redhat.com>
27083 Update copyright years.
27085 2018-01-03 Martin Liska <mliska@suse.cz>
27088 * ipa-visibility.c (function_and_variable_visibility): Skip
27089 functions with noipa attribure.
27091 2018-01-03 Jakub Jelinek <jakub@redhat.com>
27093 * gcc.c (process_command): Update copyright notice dates.
27094 * gcov-dump.c (print_version): Ditto.
27095 * gcov.c (print_version): Ditto.
27096 * gcov-tool.c (print_version): Ditto.
27097 * gengtype.c (create_file): Ditto.
27098 * doc/cpp.texi: Bump @copying's copyright year.
27099 * doc/cppinternals.texi: Ditto.
27100 * doc/gcc.texi: Ditto.
27101 * doc/gccint.texi: Ditto.
27102 * doc/gcov.texi: Ditto.
27103 * doc/install.texi: Ditto.
27104 * doc/invoke.texi: Ditto.
27106 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27108 * vector-builder.h (vector_builder::m_full_nelts): Change from
27109 unsigned int to poly_uint64.
27110 (vector_builder::full_nelts): Update prototype accordingly.
27111 (vector_builder::new_vector): Likewise.
27112 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
27113 (vector_builder::operator ==): Likewise.
27114 (vector_builder::finalize): Likewise.
27115 * int-vector-builder.h (int_vector_builder::int_vector_builder):
27116 Take the number of elements as a poly_uint64 rather than an
27118 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
27119 from unsigned int to poly_uint64.
27120 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
27121 (vec_perm_indices::new_vector): Likewise.
27122 (vec_perm_indices::length): Likewise.
27123 (vec_perm_indices::nelts_per_input): Likewise.
27124 (vec_perm_indices::input_nelts): Likewise.
27125 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
27126 number of elements per input as a poly_uint64 rather than an
27127 unsigned int. Use the original encoding for variable-length
27128 vectors, rather than clamping each individual element.
27129 For the second and subsequent elements in each pattern,
27130 clamp the step and base before clamping their sum.
27131 (vec_perm_indices::series_p): Handle polynomial element counts.
27132 (vec_perm_indices::all_in_range_p): Likewise.
27133 (vec_perm_indices_to_tree): Likewise.
27134 (vec_perm_indices_to_rtx): Likewise.
27135 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
27136 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
27137 (tree_vector_builder::new_binary_operation): Handle polynomial
27138 element counts. Return false if we need to know the number
27139 of elements at compile time.
27140 * fold-const.c (fold_vec_perm): Punt if the number of elements
27141 isn't known at compile time.
27143 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27145 * vec-perm-indices.h (vec_perm_builder): Change element type
27146 from HOST_WIDE_INT to poly_int64.
27147 (vec_perm_indices::element_type): Update accordingly.
27148 (vec_perm_indices::clamp): Handle polynomial element_types.
27149 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
27150 (vec_perm_indices::all_in_range_p): Likewise.
27151 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
27153 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
27154 polynomial vec_perm_indices element types.
27155 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
27156 * fold-const.c (fold_vec_perm): Likewise.
27157 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
27158 * tree-vect-generic.c (lower_vec_perm): Likewise.
27159 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
27160 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
27161 element type to HOST_WIDE_INT.
27163 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27164 Alan Hayward <alan.hayward@arm.com>
27165 David Sherwood <david.sherwood@arm.com>
27167 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
27168 rather than an int. Use plus_constant.
27169 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
27170 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
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 * calls.c (emit_call_1, expand_call): Change struct_value_size from
27177 a HOST_WIDE_INT to a poly_int64.
27179 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27180 Alan Hayward <alan.hayward@arm.com>
27181 David Sherwood <david.sherwood@arm.com>
27183 * calls.c (load_register_parameters): Cope with polynomial
27184 mode sizes. Require a constant size for BLKmode parameters
27185 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
27186 forces a parameter to be padded at the lsb end in order to
27187 fill a complete number of words, require the parameter size
27188 to be ordered wrt UNITS_PER_WORD.
27190 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27191 Alan Hayward <alan.hayward@arm.com>
27192 David Sherwood <david.sherwood@arm.com>
27194 * reload1.c (spill_stack_slot_width): Change element type
27195 from unsigned int to poly_uint64_pod.
27196 (alter_reg): Treat mode sizes as polynomial.
27198 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27199 Alan Hayward <alan.hayward@arm.com>
27200 David Sherwood <david.sherwood@arm.com>
27202 * reload.c (complex_word_subreg_p): New function.
27203 (reload_inner_reg_of_subreg, push_reload): Use it.
27205 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27206 Alan Hayward <alan.hayward@arm.com>
27207 David Sherwood <david.sherwood@arm.com>
27209 * lra-constraints.c (process_alt_operands): Reject matched
27210 operands whose sizes aren't ordered.
27211 (match_reload): Refer to this check here.
27213 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27214 Alan Hayward <alan.hayward@arm.com>
27215 David Sherwood <david.sherwood@arm.com>
27217 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
27218 that the mode size is in the set {1, 2, 4, 8, 16}.
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 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
27225 Use plus_constant instead of gen_rtx_PLUS.
27227 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27228 Alan Hayward <alan.hayward@arm.com>
27229 David Sherwood <david.sherwood@arm.com>
27231 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
27232 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
27233 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
27234 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
27235 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
27236 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
27237 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
27238 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
27239 * config/i386/i386.c (ix86_push_rounding): ...this new function.
27240 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
27242 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
27243 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
27244 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
27245 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
27246 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
27247 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
27248 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
27249 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
27250 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
27251 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
27253 * expr.c (emit_move_resolve_push): Treat the input and result
27254 of PUSH_ROUNDING as a poly_int64.
27255 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
27256 (emit_push_insn): Likewise.
27257 * lra-eliminations.c (mark_not_eliminable): Likewise.
27258 * recog.c (push_operand): Likewise.
27259 * reload1.c (elimination_effects): Likewise.
27260 * rtlanal.c (nonzero_bits1): Likewise.
27261 * calls.c (store_one_arg): Likewise. Require the padding to be
27262 known at compile time.
27264 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27265 Alan Hayward <alan.hayward@arm.com>
27266 David Sherwood <david.sherwood@arm.com>
27268 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
27269 Use plus_constant instead of gen_rtx_PLUS.
27271 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27272 Alan Hayward <alan.hayward@arm.com>
27273 David Sherwood <david.sherwood@arm.com>
27275 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
27276 rather than an int.
27278 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27279 Alan Hayward <alan.hayward@arm.com>
27280 David Sherwood <david.sherwood@arm.com>
27282 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
27283 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
27284 via stack temporaries. Treat the mode size as polynomial too.
27286 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27287 Alan Hayward <alan.hayward@arm.com>
27288 David Sherwood <david.sherwood@arm.com>
27290 * expr.c (expand_expr_real_2): When handling conversions involving
27291 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
27292 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
27293 as a poly_uint64 too.
27295 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27296 Alan Hayward <alan.hayward@arm.com>
27297 David Sherwood <david.sherwood@arm.com>
27299 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
27301 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27302 Alan Hayward <alan.hayward@arm.com>
27303 David Sherwood <david.sherwood@arm.com>
27305 * combine.c (can_change_dest_mode): Handle polynomial
27306 REGMODE_NATURAL_SIZE.
27307 * expmed.c (store_bit_field_1): Likewise.
27308 * expr.c (store_constructor): Likewise.
27309 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
27310 and polynomial REGMODE_NATURAL_SIZE.
27311 (gen_lowpart_common): Likewise.
27312 * reginfo.c (record_subregs_of_mode): Likewise.
27313 * rtlanal.c (read_modify_subreg_p): Likewise.
27315 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27316 Alan Hayward <alan.hayward@arm.com>
27317 David Sherwood <david.sherwood@arm.com>
27319 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
27320 numbers of elements.
27322 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27323 Alan Hayward <alan.hayward@arm.com>
27324 David Sherwood <david.sherwood@arm.com>
27326 * match.pd: Cope with polynomial numbers of vector elements.
27328 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27329 Alan Hayward <alan.hayward@arm.com>
27330 David Sherwood <david.sherwood@arm.com>
27332 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
27333 in a POINTER_PLUS_EXPR.
27335 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27336 Alan Hayward <alan.hayward@arm.com>
27337 David Sherwood <david.sherwood@arm.com>
27339 * omp-simd-clone.c (simd_clone_subparts): New function.
27340 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
27341 (ipa_simd_modify_function_body): Likewise.
27343 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27344 Alan Hayward <alan.hayward@arm.com>
27345 David Sherwood <david.sherwood@arm.com>
27347 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
27348 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
27349 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
27350 (expand_vector_condition, vector_element): Likewise.
27351 (subparts_gt): New function.
27352 (get_compute_type): Use subparts_gt.
27353 (count_type_subparts): Delete.
27354 (expand_vector_operations_1): Use subparts_gt instead of
27355 count_type_subparts.
27357 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27358 Alan Hayward <alan.hayward@arm.com>
27359 David Sherwood <david.sherwood@arm.com>
27361 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
27362 (vect_compile_time_alias): ...this new function. Do the calculation
27363 on poly_ints rather than trees.
27364 (vect_prune_runtime_alias_test_list): Update call accordingly.
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_build_slp_tree_1): Handle polynomial
27372 (vect_schedule_slp_instance): Likewise.
27374 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27375 Alan Hayward <alan.hayward@arm.com>
27376 David Sherwood <david.sherwood@arm.com>
27378 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
27379 constant and extern definitions for variable-length vectors.
27380 (vect_get_constant_vectors): Note that the number of units
27381 is known to be constant.
27383 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27384 Alan Hayward <alan.hayward@arm.com>
27385 David Sherwood <david.sherwood@arm.com>
27387 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
27388 of units as polynomial. Choose between WIDE and NARROW based
27391 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27392 Alan Hayward <alan.hayward@arm.com>
27393 David Sherwood <david.sherwood@arm.com>
27395 * tree-vect-stmts.c (simd_clone_subparts): New function.
27396 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
27398 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27399 Alan Hayward <alan.hayward@arm.com>
27400 David Sherwood <david.sherwood@arm.com>
27402 * tree-vect-stmts.c (vectorizable_call): Treat the number of
27403 vectors as polynomial. Use build_index_vector for
27404 IFN_GOMP_SIMD_LANE.
27406 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27407 Alan Hayward <alan.hayward@arm.com>
27408 David Sherwood <david.sherwood@arm.com>
27410 * tree-vect-stmts.c (get_load_store_type): Treat the number of
27411 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
27412 for variable-length vectors.
27413 (vectorizable_mask_load_store): Treat the number of units as
27414 polynomial, asserting that it is constant if the condition has
27415 already been enforced.
27416 (vectorizable_store, vectorizable_load): Likewise.
27418 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27419 Alan Hayward <alan.hayward@arm.com>
27420 David Sherwood <david.sherwood@arm.com>
27422 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
27423 of units as polynomial. Punt if we can't tell at compile time
27424 which vector contains the final result.
27426 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27427 Alan Hayward <alan.hayward@arm.com>
27428 David Sherwood <david.sherwood@arm.com>
27430 * tree-vect-loop.c (vectorizable_induction): Treat the number
27431 of units as polynomial. Punt on SLP inductions. Use an integer
27432 VEC_SERIES_EXPR for variable-length integer reductions. Use a
27433 cast of such a series for variable-length floating-point
27436 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27437 Alan Hayward <alan.hayward@arm.com>
27438 David Sherwood <david.sherwood@arm.com>
27440 * tree.h (build_index_vector): Declare.
27441 * tree.c (build_index_vector): New function.
27442 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
27443 of units as polynomial, forcibly converting it to a constant if
27444 vectorizable_reduction has already enforced the condition.
27445 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
27446 to create a {1,2,3,...} vector.
27447 (vectorizable_reduction): Treat the number of units as polynomial.
27448 Choose vectype_in based on the largest scalar element size rather
27449 than the smallest number of units. Enforce the restrictions
27452 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27453 Alan Hayward <alan.hayward@arm.com>
27454 David Sherwood <david.sherwood@arm.com>
27456 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
27457 number of units as polynomial.
27459 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27460 Alan Hayward <alan.hayward@arm.com>
27461 David Sherwood <david.sherwood@arm.com>
27463 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
27464 * target.def (autovectorize_vector_sizes): Return the vector sizes
27465 by pointer, using vector_sizes rather than a bitmask.
27466 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
27467 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
27468 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
27470 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
27471 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
27472 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
27473 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
27474 * omp-general.c (omp_max_vf): Likewise.
27475 * omp-low.c (omp_clause_aligned_alignment): Likewise.
27476 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
27477 * tree-vect-loop.c (vect_analyze_loop): Likewise.
27478 * tree-vect-slp.c (vect_slp_bb): Likewise.
27479 * doc/tm.texi: Regenerate.
27480 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
27482 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
27483 the vector size as a poly_uint64 rather than an unsigned int.
27484 (current_vector_size): Change from an unsigned int to a poly_uint64.
27485 (get_vectype_for_scalar_type): Update accordingly.
27486 * tree.h (build_truth_vector_type): Take the size and number of
27487 units as a poly_uint64 rather than an unsigned int.
27488 (build_vector_type): Add a temporary overload that takes
27489 the number of units as a poly_uint64 rather than an unsigned int.
27490 * tree.c (make_vector_type): Likewise.
27491 (build_truth_vector_type): Take the number of units as a poly_uint64
27492 rather than an unsigned int.
27494 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27495 Alan Hayward <alan.hayward@arm.com>
27496 David Sherwood <david.sherwood@arm.com>
27498 * target.def (get_mask_mode): Take the number of units and length
27499 as poly_uint64s rather than unsigned ints.
27500 * targhooks.h (default_get_mask_mode): Update accordingly.
27501 * targhooks.c (default_get_mask_mode): Likewise.
27502 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
27503 * doc/tm.texi: Regenerate.
27505 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27506 Alan Hayward <alan.hayward@arm.com>
27507 David Sherwood <david.sherwood@arm.com>
27509 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
27510 * omp-general.c (omp_max_vf): Likewise.
27511 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
27512 (expand_omp_simd): Handle polynomial safelen.
27513 * omp-low.c (omplow_simd_context): Add a default constructor.
27514 (omplow_simd_context::max_vf): Change from int to poly_uint64.
27515 (lower_rec_simd_input_clauses): Update accordingly.
27516 (lower_rec_input_clauses): Likewise.
27518 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27519 Alan Hayward <alan.hayward@arm.com>
27520 David Sherwood <david.sherwood@arm.com>
27522 * tree-vectorizer.h (vect_nunits_for_cost): New function.
27523 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
27524 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
27525 (vect_analyze_slp_cost): Likewise.
27526 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
27527 (vect_model_load_cost): Likewise.
27529 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27530 Alan Hayward <alan.hayward@arm.com>
27531 David Sherwood <david.sherwood@arm.com>
27533 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
27534 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
27535 from an unsigned int * to a poly_uint64_pod *.
27536 (calculate_unrolling_factor): New function.
27537 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
27539 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27540 Alan Hayward <alan.hayward@arm.com>
27541 David Sherwood <david.sherwood@arm.com>
27543 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
27544 from an unsigned int to a poly_uint64.
27545 (_loop_vec_info::slp_unrolling_factor): Likewise.
27546 (_loop_vec_info::vectorization_factor): Change from an int
27548 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
27549 (vect_get_num_vectors): New function.
27550 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
27551 (vect_get_num_copies): Use vect_get_num_vectors.
27552 (vect_analyze_data_ref_dependences): Change max_vf from an int *
27553 to an unsigned int *.
27554 (vect_analyze_data_refs): Change min_vf from an int * to a
27556 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
27557 than an unsigned HOST_WIDE_INT.
27558 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
27559 (vect_analyze_data_ref_dependence): Change max_vf from an int *
27560 to an unsigned int *.
27561 (vect_analyze_data_ref_dependences): Likewise.
27562 (vect_compute_data_ref_alignment): Handle polynomial vf.
27563 (vect_enhance_data_refs_alignment): Likewise.
27564 (vect_prune_runtime_alias_test_list): Likewise.
27565 (vect_shift_permute_load_chain): Likewise.
27566 (vect_supportable_dr_alignment): Likewise.
27567 (dependence_distance_ge_vf): Take the vectorization factor as a
27568 poly_uint64 rather than an unsigned HOST_WIDE_INT.
27569 (vect_analyze_data_refs): Change min_vf from an int * to a
27571 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
27572 vfm1 as a poly_uint64 rather than an int. Make the same change
27573 for the returned bound_scalar.
27574 (vect_gen_vector_loop_niters): Handle polynomial vf.
27575 (vect_do_peeling): Likewise. Update call to
27576 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
27577 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
27579 * tree-vect-loop.c (vect_determine_vectorization_factor)
27580 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
27581 (vect_get_known_peeling_cost): Likewise.
27582 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
27583 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
27584 (vect_transform_loop): Likewise. Use the lowest possible VF when
27585 updating the upper bounds of the loop.
27586 (vect_min_worthwhile_factor): Make static. Return an unsigned int
27587 rather than an int.
27588 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
27589 polynomial unroll factors.
27590 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
27591 (vect_make_slp_decision): Likewise.
27592 (vect_supported_load_permutation_p): Likewise, and polynomial
27594 (vect_analyze_slp_cost): Handle polynomial vf.
27595 (vect_slp_analyze_node_operations): Likewise.
27596 (vect_slp_analyze_bb_1): Likewise.
27597 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
27598 than an unsigned HOST_WIDE_INT.
27599 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
27600 (vectorizable_load): Handle polynomial vf.
27601 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
27603 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
27605 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27606 Alan Hayward <alan.hayward@arm.com>
27607 David Sherwood <david.sherwood@arm.com>
27609 * match.pd: Handle bit operations involving three constants
27610 and try to fold one pair.
27612 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27614 * tree-vect-loop-manip.c: Include gimple-fold.h.
27615 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
27616 niters_maybe_zero parameters. Handle other cases besides a step of 1.
27617 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
27618 Add a path that uses a step of VF instead of 1, but disable it
27620 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
27621 and niters_no_overflow parameters. Update calls to
27622 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
27623 Create a new SSA name if the latter choses to use a ste other
27624 than zero, and return it via niters_vector_mult_vf_var.
27625 * tree-vect-loop.c (vect_transform_loop): Update calls to
27626 vect_do_peeling, vect_gen_vector_loop_niters and
27627 slpeel_make_loop_iterate_ntimes.
27628 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
27629 (vect_gen_vector_loop_niters): Update declarations after above changes.
27631 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
27633 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
27634 128-bit round to integer instructions.
27635 (ceil<mode>2): Likewise.
27636 (btrunc<mode>2): Likewise.
27637 (round<mode>2): Likewise.
27639 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
27641 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
27642 unaligned VSX load/store on P8/P9.
27643 (expand_block_clear): Allow the use of unaligned VSX
27644 load/store on P8/P9.
27646 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
27648 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
27650 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
27651 swap associated with both a load and a store.
27653 2018-01-02 Andrew Waterman <andrew@sifive.com>
27655 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
27656 * config/riscv/riscv.md (clear_cache): Use it.
27658 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
27660 * web.c: Remove out-of-date comment.
27662 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
27664 * expr.c (fixup_args_size_notes): Check that any existing
27665 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
27666 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
27667 (emit_single_push_insn): ...here.
27669 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
27671 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
27672 (const_vector_encoded_nelts): New function.
27673 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
27674 (const_vector_int_elt, const_vector_elt): Declare.
27675 * emit-rtl.c (const_vector_int_elt_1): New function.
27676 (const_vector_elt): Likewise.
27677 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
27678 of CONST_VECTOR_ELT.
27680 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
27682 * expr.c: Include rtx-vector-builder.h.
27683 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
27684 directly on the tree encoding.
27685 (const_vector_from_tree): Likewise.
27686 * optabs.c: Include rtx-vector-builder.h.
27687 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
27688 sequence of "u" values.
27689 * vec-perm-indices.c: Include rtx-vector-builder.h.
27690 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
27691 directly on the vec_perm_indices encoding.
27693 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
27695 * doc/rtl.texi (const_vector): Describe new encoding scheme.
27696 * Makefile.in (OBJS): Add rtx-vector-builder.o.
27697 * rtx-vector-builder.h: New file.
27698 * rtx-vector-builder.c: Likewise.
27699 * rtl.h (rtx_def::u2): Add a const_vector field.
27700 (CONST_VECTOR_NPATTERNS): New macro.
27701 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
27702 (CONST_VECTOR_DUPLICATE_P): Likewise.
27703 (CONST_VECTOR_STEPPED_P): Likewise.
27704 (CONST_VECTOR_ENCODED_ELT): Likewise.
27705 (const_vec_duplicate_p): Check for a duplicated vector encoding.
27706 (unwrap_const_vec_duplicate): Likewise.
27707 (const_vec_series_p): Check for a non-duplicated vector encoding.
27708 Say that the function only returns true for integer vectors.
27709 * emit-rtl.c: Include rtx-vector-builder.h.
27710 (gen_const_vec_duplicate_1): Delete.
27711 (gen_const_vector): Call gen_const_vec_duplicate instead of
27712 gen_const_vec_duplicate_1.
27713 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
27714 (gen_const_vec_duplicate): Use rtx_vector_builder.
27715 (gen_const_vec_series): Likewise.
27716 (gen_rtx_CONST_VECTOR): Likewise.
27717 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
27718 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
27719 Build a new vector rather than modifying a CONST_VECTOR in-place.
27720 (handle_special_swappables): Update call accordingly.
27721 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
27722 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
27723 Build a new vector rather than modifying a CONST_VECTOR in-place.
27724 (handle_special_swappables): Update call accordingly.
27726 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
27728 * simplify-rtx.c (simplify_const_binary_operation): Use
27729 CONST_VECTOR_ELT instead of XVECEXP.
27731 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
27733 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
27734 the selector elements to be different from the data elements
27735 if the selector is a VECTOR_CST.
27736 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
27737 ssizetype for the selector.
27739 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
27741 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
27742 before testing each element individually.
27743 * tree-vect-generic.c (lower_vec_perm): Likewise.
27745 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
27747 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
27748 * selftest-run-tests.c (selftest::run_tests): Call it.
27749 * vector-builder.h (vector_builder::operator ==): New function.
27750 (vector_builder::operator !=): Likewise.
27751 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
27752 (vec_perm_indices::all_from_input_p): New function.
27753 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
27754 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
27755 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
27756 instead of reading the VECTOR_CST directly. Detect whether both
27757 vector inputs are the same before constructing the vec_perm_indices,
27758 and update the number of inputs argument accordingly. Use the
27759 utility functions added above. Only construct sel2 if we need to.
27761 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
27763 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
27764 the broadcast of the low byte.
27765 (expand_mult_highpart): Use an explicit encoding for the permutes.
27766 * optabs-query.c (can_mult_highpart_p): Likewise.
27767 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
27768 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
27769 (vectorizable_bswap): Likewise.
27770 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
27771 explicit encoding for the power-of-2 permutes.
27772 (vect_permute_store_chain): Likewise.
27773 (vect_grouped_load_supported): Likewise.
27774 (vect_permute_load_chain): Likewise.
27776 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
27778 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
27779 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
27780 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
27781 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
27782 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
27783 (vect_gen_perm_mask_any): Likewise.
27785 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
27787 * int-vector-builder.h: New file.
27788 * vec-perm-indices.h: Include int-vector-builder.h.
27789 (vec_perm_indices): Redefine as an int_vector_builder.
27790 (auto_vec_perm_indices): Delete.
27791 (vec_perm_builder): Redefine as a stand-alone class.
27792 (vec_perm_indices::vec_perm_indices): New function.
27793 (vec_perm_indices::clamp): Likewise.
27794 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
27795 (vec_perm_indices::new_vector): New function.
27796 (vec_perm_indices::new_expanded_vector): Update for new
27797 vec_perm_indices class.
27798 (vec_perm_indices::rotate_inputs): New function.
27799 (vec_perm_indices::all_in_range_p): Operate directly on the
27800 encoded form, without computing elided elements.
27801 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
27802 encoding. Update for new vec_perm_indices class.
27803 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
27804 the given vec_perm_builder.
27805 (expand_vec_perm_var): Update vec_perm_builder constructor.
27806 (expand_mult_highpart): Use vec_perm_builder instead of
27807 auto_vec_perm_indices.
27808 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
27809 vec_perm_indices instead of auto_vec_perm_indices. Use a single
27810 or double series encoding as appropriate.
27811 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
27812 vec_perm_indices instead of auto_vec_perm_indices.
27813 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
27814 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
27815 (vect_permute_store_chain): Likewise.
27816 (vect_grouped_load_supported): Likewise.
27817 (vect_permute_load_chain): Likewise.
27818 (vect_shift_permute_load_chain): Likewise.
27819 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
27820 (vect_transform_slp_perm_load): Likewise.
27821 (vect_schedule_slp_instance): Likewise.
27822 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
27823 (vectorizable_mask_load_store): Likewise.
27824 (vectorizable_bswap): Likewise.
27825 (vectorizable_store): Likewise.
27826 (vectorizable_load): Likewise.
27827 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
27828 vec_perm_indices instead of auto_vec_perm_indices. Use
27829 tree_to_vec_perm_builder to read the vector from a tree.
27830 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
27831 vec_perm_builder instead of a vec_perm_indices.
27832 (have_whole_vector_shift): Use vec_perm_builder and
27833 vec_perm_indices instead of auto_vec_perm_indices. Leave the
27834 truncation to calc_vec_perm_mask_for_shift.
27835 (vect_create_epilog_for_reduction): Likewise.
27836 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
27837 from auto_vec_perm_indices to vec_perm_indices.
27838 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
27839 instead of changing individual elements.
27840 (aarch64_vectorize_vec_perm_const): Use new_vector to install
27841 the vector in d.perm.
27842 * config/arm/arm.c (expand_vec_perm_d::perm): Change
27843 from auto_vec_perm_indices to vec_perm_indices.
27844 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
27845 instead of changing individual elements.
27846 (arm_vectorize_vec_perm_const): Use new_vector to install
27847 the vector in d.perm.
27848 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
27849 Update vec_perm_builder constructor.
27850 (rs6000_expand_interleave): Likewise.
27851 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
27852 (rs6000_expand_interleave): Likewise.
27854 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
27856 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
27857 to qimode could truncate the indices.
27858 * optabs.c (expand_vec_perm_var): Likewise.
27860 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
27862 * Makefile.in (OBJS): Add vec-perm-indices.o.
27863 * vec-perm-indices.h: New file.
27864 * vec-perm-indices.c: Likewise.
27865 * target.h (vec_perm_indices): Replace with a forward class
27867 (auto_vec_perm_indices): Move to vec-perm-indices.h.
27868 * optabs.h: Include vec-perm-indices.h.
27869 (expand_vec_perm): Delete.
27870 (selector_fits_mode_p, expand_vec_perm_var): Declare.
27871 (expand_vec_perm_const): Declare.
27872 * target.def (vec_perm_const_ok): Replace with...
27873 (vec_perm_const): ...this new hook.
27874 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
27875 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
27876 * doc/tm.texi: Regenerate.
27877 * optabs.def (vec_perm_const): Delete.
27878 * doc/md.texi (vec_perm_const): Likewise.
27879 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
27880 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
27881 expand_vec_perm for constant permutation vectors. Assert that
27882 the mode of variable permutation vectors is the integer equivalent
27883 of the mode that is being permuted.
27884 * optabs-query.h (selector_fits_mode_p): Declare.
27885 * optabs-query.c: Include vec-perm-indices.h.
27886 (selector_fits_mode_p): New function.
27887 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
27888 is defined, instead of checking whether the vec_perm_const_optab
27889 exists. Use targetm.vectorize.vec_perm_const instead of
27890 targetm.vectorize.vec_perm_const_ok. Check whether the indices
27891 fit in the vector mode before using a variable permute.
27892 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
27893 vec_perm_indices instead of an rtx.
27894 (expand_vec_perm): Replace with...
27895 (expand_vec_perm_const): ...this new function. Take the selector
27896 as a vec_perm_indices rather than an rtx. Also take the mode of
27897 the selector. Update call to shift_amt_for_vec_perm_mask.
27898 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
27899 Use vec_perm_indices::new_expanded_vector to expand the original
27900 selector into bytes. Check whether the indices fit in the vector
27901 mode before using a variable permute.
27902 (expand_vec_perm_var): Make global.
27903 (expand_mult_highpart): Use expand_vec_perm_const.
27904 * fold-const.c: Includes vec-perm-indices.h.
27905 * tree-ssa-forwprop.c: Likewise.
27906 * tree-vect-data-refs.c: Likewise.
27907 * tree-vect-generic.c: Likewise.
27908 * tree-vect-loop.c: Likewise.
27909 * tree-vect-slp.c: Likewise.
27910 * tree-vect-stmts.c: Likewise.
27911 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
27913 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
27914 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
27915 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
27916 (aarch64_vectorize_vec_perm_const): ...this new function.
27917 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
27918 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
27919 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
27920 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
27921 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
27922 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
27923 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
27925 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
27926 check for NEON modes.
27927 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
27928 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
27929 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
27930 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
27932 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
27933 the old VEC_PERM_CONST conditions.
27934 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
27935 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
27936 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
27937 (ia64_vectorize_vec_perm_const_ok): Merge into...
27938 (ia64_vectorize_vec_perm_const): ...this new function.
27939 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
27940 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
27941 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
27942 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
27943 * config/mips/mips.c (mips_expand_vec_perm_const)
27944 (mips_vectorize_vec_perm_const_ok): Merge into...
27945 (mips_vectorize_vec_perm_const): ...this new function.
27946 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
27947 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
27948 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
27949 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
27950 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
27951 (rs6000_expand_vec_perm_const): Delete.
27952 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
27954 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
27955 (altivec_expand_vec_perm_const_le): Take each operand individually.
27956 Operate on constant selectors rather than rtxes.
27957 (altivec_expand_vec_perm_const): Likewise. Update call to
27958 altivec_expand_vec_perm_const_le.
27959 (rs6000_expand_vec_perm_const): Delete.
27960 (rs6000_vectorize_vec_perm_const_ok): Delete.
27961 (rs6000_vectorize_vec_perm_const): New function.
27962 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
27963 an element count and rtx array.
27964 (rs6000_expand_extract_even): Update call accordingly.
27965 (rs6000_expand_interleave): Likewise.
27966 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
27967 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
27968 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
27969 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
27970 (rs6000_expand_vec_perm_const): Delete.
27971 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
27972 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
27973 (altivec_expand_vec_perm_const_le): Take each operand individually.
27974 Operate on constant selectors rather than rtxes.
27975 (altivec_expand_vec_perm_const): Likewise. Update call to
27976 altivec_expand_vec_perm_const_le.
27977 (rs6000_expand_vec_perm_const): Delete.
27978 (rs6000_vectorize_vec_perm_const_ok): Delete.
27979 (rs6000_vectorize_vec_perm_const): New function. Remove stray
27980 reference to the SPE evmerge intructions.
27981 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
27982 an element count and rtx array.
27983 (rs6000_expand_extract_even): Update call accordingly.
27984 (rs6000_expand_interleave): Likewise.
27985 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
27986 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
27988 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
27990 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
27992 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
27993 vector mode and that that mode matches the mode of the data
27995 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
27996 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
27997 directly using expand_vec_perm_1 when forcing selectors into
27999 (expand_vec_perm_var): New function, split out from expand_vec_perm.
28001 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
28003 * optabs-query.h (can_vec_perm_p): Delete.
28004 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
28005 * optabs-query.c (can_vec_perm_p): Split into...
28006 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
28007 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
28008 particular selector is valid.
28009 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
28010 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
28011 (vect_grouped_load_supported): Likewise.
28012 (vect_shift_permute_load_chain): Likewise.
28013 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
28014 (vect_transform_slp_perm_load): Likewise.
28015 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
28016 (vectorizable_bswap): Likewise.
28017 (vect_gen_perm_mask_checked): Likewise.
28018 * fold-const.c (fold_ternary_loc): Likewise. Don't take
28019 implementations of variable permutation vectors into account
28020 when deciding which selector to use.
28021 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
28022 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
28023 with a false third argument.
28024 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
28025 to test whether the constant selector is valid and can_vec_perm_var_p
28026 to test whether a variable selector is valid.
28028 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
28030 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
28031 * optabs-query.c (can_vec_perm_p): Likewise.
28032 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
28033 instead of vec_perm_indices.
28034 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
28035 (vect_gen_perm_mask_checked): Likewise,
28036 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
28037 (vect_gen_perm_mask_checked): Likewise,
28039 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
28041 * optabs-query.h (qimode_for_vec_perm): Declare.
28042 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
28043 (qimode_for_vec_perm): ...this new function.
28044 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
28046 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
28048 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
28049 does not have a conditional at the top.
28051 2018-01-02 Richard Biener <rguenther@suse.de>
28053 * ipa-inline.c (big_speedup_p): Fix expression.
28055 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
28058 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
28061 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
28065 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
28066 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
28067 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
28068 cond_taken_branch_cost 3->4.
28070 2018-01-01 Jakub Jelinek <jakub@redhat.com>
28072 PR tree-optimization/83581
28073 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
28074 TODO_cleanup_cfg if any changes have been made.
28076 PR middle-end/83608
28077 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
28078 convert_modes if target mode has the right side, but different mode
28081 PR middle-end/83609
28082 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
28083 last argument when extracting from CONCAT. If either from_real or
28084 from_imag is NULL, use expansion through memory. If result is not
28085 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
28086 the parts directly to inner mode, if even that fails, use expansion
28089 PR middle-end/83623
28090 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
28091 check for bswap in mode rather than HImode and use that in expand_unop
28094 Copyright (C) 2018 Free Software Foundation, Inc.
28096 Copying and distribution of this file, with or without modification,
28097 are permitted in any medium without royalty provided the copyright
28098 notice and this notice are preserved.