Remove stray reference to error_at_rich_loc
[official-gcc.git] / gcc / ChangeLog
blob82f3f02dabc398ca8a26ba85478f586fea2d1a6f
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>
8         PR target/87563
9         * tree-vectorizer.c (try_vectorize_loop_1): Don't use
10         if-conversioned loop when it contains ifn with types not
11         supported by backend.
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>
18         PR target/87572
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>
31         PR middle-end/87610
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
35         escaped.
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>
46         PR target/87599
47         * config/i386/sse.md (*vec_dupv2di): Add register source to
48         movddup.
50 2018-10-14  H.J. Lu  <hongjiu.lu@intel.com>
52         PR target/87572
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>
84         PR middle-end/81376
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
128         noinline.
130 2018-10-11  Jan Hubicka  <hubicka@ucw.cz>
132         PR target/87156
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>
138         PR middle-end/87593
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
153         function.
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>
160         PR target/87511
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
178         "repeated" elements.
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>
193         PR middle-end/87574
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>
204         PR target/87573
205         * config/i386/mmx.md (const_vector 0 -> mem splitter): New splitter.
207 2018-10-10  Jakub Jelinek  <jakub@redhat.com>
209         PR target/87550
210         * config/i386/i386-builtin.def (IX86_BUILTIN_RDPMC): Move from args set
211         to special_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
225         to SSE there.
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.
233         (main): Likewise.
234         * input.c (selftest::ebcdic_execution_charset::apply):
235         Update for renaming of...
236         (selftest::ebcdic_execution_charset::on_error): ...this, renaming
237         to...
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
250         "reason".
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
272         replacements.
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
283         information.
285 2018-10-08  Eric Botcazou  <ebotcazou@adacore.com>
287         * cgraphunit.c (cgraph_node::expand_thunk): Do not force DECL_IGNORED_P
288         on the thunk.
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.
299         
300 2018-10-08  Richard Sandiford  <richard.sandiford@arm.com>
302         PR middle-end/63155
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>
308         PR target/87517
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):
334         Open a dump scope.
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
344         bitmap bit clearing.
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>
383         PR c++/56856
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.
402         (mfloat64): 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
413         "scratch" argument.
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
440         consistently.
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
456         RPO index.
457         (add_control_edge): Likewise.
458         (ssa_propagation_engine::process_ssa_edge_worklist): Fold
459         into ...
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
463         over iteration.
464         (ssa_prop_init): Allocate new worklists, do not dump
465         immediate uses.
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
480         with BLOCK_ABSTRACT.
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
489         vector type.
491 2018-10-05  H.J. Lu  <hongjiu.lu@intel.com>
493         PR target/87522
494         * config/i386/gnu-user.h (ASM_SPEC): Don't pass -msse2avx to
495         assembler for -mavx.
496         * config/i386/gnu-user64.h (ASM_SPEC): Likewise.
498 2018-10-05  Segher Boessenkool  <segher@kernel.crashing.org>
500         PR target/87509
501         * config/rs6000/rs6000-builtin.def (RS6000_BUILTIN_SET_FPSCR_DRN): Use
502         RS6000_BTM_DFP.
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
518         -Wno-prio-ctor-dtor.
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
539         to...
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
557         TDF_COMPARE_DEBUG.
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
575         failure_at calls.
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
580         location.
581         (find_data_references_in_stmt): Convert return type from bool to
582         opt_result, converting "return false" to opt_result::failure_at
583         with a new message.
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
588         opt_result.
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
602         to opt_result.
603         (verify_data_ref_alignment): Likewise, converting dump_printf_loc
604         calls to opt_result::failure_at, using the stmt location rather
605         than vect_location.
606         (vect_verify_datarefs_alignment): Convert return type from bool
607         to opt_result.
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
611         to opt_result.
612         (vect_analyze_data_ref_accesses): Likewise, converting a
613         "return false" to a "return opt_result::failure_at", adding a
614         new message.
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
633         vect_location.
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
649         opt_loop_vec_info.
650         * tree-vect-slp.c (vect_analyze_slp): Convert return type from
651         bool to opt_result.
652         * tree-vect-stmts.c (process_use): Likewise, converting
653         dump_printf_loc call and using stmt location, rather than
654         vect_location.
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
670         to opt_result.
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
683         opt_loop_vec_info.
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"
690         sub-option.
691         * dump-context.h (dump_context::apply_dump_filter_p): New decl.
692         * dumpfile.c (dump_options): Update for renaming of MSG_ALL to
693         MSG_ALL_KINDS.
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
723         with MSG_PRIORITY_*.
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>
745         PR c/87483
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>
751         PR ipa/82625
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
759         fields.
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
766         bodies into...
767         (gcc::dump_manager::update_dfi_for_opt_info): ...this new member
768         function.
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):
786         Define.
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
800         and emit gimple.
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>
815         PR ipa/87491
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
822         target's wchar_t.
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
836         initialize it.
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.
851         (R): Likewise.
852         (D): Likewise.
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
866         precision fields.
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
879         pointer conversions.
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>
898         PR target/87474
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.
920         (isinfxf2): Ditto.
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>
934         PR middle-end/87319
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
947         state to callers.
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.
969         (jumpif): 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>
987         PR 69431
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>
1043         PR target/86486
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>
1049         PR target/86486
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>
1060         PR target/86486
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>
1070         PR target/86486
1071         * explow.c (anti_adjust_stack_and_probe_stack_clash): Support custom
1072         probe ranges.
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>
1084         PR target/86486
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>
1093         PR target/86486
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>
1100         PR target/86486
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
1122         constraints.
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
1142         checking.
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.
1151         (mmx_femms): Ditto.
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
1165         function comment.
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
1170         comment.
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
1183         function comment.
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>
1191         PR target/87370
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>
1213         PR target/87467
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
1246         do it as a .long .
1247         * config/rs6000/rs6000.h (TARGET_LWSYNC_INSTRUCTION): Delete.
1248         * config/rs6000/sync.md (*lwsync): Always generate lwsync, never do it
1249         as a .long .
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):
1321         Likewise.
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.
1381         (fldenv): Ditto.
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>
1397         PR target/87149
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>
1418         PR debug/37801
1419         PR debug/87440
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
1425         to subblocks.
1427 2018-09-27  Andrew Stubbs  <ams@codesourcery.com>
1428             Tom de Vries  <tom@codesourcery.com>
1430         PR 82089
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>
1476         PR target/87414
1477         * config/i386/i386.c: Include debug.h and dwarf2out.h.
1478         (output_indirect_thunk): Emit DW_CFA_def_cfa_offset after the
1479         call.
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.
1501         (sincosxf3): 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>
1511         PR target/87439
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>
1517         Revert
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>
1525         PR debug/87443
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>
1553         PR debug/87428
1554         PR debug/87362
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
1559         recognize it.
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>
1587         PR c/87387
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>
1594         PR target/86987
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
1600         u5bit_cint_operand.
1601         (*altivec_vsplth_internal): Ditto.
1602         (altivec_vsplth_direct): Ditto.
1603         (altivec_vspltw): Use const_0_to_3_operand instead of
1604         u5bit_cint_operand.
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>
1615         PR debug/83941
1616         * dwarf2out.c (add_AT_external_die_ref): Remove now redundant
1617         GC-ification.
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>
1627         PR target/71278
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
1650         immediately.
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>
1661         PR debug/83941
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
1665         external_die_map.
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
1671         external DIE map.
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
1685         DIEs.
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>
1694         PR fortran/87394
1695         * dbgcnt.c (dbg_cnt_process_single_pair): Return false
1696         instead of NULL.
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
1706         unused field.
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
1720         unused function.
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):
1738         Likewise.
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>
1751         PR middle-end/86078
1752         * doc/invoke.texi: Document all parameters and remove default
1753         of the parameters.
1755 2018-09-25  Ilya Leoshkevich  <iii@linux.ibm.com>
1757         PR bootstrap/87417
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>
1763         PR target/87391
1764         * config/riscv/riscv.h (STACK_BOUNDARY): Test riscv_abi == ABI_ILP32E
1765         not TARGET_RVE.
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>
1776         PR target/82699
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
1781         is true.
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
1789         types.
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>
1809         PR target/80080
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.
1817         Reword the help.
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
1823         message.
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>
1832         PR sanitizer/85774
1833         * asan.c: Make asan_handled_variables extern.
1834         * asan.h: Likewise.
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
1845         executable.
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.
1858         (k): Ditto.
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
1890         as GENERAL_REGS.
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>
1898         PR target/86798
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>
1904         PR middle-end/81035
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
1913         vr_values class.
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
1999         set sum_all.
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
2003         qualifier.
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
2008         arguments.
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
2016         histogram.
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
2022         it not obsolete.
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
2038         histogram.
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
2046         mode.
2047         * postreload-gcse.c (eliminate_partially_redundant_load): Fix
2048         GCOV coding style.
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
2070         overlapping stores.
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>
2083         PR middle-end/87054
2084         * gimplify.c (gimplify_expr): Retain alignment of
2085         addressable lvalue in dereference.
2087 2018-09-20  Alexandre Oliva <aoliva@redhat.com>
2089         PR bootstrap/87013
2090         * configure.ac: Check for .loc is_stmt support.
2091         * configure, config.in: Rebuilt.
2092         * dwarf2out.c (dwarf2out_source_line): Skip is_stmt
2093         if not supported.
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
2099         -misel=no.
2101 2018-09-20  Segher Boessenkool  <segher@kernel.crashing.org>
2103         * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
2104         VECTOR_OTHER.
2105         * config/rs6000/rs6000.c (rs6000_debug_vector_unit): Delete
2106         case VECTOR_OTHER.
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
2243         and use stlur.
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>
2281         PR other/87353
2282         * doc/invoke.texi (Link Options): Fix formatting and grammar.
2284 2018-09-18  Richard Biener  <rguenther@suse.de>
2286         PR middle-end/63155
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
2297         to __ffsdi2.
2299 2018-09-17  David Malcolm  <dmalcolm@redhat.com>
2301         * diagnostic-show-locus.c (class layout_range): Add field
2302         "m_original_idx".
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
2311         get_text call.
2312         (gcc_rich_location::add_location_if_nearby): Use 0 for
2313         original_idx.
2314         * gcc-rich-location.h (text_range_label::get_text): Update for new
2315         param.
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
2338         class.
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
2352         SImode args.
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
2374         iterating.
2376 2018-09-17  Martin Jambor  <mjambor@suse.cz>
2378         PR c/63886
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
2391         access.
2393 2018-09-17  Vineet Gupta  <vgupta@synopsys.com>
2395         * config/arc/arc.c: Object attributes for core4 not reflected
2396         correctly.
2397         * config/arc/arc.h: Don't restrict DBNZ to core3 (core4 includes
2398         core3).
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>
2440         PR middle-end/86864
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>
2446         PR middle-end/87188
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>
2456         PR c/82967
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
2477         a call to...
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>
2487         revert:
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
2499         arrays.
2500         (gimple_fold_builtin_sprintf): Propagate NO_WARNING to transformed
2501         calls.
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>
2542         PR target/87224
2543         * config/rs6000/rs6000.md (*mov<mode>_hardfloat64): Add Z to the Y
2544         alternatives.
2546 2018-09-14  Sam Tebbs  <sam.tebbs@arm.com>
2548         PR target/85628
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
2561         PR lto/87283
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
2573         formatting.
2575 2018-09-14  Richard Biener  <rguenther@suse.de>
2577         PR middle-end/63155
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
2604         correctly.
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>
2623         PR target/86812
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>
2643         PR bootstrap/87134
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>
2651         PR debug/86985
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>
2657         PR target/85628
2658         * config/aarch64/aarch64.md (*aarch64_bfxil):
2659         Define.
2660         * config/aarch64/constraints.md (Ulc): Define.
2661         * config/aarch64/aarch64-protos.h (aarch64_high_bits_all_ones_p):
2662         Define.
2663         * config/aarch64/aarch64.c (aarch64_high_bits_all_ones_p):
2664         New function.
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>
2688         PR middle-end/87290
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>
2701         PR middle-end/82853
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>
2711         PR middle-end/86336
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
2734         visited blocks.
2736 2018-09-12  Andreas Krebbel  <krebbel@linux.ibm.com>
2738         * config/s390/s390.md (PFPO_RND_MODE_DFP, PFPO_RND_MODE_BFP): New
2739         constants.
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>
2757         PR middle-end/87248
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
2774         -gno-describe-dies.
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
2783         depend on sign.
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
2872         predicate.
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>
2883         PR target/86794
2884         * config/mmix/mmix.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
2885         to speculation_safe_value_not_needed.
2887         PR target/85666
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
2901         earlyclobber.
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
2921         splitter.
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>
2933         PR target/86731
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.
2954         (fistdi2): Ditto.
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
2960         splitters.
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
2977         the init value.
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>
2993         PR target/80080
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>
3003         PR target/80080
3004         * config/s390/s390-passes.def: New file.
3005         * config/s390/s390-protos.h (class rtl_opt_pass): Add forward
3006         declaration.
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>
3027         PR target/86779
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".
3037         (ASM_SPEC): Define.
3038         (TARGET_SM35): New macro.
3039         * config/nvptx/nvptx.md (atomic_fetch_<logic><mode>): Enable with the
3040         correct predicate.
3041         * config/nvptx/nvptx.opt (ptx_isa, sm_30, sm_35): New enum and its
3042         values.
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>
3076         PR bootstrap/87134
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>
3095         PR c++/87137
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>
3101         PR bootstrap/87225
3102         * tree-vect-stmts.c (vectorizable_simd_clone_call): Fix bogus
3103         return.
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):
3111         Likewise.
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):
3121         New tuning flag.
3123 2018-09-05  Martin Liska  <mliska@suse.cz>
3125         * doc/gcov.texi: Update documentation of humar
3126         readable mode.
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>
3132         PR target/87164
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>
3139         PR debug/86593
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>
3146         PR target/87198
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
3170         fnargs.
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
3192         static.
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
3217         cleanup.
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
3224         gcc_stablesort.
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
3257         simplify result.
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
3275         executable.
3277 2018-09-03  Martin Liska  <mliska@suse.cz>
3279         PR driver/83193
3280         * common/common-target.def: Add TARGET_GET_VALID_OPTION_VALUES.
3281         * common/common-targhooks.c (default_get_valid_option_values):
3282         New function.
3283         * common/common-targhooks.h (default_get_valid_option_values):
3284         Likewise.
3285         * common/config/i386/i386-common.c: Move processor_target_table
3286         from i386.c.
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
3300         hook.
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>
3306         PR middle-end/59521
3307         * predict.c (set_even_probabilities): Add likely_edges
3308         argument and handle cases where we have precisely one
3309         likely edge.
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
3316         nodes.
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
3346         be available.
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"
3362         and "vr1".
3363         (vrp_can_optimize_bit_op): Likewise.
3364         (extract_range_from_binary_expr_1): Likewise for params "vr0_" and
3365         "vr1_".
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
3402         to sem_item.
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>
3415         PR target/86684
3416         PR target/87149
3417         * config/rs6000/rs6000.md (lround<mode>di2): Gate on TARGET_FPRND.
3419 2018-08-31  Jakub Jelinek  <jakub@redhat.com>
3421         PR middle-end/87138
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
3445         vector load modes.
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>
3464         PR lto/86517
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
3471         TYPE_STUB_DECL.
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
3500         DECL_VINDEX.
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>
3523         PR middle-end/86995
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>
3529         PR middle-end/87053
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
3542         string_constant.
3543         (c_getstr): Likewise.
3544         * tree-ssa-forwprop.c (simplify_builtin_call): Likewise.
3545         to string_constant.
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
3560         boundaries.
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>
3585         PR other/86726
3586         * invoke.texi (Optimization Options): List -ftree-scev-cprop.
3587         (-O): Ditto.
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
3593         parameters.
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
3618         invalid address.
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>
3638         PR middle-end/86631
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
3650         >= 3005 to >= 9000.
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>
3660         PR middle-end/87099
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
3692         expressions.
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
3704         functions.
3707 2018-08-27  Jeff Law  <law@redhat.com>
3708         PR tree-optimization/87110
3709         * tree-ssa-dse.c (compute_trims): Handle non-constant
3710         TYPE_SIZE_UNIT.
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>
3746         PR c++/87091
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):
3762         Likewise.
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>
3774         PR c++/87091
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>
3782         PR sanitizer/86962
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
3797         possible.
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):
3823         Likewise.
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
3886         fndecl_built_in_p.
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
3900         is_single_value_p.
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
3910         handlel by a test.
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):
3926         Make it static.
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):
3936         (expand_builtins):
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
3988         gimple functions.
3989         (switch_conversion::check_final_bb): Likewise.
3990         (switch_conversion::gather_default_values): Pass new argument
3991         to label_to_block.
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
3996         to label_to_block.
3997         (switch_decision_tree::try_switch_expansion): Use
3998         gimple_switch_default_edge.
3999         * tree-vrp.c (find_switch_asserts): Pass new argument
4000         to label_to_block.
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):
4019         New functions.
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
4036         lookup.
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
4056         check.
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>
4068         PR driver/83193
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
4074         options.
4076 2018-08-26  Jeff Law  <law@redhat.com>
4078         * config/mips/frame-header-opt.c: Include "backend.h" rather than
4079         "cfg.h"
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>
4100         PR c++/87091
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
4120         POINTER_PLUS_EXPR.
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
4128         instead of min/max.
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
4191         vectors.
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>
4203         PR debug/79342
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
4215         flag dynamically.
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>
4226         PR target/86989
4227         * config/rs6000/rs6000.c (toc_relative_expr_p): Check that the base is
4228         the TOC register.
4230 2018-08-24  Aldy Hernandez  <aldyh@redhat.com>
4232         PR 87073/bootstrap
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
4239         "permutaion".
4241 2018-08-23  Giuliano Belinassi  <giuliano.belinassi@usp.br>
4243         * genmatch.c (parser::parse_operation): Fix typo 'exapnded'
4244         to 'expanded'.
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'
4287         into 'arch'.
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
4291         'simd' into 'arch'.
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>
4301         PR target/86951
4302         * config/arm/arm-protos.h (arm_emit_speculation_barrier): New
4303         prototype.
4304         * config/arm/arm.c (speculation_barrier_libfunc): New static
4305         variable.
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
4310         DSB or ISB.
4311         (speculation_barrier_insn): Only match on Armv7 or later.
4313 2018-08-23  Richard Biener  <rguenther@suse.de>
4315         PR middle-end/87024
4316         * tree-inline.c (copy_bb): Drop unused __builtin_va_arg_pack_len
4317         calls.
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
4324         one input is given.
4326 2018-08-23  Richard Sandiford  <richard.sandiford@arm.com>
4328         PR target/85910
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
4335         behaviour.
4337 2018-08-22  Martin Sebor  <msebor@redhat.com>
4339         PR middle-end/87052
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
4354         function.
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>
4382         PR other/704
4383         * gcc-ar.c (main): Don’t try to invoke the plug-in if we’re not
4384         building it.
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>
4401         PR bootstrap/81033
4402         PR target/81733
4403         PR target/52795
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
4407         label.
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
4444         when possible.
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>
4452         PR middle-end/86121
4453         * tree-ssa-strlen.c (adjust_last_stmt): Avoid folding away undefined
4454         behaviour.
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
4495         -fdump-early-debug.
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>
4516         PR target/87033
4517         * config/rs6000/rs6000.md (extendsi<mode>2): Change constraints
4518         from 'Y' to 'YZ' to enable the LWAX instruction to be generated
4519         for indexed loads.
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>
4529         PR other/84889
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
4533         instance.
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
4538         a group.
4539         (emit_diagnostic): Add auto_diagnostic_group instance.
4540         (inform): Likewise.
4541         (inform_n): Likewise.
4542         (warning): Likewise.
4543         (warning_at): Likewise.
4544         (warning_n): Likewise.
4545         (pedwarn): Likewise.
4546         (permerror): Likewise.
4547         (error): Likewise.
4548         (error_n): Likewise.
4549         (error_at): Likewise.
4550         (sorry): 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",
4559         "end_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>
4584         PR target/87014
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>
4599         PR target/86984
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
4604         integer overflow.
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>
4628         PR target/86994
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
4645         a sentence.
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
4692         function.
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
4700         to 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
4749         Deprecated flags.
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
4762         CL_ERR_DEPRECATED.
4763         (read_cmdline_option): Report warning for OPT_SPECIAL_deprecated
4764         options.
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>
4770         PR middle-end/86505
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):
4815         Delete.
4817 2018-08-16  Segher Boessenkool  <segher@kernel.crashing.org>
4819         * config/rs6000/altivec.md: Don't set length attribute to the default
4820         value.
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>
4868         PR target/84711
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):
4890         Remove.
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
4902         for float vectors.
4904 2018-08-15  Uros Bizjak  <ubizjak@gmail.com>
4906         * config/i386/i386.c (expand_vec_perm_movs): Enable V4SFmode
4907         for TARGET_SSE.
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
4913         "m_label".
4914         (class layout): Add field "m_show_labels_p".
4915         (layout_range::layout_range): Add param "label" and use it to
4916         initialize m_label.
4917         (make_range): Pass in NULL for new "label" param of layout_range's
4918         ctor.
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
4930         "show_labels_p".
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"
4936         param.
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
4954         show_labels_p.
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>
4971         PR testsuite/86519
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>
4987         * config/darwin.c
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>
4994         PR target/81685
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
5010         patterns.
5011         (expand_vec_perm_1): Try the new method.
5013 2018-08-14  Ilya Leoshkevich  <iii@linux.ibm.com>
5015         PR target/86547
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()
5045         intrinsics.
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):
5068         New function.
5069         (s390_decompose_address): Factor out constant pool ref
5070         decomposition.
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
5099         extra_headers.
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
5105         secure attribute.
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):
5120         Remove.
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
5126         for zero value.
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>
5138         PR target/82418
5139         * config/i386/i386.md (<s>mul<mode>3_highpart): Use DWIH mode iterator
5140         instead of SWI48.
5142 2018-08-10  Martin Liska  <mliska@suse.cz>
5144         PR target/83610
5145         * builtin-types.def (BT_FN_LONG_LONG_LONG_DOUBLE): Add new
5146         function type.
5147         * builtins.c (expand_builtin_expect_with_probability):
5148         New function.
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):
5153         (fold_builtin_2):
5154         (fold_builtin_3):
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
5173         new predictor.
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
5181         early_p.
5182         (strip_predictor_early): New function.
5183         (pass_strip_predict_hints::execute): Call the function to
5184         strip predictors.
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
5196         unused variable.
5198 2018-08-09  Andreas Schwab  <schwab@linux-m68k.org>
5200         * config/m68k/m68k-protos.h (m68k_final_prescan_insn): Remove
5201         prototype.
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>
5210         PR other/84889
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>
5269         PR target/86887
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>
5277         PR c/86895
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>
5306         PR target/84332
5307         * config/s390/s390.c (s390_option_override_internal): Reduce the
5308         stack-clash-protection-probe-interval param if it would be too big
5309         for z900.
5311 2018-08-08  Andreas Schwab  <schwab@linux-m68k.org>
5313         PR target/46179
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
5319         simplify.
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>
5336         PR target/85295
5337         * config/s390/constraints.md ("NxHD0", "NxSD0"): New constraint
5338         definitions.
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
5383         usages.
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):
5393         Remove flags.
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):
5397         Remove MTN.
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):
5403         Remove enum values.
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
5423         global.
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>
5432         PR target/86838
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>
5443         PR middle-end/83023
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
5448         hit to compiler.
5450 2018-08-06  John David Anglin  <danglin@gcc.gnu.org>
5452         PR target/86785
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>
5484         PR target/86807
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
5491         assert.
5493 2018-08-06  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
5495         PR target/86662
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>
5627         PR target/86386
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
5635         NaN output to 4.
5637 2018-08-03  Sandra Loosemore  <sandra@codesourcery.com>
5639         PR target/86799
5640         * config/nios2/nios2.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
5641         Define.
5643 2018-08-03  Jeff Law  <law@redhat.com>
5645         PR target/86795
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
5681         UI_NONE.
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
5689         in fraction part.
5691 2018-08-03  Siddhesh Poyarekar  <siddhesh@sourceware.org>
5693         * config/aarch64/falkor.md (falkor_am_1_vxvy_vxvy): Move
5694         neon_dup_q to...
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
5718         parameter.
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>
5734         PR target/86790
5735         * config/m68k/m68k.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
5736         Define to speculation_safe_value_not_needed.
5738         PR target/86784
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>
5744         PR target/86660
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>
5765         PR target/86813
5766         * config/stormy16/stormy16.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
5767         Define to speculation_safe_value_not_needed.
5769         PR target/86810
5770         * config/v850/v850.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
5771         Define to speculation_safe_value_not_needed.
5773         PR target/86810
5774         * config/v850/v850.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
5775         Define to speculation_safe_value_not_needed.
5777         PR target/86803
5778         * config/rx/rx.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
5779         Define to speculation_safe_value_not_needed.
5781         PR target/86797
5782         * config/msp430/msp430.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
5783         Define to speculation_safe_value_not_needed.
5785         PR target/86791
5786         * config/mcore/mcore.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
5787         Define to speculation_safe_value_not_needed.
5789         PR target/86789
5790         * config/m32r/m32r.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
5791         Define to speculation_safe_value_not_needed.
5793         PR target/86787
5794         * config/iq2000/iq2000.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
5795         Define to speculation_safe_value_not_needed.
5797         PR target/86782
5798         * config/frv/frv.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define to
5799         speculation_safe_value_not_needed.
5801         PR target/86781
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.
5810         (mapping): Declare.
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
5824         functions.
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>
5884         PR target/86014
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.
5892         (main): Call it.
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):
5901         Likewise.
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
5910         to hold it.
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
5923         to this callsite.
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
5951         parameter.
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
5974         code...
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
6021         argument.
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
6036         argument.
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>
6049         PR target/86800
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
6056         as base and offset.
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>
6068         PR value-prof/35543
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>
6094         PR bootstrap/86724
6095         * graphite.h: Include isl/id.h and isl/space.h to allow build
6096         with ISL 0.20.
6098 2018-08-01  Jan Willem Jagersma  <jwjagersma@gmail.com>
6100         PR target/86651
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
6117         available.
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
6130         bb_discriminator.
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
6138         ATTRIBUTE_UNUSED.
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
6146         "m_test_pp_flags".
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
6152         m_test_pp.
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
6169         optinfo::add_tree.
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
6190         via emit_item.
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>
6236         PR middle-end/86705
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
6259         expand.
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
6265         function.
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
6276         prototype.
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
6321         list.
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
6338         hooks.
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
6351         dump_gimple_expr.
6352         (dump_context::dump_generic_expr_loc): Likewise, using
6353         dump_generic_expr.
6354         (dump_context::dump_printf_loc_va): Likewise, using
6355         dump_printf_va.
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
6417         new_stmt_vec_info.
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
6457         of DR_VECT_AUX.
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
6464         argument.
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
6471         DR_VECT_AUX.
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
6479         dr_vec_info.
6480         * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Update
6481         accordingly.
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
6516         above changes.
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
6525         data_reference.
6526         * tree-vect-loop-manip.c (get_misalign_in_elems)
6527         (vect_gen_prolog_loop_niters): Use dr_vec_infos.  Update calls after
6528         above changes.
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
6549         of data_ref.
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
6599         vect_is_simple_use.
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
6682         of a gphi.
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
6704         stmt_vec_infos.
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
6714         stmt_vec_info.
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
6729         chains.
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
6787         stmts.
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
6797         null arguments.
6798         (vect_find_last_scalar_stmt_in_slp): Return a stmt_vec_info instead
6799         of a gimple stmt.
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
6823         accordingly.
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
6838         accordingly.
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
6868         than a gimple stmt.
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
6893         vec<stmt_vec_info>.
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
6935         than gimple stmts.
6936         * tree-parloops.c (valid_reduction_p): Take a stmt_vec_info instead
6937         of a gimple stmt.
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
6941         than gimple stmts.
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
6953         accordingly.
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
6961         vec<stmt_vec_info>.
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
6983         to a stmt_vec_info.
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
7073         _stmt_vec_info *.
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
7093         of a gimple *.
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
7104         of vinfo_for_stmt.
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
7144         had been created.
7146 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
7148         * tree-vect-loop.c (vect_valid_reduction_input_p): New function,
7149         split out from...
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
7160         vect_free_slp_tree.
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
7172         a reduction group.
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>
7250         PR target/86640
7251         * config/arm/arm.c (arm_block_set_aligned_vect): Use gen_int_mode
7252         instead of GEN_INT.
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>
7268         PR target/86673
7269         * doc/extend.texi (Global Register Variables): Discourage use of type
7270         qualifiers.
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>
7280         PR target/86547
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
7302         used.
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
7308         fusion mask bit.
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
7319         -mcmodel= switch.
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
7328         UNSPEC.
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
7368         MEM_REF.
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>
7374         PR middle-end/86660
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
7377         kinds.
7379 2018-07-26  Martin Liska  <mliska@suse.cz>
7381         PR lto/86548
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
7414         accesses as well.
7415         (arc_is_uncached_mem_p): uncached applies to both the variable and
7416         the pointer.
7418 2018-07-25  Claudiu Zissulescu  <claziss@synopsys.com>
7420         * config/arc/arc.h (ADDITIONAL_REGISTER_NAMES): Add additional
7421         register names.
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
7431         return type.
7433 2018-07-25  Richard Biener  <rguenther@suse.de>
7435         PR debug/86654
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>
7447         PR middle-end/86645
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>
7453         PR sanitizer/79635
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
7480         assertion.
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
7488         times.
7489         (optrecord_json_writer::inlining_chain_to_json): Fix the check for
7490         UNKNOWN_LOCATION, to use LOCATION_LOCUS to look through ad-hoc
7491         wrappers.
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>
7510         PR middle-end/86627
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>
7521         PR c/86617
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>
7551         PR bootstrap/86621
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>
7567         PR middle-end/82063
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
7572         variables.
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.
7576         Initialize.
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
7589         PTRDIFF_MAX.
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
7592         for alloca(0).
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
7596         to HOST_WIDE_INT.
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
7600         ByteSize flags.
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
7635         two operands.
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.
7654         * json.h: 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>
7678         PR debug/86585
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
7697         globals.
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
7704         hash element.
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
7710         last_inserted_nary.
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
7715         last_inserted_phi.
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",
7743         "tree-pass.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
7766         need be.
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
7871         extension.
7872         * config/aarch64/aarch64.h (AARCH64_FL_PROFILE): New.
7873         * doc/invoke.texi (aarch64-feature-modifiers): New entry for profile
7874         extension.
7876 2018-07-19  Andre Vieira  <andre.simoesdiasvieira@arm.com>
7878         PR target/83009
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
7885         Umq with Umn.
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.
7899         (Umq): Remove.
7901 2018-07-19  Richard Biener  <rguenther@suse.de>
7903         * tree-ssa-sccvn.h (struct vn_phi_s): Make phiargs member
7904         a trailing array.
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>
7931         PR target/86560
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
7935         type attribute.
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>
7962         PR middle-end/86544
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
7974         ISA 2.07".
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
7984         EXACT_DIV_EXPR.
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>
8019         PR lto/86456
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
8024         with fat LTO part.
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
8035         alignment to 16.
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>
8042         PR middle-end/86542
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
8079         directly.
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
8083         align_flags class.
8084         * config/m68k/m68k.c: Do not use removed align_labels_value and
8085         align_loops_value.
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
8101         static function.
8102         * config/rx/rx.c (rx_align_for_label): Change return type
8103         to align_flags.
8104         (rx_max_skip_for_label): Remove TARGET_ASM_*_ALIGN_MAX_SKIP
8105         macro definitions.
8106         into align_flags class.
8107         (LABEL_ALIGN): Likewise.
8108         (LOOP_ALIGN): Likewise.
8109         * config/s390/s390.c (s390_label_align): Use align_flags
8110         class member.
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>
8187         PR middle-end/86539
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>
8223         PR middle-end/86528
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>
8236         PR target/86511
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
8259         CPUs.
8260         * config/s390/s390.opt: Add the new option.
8262 2018-07-16  Richard Biener  <rguenther@suse.de>
8264         PR lto/86523
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>
8270         PR ipa/86529
8271         * ipa-pure-const.c (malloc_candidate_p): Revert ::get
8272         to ::get_create.
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
8281         for ARCHS4x.
8282         * config/arc/arc-cpus.def (hs4x): New cpu.
8283         (hs4xd): Likewise.
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,
8289         fpu_cvt.
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>
8300         PR debug/86455
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
8306         early clobber.
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>
8337         PR target/85993
8338         * config/sh/sh.c (output_mi_thunk): Remove dead conditional
8339         block.
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
8352         copying section.
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>
8364         PR middle-end/78809
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):
8374         New option.
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>
8391         PR middle-end/85974
8392         * match.pd (addr1 - addr2): Allow either of the operand to
8393         have a conversion.
8395 2018-07-13  Tom de Vries  <tdevries@suse.de>
8397         * tree-inline.c (remap_ssa_name): Save and reuse debug exprs generated
8398         in remap_ssa_name.
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>
8412         PR debug/86452
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>
8419         PR middle-end/86489
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
8428         function.
8429         (riscv_get_interrupt_type): New function.
8430         (riscv_merge_decl_attributes): New function, checking interrupt type is
8431         same.
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
8437         directive.
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>
8449         PR c/86453
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
8456          parameter.
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
8479         internal functions.
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
8490         name change.
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
8521         internal functions.
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
8533         operands.
8534         (gimple_match_op::set_op): Likewise.
8535         (gimple_resimplify5): Declare.
8536         * genmatch.c (decision_tree::gen): Generate simplifications for
8537         5 operands.
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
8550         internal functions.
8551         * config/aarch64/aarch64.c (aarch64_preferred_else_value): Handle
8552         ternary operations.
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
8576         gimple_match_cond.
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
8592         constructor.
8593         * match.pd: Fold VEC_COND_EXPRs of an IFN_COND_* call to a new
8594         IFN_COND_* call.
8595         * config/aarch64/aarch64.c (aarch64_preferred_else_value): New
8596         function.
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
8602         DECL_FCONTEXT
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>
8611         PR debug/86462
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>
8617         PR target/84829
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>
8635         PR middle-end/86479
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
8643         here.
8644         (vn_nary_build_or_lookup_1): Adjust.
8645         (try_to_simplify): Likewise.
8646         * gimple-match-head.c (gimple_resimplify1): Instead apply one
8647         here.
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
8658         __U argument.
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
8672         __U argument.
8673         (_mm256_cmp_epi8_mask): Use __mmask32 return type instead of
8674         __mmask16.
8675         (_mm_mask_cmp_epu8_mask): Use __mmask16 instead of __mmask8 for __U
8676         argument.
8677         (_mm256_mask_cmp_epu8_mask): Use __mmask32 instead of __mmask16 for
8678         __U argument.
8679         (_mm256_cmp_epu8_mask): Use __mmask32 return type instead of
8680         __mmask16.
8681         (_mm_mask_cmp_epi16_mask): Cast last argument to __mmask8 instead
8682         of __mmask16.
8683         (_mm256_mask_cvtepi8_epi16): Use __mmask16 instead of __mmask32 for
8684         __U argument.
8685         (_mm_mask_cvtepi8_epi16): Use __mmask8 instead of __mmask32 for
8686         __U argument.
8687         (_mm256_mask_cvtepu8_epi16): Use __mmask16 instead of __mmask32 for
8688         __U argument.
8689         (_mm_mask_cvtepu8_epi16): Use __mmask8 instead of __mmask32 for
8690         __U argument.
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
8694         to __mmask32.
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
8698         to __mmask16.
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
8702         to __mmask32.
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
8706         to __mmask16.
8707         * config/i386/avx512vbmi2vlintrin.h (_mm_mask_shrdi_epi32,
8708         _mm_mask_shldi_epi32): Cast last argument to __mmask8 instead of
8709         __mmask16.
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
8715         for __U argument.
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
8725         alphabetical order.
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
8737         external decls.
8738         (write_ts_decl_non_common_tree_pointers): Do not stream
8739         DECL_ORIGINAL_TYPE
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>
8765         PR debug/86457
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
8770         type units.
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>
8780         PR debug/86459
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>
8790         PR c++/86443
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>
8805         PR middle-end/77357
8806         PR middle-end/86428
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.
8829         (MIN_SOB): Ditto.
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
8834         extend.
8835         (zero_extendhisi2): Remove.
8836         (shift patterns): Add CC handling variants.
8837         (bswaphi2): New.
8838         (bswapsi2): New.
8839         (rothi3): New.
8840         (define_peephole2): New peephole to recognize mov that sets CC for
8841         subsequent test.
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,
8847         wi::add.
8849 2018-07-09  Jakub Jelinek  <jakub@redhat.com>
8851         PR c/86420
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>
8860         PR debug/86413
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
8881         overflow_type.
8882         (combine_bound): Use wide-int overflow calculation instead of
8883         rolling our own.
8884         * calls.c (maybe_warn_alloc_args_overflow): Change overflow type to
8885         overflow_type.
8886         * fold-const.c (int_const_binop_2): Same.
8887         (extract_muldiv_1): Same.
8888         (fold_div_compare): Same.
8889         (fold_abs_const): Same.
8890         * match.pd: Same.
8891         * poly-int.h (add): Same.
8892         (sub): Same.
8893         (neg): Same.
8894         (mul): 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>
8944         PR target/86324
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>
8986         PR c++/86400
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>
9018         PR target/84711
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>
9024         PR target/84711
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>
9034         Revert
9035         2018-07-04  Maxim Ostapenko  <m.ostapenko@samsung.com>
9037         PR sanitizer/84250
9038         * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
9039         libasan.
9040         * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
9042 2018-07-04  Maxim Ostapenko  <m.ostapenko@samsung.com>
9044         PR sanitizer/84250
9045         * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
9046         libasan.
9047         * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
9049 2018-07-04  Eric Botcazou  <ebotcazou@adacore.com>
9051         PR middle-end/86380
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>
9067         PR middle-end/66240
9068         PR target/45996
9069         PR c/84100
9070         * common.opt: Rename align options with 'str_' prefix.
9071         * common/config/i386/i386-common.c (set_malign_value): New
9072         function.
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
9077         to string.
9078         * config/alpha/alpha.c (alpha_override_options_after_change):
9079         Likewise.
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
9092         string.
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
9111         type of constants.
9112         (mips_option_override): Set default values for options.
9113         * config/powerpcspe/powerpcspe.c (rs6000_option_override_internal):
9114         Likewise.
9115         * config/rs6000/rs6000.c (rs6000_option_override_internal):
9116         Likewise.
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
9122         macros.
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
9127         values for options.
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
9134         align_flags.
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
9155         topleev.h.
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
9159         of string types.
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
9164         value ranges.
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
9178         argument.
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>
9207         PR target/85620
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
9216         precondition.
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>
9259         PR target/85694
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
9268         UNSPEC_URHADD.
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>
9285         PR middle-end/86202
9286         * gimple-fold.c (size_must_be_zero_p): Check the type of the size.
9288 2018-07-03  Richard Biener  <rguenther@suse.de>
9290         PR ipa/86389
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
9322         directly.
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
9344         is null.
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
9374         functions.
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.
9410         Add cast_forwprop.
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
9505         in dump string.
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
9524         code...
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>
9534         PR ipa/86279
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>
9540         PR ipa/86323
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
9573         -mdec-asm.
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.
9595         (debug_path): New.
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.
9600         Add new argument.
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,
9616         split out from...
9617         (vect_recog_rotate_pattern): ...here.
9618         (vect_convert_input): Try to insert casts of invariants in the
9619         preheader.
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
9627         STMT_VINFO_VECTYPE.
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
9657         PLUS_EXPR operand.
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
9669         checks.
9670         (vect_init_pattern_stmt, vect_set_pattern_stmt): New functions,
9671         split out from...
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
9675         the original one.
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
9711         overload.
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>
9753         PR lto/85759
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
9784         AUTO_DUMP_SCOPE.
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>
9799         PR target/86348
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
9823         set_alt_dump_file.
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"
9829         global.
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
9871         clusters.
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
9876         clusters.
9877         (bit_test_cluster::is_beneficial): Bail out for such case.
9878         (switch_decision_tree::analyze_switch_statement):
9879         Fix comment.
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
9886         is set.
9887         * opt-suggestions.c (option_proposer::get_completions):
9888         New function.
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.
9958         (CCFP): Remove.
9959         * config/pdp11/pdp11-protos.h (int_no_side_effect_operand): New
9960         function.
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
9969         from output.
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.
9990         (reg_class): 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
10004         handling.
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
10015         exclusive options.
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
10042         patterns.
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
10051         after reload.
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
10059         reload.
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.
10074         (trfsr): Likewise.
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>
10136         PR target/85769
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
10142         comment.
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>
10159         PR target/86285
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
10177         location_t.
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
10182         "selftest.h".
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
10219         gimple stmt.
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
10225         return type.
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
10234         gimple stmt.
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
10245         location_t
10246         (compute_branch_probabilities): Likewise.
10247         * selftest-run-tests.c (selftest::run_tests): Call
10248         dumpfile_c_tests.
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):
10262         Likewise.
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
10267         vect_location.
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
10271         of vect_location.
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
10291         call_stmt.
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>
10300         PR target/86314
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
10303         operands.
10305 2018-06-26  Richard Biener  <rguenther@suse.de>
10307         PR tree-optimization/86287
10308         PR bootstrap/86316
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
10316         isn't valid.
10318 2018-06-26  Alexandre Oliva <aoliva@redhat.com>
10320         PR debug/86064
10321         * dwarf2out.c (loc_list_has_views): Adjust comments.
10322         (dw_loc_list): Split single cross-partition range with
10323         nonzero locview.
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
10331         test.
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
10336         (recipsf2_insn).
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):
10371         Likewise.
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
10435         datarefs lazily.
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
10442         transform.
10443         * tree-vect-stmts.c (ensure_base_align): Adjust for DR_AUX
10444         change.
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>
10473         PR debug/86257
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>
10479         PR C++/86082
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>
10514         PR target/86222
10515         * config/rs6000/rs6000-string.c (expand_strn_compare): Handle -m32
10516         correctly.
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):
10524         New declaration.
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>
10550         PR target/85994
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
10568         than once.
10570 2018-06-21  Eric Botcazou  <ebotcazou@adacore.com>
10572         * cfgrtl.c (fixup_reorder_chain): Do not emit NOPs in DECL_IGNORED_P
10573         functions.
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
10609         comment.
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
10616         deleting it.
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
10648         NULL DR_STEP.
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
10654         or scatters.
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
10683         enter/exit data.
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,
10688         FINALIZE}
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,
10692         FINALIZE}.
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,
10698         FINALIZE}.
10699         (omp_clause_code_name): Likewise.
10701 2018-06-20  Jakub Jelinek  <jakub@redhat.com>
10703         PR debug/86194
10704         * var-tracking.c (use_narrower_mode_test): Check if shift amount can
10705         be narrowed.
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
10735         128-bit modes.
10736         * config/aarch64/aarch64-simd.md (load_pair<VQ:mode><VQ2:mode>):
10737         New pattern.
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):
10749         New.
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
10753         hierarchy.
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.
10799         (PRINT_CASE): 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
10946         live or relevant.
10947         * tree-vect-loop.c (vect_transform_loop_stmt): New function,
10948         split out from...
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
10984         constant.
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
10990         blocks.
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
10997         get.
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
11017         of get_create.
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
11031         of get_create.
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>
11067         PR target/86197
11068         * config/rs6000/rs6000.md (rs6000_discover_homogeneous_aggregate): An
11069         ieee128 argument takes up only one (vector) register, not two (floating
11070         point) registers.
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"
11097         param.
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
11108         param.
11109         (vax_output_int_subtract): Likewise.
11110         * config/vax/vax.c (vax_output_int_add): Likewise for param
11111         "insn".
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
11123         a cast.
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>
11136         PR target/85358
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
11162         temporary.
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
11225         here to ...
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>
11235         PR 84195
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
11261         referenced VLAs.
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):
11325         Adjust callers.
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>
11375         PR 84195
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>
11401         PR target/85968
11402         * config/arc/arc.c (arc_return_address_register): Fix
11403         if-condition.
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.
11431         PR target/85945
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>
11444         PR ipa/86124
11445         * tree-ssa-struct-alias.c (create_variable_info_for): Handle
11446         NULL cgraph_node.
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
11476         of integer types.
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
11497         "label".
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
11505         rtx to rtx_insn *.
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
11510         braf_label.
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"
11516         to "rtx_insn *".
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"
11524         to "rtx_insn *".
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
11528         "rtx_insn *".
11529         (emit_insn_after_1): Strengthen "after" param from "rtx" to
11530         "rtx_insn *".
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
11548         "location_t".
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
11560         "rtx_insn *".
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
11570         "rtx_insn *".
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
11596         in.
11598 2018-06-13  Eric Botcazou  <ebotcazou@adacore.com>
11600         PR target/86048
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
11614         and size_reason.
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
11638         ARC cores.
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>
11647         PR target/86076
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
11661         are poly_ints.
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
11678         poly_int_rtx_p.
11679         (set_noop_p): Use rtx_to_poly_int64 for the elements selected
11680         by a VEC_SELECT.
11681         * simplify-rtx.c (avoid_constant_pool_reference): Use strip_offset.
11682         (simplify_binary_operation_1): Extend CONST_INT handling to
11683         poly_int_rtx_p.
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
11687         poly_int64.
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>
11701         PR other/69968
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
11710         "v_one_ago".
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>
11739         PR c/85931
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>
11749         PR target/85990
11750         * config/i386/gnu-user.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
11751         Update comments.
11752         * config/i386/gnu-user64.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
11753         Likewise.
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
11764         and report error.
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
11786         field.
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
11806         mips64r6.
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.
11856         <'o'>: Likewise.
11857         <'@'>: New case.
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>
11892         PR target/85755
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>
11915         PR target/85755
11916         * config/rs6000/rs6000.c (mem_operand_gpr): Enable PRE_INC and PRE_DEC
11917         addresses.
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
11926         TARGET_ELF.
11928 2018-06-08  Martin Liska  <mliska@suse.cz>
11930         * tree-cfg.h (debug_function): Fix argument type to match
11931         implementation.
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
11947         rather than 0.
11948         (optinfo_verbosity_options): Likewise.
11949         (optgroup_options): Convert to kv_pair<optgroup_flags_t>; use
11950         OPTGROUP_NONE.
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
11954         "dump_kind" param.
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
11963         optgroup_options.
11964         (opt_info_switch_p): Convert optgroup_flags from int to
11965         optgroup_flags_t.
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
11987         optgroup_flags_t.
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
11992         optgroup_flags_t.
11993         (opt_info_enable_passes): Likewise.
11994         * early-remat.c (early_remat::dump_edge_list): Use TDF_NONE rather
11995         than 0.
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.
12005         (debug): Likewise.
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
12010         than 0.
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
12059         unique number.
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
12095         register hooks.
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
12121         to ...
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
12139         of struct.
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
12158         instead.
12159         (pass_ipa_pure_const::pass_ipa_pure_const): Simplify
12160         constructor.
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
12195         HSA_INVALID.
12196         (hsa_function_summary::hsa_function_summary): Use the new enum
12197         value.
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
12209         of get.
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.
12279         (get): Likewise.
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
12286         declaration.
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.
12298         (get): 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.
12367         (VOID): Likewise.
12368         (PVOID): Likewise.
12369         (ULONG): Likewise.
12370         * config/i386/i386-builtin.def (BDESC_END): Likewise.
12371         (BDESC_FIRST): Likewise.
12372         (BDESC): 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
12608         PROCESSOR_TREMONT.
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
12666         specific.
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
12673         vec_xst.
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
12690         name.
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.
12695         PR target/85657
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
12704         error message.
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>
12728         PR target/63177
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>
12772         PR target/79924
12773         * config/aarch64/aarch64-protos.h (aarch64_err_no_fpadvsimd): Remove
12774         second argument.
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
12796         __typeof__.
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
12803         is extended.
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>
12839         PR target/81497
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
12847         void intrinsics.
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
12861         summary.
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
12871         gcov_ctr_summary.
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
12875         struct.
12876         * lto-streamer.h (struct GTY): Make profile_info gcov_summary
12877         struct.
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
12891         tbeginc builtin.
12893 2018-06-04  Jim Wilson  <jimw@sifive.com>
12895         * config/riscv/riscv-protos.h (riscv_expand_epilogue): Change bool arg
12896         to int.
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
12904         EXCEPTION_RETURN.
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>
12938         PR target/85832
12939         PR target/86036
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>
12970         PR target/86003
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, -].
12981         PR c++/86025
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
13023         TARGET_LINUX_ABI.
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
13031         spec content.
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
13048         optimization.
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
13052         optmization pass.
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.
13060         (epilogue): Ditto.
13061         (return): Ditto.
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
13102         mnemonics.
13103         (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
13104         mnemonics.
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
13110         variable.
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
13150         modify DR_STMT.
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>
13173         PR ipa/85960
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
13183         aliases.
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
13207         calls.
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>
13215         PR target/85984
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>
13228         PR c/82063
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>
13235         PR target/85829
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.
13251         (st1x2): Likewise.
13252         (st1x3): Likewise.
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
13372         in saphira.
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
13381         is in memory.
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
13391         link.
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>
13419         Revert:
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
13446         r=w alternative.
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
13461         division.
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>
13479         PR target/85950
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}
13482         sequence.
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>
13489         PR testsuite/85888
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
13507         env variables.
13509 2018-05-29  Jakub Jelinek  <jakub@redhat.com>
13511         * tree-cfg.c (verify_gimple_assign_unary): Add checking for
13512         VEC_UNPACK_*_EXPR.
13513         (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
13514         VEC_PACK_*_EXPR.
13516         PR target/85918
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
13536         optabs.
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
13542         type.
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
13550         FIX_TRUNC_EXPR.
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
13556         mode attributes.
13557         (vec_pack<floatprefix>_float_<mode>): New expander.
13558         (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
13559         attributes.
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}):
13565         Document.
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
13574         member.
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
13598         vector.
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
13616         and later.
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
13637         for armv5te.
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
13656         implementation.
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
13669         pattern.
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>
13691         PR target/85918
13692         * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
13693         attributes.
13694         * config/i386/sse.md
13695         (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
13696         Rename to ...
13697         (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
13698         ... this.
13699         (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
13700         Rename to ...
13701         (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
13702         ... this.
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
13710         to ...
13711         (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
13712         ... this.
13713         (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
13714         Rename to ...
13715         (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
13716         ... this.
13717         (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
13718         Rename to ...
13719         (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
13720         ... this.
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>
13736         PR target/85900
13737         PR target/85345
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.
13761         (GP_REGNUM): New.
13762         (riscv_frflags, riscv_fsflags): Use tab after opcode.
13763         (riscv_mret): New.
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>
13779         PR target/83628
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,
13793         defaulted to true.
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
13798         and pass it along.
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
13804         couldn't handle.
13805         (vect_analyze_data_refs): Do not give up on not analyzable
13806         DRs for BB vectorization.  Remove code truncating the dataref
13807         vector.
13809 2018-05-25  Jakub Jelinek  <jakub@redhat.com>
13811         PR target/85832
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
13820         parts.
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>
13879         PR c++/85912
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):
13897         New unspecs.
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
13902         UNSPEC_COND_DIV.
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):
13940         New unspecs.
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):
13947         Declare.
13948         * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
13949         function.
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.
13968         Rename to...
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
13974         SECTION_EXCLUDE.
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
13988         instead of 2.
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
14004         attribute.
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
14013         earlier operands.
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>
14034         PR target/85900
14035         PR target/85345
14036         * varasm.c (assemble_alias): Check ifunc_resolver only on
14037         FUNCTION_DECL.
14039 2018-05-24  Uros Bizjak  <ubizjak@gmail.com>
14041         PR target/85903
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
14048         global.
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>
14057         PR target/83009
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
14113         gimple_simplify.
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>
14134         PR target/78849
14135         * tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
14136         types.
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
14153         false.
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
14187         isa_bit_mode26.
14189 2018-05-23  Uros Bizjak  <ubizjak@gmail.com>
14191         * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
14192         New insn pattern.
14193         (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
14194         Rewrite expander pattern.  Emit gen_floatunssi<mode>2_i387_with_xmm
14195         for non-SSE modes.
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
14233         interpretation.
14234         (slsr_process_copy): Likewise.
14235         (dump_candidate): Dump first_interp field.
14236         (replace_mult_candidate): Process all interpretations, not just
14237         subsequent ones.
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):
14244         Add new boolean.
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>
14250         PR target/84882
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
14259         no-strict-align.
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
14281         constants.
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>
14290         PR target/85345
14291         * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
14292         attribute.
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
14304         attribute.
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
14331         code.
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):
14337         Define prototype.
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
14343         and ldp.
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):
14348         Delete.
14349         (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
14350         Delete.
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):
14355         Add size check.
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):
14366         New predicate.
14368 2018-05-22  Martin Sebor  <msebor@redhat.com>
14370         PR c/85623
14371         * calls.c (maybe_warn_nonstring_arg): Use string length to set
14372         or ajust the presumed bound on an operation to avoid unnecessary
14373         warnings.
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>
14394         PR ipa/85607
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>
14407         PR target/85657
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
14417         makean alias.
14418         * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
14419         KFmode.
14420         (IFKF_reg): New attributes to give the register constraints for
14421         IFmode and KFmode.
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>
14435         PR gcc/84923
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.
14444         (ABDL2): Likewise.
14445         (ADALP): Likewise.
14446         (sur): Add mappings for the above.
14447         * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
14448         New define_insn.
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
14464         eor3q<mode>4.
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,
14469         vbcaxq_s64): New.
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
14476         arc*-*-linux*.
14477         * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
14478         MULTILIB_DIRNAMES
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
14537         MASK_HW_ABS.
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
14596         UNSPEC_ADD32.
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
14603         code generation.
14604         * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
14605         optimization.
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
14610         predicate.
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
14655         model.
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>
14682         PR c++/82899
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>
14692         PR bootstrap/85838
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.
14698         (ARMv4): Update.
14699         (ARMv2, ARMv3, ARMv3m): Delete fgroups.
14700         (ARMv6m): Update.
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):
14705         Delete cpus.
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
14742         arm_arch5.
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):
14753         Likewise.
14754         * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
14755         arm_arch5e.
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
14784         unreachable path.
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
14798         the option.
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
14819         method.
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
14851         two instances.
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
14866         value.
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
14871         'limit_high'.
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.
14946         Add untyped.
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>
14985         PR target/85698
14986         * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
14987         operand.
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>
15036         PR target/85323
15037         * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
15038         even if the mask is not all ones.
15040         PR target/85323
15041         * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
15042         vector.
15043         (ix86_gimple_fold_builtin): Likewise.
15045         PR target/85323
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
15067         "goldmont-plus".
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
15093         asterisk to name.
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
15174         zero.
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
15211         limiting the walk.
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,
15218         split out from...
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>
15280         PR ipa/85734
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):
15292         New pattern.
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>
15313         PR target/85756
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
15328         -mcldemote.
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.
15338         (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
15371         pipeline.
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
15385         pipeline.
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
15398         pipeline.
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):
15413         New constraints.
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
15436         extension.
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
15451         subsection.
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):
15455         New subsubsection.
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>
15463         PR ipa/85655
15464         * ipa-cp.c (intersect_with_plats): Check that the lattice contains
15465         single const.
15467 2018-05-11  Richard Earnshaw  <rearnsha@arm.com>
15469         PR target/85733
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
15481         function type.
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>
15504         PR target/85606
15505         * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
15506         equivalent.
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>
15523         PR sanitizer/85556
15524         * doc/extend.texi: Document LLVM style format for no_sanitize
15525         attribute.
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>
15562         PR target/85693
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
15582         support.
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
15602         comment.
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):
15618         Adjust.
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>
15641         PR target/84797
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
15650         vector.
15651         (vect_bb_vectorization_profitable_p): Adjust.  Compute
15652         actual scalar cost using the cost vector and the add_stmt_cost
15653         machinery.
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,
15686         split out from...
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>
15694         PR target/85626
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>
15712         Revert:
15713         * doc/extend.texi (PowerPC Built-in Functions): Rename this
15714         subsection.
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):
15718         New subsubsection.
15719         (Basic PowerPC Built-in Functions Available on ISA 2.05): New
15720         subsubsection.
15721         (Basic PowerPC Built-in Functions Available on ISA 2.06): New
15722         subsubsection.
15723         (Basic PowerPC Built-in Functions Available on ISA 2.07): New
15724         subsubsection.
15725         (Basic PowerPC Built-in Functions Available on ISA 3.0): New
15726         subsubsection.
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
15737         subsection.
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):
15741         New subsubsection.
15742         (Basic PowerPC Built-in Functions Available on ISA 2.05): New
15743         subsubsection.
15744         (Basic PowerPC Built-in Functions Available on ISA 2.06): New
15745         subsubsection.
15746         (Basic PowerPC Built-in Functions Available on ISA 2.07): New
15747         subsubsection.
15748         (Basic PowerPC Built-in Functions Available on ISA 3.0): New
15749         subsubsection.
15751 2018-05-08  Jakub Jelinek  <jakub@redhat.com>
15753         PR target/85683
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>
15781         PR target/85572
15782         * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
15783         E_V4DImode.
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.
15788         PR target/85317
15789         * config/i386/i386.c (ix86_fold_builtin): Handle
15790         IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
15792         PR target/85480
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>
15798         PR target/85658
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
15824         to...
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
15838         functions.
15839         (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
15840         and UNORDERED.
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>
15863         PR bootstrap/85681
15864         Revert:
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
15876         to false.
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
15907         to 512.
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
15920         to false.
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>
15950         PR c++/85659
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
15953         BLKmode.
15955 2018-05-05  Roland McGrath  <mcgrathr@google.com>
15957         PR other/77609
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>
15970         PR target/85653
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
15996         Xilinx FP support.
15997         * config.gcc (powerpc-xilinx-eabi*): Remove.
15998         * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
15999         support.
16000         (fusion_addis_mem_combo_load): Ditto.
16001         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
16002         FP support.
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
16008         support.
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>
16053         PR libgomp/85639
16054         * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
16055         if ignore == 0.
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
16073         overflow.
16074         (fold_negate_expr_1): Likewise.
16076 2018-05-04  Jakub Jelinek  <jakub@redhat.com>
16078         PR libstdc++/85466
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>
16097         PR target/85530
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>
16103         PR testsuite/85106
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
16112         loops.
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>
16130         PR target/85582
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>
16140         PR libgomp/82428
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
16154         directly.
16156 2018-05-02  Tom de Vries  <tom@codesourcery.com>
16158         PR testsuite/85106
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>
16164         PR testsuite/85106
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
16176         float support.
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
16182         VECTOR_PAIRED.
16183         * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
16184         paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
16185         declarations.
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>
16207         PR target/85582
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
16218         not strided.
16220 2018-05-02  Tom de Vries  <tom@codesourcery.com>
16222         PR lto/85451
16223         * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
16224         error message.
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>
16233         PR lto/85451
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>
16239         PR other/83786
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.
16261         PR web/85578
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>
16267         PR c++/85523
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
16280         "val1" and "val2".
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
16300         declarations.
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
16306         of clobber.
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
16320         private.
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"
16332         param.
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
16402         pattern.
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):
16439         Add declaration.
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
16450         addresses cases.
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
16456         'r20' assignation.
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
16465         relative mode.
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
16468         flag_pic.
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
16483         redundant check.
16484         * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
16486 2018-04-30  Richard Biener  <rguenther@suse.de>
16488         PR bootstrap/85571
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>
16507         PR target/84431
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>
16521         PR ipa/85549
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
16528         function.
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
16539         fix for ILP32.
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>
16560         PR target/85512
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.
16606         (Yj): Remove.
16607         (Ym): Remove.
16608         (Yn): 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>
16622         PR target/84952
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>
16629         PR target/84025
16630         * config/nvptx/nvptx.c (needs_neutering_p): New function.
16631         (nvptx_single): Use needs_neutering_p to skip over insns that do not
16632         need neutering.
16634 2018-04-26  Richard Biener <rguenther@suse.de>
16635             Tom de Vries  <tom@codesourcery.com>
16637         PR lto/85422
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
16648         processing.
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
16679         shorter ones.
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>
16694         PR target/85473
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>
16735         PR target/85512
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>
16744         PR target/85485
16745         * common/config/i386/i386-common.c (ix86_handle_option): Don't
16746         handle OPT_mcet.
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>
16753         PR target/85485
16754         * doc/install.texi: Remove -mcet from bootstrap-cet.
16756 2018-04-24  Jakub Jelinek  <jakub@redhat.com>
16758         PR target/85511
16759         * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
16760         __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
16761         if TARGET_64BIT.
16763         PR target/85503
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>
16774         PR target/85508
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>
16786         PR target/85491
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>
16792         PR target/84828
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.
16814         (seti): Likewise.
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):
16847         Declare.
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
16881         unaligned access.
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>
16897         PR c/85365
16898         * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
16899         for null pointers.
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>
16905         PR target/85456
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>
16916         PR target/83402
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>
16923         PR target/85445
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
16928         calls.
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>
16934         PR target/85469
16935         * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
16936         Removed.
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
16940         __SHSTK__.
16941         * config/i386/driver-i386.c (host_detect_local_cpu): Remove
16942         has_ibt and ibt.
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.
16954         (mibt): Removed.
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>
16966         ipa/85449
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>
16972         ipa/85447
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
16979         by -O3.
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>
16990         PR target/85397
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>
16999         PR target/85404
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>
17005         PR target/85417
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
17011         TARGET_IBT.
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
17017         -fcf-protection.
17018         (incssp<mode>): Likewise.
17019         (nop_endbr): Likewise.
17020         * config/i386/i386.opt (mcet): Change help message to built-in
17021         functions only.
17022         (mibt): Likewise.
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
17039         and -mmvodir64b.
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
17047         and -mmovdiri.
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
17085         --enable-cet=auto.
17087 2018-04-18  Martin Liska  <mliska@suse.cz>
17089         * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
17090         style.
17092 2018-04-18  Martin Liska  <mliska@suse.cz>
17094         Revert
17095         2018-03-02  Eric Botcazou  <ebotcazou@adacore.com>
17097         PR ipa/83983
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>
17103         Revert
17104         2018-03-13  Eric Botcazou  <ebotcazou@adacore.com>
17106         PR lto/84805
17107         * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
17108         incomplete types.
17110 2018-04-18  H.J. Lu  <hongjiu.lu@intel.com>
17112         PR target/85388
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>
17118         PR jit/85384
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>
17126         PR target/81084
17127         * config.gcc: Obsolete powerpc*-*-*spe*.
17129 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
17131         PR debug/84637
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
17135         type.
17137 2018-04-17  Jim Wilson  <jimw@sifive.com>
17139         PR 84856
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.
17149         PR sanitizer/85230
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):
17160         New prototype.
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
17165         function.
17167 2018-04-17  Michael Meissner  <meissner@linux.vnet.ibm.com>
17169         PR target/85424
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"
17179         attributes.
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"
17186         attributes.
17187         (vec_extract_hi_v32hi): Merge all alternatives into one, use
17188         "sselog1" type instead of "sselog".  Remove explicit "memory"
17189         attribute.
17190         (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
17191         use "sselog1" type instead of "sselog".  Remove explicit "memory"
17192         attribute.
17193         (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
17194         "type", "prefix", "prefix_extra", "length_immediate" and "mode"
17195         attributes.
17196         (vec_extract_hi_v64qi): Merge all alternatives into one, use
17197         "sselog1" type instead of "sselog".  Remove explicit "memory"
17198         attribute.
17199         (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
17200         use "sselog1" type instead of "sselog".  Remove explicit "memory"
17201         attribute.
17203         PR target/85430
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
17208         on a SUBREG.
17210 2018-04-17  Martin Jambor  <mjambor@suse.cz>
17212         PR ipa/85421
17213         * ipa-cp.c (create_specialized_node): Call
17214         expand_all_artificial_thunks if necessary.
17216 2018-04-17  Martin Liska  <mliska@suse.cz>
17218         PR lto/85405
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>
17224         PR target/85281
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
17236         -masm=intel.
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
17245         -masm=intel.
17246         (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
17247         -masm=intel.
17248         (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
17249         -masm=intel.
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
17252         %g1.
17253         (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
17254         -masm=intel.
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>
17272         PR lto/85405
17273         * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
17275 2018-04-17  Martin Liska  <mliska@suse.cz>
17277         PR ipa/85329
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
17282         string.
17283         (expand_target_clones): Handle it.
17284         (ipa_target_clone): Make redirection just for target_clones
17285         functions.
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>
17296         PR target/83660
17297         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
17298         vec_extract expression as having side effects to make sure it gets
17299         a cleanup point.
17301 2018-04-16  H.J. Lu  <hongjiu.lu@intel.com>
17303         PR target/85403
17304         * config/i386/i386.c (get_builtin_code_for_version): Check
17305         error_mark_node.
17307 2018-04-16  Olga Makhotina  <olga.makhotina@intel.com>
17309         PR target/84331
17310         * config.gcc: Support "skylake".
17311         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
17312         PROCESSOR_SKYLAKE.
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>
17327         PR c++/85112
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>
17334         PR target/84413
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>
17340         PR target/85293
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>
17347         PR target/83402
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
17357         progress.
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
17367         basic block.
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>
17387         PR lto/71991
17388         * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
17389         always inline.
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>
17403         PR target/85291
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>
17411         PR debug/83157
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>
17431         PR target/85238
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
17448         streaming out.
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>
17454         PR lto/85371
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
17458         split-dwarf.
17459         (dwarf2out_early_finish): Likewise.
17461 2018-04-12  Jakub Jelinek  <jakub@redhat.com>
17463         PR target/85328
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>
17490         PR target/85296
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>
17499         PR target/85321
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>
17524         PR target/84301
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>
17536         PR ipa/84149
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>
17556         PR lto/85339
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>
17564         PR target/85281
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>
17577         PR debug/85302
17578         * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
17579         SIZEP is NULL.
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
17583         call size_of_locs.
17585 2018-04-11  Thomas Preud'homme  <thomas.preudhomme@arm.com>
17587         PR target/85261
17588         * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
17589         into register.
17591 2018-04-10  Aaron Sawdey  <acsawdey@linux.ibm.com>
17593         PR target/85321
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>
17600         PR target/85287
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
17603         a register.
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>
17612         PR target/85321
17613         * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
17614         the help text.
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>
17660         PR target/85177
17661         PR target/85255
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>
17680         PR target/85173
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>
17687         PR lto/85078
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;
17691         walk basetypes.
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
17697         options.
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>
17704         PR rtl/84058
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
17742         shifter option.
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>
17753         PR target/84041
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
17795         into tm_file.
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
17820         pipeline.
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
17833         pipeline.
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
17846         pipeline.
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
17862         pipeline.
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
17883         align case.
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
17956         and memory sync.
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
17981         macros.
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>
18025         PR target/85196
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
18054         file.
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>
18067         PR debug/85252
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>
18091         PR sanitizer/85213
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
18094         recursive calls.
18095         (fold_comparison): Adjust twoval_comparison_p caller, don't handle
18096         save_p here.
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
18104         recursion.
18106 2018-04-06  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
18108         * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
18109         instructions.
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
18134         patterns.
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
18139         ARCH_V3S.
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
18149         configuration.
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>
18164         PR target/85193
18165         * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
18167 2018-04-05  Tom de Vries  <tom@codesourcery.com>
18169         PR target/85204
18170         * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
18171         cond jump.
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
18188         operands.
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>
18245         PR target/85203
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>
18264         PR lto/85176
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
18304         patterns.
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
18351         than vector.
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>
18366         PR target/85169
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>
18385         PR ipa/84947
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
18411         example.
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
18460         of V.
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>
18466         PR target/83315
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>
18472         PR target/80546
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>
18484         PR target/83822
18485         * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
18486         condition.
18487         * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
18488         condition.
18490 2018-03-30  Julia Koval  <julia.koval@intel.com>
18492         PR target/84413
18493         * x86-tune.def (movx, partial_reg_dependency): Enable for
18494         m_SKYLAKE_AVX512.
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>
18504         PR lto/84995.
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>.
18513         PR c/85094
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
18517         checking.
18519 2018-03-28  Peter Bergner  <bergner@vnet.ibm.com>
18521         PR target/84912
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
18529         definition.
18530         (DIVDE): Use it.
18531         (DIVDEU): Likewise.
18533 2018-03-28  Carl Love  <cel@us.ibm.com>
18535         Revert
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
18541         fctiw instruction.
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>
18552         PR target/84912
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
18563         builtin function.
18564         (__builtin_divweuo): Likewise.
18565         (__builtin_divdeo): Likewise.
18566         (__builtin_divdeuo): Likewise.
18568 2018-03-28  Jakub Jelinek  <jakub@redhat.com>
18570         PR target/85095
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>
18581         PR other/84819
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>
18590         PR sanitizer/85081
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>
18598         PR target/84988
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>
18608         PR target/84914
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>
18616         PR target/85044
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>
18623         PR target/81863
18624         * config/arm/arm.c (arm_valid_symbolic_address): Handle
18625         arm_word_relocations.
18627 2018-03-27  Cesar Philippidis  <cesar@codesourcery.com>
18629         PR target/85056
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):
18643         Valueize the VUSE.
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>
18653         PR target/85073
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
18661         of ...
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>
18691         PR target/85026
18692         * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
18693         Clean up attributes.
18695 2018-03-23  Richard Biener  <rguenther@suse.de>
18697         PR debug/85020
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.
18711         PR sanitizer/85029
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):
18721         Remove prototype.
18722         * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
18723         function.
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>
18732         PR target/84826
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>
18742         PR target/84760
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>
18769         PR sanitizer/85018
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
18798         subtraction.
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>
18814         PR jit/84288
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>
18843         PR ipa/84963
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>
18849         PR target/84988
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>
18855         PR target/84838
18856         * Minor grammar fixes for x86 options.
18858 2018-03-20  Jakub Jelinek  <jakub@redhat.com>
18860         PR debug/84875
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>
18866         PR target/83789
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>
18886         PR target/82518
18887         * config/arm/arm.c (arm_array_mode_supported_p): Return false for
18888         BYTES_BIG_ENDIAN.
18890 2018-03-20  Richard Biener  <rguenther@suse.de>
18892         PR target/84986
18893         * config/i386/i386.c (ix86_add_stmt_cost): Only cost
18894         sign-conversions as zero, fall back to standard scalar_stmt
18895         cost for the rest.
18897 2018-03-20  Martin Liska  <mliska@suse.cz>
18899         PR ipa/84825
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>
18905         PR target/84990
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.
18911         PR target/84845
18912         * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
18913         to ...
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]
18916         itself.
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>
18926         PR target/82989
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>
18933         PR target/84952
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>
18939         PR target/84954
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>
18945         PR target/84945
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.
18950         PR c/84953
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.
18960         PR sanitizer/78651
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>
18970         PR sanitizer/78651
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>
18976         PR target/81647
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>
18982         PR bootstrap/84856
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
18993         STACK_BOUNDARY.
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>
19017         PR target/84711
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
19024         matches >= 0.
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>
19064         PR target/84902
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>
19078         PR target/84899
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.
19083         PR c/84909
19084         * hsa-gen.c (mem_type_for_type): Fix comment typo.
19085         * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
19086         Likewise.
19087         * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
19088         Likewise.
19090 2018-03-16  Vladimir Makarov  <vmakarov@redhat.com>
19092         PR target/84876
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>
19100         PR ipa/84833
19101         * multiple_target.c (create_dispatcher_calls): Redirect
19102         reference in the symbol table.
19104 2018-03-16  Martin Liska  <mliska@suse.cz>
19106         PR ipa/84722
19107         * multiple_target.c (create_dispatcher_calls): Redirect also
19108         an alias.
19110 2018-03-16  Jakub Jelinek  <jakub@redhat.com>
19112         PR c++/79937
19113         PR c++/82410
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
19127         1 << 3.
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
19136         split predicate.
19138 2018-03-15  Jakub Jelinek  <jakub@redhat.com>
19140         PR c++/79085
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>
19146         PR target/84574
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
19154         function return.
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>
19172         PR c/84852
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>
19183         PR target/84711
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>
19189         PR target/68256
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>
19196         PR target/84860
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>
19207         PR target/83451
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
19219         UNSPEC_VPERMXOR.
19220         * config/doc/extend.texi: Add prototypes for vec_permxor.
19222 2018-03-14  David Malcolm  <dmalcolm@redhat.com>
19224         PR c/84852
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
19231         linenum_type.
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.
19274         Revert
19275         * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
19276         member.
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.
19283         (icelake): Remove.
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
19291         icelake server.
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>
19301         PR sanitizer/83392
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.
19306         PR target/84844
19307         Revert
19308         2017-04-20  Uros Bizjak  <ubizjak@gmail.com>
19310         PR target/78090
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>
19324         PR target/83712
19325         * lra-assigns.c (find_all_spills_for): Ignore uninteresting
19326         pseudos.
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
19330         necessary.
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
19342         live ranges.
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>
19353         PR target/84743
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
19361         for big-endian.
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
19371         unsigned packs.
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
19398         define_expand.
19399         (*<su>mul<mode>3_highpart): New define_insn.
19401 2018-03-13  Eric Botcazou  <ebotcazou@adacore.com>
19403         PR lto/84805
19404         * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
19405         incomplete types.
19407 2018-03-13  Martin Liska  <mliska@suse.cz>
19409         PR ipa/84658.
19410         * (sem_item_optimizer::sem_item_optimizer): Initialize new
19411         vector.
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
19428         that.
19430         PR target/84827
19431         * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
19432         pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
19434         PR target/84828
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.
19439         PR target/84786
19440         * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
19441         on the last operand.
19443         PR c++/84704
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>
19514         PR debug/58150
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
19533         function.
19534         * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
19535         * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
19536         definition.
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
19542         function.
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>
19563         Reverting patch:
19564         2018-03-09  Vladimir Makarov  <vmakarov@redhat.com>
19566         PR target/83712
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
19570         reg class.
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
19582         live ranges.
19584 2018-03-10  H.J. Lu  <hongjiu.lu@intel.com>
19586         PR target/84807
19587         * config/i386/i386.opt: Replace Enforcment with Enforcement.
19589 2018-03-10  Alexandre Oliva  <aoliva@redhat.com>
19591         PR debug/84620
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
19607         zero_view_p.
19609 2018-03-09  Peter Bergner  <bergner@vnet.ibm.com>
19611         PR target/83969
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>
19620         PR target/84772
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.
19625         PR c++/84767
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):
19633         Remove dead code.
19634         (builtin_access::generic_overlap): Be prepared to handle non-array
19635         base objects.
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.
19646         Adjust callers.
19647         * recog.c (preprocess_constraints): Test address_operand for
19648         CT_ADDRESS constraints.
19650 2018-03-09  Vladimir Makarov  <vmakarov@redhat.com>
19652         PR target/83712
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
19656         reg class.
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
19668         live ranges.
19670 2018-03-09  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
19672         PR target/83193
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>
19684         PR jit/64089
19685         PR jit/84288
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.
19697         Revert
19698         PR tree-optimization/84178
19699         * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
19700         to caller.
19701         (version_loop_for_if_conversion): Delay update_ssa call.
19702         (tree_if_conversion): Delay update_ssa until after predicate
19703         insertion.
19705 2018-03-09  Eric Botcazou  <ebotcazou@adacore.com>
19707         PR target/84763
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>
19713         PR debug/84456
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>
19723         PR target/84748
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
19732         for update.
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
19738         to caller.
19739         (version_loop_for_if_conversion): Delay update_ssa call.
19740         (tree_if_conversion): Delay update_ssa until after predicate
19741         insertion.
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>
19770         PR debug/84404
19771         PR debug/84408
19772         * dwarf2out.c (struct dw_line_info_table): Update comments for
19773         view == -1.
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>
19804         PR target/82411
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>
19819         PR target/84277
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>
19847         PR fortran/84565
19848         * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
19849         aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
19851         PR c++/84704
19852         * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
19853         on tmp_var.
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
19865         sprint_ul.
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>
19872         PR target/84710
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):
19883         Remove.
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
19906         assertion failure.
19908         PR tree-optimization/84687
19909         * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
19910         on new_node->decl.
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>
19922         PR target/84700
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
19925         are equal to x.
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>
19934         PR target/84564
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.
19939         PR target/84524
19940         * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
19941         orig,vex.
19942         (*<plusminus_insn><mode>3): Likewise.  Remove <mask_operand3> uses.
19944 2018-03-05  Peter Bergner  <bergner@vnet.ibm.com>
19946         PR target/84264
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
19971         and -mwbnoinvd.
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
19989         wbnoinvdintrin.h.
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
19996         member.
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
20001         once.
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
20067         implementation.
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
20107         predicate.
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
20146         argument.
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,
20153         use ngettext.
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>
20172         PR ipa/84628
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.
20175         Formatting fixes.
20177 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
20179         PR target/56540
20180         * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
20181         __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
20183         PR target/56540
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>
20192         PR ipa/83983
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>
20211         PR target/84614
20212         * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
20213         prototypes.
20214         * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
20215         comments.
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
20220         prev_real_insn.
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
20225         zero vector.
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
20232         exprs per value.
20233         (clean): Likewise.
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>
20253         PR target/82005
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>
20260         PR sanitizer/82484
20261         * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
20262         volatile arguments.
20264 2018-03-01  Richard Biener  <rguenther@suse.de>
20266         PR debug/84645
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
20273         XINT with INTVAL.
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
20307         where possible.
20309         PR target/52991
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)
20314         as the alignment.
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>
20334         PR testsuite/84597
20335         * timevar.c (timer::print): Fix format to properly print 100%
20336         values.
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
20353         reg only.
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>
20359         PR c++/83871
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>
20414         PR target/84039
20415         * config/i386/constraints.md (Bs): Replace
20416         ix86_indirect_branch_register with
20417         TARGET_INDIRECT_BRANCH_REGISTER.
20418         (Bw): Likewise.
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>
20457         PR debug/84545
20458         * final.c (rest_of_clean_state): Also look for calls inside sequences.
20460 2018-02-26  H.J. Lu  <hongjiu.lu@intel.com>
20462         PR target/84530
20463         * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
20464         the bool argument.
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>
20487         PR bootstrap/84405
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>
20497         PR target/84521
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
20506         volatile memory.
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
20534         BU_P8V_OVERLOAD_2.
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>
20541         PR target/81572
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
20544         LRA_UNKNOWN_ALT.
20545         * lra-constraints.c (curr_insn_transform): Set up
20546         LRA_NON_CLOBBERED_ALT for moves processed on the fast path.  Use
20547         LRA_UNKNOWN_ALT.
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>
20567         PR target/84176
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.
20577         (): 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>
20589         PR target/83335
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>
20610         PR driver/83193
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>
20616         PR driver/83193
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>
20622         PR target/84502
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
20633         fails.
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>
20669         PR c/84229
20670         * ipa-cp.c (determine_versionability): Do not version functions caling
20671         va_arg_pack.
20673 2018-02-21  Martin Liska  <mliska@suse.cz>
20675         PR driver/83193
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>
20683         PR driver/83193
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
20694         false.
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
20701         false.
20702         (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
20703         (gimple_fold_builtin_strlen): Pass true as last argument to
20704         get_range_strlen.
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
20750         if not optimizing.
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
20759         different.
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>
20769         PR driver/83193
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>
20775         PR c/84310
20776         PR target/79747
20777         * final.c (shorten_branches): Build align_tab array with one
20778         more element.
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>
20787         PR target/84146
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
20799         on call_insn.
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
20818         reg note.
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):
20823         Likewise.
20824         * jump.c (cleanup_barriers, delete_related_insns): Likewise.
20825         * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
20827         PR c++/84444
20828         * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
20829         is ADDR_EXPR.
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
20834         is NULL.
20836 2018-02-19  Martin Liska  <mliska@suse.cz>
20838         PR sanitizer/82183
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
20846         to prevent ubsan.
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
20852         BU_P8V_OVERLOAD_1.
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>
20862         PR other/80589
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
20870         e500 family CPUs.
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>
20879         PR ipa/84425
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
20913         vec_insert4b.
20915 2018-02-16  Carl Love  <cel@us.ibm.com>
20917         * config/rs6000/altivec.h: Add builtin names vec_extract4b
20918         vec_insert4b.
20919         * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
20920         definitions.
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>
20943         PR target/83831
20944         * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
20945         to QImode.
20947 2018-02-16  Richard Biener  <rguenther@suse.de>
20949         PR tree-optimization/84037
20950         PR tree-optimization/84016
20951         PR target/82862
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
20956         elementwise loads.
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>
20967         PR sanitizer/84307
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>
20978         PR target/84272
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.
20985         PR bootstrap/84405
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
20991         is defined.
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
20997         recursive calls.
20998         (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
20999         callers.
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
21055         already been set.
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>
21061         PR c/84108
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>
21067         PR target/83984
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>
21079         PR target/84220
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>
21086         PR target/84239
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>
21108         PR target/83831
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
21111         declarations.
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
21116         functions.
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>
21126         PR target/79242
21127         * machmode.def: Define a complex mode for PARTIAL_INT.
21128         * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
21129         MODE_PARTIAL_INT.
21130         * doc/rtl.texi: Document CSPImode.
21131         * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
21132         handling.
21133         (msp430_hard_regno_nregs_with_padding): Likewise.
21135 2018-02-13  Peter Bergner  <bergner@vnet.ibm.com>
21137         PR target/84279
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>
21151         PR c/82210
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
21166         ARGS as unused.
21168 2018-02-13  Alexandre Oliva  <aoliva@redhat.com>
21170         PR debug/84342
21171         PR debug/84319
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
21204         disabled.
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>
21217         PR sanitizer/84340
21218         * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
21220 2018-02-13  Martin Jambor  <mjambor@suse.cz>
21222         PR c++/83990
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>
21242         PR target/84335
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.
21255         Formatting fixes.
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
21261         inline function.
21262         * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
21263         inline function.
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.
21273         PR target/84336
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>
21279         PR target/83760
21280         * config/sh/sh.c (find_barrier): Consider a sibling call
21281         a barrier as well.
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
21294         for all instances.
21295         * tree-vect-stmts.c (vect_model_simple_cost): Make early
21296         out an assert.
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
21300         SLP stmts.
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>
21313         PR sanitizer/84307
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
21326         default values.
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>
21385         PR target/84266
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
21403         function label.
21405 2018-02-10  Alan Modra  <amodra@gmail.com>
21407         PR target/84300
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>
21413         PR sanitizer/83987
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>
21436         PR target/84226
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>
21453         PR target/83926
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
21458         mode.
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>
21469         PR lto/84213
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>
21482         PR target/82641
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>
21488         PR target/PR84295
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>
21502         PR sanitizer/84285
21503         * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
21504         STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
21505         -static-lib*san.
21507         PR debug/84252
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
21514         TREE_READONLY bit.
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>
21521         PR c++/83659
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
21529         markers.
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
21551         markers.
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
21642         needed.
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>
21679         PR target/83008
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
21693         get_range_strlen.
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>
21716         PR ipa/81360
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):
21736         Define variables.
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
21760         macros.
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
21775         automatically.
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
21782         line thunks.
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
21794         conversion.
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
21799         option values.
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>
21846         PR target/84113
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
21855         options.
21857         PR target/84154
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
21893         [a-zA-Z].
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
21904         ivexpr_map.
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>
21921         PR target/84248
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>
21927         PR libgomp/84217
21928         * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
21929         enough.
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
21935         this place.
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>
21948         PR target/84209
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>
21964         PR target/84146
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>
21979         PR target/84154
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
22003         Undocumented.
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>
22014         PR target/84145
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>
22026         PR target/82641
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
22045         inlining context.
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
22095         options.
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>
22129         PR target/84066
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>
22135         PR target/56010
22136         PR target/83743
22137         * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
22138         #include "opts.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>
22147         PR target/84089
22148         * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
22150 2018-02-01  Jeff Law  <law@redhat.com>
22152         PR target/84128
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):
22174         ...here.
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
22181         cannot occur.
22183 2018-02-01  Renlin Li  <renlin.li@arm.com>
22185         PR target/83370
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
22203         in RPO order.
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
22213         predicate.
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
22219         functions.
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,
22232         not just bytes.
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>
22245         PR tearget/83845
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>
22270         PR target/82444
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>
22288         PR target/83618
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>
22295         PR lto/84105
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>
22301         Revert
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
22330         accesses.
22331         (arc_in_small_data_p): Do not consider for small data the uncached
22332         types.
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>
22341         PR c/84100
22342         * common.opt (falign-functions=, falign-jumps=, falign-labels=,
22343         falign-loops=): Add Optimization flag.
22345 2018-01-30  Jeff Law  <law@redhat.com>
22347         PR target/84064
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>
22357         PR target/40411
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>
22363         PR target/84112
22364         * lra-constraints.c (curr_insn_transform): Process AND in the
22365         address.
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
22378         copies.
22380 2018-01-30  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
22382         PR target/83758
22383         * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
22384         a reg rtx.
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
22392         SSA form.
22394 2018-01-30  Jan Hubicka  <hubicka@ucw.cz>
22396         PR ipa/81360
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
22409         zeros correctly.
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>
22420         PR bootstrap/84017
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
22427         pattern.
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
22439         of usage.
22441 2018-01-29  Michael Meissner  <meissner@linux.vnet.ibm.com>
22443         PR target/81550
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>
22454         PR bootstrap/80867
22455         * tree-vect-stmts.c (vectorizable_call): Don't call
22456         targetm.vectorize_builtin_md_vectorized_function if callee is
22457         NULL.
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
22494         debug insns.
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
22501         specified.
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>
22518         PR target/81763
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.
22536         PR c/83989
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
22547         value.
22548         (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
22549         support.
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
22565         architectures.
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
22589         secure.
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
22613         operand.
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.
22636         (normw): Likewise.
22637         (swap): Likewise.
22638         (divaw): Likewise.
22639         (flag): Likewise.
22640         (sr): Likewise.
22641         (kflag): Likewise.
22642         (ffs): Likewise.
22643         (ffs_f): Likewise.
22644         (fls): 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
22650         operand.
22651         (subsf3_fpx): Likewise.
22652         (mulsf3_fpx): Likewise.
22653         * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
22654         asm operand.
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>
22696         PR target/84033
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
22708         node->count.
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
22716         attribute.
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>
22723         PR target/84028
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>
22729         PR target/68467
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>
22735         PR target/83994
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
22741         assertion.
22743 2018-01-24  Vladimir Makarov  <vmakarov@redhat.com>
22745         PR target/84014
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
22753         here.
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
22769         with zero counts.
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>
22782         PR target/83589
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>
22792         PR target/81352
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) .. }
22801         operands.
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
22821         member function.
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>
22848         PR target/83905
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
22910         "reachability".
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.
22914         * tree-vrp.c
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
22918         REACHABLE_BLOCKS.
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
22925         coverage.
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>
22934         PR testsuite/83888
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>
22968         PR c++/83918
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
22989         decl.
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
23000         declaration.
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>
23006         PR target/83862
23007         * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
23008         no longer used.
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>
23078         PR gcc/77734
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>
23086         PR target/83946
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>
23102         PR target/80870
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.
23135         PR target/83930
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>
23141         PR debug/81570
23142         PR debug/83728
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):
23153         Likewise.
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>
23166         PR target/83920
23167         * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
23169 2018-01-19  Cesar Philippidis  <cesar@codesourcery.com>
23171         PR target/83790
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
23203         suboption.
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
23225         instead of NEG.
23227 2018-01-18  Jakub Jelinek  <jakub@redhat.com>
23229         PR sanitizer/81715
23230         PR testsuite/83882
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>
23241         PR target/82964
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>
23247         PR ipa/82256
23248         patch by PaX Team
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>
23258         PR ipa/83619
23259         * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
23260         frequencies.
23262 2018-01-18  Boris Kolpackov  <boris@codesynthesis.com>
23264         PR other/70268
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
23280         remap_filename().
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
23295         lfence is emitted.
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>
23310         PR c/61240
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>
23337         PR ipa/83051
23338         * ipa-inline.c (flatten_function): Do not overwrite final inlining
23339         failure.
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>
23357         PR lto/83121
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):
23370         Add declaration.
23371         * config/aarch64/constraints.md (aarch64_movti_operand):
23372         Limit immediates.
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
23380         lxvll.
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>
23404         PR target/83546
23405         * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
23406         to PTA_SILVERMONT.
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
23421         multilib options.
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
23429         copies.
23431         * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
23432         64 bits.
23433         * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
23434         128 bits.
23436         * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
23437         variables.
23439         * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
23440         return value.
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
23483         option.
23485 2018-01-16  Artyom Skrobov tyomitch@gmail.com
23487         * caller-save.c (insert_save): Drop unnecessary parameter.  All
23488         callers updated.
23490 2018-01-16  Jakub Jelinek  <jakub@redhat.com>
23491             Richard Biener  <rguenth@suse.de>
23493         PR libgomp/83590
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>
23516         PR c/83844
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
23522         into MULT_EXPR.
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>
23530         PR target/83858
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
23567         !TARGET_MUL.
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.
23615         (Bw): Likewise.
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.
23631         (Bw): Likewise.
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>
23654         PR target/83839
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>
23666         PR target/83687
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>
23690         PR lto/83804
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>
23707         PR other/83508
23708         * builtins.c (check_access): Avoid warning when the no-warning bit
23709         is set.
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>
23718         PR target/83013
23719         * config/microblaze/microblaze.c (microblaze_asm_output_ident):
23720         Use .pushsection/.popsection.
23722 2018-01-14  Martin Sebor  <msebor@redhat.com>
23724         PR c++/81327
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.
23755         (Bw): Likewise.
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
23780         return.
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
23823         needed.
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
23829         to true.
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.
23836         (keep): Likewise.
23837         (thunk): Likewise.
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>
23845         PR ipa/83051
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
23853         can_inline_edge_p.
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):
23859         New function.
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
23906         access size.
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
23911         tree-ssanames.h.
23912         (runtime_alias_check_p): Allow runtime alias checks with
23913         variable strides.
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
23918         segment lengths.
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
23928         constructor.
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
23947         function.
23948         (vect_loop_versioning): Call it.
23949         * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
23950         when retrying.
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
23959         optabs.
23960         * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
23961         Document.
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
23967         functions.
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
24009         function.
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
24026         null target hooks.
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
24050         functions.
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
24090         functions.
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
24145         accordingly.
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
24171         function.
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
24197         not a power of 2.
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,
24221         split out from...
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
24226         pattern.
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):
24257         New function.
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
24282         mask_skip_niters.
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
24290         the loop bounds.
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,
24301         split out from...
24302         (vect_analyze_loop_2): ...here. Don't check the vectorization
24303         factor against the number of loop iterations if the loop is
24304         fully-masked.
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
24316         mem_type field.
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
24346         TARGET_MEM_REFs.
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
24373         than assignments.
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
24424         loops.
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
24453         booleans.
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
24566         parameter.
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
24593         modes for SVE.
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>):
24608         New patterns.
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
24630         type sizes.
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
24654         enum.
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
24687         left one place.
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
24704         SVE macros.
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
24726         for predicates.
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
24731         functions.
24732         (aarch64_add_offset): Add a temp2 parameter.  Assert that temp1
24733         does not overlap dest if the function is frame-related.  Handle
24734         SVE constants.
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
24747         functions above.
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
24780         data modes.
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
24798         SVE modes.
24799         (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
24800         if SVE is enabled.
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
24819         the vector size.
24820         (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
24821         (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
24822         functions.
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
24828         for SVE modes.
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
24834         MAX_VECT_LEN.
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
24859         constraints.
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
24873         mode attributes.
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
24881         attributes.
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
24923         TARGET_SVE.
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>
24959         PR c++/83778
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
24979         parts.  Simplify.
24980         (load_multiple): Delete.
24981         (*ldmsi8): Delete.
24982         (*ldmsi7): Delete.
24983         (*ldmsi6): Delete.
24984         (*ldmsi5): Delete.
24985         (*ldmsi4): Delete.
24986         (*ldmsi3): Delete.
24987         (store_multiple): Delete.
24988         (*stmsi8): Delete.
24989         (*stmsi7): Delete.
24990         (*stmsi6): Delete.
24991         (*stmsi5): Delete.
24992         (*stmsi4): Delete.
24993         (*stmsi3): 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.
25013         PR debug/81155
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>
25019         PR target/83737
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>
25032         PR target/83628
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
25045         INTEGER_CSTs.
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
25080         and pass the info.
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>
25089         PR target/81616
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
25103         bits.
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>
25112         PR debug/83157
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>
25118         PR target/81819
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
25130         parts.
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>
25141         PR target/83368
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>
25176         PR target/40411
25177         * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
25178         -symbolic.
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>
25184         PR ipa/83054
25185         * ipa-devirt.c (final_warning_record::grow_type_warnings):
25186         New function.
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):
25205         Likewise.
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
25212         PowerPC systems.
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
25227         enum value.
25228         (speculation_barrier): New define_insn.
25229         * doc/extend.texi: Document __builtin_speculation_barrier.
25231 2018-01-11  Jakub Jelinek  <jakub@redhat.com>
25233         PR target/83203
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
25237         iterators.
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>
25247         PR target/83330
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>
25253         PR target/82682
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>):
25293         New define_expand.
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
25312         when appropriate.
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
25328         define_expand.
25329         (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
25330         vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
25331         New define_insn.
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
25336         documentation.
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>
25354         PR target/81821
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
25370         field.
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
25389         mode.
25390         (aarch64_builtin_vectorized_function): Check for a constant number
25391         of units.
25392         * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
25393         GET_MODE_SIZE.
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
25434         GET_MODE_NUNITS.
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
25448         than an int.
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>
25501         PR target/83514
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>
25512         PR target/82096
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
25527         profile.
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
25536         partition fixup.
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.
25549         (f16quad): Ditto.
25550         (f16mac1): 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):
25752         New flags.
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
25761         on TARGET_SHA2.
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
25776         as string arrays.
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
25797         function.
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
25819         string.
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
25831         empty blocks.
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>
25839         PR target/83399
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>
25862         PR c++/43486
25863         * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
25864         "public_flag".
25865         * tree.c (tree_nop_conversion): Return true for location wrapper
25866         nodes.
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>
25879         PR target/83735
25880         * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
25881         stack_realign_offset for the largest alignment of stack slot
25882         actually used.
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
25893         account.
25895 2018-01-10  Segher Boessenkool  <segher@kernel.crashing.org>
25897         PR target/83629
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>
25903         PR debug/83765
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
25906         origin.
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>
25916         PR target/81616
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>
25923         PR bootstrap/82831
25924         * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
25925         * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
25926         partitioning.
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.
25939         Revert:
25940         * common.md (E, F): Use CONSTANT_P instead of checking for
25941         CONST_VECTOR.
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,
25954         V4SI, V4SF types.
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
25964         function.
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>
25970         PR target/83628
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
25984         generating rtl if
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
25989         generated.
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,
26005         move prologue...
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>
26019         PR sanitizer/82517
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().
26026         PR target/83738
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,
26032         OS_main.
26033         (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
26034         attribute.
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>
26051         PR ipa/80763
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>
26057         PR target/83507
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
26065         do_report_bug.
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
26077         function.
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>
26092         PR target/83677
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
26104         cfg_altered.
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
26110         splitting insns.
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
26131         to generate rtl.
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
26149         cache.
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>
26164         PR lto/83719
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
26184         munaligned-access.
26186 2018-01-08  Sebastian Huber  <sebastian.huber@embedded-brains.de>
26188         PR target/83681
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
26196         USED_FOR_TARGET.
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
26214         forwarder blocks.
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
26220         of paths.
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>
26244         PR debug/83480
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>
26259         PR target/83604
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
26281         exceptions.
26282         (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
26283         processing.
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
26295         defined.
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.
26300         PR target/83604
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
26323         POLY_INT_CST.
26325 2018-01-05  Sudakshina Das  <sudi.das@arm.com>
26327         PR target/82439
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
26336         can throw.
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>
26346         PR target/83554
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>
26355         PR bootstrap/81926
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.
26359         Include output.h.
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>
26388         PR debug/83585
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>
26394         PR target/83680
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>
26400         PR target/83387
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>
26406         PR debug/83666
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>
26412         PR target/83675
26413         * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
26414         TARGET_VIS2.
26416 2018-01-04  Uros Bizjak  <ubizjak@gmail.com>
26418         PR target/83628
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
26439         gcov file.
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>
26445         PR ipa/82352
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,
26471         split out from..,
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,
26484         split out from...
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,
26491         split out from...
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
26498         vect_def_type.
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
26501         vect_def_type.
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
26518         isn't available.
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,
26543         if defined.
26544         (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
26545         if nonzero.
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
26554         arguments.
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
26582         accordingly.
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
26595         parameter.
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
26607         if adj_nunits.
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
26628         MODE_VECTOR_BOOL.
26629         * lto-streamer-in.c (lto_input_mode_table): Likewise.
26630         * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
26631         Likewise.
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
26650         poly_uint16_pod.
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
26661         for GET_MODE_SIZE.
26662         * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
26663         for GET_MODE_SIZE.
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
26745         unsigned short.
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
26783         poly_uint16_pod.
26784         (mode_to_precision): Return a poly_uint16 rather than an unsigned
26785         short.
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
26792         initializer.
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
26798         as polynomial.
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
26839         VECTOR_CST_NELTS.
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):
26891         Likewise.
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
26909         poly_uint16_pod.
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):
26914         New typedefs.
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
26920         poly_uint16.
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>
26985         PR target/83641
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.
26991         PR target/83641
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>
26998         PR debug/83621
26999         * cfgexpand.c (expand_debug_expr): Return NULL if mode is
27000         BLKmode for ternary, binary or unary expressions.
27002         PR debug/83645
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>
27024         PR c/83559
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
27046         of the residual.
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
27061         to false.
27062         (strlen_dom_walker::before_dom_children): Call
27063         gimple_purge_dead_eh_edges. Dump tranformation with details
27064         dump flags.
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>
27071         PR ipa/83549
27072         * cif-code.def (VARIADIC_THUNK): New enum value.
27073         * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
27074         thunks.
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>
27087         PR ipa/83594
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
27117         unsigned int.
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
27152         than shwi trees.
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
27241         a poly_int64.
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
27252         function.
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
27371         numbers of units.
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
27389         on multiple_p.
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
27434         reductions.
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
27450         relied on above.
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):
27469         Likewise.
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
27481         to a poly_uint64.
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
27547         to a poly_uint64.
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
27555         poly_uint64 *.
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
27570         poly_uint64 *.
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
27578         be constant.
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
27593         vf too.
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
27602         a poly_uint64.
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
27619         for now.
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):
27649         New function.
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
27866         declaration.
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):
27912         Delete.
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
27924         into...
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
27931         into...
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):
27953         Delete.
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
27987         new function.
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
27994         being permuted.
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
27998         registers.
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>
28057         PR target/81616
28058         * config/i386/x86-tune-costs.h: Increase cost of integer load costs
28059         for generic 4->6.
28061 2018-01-02  Jan Hubicka  <hubicka@ucw.cz>
28063         PR target/81616
28064         Generic tuning.
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
28079         class.
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
28087         through memory.
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
28092         too.
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.