diagnostics: add minimum width to left margin for line numbers
[official-gcc.git] / gcc / ChangeLog
blob789e43b238848bafc65c297f45d253ee4ed1d24d
1 2018-10-15  David Malcolm  <dmalcolm@redhat.com>
3         * common.opt (fdiagnostics-minimum-margin-width=): New option.
4         * diagnostic-show-locus.c (layout::layout): Apply the minimum
5         margin width.
6         (layout::start_annotation_line): Only print up to 3 of the
7         margin character, to avoid touching the left-hand side.
8         (selftest::test_diagnostic_show_locus_fixit_lines): Update for
9         minimum margin width, as set by test_diagnostic_context's ctor.
10         (selftest::test_fixit_insert_containing_newline): Likewise.
11         (selftest::test_fixit_insert_containing_newline_2): Likewise.
12         (selftest::test_line_numbers_multiline_range): Clear
13         dc.min_margin_width.
14         * diagnostic.c (diagnostic_initialize): Initialize
15         min_margin_width.
16         * diagnostic.h (struct diagnostic_context): Add field
17         "min_margin_width".
18         * doc/invoke.texi: Add -fdiagnostics-minimum-margin-width=.
19         * opts.c (common_handle_option): Handle
20         OPT_fdiagnostics_minimum_margin_width_.
21         * selftest-diagnostic.c
22         (selftest::test_diagnostic_context::test_diagnostic_context):
23         Initialize min_margin_width to 6.
24         * toplev.c (general_init): Initialize global_dc->min_margin_width.
26 2018-10-15  David Malcolm  <dmalcolm@redhat.com>
28         * gcc-rich-location.h (gcc_rich_location::add_location_if_nearby):
29         Fix usage of "error_at_rich_loc" in the comment.
31 2018-10-15  Renlin Li  <renlin.li@arm.com>
33         PR target/87563
34         * tree-vectorizer.c (try_vectorize_loop_1): Don't use
35         if-conversioned loop when it contains ifn with types not
36         supported by backend.
37         * internal-fn.c (expand_direct_optab_fn): Add an assert.
38         (direct_internal_fn_supported_p): New helper function.
39         * internal-fn.h (direct_internal_fn_supported_p): Declare.
41 2018-10-15  Jakub Jelinek  <jakub@redhat.com>
43         PR target/87572
44         * common/config/i386/i386-common.c (OPTION_MASK_ISA_AVX512F_UNSET):
45         Remove OPTION_MASK_ISA_AVX5124FMAPS_UNSET and
46         OPTION_MASK_ISA_AVX5124VNNIW_UNSET.
48 2018-10-15  Bin Cheng  <bin.cheng@linux.alibaba.com>
50         PR tree-optimization/87022
51         * tree-loop-distribution.c (pg_add_dependence_edges): Check all
52         bits in dist vector rather than the first one.
54 2018-10-15  Richard Biener  <rguenther@suse.de>
56         PR middle-end/87610
57         * tree-ssa-structalias.c (struct vls_data): Add escaped_p member.
58         (visit_loadstore): When a used restrict tag escaped verify that
59         the points-to solution of "other" pointers do not include
60         escaped.
61         (compute_dependence_clique): If a used restrict tag escaped
62         communicated that down to visit_loadstore.
64 2018-10-15  Andreas Krebbel  <krebbel@linux.ibm.com>
66         * config/s390/s390.c (s390_expand_vec_init): Force vector element
67         into reg if it isn't a general operand.
69 2018-10-14  H.J. Lu  <hongjiu.lu@intel.com>
71         PR target/87599
72         * config/i386/sse.md (*vec_dupv2di): Add register source to
73         movddup.
75 2018-10-14  H.J. Lu  <hongjiu.lu@intel.com>
77         PR target/87572
78         * common/config/i386/i386-common.c (OPTION_MASK_ISA_AVX512F_UNSET):
79         Add OPTION_MASK_ISA_AVX512IFMA_UNSET,
80         OPTION_MASK_ISA_AVX5124FMAPS_UNSET and
81         OPTION_MASK_ISA_AVX5124VNNIW_UNSET.
83 2018-10-13  Eric Botcazou  <ebotcazou@adacore.com>
85         * dwarf2cfi.c (struct dw_trace_info): Add args_size_defined_for_eh.
86         (notice_args_size): Set it in the current trace if no insn that can
87         throw internally has been seen yet.
88         (connect_traces): When connecting args_size between traces, allow the
89         incoming values not to match if there is an insn setting it before the
90         first insn that can throw internally; in that case, force the creation
91         of a CFI note on this latter insn.
93 2018-10-13  Jonathan Wakely  <jwakely@redhat.com>
95         * opt-problem.h (opt_wrapper): Use template-argument-list when naming
96         the base class, because using the injected-class-name was not clearly
97         specified until DR 176.
99 2018-10-12  Paul Koning  <ni1d@arrl.net>
101         * config/pdp11/pdp11.md (doloop_end): New expander.
102         (doloop_end_insn): renamed from "doloop_end".
103         (addqi3): New pattern.
104         (subqi3): New pattern.
105         * config/pdp11/predicates.md (incdec_operand): New predicate.
107 2018-10-12  Yury Gribov  <tetra2005@gmail.com>
109         PR middle-end/81376
110         * real.c (format_helper::can_represent_integral_type_p): New function
111         * real.h (format_helper::can_represent_integral_type_p): Ditto.
112         * match.pd: New pattern.
114 2018-10-12  Alexandre Oliva <oliva@adacore.com>
116         * configure.ac: Introduce --enable-large-address-aware
117         to define MINGW_DEFAULT_LARGE_ADDR_AWARE.
118         * doc/install.texi: Document it.
119         * configure, config.in: Rebuilt.
120         * config/i386/mingw32.h (LINK_SPEC_LARGE_ADDR_AWARE): Define,
121         based on MINGW_DEFAULT_LARGE_ADDR_AWARE.
122         (LINK_SPEC): Insert it.
123         * config/i386/mingw-264.h: Likewise.
125         * cygming.h (HAVE_GAS_ALIGNED_COMM): Fallback-define.
127 2018-10-12  Peter Bergner  <bergner@linux.ibm.com>
129         PR rtl-optimization/87600
130         * ira-lives (non_conflicting_reg_copy_p): Disable for non LRA targets.
132 2018-10-12  Paul Koning  <ni1d@arrl.net>
134         * doc/md.texi (doloop_end): Document that the pattern code may
135         need to check operand mode.
137 2018-10-12  Wilco Dijkstra  <wdijkstr@arm.com>
139         * config/aarch64/aarch64.md (zero_extendsidi2_aarch64): Add alternatives
140         to zero-extend between int and floating-point registers.
141         (load_pair_zero_extendsidi2_aarch64): Add alternative for zero-extended
142         ldp into floating-point registers.  Add type and arch attributes.
143         (zero_extend<SHORT:mode><GPI:mode>2_aarch64): Add arch attribute.
144         Use f_loads for type attribute.
146 2018-10-11  Martin Sebor  <msebor@redhat.com>
148         * doc/extend.texi (attribute packed): Correct typos.
150 2018-10-11  Martin Sebor  <msebor@redhat.com>
152         * doc/extend.texi (attribute flatten): Mention interaction with
153         noinline.
155 2018-10-11  Jan Hubicka  <hubicka@ucw.cz>
157         PR target/87156
158         * cgraphclones.c (cgraph_node::create_version_clone_with_body):
159         Set new_decl virtual flag to zero.
161 2018-10-11  Martin Sebor  <msebor@redhat.com>
163         PR middle-end/87593
164         * doc/extend.texi (attribute format_arg): Discuss using multiple
165         attributes on a single function.
167 2018-10-11  Giuliano Belinassi  <giuliano.belinassi@usp.br>
169         PR tree-optimization/86829
170         * match.pd (sin (atan (x))): New simplification rules.
171         (cos (atan (x))): Likewise.
172         * real.c (build_sinatan_real): New function.
173         * real.h (build_sinatan_real): Prototype.
175 2018-10-11  Will Schmidt <will_schmidt@vnet.ibm.com>
177         * config/rs6000/rs6000.c (map_to_integral_tree_type): New helper
178         function.
179         (fold_mergeeo_helper): New helper function.
180         (rs6000_gimple_fold_builtin): Add hooks for vec_mergee and vec_mergeo
181         intrinsics.  Correct some whitespace indentation issues.
183 2018-10-11  Wilco Dijkstra  <wdijkstr@arm.com>
185         PR target/87511
186         * config/aarch64/aarch64.c (aarch64_mask_and_shift_for_ubfiz_p):
187         Use HOST_WIDE_INT_1U for shift.
189 2018-10-11  Doug Rupp  <rupp@adacore.com>
190             Olivier Hainque  <hainque@adacore.com>
192         * config/rs6000/vxworks.h (VXWORKS_RELAX_LINK_SPEC): New macro.
193         Pass --relax to the linker for RTPs.
194         (LINK_SPEC): Append VXWORKS_RELAX_LINK_SPEC.
196 2018-10-11  Andrew Stubbs  <ams@codesourcery.com>
197             Jan Hubicka  <jh@suse.cz>
198             Martin Jambor  <mjambor@suse.cz>
200         * print-rtl.c (print_rtx_operand_codes_E_and_V): Print how many times
201         the same elements are repeated rather than printing all of them.
202         * read-rtl.c (rtx_reader::read_rtx_operand): Recognize and expand
203         "repeated" elements.
204         * read-rtl-function.c (test_loading_repeat): New function.
205         (read_rtl_function_c_tests): Call test_loading_repeat.
206         * rtl-tests.c (test_dumping_repeat): New function.
207         (rtl_tests_c_tests): Call test_dumping_repeat.
209 2018-10-11  Richard Biener  <rguenther@suse.de>
211         * config/i386/x86-tune-costs.h (bdver?_memcpy, bdver?_memset,
212         bdver?_cost): Unify to ...
213         (bdver_memcpy, bdver_memset, bdver_cost): ... this.
214         * config/i386/i386.c (processor_cost_table): Adjust.
216 2018-10-10  Eric Botcazou  <ebotcazou@adacore.com>
218         PR middle-end/87574
219         * cgraphunit.c (cgraph_node::expand_thunk): Force DECL_IGNORED_P on
220         the thunk when expanding to GIMPLE.
222 2018-10-10  Bernd Edlinger  <bernd.edlinger@hotmail.de>
224         * varasm.c (mergeable_string_section): Don't try to move zero-length
225         strings to the merge section.
227 2018-10-10  Uros Bizjak  <ubizjak@gmail.com>
229         PR target/87573
230         * config/i386/mmx.md (const_vector 0 -> mem splitter): New splitter.
232 2018-10-10  Jakub Jelinek  <jakub@redhat.com>
234         PR target/87550
235         * config/i386/i386-builtin.def (IX86_BUILTIN_RDPMC): Move from args set
236         to special_args set.
238 2018-10-10  Richard Biener  <rguenther@suse.de>
240         * config/i386/sse.md (reduc_plus_scal_v8df, reduc_plus_scal_v4df,
241         reduc_plus_scal_v2df, reduc_plus_scal_v16sf, reduc_plus_scal_v8sf,
242         reduc_plus_scal_v4sf): Merge into pattern reducing to half width
243         and recursing and pattern terminating the recursion on SSE
244         vector width using ix86_expand_reduc.
245         (reduc_sminmax_scal_<mode>): Split into part reducing to half
246         width and recursing and SSE2 vector variant doing the final
247         reduction with ix86_expand_reduc.
248         (reduc_uminmax_scal_<mode>): Likewise for the AVX512 variants
249         with terminating the recursion at AVX level, splitting that
250         to SSE there.
252 2018-10-09  David Malcolm  <dmalcolm@redhat.com>
254         * genmatch.c (error_cb): Rename to...
255         (diagnostic_cb): ...this, converting int params to enums.
256         (fatal_at): Update for renaming.
257         (warning_at): Likewise.
258         (main): Likewise.
259         * input.c (selftest::ebcdic_execution_charset::apply):
260         Update for renaming of...
261         (selftest::ebcdic_execution_charset::on_error): ...this, renaming
262         to...
263         (selftest::ebcdic_execution_charset::on_diagnostic): ...this,
264         converting level and reason to enums.
265         (class selftest::lexer_error_sink): Rename to...
266         (class selftest::lexer_test_options): ...this, renaming field
267         "m_errors" to "m_diagnostics".
268         (selftest::lexer_test_options::apply): Update for renaming of...
269         (selftest::lexer_test_options::on_error): ...this, renaming to...
270         (selftest::lexer_test_options::on_diagnostic): ...this
271         converting level and reason to enums.
272         (selftest::test_lexer_string_locations_raw_string_unterminated):
273         Update for renamings.
274         * opth-gen.awk (struct cpp_reason_option_codes_t): Use enum for
275         "reason".
277 2018-10-09  Paul A. Clarke  <pc@us.ibm.com>
279         * config.gcc (powerpc*-*-*): Add pmmintrin.h to extra_headers.
280         * config/rs6000/pmmintrin.h: New file.
282 2018-10-09  Eric Botcazou  <ebotcazou@adacore.com>
284         PR tree-optimization/86659
285         * gimple-match.h (gimple_match_op constructors): Initialize reverse.
287 2018-10-09  Richard Biener  <rguenther@suse.de>
289         PR tree-optimization/63155
290         * tree-ssa-structalias.c: Include tree-ssa.h.
291         (get_constraint_for_ssa_var): For undefs return nothing_id.
292         (find_func_aliases): Cleanup PHI handling.
294 2018-10-09  Richard Biener  <rguenther@suse.de>
296         * tree-vectorizer.c (dump_stmt_cost): Fix cut&paste missing
297         replacements.
299 2018-10-09  Martin Liska  <mliska@suse.cz>
301         * asan.c (asan_emit_stack_protection): If a stack variable
302         is located in a same file as current function, then emit
303         line info into variable definition string.
305 2018-10-08  Eric Botcazou  <ebotcazou@adacore.com>
307         * print-rtl.c (rtx_writer::print_rtx_operand_code_i): Print column
308         information.
310 2018-10-08  Eric Botcazou  <ebotcazou@adacore.com>
312         * cgraphunit.c (cgraph_node::expand_thunk): Do not force DECL_IGNORED_P
313         on the thunk.
315 2018-10-08  Eric Botcazou  <ebotcazou@adacore.com>
317         PR tree-optimization/86659
318         * gimple-match.h (struct gimple_match_op): Add reverse field.
319         (gimple_match_op::set_op): New overloaded method.
320         * gimple-match-head.c (maybe_build_generic_op) <BIT_FIELD_REF>: Set
321         the REF_REVERSE_STORAGE_ORDER flag on the value.
322         (gimple_simplify) <GIMPLE_ASSIGN>: For BIT_FIELD_REF, propagate the
323         REF_REVERSE_STORAGE_ORDER flag and avoid simplifying if it is set.
324         
325 2018-10-08  Richard Sandiford  <richard.sandiford@arm.com>
327         PR middle-end/63155
328         * gimple-ssa-backprop.c (backprop::intersect_uses): Use
329         FOR_EACH_IMM_USE_FAST instead of FOR_EACH_IMM_USE_STMT.
331 2018-10-08  H.J. Lu  <hongjiu.lu@intel.com>
333         PR target/87517
334         * config/i386/avx512fintrin.h (_mm512_mask_fmaddsub_round_pd):
335         Defined with __builtin_ia32_vfmaddsubpd512_mask.
337 2018-10-08  Richard Biener  <rguenther@suse.de>
339         * config/i386/x86-tune-costs.h (znver1_cost): Make AVX256 vector loads
340         cost the same as AVX128 ones.
342 2018-10-08  Paul Koning  <ni1d@arrl.net>
344         * config/pdp11/pdp11-protos.h (output_block_move): Remove.
345         (expand_block_move): New function.
346         * config/pdp11/pdp11.c (output_block_move): Remove.
347         (expand_block_move): New function.
348         * config/pdp11/pdp11.h (MOVE_RATIO): New definition.
349         * config/pdp11/pdp11.md (movmemhi): Use expand_block_move.
350         (*movmemhi1): Remove.
352 2018-10-08  Robin Dapp  <rdapp@linux.ibm.com>
354         * config/s390/2827.md: Increase latencies for some FP instructions.
356 2018-10-08  Richard Biener  <rguenther@suse.de>
358         * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
359         Open a dump scope.
360         * tree-vectorizer.c (dump_stmt_cost): Add cost param and dump it.
361         * tree-vectorizer.h (dump_stmt_cost): Adjust.
362         (add_stmt_cost): Dump return value of the hook.
364 2018-10-08  Richard Biener  <rguenther@suse.de>
366         PR tree-optimization/63155
367         * tree-ssa-propagate.c (add_ssa_edge): Do cheap check first.
368         (ssa_propagation_engine::ssa_propagate): Remove redundant
369         bitmap bit clearing.
371 2018-10-05  Peter Bergner  <bergner@linux.ibm.com>
373         PR rtl-optimization/86939
374         PR rtl-optimization/87479
375         * ira.h (non_conflicting_reg_copy_p): New prototype.
376         * ira-lives.c (ignore_reg_for_conflicts): New static variable.
377         (make_hard_regno_dead): Don't add conflicts for register
378         ignore_reg_for_conflicts.
379         (make_object_dead): Likewise.
380         (non_conflicting_reg_copy_p): New function.
381         (process_bb_node_lives): Set ignore_reg_for_conflicts for copies.
382         Remove special conflict handling of REAL_PIC_OFFSET_TABLE_REGNUM.
383         * lra-lives.c (ignore_reg_for_conflicts): New static variable.
384         (make_hard_regno_dead): Don't add conflicts for register
385         ignore_reg_for_conflicts.  Remove special conflict handling of
386         REAL_PIC_OFFSET_TABLE_REGNUM.  Remove now unused argument
387         check_pic_pseudo_p and update callers.
388         (mark_pseudo_dead): Don't add conflicts for register
389         ignore_reg_for_conflicts.
390         (process_bb_lives): Set ignore_reg_for_conflicts for copies.
392 2018-10-05  Andrew Waterman  <andrew@sifive.com>
393             Jim Wilson  <jimw@sifive.com>
395         * config/riscv/riscv.md (f<quiet_pattern>_quiet<ANYF:mode><X:mode>4):
396         Add define_expand.  Add ! HONOR_SNANS check to current pattern.  Add
397         new pattern using HONOR_SNANS that emits one extra instruction.
399 2018-10-05  Segher Boessenkool  <segher@kernel.crashing.org>
401         * config/rs6000/rs6000.md (unnamed mfcr scc_comparison_operator
402         patterns): Merge SI and DI patterns to a GPR pattern.
403         (unnamed define_insn and define_split for record form of that): Merge
404         to a single define_insn_and_split pattern.
406 2018-10-05  David Malcolm  <dmalcolm@redhat.com>
408         PR c++/56856
409         * input.c (expand_location_to_spelling_point): Add param "aspect"
410         and use rather than hardcoding LOCATION_ASPECT_CARET.
411         (get_substring_ranges_for_loc): Handle the case of a single token
412         within a macro expansion.
413         * input.h (expand_location_to_spelling_point): Add "aspect" param,
414         defaulting to LOCATION_ASPECT_CARET.
416 2018-10-05  Paul Koning  <ni1d@arrl.net>
418         * config/pdp11/pdp11.c (TARGET_CXX_GUARD_TYPE): Define.
419         (TARGET_CXX_CLASS_DATA_ALWAYS_COMDAT): Define.
420         (TARGET_CXX_LIBRARY_RTTI_COMDAT): Define.
421         (pdp11_guard_type): New function.
423 2018-10-05  Paul Koning  <ni1d@arrl.net>
425         * config/pdp11/pdp11.h (FLOAT_TYPE_SIZE): Always 32.
426         * config/pdp11/pdp11.opt (mfloat32): Remove.
427         (mfloat64): Remove.
428         * doc/invoke.texi (pdp11 -mfloat32): Remove:
429         (pdp11 -mfloat64): Remove.
431 2018-10-05  Uros Bizjak  <ubizjak@gmail.com>
433         * config/i386/i386.md (*cmpxf_cc_i387): Remove pattern.
434         (*cmp<mode>_cc_i387): Ditto.
435         (*cmpu<mode>_cc_i387): Ditto.
436         (*cmp<X87MODEF:mode>_<SWI24:mode>_cc_i387): Ditto.
437         * config/i386/i386.c (ix86_expand_fp_compare): Remove
438         "scratch" argument.
439         <case IX86_FPCMP_SAHF>: Do not generate pattern with HImode clobber.
440         Emit x86_sahf_1 pattern.
441         (ix86_expand_compare): Update call to ix86_expand_fp_compare.
442         (ix86_expand_carry_flag_compare): Ditto.
444 2018-10-05  Uros Bizjak  <ubizjak@gmail.com>
446         * config/i386/i386.md (*cmpxf_i387): Change operand 2 predicate
447         to reg_or_0_operand.  Add "C" constraint.
448         (*cmpxf_cc_i387): Ditto.
449         (*cmp<mode>_i387): Change operand 2 predicate
450         to nonimm_or_0_operand.  Add "C" constraint.
451         (*cmp<mode>_cc_i387): Ditto.
452         (*cmp<mode>_0_i387): Remove insn pattern.
453         (*cmp<mode>_0_cc_i387): Ditto.
455 2018-10-05  Uros Bizjak  <ubizjak@gmail.com>
457         * config/i386/constraints.md ("C"): Do not depend on TARGET_SSE.
458         * config/i386/predicates.md (nonimm_or_0_operand): Rename
459         from vector_move_operand.  Update all uses.
461 2018-10-05  Martin Sebor  <msebor@redhat.com>
463         PR tree-optimization/87490
464         * builtins.c (expand_builtin_strnlen): Handle a null data.decl
465         consistently.
467 2018-10-05  Richard Biener  <rguenther@suse.de>
469         PR tree-optimization/63155
470         * tree-ssa-ccp.c (ccp_propagate::visit_phi): Avoid excess
471         vertical space in dumpfiles.
472         * tree-ssa-propagate.h
473         (ssa_propagation_engine::process_ssa_edge_worklist): Remove.
474         * tree-ssa-propagate.c (cfg_blocks_back): New global.
475         (ssa_edge_worklist_back): Likewise.
476         (curr_order): Likewise.
477         (cfg_blocks_get): Remove abstraction.
478         (cfg_blocks_add): Likewise.
479         (cfg_blocks_empty_p): Likewise.
480         (add_ssa_edge): Add to current or next worklist based on
481         RPO index.
482         (add_control_edge): Likewise.
483         (ssa_propagation_engine::process_ssa_edge_worklist): Fold
484         into ...
485         (ssa_propagation_engine::ssa_propagate): ... here.  Unify
486         iteration from CFG and SSA edge worklist so we process
487         everything in RPO order, prioritizing forward progress
488         over iteration.
489         (ssa_prop_init): Allocate new worklists, do not dump
490         immediate uses.
491         (ssa_prop_fini): Free new worklists.
493 2018-10-05  Richard Biener  <rguenther@suse.de>
495         * tree-core.h (tree_block::abstract_flag): Remove.
496         (tree_block::block_num): Make full 32bits.
497         * tree.def (BLOCK): Remove docs about BLOCK_ABSTRACT.
498         * tree.h (BLOCK_ABSTRACT): Remove.
499         * dwarf2out.c (gen_lexical_block_die): Remove dead code
500         resulting from BLOCK_ABSTRACT being always false.
501         (gen_inlined_subroutine_die): Likewise.
502         (gen_block_die): Likewise.
503         * tree.c (block_ultimate_origin): Likewise.
504         * tree-pretty-print.c (dump_block_node): Remove code dealing
505         with BLOCK_ABSTRACT.
506         * tree-ssa-live.c (dump_scope_block): Likewise.
507         * tree-streamer-in.c (unpack_ts_block_value_fields): Likewise.
508         * tree-streamer-out.c (pack_ts_block_value_fields): Likewise.
510 2018-10-05   Richard Biener  <rguenther@suse.de>
512         * config/i386/i386.c (ix86_add_stmt_cost): When scalar cost
513         is asked for initialize mode to the component mode of the
514         vector type.
516 2018-10-05  H.J. Lu  <hongjiu.lu@intel.com>
518         PR target/87522
519         * config/i386/gnu-user.h (ASM_SPEC): Don't pass -msse2avx to
520         assembler for -mavx.
521         * config/i386/gnu-user64.h (ASM_SPEC): Likewise.
523 2018-10-05  Segher Boessenkool  <segher@kernel.crashing.org>
525         PR target/87509
526         * config/rs6000/rs6000-builtin.def (RS6000_BUILTIN_SET_FPSCR_DRN): Use
527         RS6000_BTM_DFP.
528         * config/rs6000/rs6000.md (rs6000_set_fpscr_rn): Require the operand
529         to be DImode.  When using mffscrn, force the operand to a register.
531 2018-10-04  Uros Bizjak  <ubizjak@gmail.com>
533         * config/i386/i386.md (*fop_<X87MODEF:mode>_2_i387): Macroize insn
534         from *fop_<MODEF:mode>_2_i387 and *fop_xf_2_i387 using
535         X87MODEF mode iterator.
536         (*fop_<X87MODEF:mode>_3_i387): Macroize insn from
537         *fop_<MODEF:mode>_3_i387 and *fop_xf_3_i387 using
538         X87MODEF mode iterator.
540 2018-10-04  Vinay Kumar  <vinay.kumar@blackfigtech.com>
542         * doc/invoke.texi (-Wno-prio-ctor-dtor): Document new warning
543         -Wno-prio-ctor-dtor.
545 2018-10-04  David Malcolm  <dmalcolm@redhat.com>
547         * Makefile.in (OBJS): Add opt-problem.o.
548         * dump-context.h: Include "selftest.h.
549         (selftest::temp_dump_context): New forward decl.
550         (class dump_context): Make friend of class
551         selftest::temp_dump_context.
552         (dump_context::dump_loc_immediate): New decl.
553         (class dump_pretty_printer): Move here from dumpfile.c.
554         (class temp_dump_context): Move to namespace selftest.
555         (temp_dump_context::temp_dump_context): Add param
556         "forcibly_enable_dumping".
557         (selftest::verify_dumped_text):
558         (ASSERT_DUMPED_TEXT_EQ): Move here from dumpfile.c.
559         (selftest::verify_item):
560         (ASSERT_IS_TEXT): Move here from dumpfile.c.
561         (ASSERT_IS_TREE): Likewise.
562         (ASSERT_IS_GIMPLE): Likewise.
563         * dumpfile.c (dump_context::dump_loc): Move immediate dumping
564         to...
565         (dump_context::dump_loc_immediate): ...this new function.
566         (class dump_pretty_printer): Move to dump-context.h.
567         (dump_switch_p_1): Don't enable MSG_PRIORITY_REEMITTED.
568         (opt_info_switch_p_1): Enable MSG_PRIORITY_REEMITTED.
569         (temp_dump_context::temp_dump_context): Move to "selftest"
570         namespace.  Add param "forcibly_enable_dumping", and use it to
571         conditionalize the use of m_pp;
572         (selftest::verify_dumped_text): Make non-static.
573         (ASSERT_DUMPED_TEXT_EQ): Move to dump-context.h.
574         (selftest::verify_item): Make non-static.
575         (ASSERT_IS_TEXT): Move to dump-context.h.
576         (ASSERT_IS_TREE): Likewise.
577         (ASSERT_IS_GIMPLE): Likewise.
578         (selftest::test_capture_of_dump_calls): Pass "true" for new
579         param of temp_dump_context.
580         * dumpfile.h (enum dump_flag): Add MSG_PRIORITY_REEMITTED, adding
581         it to MSG_ALL_PRIORITIES.  Update values of TDF_COMPARE_DEBUG and
582         TDF_COMPARE_DEBUG.
583         * opt-problem.cc: New file.
584         * opt-problem.h: New file.
585         * optinfo-emit-json.cc
586         (selftest::test_building_json_from_dump_calls): Pass "true" for
587         new param of temp_dump_context.
588         * optinfo.cc (optinfo_kind_to_dump_flag): New function.
589         (optinfo::emit_for_opt_problem): New function.
590         (optinfo::emit): Clarity which emit_item is used.
591         * optinfo.h (optinfo::get_dump_location): New accessor.
592         (optinfo::emit_for_opt_problem): New decl.
593         (optinfo::emit): Make const.
594         * selftest-run-tests.c (selftest::run_tests): Call
595         selftest::opt_problem_cc_tests.
596         * selftest.h (selftest::opt_problem_cc_tests): New decl.
597         * tree-data-ref.c (dr_analyze_innermost): Convert return type from
598         bool to opt_result, converting fprintf messages to
599         opt_result::failure_at calls.  Add "stmt" param for use by the
600         failure_at calls.
601         (create_data_ref): Pass "stmt" to the dr_analyze_innermost call.
602         (runtime_alias_check_p): Convert return type from bool to
603         opt_result, converting dump_printf calls to
604         opt_result::failure_at, using the statement DDR_A for their
605         location.
606         (find_data_references_in_stmt): Convert return type from bool to
607         opt_result, converting "return false" to opt_result::failure_at
608         with a new message.
609         * tree-data-ref.h: Include "opt-problem.h".
610         (dr_analyze_innermost): Convert return type from bool to opt_result,
611         and add a const gimple * param.
612         (find_data_references_in_stmt): Convert return type from bool to
613         opt_result.
614         (runtime_alias_check_p): Likewise.
615         * tree-predcom.c (find_looparound_phi): Pass "init_stmt" to
616         dr_analyze_innermost.
617         * tree-vect-data-refs.c (vect_mark_for_runtime_alias_test):
618         Convert return type from bool to opt_result, adding a message for
619         the PARAM_VECT_MAX_VERSION_FOR_ALIAS_CHECKS zero case.
620         (vect_analyze_data_ref_dependence): Convert return type from bool
621         to opt_result.  Change sense of return type from "false"
622         effectively meaning "no problems" to "false" meaning a problem,
623         so that "return false" becomes "return opt_result::success".
624         Convert "return true" calls to opt_result::failure_at, using
625         the location of statement A rather than vect_location.
626         (vect_analyze_data_ref_dependences): Convert return type from bool
627         to opt_result.
628         (verify_data_ref_alignment): Likewise, converting dump_printf_loc
629         calls to opt_result::failure_at, using the stmt location rather
630         than vect_location.
631         (vect_verify_datarefs_alignment): Convert return type from bool
632         to opt_result.
633         (vect_enhance_data_refs_alignment): Likewise.  Split local "stat"
634         into multiple more-tightly-scoped copies.
635         (vect_analyze_data_refs_alignment): Convert return type from bool
636         to opt_result.
637         (vect_analyze_data_ref_accesses): Likewise, converting a
638         "return false" to a "return opt_result::failure_at", adding a
639         new message.
640         (vect_prune_runtime_alias_test_list): Convert return type from
641         bool to opt_result, converting dump_printf_loc to
642         opt_result::failure_at.  Add a %G to show the pertinent statement,
643         and use the stmt's location rather than vect_location.
644         (vect_find_stmt_data_reference): Convert return type from
645         bool to opt_result, converting dump_printf_loc to
646         opt_result::failure_at, using stmt's location.
647         (vect_analyze_data_refs):  Convert return type from bool to
648         opt_result.  Convert "return false" to "return
649         opt_result::failure_at", adding messages as needed.
650         * tree-vect-loop.c (vect_determine_vf_for_stmt_1): Convert return
651         type from bool to opt_result.
652         (vect_determine_vf_for_stmt): Likewise.
653         (vect_determine_vectorization_factor): Likewise, converting
654         dump_printf_loc to opt_result::failure_at, using location of phi
655         rather than vect_location.
656         (vect_analyze_loop_form_1): Convert return type from bool to
657         opt_result, converting dump_printf_loc calls, retaining the use of
658         vect_location.
659         (vect_analyze_loop_form): Convert return type from loop_vec_info
660         to opt_loop_vec_info.
661         (vect_analyze_loop_operations): Convert return type from bool to
662         opt_result, converting dump_printf_loc calls, using the location
663         of phi/stmt rather than vect_location where available.  Convert
664         various "return false" to "return opt_result::failure_at" with
665         "unsupported phi" messages.
666         (vect_get_datarefs_in_loop): Convert return type from bool to
667         opt_result.  Add a message for the
668         PARAM_LOOP_MAX_DATAREFS_FOR_DATADEPS failure.
669         (vect_analyze_loop_2): Convert return type from bool to
670         opt_result.  Ensure "ok" is set to a opt_result::failure_at before
671         each "goto again;", adding new messages where needed.
672         Add "unsupported grouped {store|load}" messages.
673         (vect_analyze_loop): Convert return type from loop_vec_info to
674         opt_loop_vec_info.
675         * tree-vect-slp.c (vect_analyze_slp): Convert return type from
676         bool to opt_result.
677         * tree-vect-stmts.c (process_use): Likewise, converting
678         dump_printf_loc call and using stmt location, rather than
679         vect_location.
680         (vect_mark_stmts_to_be_vectorized): Likeise.
681         (vect_analyze_stmt): Likewise, adding a %G.
682         (vect_get_vector_types_for_stmt): Convert return type from bool to
683         opt_result, converting dump_printf_loc calls and using stmt
684         location, rather than vect_location.
685         (vect_get_mask_type_for_stmt): Convert return type from tree to
686         opt_tree, converting dump_printf_loc calls and using stmt location.
687         * tree-vectorizer.c: Include "opt-problem.h.
688         (try_vectorize_loop_1): Flag "Analyzing loop at" dump message as
689         MSG_PRIORITY_INTERNALS.  Convert local "loop_vinfo" from
690         loop_vec_info to opt_loop_vec_info.  If if fails, and dumping is
691         enabled, use it to report at the top level "couldn't vectorize
692         loop" followed by the problem.
693         * tree-vectorizer.h (opt_loop_vec_info): New typedef.
694         (vect_mark_stmts_to_be_vectorized): Convert return type from bool
695         to opt_result.
696         (vect_analyze_stmt): Likewise.
697         (vect_get_vector_types_for_stmt): Likewise.
698         (tree vect_get_mask_type_for_stmt): Likewise.
699         (vect_analyze_data_ref_dependences): Likewise.
700         (vect_enhance_data_refs_alignment): Likewise.
701         (vect_analyze_data_refs_alignment): Likewise.
702         (vect_verify_datarefs_alignment): Likewise.
703         (vect_analyze_data_ref_accesses): Likewise.
704         (vect_prune_runtime_alias_test_list): Likewise.
705         (vect_find_stmt_data_reference): Likewise.
706         (vect_analyze_data_refs): Likewise.
707         (vect_analyze_loop): Convert return type from loop_vec_info to
708         opt_loop_vec_info.
709         (vect_analyze_loop_form): Likewise.
710         (vect_analyze_slp): Convert return type from bool to opt_result.
712 2018-10-04  David Malcolm  <dmalcolm@redhat.com>
714         * doc/invoke.texi (-fopt-info): Document new "internals"
715         sub-option.
716         * dump-context.h (dump_context::apply_dump_filter_p): New decl.
717         * dumpfile.c (dump_options): Update for renaming of MSG_ALL to
718         MSG_ALL_KINDS.
719         (optinfo_verbosity_options): Add "internals".
720         (kind_as_string): Update for renaming of MSG_ALL to MSG_ALL_KINDS.
721         (dump_context::apply_dump_filter_p): New member function.
722         (dump_context::dump_loc): Use apply_dump_filter_p rather than
723         explicitly masking the dump_kind.
724         (dump_context::begin_scope): Increment the scope depth first.  Use
725         apply_dump_filter_p rather than explicitly masking the dump_kind.
726         (dump_context::emit_item): Use apply_dump_filter_p rather than
727         explicitly masking the dump_kind.
728         (dump_dec): Likewise.
729         (dump_hex): Likewise.
730         (dump_switch_p_1): Default to MSG_ALL_PRIORITIES.
731         (opt_info_switch_p_1): Default to MSG_PRIORITY_USER_FACING.
732         (opt_info_switch_p): Update handling of default
733         MSG_OPTIMIZED_LOCATIONS to cope with default of
734         MSG_PRIORITY_USER_FACING.
735         (dump_basic_block): Use apply_dump_filter_p rather than explicitly
736         masking the dump_kind.
737         (selftest::test_capture_of_dump_calls): Update test_dump_context
738         instances to use MSG_ALL_KINDS | MSG_PRIORITY_USER_FACING rather
739         than MSG_ALL.  Generalize scope test to be run at all four
740         combinations of with/without MSG_PRIORITY_USER_FACING and
741         MSG_PRIORITY_INTERNALS, adding examples of explicit priority
742         for each of the two values.
743         * dumpfile.h (enum dump_flag): Add comment about the MSG_* flags.
744         Rename MSG_ALL to MSG_ALL_KINDS.  Add MSG_PRIORITY_USER_FACING,
745         MSG_PRIORITY_INTERNALS, and MSG_ALL_PRIORITIES, updating the
746         values for TDF_COMPARE_DEBUG and TDF_ALL_VALUES.
747         (AUTO_DUMP_SCOPE): Add a note to the comment about the interaction
748         with MSG_PRIORITY_*.
749         * tree-vect-loop-manip.c (vect_loop_versioning): Mark versioning
750         dump messages as MSG_PRIORITY_USER_FACING.
751         * tree-vectorizer.h (DUMP_VECT_SCOPE): Add a note to the comment
752         about the interaction with MSG_PRIORITY_*.
754 2018-10-04  Bernd Edlinger  <bernd.edlinger@hotmail.de>
756         * varasm.c (output_constant): Add new parameter merge_strings.
757         Make strings properly zero terminated in merge string sections.
758         (mergeable_string_section): Don't fail if the last char is non-zero.
759         (assemble_variable_contents): Handle merge string sections.
760         (assemble_variable): Likewise.
761         (assemble_constant_contents): Likewise.
762         (output_constant_def_contents): Likewise.
763         (output_constructor_array_range,
764         output_constructor_regular_field): Adjust call to output_constant.
765         (output_object_block): Adjust call to assemble_constant_contents
766         and assemble_variable_contents.
768 2018-10-04  Martin Liska  <mliska@suse.cz>
770         PR c/87483
771         * cgraphunit.c (process_function_and_variable_attributes):
772         Warn about a function with alias attribute and a body.
774 2018-10-04  Martin Liska  <mliska@suse.cz>
776         PR ipa/82625
777         * multiple_target.c (redirect_to_specific_clone): New function.
778         (ipa_target_clone): Use it.
779         * tree-inline.c: Fix comment.
781 2018-10-04  David Malcolm  <dmalcolm@redhat.com>
783         * dumpfile.c (gcc::dump_manager::dump_manager): Initialize new
784         fields.
785         (gcc::dump_manager::~dump_manager): Free m_optinfo_filename.
786         (gcc::dump_manager::register_pass): New member function, adapted
787         from loop body in gcc::pass_manager::register_pass, adding a
788         call to update_dfi_for_opt_info.
789         (gcc::dump_manager::opt_info_enable_passes): Store the
790         -fopt-info options into the new fields.  Move the loop
791         bodies into...
792         (gcc::dump_manager::update_dfi_for_opt_info): ...this new member
793         function.
794         * dumpfile.h (struct opt_pass): New forward decl.
795         (gcc::dump_manager::register_pass): New decl.
796         (gcc::dump_manager::update_dfi_for_opt_info): New decl.
797         (class gcc::dump_manager): Add fields "m_optgroup_flags",
798         "m_optinfo_flags", and "m_optinfo_filename".
799         * passes.c (gcc::pass_manager::register_pass): Move all of the
800         dump-handling code to gcc::dump_manager::register_pass.
802 2018-10-04  Peter Bergner  <bergner@linux.ibm.com>
804         PR rtl-optimization/87466
805         * target.def (setjmp_preserves_nonvolatile_regs_p): New target hook.
806         * doc/tm.texi.in (TARGET_SETJMP_PRESERVES_NONVOLATILE_REGS_P): New hook.
807         * doc/tm.texi: Regenerate.
808         * ira-lives.c (process_bb_node_lives): Use the new target hook.
809         * lra-lives.c (process_bb_lives): Likewise.
810         * config/rs6000/rs6000.c (TARGET_SETJMP_PRESERVES_NONVOLATILE_REGS_P):
811         Define.
813 2018-10-04  Tamar Christina  <tamar.christina@arm.com>
815         * params.c (add_params): Fix initialization.
817 2018-10-04  Martin Liska  <mliska@suse.cz>
819         PR gcov-profile/84107
820         * tree-profile.c (init_ic_make_global_vars):
821         Remove ic_void_ptr_var and ic_gcov_type_ptr_var.
822         Come up with new ic_tuple* variables.  Emit
823         __gcov_indirect_call{,_topn} variables.
824         (gimple_gen_ic_profiler): Access the variable
825         and emit gimple.
826         (gimple_gen_ic_func_profiler): Access
827         __gcov_indirect_call.callee field.
828         (gimple_init_gcov_profiler): Use ptr_type_node.
829         * value-prof.c (gimple_ic): Use ptr_type_node.
831 2018-10-04  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
833         PR tree-optimization/85787
834         * ipa-pure-const.c (malloc_candidate_p_1): Move most of malloc_candidate_p
835         into this function and add support for detecting multiple phis.
836         (DUMP_AND_RETURN): Move from malloc_candidate_p into top-level macro.
838 2018-10-04  Martin Liska  <mliska@suse.cz>
840         PR ipa/87491
841         * ipa-inline.c (inline_to_all_callers_1):
842         Call ultimate_alias_target for node being inlined.
844 2018-10-03  Jeff Law  <law@redhat.com>
846         * gimple-ssa-sprintf.c (format_string): Do not hardcode size of
847         target's wchar_t.
848         * tree.c (get_typenode_from_name): Moved from fortran/trans-types.c.
849         * tree.h (get_typenode_from_name): Prototype.
851 2018-10-03  Uros Bizjak  <ubizjak@gmail.com>
853         * config/i386/i386.md (*cmp<X87MODEF:mode>_<SWI24:mode>_i387):
854         Change operand 2 predicate to nonimmediate_operand.
855         (*cmp<X87MODEF:mode>_<SWI24:mode>_cc_i387): Ditto.
857 2018-10-03  Martin Sebor  <msebor@redhat.com>
858             Jeff Law  <law@redhat.com>
860         * gimple-ssa-sprintf.c (struct fmtresult): Add new member and
861         initialize it.
862         (get_string_length): Detect unterminated arrays.
863         (format_string): Same.
864         (format_directive): Warn about unterminated arrays.
865         (handle_gimple_call): Mark statements with no_warning as needed.
867 2018-10-03  Jim Wilson  <jimw@sifive.com>
869         * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): For ABI_ILP32E,
870         also define __riscv_abi_rve.  Delete trailing white space.
872 2018-10-03  Paul Koning  <ni1d@arrl.net>
874         Enable LRA register allocator for PDP11.
875         * config/pdp11/constraints.md (Q): Use define_memory_constraint.
876         (R): Likewise.
877         (D): Likewise.
878         * config/pdp11/pdp11.c (pdp11_lra_p): New function.
879         * config/pdp11/pdp11.opt (-mlra): New option.
880         * doc/invoke.texi (PDP-11 Options): Document -mlra.
882 2018-10-03  Uros Bizjak  <ubizjak@gmail.com>
884         * config/i386/i386.md (*<absneg:code>extendsfdf2): Remove.
885         (*<absneg:code>extend<mode>xf2): Ditto.
887 2018-10-03  Aldy Hernandez  <aldyh@redhat.com>
889         PR tree-optimization/87415
890         * tree-vrp.c (set_value_range_with_overflow): Special case one bit
891         precision fields.
893 2018-10-02  Jeff Law  <law@redhat.com>
895         * gimple-fold.c (get_range_strlen): Only set *nonstr when
896         an unterminated string is discovered.  Bubble up range
897         even for unterminated strings.
898         (gimple_fold_builtin_strlen): Do not fold if get_range_strlen
899         indicates the string was not terminated via NONSTR.
901 2018-10-03  Aldy Hernandez  <aldyh@redhat.com>
903         * tree-vrp.c (extract_range_from_unary_expr): Special case all
904         pointer conversions.
905         Do not do anything special for anti-ranges.
907 2018-10-03  Jérôme Lambourg  <lambourg@adacore.com>
909         * config/arm/vxworks.h (ARM_TARGET2_DWARF_FORMAT): Adjust to
910         DW_EH_PE_pcrel | DW_EH_PE_indirect for RTPs.
912 2018-10-03  Martin Liska  <mliska@suse.cz>
914         PR gcov-profile/86109
915         * coverage.c (coverage_begin_function): Do not
916         mark lambdas as artificial.
917         * tree-core.h (struct GTY): Remove tm_clone_flag
918         and introduce new lambda_function.
919         * tree.h (DECL_LAMBDA_FUNCTION): New macro.
921 2018-10-02  Aaron Sawdey  <acsawdey@linux.ibm.com>
923         PR target/87474
924         * config/rs6000/rs6000-string.c (expand_strn_compare): Check that both
925         P8_VECTOR and VSX are enabled.
927 2018-10-02  Andreas Krebbel  <krebbel@linux.ibm.com>
929         * config/s390/driver-native.c (s390_host_detect_local_cpu): Add
930         0x3907 as CPU model number.
932 2018-10-02  Andreas Krebbel  <krebbel@linux.ibm.com>
934         * common/config/s390/s390-common.c: Rename PF_ARCH12 to PF_Z14.
935         * config/s390/s390.h (enum processor_flags): Rename PF_ARCH12 to
936         PF_Z14.  Rename TARGET_CPU_ARCH12 to TARGET_CPU_Z14,
937         TARGET_CPU_ARCH12_P to TARGET_CPU_Z14_P, TARGET_ARCH12 to
938         TARGET_Z14, and TARGET_ARCH12_P to TARGET_Z14_P.
939         * config/s390/s390.md: Likewise. Rename also the cpu attribute
940         value from arch12 to z14.
942 2018-10-02  Uros Bizjak  <ubizjak@gmail.com>
944         * config/i386/i386.md (fxam<mode>2_i387_with_temp): Remove.
945         (isinfxf2): Ditto.
946         (isinf<mode>2): Ditto.
948 2018-10-02  Uros Bizjak  <ubizjak@gmail.com>
950         * config/i386/i386.c (ix86_emit_i387_round): Extend op1 to XFmode
951         before emitting fxam.  Perform calculations in XFmode.
953 2018-10-02  Marc Glisse  <marc.glisse@inria.fr>
955         * match.pd (((X /[ex] A) +- B) * A): New transformation.
957 2018-10-02  Marc Glisse  <marc.glisse@inria.fr>
959         PR middle-end/87319
960         * fold-const.c (fold_plusminus_mult_expr): Handle complex and vectors.
961         * tree.c (signed_or_unsigned_type_for): Handle complex.
963 2018-10-02  Jeff Law  <law@redhat.com>
965         * gimple-fold.c (get_range_strlen): Remove dead code.
967 2018-10-02  Martin Sebor  <msebor@redhat.com>
968             Jeff Law  <law@redhat.com>
970         * builtins.c (unterminated_array): Add new arguments.
971         If argument is not terminated, bubble up size and exact
972         state to callers.
973         (expand_builtin_strnlen): Detect, avoid expanding
974         and diagnose unterminated arrays.
975         (c_strlen): Fill in offset of start of unterminated strings.
976         * builtins.h (unterminated_array): Update prototype.
978 2018-10-02  Richard Biener  <rguenther@suse.de>
980         * config/i386/sse.md (reduc_plus_scal_v4df): Avoid the use
981         of haddv4df, first reduce to SSE width and exploit the fact
982         that we only need element zero with the reduction result.
983         (reduc_plus_scal_v2df): Likewise.
985 2018-10-02  Eric Botcazou  <ebotcazou@adacore.com>
987         * dojump.h (do_jump): Delete.
988         (do_jump_1): Likewise.
989         (split_comparison): Move around.
990         * dojump.c (do_jump): Make static.
991         (do_jump_1): Likewise.
992         (jumpifnot): Move around.
993         (jumpifnot_1): Likewise.
994         (jumpif): Likewise.
995         (jumpif_1): Likewise.
996         * expr.c (expand_expr_real_1): Call jumpif[not] instead of do_jump.
998 2018-10-02  Eric Botcazou  <ebotcazou@adacore.com>
1000         * reorg.c (make_return_insns): Use emit_copy_of_insn_after for the
1001         insns in the delay slot and add_insn_after for the jump insn.
1003 2018-10-02  Richard Biener  <rguenther@suse.de>
1005         * tree-inline.c (expand_call_inline): Use the location of
1006         the callee declaration for the inline-entry marker.
1007         * final.c (notice_source_line): Remove special-casing of
1008         NOTE_INSN_INLINE_ENTRY.
1010 2018-10-01  Carl Love  <cel@us.ibm.com>
1012         PR 69431
1013         * config/rs6000/rs6000-builtin.def (__builtin_mffsl): New.
1014         (__builtin_mtfsb0): New.
1015         (__builtin_mtfsb1): New.
1016         ( __builtin_set_fpscr_rn): New.
1017         (__builtin_set_fpscr_drn): New.
1018         * config/rs6000/rs6000.c (rs6000_expand_mtfsb_builtin): Add.
1019         (rs6000_expand_set_fpscr_rn_builtin): Add.
1020         (rs6000_expand_set_fpscr_drn_builtin): Add.
1021         (rs6000_expand_builtin): Add case statement entries for
1022         RS6000_BUILTIN_MTFSB0, RS6000_BUILTIN_MTFSB1,
1023         RS6000_BUILTIN_SET_FPSCR_RN, RS6000_BUILTIN_SET_FPSCR_DRN,
1024         RS6000_BUILTIN_MFFSL.
1025         (rs6000_init_builtins): Add ftype initialization and def_builtin
1026         calls for __builtin_mffsl, __builtin_mtfsb0, __builtin_mtfsb1,
1027         __builtin_set_fpscr_rn, __builtin_set_fpscr_drn.
1028         * config/rs6000.md (rs6000_mtfsb0, rs6000_mtfsb1, rs6000_mffscrn,
1029         rs6000_mffscdrn): Add define_insn.
1030         (rs6000_set_fpscr_rn, rs6000_set_fpscr_drn): Add define_expand.
1031         * doc/extend.texi: Add documentation for the builtins.
1033 2018-10-01  Richard Biener  <rguenther@suse.de>
1035         PR tree-optimization/87465
1036         * tree-ssa-loop-ivcanon.c (tree_estimate_loop_size): Fix typo
1037         causing branch miscounts.
1039 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
1041         * common/config/aarch64/aarch64-common.c (TARGET_OPTION_DEFAULT_PARAM,
1042         aarch64_option_default_param):  New.
1043         (params.h): Include.
1044         (TARGET_OPTION_VALIDATE_PARAM, aarch64_option_validate_param): New.
1045         * config/aarch64/aarch64.c (aarch64_override_options_internal): Simplify
1046         stack-clash protection validation code.
1048 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
1050         * params.c (validate_param): New.
1051         (add_params): Use it.
1052         (set_param_value): Refactor param validation into validate_param.
1053         (diagnostic.h): Include.
1054         * diagnostic.h (diagnostic_ready_p): New.
1056 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
1058         * params.c (set_param_value):
1059         Add index of parameter being validated.
1060         * common/common-target.def (option_validate_param): New.
1061         * common/common-targhooks.h (default_option_validate_param): New.
1062         * common/common-targhooks.c (default_option_validate_param): New.
1063         * doc/tm.texi.in (TARGET_OPTION_VALIDATE_PARAM): New.
1064         * doc/tm.texi: Regenerate.
1066 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
1068         PR target/86486
1069         * config/aarch64/aarch64.c (aarch64_override_options_internal):
1070         Add validation for stack-clash parameters and set defaults.
1072 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
1074         PR target/86486
1075         * configure.ac: Add stack-clash-protection-guard-size.
1076         * doc/install.texi: Document it.
1077         * config.in (DEFAULT_STK_CLASH_GUARD_SIZE): New.
1078         * params.def: Update comment for guard-size.
1079         (PARAM_STACK_CLASH_PROTECTION_GUARD_SIZE,
1080         PARAM_STACK_CLASH_PROTECTION_PROBE_INTERVAL): Update description.
1081         * configure: Regenerate.
1083 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
1085         PR target/86486
1086         * config/aarch64/aarch64.h (STACK_CLASH_MIN_BYTES_OUTGOING_ARGS,
1087         STACK_DYNAMIC_OFFSET): New.
1088         * config/aarch64/aarch64.c (aarch64_layout_frame):
1089         Update outgoing args size.
1090         (aarch64_stack_clash_protection_alloca_probe_range,
1091         TARGET_STACK_CLASH_PROTECTION_ALLOCA_PROBE_RANGE): New.
1093 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
1095         PR target/86486
1096         * explow.c (anti_adjust_stack_and_probe_stack_clash): Support custom
1097         probe ranges.
1098         * target.def (stack_clash_protection_alloca_probe_range): New.
1099         (stack_clash_protection_final_dynamic_probe): Remove.
1100         * targhooks.h (default_stack_clash_protection_alloca_probe_range) New.
1101         (default_stack_clash_protection_final_dynamic_probe): Remove.
1102         * targhooks.c: Likewise.
1103         * doc/tm.texi.in (TARGET_STACK_CLASH_PROTECTION_ALLOCA_PROBE_RANGE): New.
1104         (TARGET_STACK_CLASH_PROTECTION_FINAL_DYNAMIC_PROBE): Remove.
1105         * doc/tm.texi: Regenerate.
1107 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
1109         PR target/86486
1110         * config/aarch64/aarch64-protos.h (aarch64_output_probe_sve_stack_clash): New.
1111         * config/aarch64/aarch64.c (aarch64_output_probe_sve_stack_clash,
1112         aarch64_clamp_to_uimm12_shift): New.
1113         (aarch64_allocate_and_probe_stack_space): Add SVE specific section.
1114         * config/aarch64/aarch64.md (probe_sve_stack_clash): New.
1116 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
1118         PR target/86486
1119         * config/aarch64/aarch64.c (aarch64_layout_frame): Add assert.
1121 2018-10-01  Jeff Law  <law@redhat.com>
1122             Richard Sandiford <richard.sandiford@linaro.org>
1123             Tamar Christina  <tamar.christina@arm.com>
1125         PR target/86486
1126         * config/aarch64/aarch64.md
1127         (probe_stack_range): Add k (SP) constraint.
1128         * config/aarch64/aarch64.h (STACK_CLASH_CALLER_GUARD,
1129         STACK_CLASH_MAX_UNROLL_PAGES): New.
1130         * config/aarch64/aarch64.c (aarch64_output_probe_stack_range): Emit
1131         stack probes for stack clash.
1132         (aarch64_allocate_and_probe_stack_space): New.
1133         (aarch64_expand_prologue): Use it.
1134         (aarch64_expand_epilogue): Likewise and update IP regs re-use criteria.
1135         (aarch64_sub_sp): Add emit_move_imm optional param.
1137 2018-10-01  MCC CS <deswurstes@users.noreply.github.com>
1139         PR tree-optimization/87261
1140         * match.pd: Remove trailing whitespace.
1141         Add (x & y) | ~(x | y) -> ~(x ^ y),
1142         (~x | y) ^ (x ^ y) -> x | ~y and (x ^ y) | ~(x | y) -> ~(x & y)
1144 2018-10-01  Claudiu Zissulescu  <claziss@synopsys.com>
1146         * config/arc/arc.md (*add_n): Clean up pattern, update instruction
1147         constraints.
1148         (ashlsi3_insn): Update instruction constraints.
1149         (ashrsi3_insn): Likewise.
1150         (rotrsi3): Likewise.
1151         (add_shift): Likewise.
1152         * config/arc/constraints.md (Csz): New 32 bit constraint. It
1153         avoids placing in the limm field small constants which, otherwise,
1154         could end into a small instruction.
1156 2018-10-01  Claudiu Zissulescu  <claziss@synopsys.com>
1158         * config/arc/arc.md (maddsidi4_split): Don't use dmac if the
1159         destination register is not odd-even.
1160         (umaddsidi4_split): Likewise.
1162 2018-10-01  Richard Biener  <rguenther@suse.de>
1164         * tree-inline.c (expand_call_inline): Store origin of fn
1165         in BLOCK_ABSTRACT_ORIGIN for the inline BLOCK.
1166         * tree.c (block_ultimate_origin): Simplify and do some
1167         checking.
1169 2018-09-30  Uros Bizjak  <ubizjak@gmail.com>
1171         * config/i386/mmx.md (EMMS): New int iterator.
1172         (emms): New int attribute.
1173         (mmx_<emms>): Macroize insn from *mmx_emms and *mmx_femms using
1174         EMMS int iterator.  Explicitly declare clobbers.
1175         (mmx_emms): Remove expander.
1176         (mmx_femms): Ditto.
1177         * config/i386/predicates.md (emms_operation): Remove predicate.
1178         (vzeroall_pattern): New predicate.
1179         (vzeroupper_pattern): Rename from vzeroupper_operation.
1180         * config/i386/i386.c (ix86_avx_u128_mode_after): Use
1181         vzeroupper_pattern and vzeroall_pattern predicates.
1183 2018-09-30  Peter Bergner  <bergner@linux.ibm.com>
1185         PR rtl-optimization/86939
1186         * ira-lives.c (make_hard_regno_born): Rename from this...
1187         (make_hard_regno_live): ... to this.  Remove update to conflict
1188         information.  Update function comment.
1189         (make_hard_regno_dead): Add conflict information update.  Update
1190         function comment.
1191         (make_object_born): Rename from this...
1192         (make_object_live): ... to this.  Remove update to conflict information.
1193         Update function comment.
1194         (make_object_dead):  Add conflict information update.  Update function
1195         comment.
1196         (mark_pseudo_regno_live): Call make_object_live.
1197         (mark_pseudo_regno_subword_live): Likewise.
1198         (mark_hard_reg_dead): Update function comment.
1199         (mark_hard_reg_live): Call make_hard_regno_live.
1200         (process_bb_node_lives): Likewise.
1201         * lra-lives.c (make_hard_regno_born): Rename from this...
1202         (make_hard_regno_live): ... to this.  Remove update to conflict
1203         information.  Remove now uneeded check_pic_pseudo_p argument.
1204         Update function comment.
1205         (make_hard_regno_dead): Add check_pic_pseudo_p argument and add update
1206         to conflict information.  Update function comment.
1207         (mark_pseudo_live): Remove update to conflict information.  Update
1208         function comment.
1209         (mark_pseudo_dead): Add conflict information update.
1210         (mark_regno_live): Call make_hard_regno_live.
1211         (mark_regno_dead): Call make_hard_regno_dead with new arguement.
1212         (process_bb_lives): Call make_hard_regno_live and make_hard_regno_dead.
1214 2018-09-29  H.J. Lu  <hongjiu.lu@intel.com>
1216         PR target/87370
1217         * config/i386/i386.c (construct_container): Use TImode for
1218         BLKmode values in 2 integer registers.
1220 2018-09-29  Jeff Law  <law@redhat.com>
1222         * builtins.c (unterminated_array): Pass in c_strlen_data * to
1223         c_strlen rather than just a tree *.
1224         (c_strlen): Change NONSTR argument to a c_strlen_data pointer.
1225         Update recursive calls appropriately.  If caller did not provide a
1226         suitable data pointer, create a local one.  When a non-terminated
1227         string is discovered, bubble up information about the string via the
1228         c_strlen_data object.
1229         * builtins.h (c_strlen): Update prototype.
1230         (c_strlen_data): New structure.
1231         * gimple-fold.c (get_range_strlen): Update calls to c_strlen.
1232         For a type 2 call, if c_strlen indicates a non-terminated string
1233         use the length of the non-terminated string.
1234         (gimple_fold_builtin_stpcpy): Update calls to c_strlen.
1236 2018-09-29  Jakub Jelinek  <jakub@redhat.com>
1238         PR target/87467
1239         * config/i386/avx512fintrin.h (_mm512_abs_pd, _mm512_mask_abs_pd): Use
1240         __m512d type for __A argument rather than __m512.
1242 2018-09-28  John David Anglin  <danglin@gcc.gnu.org>
1244         * match.pd (simple_comparison): Don't optimize if either operand is
1245         a function pointer when target needs function pointer canonicalization.
1247 2018-09-28  Segher Boessenkool  <segher@kernel.crashing.org>
1249         * config/rs6000/driver-rs6000.c (asm_names): Adjust the entries for
1250         power5 .. power9 to remove indirection.
1251         * config/rs6000/rs6000.h (ASM_CPU_POWER5_SPEC, ASM_CPU_POWER6_SPEC,
1252         ASM_CPU_POWER7_SPEC, ASM_CPU_POWER8_SPEC, ASM_CPU_POWER9_SPEC,
1253         ASM_CPU_476_SPEC): Delete.
1254         (ASM_CPU_SPEC): Adjust.
1255         (EXTRA_SPECS): Delete asm_cpu_power5, asm_cpu_power6, asm_cpu_power7,
1256         asm_cpu_power8, asm_cpu_power9, asm_cpu_476.
1258 2018-09-28  Segher Boessenkool  <segher@kernel.crashing.org>
1260         * config.in: Delete HAVE_AS_DCI.
1261         * config/powerpcspe/powerpcspe.h: Treat HAVE_AS_DCI as always true.
1262         * config/rs6000/rs6000.h: Ditto.
1263         * configure.ac: Delete HAVE_AS_DCI.
1264         * configure: Regenerate.
1266 2018-09-28  Segher Boessenkool  <segher@kernel.crashing.org>
1268         * config.in (HAVE_AS_LWSYNC): Delete.
1269         * config/powerpcspe/powerpcspe.h (TARGET_LWSYNC_INSTRUCTION): Delete.
1270         * config/powerpcspe/sync.md (*lwsync): Always generate lwsync, never
1271         do it as a .long .
1272         * config/rs6000/rs6000.h (TARGET_LWSYNC_INSTRUCTION): Delete.
1273         * config/rs6000/sync.md (*lwsync): Always generate lwsync, never do it
1274         as a .long .
1275         * configure.ac: Delete HAVE_AS_LWSYNC.
1276         * configure: Regenerate.
1278 2018-09-28  Eric Botcazou  <ebotcazou@adacore.com>
1279             Pierre-Marie de Rodat  <derodat@adacore.com>
1281         * calls.c (expand_call): Try to do a tail call for thunks at -O0 too.
1282         * cgraph.h (struct cgraph_thunk_info): Add indirect_offset.
1283         (cgraph_node::create_thunk): Add indirect_offset parameter.
1284         (thunk_adjust): Likewise.
1285         * cgraph.c (cgraph_node::create_thunk): Add indirect_offset parameter
1286         and initialize the corresponding field with it.
1287         (cgraph_node::dump): Dump indirect_offset field.
1288         * cgraphclones.c (duplicate_thunk_for_node): Deal with indirect_offset.
1289         * cgraphunit.c (cgraph_node::analyze): Be prepared for external thunks.
1290         (thunk_adjust): Add indirect_offset parameter and deal with it.
1291         (cgraph_node::expand_thunk): Deal with the indirect_offset field and
1292         pass it to thunk_adjust.  Do not call the target hook if it's non-zero
1293         or if the thunk is external or local.  Fix formatting.  Do not chain
1294         the RESULT_DECL to BLOCK_VARS.  Pass the static chain to the target,
1295         if any, in the GIMPLE representation.
1296         * ipa-icf.c (sem_function::equals_wpa): Deal with indirect_offset.
1297         * lto-cgraph.c (lto_output_node): Write indirect_offset field.
1298         (input_node): Read indirect_offset field.
1299         * tree-inline.c (expand_call_inline): Pass indirect_offset field in the
1300         call to thunk_adjust.
1301         * tree-nested.c (struct nesting_info): Add thunk_p field.
1302         (create_nesting_tree): Set it.
1303         (convert_all_function_calls): Copy static chain from targets to thunks.
1304         (finalize_nesting_tree_1): Return early for thunks.
1305         (unnest_nesting_tree_1): Do not finalize thunks.
1306         (gimplify_all_functions): Do not gimplify thunks.
1308 2018-09-28  David Malcolm  <dmalcolm@redhat.com>
1310         * opt-suggestions.c (option_proposer::build_option_suggestions):
1311         Release "option_values".
1313 2018-09-28  David Malcolm  <dmalcolm@redhat.com>
1315         * coverage.c (get_coverage_counts): Convert problem-reporting dump
1316         messages from MSG_OPTIMIZED_LOCATIONS to MSG_MISSED_OPTIMIZATION.
1317         * dumpfile.c (kind_as_string): New function.
1318         (dump_loc): Rather than a hardcoded prefix of "note: ", use
1319         kind_as_string to vary the prefix based on dump_kind.
1320         (selftest::test_capture_of_dump_calls): Update for above.
1322 2018-09-28  Uros Bizjak  <ubizjak@gmail.com>
1324         * config/i386/i386.h (SSE_REGNO): Fix check for FIRST_REX_SSE_REG.
1325         (GET_SSE_REGNO): Rename from SSE_REGNO.  Update all uses for rename.
1327 2018-09-28  Uros Bizjak  <ubizjak@gmail.com>
1329         * config/i386/i386.h (CC_REGNO): Remove FPSR_REG.
1330         * config/i386/i386.c (ix86_fixed_condition_code_regs): Use
1331         INVALID_REGNUM instead of FPSR_REG.
1332         (ix86_md_asm_adjust): Do not clobber FPSR_REG.
1333         * config/i386/i386.md: Update comment of FP compares.
1334         (fldenv): Do not clobber FPSR_REG.
1336 2018-09-28  Richard Biener  <rguenther@suse.de>
1338         * tree.h (BLOCK_ORIGIN): New.
1339         * omp-expand.c (grid_expand_target_grid_body): Assign
1340         BLOCK_ORIGIN to BLOCK_ABSTRACT_ORIGIN.
1341         * tree-inline.c (remap_block): Likewise.
1342         * auto-profile.c (get_function_decl_from_block): Simplify
1343         by eliding the BLOCK_ABSTRACT_ORIGIN chasing.
1344         * langhooks.c (lhd_print_error_function): Likewise.
1345         * optinfo-emit-json.cc (optrecord_json_writer::inlining_chain_to):
1346         Likewise.
1347         * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
1348         * tree.c (block_nonartificial_location): Likewise.
1349         (block_ultimate_origin): Likewise.
1350         * tree-pretty-print.c (percent_K_format): Likewise.  Remove
1351         no longer needed LTO case.
1353 2018-09-28  Andrew Stubbs  <ams@codesourcery.com>
1354             Jan Hubicka  <jh@suse.cz>
1355             Martin Jambor  <mjambor@suse.cz>
1357         * simplify-rtx.c (simplify_merge_mask): New function.
1358         (simplify_ternary_operation): Use it, also see if VEC_MERGEs with the
1359         same masks are used in op1 or op2.
1360         (test_vec_merge): New function.
1361         (test_vector_ops): Call test_vec_merge.
1363 2018-09-28  Eric Botcazou  <ebotcazou@adacore.com>
1365         * config/sparc/sparc-protos.h (sparc_branch_cost): Declare.
1366         * config/sparc/sparc.h (BRANCH_COST): Call sparc_branch_cost.
1367         * config/sparc/sparc.c (struct processor_costs): Add branch_cost field.
1368         (cypress_costs): Set it.
1369         (supersparc_costs): Likewise.
1370         (hypersparc_costs): Likewise.
1371         (leon_cost): Likewise.
1372         (leon3_costs): Likewise.
1373         (sparclet_costs): Likewise.
1374         (ultrasparc_costs): Likewise.
1375         (ultrasparc_costs): Likewise.
1376         (niagara_costs): Likewise.
1377         (niagara2_costs): Likewise.
1378         (niagara3_costs): Likewise.
1379         (niagara4_costs): Likewise.
1380         (niagara7_costs): Likewise.
1381         (m8_costs): Likewise.
1382         (TARGET_CAN_FOLLOW_JUMP): Define.
1383         (pass_work_around_errata::gate): Minor tweak.
1384         (sparc_option_override): Remove MASK_FSMULD mask for V7 processors.
1385         Do not set both MASK_VIS4 and MASK_VIS4B for M8 processor.
1386         Automaitcally clear MASK_FSMULD mask for V7 processors.
1387         (sparc_can_follow_jump): New static function.
1388         (output_ubranch): Deal with CROSSING_JUMP_P.
1389         (sparc_use_sched_lookahead): Rewrite using switch statement.
1390         (sparc_issue_rate): Reorder.
1391         (sparc_branch_cost): New function.
1393 2018-09-27  Martin Sebor  <msebor@redhat.com>
1395         * tree.h (tree_to_shwi): Add attribute nonnull and pure.
1396         (tree_to_poly_int64, tree_to_uhwi, tree_to_poly_uint64): Same.
1397         (int_fits_type_p): Same.
1399 2018-09-27  Uros Bizjak  <ubizjak@gmail.com>
1401         * config/i386/i386.md (FPCR_REG): Remove.
1402         (UNSPEC_FLDCW): Remove.
1403         (x86_fnstcw_1): Use (const_int 0) instead of FPCR_REG.
1404         (x86_fldcw_1): Remove insn pattern.
1405         (fnstenv): Do not clobber FPCR_REG.
1406         (fldenv): Ditto.
1407         * config/i386/i386.h (FIXED_REGISTERS) Remove fpsr register.
1408         (CALL_USED_REGISTERS): Ditto.
1409         (REG_ALLOC_ORDER): Ditto.
1410         (REG_CLASS_CONTENTS): Ditto.
1411         (HI_REGISTER_NAMES): Ditto.
1412         (ADDITIONAL_REGISTER_NAMES): Use defines instead
1413         of numerical constants.
1414         * config/i386/i386.c (regclass_map): Remove fpsr register.
1415         (dbx_register_map): Ditto.
1416         (dbx64_register_map): Ditto.
1417         (svr4_dbx_register_map): Ditto.
1418         (print_reg): Do not handle FPCR_REG.
1420 2018-09-27  Segher Boessenkool  <segher@kernel.crashing.org>
1422         PR target/87149
1423         * config.in (HAVE_AS_CMPB, HAVE_AS_DFP, HAVE_AS_FPRND, HAVE_AS_MFPGPR,
1424         HAVE_AS_POPCNTB, HAVE_AS_POPCNTD, HAVE_AS_POWER8, HAVE_AS_POWER9):
1425         Delete, always treat as true.
1426         * config/powerpcspe/powerpcspe.c (rs6000_option_override_internal):
1427         Ditto.  Simplify remaining code.
1428         * config/powerpcspe/powerpcspe.h: Ditto.
1429         * config/rs6000/rs6000.c (rs6000_option_override_internal): Ditto.
1430         Simplify remaining code.
1431         (rs6000_expand_builtin): Ditto.
1432         * config/rs6000/rs6000.h: Ditto.
1433         * configure.ac: Ditto.
1434         * configure: Regenerate.
1436 2018-09-27  Martin Liska  <mliska@suse.cz>
1438         * coverage.c (get_coverage_counts): Revert the formatting
1439         of missing profile opt info.
1441 2018-09-27  Richard Biener  <rguenther@suse.de>
1443         PR debug/37801
1444         PR debug/87440
1445         * dwarf2out.c (set_block_origin_self): Do not mark outermost
1446         block as we do not output that.
1447         (gen_inlined_subroutine_die): Elide the originally outermost
1448         block, matching what we do for concrete instances.
1449         (decls_for_scope): Add parameter specifying whether to recurse
1450         to subblocks.
1452 2018-09-27  Andrew Stubbs  <ams@codesourcery.com>
1453             Tom de Vries  <tom@codesourcery.com>
1455         PR 82089
1457         * expmed.c (emit_cstore): Fix handling of result_mode == BImode and
1458         STORE_FLAG_VALUE == 1.
1460 2018-09-27  Andreas Krebbel  <krebbel@linux.ibm.com>
1462         * config/s390/s390.md (PPA_TX_ABORT, PPA_OOO_BARRIER): New
1463         constant definitions.
1464         ("tx_assist"): Replace magic number with PPA_TX_ABORT.
1465         ("*ppa"): Enable pattern also for -march=zEC12 -mno-htm.
1466         ("speculation_barrier"): New expander definition.
1468 2018-09-26  Indu Bhagat  <indu.bhagat@oracle.com>
1470         PR gcov-profile/86957
1471         * common.opt: New warning option -Wmissing-profile.
1472         * coverage.c (get_coverage_counts): Add warning for missing .gcda file.
1473         * doc/invoke.texi: Document -Wmissing-profile.
1475 2018-09-26  Jim Wilson  <jimw@sifive.com>
1477         * config/riscv/riscv.md (subsi3_extended2): Add J constraint.
1478         (negdi2, negsi2, negsi2_extended, negsi2_extended2): New.
1480 2018-09-26  Martin Sebor  <msebor@redhat.com>
1482         * tree.c (zerop): Change return type to bool.
1483         (integer_zerop, integer_onep, integer_each_onep): Same.
1484         (integer_all_onesp, integer_minus_onep, integer_pow2p): Same.
1485         (integer_nonzerop, integer_truep, tree_ctz, real_zerop): Same.
1486         (real_onep, real_minus_onep, chain_index): Same.
1487         (print_type_hash_statistics, type_list_equal): Same.
1488         * tree.h (zerop): Same.
1489         (zerop, integer_zerop, integer_onep, integer_each_onep): Same.
1490         (integer_all_onesp, integer_minus_onep, integer_pow2p): Same.
1491         (integer_nonzerop, integer_truep, tree_ctz, real_zerop): Same.
1492         (real_onep, real_minus_onep, chain_index): Same.
1493         (print_type_hash_statistics, type_list_equal): Same.
1495 2018-09-26  Jim Wilson  <jimw@sifive.com>
1497         * config/riscv/riscv.h (FUNCTION_ARG_REGNO_P): Fix comment.
1499 2018-09-26  Jakub Jelinek  <jakub@redhat.com>
1501         PR target/87414
1502         * config/i386/i386.c: Include debug.h and dwarf2out.h.
1503         (output_indirect_thunk): Emit DW_CFA_def_cfa_offset after the
1504         call.
1506 2018-09-25  Andrew Stubbs  <ams@codesourcery.com>
1508         * builtins.c (get_builtin_sync_mem): Force address mode conversion.
1510 2018-09-26  Uros Bizjak  <ubizjak@gmail.com>
1512         * config/i386/i386.h (enum reg_class): Remove FP_TOP_SSE_REGS
1513         and FP_SECOND_SSE_REGS.
1514         (REG_CLASS_NAMES): Ditto.
1515         (REG_CLASS_CONTENTS): Ditto.
1516         * config/i386/i386.c (ix86_preferred_reload_class) Do not handle
1517         FP_TOP_SSE_REGS and FP_SECOND_SSE_REGS classes.
1518         (ix86_preferred_output_reload_class): Ditto.
1519         * config/i386/i386.md (fix_trunc<mode>_i387_fisttp): Change "=&1f"
1520         clobber constraint to "=&f".
1521         (fix_truncdi_i387): Ditto.
1522         (lrintxfdi2): Ditto.
1523         (fistdi2_<rounding>): Ditto.
1524         (fpremxf4_i387): Change "=u" constraint to "=f".
1525         (fprem1xf4_i387): Ditto.
1526         (sincosxf3): Ditto.
1527         (fptanxf4_i387): Ditto.
1528         (fxtractxf3_i387): Ditto.
1529         (fscalexf4_i387): Ditto.
1530         (atan2xf3): Change "u" constraint to "f".
1531         (fyl2xxf3_i387): Ditto.
1532         (fyl2xp1xf3_i387): Ditto.
1534 2018-09-26  Uros Bizjak  <ubizjak@gmail.com>
1536         PR target/87439
1537         * config/i386/i386.h (NUM_MODES_FOR_MODE_SWITCHING): Update
1538         for removed I387_MASK_PM entity.
1541 2018-09-26  Jeff Law  <law@redhat.com>
1542         Revert
1543         2018-09-26  Alexey Neyman  <stilor@att.net>
1545         * graphite.h: Include <isl/id.h> and <isl/space.h>; these
1546         headers are no longer pulled in by <isl/val.h>.
1548 2018-09-26  Richard Biener  <rguenther@suse.de>
1550         PR debug/87443
1551         * dwarf2out.c (gen_lexical_block_die): Do not equate inline
1552         or concrete instance DIE to the tree.  Create abstract origin
1553         attributes also for concrete instances.
1555 2018-09-26  Alexey Neyman  <stilor@att.net>
1557         * graphite.h: Include <isl/id.h> and <isl/space.h>; these
1558         headers are no longer pulled in by <isl/val.h>.
1560 2018-09-26  Matthew Malcomson  <matthew.malcomson@arm.com>
1562         * config/arm/arm.c (arm_split_compare_and_swap, arm_split_atomic_op):
1563         Use new helper functions.
1564         * config/arm/sync.md (atomic_load<mode>, atomic_store<mode>):
1565         Use new helper functions.
1566         * config/arm/aarch-common-protos.h (aarch_mm_needs_acquire,
1567         aarch_mm_needs_release): New declarations.
1568         * config/arm/aarch-common.c (aarch_mm_needs_acquire,
1569         aarch_mm_needs_release): New.
1571 2018-09-26  Eric Botcazou  <ebotcazou@adacore.com>
1573         * config/arm/arm.c (arm_reorg): Skip Thumb reorg pass for thunks.
1574         (arm32_output_mi_thunk): Deal with long calls.
1576 2018-09-26  Richard Biener  <rguenther@suse.de>
1578         PR debug/87428
1579         PR debug/87362
1580         * tree-inline.c (expand_call_inline): When the location
1581         of the call is UNKNOWN_LOCATION use DECL_SOURCE_LOCATION
1582         or BUILTINS_LOCATION for the BLOCK_SOURCE_LOCATION of
1583         the inserted BLOCK to make inlined_function_outer_scope_p
1584         recognize it.
1585         * dwarf2out.c (add_call_src_coords_attributes): Do not add
1586         coords for reserved locations.
1588 2018-09-25  Segher Boessenkool  <segher@kernel.crashing.org>
1590         * config/rs6000/rs6000.md (*movcc_internal1): Use set_attr_alternative.
1591         (*call_indirect_nonlocal_sysv<mode>): Ditto.
1592         (*call_value_indirect_nonlocal_sysv<mode>): Ditto.
1593         (*sibcall_nonlocal_sysv<mode>): Ditto.
1594         (*sibcall_value_nonlocal_sysv<mode>): Ditto.
1595         (<bd>_<mode>): Ditto.
1596         (<bd>tf_<mode>): Ditto.
1598 2018-09-25  Segher Boessenkool  <segher@kernel.crashing.org>
1600         * config/rs6000/altivec.md (*altivec_mov<mode>): Write the output
1601         control string as a list of templates instead of as C code.
1602         (*altivec_movti): Ditto.
1603         * config/rs6000/darwin.md (movdf_low_di): Ditto.
1605 2018-09-25  Jim Wilson  <jimw@sifive.com>
1607         * config/riscv/riscv.c (riscv_split_symbol): Mark auipc label as weak
1608         when target symbol is weak.
1610 2018-09-25  Bernd Edlinger  <bernd.edlinger@hotmail.de>
1612         PR c/87387
1613         * builtins.c (unterminated_array): Simplify.
1614         * expr.c (string_constant): Handle SSA_NAME.  Add more exceptions
1615         where pointer arithmetic is safe.
1617 2018-09-25  Segher Boessenkool  <segher@kernel.crashing.org>
1619         PR target/86987
1620         * config/rs6000/altivec.md (altivec_vspltb): Use
1621         const_0_to_15_operand instead of u5bit_cint_operand.
1622         (*altivec_vspltb_internal): Ditto.
1623         (altivec_vspltb_direct): Ditto.
1624         (altivec_vsplth): Use const_0_to_7_operand instead of
1625         u5bit_cint_operand.
1626         (*altivec_vsplth_internal): Ditto.
1627         (altivec_vsplth_direct): Ditto.
1628         (altivec_vspltw): Use const_0_to_3_operand instead of
1629         u5bit_cint_operand.
1630         (*altivec_vspltw_internal): Ditto.
1631         (altivec_vspltw_direct): Ditto.
1632         (altivec_vspltsf): Ditto.
1633         (*altivec_vspltsf_internal): Ditto.
1634         * config/rs6000/rs6000.c (rs6000_expand_binop_builtin): Handle the
1635         various splats with the proper size immediate.  Reorder the various
1636         cases by ascending size of immediate, and put all such together.
1638 2018-09-25  Richard Biener  <rguenther@suse.de>
1640         PR debug/83941
1641         * dwarf2out.c (add_AT_external_die_ref): Remove now redundant
1642         GC-ification.
1643         (maybe_create_die_with_external_ref): Do not create
1644         DW_TAG_imported_unit here.
1645         (add_abstract_origin_attribute): Handle external BLOCK refs.
1646         (dwarf2out_abstract_function): Simplify LTO case.
1647         (dwarf2out_early_finish): Create DW_TAG_imported_unit explicitely
1648         rather than using maybe_create_die_with_external_ref.
1650 2018-09-25  Uros Bizjak  <ubizjak@gmail.com>
1652         PR target/71278
1653         * config/i386/i386.md (frndintxf2_mask_pm): Remove.
1654         (frndintxf2_mask_pm_i387): Ditto.
1655         (nearbyintxf2): Rewrite expander pattern to match rintxf2.
1656         Enable for !flag_trapping_math.
1657         (nearbyint<mode>2): Enable x87 modes for !flag_trapping_math.
1658         Enable SSE modes for TARGET_SSE4_1 and expand them with round insn.
1659         Change operand 1 predicate to nonimmediate_operand.
1660         (attr "i387_cw"): Remove mask_pm.
1661         * config/i386/i386.h (enum ix86_stack_slot): Remove SLOT_CW_MASK_PM.
1662         (enum ix86_entity): Remove I387_MASK_PM.
1663         * config/i386/i386.c (ix86_i387_mode_needed): Do not
1664         handle I387_MASK_PM.
1665         (ix86_mode_needed): Ditto.
1666         (ix86_mode_after): Ditto.
1667         (ix86_mode_entry): Ditto.
1668         (ix86_mode_exit): Ditto.
1669         (emit_i387_cw_initialization): Do not handle I387_CW_MASK_PM.
1671 2018-09-25  Jakub Jelinek  <jakub@redhat.com>
1673         * vr-values.c (vr_values::vr_values): Initialize to_remove_edges and
1674         to_update_switch_stmts to vNULL instead of calling create on them
1675         immediately.
1677 2018-09-25  Richard Biener  <rguenther@suse.de>
1679         PR tree-optimization/87402
1680         * tree-ssa-sccvn.c (SSA_VISITED): Remove unused function.
1681         (visit_phi): Re-instantiate handling of supposed to be VARYING
1682         but non-VARYING backedge value.
1684 2018-09-25  Richard Biener  <rguenther@suse.de>
1686         PR debug/83941
1687         * dwarf2out.c (struct sym_off_pair): New.
1688         (external_die_map): New global.
1689         (lookup_decl_die): When in LTO create DIEs lazily from the
1690         external_die_map.
1691         (lookup_block_die): New function, create DIEs lazily in LTO.
1692         (equate_block_to_die): New function.
1693         (dwarf2out_die_ref_for_decl): During WPA get the association
1694         from the external DIE map.
1695         (dwarf2out_register_external_die): Record mapping into the
1696         external DIE map.
1697         (maybe_create_die_with_external_ref): New function split out from
1698         DIE generation part of old dwarf2out_register_external_die.
1699         (add_abstract_origin_attribute): Do not return the DIE.  When
1700         in LTO reference externals directly.
1701         (dwarf2out_abstract_function): When in LTO ignore calls for
1702         decls with external DIEs (already present abstract instances).
1703         (gen_call_site_die): Adjust.
1704         (add_high_low_attributes): Likewise.
1705         (gen_lexical_block_die): Likewise.
1706         (gen_inlined_subroutine_die): Likewie.
1707         (gen_block_die): Likewise.
1708         (dwarf2out_inline_entry): Likewise.
1709         (dwarf2out_early_finish): In LTRANS phase create DW_TAG_imported_unit
1710         DIEs.
1712 2018-09-25  Martin Liska  <mliska@suse.cz>
1714         * ipa-fnsummary.c (estimate_node_size_and_time): Scale by two
1715         integers and not by a float value.
1717 2018-09-25  Martin Liska  <mliska@suse.cz>
1719         PR fortran/87394
1720         * dbgcnt.c (dbg_cnt_process_single_pair): Return false
1721         instead of NULL.
1722         * dumpfile.c (dump_enable_all): Remove extra parenthesis.
1723         * gcov-tool.c: Declare the function with ATTRIBUTE_NORETURN.
1724         * godump.c (go_format_type): Remove extra parenthesis.
1726 2018-09-25  Martin Liska  <mliska@suse.cz>
1728         * alias.c (set_dest_equal_p): Remove unused function.
1729         * config/i386/i386.c (def_builtin_pure2): Likewise.
1730         * diagnostic-show-locus.c (class layout): Remove
1731         unused field.
1732         (layout::layout): Likewise here.
1733         * dump-context.h (class temp_dump_context): Likewise.
1734         * dwarf2out.c (add_AT_fde_ref): Remove unused function.
1735         (add_AT_loclistsptr): Likewise.
1736         (add_AT_offset): Likewise.
1737         (get_AT_hi_pc): Likewise.
1738         (is_comdat_die): Likewise.
1739         (type_is_enum): Likewise.
1740         (ceiling): Likewise.
1741         (add_AT_vms_delta): Likewise.
1742         (is_class_die): Likewise.
1743         * edit-context.c (class line_event): Remove unused field.
1744         * graphite-sese-to-poly.c (tree_int_to_gmp): Remove
1745         unused function.
1746         * ipa-cp.c (ipa_get_vr_lat): Likewise.
1747         * lra-constraints.c (ok_for_index_p_nonstrict): Likewise.
1748         (ok_for_base_p_nonstrict): Likewise.
1749         * tree-chrec.c (is_not_constant_evolution): Likewise.
1750         (chrec_fold_poly_cst): Likewise.
1751         * tree-if-conv.c (has_pred_critical_p): Likewise.
1752         * tree-ssa-coalesce.c (print_exprs): Likewise.
1753         * tree-ssa-pre.c (bitmap_set_contains_expr): Likewise.
1754         * tree-ssa-uninit.c (is_and_or_or_p): Likewise.
1755         * tree-vrp.c (value_ranges_intersect_p): Likewise.
1756         (value_range_nonnegative_p): Likewise.
1758 2018-09-25  Martin Liska  <mliska@suse.cz>
1760         * config/powerpcspe/powerpcspe.c (rs6000_output_function_epilogue):
1761         Do not handle "GNU Pascal".
1762         * config/rs6000/rs6000.c (rs6000_output_function_epilogue):
1763         Likewise.
1764         * config/sparc/sparc.c (sparc_pass_by_reference): Remove Pascal
1765         from documentation. Likewise.
1766         * dbxout.c (dbxout_range_type): Likewise.
1767         * doc/cpp.texi: Likewise.
1768         * doc/extend.texi: Likewise.
1769         * doc/frontends.texi: Likewise.
1770         * doc/invoke.texi: Remove Pascal entry.
1771         * tree.def (CLEANUP_POINT_EXPR): Likewise.
1772         * doc/rtl.texi (MODE_FUNCTION): Remove not used entry.
1774 2018-09-25  Martin Liska  <mliska@suse.cz>
1776         PR middle-end/86078
1777         * doc/invoke.texi: Document all parameters and remove default
1778         of the parameters.
1780 2018-09-25  Ilya Leoshkevich  <iii@linux.ibm.com>
1782         PR bootstrap/87417
1783         * rtl.c (rtx_code_size): Take into account that EQ_ATTR_ALT
1784         contains HOST_WIDE_INTs when computing its size.
1786 2018-09-24  Jim Wilson  <jimw@sifive.com>
1788         PR target/87391
1789         * config/riscv/riscv.h (STACK_BOUNDARY): Test riscv_abi == ABI_ILP32E
1790         not TARGET_RVE.
1791         (ABI_STACK_BOUNDARY, MAX_ARGS_IN_REGISTERS): Likewise.
1793 2018-09-24  Andrew Pinski  <apinski@marvell.com>
1795         *  config/aarch64/aarch64.c (aarch_macro_fusion_pair_p): Don't
1796         access prev before checking it for NULLness in the
1797         AARCH64_FUSE_CMP_BRANCH case.
1799 2018-09-24  H.J. Lu  <hongjiu.lu@intel.com>
1801         PR target/82699
1802         * config/i386/i386.c (rest_of_insert_endbranch): Set
1803         endbr_queued_at_entrance to true and don't insert ENDBR if
1804         x86_function_profiler will be called.
1805         (x86_function_profiler): Insert ENDBR if endbr_queued_at_entrance
1806         is true.
1807         * config/i386/i386.h (machine_function): Add
1808         endbr_queued_at_entrance.
1810 2018-09-24  Ilya Leoshkevich  <iii@linux.ibm.com>
1812         * genattrtab.c (mk_attr_alt): Use alternative_mask.
1813         (attr_rtx_1): Adjust caching to match the new EQ_ATTR_ALT field
1814         types.
1815         (check_attr_test): Use alternative_mask.
1816         (get_attr_value): Likewise.
1817         (compute_alternative_mask): Use alternative_mask and XWINT.
1818         (make_alternative_compare): Use alternative_mask.
1819         (attr_alt_subset_p): Use XWINT.
1820         (attr_alt_subset_of_compl_p): Likewise.
1821         (attr_alt_intersection): Use alternative_mask and XWINT.
1822         (attr_alt_union): Likewise.
1823         (attr_alt_complement): Use HOST_WIDE_INT and XWINT.
1824         (mk_attr_alt): Use alternative_mask and HOST_WIDE_INT.
1825         (simplify_test_exp): Use alternative_mask and XWINT.
1826         (write_test_expr): Use alternative_mask and XWINT, adjust bit
1827         number calculation to support 64 bits.  Generate code that
1828         checks 64-bit masks.
1829         (main): Use alternative_mask.
1830         * rtl.def (EQ_ATTR_ALT): Change field types from ii to ww.
1832 2018-09-24  Ilya Leoshkevich  <iii@linux.ibm.com>
1834         PR target/80080
1835         * config/s390/s390.c (s390_emit_epilogue): Do not use PARALLEL
1836         RETURN+USE when returning via %r14.
1838 2018-09-24  Martin Liska  <mliska@suse.cz>
1840         * gcov.c (output_lines): Print colorization legend
1841         for both flag_use_colors and flag_use_hotness_colors.
1842         Reword the help.
1844 2018-09-24  Martin Liska  <mliska@suse.cz>
1846         * coverage.c (get_coverage_counts): Use warning_at
1847         with current_function_decl location. Use %qD in warning
1848         message.
1850 2018-09-24  Martin Liska  <mliska@suse.cz>
1852         * memory-block.h (memory_block_pool::release): Annotate with
1853         valgrind that the memory is not accessible.
1855 2018-09-24  Martin Liska  <mliska@suse.cz>
1857         PR sanitizer/85774
1858         * asan.c: Make asan_handled_variables extern.
1859         * asan.h: Likewise.
1860         * cfgexpand.c (expand_stack_vars): Make sure
1861         a representative is unpoison if another
1862         variable in the partition is handled by
1863         use-after-scope sanitization.
1865 2018-09-24  Richard Biener  <rguenther@suse.de>
1867         PR tree-optimization/63155
1868         * tree-ssa-propagate.c (add_ssa_edge): Avoid adding PHIs to
1869         the worklist when the edge of the respective argument isn't
1870         executable.
1872 2018-09-23   Uros Bizjak  <ubizjak@gmail.com>
1874         * config/i386/i386.h (enum reg_class): Rename MASK_REGS to
1875         ALL_MASK_REGS and MASK_EVEX_REGS to MASK_REGS.
1876         (MASK_CLASS_P): Update for rename.
1877         (MAYBE_MASK_CLASS_P): Ditto.
1878         (REG_CLASS_NAMES): Update.
1879         (REG_CLASS_CONTENT): Update.
1880         * config/i386/i386.c (regclass_map): Update for MASK_REG
1881         and ALL_MASK_REGS rename.
1882         * config/i386/constraints.md (Yk): Update for rename.
1883         (k): Ditto.
1885 2018-09-23   Uros Bizjak  <ubizjak@gmail.com>
1887         * config/i386/i386.h (enum reg_class): Remove
1888         EVEX_SSE_REGS and MOD4_SSE_REGS.
1889         (REG_CLASS_NAMES): Update.
1890         (REG_CLASS_CONTENT): Update.
1891         * config/i386/i386.c (regclass_map): Declare AVX-512 SSE
1892         registers as ALL_SSE_REGS.
1893         (ix86_additional_allocno_class_p): Remove.
1894         (TARGET_ADDITIONAL_ALLOCNO_CLASS_P): Remove.
1895         (ix86_register_priority): Lower priority of EVEX SSE registers.
1896         Use IN_RANGE macro where appropriate.
1897         (ix86_hard_regno_mode_ok): Merge AVX-5124FMAPS and
1898         AVX-5124VNNIW checks.
1899         (ix86_modes_tieable_p): Tie 512-bit SSE modes.
1900         * config/i386/sse.md (avx5124fmaddps_4fmaddps)
1901         (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddps_maskz)
1902         (avx5124fmaddps_4fmaddss, avx5124fmaddps_4fmaddss_mask)
1903         (avx5124fmaddps_4fmaddss_maskz, avx5124fmaddps_4fnmaddps)
1904         (avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddps_maskz)
1905         (avx5124fmaddps_4fnmaddss, avx5124fmaddps_4fnmaddss_mask)
1906         (avx5124fmaddps_4fnmaddss_maskz, avx5124vnniw_vp4dpwssd)
1907         (avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssd_maskz)
1908         (avx5124vnniw_vp4dpwssds, avx5124vnniw_vp4dpwssds_mask)
1909         (avx5124vnniw_vp4dpwssds_maskz): Use "v" instead of "Yh" constraint.
1910         * config/i386/constraints.md (Yh): Remove.
1912 2018-09-23   Uros Bizjak  <ubizjak@gmail.com>
1914         * config/i386/i386.c (regclass_map): Declare integer REX registers
1915         as GENERAL_REGS.
1917 2018-09-23  Gerald Pfeifer  <gerald@pfeifer.com>
1919         * doc/service.texi (Service): Switch the fsf.org link to https.
1921 2018-09-22  Chung-Ju Wu  <jasonwucj@gmail.com>
1923         PR target/86798
1924         * config/nds32/nds32.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
1925         Define to speculation_safe_value_not_needed.
1927 2018-09-21  Florian Weimer  <fweimer@redhat.com>
1929         PR middle-end/81035
1930         * doc/extend.texi (Common Function Attributes): Mention that
1931         noreturn suppresses tail call optimization.
1933 2018-09-21  Jeff Law  <law@redhat.com>
1935         * gimple-ssa-evrp.c (evrp_dom_walker::cleanup): Call
1936         vr_values::cleanup_edges_and_switches.
1937         * tree-vrp.c (to_remove_edges, to_update_switch_stmts): Moved into
1938         vr_values class.
1939         (identify_jump_threads): Remove EDGE_IGNORE handling.
1940         (execute_vrp): Move handling of to_remove_edges and
1941         to_update_switch_stmts into vr_values class member functions.
1942         * tree-vrp.h (switch_update, to_remove_edges): Remove declarations.
1943         (to_update_switch_stmts): Likewise.
1944         * vr-values.c: Include cfghooks.h.
1945         (vr_values::vr_values): Initialize to_remove_edges and
1946         to_update_switch_stmts.
1947         (vr_values::~vr_values): Verify to_remove_edges and
1948         to_update_switch_stmts are empty.
1949         (vr_values::simplify_switch_using_ranges): Set EDGE_IGNORE as needed.
1950         (vr_values::cleanup_edges_and_switches): New member function.
1951         * vr-values.h (vr_values): Add cleanup_edges_and_switches member
1952         function.  Add new data members.
1954 2018-09-21  David Malcolm  <dmalcolm@redhat.com>
1956         PR tree-optimization/87309
1957         * dumpfile.c (dump_context::begin_scope): Filter the dump_loc
1958         calls with pflags and alt_flags.
1959         (selftest::test_capture_of_dump_calls): Add test of interaction of
1960         MSG_OPTIMIZED_LOCATIONS with AUTO_DUMP_SCOPE.
1962 2018-09-21  Olivier Hainque  <hainque@adacore.com>
1964         * config.gcc: Factorize and comment inclusion of vxworks-dummy.h.
1966 2018-09-21  Olivier Hainque  <hainque@adacore.com>
1968         * config/vxworks.h (CLEAR_INSN_CACHE): #define to 1.
1970 2018-09-21  Olivier Hainque  <hainque@adacore.com>
1972         * config/vxworks.h (VXWORKS_LIBGCC_SPEC): Remove -lc_internal.
1973         Merge block comment with the one ahead of VXWORKS_LIBS_RTP. Then:
1974         (VXWORKS_LIBS_RTP): Minor reordering.
1976 2018-09-21  Olivier Hainque  <hainque@adacore.com>
1978         * config/vxworks.h (STARTFILE_PREFIX_SPEC): Define.
1979         (VXWORKS_LIBS_DIR_RTP): Remove definition and use.
1981 2018-09-21  Olivier Hainque  <hainque@adacore.com>
1983         * config/vxworks.h (SIZE_TYPE): Account for TARGET_VXWORKS64.
1984         (PTRDIFF_TYPE): Likewise.
1986 2018-09-21  Olivier Hainque  <hainque@adacore.com>
1988         * config.gcc: Enforce def of TARGET_VXWORKS64 to 1 from
1989         triplet, similar to support for VxWorks7.
1990         * config/vxworks-dummy.h: Provide a default definition
1991         of TARGET_VXWORKS64 to 0.
1993 2018-09-21  Olivier Hainque  <hainque@adacore.com>
1995         * config/vxworks.h (TARGET_VXWORKS7): Move default definition ...
1996         * config/vxworks-dummy.h: here.
1998 2018-09-21  Olivier Hainque  <hainque@adacore.com>
2000         * config.gcc: Prepend vxworks-dummy.h to tm_file for powerpc*
2002 2018-09-21  Shaokun Zhang  <zhangshaokun@hisilicon.com>
2003             Bo Zhou  <zbo.zhou@hisilicon.com>
2005         * config/aarch64/aarch64-cores.def (tsv110): New CPU.
2006         * config/aarch64/aarch64-tune.md: Regenerated.
2007         * doc/invoke.texi (AArch64 Options/-mtune): Add "tsv110".
2008         * config/aarch64/aarch64.c (tsv110_tunings): New tuning table.
2009         * config/aarch64/aarch64-cost-tables.h: Add "tsv110" extra costs.
2011 2018-09-21  Andrew Stubbs  <ams@codesourcery.com>
2012             Julian Brown  <julian@codesourcery.com>
2014         * builtins.c (get_builtin_sync_mem): Handle address spaces.
2016 2018-09-21  Eric Botcazou  <ebotcazou@adacore.com>
2018         * config/rs6000/rs6000.c (rs6000_function_ok_for_sibcall): Return false
2019         if the call takes a static chain.
2021 2018-09-21  Martin Liska  <mliska@suse.cz>
2023         * auto-profile.c (autofdo_source_profile::read): Do not
2024         set sum_all.
2025         (read_profile): Do not add working sets.
2026         (read_autofdo_file): Remove sum_all.
2027         (afdo_callsite_hot_enough_for_early_inline): Remove const
2028         qualifier.
2029         * coverage.c (struct counts_entry): Remove gcov_summary.
2030         (read_counts_file): Read new GCOV_TAG_OBJECT_SUMMARY,
2031         do not support GCOV_TAG_PROGRAM_SUMMARY.
2032         (get_coverage_counts): Remove summary and expected
2033         arguments.
2034         * coverage.h (get_coverage_counts): Likewise.
2035         * doc/gcov-dump.texi: Remove -w option.
2036         * gcov-dump.c (dump_working_sets): Remove.
2037         (main): Do not support '-w' option.
2038         (print_usage): Likewise.
2039         (tag_summary): Likewise.
2040         * gcov-io.c (gcov_write_summary): Do not dump
2041         histogram.
2042         (gcov_read_summary): Likewise.
2043         (gcov_histo_index): Remove.
2044         (gcov_histogram_merge): Likewise.
2045         (compute_working_sets): Likewise.
2046         * gcov-io.h (GCOV_TAG_OBJECT_SUMMARY): Mark
2047         it not obsolete.
2048         (GCOV_TAG_PROGRAM_SUMMARY): Mark it obsolete.
2049         (GCOV_TAG_SUMMARY_LENGTH): Adjust.
2050         (GCOV_HISTOGRAM_SIZE): Remove.
2051         (GCOV_HISTOGRAM_BITVECTOR_SIZE): Likewise.
2052         (struct gcov_summary): Simplify rapidly just
2053         to runs and sum_max fields.
2054         (gcov_histo_index): Remove.
2055         (NUM_GCOV_WORKING_SETS): Likewise.
2056         (compute_working_sets): Likewise.
2057         * gcov-tool.c (print_overlap_usage_message): Remove
2058         trailing empty line.
2059         * gcov.c (read_count_file): Read GCOV_TAG_OBJECT_SUMMARY.
2060         (output_lines): Remove program related line.
2061         * ipa-profile.c (ipa_profile): Do not consider GCOV histogram.
2062         * lto-cgraph.c (output_profile_summary): Do not stream GCOV
2063         histogram.
2064         (input_profile_summary): Do not read it.
2065         (merge_profile_summaries): And do not merge it.
2066         (input_symtab): Do not call removed function.
2067         * modulo-sched.c (sms_schedule): Do not print sum_max.
2068         * params.def (HOT_BB_COUNT_FRACTION): Reincarnate param that was
2069         removed when histogram method was invented.
2070         (HOT_BB_COUNT_WS_PERMILLE): Mention that it's used only in LTO
2071         mode.
2072         * postreload-gcse.c (eliminate_partially_redundant_load): Fix
2073         GCOV coding style.
2074         * predict.c (get_hot_bb_threshold): Use HOT_BB_COUNT_FRACTION
2075         and dump selected value.
2076         * profile.c (add_working_set): Remove.
2077         (get_working_sets): Likewise.
2078         (find_working_set): Likewise.
2079         (get_exec_counts): Do not work with working sets.
2080         (read_profile_edge_counts): Do not inform as sum_max is removed.
2081         (compute_branch_probabilities): Likewise.
2082         (compute_value_histograms): Remove argument for call of
2083         get_coverage_counts.
2084         * profile.h: Do not make gcov_summary const.
2086 2018-09-21  Monk Chiang  <sh.chiang04@gmail.com>
2088         * config.gcc (nds32*-*-*): Set TARGET_DEFAULT_TLSDESC_TRAMPOLINE=0.
2090 2018-09-21  Eric Botcazou  <ebotcazou@adacore.com>
2092         PR tree-optimization/86990
2093         * gimple-ssa-store-merging.c (imm_store_chain_info:coalesce_immediate):
2094         Check that the entire merged store group is made of constants only for
2095         overlapping stores.
2097 2018-09-20  Allan Sandfeld Jensen  <allan.jensen@qt.io>
2099         * gcc.c (LINK_COMMAND_SPEC): Handle -r like -nostdlib.
2100         (VTABLE_VERIFICATION_SPEC): Likewise.
2101         (SANITIZER_EARLY_SPEC): Likewise.
2102         (SANITIZER_SPEC): Likewise.
2103         * config/darwin.h (LINK_COMMAND_SPEC): Likewise.
2104         * doc/invoke.texi (Link Options): Document -r.
2106 2018-09-20  Richard Biener <rguenther@suse.de>
2108         PR middle-end/87054
2109         * gimplify.c (gimplify_expr): Retain alignment of
2110         addressable lvalue in dereference.
2112 2018-09-20  Alexandre Oliva <aoliva@redhat.com>
2114         PR bootstrap/87013
2115         * configure.ac: Check for .loc is_stmt support.
2116         * configure, config.in: Rebuilt.
2117         * dwarf2out.c (dwarf2out_source_line): Skip is_stmt
2118         if not supported.
2120 2018-09-20  Segher Boessenkool  <segher@kernel.crashing.org>
2122         * config/rs6000/rs6000.opt (misel=no, misel=yes): Delete.
2123         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -misel=yes and
2124         -misel=no.
2126 2018-09-20  Segher Boessenkool  <segher@kernel.crashing.org>
2128         * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
2129         VECTOR_OTHER.
2130         * config/rs6000/rs6000.c (rs6000_debug_vector_unit): Delete
2131         case VECTOR_OTHER.
2133 2018-09-20  Marek Polacek  <polacek@redhat.com>
2135         * doc/invoke.texi: Add -Wno-init-list-lifetime to C++ Language Options.
2137 2018-09-20  Richard Sandiford  <richard.sandiford@arm.com>
2139         PR tree-optimization/87288
2140         * tree-vect-loop.c (vect_analyze_loop_2): Take PEELING_FOR_GAPS
2141         into account when determining PEELING_FOR_NITERS.
2143 2018-09-20  Richard Sandiford  <richard.sandiford@arm.com>
2145         PR tree-optimization/86877
2146         * tree-vect-loop.c (vect_analyze_loop_2): Call
2147         vect_verify_datarefs_alignment.
2149 2018-09-19  Marek Polacek  <polacek@redhat.com>
2151         * doc/invoke.texi: Document -Wclass-conversion.
2153 2018-09-19  John David Anglin  <danglin@gcc.gnu.org>
2155         * config/pa/pa.c (pa_adjust_priority): Delete.
2156         (TARGET_SCHED_ADJUST_PRIORITY): Delete define.
2158         * config/pa/pa.md (atomic_storeqi): Restore deleted expander.
2159         (atomic_storehi): Likewise.
2160         (atomic_storesi): Likewise.
2161         (atomic_loaddi): Restore compare and swap exchange loop code.
2163 2018-09-19  Segher Boessenkool  <segher@kernel.crashing.org>
2165         PR rtl-optimization/86902
2166         * combine.c (try_combine): When changing the CC mode used, don't change
2167         an unrelated mode in other_insn to that new CC mode.
2169 2018-09-19  David Malcolm  <dmalcolm@redhat.com>
2171         * tree-data-ref.c (runtime_alias_check_p): Use formatted printing
2172         with %T in place of calls to dump_generic_expr.
2173         (prune_runtime_alias_test_list): Likewise.
2174         (create_runtime_alias_checks): Likewise.
2175         * tree-vect-data-refs.c (vect_check_nonzero_value): Likewise.
2176         (vect_analyze_data_ref_dependence): Likewise.
2177         (vect_slp_analyze_data_ref_dependence): Likewise.
2178         (vect_record_base_alignment): Likewise.  Use %G in place of call
2179         to dump_gimple_stmt.
2180         (vect_compute_data_ref_alignment): Likewise.
2181         (verify_data_ref_alignment): Likewise.
2182         (vect_find_same_alignment_drs): Likewise.
2183         (vect_analyze_group_access_1): Likewise.
2184         (vect_analyze_data_ref_accesses): Likewise.
2185         (dependence_distance_ge_vf): Likewise.
2186         (dump_lower_bound): Likewise.
2187         (vect_prune_runtime_alias_test_list): Likewise.
2188         (vect_find_stmt_data_reference): Likewise.
2189         (vect_analyze_data_refs): Likewise.
2190         (vect_create_addr_base_for_vector_ref): Likewise.
2191         (vect_create_data_ref_ptr): Likewise.
2192         * tree-vect-loop-manip.c (vect_set_loop_condition): Likewise.
2193         (vect_can_advance_ivs_p): Likewise.
2194         (vect_update_ivs_after_vectorizer): Likewise.
2195         (vect_gen_prolog_loop_niters): Likewise.
2196         (vect_prepare_for_masked_peels): Likewise.
2197         * tree-vect-loop.c (vect_determine_vf_for_stmt): Likewise.
2198         (vect_determine_vectorization_factor): Likewise.
2199         (vect_is_simple_iv_evolution): Likewise.
2200         (vect_analyze_scalar_cycles_1): Likewise.
2201         (vect_analyze_loop_operations): Likewise.
2202         (report_vect_op): Likewise.
2203         (vect_is_slp_reduction): Likewise.
2204         (check_reduction_path): Likewise.
2205         (vect_is_simple_reduction): Likewise.
2206         (vect_create_epilog_for_reduction): Likewise.
2207         (vect_finalize_reduction:): Likewise.
2208         (vectorizable_induction): Likewise.
2209         (vect_transform_loop_stmt): Likewise.
2210         (vect_transform_loop): Likewise.
2211         (optimize_mask_stores): Likewise.
2212         * tree-vect-patterns.c (vect_pattern_detected): Likewise.
2213         (vect_split_statement): Likewise.
2214         (vect_recog_over_widening_pattern): Likewise.
2215         (vect_recog_average_pattern): Likewise.
2216         (vect_determine_min_output_precision_1): Likewise.
2217         (vect_determine_precisions_from_range): Likewise.
2218         (vect_determine_precisions_from_users): Likewise.
2219         (vect_mark_pattern_stmts): Likewise.
2220         (vect_pattern_recog_1): Likewise.
2221         * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
2222         (vect_record_max_nunits): Likewise.
2223         (vect_build_slp_tree_1): Likewise.
2224         (vect_build_slp_tree_2): Likewise.
2225         (vect_print_slp_tree): Likewise.
2226         (vect_analyze_slp_instance): Likewise.
2227         (vect_detect_hybrid_slp_stmts): Likewise.
2228         (vect_detect_hybrid_slp_1): Likewise.
2229         (vect_slp_analyze_operations): Likewise.
2230         (vect_slp_analyze_bb_1): Likewise.
2231         (vect_transform_slp_perm_load): Likewise.
2232         (vect_schedule_slp_instance): Likewise.
2233         * tree-vect-stmts.c (vect_mark_relevant): Likewise.
2234         (vect_mark_stmts_to_be_vectorized): Likewise.
2235         (vect_init_vector_1): Likewise.
2236         (vect_get_vec_def_for_operand): Likewise.
2237         (vect_finish_stmt_generation_1): Likewise.
2238         (vect_check_load_store_mask): Likewise.
2239         (vectorizable_call): Likewise.
2240         (vectorizable_conversion): Likewise.
2241         (vectorizable_operation): Likewise.
2242         (vectorizable_load): Likewise.
2243         (vect_analyze_stmt): Likewise.
2244         (vect_is_simple_use): Likewise.
2245         (vect_get_vector_types_for_stmt): Likewise.
2246         (vect_get_mask_type_for_stmt): Likewise.
2247         * tree-vectorizer.c (increase_alignment): Likewise.
2249 2018-09-19  Andrew Stubbs  <ams@codesourcery.com>
2251         * doc/rtl.texi: Adjust vec_select description.
2252         * simplify-rtx.c (simplify_binary_operation_1): Allow VEC_SELECT to use
2253         non-constant selectors.
2255 2018-09-19  Matthew Malcomson  <matthew.malcomson@arm.com>
2257         * config/aarch64/aarch64-protos.h
2258         (aarch64_offset_9bit_signed_unscaled_p): New declaration.
2259         * config/aarch64/aarch64.md (arches): New "rcpc8_4" attribute value.
2260         (arch_enabled): Add check for "rcpc8_4" attribute value of "arch".
2261         * config/aarch64/aarch64.h (AARCH64_FL_RCPC8_4): New bitfield.
2262         (AARCH64_FL_FOR_ARCH8_4): Include AARCH64_FL_RCPC8_4.
2263         (AARCH64_FL_PROFILE): Move index so flags are ordered.
2264         (AARCH64_ISA_RCPC8_4): New flag.
2265         * config/aarch64/aarch64.c (offset_9bit_signed_unscaled_p): Renamed
2266         to aarch64_offset_9bit_signed_unscaled_p.
2267         * config/aarch64/atomics.md (atomic_store<mode>): Allow offset
2268         and use stlur.
2269         * config/aarch64/constraints.md (Ust): New constraint.
2270         * config/aarch64/predicates.md.
2271         (aarch64_9bit_offset_memory_operand): New predicate.
2272         (aarch64_rcpc_memory_operand): New predicate.
2274 2018-09-19  Eric Botcazou  <ebotcazou@adacore.com>
2276         PR rtl-optimization/87361
2277         * rtlanal.c (nonzero_bits1): Revert accidental change.
2279 2018-09-19  Richard Biener  <rguenther@suse.de>
2281         PR tree-optimization/87349
2282         PR tree-optimization/87342
2283         * tree-ssa-sccvn.c (do_rpo_vn): Iterate max_rpo computation.
2285 2018-09-18  Marek Polacek  <polacek@redhat.com>
2287         P1064R0 - Allowing Virtual Function Calls in Constant Expressions
2288         * gimple-fold.c (gimple_get_virt_method_for_vtable): Adjust assert.
2290 2018-09-18  Segher Boessenkool  <segher@kernel.crashing.org>
2292         * config/rs6000/rs6000.md: Remove old "Cygnus sibcall" comment.
2294 2018-09-18  Segher Boessenkool  <segher@kernel.crashing.org>
2296         PR rtl-optimization/86882
2297         * rtlanal.c (reg_overlap_mentioned_p): Handle CLOBBER.
2299 2018-09-18  Uros Bizjak  <ubizjak@gmail.com>
2301         * config/i386/i386.md (*<code>extend<mode>xf2): Macroize insn from
2302         *<code>extendsfxf2 and *<code>extenddfxf2 using MODEF mode iterator.
2304 2018-09-18  Jonathan Wakely  <jwakely@redhat.com>
2306         PR other/87353
2307         * doc/invoke.texi (Link Options): Fix formatting and grammar.
2309 2018-09-18  Richard Biener  <rguenther@suse.de>
2311         PR middle-end/63155
2312         * tree-ssa-coalesce.c (tree_int_map_hasher): Remove.
2313         (compute_samebase_partition_bases): Likewise.
2314         (coalesce_ssa_name): Always use compute_optimized_partition_bases.
2315         (gimple_can_coalesce_p): Simplify.
2317 2018-09-18  Hans-Peter Nilsson  <hp@bitrange.com>
2319         Handle a library implementation of ffs calling __builtin_ffs.
2320         * config/mmix/mmix.c (TARGET_INIT_LIBFUNCS): Override with...
2321         (mmix_init_libfuncs): New function: make __builtin_ffs expand
2322         to __ffsdi2.
2324 2018-09-17  David Malcolm  <dmalcolm@redhat.com>
2326         * diagnostic-show-locus.c (class layout_range): Add field
2327         "m_original_idx".
2328         (layout_range::layout_range): Add "original_idx" param and use it
2329         to initialize new field.
2330         (make_range): Use 0 for original_idx.
2331         (layout::layout): Pass in index to calls to
2332         maybe_add_location_range.
2333         (layout::maybe_add_location_range): Add param "original_idx" and
2334         pass it on to layout_range.
2335         (layout::print_any_labels): Pass on range->m_original_idx to
2336         get_text call.
2337         (gcc_rich_location::add_location_if_nearby): Use 0 for
2338         original_idx.
2339         * gcc-rich-location.h (text_range_label::get_text): Update for new
2340         param.
2341         (range_label_for_type_mismatch::get_text): Likewise.
2343 2018-09-17  Uros Bizjak  <ubizjak@gmail.com>
2345         * config/i386/i386.c (ix86_emit_i387_log1p): Emit fldln2 earlier.
2347 2018-09-17  David Malcolm  <dmalcolm@redhat.com>
2349         * gimple-ssa-sprintf.c (fmtwarn): Update for introduction of
2350         format_string_diagnostic_t.
2351         (fmtwarn_n): Likewise.
2352         * substring-locations.c
2353         (format_string_diagnostic_t::format_string_diagnostic_t) New ctor.
2354         (format_warning_n_va): Convert to...
2355         (format_string_diagnostic_t::emit_warning_n_va): ...this.
2356         (format_warning_va): Convert to...
2357         (format_string_diagnostic_t::emit_warning_va): ...this.
2358         (format_warning_at_substring): Convert to...
2359         (format_string_diagnostic_t::emit_warning): ...this.
2360         (format_warning_at_substring_n): Convert to...
2361         (format_string_diagnostic_t::emit_warning_n): ...this.
2362         * substring-locations.h (class format_string_diagnostic_t): New
2363         class.
2364         (format_warning_va): Convert to
2365         format_string_diagnostic_t::emit_warning_va.
2366         (format_warning_n_va): Convert to
2367         format_string_diagnostic_t::emit_warning_n_va.
2368         (format_warning_at_substring): Convert to
2369         format_string_diagnostic_t::emit_warning.
2370         (format_warning_at_substring_n): Convert to
2371         format_string_diagnostic_t::emit_warning_n.
2373 2018-09-17  Cesar Philippidis  <cesar@codesourcery.com>
2374             Bernd Schmidt <bernds_cb1@t-online.de>
2376         * config/nvptx/nvptx.md (atomic_fetch_<logic><mode>): Enable with
2377         SImode args.
2379 2018-09-17  Uros Bizjak  <ubizjak@gmail.com>
2381         * config/i386/i386.md (truncxf<mode>2_i387_noop_unspec): Change
2382         operand 0 predicate to nonimmediate operand.
2383         (rint<mode>2_frndint): Remove insn pattern.
2384         (rint<mode>2): Change operand 1 predicate to general_operand.
2385         Extend operand 1 to XFmode and generate rintxf2 insn.
2386         (frndintxf2_<rounding>): Rename from frndint<mode>2_<rounding>.
2387         Do not use X87MODEF mode macro.
2388         (frndintxf2_<rounding>_i387): Rename from
2389         frndint<mode>2_<rounding>_i387.  Do not use X87MODEF mode macro.
2390         (<rounding_insn><mode>2): For non-SSE modes, extend operand 1
2391         to XFmode and generate significandxf3 insn.
2393 2018-09-17  Richard Biener  <rguenther@suse.de>
2395         PR tree-optimization/87328
2396         * tree-ssa-sccvn.c (process_bb): Remove assertion about not
2397         visiting unexecutable backedges when not iterating.
2398         (do_rpo_vn): Mark all edges not executable even when not
2399         iterating.
2401 2018-09-17  Martin Jambor  <mjambor@suse.cz>
2403         PR c/63886
2404         * doc/invoke.texi (Warning Options): Likewise.
2406 2018-09-17  Richard Biener  <rguenther@suse.de>
2408         PR tree-optimization/87301
2409         * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_cleanup): Properly
2410         clean EH info from leftover copy assignments.
2412 2018-09-17  Martin Liska  <mliska@suse.cz>
2414         PR gcov-profile/85871
2415         * gcov.c (output_intermediate_file): Fix out of bounds
2416         access.
2418 2018-09-17  Vineet Gupta  <vgupta@synopsys.com>
2420         * config/arc/arc.c: Object attributes for core4 not reflected
2421         correctly.
2422         * config/arc/arc.h: Don't restrict DBNZ to core3 (core4 includes
2423         core3).
2425 2018-09-17  Alexey Brodkin  <abrodkin@synopsys.com>
2427         * config/arc/linux.h (LINK_EH_SPEC): Add missing space.
2429 2018-09-17  Martin Liska  <mliska@suse.cz>
2431         * doc/gcov.texi: Document new option --use-hotness-colors.
2432         * gcov.c (struct source_info): Declare new field.
2433         (source_info::source_info): Set default for maximum_count.
2434         (print_usage): Add new -q option.
2435         (process_args): Process it.
2436         (accumulate_line_info): Save src->maximum_count.
2437         (output_line_beginning): Make color line number if
2438         flag_use_hotness_colors is set.
2439         (output_line_details): Pass default argument value.
2440         (output_lines): Pass src->maximum_count.
2442 2018-09-17  Martin Liska  <mliska@suse.cz>
2444         * common/config/i386/i386-common.c (ix86_get_valid_option_values):
2445         Use processor_names table.
2446         * config/i386/i386.c (ix86_default_align): Use
2447         processor_cost_table for alignment values.
2448         (ix86_option_override_internal): Use processor_names.
2449         (ix86_function_specific_print): Likewise.
2450         * config/i386/i386.h (struct processor_costs):
2451         Add alignment values.
2452         (struct ptt): Remove and replace with const char *.
2453         * config/i386/x86-tune-costs.h (struct processor_costs):
2454         Declare default alignments for all costs.
2456 2018-09-17  Aldy Hernandez  <aldyh@redhat.com>
2458         * tree-vrp.c (extract_range_from_unary_expr): Do not special case
2459         symbolics or VR_VARYING ranges for ABS_EXPR.
2460         * wide-int-range.cc (wide_int_range_abs): Return positive numbers
2461         when range will wrap.
2463 2018-09-15  Eric Botcazou  <ebotcazou@adacore.com>
2465         PR middle-end/86864
2466         * cfgexpand.c (expand_gimple_basic_block): Be prepared for a BARRIER
2467         before and after a JUMP_TABLE_DATA.
2469 2018-09-14  John David Anglin  <danglin@gcc.gnu.org>
2471         PR middle-end/87188
2472         * dojump.c (do_compare_and_jump): Canonicalize function pointers
2473         when one operand is a function pointer.  Use POINTER_TYPE_P and
2474         FUNC_OR_METHOD_TYPE_P.
2475         * expr.c (do_store_flag): Use POINTER_TYPE_P and FUNC_OR_METHOD_TYPE_P.
2476         * fold-const.c (build_range_check): Likewise.
2477         * match.pd (simple_comparison): Likewise.
2479 2018-09-14  David Malcolm  <dmalcolm@redhat.com>
2481         PR c/82967
2482         * spellcheck.c (get_edit_distance_cutoff): New function.
2483         (selftest::test_edit_distance_unit_test_oneway): Rename to...
2484         (selftest::test_get_edit_distance_one_way): ...this.
2485         (selftest::test_get_edit_distance_unit): Rename to...
2486         (selftest::test_get_edit_distance_both_ways): ...this.
2487         (selftest::test_edit_distances): Move tests to this new function,
2488         and test some more pairs of strings.  Update for above renaming.
2489         (selftest::get_old_cutoff): New function.
2490         (selftest::test_get_edit_distance_cutoff): New function.
2491         (selftest::assert_suggested_for): New function.
2492         (ASSERT_SUGGESTED_FOR): New macro.
2493         (selftest::assert_not_suggested_for): New function.
2494         (ASSERT_NOT_SUGGESTED_FOR): New macro.
2495         (selftest::test_suggestions): New function.
2496         (selftest::spellcheck_c_tests): Move test_get_edit_distance_unit
2497         tests to selftest::test_edit_distances and call it.  Add calls to
2498         selftest::test_get_edit_distance_cutoff and
2499         selftest::test_suggestions.
2500         * spellcheck.h (get_edit_distance_cutoff): New function declaration.
2501         (best_match::consider): Replace hard-coded cutoff calculation with
2502         a call to...
2503         (best_match::get_cutoff): New declaration.
2504         (best_match::get_best_meaningful_candidate): Likewise.
2506 2018-09-14  Bernd Edlinger  <bernd.edlinger@hotmail.de>
2508         * builtins.c (fold_builtin_strlen): Remove TODO comment.
2510 2018-09-14  Bernd Edlinger  <bernd.edlinger@hotmail.de>
2512         revert:
2513         2018-07-30  Bernd Edlinger  <bernd.edlinger@hotmail.de>
2515         * tree-ssa-forwprop.c (simplify_builtin_call): Don't create a not NUL
2516         terminated string literal.
2518 2018-09-14  Martin Sebor  <msebor@redhat.com>
2520         * builtins.c (unterminated_array): Handle ARRAY_REF.
2521         (expand_builtin_stpcpy_1): Detect unterminated char arrays.
2522         * builtins.h (unterminated_array): Declare extern.
2523         * gimple-fold.c (gimple_fold_builtin_stpcpy): Detect unterminated
2524         arrays.
2525         (gimple_fold_builtin_sprintf): Propagate NO_WARNING to transformed
2526         calls.
2528 2018-09-14  Martin Sebor  <msebor@redhat.com>
2529             Jeff Law  <law@redhat.com>
2531         * builtins.c (unterminated_array): New.
2532         (expand_builtin_strcpy): Adjust.
2533         (expand_builtin_strcpy_args): Detect unterminated arrays.
2534         * gimple-fold.c (get_maxval_strlen): Add argument.  Detect
2535         unterminated arrays.
2536         * gimple-fold.h (get_maxval_strlen): Add argument.
2537         (gimple_fold_builtin_strcpy): Detec unterminated arrays.
2539         * gimple-fold.c (get_range_strlen): Add argument.
2540         (get_maxval_strlen): Adjust.
2541         * gimple-fold.h (get_range_strlen): Add argument.
2543 2018-09-14  Wei Xiao  <wei3.xiao@intel.com>
2545         * config/i386/movdirintrin.h: Fix copyright year.
2547 2018-09-14  Uros Bizjak  <ubizjak@gmail.com>
2549         * reg-stack.c: Include regs.h.
2550         (replace_reg): Assert that mode is MODE_FLOAT or MODE_COMPLEX_FLOAT.
2551         (emit_pop_insn): Default pop insn mode to the reg_raw_mode of
2552         FIRST_STACK_REG, not DFmode.
2553         (emit_swap_insn): Default swap insn mode to the reg_raw_mode of
2554         FIRST_STACK_REG, not XFmode.  Explicitly construct swap RTX.
2555         (change stack): Default register mode to the reg_raw_mode of
2556         FIRST_STACK_REG, not DFmode.
2557         * config/i386/i386.md (*swap<mode>): Remove insn pattern.
2558         (*swapxf): Rename from swapxf.
2560 2018-09-14  Carl Love  <cel@us.ibm.com>
2562         * config/rs6000/emmintrin.h: Add _MM_SHUFFLE2.
2563         * config/rs6000/xmmintrin.h: Add _MM_SHUFFLE.
2565 2018-09-14  Segher Boessenkool  <segher@kernel.crashing.org>
2567         PR target/87224
2568         * config/rs6000/rs6000.md (*mov<mode>_hardfloat64): Add Z to the Y
2569         alternatives.
2571 2018-09-14  Sam Tebbs  <sam.tebbs@arm.com>
2573         PR target/85628
2574         * config/aarch64/aarch64.md (*aarch64_bfxilsi_uxtw): Define.
2576 2018-09-14  Jason Merrill  <jason@redhat.com>
2578         Fix --enable-gather-detailed-mem-stats.
2579         * hash-table.c (hash_table_usage): Change from variable to function.
2580         * hash-table.h: Adjust.
2581         * Makefile.in: Add missing dependencies on hash-table.h.
2583 2018-09-14  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
2585         PR tree-optimization/87259
2586         PR lto/87283
2587         (pass_cse_reciprocals::execute): Run optimize_recip_sqrt after
2588         execute_cse_reciprocals_1 has tried transforming.
2590 2018-09-14  Aldy Hernandez  <aldyh@redhat.com>
2592         * tree-vrp.c (extract_range_from_binary_expr_1): Normalize
2593         VR_VARYING for PLUS/MINUS_EXPR.
2595 2018-09-14  Ilya Leoshkevich  <iii@linux.ibm.com>
2597         * config/s390/s390-passes.def (INSERT_PASS_BEFORE): Improve
2598         formatting.
2600 2018-09-14  Richard Biener  <rguenther@suse.de>
2602         PR middle-end/63155
2603         * tree-ssa-coalesce.c (ssa_conflicts_merge): Remove conflict
2604         bits for the merged partition.
2606 2018-09-13  Martin Sebor  <msebor@redhat.com>
2607             Bernd Edlinger  <bernd.edlinger@hotmail.de>
2609         * builtins.h (c_srlen): Add argument.
2610         * builtins.c (warn_string_no_nul): New function.
2611         (c_strlen): Add argument and use it.  Update recursive calls.
2612         Pass DECL argument to string_constant to get info on non
2613         terminated strings.  Update *NONSTR as needed.
2614         (fold_builtin_strlen): Add argument to calls to c_strlen.
2615         Warn for unterminated arrays.
2616         (warn_string_no_null): Add prototype.
2617         * expr.c (string_constant): Update arguments.  Update recursive
2618         calls appropriately.  Detect missing NUL terminator and outermost
2619         declaration its missing in.
2620         Improve checks for arrays with nonzero lower bound or elements
2621         that are not a single byte.  Simplify offset computation.
2622         Simplify checks for non-NUL terminated strings.
2623         * gimple-fold.c (get_range_strlen): Add argument to c_strlen call.
2624         * gimple-ssa-sprintf.c (get_string_length): Remove unnecessary code.
2626 2018-09-13  Bernd Edlinger  <bernd.edlinger@hotmail.de>
2628         * builtins.c (c_strlen): Handle not zero terminated STRING_CSTs
2629         correctly.
2630         * fold-const.c (c_getstr): Fix function comment.  Remove unused third
2631         argument.  Fix range checks.
2632         * fold-const.h (c_getstr): Adjust protoype.
2633         * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid folding when
2634         string is constant but contains no NUL byte.
2636         * expr.c (string_constant): Adjust function comment.
2637         Remove bogus check for zero termination.
2639         * fold-const.c (c_getstr): Clamp STRING_LENGTH to STRING_SIZE.
2641         * varasm.c (compare_constant): Compare type size of STRING_CSTs.
2642         (get_constant_size): Don't make STRING_CSTs larger than they are.
2643         (check_string_literal): New check function for STRING_CSTs.
2644         (output_constant): Use it.
2646 2018-09-13  Eric Botcazou  <ebotcazou@adacore.com>
2648         PR target/86812
2649         * config/visium/visium.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define.
2651 2018-09-13  Richard Biener  <rguenther@suse.de>
2653         PR tree-optimization/87263
2654         * tree-ssa-sccvn.c (visit_phi): Revert some earlier changes.
2655         (struct unwind_state): Add max_rpo field.
2656         (do_rpo_vn): Allow up-to-date loop state to be used when not iterating.
2657         Compute max_rpo, the max RPO number a block can be backwards reached
2658         from.  Re-write non-iterating mode to a RPO ordered worklist approach,
2659         separating it from the iterating mode.
2661 2018-09-13  Vlad Lazar  <vlad.lazar@arm.com>
2663         * haifa-sched.c (rank_for_schedule): Schedule by INSN_COST.
2664         (rfs_decision): New scheduling decision.
2666 2018-09-13  Richard Biener  <rguenther@suse.de>
2668         PR bootstrap/87134
2669         * tree-ssa-sccvn.c (vn_nary_op_insert_into): Fix assert.
2670         (vn_nary_op_insert_pieces_predicated): Do not write useless
2671         valid_dominated_by_p entry outside of the allocated storage.
2673 2018-09-13  Omar Sandoval  <osandov@osandov.com>
2674             Tom de Vries  <tdevries@suse.de>
2676         PR debug/86985
2677         * dwarf2out.c (is_c): New function.
2678         (add_subscript_info): Add DW_AT_count of 0 for C zero-length arrays.
2680 2018-09-13  Sam Tebbs  <sam.tebbs@arm.com>
2682         PR target/85628
2683         * config/aarch64/aarch64.md (*aarch64_bfxil):
2684         Define.
2685         * config/aarch64/constraints.md (Ulc): Define.
2686         * config/aarch64/aarch64-protos.h (aarch64_high_bits_all_ones_p):
2687         Define.
2688         * config/aarch64/aarch64.c (aarch64_high_bits_all_ones_p):
2689         New function.
2691 2018-09-13  Vlad Lazar  <vlad.lazar@arm.com>
2693         * config/aarch64/aarch64.h (TARGET_COMPUTE_FRAME_LAYOUT): Define.
2694         * config/aarch64/aarch64.c (aarch64_expand_prologue): Remove
2695         aarch64_layout_frame call.
2696         (aarch64_expand_epilogue): Likewise.
2697         (aarch64_initial_elimination_offset): Likewise.
2698         (aarch64_get_separate_components): Likewise.
2699         (aarch64_use_return_insn_p): Likewise.
2700         (aarch64_layout_frame): Remove unneeded check.
2702 2018-09-13  Jakub Jelinek  <jakub@redhat.com>
2704         * configure.ac: Only append
2705         " : (reconfigured) $TOPLEVEL_CONFIGURE_ARGUMENTS" to
2706         gcc_config_arguments if it was never reconfigured or last reconfigure
2707         was with different arguments.
2708         * configure: Regenerated.
2710 2018-09-13  Jakub Jelinek  <jakub@redhat.com>
2711             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
2713         PR middle-end/87290
2714         * expr.c (maybe_optimize_pow2p_mod_cmp): New function.
2715         (maybe_optimize_mod_cmp): Use it if integer_pow2p treeop1.
2717 2018-09-13  Jakub Jelinek  <jakub@redhat.com>
2719         PR tree-optimization/87287
2720         * fold-const.c (fold_binary_loc) <case EQ_EXPR>: Move signed modulo
2721         X % C == 0 to X % (unsigned) C == 0 optimization to ...
2722         * match.pd (X % C == 0): ... here.  New optimization.
2724 2018-09-12  Jakub Jelinek  <jakub@redhat.com>
2726         PR middle-end/82853
2727         * expr.h (maybe_optimize_mod_cmp): Declare.
2728         * expr.c (mod_inv): New function.
2729         (maybe_optimize_mod_cmp): New function.
2730         (do_store_flag): Use it.
2731         * cfgexpand.c (expand_gimple_cond): Likewise.
2733 2018-09-09  Cesar Philippidis  <cesar@codesourcery.com>
2734             Julian Brown  <julian@codesourcery.com>
2736         PR middle-end/86336
2737         * gimplify.c (gimplify_scan_omp_clauses): Set
2738         target_firstprivatize_array_bases in OpenACC parallel and kernels
2739         region contexts.  Remove GOMP_MAP_FIRSTPRIVATE_REFERENCE clauses from
2740         OpenACC data regions.
2742 2018-09-12  Uros Bizjak  <ubizjak@gmail.com>
2744         * config/i386/i386.md (sqrt_extend<mode>xf3_i387): Remove.
2745         (sqrt<mode>2): Extend operand 1 to XFmode and generate sqrtxf3 insn.
2747 2018-09-12  Richard Biener  <rguenther@suse.de>
2749         PR tree-optimization/87280
2750         * tree-ssa-sccvn.c (process_bb): Handle the case of executable
2751         edge but unreachable target.
2752         (do_rpo_vn): For conservatively handling a PHI only mark
2753         the backedge executable but not the block reachable.
2755 2018-09-12  Richard Biener  <rguenther@suse.de>
2757         PR tree-optimization/87266
2758         * tree-ssa-sccvn.c (do_rpo_vn): Always iterate to not yet
2759         visited blocks.
2761 2018-09-12  Andreas Krebbel  <krebbel@linux.ibm.com>
2763         * config/s390/s390.md (PFPO_RND_MODE_DFP, PFPO_RND_MODE_BFP): New
2764         constants.
2765         ("trunc<BFP:mode><DFP_ALL:mode>2")
2766         ("trunc<DFP_ALL:mode><BFP:mode>2")
2767         ("extend<BFP:mode><DFP_ALL:mode>2")
2768         ("extend<DFP_ALL:mode><BFP:mode>2"): Set proper rounding mode
2769         according to the target operand type.
2771 2018-09-12  Jakub Jelinek  <jakub@redhat.com>
2772             Andreas Krebbel  <krebbel@linux.ibm.com>
2774         PR tree-optimization/86844
2775         * gimple-ssa-store-merging.c
2776         (imm_store_chain_info::coalesce_immediate): For overlapping stores, if
2777         there are any overlapping stores in between them, make sure they are
2778         also coalesced or we give up completely.
2780 2018-09-12  Jakub Jelinek  <jakub@redhat.com>
2782         PR middle-end/87248
2783         * fold-const.c (fold_ternary_loc) <case COND_EXPR>: Verify also that
2784         BIT_AND_EXPR's second operand is a power of two.  Formatting fix.
2786 2018-09-12  Tom de Vries  <tdevries@suse.de>
2788         * common.opt (gdescribe-dies): Add option.
2789         * dwarf2out.c (add_name_and_src_coords_attributes): Add description
2790         attribute for artifical and nameless decls.
2791         (dwarf2out_register_external_die): Add description attribute to
2792         external reference die.
2793         (add_desc_attribute): New functions.
2794         (gen_subprogram_die): Add description attribute to
2795         DW_TAG_call_site_parameter.
2796         * tree-pretty-print.c (print_generic_expr_to_str): New function.
2797         * tree-pretty-print.h (print_generic_expr_to_str): Declare.
2798         * doc/invoke.texi (@item Debugging Options): Add -gdescribe-dies and
2799         -gno-describe-dies.
2800         (@item -gdescribe-dies): Add.
2802 2018-09-12  Aldy Hernandez  <aldyh@redhat.com>
2804         * tree-vrp.c (vrp_shift_undefined_p): Remove.
2805         (extract_range_from_binary_expr_1: Call
2806         wide_int_range_shift_undefined_p instead of vrp_shift_undefined_p.
2807         * wide-int-range.h (wide_int_range_shift_undefined_p): Do not
2808         depend on sign.
2810 2018-09-12  Aldy Hernandez  <aldyh@redhat.com>
2812         * gimple-ssa-warn-alloca.c
2813         (alloca_type_and_limit::alloca_type_and_limit): Initialize limit
2814         field for ALLOCA_BOUND_*_LARGE.
2816 2018-09-11  Nathan Sidwell  <nathan@acm.org>
2818         * gcc.c (load_specs, execute, run_attempt): Use %qs not '%s'.
2820 2018-09-11  Uros Bizjak  <ubizjak@gmail.com>
2822         * reg-stack.c (subst_asm_stack_regs): Call replace_reg also
2823         for clobbers.  Remove obsolete comment.
2825 2018-09-11  Uros Bizjak  <ubizjak@gmail.com>
2827         * config/i386/i386.md (define_attr "type"): Remove mpxmov, mpxmk,
2828         mpxchk, mpxld and mpxst types.
2829         (define_attr length_immediate): Remove all processing of mpx types.
2830         (define_attr prefix_0f): Ditto.
2831         (define_attr memory): Ditto.
2833 2018-09-11  Uros Bizjak  <ubizjak@gmail.com>
2835         * config/i386/i386.md (fyl2x_extend<mode>xf3_i387): Remove.
2836         (log<mode>2): Change operand 1 predicate to general_operand.
2837         Extend operand 1 to XFmode and generate logxf3 insn.
2838         (log10<mode>2): Change operand 1 predicate to general_operand.
2839         Extend operand 1 to XFmode and generate log10xf3 insn.
2840         (log2<mode>2): Change operand 1 predicate to general_operand.
2841         Extend operand 1 to XFmode and generate log2xf3 insn.
2842         (fyl2xp1_extend<mode>xf3_i387): Remove.
2843         (log1p<mode>2): Change operand 1 predicate to general_operand.
2844         Extend operand 1 to XFmode and generate log1pxf3 insn.
2845         (fxtract_extend<mode>xf3_i387): Remove.
2846         (logb<mode>2): Change operand 1 predicate to general_operand.
2847         Extend operand 1 to XFmode and generate logbxf3 insn.
2848         (ilogb<mode>2): Change operand 1 predicate to general_operand.
2849         Extend operand 1 to XFmode and generate fxtractxf3_i387 insn.
2850         (significand<mode>2): Change operand 1 predicate to general_operand.
2851         Extend operand 1 to XFmode and generate significandxf3 insn.
2853 2018-09-11  Nathan Sidwell  <nathan@acm.org>
2855         * gcc.c (perror_with_name, pfatal_with_name): Delete.
2856         (load_specs): Use fatal_error.
2857         (DELETE_IF_ORDINARY, process_command): Use error.
2858         (execute, run_attempt): Use fatal_error.
2860 2018-09-11  Andrew Stubbs  <ams@codesourcery.com>
2862         * diagnostic-core.h (sorry_at): New prototype.
2863         * diagnostic.c (sorry_at): New function.
2865 2018-09-11  Aldy Hernandez  <aldyh@redhat.com>
2867         * tree-vrp.c (extract_range_from_binary_expr_1): Treat all divisions
2868         by zero as VR_UNDEFINED.
2870 2018-09-10  Uros Bizjak  <ubizjak@gmail.com>
2872         * config/i386/i386.md (<sincos>xf2): Rename from *<sincos>xf2_i387.
2873         (*<sincos>_extend<mode>xf2_i387): Remove insn pattern.
2874         (<sincos>mode2): New expander.
2875         (sincos_extend<mode>xf3_i387): Remove insn pattern.
2876         (sincos -> sin, cos splitters): Remove splitter patterns.
2877         (sincos<mode>3): Change operand 2 predicate to general_operand.
2878         Extend operand 2 to XFmode and generate sincosxf3 insn.
2879         (fptanxf4_i387): Change mode of operands 0 and 3 to SFmode.
2880         Change operand 3 predicate to const1_operand.
2881         (fptan_extend<mode>xf4_i387): Remove insn pattern.
2882         (tanxf2): Update operands in the call to fptanxf4_i387.
2883         (tan<mode>2): Change operand 1 predicate to general_operand.
2884         Extend operand 1 to XFmode and generate tanxf3 insn.
2885         (atan2xf3): Rename from *fpatanxf3_i387.
2886         (fpatan_extend<mode>xf3_i387): Remove insn pattern.
2887         (atan2xf3): Remove expander.
2888         (atan2<mode<3):  Change operand 1 and 2 predicates to general_operand.
2889         Extend operands 1 and 2 to XFmode and generate atan2xf3 insn.
2890         (atan<mode>2): Change operand 1 predicate to general_operand.
2891         Extend operand 1 to XFmode and generate atanxf3 insn.
2893 2018-09-10  Uros Bizjak  <ubizjak@gmail.com>
2895         * config/i386/i386.md (x87/SSE constant load splitter): Use
2896         memory_operand instead of nonimmediate_operand for input operand
2897         predicate.
2899 2018-09-09  Uros Bizjak  <ubizjak@gmail.com>
2901         * config/i386/i386.md (float partial SSE register stall splitter): Move
2902         splitter near its instruction pattern.
2903         (float_extend partial SSE register stall splitter): Ditto.
2904         (float_truncate partial SSE register stall splitter): Ditto.
2906 2018-09-09  Hans-Peter Nilsson  <hp@bitrange.com>
2908         PR target/86794
2909         * config/mmix/mmix.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
2910         to speculation_safe_value_not_needed.
2912         PR target/85666
2913         * config/mmix/mmix.c (mmix_assemble_integer): Handle byte-size
2914         non-CONST_INT rtx:es using assemble_integer_with_op ".byte".
2915         (MMIX_CFUN_NEEDS_SAVED_EH_RETURN_ADDRESS): Don't call
2916         leaf_function_p, instead use has_hard_reg_initial_val.
2918 2018-09-09  Nathan Sidwell  <nathan@acm.org>
2920         * gcc.h (pfatal_with_name): Don't declare here.
2921         * gcc.c (pfatal_with_name): Make static.
2923 2018-09-09  Xianmiao Qu  <xianmiao_qu@c-sky.com>
2925         * config/csky/csky.md (*cskyv2_adddi3, *cskyv2_subdi3): Add
2926         earlyclobber.
2928 2018-09-08  John David Anglin  <danglin@gcc.gnu.org>
2930         PR rtl-optimization/85458
2931         * sel-sched.c (sel_target_adjust_priority): Allow backend adjust
2932         priority hook to reduce the priority of EXPR.
2934 2018-09-07  Uros Bizjak  <ubizjak@gmail.com>
2936         * config/i386/i386.md (float<SWI48x:mode><MODEF:mode>2) Enable
2937         DImode for x87 on 32bit targets.  Conditionally disable x87 modes
2938         with X87_ENABLE_FLOAT.  Remove preparation code.
2939         (*float<SWI48:mode><MODEF:mode>2): Rename from
2940         *float<SWI48:mode><MODEF:mode>2_mixed.  Handle x87, SSE and mixed
2941         math using "enabled" attribute.
2942         (*floatdi<MODEF:mode>2_i387): Rename from
2943         *float<SWI48x:mode><MODEF:mode>2_i387.  Handle only DImode and
2944         enable for 32bit targets only.
2945         (floatdi<X87MODEF:mode>2_i387_with_xmm pre-reload splitter): New
2946         splitter.
2947         (floatdi<X87MODEF:mode>2_i387_with_xmm): Use register_operand
2948         as operand 1 predicate.  Rewrite as define_insn_and_split.
2949         (floatdi<X87MODEF:mode>2_i387_with_xmm memory input splitter): Remove.
2951 2018-09-06  Uros Bizjak  <ubizjak@gmail.com>
2953         * reg-stack.c (get_true_reg) <case FLOAT_TRUNCATE>: Reorder
2954         to fallthru to FLOAT case.
2956 2018-09-06  Will Schmidt  <will_schmidt@vnet.ibm.com>
2958         PR target/86731
2959         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Update logic
2960         around folding of vec_sl to handle out of range shift values.
2962 2018-09-06  Uros Bizjak  <ubizjak@gmail.com>
2964         * config/i386/i386.md (fix_trunc<mode>_fisttp_i387_1): Remove.
2965         Update callers to gen_fix_trunc<mode>_i387_fisttp
2966         (fix_trunc<mode>_i387_fisttp): Change operand 0 predicate to
2967         nonimmediate_operand.
2968         (fix_trunc<mode>_i387_fisttp_with_temp): Remove insn pattern
2969         and corresponding splitters.
2970         (*fix_trunc<mode>_i387_1): Always expand with fix_trunc<mode>_i387.
2971         (fix_truncdi_i387): Change operand 0 predicate to nonimmediate_operand.
2972         (fix_truncdi_i387_with_temp): Remove insn pattern
2973         and corresponding splitters.
2974         (fix_trunc<mode>_i387): Change operand 0 predicate to
2975         nonimmediate_operand.
2976         (fix_trunc<mode>_i387_with_temp): Remove insn pattern
2977         and corresponding splitters.
2978         (*fistdi2_1): Remove.
2979         (fistdi2): Ditto.
2980         (fistdi2_with_temp): Remove insn pattern and corresponding splitters.
2981         (lrintxfdi2): Remove expander.  Reimplement as define_insn.
2982         (*fist<mode>2_1): Remove.
2983         (fist<mode>2): Ditto.
2984         (fist<mode>2_with_temp): Remove insn pattern and corresponding
2985         splitters.
2986         (lrintxf<mode>2): Remove expander.  Reimplement as define_insn.
2987         (*fist<mode>2_<rounding>_1): Always expand with fist<mode>2_<rounding>.
2988         (fistdi2_<rounding>): Change operand 0 predicate to
2989         nonimmediate_operand.
2990         (fistdi2_<rounding>_with_temp): Remove insn pattern
2991         and corresponding splitters.
2992         (fist<mode>2_<rounding>): Change operand 0 predicate to
2993         nonimmediate_operand.
2994         (fist<mode>2_<rounding>_with_temp): Remove insn pattern
2995         and corresponding splitters.
2997         (*fixuns_trunc<mode>si2_avx512f_zext): Depend on TARGET_SSE_MATH.
2999 2018-09-06  Bernd Edlinger  <bernd.edlinger@hotmail.de>
3001         * varasm.c (output_constructor_regular_field): Check TYPE_SIZE_UNIT of
3002         the init value.
3004 2018-09-06  Will Schmidt  <will_schmidt@vnet.ibm.com>
3006         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support for
3007         early gimple folding of vec_splat().
3008         * tree-vect-generic.c: Remove static from tree_vec_extract() definition.
3009         * gimple-fold.h: Add an extern define for tree_vec_extract().
3011 2018-09-06  Will Schmidt  <will_schmidt@vnet.ibm.com>
3013         * config/rs6000/rs6000.c (fold_mergehl_helper): Add types_compatible_p
3014         wrappers around TREE_TYPE comparisons.
3016 2018-09-06  Ilya Leoshkevich  <iii@linux.ibm.com>
3018         PR target/80080
3019         * config/s390/predicates.md: Add nonsym_memory_operand.
3020         * config/s390/s390.c (s390_legitimize_cs_operand): If operand
3021         contains a SYMBOL_REF, load it into an intermediate pseudo.
3022         (s390_emit_compare_and_swap): Legitimize operand.
3023         * config/s390/s390.md: Use the new nonsym_memory_operand
3024         with UNSPECV_CAS patterns.
3026 2018-09-06  Ilya Leoshkevich  <iii@linux.ibm.com>
3028         PR target/80080
3029         * config/s390/s390-passes.def: New file.
3030         * config/s390/s390-protos.h (class rtl_opt_pass): Add forward
3031         declaration.
3032         (make_pass_s390_early_mach): Add declaration.
3033         * config/s390/s390.c (make_pass_s390_early_mach):
3034         (s390_option_override): Remove dynamic registration.
3035         * config/s390/t-s390: Add s390-passes.def.
3037 2018-09-06  Ilya Leoshkevich  <iii@linux.ibm.com>
3039         * config/s390/s390.c (s390_decompose_constant_pool_ref):
3040         Remove UNSPEC_LTREL_BASE check.
3041         (annotate_constant_pool_refs): Likewise.
3042         (find_constant_pool_ref): Likewise.
3043         (find_ltrel_base): Removed.
3044         (replace_ltrel_base): Removed.
3045         (s390_mainpool_finish): Remove replace_ltrel_base call.
3046         (s390_chunkify_start): Remove pending LTREL_BASE logic.
3047         (s390_chunkify_finish): Remove replace_ltrel_base call.
3048         * config/s390/s390.md: Remove UNSPEC_LTREL_BASE.
3050 2018-09-06  Hans-Peter Nilsson  <hp@axis.com>
3052         PR target/86779
3053         * config/cris/cris.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
3054         to speculation_safe_value_not_needed.
3056 2018-09-05  Cesar Philippidis  <cesar@codesourcery.com>
3057             Bernd Schmidt  <bernds_cb1@t-online.de>
3059         * config/nvptx/nvptx-opts.h: New file.
3060         * config/nvptx/nvptx.c (nvptx_file_start): Print the correct .target.
3061         * config/nvptx/nvptx.h: Include "nvptx-opts.h".
3062         (ASM_SPEC): Define.
3063         (TARGET_SM35): New macro.
3064         * config/nvptx/nvptx.md (atomic_fetch_<logic><mode>): Enable with the
3065         correct predicate.
3066         * config/nvptx/nvptx.opt (ptx_isa, sm_30, sm_35): New enum and its
3067         values.
3068         (misa=): New option.
3069         * doc/invoke.texi (Nvidia PTX Options): Document -misa.
3071 2018-09-05  Uros Bizjak  <ubizjak@gmail.com>
3073         * config/i386/i386.md (truncdfsf2): Remove expander.
3074         (truncdfsf2_with_temp): Ditto.
3075         (truncxf<mode>2): Ditto.
3076         (*truncdfsf_fast_mixed): Remove insn pattern.
3077         (*truncdfsf_fast_i387): Ditto.
3078         (*truncdfsf_mixed): Ditto.
3079         (*truncdfsf_i387): Ditto.
3080         (*truncdfsf2_i387_1): Ditto.
3081         (*truncxfsf2_mixed): Ditto.
3082         (*truncxfdf2_mixed): Ditto.
3083         (*truncxf<mode>2_i387_noop): Ditto. Update callers
3084         to call gen_truncxf<mode>2 instead.
3085         (*truncxf<mode>2_i387): Remove.
3086         (reg->reg splitters): Remove splitter pattern.
3087         (reg->mem splitters): Ditto.
3089         (truncdfsf2): New insn pattern.
3090         (truncxf<mode>2): Ditto.
3092 2018-09-05  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
3094         * tree-ssa-math-opts.c (is_mult_by): New function.
3095         (is_square_of): Use the above.
3096         (optimize_recip_sqrt): New function.
3097         (pass_cse_reciprocals::execute): Use the above.
3099 2018-09-05  Richard Biener  <rguenther@suse.de>
3101         PR bootstrap/87134
3102         * tree-ssa-sccvn.c (rpo_elim::eliminate_push_avail): Make sure
3103         to zero-init the emplaced vec.
3105 2018-09-05  Martin Liska  <mliska@suse.cz>
3107         PR tree-optimization/87205
3108         * tree-switch-conversion.c (pass_lower_switch::execute):
3109         Group cases for switch statements.
3111 2018-09-05  Richard Biener  <rguenther@suse.de>
3113         PR tree-optimization/87217
3114         * tree-ssa-sccvn.c (vuse_valueize): New.
3115         (vn_reference_lookup_pieces): Use it.
3116         (vn_reference_lookup): Likewise.
3118 2018-09-05  Nathan Sidwell  <nathan@acm.org>
3120         PR c++/87137
3121         * stor-layout.c (place_field): Scan forwards to check last
3122         bitfield when ms_bitfield_placement is in effect.
3124 2018-09-05  Richard Biener  <rguenther@suse.de>
3126         PR bootstrap/87225
3127         * tree-vect-stmts.c (vectorizable_simd_clone_call): Fix bogus
3128         return.
3130 2018-09-05  Siddhesh Poyarekar  <siddhesh@sourceware.org>
3131             Kugan Vivekanandarajah  <kugan.vivekanandarajah@linaro.org>
3133         * config/aarch64/falkor-tag-collision-avoidance.c: New file.
3134         * config.gcc (extra_objs): Build it.
3135         * config/aarch64/t-aarch64 (falkor-tag-collision-avoidance.o):
3136         Likewise.
3137         * config/aarch64/aarch64-passes.def
3138         (pass_tag_collision_avoidance): New pass.
3139         * config/aarch64/aarch64.c (qdf24xx_tunings): Add
3140         AARCH64_EXTRA_TUNE_RENAME_LOAD_REGS to tuning_flags.
3141         (aarch64_classify_address): Remove static qualifier.
3142         (aarch64_address_info, aarch64_address_type): Move to...
3143         * config/aarch64/aarch64-protos.h: ... here.
3144         (make_pass_tag_collision_avoidance): New function.
3145         * config/aarch64/aarch64-tuning-flags.def (rename_load_regs):
3146         New tuning flag.
3148 2018-09-05  Martin Liska  <mliska@suse.cz>
3150         * doc/gcov.texi: Update documentation of humar
3151         readable mode.
3152         * gcov.c (format_count): Print one decimal place, it provides
3153         more fine number of situations like '1G' vs. '1.4G'.
3155 2018-09-05  Martin Liska  <mliska@suse.cz>
3157         PR target/87164
3158         * config/rs6000/rs6000.opt: Mark the option as Deprecated.
3159         * optc-gen.awk: Allow 'Var' for Deprecated options in order
3160         to generate a MASK value.
3162 2018-09-04  H.J. Lu  <hongjiu.lu@intel.com>
3164         PR debug/86593
3165         * dwarf2out.c (based_loc_descr): Allow hard frame pointer even
3166         if frame pointer isn't used.
3167         (compute_frame_pointer_to_fb_displacement): Likewise.
3169 2018-09-04  Jakub Jelinek  <jakub@redhat.com>
3171         PR target/87198
3172         * common/config/i386/i386-common.c (OPTION_MASK_ISA_XSAVEOPT_SET,
3173         OPTION_MASK_ISA_XSAVES_SET, OPTION_MASK_ISA_XSAVEC_SET): Use
3174         OPTION_MASK_ISA_XSAVE_SET instead of OPTION_MASK_ISA_XSAVE.
3175         (OPTION_MASK_ISA_XSAVE_UNSET): Add OPTION_MASK_ISA_XSAVES_UNSET
3176         and OPTION_MASK_ISA_XSAVEC_UNSET.
3178 2018-09-04  Max Filippov  <jcmvbkbc@gmail.com>
3180         * config/xtensa/xtensa.c (xtensa_expand_atomic): Reorder AND and
3181         XOR operations in NAND case.
3183 2018-09-04  Aldy Hernandez  <aldyh@redhat.com>
3185         * wide-int-range.cc (wide_int_range_convert): New.
3186         * wide-int-range.h (wide_int_range_convert): New.
3187         * tree-vrp.c (extract_range_from_unary_expr): Abstract wide int
3188         code into wide_int_range_convert.
3189         (extract_range_into_wide_ints): Do not munge anti range constants
3190         into the entire domain.  Just return the range back.
3192 2018-09-04  Martin Liska  <mliska@suse.cz>
3194         * genmatch.c (output_line_directive): Add new argument
3195         fnargs.
3196         (dt_simplify::gen_1): Encapsulate dump within __builtin_expect.
3198 2018-09-04  Jonathan Wakely  <jwakely@redhat.com>
3200         * doc/invoke.texi (Option Summary): Add whitespace.
3202         * doc/invoke.texi (Option Summary): Add -Waligned-new.
3204 2018-09-04  Richard Biener  <rguenther@suse.de>
3206         PR tree-optimization/87211
3207         * tree-ssa-sccvn.c (visit_phi): When value-numbering to a
3208         backedge value we're supposed to treat as VARYING also number
3209         the PHI to VARYING in case it got a different value-number already.
3211 2018-09-04  Aldy Hernandez  <aldyh@redhat.com>
3213         * tree-vrp.c (vrp_can_optimize_bit_op): Remove.
3214         (extract_range_from_binary_expr_1): Do not call
3215         vrp_can_optimize_bit_op.
3216         * wide-int-range.cc (wide_int_range_can_optimize_bit_op): Make
3217         static.
3218         (wide_int_range_get_mask_and_bounds): New.
3219         (wide_int_range_optimize_bit_op): New.
3220         (wide_int_range_bit_ior): Call wide_int_range_optimize_bit_op.
3221         (wide_int_range_bit_and): Same.
3222         * wide-int-range.h (wide_int_range_can_optimize_bit_op): Remove.
3223         (wide_int_range_optimize_bit_op): New.
3224         (wide_int_range_get_mask_and_bounds): New.
3226 2018-09-04  Richard Biener  <rguenther@suse.de>
3228         PR tree-optimization/87176
3229         * tree-ssa-sccvn.c (visit_phi): Remove redundant allsame
3230         variable.  When value-numbering a virtual PHI node make sure
3231         to not value-number to the backedge value.
3233 2018-09-04  Jonathan Wakely  <jwakely@redhat.com>
3235         * doc/extend.texi (Long Long, Hex Floats): Document support for
3236         long long and hex floats in more recent versions of ISO C++.
3238 2018-09-03  Richard Biener  <rguenther@suse.de>
3240         PR tree-optimization/87177
3241         * tree-ssa-sccvn.c (vuse_ssa_val): Revert previous change, keep
3242         cleanup.
3244 2018-09-03  Alexander Monakov  <amonakov@ispras.ru>
3246         * bb-reorder.c (edge_order): Convert to C-qsort-style
3247         tri-state comparator.
3248         (reorder_basic_blocks_simple): Change std::stable_sort to
3249         gcc_stablesort.
3251 2018-09-03  Alexander Monakov  <amonakov@ispras.ru>
3253         * tree-loop-distribution.c (offset_cmp): Convert to C-qsort-style
3254         tri-state comparator.
3255         (fuse_memset_builtins): Change std::stable_sort to gcc_stablesort.
3257 2018-09-03  Alexander Monakov  <amonakov@ispras.ru>
3259         * sort.cc (struct sort_ctx): New field 'nlim'.  Use it...
3260         (mergesort): ... here as maximum count for using netsort.
3261         (gcc_qsort): Set nlim to 3 if stable sort is requested.
3262         (gcc_stablesort): New.
3263         * system.h (gcc_stablesort): Declare.
3265 2018-09-03  Alexander Monakov  <amonakov@ispras.ru>
3267         * sort.cc (gcc_qsort) [CHECKING_P]: Call qsort_chk.
3268         * system.h (qsort): Always redirect to gcc_qsort.  Update comment.
3269         * vec.c (qsort_chk): Do not call gcc_qsort.  Update comment.
3271 2018-09-03  Segher Boessenkool  <segher@kernel.crashing.org>
3273         * config/rs6000/rs6000.md (*mov<mode>_hardfloat32): Remove %U from the
3274         lxsdx and stxsdx alternatives.
3275         (*mov<mode>_hardfloat64): Ditto.
3276         * config/rs6000/vsx.md (*vsx_extract_<mode>_store): Ditto.
3278 2018-09-03  Richard Biener  <rguenther@suse.de>
3280         PR tree-optimization/87200
3281         * tree-ssa-sccvn.c (vn_nary_build_or_lookup_1): Valueize a
3282         simplify result.
3284 2018-09-03  Martin Liska  <mliska@suse.cz>
3286         PR tree-optimization/87201
3287         * tree-switch-conversion.c (switch_decision_tree::balance_case_nodes):
3288         Fix parenthesis in an expression.
3290 2018-09-03  Richard Biener  <rguenther@suse.de>
3292         PR tree-optimization/87197
3293         * tree-ssa-sccvn.c (vn_nary_build_or_lookup_1): Mark the new def
3294         visited.  CSE the VN_INFO hashtable lookup.
3296         PR tree-optimization/87169
3297         * tree-ssa-sccvn.c (do_rpo_vn): When marking loops for not
3298         iterating make sure there's no extra backedges from irreducible
3299         regions feeding the header.  Mark the destination block
3300         executable.
3302 2018-09-03  Martin Liska  <mliska@suse.cz>
3304         PR driver/83193
3305         * common/common-target.def: Add TARGET_GET_VALID_OPTION_VALUES.
3306         * common/common-targhooks.c (default_get_valid_option_values):
3307         New function.
3308         * common/common-targhooks.h (default_get_valid_option_values):
3309         Likewise.
3310         * common/config/i386/i386-common.c: Move processor_target_table
3311         from i386.c.
3312         (ix86_get_valid_option_values): New function.
3313         (TARGET_GET_VALID_OPTION_VALUES): New macro.
3314         * config/i386/i386.c (struct ptt): Move to i386-common.c.
3315         (PTA_*): Move all defined masks into i386-common.c.
3316         (ix86_function_specific_restore): Use new processor_cost_table.
3317         * config/i386/i386.h (struct ptt): Moved from i386.c.
3318         (struct pta): Likewise.
3319         * doc/tm.texi: Document new TARGET_GET_VALID_OPTION_VALUES.
3320         * doc/tm.texi.in: Likewise.
3321         * opt-suggestions.c (option_proposer::suggest_option):
3322         Pass prefix to build_option_suggestions.
3323         (option_proposer::get_completions): Likewise.
3324         (option_proposer::build_option_suggestions): Use the new target
3325         hook.
3326         * opts.c (struct option_help_tuple): New struct.
3327         (print_filtered_help): Use the new target hook.
3329 2018-09-03  Martin Liska  <mliska@suse.cz>
3331         PR middle-end/59521
3332         * predict.c (set_even_probabilities): Add likely_edges
3333         argument and handle cases where we have precisely one
3334         likely edge.
3335         (combine_predictions_for_bb): Catch also likely_edges.
3336         (tree_predict_by_opcode): Handle gswitch statements.
3337         * tree-cfg.h (find_case_label_for_value): New declaration.
3338         (find_taken_edge_switch_expr): Likewise.
3339         * tree-switch-conversion.c (switch_decision_tree::balance_case_nodes):
3340         Find pivot in decision tree based on probabily, not by number of
3341         nodes.
3343 2018-09-02  Gerald Pfeifer  <gerald@pfeifer.com>
3345         * doc/standards.texi (Standards): Update Objective-C reference.
3347 2018-09-01  Gerald Pfeifer  <gerald@pfeifer.com>
3349         * doc/install.texi (Prerequisites): Update link for MPC.
3351 2018-09-01  Michael Matz  <matz@suse.de>
3353         PR tree-optimization/87074
3354         * gimple-loop-jam.c (unroll_jam_possible_p): Check loop exit
3355         PHIs for outer-loop uses.
3357 2018-09-01  Gerald Pfeifer  <gerald@pfeifer.com>
3359         * doc/generic.texi (OpenMP): Adjust link to openmp.org.
3360         * doc/invoke.texi (C Dialect Options): Ditto.
3362 2018-09-01  Gerald Pfeifer  <gerald@pfeifer.com>
3364         * doc/install.texi (Prerequisites): Adjust link mpfr.org.
3366 2018-08-31  Richard Biener  <rguenther@suse.de>
3368         PR tree-optimization/87168
3369         * tree-ssa-sccvn.c (SSA_VAL): Add visited output parameter.
3370         (rpo_elim::eliminate_avail): When OP was not visited it must
3371         be available.
3373 2018-08-31  David Malcolm  <dmalcolm@redhat.com>
3375         * tree-vrp.c (copy_value_range): Convert param "from" from
3376         "value_range *" to "const value_range *".
3377         (range_is_null): Likewise for param "vr".
3378         (range_int_cst_p): Likewise.
3379         (range_int_cst_singleton_p): Likewise.
3380         (symbolic_range_p): Likewise.
3381         (value_ranges_intersect_p): Likewise for both params.
3382         (value_range_nonnegative_p): Likewise for param "vr".
3383         (value_range_constant_singleton): Likewise.
3384         (vrp_set_zero_nonzero_bits): Likewise for param "ar".
3385         (extract_range_into_wide_ints): Likewise for param "vr".
3386         (extract_range_from_multiplicative_op): Likewise for params "vr0"
3387         and "vr1".
3388         (vrp_can_optimize_bit_op): Likewise.
3389         (extract_range_from_binary_expr_1): Likewise for params "vr0_" and
3390         "vr1_".
3391         (extract_range_from_unary_expr): Likewise.
3392         (debug_value_range): Likewise for param "vr".
3393         (value_range::dump): Add "const" qualifier.
3394         (vrp_prop::check_array_ref): Convert local "vr" from
3395         "value_range *" to "const value_range *".
3396         (vrp_prop::check_mem_ref): Likewise.
3397         (vrp_prop::visit_stmt): Likewise for local "old_vr".
3398         (vrp_intersect_ranges_1): Likewise for param "vr_1".
3399         (vrp_intersect_ranges): Likewise.
3400         (simplify_stmt_for_jump_threading): Likewise for local "vr".
3401         (vrp_prop::vrp_finalize): Likewise.
3402         * tree-vrp.h (value_range::dump): Add "const" qualifier.
3403         (vrp_intersect_ranges): Add "const" qualifier to params as above.
3404         (extract_range_from_unary_expr): Likewise.
3405         (value_range_constant_singleton): Likewise.
3406         (symbolic_range_p): Likewise.
3407         (copy_value_range): Likewise.
3408         (extract_range_from_binary_expr_1): Likewise.
3409         (range_int_cst_p): Likewise.
3410         (vrp_set_zero_nonzero_bits): Likewise.
3411         (range_int_cst_singleton_p): Likewise.
3413 2018-08-31  Vlad Lazar  <vlad.lazar@arm.com>
3415         * config/aarch64/arm_neon.h (vabsd_s64): New.
3416         (vnegd_s64): Likewise.
3418 2018-08-31  Martin Jambor  <mjambor@suse.cz>
3420         * ipa-cp.c (estimate_local_effects): Replace wrong MAX with MIN.
3422 2018-08-31  Martin Liska  <mliska@suse.cz>
3424         * ipa-icf.c (sem_item::add_type): Use
3425         sem_item::m_type_hash_cache.
3426         * ipa-icf.h: Move the cache from sem_item_optimizer
3427         to sem_item.
3429 2018-08-31  Nathan Sidwell  <nathan@acm.org>
3431         * doc/extend.texi (Backwards Compatibility): Remove implicit
3432         extern C leeway of () being (...).
3434 2018-08-31  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
3436         * ipa-inline.c (can_inline_edge_by_limits_p): Fix typos in comment.
3438 2018-08-31  Segher Boessenkool  <segher@kernel.crashing.org>
3440         PR target/86684
3441         PR target/87149
3442         * config/rs6000/rs6000.md (lround<mode>di2): Gate on TARGET_FPRND.
3444 2018-08-31  Jakub Jelinek  <jakub@redhat.com>
3446         PR middle-end/87138
3447         * expmed.c (expand_mult_const): Use immed_wide_int_const instead of
3448         gen_int_mode.  Formatting fixes.
3450 2018-08-30  Sandra Loosemore  <sandra@codesourcery.com>
3452         * target.def (custom_function_descriptors): Improve documentation.
3453         * doc/tm.texi.in (Trampolines): Expand discussion of function
3454         descriptors and move TARGET_CUSTOM_FUNCTION_DESCRIPTORS to the
3455         beginning of the section.
3456         * doc/tm.texi: Regenerated.
3458 2018-08-30  Jose E. Marchesi  <jose.marchesi@oracle.com>
3460         * cfg.h (class auto_edge_flag): Spell out the template-id of the
3461         base class in the initializer list.  This is a workaround for
3462         building with older GCC.
3463         (class auto_bb_flag): Likewise.
3465 2018-08-30  Aaron Sawdey  <acsawdey@linux.ibm.com>
3467         * config/rs6000/altivec.md (altivec_eq<mode>): Remove star.
3468         (altivec_vcmpequ<VI_char>_p): Remove star.
3469         * config/rs6000/rs6000-string.c (do_load_for_compare): Support
3470         vector load modes.
3471         (expand_strncmp_vec_sequence): New function.
3472         (emit_final_str_compare_vec): New function.
3473         (expand_strn_compare): Add support for vector strncmp.
3474         * config/rs6000/rs6000.opt (-mstring-compare-inline-limit): Change
3475         length specification to bytes.
3476         * config/rs6000/vsx.md (vsx_ld_elemrev_v16qi_internal): Remove star.
3477         (vcmpnezb_p): New pattern.
3478         * doc/invoke.texi (RS/6000 and PowerPC Options): Update documentation
3479         for option -mstring-compare-inline-limit.
3481 2018-08-30  Thiago Macieira  <thiago.macieira@intel.com>
3483         * config/i386/i386.c (PTA_WESTMERE): Remove PTA_AES.
3484         (PTA_SKYLAKE): Add PTA_AES.
3485         (PTA_GOLDMONT): Likewise.
3487 2018-08-29  Jan Hubicka  <jh@suse.cz>
3489         PR lto/86517
3490         * lto-opts.c (lto_write_options): Always stream PIC/PIE mode.
3491         * lto-wrapper.c (merge_and_complain): Fix merging of PIC/PIE.
3493 2018-08-29  Jan Hubicka  <jh@suse.cz>
3495         * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not follow
3496         TYPE_STUB_DECL.
3497         (hash_tree): Do not visit TYPE_STUB_DECL.
3498         * tree-streamer-out.c (write_ts_type_common_tree_pointers): Do not
3499         stream TYPE_STUB_DECL.
3500         * tree-streamer-in.c (lto_input_ts_type_common_tree_pointers): Likewise.
3501         * ipa-utils.h (type_with_linkage_p): Do not rely on TYPE_STUB_DECL
3502         after free_lang_data.
3503         (type_in_anonymous_namespace_p): Likewise.
3505 2018-08-29  Jan Hubicka  <jh@suse.cz>
3507         * sreal.h (SREAL_PART_BITS): Change to 31; remove seemingly unnecessary
3508         comment that it has to be even number.
3509         (class sreal): Change m_sig type to int32_t.
3510         * sreal.c (sreal::dump, sreal::to_int, opreator+, operator-): Use
3511         int64_t for temporary calculations.
3512         (sreal_verify_basics): Drop one bit from minimum and maximum.
3514 2018-08-30  Richard Biener  <rguenther@suse.de>
3516         PR tree-optimization/87147
3517         * tree-ssa-sccvn.c (SSA_VISITED): New function.
3518         (visit_phi): When the degenerate result is from the backedge and
3519         we didn't visit its definition yet drop to VARYING.
3520         (do_rpo_vn): Properly mark blocks with incoming backedges as executable.
3522 2018-08-29  Jan Hubicka  <jh@suse.cz>
3524         * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not walk
3525         DECL_VINDEX.
3526         (hash_tree): Likewise.
3528 2018-08-29  Jan Hubicka  <jh@suse.cz>
3530         * tree.c (find_decls_types_r): Walk also TYPE_NEXT_PTR_TO
3531         and TYPE_NEXT_REF_TO.
3533 2018-08-29  Jan Hubicka  <jh@suse.cz>
3535         * sreal.h (SREAL_PART_BITS): Change to 31; remove seemingly unnecessary
3536         comment that it has to be even number.
3537         (class sreal): Change m_sig type to int32_t.
3538         * sreal.c (sreal::dump, sreal::to_int, opreator+, operator-): Use
3539         int64_t for temporary calculations.
3540         (sreal_verify_basics): Drop one bit from minimum and maximum.
3542 2018-08-30  Tamar Christina  <tamar.christina@arm.com>
3544         * config/aarch64/aarch64.c (aarch64_expand_movmem): Set TImode max.
3546 2018-08-30  Vlad Lazar  <vlad.lazar@arm.com>
3548         PR middle-end/86995
3549         * expmed.c (canonicalize_comparison): Use wi::sub instead of wi::add
3550         if to_add is negative.
3552 2018-08-29  Bernd Edlinger  <bernd.edlinger@hotmail.de>
3554         PR middle-end/87053
3555         * builtins.c (c_strlen): Improve range checks.
3557 2018-08-29  Martin Sebor  <msebor@redhat.com>
3558             Jeff Law  <law@redhat.com>
3560         PR tree-optimization/86714
3561         PR tree-optimization/86711
3562         * builtins.c (c_strlen): Add arguments to call to string_constant.
3563         * expr.c (string_constant): Add argument.  Detect missing nul
3564         terminator and outermost declaration it's missing in.
3565         * expr.h (string_constant): Add argument.
3566         * fold-const.c (read_from_constant_string): Add arguments to call to
3567         string_constant.
3568         (c_getstr): Likewise.
3569         * tree-ssa-forwprop.c (simplify_builtin_call): Likewise.
3570         to string_constant.
3571         * tree-ssa-strlen.c (get_stridx): Likewise.
3573 2018-08-29  Jan Hubicka  <jh@suse.cz>
3575         * tree-streamer-in.c (lto_input_ts_function_decl_tree_pointers):
3576         Do not stream DECL_VINDEX.
3577         * tree-streamer-out.c (write_ts_function_decl_tree_pointers): Likewise.
3578         * tree.c (free_lang_data_in_decl): Clear DECL_VINDEX.
3579         (decl_function_context): Use DECL_VIRTUAL_P rather than DECL_VINDEX.
3581 2018-08-29  Richard Biener  <rguenther@suse.de>
3583         * tree-ssa-sccvn.c (vuse_ssa_val): Return NULL for unvisited
3584         virtual operands that are not default defs to honor region
3585         boundaries.
3586         (rpo_vn_valueize): Remove ineffective code here.
3588 2018-08-29  Richard Biener  <rguenther@suse.de>
3590         PR tree-optimization/87132
3591         * tree-ssa-alias.c (get_continuation_for_phi): Do not translate
3592         when skipping defs reachable over backedges.
3594 2018-08-29  Richard Biener  <rguenther@suse.de>
3596         * tree-core.h: Document use of deprecated_flag in SSA_NAME.
3597         * tree.h (SSA_NAME_POINTS_TO_READONLY_MEMORY): Define.
3598         * tree-into-ssa.c (pass_build_ssa::execute): Initialize
3599         function parameters SSA_NAME_POINTS_TO_READONLY_MEMORY from fnspec.
3600         * tree-ssa-sccvn.c (const_parms, init_const_parms): Remove.
3601         (vn_reference_lookup_3): Remove use of const_parms.
3602         (free_rpo_vn): Do not free const_parms.
3603         (do_rpo_vn): Do not call init_const_parms.
3604         * tree-ssa-alias.c (refs_may_alias_p_1): Honor
3605         SSA_NAME_POINTS_TO_READONLY_MEMORY.
3606         (call_may_clobber_ref_p_1): Likewise.
3608 2018-08-29  Alexander Monakov  <amonakov@ispras.ru>
3610         PR other/86726
3611         * invoke.texi (Optimization Options): List -ftree-scev-cprop.
3612         (-O): Ditto.
3613         (-ftree-scev-cprop): Document.
3615 2018-08-29  Jan Hubicka  <jh@suse.cz>
3617         * sreal.h (normalize, normalize_up, normalize_down): Add new_sig/new_exp
3618         parameters.
3619         (sreal constructor): Update.
3620         * sreal.c (sreal:operator+, sreal:operator-, sreal:operator*,
3621         sreal:operator/): Update.
3623 2018-08-29  Martin Liska  <mliska@suse.cz>
3625         * tree-switch-conversion.c (switch_conversion::expand):
3626         Strenghten assumption about gswitch statements.
3628 2018-08-29  Richard Biener  <rguenther@suse.de>
3630         PR tree-optimization/87117
3631         * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_stmt): Only
3632         re-value-number released SSA VDEFs.
3634 2018-08-29  Richard Biener  <rguenther@suse.de>
3636         PR tree-optimization/87126
3637         * tree-ssa-sccvn.c (vn_reference_insert): Remove assert.
3639 2018-08-28  Jim Wilson  <jimw@sifive.com>
3641         * config/riscv/pic.md: Rewrite.
3642         * config/riscv/riscv.c (riscv_address_insns): Return cost of 3 for
3643         invalid address.
3644         * config/riscv/riscv.md (ZERO_EXTEND_LOAD): Delete.
3645         (SOFTF, default_load, softload, softstore): New.
3647 2018-08-28  Jeff Law  <law@redhat.com>
3649         * fold-const.c (fold_binary_loc): Remove recently added assert.
3651 2018-08-28  Joern Rennecke  <joern.rennecke@riscy-ip.com>
3653         * genpreds.c (write_predicate_subfunction): Also add ATTRIBUTE_UNUSED
3654         to OP parmeter of generated function.
3656 2018-08-28  MCC CS  <deswurstes@users.noreply.github.com>
3658         PR tree-optimization/87009
3659         * match.pd: Add boolean optimizations.
3661 2018-08-28  Martin Sebor  <msebor@redhat.com>
3663         PR middle-end/86631
3664         * calls.c (alloc_max_size): Treat HOST_WIDE_INT special.
3665         * gimple-ssa-warn-alloca.c (adjusted_warn_limit): New function.
3666         (pass_walloca::gate): Use it.
3667         (alloca_call_type): Same.
3668         (pass_walloca::execute): Same.
3669         * stor-layout.c (layout_decl): Treat HOST_WIDE_INT special.
3671 2018-08-28  David Malcolm  <dmalcolm@redhat.com>
3673         * dumpfile.h (ATTRIBUTE_GCC_DUMP_PRINTF): Change version check on
3674         GCC_VERSION for usage of "__gcc_dump_printf__" format from
3675         >= 3005 to >= 9000.
3677 2018-08-28  Richard Biener  <rguenther@suse.de>
3679         PR tree-optimization/87124
3680         * tree-ssa-sccvn.c (vn_lookup_simplify_result): Guard against
3681         constants before looking up avail.
3683 2018-08-28  Jakub Jelinek  <jakub@redhat.com>
3685         PR middle-end/87099
3686         * calls.c (maybe_warn_nonstring_arg): Punt early if
3687         warn_stringop_overflow is zero.  Don't call get_range_strlen
3688         on 3rd argument, keep iterating until lenrng[1] is INTEGER_CST.
3689         Swap comparison operands to have constants on rhs.  Only use
3690         lenrng[1] if non-NULL and INTEGER_CST.  Don't uselessly
3691         increment lenrng[0].
3693 2018-08-28  Richard Sandiford  <richard.sandiford@arm.com>
3695         * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Fix unguarded
3696         use of tree_to_shwi.  Remove duplicated test for the size being
3697         a whole number of bytes.
3699 2018-08-28  Richard Biener  <rguenther@suse.de>
3701         PR tree-optimization/87117
3702         * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_cleanup):
3703         Handle removed stmt without LHS (GIMPLE_NOP).
3705 2018-08-28  Richard Biener  <rguenther@suse.de>
3707         PR tree-optimization/87117
3708         * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Exclude
3709         void which is is_gimple_reg_type by checking for COMPLETE_TYPE_P.
3711 2018-08-28  Richard Biener  <rguenther@suse.de>
3713         PR tree-optimization/87117
3714         * tree-ssa-pre.c (compute_avail): Do not make expressions
3715         with predicated values available.
3716         (get_expr_value_id): Assert we do not run into predicated value
3717         expressions.
3719 2018-08-28  Richard Biener  <rguenther@suse.de>
3721         PR tree-optimization/87117
3722         * tree-ssa-operands.c (add_stmt_operand): STRING_CST may
3723         get virtual operands.
3724         (get_expr_operands): Handle STRING_CST like other decls.
3726 2018-08-28  Martin Liska  <mliska@suse.cz>
3728         * tree.h: Update documentation of fndecl_built_in_p
3729         functions.
3732 2018-08-27  Jeff Law  <law@redhat.com>
3733         PR tree-optimization/87110
3734         * tree-ssa-dse.c (compute_trims): Handle non-constant
3735         TYPE_SIZE_UNIT.
3737 2018-08-27  Martin Sebor  <msebor@redhat.com>
3739         PR tree-optimization/86914
3740         * tree-ssa-strlen.c (maybe_set_strlen_range): Avoid MEM_REF.
3742 2018-08-27  Martin Sebor  <msebor@redhat.com>
3744         PR tree-optimization/87112
3745         * builtins.c (expand_builtin_strnlen): Convert c_strlen result to
3746         the type of the bound argument.
3748 2018-08-27  Jeff Law  <law@redhat.com>
3750         * tree-ssa-dse.c (compute_trims): Handle case where the reference's
3751         type does not have a TYPE_SIZE_UNIT.
3753 2018-08-27  Steve Ellcey  <sellcey@cavium.com>
3755         * config/aarch64/aarch64-speculation.cc: Replace include of cfg.h
3756         with include of backend.h.
3758 2018-08-27  Richard Biener  <rguenther@suse.de>
3760         PR tree-optimization/86927
3761         * tree-vect-loop.c (vect_create_epilog_for_reduction): Properly
3762         use const cond reduction code.
3764 2018-08-27  Alexander Monakov  <amonakov@ispras.ru>
3766         PR tree-optimization/85758
3767         * match.pd ((X & Y) ^ Y): Add :s qualifier to inner expression.
3769 2018-08-27  David Malcolm  <dmalcolm@redhat.com>
3771         PR c++/87091
3772         * diagnostic-show-locus.c (class layout_range): Update for
3773         conversion of show_caret_p to a tri-state.
3774         (layout_range::layout_range): Likewise.
3775         (make_range): Likewise.
3776         (layout::maybe_add_location_range): Likewise.
3777         (layout::should_print_annotation_line_p): Don't show annotation
3778         lines for ranges that are SHOW_LINES_WITHOUT_RANGE.
3779         (layout::get_state_at_point): Update for conversion of
3780         show_caret_p to a tri-state.  Bail out early for
3781         SHOW_LINES_WITHOUT_RANGE, so that such ranges don't affect
3782         underlining or source colorization.
3783         (gcc_rich_location::add_location_if_nearby): Update for conversion
3784         of show_caret_p to a tri-state.
3785         (selftest::test_one_liner_multiple_carets_and_ranges): Likewise.
3786         (selftest::test_one_liner_fixit_replace_equal_secondary_range):
3787         Likewise.
3788         (selftest::test_one_liner_labels): Likewise.
3789         * gcc-rich-location.c (gcc_rich_location::add_expr): Update for
3790         conversion of show_caret_p to a tri-state.
3791         * pretty-print.c (text_info::set_location): Likewise.
3792         * pretty-print.h (text_info::set_location): Likewise.
3793         * substring-locations.c (format_warning_n_va): Likewise.
3794         * tree-diagnostic.c (default_tree_printer): Likewise.
3795         * tree-pretty-print.c (newline_and_indent): Likewise.
3797 2018-08-27  David Malcolm  <dmalcolm@redhat.com>
3799         PR c++/87091
3800         * diagnostic-show-locus.c (get_line_span_for_fixit_hint): Show the
3801         line above for line-insertion fix-it hints.
3802         (selftest::test_fixit_insert_containing_newline): Update the
3803         expected results, and add a test with line-numbering enabled.
3805 2018-08-27  Martin Liska  <mliska@suse.cz>
3807         PR sanitizer/86962
3808         * sanopt.c (sanitize_rewrite_addressable_params): Ignore
3809         params with DECL_HAS_VALUE_EXPR_P.
3811 2018-08-27  Martin Liska  <mliska@suse.cz>
3813         * config/i386/i386.c (ix86_expand_set_or_movmem): Dump
3814         selected expansion strategy.
3816 2018-08-27  Martin Liska  <mliska@suse.cz>
3818         * builtins.h (is_builtin_fn): Remove and fndecl_built_in_p.
3819         * builtins.c (is_builtin_fn): Likewise.
3820         * attribs.c (diag_attr_exclusions): Use new function
3821         fndecl_built_in_p and remove check for FUNCTION_DECL if
3822         possible.
3823         (builtin_mathfn_code): Likewise.
3824         (fold_builtin_expect): Likewise.
3825         (fold_call_expr): Likewise.
3826         (fold_builtin_call_array): Likewise.
3827         (fold_call_stmt): Likewise.
3828         (set_builtin_user_assembler_name): Likewise.
3829         (is_simple_builtin): Likewise.
3830         * calls.c (gimple_alloca_call_p): Likewise.
3831         (maybe_warn_nonstring_arg): Likewise.
3832         * cfgexpand.c (expand_call_stmt): Likewise.
3833         * cgraph.c (cgraph_update_edges_for_call_stmt_node): Likewise.
3834         (cgraph_edge::verify_corresponds_to_fndecl): Likewise.
3835         (cgraph_node::verify_node): Likewise.
3836         * cgraphclones.c (build_function_decl_skip_args): Likewise.
3837         (cgraph_node::create_clone): Likewise.
3838         * config/arm/arm.c (arm_insert_attributes): Likewise.
3839         * config/i386/i386.c (ix86_gimple_fold_builtin): Likewise.
3840         * dse.c (scan_insn): Likewise.
3841         * expr.c (expand_expr_real_1): Likewise.
3842         * fold-const.c (operand_equal_p): Likewise.
3843         (fold_binary_loc): Likewise.
3844         * gimple-fold.c (gimple_fold_stmt_to_constant_1): Likewise.
3845         * gimple-low.c (lower_stmt): Likewise.
3846         * gimple-pretty-print.c (dump_gimple_call): Likewise.
3847         * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
3848         Likewise.
3849         * gimple.c (gimple_build_call_from_tree): Likewise.
3850         (gimple_call_builtin_p): Likewise.
3851         (gimple_call_combined_fn): Likewise.
3852         * gimplify.c (gimplify_call_expr): Likewise.
3853         (gimple_boolify): Likewise.
3854         (gimplify_modify_expr): Likewise.
3855         (gimplify_addr_expr): Likewise.
3856         * hsa-gen.c (gen_hsa_insns_for_call): Likewise.
3857         * ipa-cp.c (determine_versionability): Likewise.
3858         * ipa-fnsummary.c (compute_fn_summary): Likewise.
3859         * ipa-param-manipulation.c (ipa_modify_formal_parameters): Likewise.
3860         * ipa-split.c (visit_bb): Likewise.
3861         (split_function): Likewise.
3862         * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
3863         * lto-cgraph.c (input_node): Likewise.
3864         * lto-streamer-out.c (write_symbol): Likewise.
3865         * omp-low.c (setjmp_or_longjmp_p): Likewise.
3866         (lower_omp_1): Likewise.
3867         * predict.c (strip_predict_hints): Likewise.
3868         * print-tree.c (print_node): Likewise.
3869         * symtab.c (symtab_node::output_to_lto_symbol_table_p): Likewise.
3870         * trans-mem.c (is_tm_irrevocable): Likewise.
3871         (is_tm_load): Likewise.
3872         (is_tm_simple_load): Likewise.
3873         (is_tm_store): Likewise.
3874         (is_tm_simple_store): Likewise.
3875         (is_tm_abort): Likewise.
3876         (tm_region_init_1): Likewise.
3877         * tree-call-cdce.c (gen_shrink_wrap_conditions): Likewise.
3878         * tree-cfg.c (verify_gimple_call): Likewise.
3879         (move_stmt_r): Likewise.
3880         (stmt_can_terminate_bb_p): Likewise.
3881         * tree-eh.c (lower_eh_constructs_2): Likewise.
3882         * tree-if-conv.c (if_convertible_stmt_p): Likewise.
3883         * tree-inline.c (remap_gimple_stmt): Likewise.
3884         (copy_bb): Likewise.
3885         (estimate_num_insns): Likewise.
3886         (fold_marked_statements): Likewise.
3887         * tree-sra.c (scan_function): Likewise.
3888         * tree-ssa-ccp.c (surely_varying_stmt_p): Likewise.
3889         (optimize_stack_restore): Likewise.
3890         (pass_fold_builtins::execute): Likewise.
3891         * tree-ssa-dce.c (mark_stmt_if_obviously_necessary): Likewise.
3892         (mark_all_reaching_defs_necessary_1): Likewise.
3893         * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Likewise.
3894         * tree-ssa-forwprop.c (simplify_builtin_call): Likewise.
3895         (pass_forwprop::execute): Likewise.
3896         * tree-ssa-loop-im.c (stmt_cost): Likewise.
3897         * tree-ssa-math-opts.c (pass_cse_reciprocals::execute): Likewise.
3898         * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Likewise.
3899         * tree-ssa-strlen.c (get_string_length): Likewise.
3900         * tree-ssa-structalias.c (handle_lhs_call): Likewise.
3901         (find_func_aliases_for_call): Likewise.
3902         * tree-ssa-ter.c (find_replaceable_in_bb): Likewise.
3903         * tree-stdarg.c (optimize_va_list_gpr_fpr_size): Likewise.
3904         * tree-tailcall.c (find_tail_calls): Likewise.
3905         * tree.c (need_assembler_name_p): Likewise.
3906         (free_lang_data_in_decl): Likewise.
3907         (get_call_combined_fn): Likewise.
3908         * ubsan.c (is_ubsan_builtin_p): Likewise.
3909         * varasm.c (incorporeal_function_p): Likewise.
3910         * tree.h (DECL_BUILT_IN): Remove and replace with
3911         fndecl_built_in_p.
3912         (DECL_BUILT_IN_P): Transfort to fndecl_built_in_p.
3913         (fndecl_built_in_p): New.
3915 2018-08-27  Martin Liska  <mliska@suse.cz>
3917         PR tree-optimization/86847
3918         * tree-switch-conversion.c (switch_decision_tree::dump_case_nodes):
3919         Dump also subtree probability.
3920         (switch_decision_tree::do_jump_if_equal): New function.
3921         (switch_decision_tree::emit_case_nodes): Handle special
3922         situations in balanced tree that can be emitted much simpler.
3923         Fix calculation of probabilities that happen in tree expansion.
3924         * tree-switch-conversion.h (struct cluster): Add
3925         is_single_value_p.
3926         (struct simple_cluster): Likewise.
3927         (struct case_tree_node): Add new function has_child.
3928         (do_jump_if_equal): New.
3930 2018-08-27  Martin Liska  <mliska@suse.cz>
3932         * tree-switch-conversion.c (bit_test_cluster::find_bit_tests):
3933         Add new argument to bit_test_cluster constructor.
3934         (bit_test_cluster::emit): Set bits really number of values
3935         handlel by a test.
3936         (bit_test_cluster::hoist_edge_and_branch_if_true): Add
3937         probability argument.
3938         * tree-switch-conversion.h (struct bit_test_cluster):
3939         Add m_handles_entire_switch.
3941 2018-08-27  Martin Liska  <mliska@suse.cz>
3943         PR tree-optimization/86702
3944         * tree-switch-conversion.c (jump_table_cluster::emit):
3945         Make probabilities even for values in jump table
3946         according to number of cases handled.
3947         (switch_decision_tree::compute_cases_per_edge): Pass
3948         argument to reset_out_edges_aux function.
3949         (switch_decision_tree::analyze_switch_statement): Likewise.
3950         * tree-switch-conversion.h (switch_decision_tree::reset_out_edges_aux):
3951         Make it static.
3953 2018-08-27  Martin Liska  <mliska@suse.cz>
3955         * cfgexpand.c (expand_asm_stmt): Use label_to_block and pass
3956         cfun argument explicitly.
3957         * gimple-pretty-print.c (dump_gimple_switch): Likewise.
3958         * hsa-gen.c (gen_hsa_insns_for_switch_stmt): Use new
3959         function gimple_switch_default_bb.
3960         (convert_switch_statements):
3961         (expand_builtins):
3962         * ipa-fnsummary.c (set_switch_stmt_execution_predicate):
3963         * stmt.c (label_to_block_fn): Use label_to_block and pass
3964         cfun argument explicitly and use gimple_switch_label_bb.
3965         (expand_case): Likewise.
3966         * tree-cfg.c (lower_phi_internal_fn): Use label_to_block and pass
3967         cfun argument explicitly. Likewise.
3968         (make_edges_bb): Likewise.
3969         (make_cond_expr_edges): Likewise.
3970         (get_cases_for_edge): Likewise.
3971         (make_gimple_switch_edges): Likewise.
3972         (label_to_block_fn): Likewise.
3973         (label_to_block): Likewise.
3974         (make_goto_expr_edges): Likewise.
3975         (make_gimple_asm_edges): Likewise.
3976         (main_block_label): Likewise.
3977         (group_case_labels_stmt): Likewise.
3978         (find_taken_edge_computed_goto): Likewise.
3979         (find_taken_edge_switch_expr): Likewise.
3980         (gimple_verify_flow_info): Likewise.
3981         (gimple_redirect_edge_and_branch): Likewise.
3982         (gimple_switch_label_bb): New function.
3983         (gimple_switch_default_bb): Likewise.
3984         (gimple_switch_edge): Likewise.
3985         (gimple_switch_default_edge): Likewise.
3986         * tree-cfg.h (label_to_block_fn): Remove and replace ...
3987         (label_to_block): ... with this.
3988         (gimple_switch_label_bb): New.
3989         (gimple_switch_default_bb): Likewise.
3990         (gimple_switch_edge): Likewise.
3991         (gimple_switch_default_edge): Likewise.
3992         * tree-cfgcleanup.c (convert_single_case_switch): Use
3993         new gimple functions and pass new argument to label_to_block.
3994         (cleanup_control_flow_bb):
3995         * tree-eh.c (make_eh_dispatch_edges): Use label_to_block and pass
3996         cfun argument explicitly.
3997         (make_eh_edges): Likewise.
3998         (redirect_eh_dispatch_edge): Likewise.
3999         (lower_resx): Likewise.
4000         (lower_eh_dispatch): Likewise.
4001         (maybe_remove_unreachable_handlers): Likewise.
4002         (unsplit_eh): Likewise.
4003         (cleanup_empty_eh): Likewise.
4004         (verify_eh_edges): Likewise.
4005         (verify_eh_dispatch_edge): Likewise.
4006         * tree-ssa-dom.c (record_edge_info): Likewise.
4007         * tree-ssa-forwprop.c (simplify_gimple_switch_label_vec): Likewise.
4008         * tree-ssa-threadedge.c (thread_around_empty_blocks): Likewise.
4009         (thread_through_normal_block): Likewise.
4010         * tree-ssa-uncprop.c (associate_equivalences_with_edges): Likewise.
4011         * tree-ssa-uninit.c (convert_control_dep_chain_into_preds):
4012         * tree-switch-conversion.c (switch_conversion::collect): Use new
4013         gimple functions.
4014         (switch_conversion::check_final_bb): Likewise.
4015         (switch_conversion::gather_default_values): Pass new argument
4016         to label_to_block.
4017         (switch_conversion::build_constructors): Likewise.
4018         (switch_decision_tree::compute_cases_per_edge): Use new
4019         gimple_switch_edge function.
4020         (switch_decision_tree::analyze_switch_statement): Pass new argument
4021         to label_to_block.
4022         (switch_decision_tree::try_switch_expansion): Use
4023         gimple_switch_default_edge.
4024         * tree-vrp.c (find_switch_asserts): Pass new argument
4025         to label_to_block.
4026         * vr-values.c (vr_values::vrp_visit_switch_stmt): Likewise.
4027         (vr_values::simplify_switch_using_ranges): Likewise.
4029 2018-08-27  Richard Biener  <rguenther@suse.de>
4031         * cfganal.h (rev_post_order_and_mark_dfs_back_seme): Declare.
4032         * cfganal.c (rev_post_order_and_mark_dfs_back_seme): New function.
4034         * tree-ssa-sccvn.h (struct vn_pval): New structure.
4035         (struct vn_nary_op_s): Add unwind_to member.  Add
4036         predicated_values flag and put result into a union together
4037         with a linked list of vn_pval.
4038         (struct vn_ssa_aux): Add name member to make maintaining
4039         a map of SSA name to vn_ssa_aux possible.  Remove no longer
4040         needed info, dfsnum, low, visited, on_sccstack, use_processed
4041         and range_info_anti_range_p members.
4042         (run_scc_vn, vn_eliminate, free_scc_vn, vn_valueize): Remove.
4043         (do_rpo_vn, run_rpo_vn, eliminate_with_rpo_vn, free_rpo_vn):
4044         New functions.
4045         (vn_valueize): New global.
4046         (vn_context_bb): Likewise.
4047         (VN_INFO_RANGE_INFO, VN_INFO_ANTI_RANGE_P, VN_INFO_RANGE_TYPE,
4048         VN_INFO_PTR_INFO): Remove.
4049         * tree-ssa-sccvn.c: ... (rewrite)
4050         (pass_fre::execute): For -O2+ initialize loops and run
4051         RPO VN in optimistic mode (iterating).  For -O1 and -Og
4052         run RPO VN in non-optimistic mode.
4053         * params.def (PARAM_SCCVN_MAX_SCC_SIZE): Remove.
4054         (PARAM_RPO_VN_MAX_LOOP_DEPTH): Add.
4055         * doc/invoke.texi (sccvn-max-scc-size): Remove.
4056         (rpo-vn-max-loop-depth): Document.
4057         * tree-ssa-alias.c (walk_non_aliased_vuses): Stop walking
4058         when valuezing the VUSE signals we walked out of the region.
4059         * tree-ssa-pre.c (phi_translate_1): Ignore predicated values.
4060         (phi_translate): Set VN context block to use for availability
4061         lookup.
4062         (compute_avail): Likewise.
4063         (pre_valueize): New function.
4064         (pass_pre::execute): Adjust to the RPO VN API.
4066         * tree-ssa-loop-ivcanon.c: Include tree-ssa-sccvn.h.
4067         (propagate_constants_for_unrolling): Remove.
4068         (tree_unroll_loops_completely): Perform value-numbering
4069         on the unrolled bodies loop parent.
4071 2018-08-27  Richard Biener  <rguenther@suse.de>
4073         * tree-ssa-pre.c (compute_antic): Re-use inverted postorder
4074         for partial antic compute.
4076 2018-08-27  Jakub Jelinek  <jakub@redhat.com>
4078         PR rtl-optimization/87065
4079         * combine.c (simplify_if_then_else): Formatting fix.
4080         (if_then_else_cond): Guard MULT optimization with SCALAR_INT_MODE_P
4081         check.
4082         (known_cond): Don't return const_true_rtx for vector modes.  Use
4083         CONST0_RTX instead of const0_rtx.  Formatting fixes.
4085 2018-08-27  Martin Liska  <mliska@suse.cz>
4087         PR gcov-profile/87069
4088         * gcov.c (process_file): Record files already processed
4089         and warn about a file being processed multiple times.
4091 2018-08-27  Martin Liska  <mliska@suse.cz>
4093         PR driver/83193
4094         * config/aarch64/aarch64.c (aarch64_override_options_internal):
4095         Set default values for x_aarch64_*_string strings.
4096         * config/aarch64/aarch64.opt: Remove --{march,mcpu,mtune}==
4097         prefix.  For -mabi do not print '=ABI' in help and use
4098         <option_value> format for -msve-vector-bits and -moverride
4099         options.
4101 2018-08-26  Jeff Law  <law@redhat.com>
4103         * config/mips/frame-header-opt.c: Include "backend.h" rather than
4104         "cfg.h"
4106 2018-08-26  Marek Polacek  <polacek@redhat.com>
4108         PR c++/87029, Implement -Wredundant-move.
4109         * doc/invoke.texi: Document -Wredundant-move.
4111 2018-08-25  Martin Sebor  <msebor@redhat.com>
4113         PR tree-optimization/87059
4114         * builtins.c (expand_builtin_strncmp): Convert MIN_EXPR operand
4115         to the same type as the other.
4116         * fold-const.c (fold_binary_loc): Assert expectation.
4118 2018-08-25  Iain Sandoe  <iain@sandoe.co.uk>
4120         * config/darwin.c (machopic_legitimize_pic_address): Clean up
4121         extraneous parentheses, dead code section and formatting.
4123 2018-08-24  David Malcolm  <dmalcolm@redhat.com>
4125         PR c++/87091
4126         * diagnostic-show-locus.c (layout::layout): Ensure the margin is
4127         wide enough for jumps in the line-numbering to be visible.
4128         (layout::print_gap_in_line_numbering): New member function.
4129         (layout::calculate_line_spans): When using line numbering, merge
4130         line spans that are only 1 line apart.
4131         (diagnostic_show_locus): When printing line numbers, show gaps in
4132         line numbering directly, rather than printing headers.
4133         (selftest::test_diagnostic_show_locus_fixit_lines): Add test of
4134         line-numbering with multiple line spans.
4135         (selftest::test_fixit_insert_containing_newline_2): Add test of
4136         line-numbering, in which the spans are close enough to be merged.
4138 2018-08-24  Aldy Hernandez  <aldyh@redhat.com>
4140         * gimple-ssa-evrp-analyze.c (set_ssa_range_info): Pass value_range
4141         to range_includes_zero_p.  Do not special case VR_ANTI_RANGE.
4142         * tree-vrp.c (range_is_nonnull): Remove.
4143         (range_includes_zero_p): Accept value_range instead of min/max.
4144         (extract_range_from_binary_expr_1): Do not early bail on
4145         POINTER_PLUS_EXPR.
4146         Use range_includes_zero_p instead of range_is_nonnull.
4147         (extract_range_from_unary_expr): Use range_includes_zero_p instead
4148         of range_is_nonnull.
4149         (vrp_meet_1): Pass value_range to range_includes_zero_p.  Do not
4150         special case VR_ANTI_RANGE.
4151         (vrp_finalize): Same.
4152         * tree-vrp.h (range_includes_zero_p): Pass value_range as argument
4153         instead of min/max.
4154         (range_is_nonnull): Remove.
4155         * vr-values.c (vrp_stmt_computes_nonzero): Use
4156         range_includes_zero_p instead of range_is_nonnull.
4157         (extract_range_basic): Pass value_range to range_includes_zero_p
4158         instead of range_is_nonnull.
4160 2018-08-24  Uros Bizjak  <ubizjak@gmail.com>
4162         * emit-rtl.c (init_emit_once): Do not emit MODE_POINTER_BOUNDS RTXes.
4163         * emit-rtl.h (rtl_data): Remove return_bnd.
4164         * explow.c (trunc_int_for_mode): Do not handle POINTER_BOUNDS_MODE_P.
4165         * function.c (diddle_return_value): Do not handle crtl->return_bnd.
4166         * genmodes.c (complete_mode): Do not handle MODE_POINTER_BOUNDS.
4167         (POINTER_BOUNDS_MODE): Remove definition.
4168         (make_pointer_bounds_mode): Remove.
4169         (get_mode_class): Do not handle MODE_POINTER_BOUNDS.
4170         * machmode.h (POINTER_BOUNDS_MODE_P): Remove definition.
4171         (scalare_mode::includes_p): Do not handle MODE_POINTER_BOUNDS.
4172         * mode-classes.def: Do not define MODE_POINTER_BOUNDS.
4173         * stor-layout.c (int_mode_for_mode): Do not handle MODE_POINTER_BOUNDS.
4174         * tree-core.h (enum tree_index): Remove TI_POINTER_BOUNDS_TYPE.
4175         * varasm.c (output_constant_pool_2): Do not handle MODE_POINTER_BOUNDS.
4177         * config/i386/i386-modes.def (BND32, BND64): Remove.
4178         * config/i386/i386.c (dbx_register_map): Remove bound registers.
4179         (dbx64_register_map): Ditto.
4180         (svr4_dbx_register_map): Ditto.
4181         (indirect_thunk_bnd_needed): Remove.
4182         (indirect_thunks_bnd_used): Ditto.
4183         (indirect_return_bnd_needed): Ditto.
4184         (indirect_return_via_cx_bnd): Ditto.
4185         (enum indirect_thunk_prefix): Remove indirect_thunk_prefix_bnd.
4186         (indirect_thunk_name): Remove handling of indirect_thunk_prefix_bnd.
4187         (output_indirect_thunk): Ditto.  Remove need_prefix argument.
4188         (output_indirect_thunk_function): Remove handling of
4189         indirect_return_bnd_needed, indirect_return_via_cx_bnd,
4190         indirect_thunk_bnd_needed and indirect_thunks_bnd_used variables.
4191         (ix86_save_reg): Remove handling of crtl->return_bnd.
4192         (ix86_legitimate_constant_p): Remove handling of POINTER_BOUNDS_MODE_P.
4193         (ix86_print_operand_address_as): Remove handling of UNSPEC_BNDMK_ADDR
4194         and UNSPEC_BNDLX_ADDR.
4195         (ix86_output_indirect_branch_via_reg): Remove handling of
4196         indirect_thunk_prefix_bnd.
4197         (ix86_output_indirect_branch_via_push): Ditto.
4198         (ix86_output_function_return): Ditto.
4199         (ix86_output_indirect_function_return): Ditto.
4200         (avoid_func_arg_motion): Do not handle UNSPEC_BNDSTX.
4201         * config/i386/i386.h (FIXED_REGISTERS): Remove bound registers.
4202         (CALL_USED_REGISTERS): Ditto.
4203         (REG_ALLOC_ORDER): Update for removal of bound registers.
4204         (HI_REGISTER_NAMES): Ditto.
4205         * config/i386/i386.md (UNSPEC_BNDMK, UNSPEC_BNDMK_ADDR, UNSPEC_BNDSTX)
4206         (UNSPEC_BNDLDX, UNSPEC_BNDLDX_ADDR, UNSPEC_BNDCL, UNSPEC_BNDCU)
4207         (UNSPEC_BNDCN, UNSPEC_MPX_FENCE): Remove.
4208         (BND0_REG, BND1_REG, BND2_REG, BND3_REG): Remove
4209         (FIRST_PSEUDO_REG): Update.
4210         (BND): Remove mode iterator.
4211         * config/i386/predicates.md (bnd_mem_operator): Remove.
4213 2018-08-24  Richard Sandiford  <richard.sandiford@arm.com>
4215         * tree-vect-stmts.c (vectorizable_bswap): Handle variable-length
4216         vectors.
4218 2018-08-24  Richard Sandiford  <richard.sandiford@arm.com>
4220         * tree-vect-slp.c (vect_transform_slp_perm_load): Separate out
4221         the case in which the permute needs only a single element and
4222         repeats for every vector of the result.  Extend that case to
4223         handle variable-length vectors.
4224         * tree-vect-stmts.c (vectorizable_load): Update accordingly.
4226 2018-08-24  H.J. Lu  <hongjiu.lu@intel.com>
4228         PR debug/79342
4229         * dwarf2out.c (save_macinfo_strings): Call set_indirect_string
4230         on DW_MACINFO_start_file for -gsplit-dwarf -g3.
4232 2018-08-24  Richard Biener  <rguenther@suse.de>
4234         * cfg.h (struct control_flow_graph): Add edge_flags_allocated and
4235         bb_flags_allocated members.
4236         (auto_flag): New RAII class for allocating flags.
4237         (auto_edge_flag): New RAII class for allocating edge flags.
4238         (auto_bb_flag): New RAII class for allocating bb flags.
4239         * cfgloop.c (verify_loop_structure): Allocate temporary edge
4240         flag dynamically.
4241         * cfganal.c (dfs_enumerate_from): Remove use of visited sbitmap
4242         in favor of temporarily allocated BB flag.
4243         * hsa-brig.c: Re-order includes.
4244         * hsa-dump.c: Likewise.
4245         * hsa-regalloc.c: Likewise.
4246         * print-rtl.c: Likewise.
4247         * profile-count.c: Likewise.
4249 2018-08-24  Segher Boessenkool  <segher@kernel.crashing.org>
4251         PR target/86989
4252         * config/rs6000/rs6000.c (toc_relative_expr_p): Check that the base is
4253         the TOC register.
4255 2018-08-24  Aldy Hernandez  <aldyh@redhat.com>
4257         PR 87073/bootstrap
4258         * wide-int-range.cc (wide_int_range_div): Do not ignore result
4259         from wide_int_range_multiplicative_op.
4261 2018-08-23  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
4263         * tree-vect-data-refs.c (vect_grouped_store_supported): Fix typo
4264         "permutaion".
4266 2018-08-23  Giuliano Belinassi  <giuliano.belinassi@usp.br>
4268         * genmatch.c (parser::parse_operation): Fix typo 'exapnded'
4269         to 'expanded'.
4271 2018-08-23  Alexander Monakov  <amonakov@ispras.ru>
4273         * tree-scalar-evolution.c (final_value_replacement_loop): Dump
4274         full GENERIC expression used for replacement.
4276 2018-08-23  Aldy Hernandez  <aldyh@redhat.com>
4278         * tree-vrp.c (abs_extent_range): Remove.
4279         (extract_range_into_wide_ints): Pass wide ints by reference.
4280         (extract_range_from_binary_expr_1): Rewrite the *DIV_EXPR code.
4281         Pass wide ints by reference in all calls to
4282         extract_range_into_wide_ints.
4283         * wide-int-range.cc (wide_int_range_div): New.
4284         * wide-int-range.h (wide_int_range_div): New.
4285         (wide_int_range_includes_zero_p): New.
4286         (wide_int_range_zero_p): New.
4288 2018-08-23  Matthew Malcomson  <matthew.malcomson@arm.com>
4290         * config/aarch64/aarch64.md (arches): New enum.
4291         (arch): New enum attr.
4292         (arch_enabled): New attr.
4293         (enabled): Now uses arch_enabled only.
4294         (simd, sve, fp16): Removed attribute.
4295         (fp): Attr now defined in terms of 'arch'.
4296         (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64, *movti_aarch64,
4297         *movhf_aarch64, <optab><fcvt_target><GPF:mode>2,
4298         <FCVT_F2FIXED:fcvt_fixed_insn><GPF:mode>3,
4299         <FCVT_FIXED2F:fcvt_fixed_insn><GPI:mode>3): Merge 'fp' and 'simd'
4300         attributes into 'arch'.
4301         (*movsf_aarch64, *movdf_aarch64, *movtf_aarch64, *add<mode>3_aarch64,
4302         subdi3, neg<mode>2, <optab><mode>3, one_cmpl<mode>2,
4303         *<NLOGICAL:optab>_one_cmpl<mode>3, *xor_one_cmpl<mode>3,
4304         *aarch64_ashl_sisd_or_int_<mode>3, *aarch64_lshr_sisd_or_int_<mode>3,
4305         *aarch64_ashr_sisd_or_int_<mode>3, *aarch64_sisd_ushl): Convert use of
4306         'simd' attribute into 'arch'.
4307         (load_pair_sw_<SX:mode><SX2:mode>, load_pair_dw_<DX:mode><DX2:mode>,
4308         store_pair_sw_<SX:mode><SX2:mode>, store_pair_dw_<DX:mode><DX2:mode>):
4309         Convert use of 'fp' attribute to 'arch'.
4310         * config/aarch64/aarch64-simd.md (move_lo_quad_internal_<mode>,
4311         move_lo_quad_internal_<mode>): (different modes) Merge 'fp' and 'simd'
4312         into 'arch'.
4313         (move_lo_quad_internal_be_<mode>, move_lo_quad_internal_be_<mode>):
4314         (different modes) Merge 'fp' and 'simd' into 'arch'.
4315         (*aarch64_combinez<mode>, *aarch64_combinez_be<mode>): Merge 'fp' and
4316         'simd' into 'arch'.
4318 2018-08-23  Segher Boessenkool  <segher@kernel.crashing.org>
4320         PR rtl-optimization/87026
4321         * expmed.c (canonicalize_comparison): If we can no longer create
4322         pseudoregisters, don't.
4324 2018-08-23  Richard Earnshaw  <rearnsha@arm.com>
4326         PR target/86951
4327         * config/arm/arm-protos.h (arm_emit_speculation_barrier): New
4328         prototype.
4329         * config/arm/arm.c (speculation_barrier_libfunc): New static
4330         variable.
4331         (arm_init_libfuncs): Initialize it.
4332         (arm_emit_speculation_barrier): New function.
4333         * config/arm/arm.md (speculation_barrier): Call
4334         arm_emit_speculation_barrier for architectures that do not have
4335         DSB or ISB.
4336         (speculation_barrier_insn): Only match on Armv7 or later.
4338 2018-08-23  Richard Biener  <rguenther@suse.de>
4340         PR middle-end/87024
4341         * tree-inline.c (copy_bb): Drop unused __builtin_va_arg_pack_len
4342         calls.
4344 2018-08-23  Richard Sandiford  <richard.sandiford@arm.com>
4346         * config/aarch64/aarch64.c (aarch64_evpc_sve_tbl): Fix handling
4347         of single-vector TBLs.
4348         (aarch64_vectorize_vec_perm_const): Set one_vector_p when only
4349         one input is given.
4351 2018-08-23  Richard Sandiford  <richard.sandiford@arm.com>
4353         PR target/85910
4354         * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const_1): Fix
4355         aarch64_evpc_tbl guard.
4357 2018-08-22  Bernd Edlinger  <bernd.edlinger@hotmail.de>
4359         * tree-ssa-dse.c (compute_trims): Avoid folding away undefined
4360         behaviour.
4362 2018-08-22  Martin Sebor  <msebor@redhat.com>
4364         PR middle-end/87052
4365         * tree-pretty-print.c (pretty_print_string): Add argument.
4366         (dump_generic_node): Call to pretty_print_string with string size.
4368 2018-08-22  Segher Boessenkool  <segher@kernel.crashing.org>
4370         PR rtl-optimization/86771
4371         * combine.c (try_combine): Do not allow splitting a resulting PARALLEL
4372         of two SETs into those two SETs, one to be placed at i2, if that SETs
4373         destination is modified between i2 and i3.
4375 2018-08-22  Richard Sandiford  <richard.sandiford@arm.com>
4377         PR tree-optimization/86725
4378         * tree-vect-loop.c (vect_inner_phi_in_double_reduction_p): New
4379         function.
4380         (vect_analyze_scalar_cycles_1): Check it.
4382 2018-08-22  Richard Sandiford  <richard.sandiford@arm.com>
4384         PR tree-optimization/86725
4385         * tree-vect-loop.c (vect_is_simple_reduction): When treating
4386         an outer loop phi as a double reduction, make sure that the
4387         single user of the phi result is an inner loop phi.
4389 2018-08-22  Richard Sandiford  <richard.sandiford@arm.com>
4391         * tree-vect-data-refs.c (vect_analyze_group_access_1): Convert
4392         grouped stores with gaps to a strided group.
4394 2018-08-22  Richard Sandiford  <richard.sandiford@arm.com>
4396         * tree-vect-stmts.c (get_group_load_store_type)
4397         (get_load_store_type): Only test STMT_VINFO_STRIDED_P for the
4398         first statement in a group.
4400 2018-08-22  Iain Sandoe  <iain@sandoe.co.uk>
4402         * config/darwin.h (LINK_COMMAND_SPEC_A): Sync LTO options with
4403         the sequence used in gcc/gcc.c.
4405 2018-08-22  Iain Sandoe  <iain@sandoe.co.uk>
4407         PR other/704
4408         * gcc-ar.c (main): Don’t try to invoke the plug-in if we’re not
4409         building it.
4411 2018-08-22  Iain Sandoe  <iain@sandoe.co.uk>
4413         * config/darwin10.h (LINK_GCC_C_SEQUENCE_SPEC): Adjust to use the
4414         Darwin10-specific unwinder-shim.
4415         * config/darwin12.h (LINK_GCC_C_SEQUENCE_SPEC): Remove.
4416         * config/rs6000/darwin.h (DARWIN_CRT1_SPEC, DARWIN_DYLIB1_SPEC):
4417         New to cater for Darwin10 Rosetta.
4419 2018-08-22  Iain Sandoe  <iain@sandoe.co.uk>
4421         * config/i386/i386.c (ix86_output_addr_diff_elt): Move the MACH-O
4422         specific test before the one for HAVE_AS_GOTOFF_IN_DATA.
4424 2018-08-22  Iain Sandoe  <iain@sandoe.co.uk>
4426         PR bootstrap/81033
4427         PR target/81733
4428         PR target/52795
4429         * gcc/dwarf2out.c (FUNC_SECOND_SECT_LABEL): New.
4430         (dwarf2out_switch_text_section): Generate a local label for the
4431         second function sub-section and apply it as the second FDE start
4432         label.
4433         * gcc/final.c (final_scan_insn_1): Emit second FDE label after the
4434         second sub-section start.
4436 2018-08-22  Richard Biener  <rguenther@suse.de>
4438         PR tree-optimization/86988
4439         * tree-vrp.c (vrp_prop::check_mem_ref): Bail out on VLAs.
4441 2018-08-22  Richard Biener  <rguenther@suse.de>
4443         PR tree-optimization/86945
4444         * tree-cfg.c (generate_range_test): Use unsigned arithmetic.
4446 2018-08-22  Alexandre Oliva <oliva@adacore.com>
4448         * config/rs6000/rs6000.c (SMALL_DATA_RELOC, SMALL_DATA_REG): Add
4449         a comment about how uses of r2 for .sdata2 come about.
4451 2018-08-22  Alexandre Oliva <aoliva@redhat.com>
4453         * tree-ssa-reassoc.c (is_reassociable_op): Fix cut&pasto.
4455 2018-08-21  Marek Polacek  <polacek@redhat.com>
4457         PR c++/86981, Implement -Wpessimizing-move.
4458         * doc/invoke.texi: Document -Wpessimizing-move.
4460 2018-08-21  Jan Hubicka  <jh@suse.cz>
4462         * tree.c (find_decls_types_r): Do not check for redundant typedefs.
4463         * tree.h (is_redundant_typedef): Remove.
4464         * dwarf2out.c (is_redundant_typedef): Turn into static function.
4466 2018-08-21  Jan Hubicka  <jh@suse.cz>
4468         * tree.c (free_lang_data_in_decl): Remove types from DECL_CONTEXT
4469         when possible.
4471 2018-08-21  Tamar Christina  <tamar.christina@arm.com>
4473         * expmed.c (extract_low_bits): Reject invalid subregs early.
4475 2018-08-21  Bernd Edlinger  <bernd.edlinger@hotmail.de>
4477         PR middle-end/86121
4478         * tree-ssa-strlen.c (adjust_last_stmt): Avoid folding away undefined
4479         behaviour.
4481 2018-08-21  Rasmus Villemoes  <rv@rasmusvillemoes.dk>
4483         * config/vxworks.h: Guard vxworks_asm_out_constructor and
4484         vxworks_asm_out_destructor by !HAVE_INITFINI_ARRAY_SUPPORT
4485         * config/vxworks.c: Likewise.
4487 2018-08-21  Rasmus Villemoes  <rv@rasmusvillemoes.dk>
4489         * config/vxworks.c: Set targetm.have_ctors_dtors
4490         if HAVE_INITFINI_ARRAY_SUPPORT.
4491         * config/vxworks.h: Set SUPPORTS_INIT_PRIORITY
4492         if HAVE_INITFINI_ARRAY_SUPPORT.
4494 2018-08-21  Rasmus Villemoes  <rv@rasmusvillemoes.dk>
4496         * config/vxworks.h: Add $(WIND_BASE)/target/h/wrn/coreip to
4497         default search path for VxWorks < 7.
4499 2018-08-21  Bernd Edlinger  <bernd.edlinger@hotmail.de>
4501         * gimple-ssa-sprintf.c (decl_constant_value): Remove.
4502         (get_format_string): Refer to c_getstr.
4504 2018-08-21  Tom de Vries  <tdevries@suse.de>
4506         * cgraph.h (debuginfo_early_init, debuginfo_init, debuginfo_fini)
4507         (debuginfo_start, debuginfo_stop, debuginfo_early_start)
4508         (debuginfo_early_stop): Declare.
4509         * cgraphunit.c (debuginfo_early_init, debuginfo_init, debuginfo_fini)
4510         (debuginfo_start, debuginfo_stop, debuginfo_early_start)
4511         (debuginfo_early_stop): New function.
4512         (symbol_table::finalize_compilation_unit): Call debuginfo_early_start
4513         and debuginfo_early_stop.
4514         * dwarf2out.c (dwarf2out_finish, dwarf2out_early_finish): Dump dwarf.
4515         * toplev.c (compile_file): Call debuginfo_start and debuginfo_stop.
4516         (general_init): Call debuginfo_early_init.
4517         (finalize): Call debuginfo_fini.
4518         (do_compile): Call debuginfo_init.
4519         * doc/invoke.texi (@gccoptlist): Add -fdump-debug and
4520         -fdump-early-debug.
4521         (@item -fdump-debug, @item -fdump-earlydebug): Add.
4523 2018-08-21  Tom de Vries  <tdevries@suse.de>
4525         * dwarf2out.c (print_dw_val, print_loc_descr, print_die): Handle
4526         flag_dump_noaddr and flag_dump_unnumbered.
4528 2018-08-21  Aldy Hernandez  <aldyh@redhat.com>
4530         * wide-int-range.cc (wide_int_range_abs): New.
4531         (wide_int_range_order_set): Rename from wide_int_range_min_max.
4532         * wide-int-range.h (wide_int_range_abs): New.
4533         (wide_int_range_min_max): New.
4534         * tree-vrp.c (extract_range_from_unary_expr): Rewrite ABS_EXPR
4535         case to call wide_int_range_abs.
4536         Rewrite MIN/MAX_EXPR to call wide_int_range_min_max.
4537         (extract_range_from_abs_expr): Delete.
4539 2018-08-20  Michael Meissner  <meissner@linux.ibm.com>
4541         PR target/87033
4542         * config/rs6000/rs6000.md (extendsi<mode>2): Change constraints
4543         from 'Y' to 'YZ' to enable the LWAX instruction to be generated
4544         for indexed loads.
4546 2018-08-20  Nathan Sidwell  <nathan@acm.org>
4547             Jeff Law <law@redhat.com>
4549         * config/s390/s390-c (s390_macro_to_expand): Use cpp_macro_p.
4550         * config/spu/spu-c.c (spu_macro_to_expand): Likewise.
4552 2018-08-20  David Malcolm  <dmalcolm@redhat.com>
4554         PR other/84889
4555         * attribs.c (diag_attr_exclusions): Add auto_diagnostic_group instance.
4556         (decl_attributes): Likewise.
4557         * calls.c (maybe_warn_nonstring_arg): Add auto_diagnostic_group
4558         instance.
4559         * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
4560         * diagnostic-core.h (class auto_diagnostic_group): New class.
4561         * diagnostic.c (diagnostic_initialize): Initialize the new fields.
4562         (diagnostic_report_diagnostic): Handle the first diagnostics within
4563         a group.
4564         (emit_diagnostic): Add auto_diagnostic_group instance.
4565         (inform): Likewise.
4566         (inform_n): Likewise.
4567         (warning): Likewise.
4568         (warning_at): Likewise.
4569         (warning_n): Likewise.
4570         (pedwarn): Likewise.
4571         (permerror): Likewise.
4572         (error): Likewise.
4573         (error_n): Likewise.
4574         (error_at): Likewise.
4575         (sorry): Likewise.
4576         (fatal_error): Likewise.
4577         (internal_error): Likewise.
4578         (internal_error_no_backtrace): Likewise.
4579         (auto_diagnostic_group::auto_diagnostic_group): New ctor.
4580         (auto_diagnostic_group::~auto_diagnostic_group): New dtor.
4581         * diagnostic.h (struct diagnostic_context): Add fields
4582         "diagnostic_group_nesting_depth",
4583         "diagnostic_group_emission_count", "begin_group_cb",
4584         "end_group_cb".
4585         * gimple-ssa-isolate-paths.c (find_implicit_erroneous_behavior):
4586         Add auto_diagnostic_group instance(s).
4587         (find_explicit_erroneous_behavior): Likewise.
4588         * gimple-ssa-warn-alloca.c (pass_walloca::execute): Likewise.
4589         * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Likewise.
4590         * gimplify.c (warn_implicit_fallthrough_r): Likewise.
4591         (gimplify_va_arg_expr): Likewise.
4592         * hsa-gen.c (HSA_SORRY_ATV): Likewise.
4593         (HSA_SORRY_AT): Likewise.
4594         * ipa-devirt.c (compare_virtual_tables): Likewise.
4595         (warn_odr): Likewise.
4596         * multiple_target.c (expand_target_clones): Likewise.
4597         * opts-common.c (cmdline_handle_error): Likewise.
4598         * reginfo.c (globalize_reg): Likewise.
4599         * substring-locations.c (format_warning_n_va): Likewise.
4600         * tree-inline.c (expand_call_inline): Likewise.
4601         * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Likewise.
4602         * tree-ssa-loop-niter.c
4603         (do_warn_aggressive_loop_optimizations): Likewise.
4604         * tree-ssa-uninit.c (warn_uninit): Likewise.
4605         * tree.c (warn_deprecated_use): Likewise.
4607 2018-08-20  H.J. Lu  <hongjiu.lu@intel.com>
4609         PR target/87014
4610         * config/i386/i386.md (eh_return): Always update EH return
4611         address in word_mode.
4613 2018-08-20  Chung-Lin Tang  <cltang@codesourcery.com>
4615         * targhooks.c (std_gimplify_va_arg_expr): Properly handle case of when
4616         TARGET_SPLIT_COMPLEX_ARG is defined.
4618 2018-08-20  Bernd Edlinger  <bernd.edlinger@hotmail.de>
4620         * expr.c (store_field): Change gcc_assert to gcc_checking_assert.
4622 2018-08-20  Bernd Edlinger  <bernd.edlinger@hotmail.de>
4624         PR target/86984
4625         * expr.c (expand_assignment): Assert that bitpos is positive.
4626         (store_field): Likewise
4627         (expand_expr_real_1): Make sure that bitpos is positive.
4628         * config/alpha/alpha.h (CONSTANT_ADDRESS_P): Avoid signed
4629         integer overflow.
4631 2018-08-20  Nathan Sidwell  <nathan@acm.org>
4633         * Makefile.in (CPP_ID_DATA_H): Delete.
4634         (CPP_INTERNAL_H): Don't add it.
4635         (GTFILES): Replace CPP_ID_DATA_H with CPPLIB_H.
4636         * gengtype.c (open_base_files): Replace cpp-id-data.h with cpplib.h
4638 2018-08-20  Richard Biener  <rguenther@suse.de>
4640         PR tree-optimization/78655
4641         * tree-vrp.c (extract_range_from_binary_expr_1): Make
4642         pointer + offset nonnull if either operand is nonnull work.
4644 2018-08-20  Tom de Vries  <tdevries@suse.de>
4646         * dwarf2out.c (add_scalar_info): Don't add reference to existing die
4647         unless the referenced die describes the added property using
4648         DW_AT_location or DW_AT_const_value.  Fall back to exprloc case.
4649         Otherwise, add a DW_AT_location to the referenced die.
4651 2018-08-19  Uros Bizjak  <ubizjak@gmail.com>
4653         PR target/86994
4654         * config/i386/i386.c (ix86_rtx_costs) [case SET]: Check source for
4655         register_operand when calling ix86_set_reg_reg_cost.
4656         [case CONST_INT, case CONST, case LABEL_REF, case SYMBOL_REF]:
4657         Set *total to 0 for operands that satisfy x86_64_immediate_operand
4658         predicate and to 1 otherwise.
4660 2018-08-18  Iain Sandoe  <iain@sandoe.co.uk>
4662         * config/darwin.c (darwin_override_options): If -gsplit-dwarf is set,
4663         emit a diagnostic that it is not supported and reset the option.
4664         * config/darwin.h (DRIVER_SELF_SPECS): Note that gsplit-dwarf is not
4665         supported and consume the option.  (ASM_FINAL_SPEC): New.
4667 2018-08-17  Segher Boessenkool  <segher@kernel.crashing.org>
4669         * doc/md.texi (Patterns): Use @ref instead of @xref in the middle of
4670         a sentence.
4672 2018-08-17  Sandra Loosemore  <sandra@codesourcery.com>
4674         C-SKY port: Documentation
4676         * doc/extend.texi (C-SKY Function Attributes): New section.
4677         * doc/invoke.texi (Option Summary): Add C-SKY options.
4678         (C-SKY Options): New section.
4679         * doc/md.texi (Machine Constraints): Document C-SKY constraints.
4681 2018-08-17  Jojo  <jijie_rong@c-sky.com>
4682             Huibin Wang  <huibin_wang@c-sky.com>
4683             Sandra Loosemore  <sandra@codesourcery.com>
4684             Chung-Lin Tang  <cltang@codesourcery.com>
4686         C-SKY port: Backend implementation
4688         * config/csky/*: New.
4689         * common/config/csky/*: New.
4691 2018-08-17  Jojo  <jijie_rong@c-sky.com>
4692             Huibin Wang  <huibin_wang@c-sky.com>
4693             Sandra Loosemore  <sandra@codesourcery.com>
4694             Chung-Lin Tang  <cltang@codesourcery.com>
4695             Andrew Jenner  <andrew@codesourcery.com>
4697         C-SKY port: Configury
4699         * config.gcc (csky-*-*): New.
4700         * configure.ac: Add csky to targets for dwarf2 debug_line support.
4701         * configure: Regenerated.
4703 2018-08-17  David Malcolm  <dmalcolm@redhat.com>
4705         * dump-context.h: Include "dumpfile.h".
4706         (dump_context::dump_printf_va): Convert final param from va_list
4707         to va_list *.  Convert from ATTRIBUTE_PRINTF to
4708         ATTRIBUTE_GCC_DUMP_PRINTF.
4709         (dump_context::dump_printf_loc_va): Likewise.
4710         * dumpfile.c: Include "stringpool.h".
4711         (make_item_for_dump_printf_va): Delete.
4712         (make_item_for_dump_printf): Delete.
4713         (class dump_pretty_printer): New class.
4714         (dump_pretty_printer::dump_pretty_printer): New ctor.
4715         (dump_pretty_printer::emit_items): New member function.
4716         (dump_pretty_printer::emit_any_pending_textual_chunks): New member
4717         function.
4718         (dump_pretty_printer::emit_item): New member function.
4719         (dump_pretty_printer::stash_item): New member function.
4720         (dump_pretty_printer::format_decoder_cb): New member function.
4721         (dump_pretty_printer::decode_format): New member function.
4722         (dump_context::dump_printf_va): Reimplement in terms of
4723         dump_pretty_printer.
4724         (dump_context::dump_printf_loc_va): Convert final param from va_list
4725         to va_list *.
4726         (dump_context::begin_scope): Reimplement call to
4727         make_item_for_dump_printf.
4728         (dump_printf): Update for change to dump_printf_va.
4729         (dump_printf_loc): Likewise.
4730         (selftest::test_capture_of_dump_calls): Convert "stmt" from
4731         greturn * to gimple *.  Add a test_decl.  Add tests of dump_printf
4732         with %T, %E, and %G.
4733         * dumpfile.h (ATTRIBUTE_GCC_DUMP_PRINTF): New macro.
4734         (dump_printf): Replace ATTRIBUTE_PRINTF_2 with
4735         ATTRIBUTE_GCC_DUMP_PRINTF (2, 3).
4736         (dump_printf_loc): Replace ATTRIBUTE_PRINTF_3 with
4737         ATTRIBUTE_GCC_DUMP_PRINTF (3, 0).
4738         * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Convert
4739         use of HOST_WIDE_INT_PRINT_DEC on unsigned HOST_WIDE_INT "count"
4740         within a dump_printf_loc call to "%wu".
4741         (vector_alignment_reachable_p): Merge two dump_printf[_loc] calls,
4742         converting a use of HOST_WIDE_INT_PRINT_DEC to "%wd".  Add a
4743         missing space after "=".
4744         * tree-vect-loop.c (vect_analyze_loop_2) Within a dump_printf
4745         call, convert use of HOST_WIDE_INT_PRINT_DEC to "%wd".
4746         * tree-vect-slp.c (vect_slp_bb): Within a dump_printf_loc call,
4747         convert use of HOST_WIDE_INT_PRINT_UNSIGNED to "%wu".
4748         * tree-vectorizer.c (try_vectorize_loop_1): Likewise.  Remove
4749         duplicate "vectorized" from message.
4751 2018-08-17  Szabolcs Nagy  <szabolcs.nagy@arm.com>
4753         * config/arm/arm-builtins.c (arm_init_simd_builtin_types): Clear
4754         polyNxK_t element's TYPE_STRING_FLAG.
4756 2018-08-17  Segher Boessenkool  <segher@kernel.crashing.org>
4758         * config/rs6000/rs6000.md (*cbranch, *creturn): Name these patterns
4759         (they were unnamed before).  Fix comments.
4761 2018-08-17  Nathan Sidwell  <nathan@acm.org>
4763         * cppbuiltin.c: Include "cpplib.h", not "cpp-id-data.h".
4765 2018-08-17  Richard Biener  <rguenther@suse.de>
4767         PR tree-optimization/86841
4768         * wide-int-range.cc (wide_int_range_lshift): Use to_uhwi.
4770 2018-08-17  Martin Liska  <mliska@suse.cz>
4772         * common.opt: Remove Warn, Init and Report for options with
4773         Ignore/Deprecated flag. Warning is done automatically for
4774         Deprecated flags.
4775         * config/i386/i386.opt: Likewise.
4776         * config/ia64/ia64.opt: Likewise.
4777         * config/rs6000/rs6000.opt: Likewise.
4778         * cppbuiltin.c (define_builtin_macros_for_compilation_flags):
4779         Remove usage of flag_check_pointer_bounds.
4780         * lto-wrapper.c (merge_and_complain): Do not handle
4781         OPT_fcheck_pointer_bounds.
4782         (append_compiler_options): Likewise.
4783         * opt-functions.awk: Do not handle Deprecated.
4784         * optc-gen.awk: Check that Var, Report and Init are not
4785         used for an option with Ignore/Deprecated flag.
4786         * opts-common.c (decode_cmdline_option): Do not report
4787         CL_ERR_DEPRECATED.
4788         (read_cmdline_option): Report warning for OPT_SPECIAL_deprecated
4789         options.
4790         * opts.h (struct cl_option): Remove cl_deprecated flag.
4791         (CL_ERR_DEPRECATED): Remove error enum value.
4793 2018-08-17  Richard Biener  <rguenther@suse.de>
4795         PR middle-end/86505
4796         * tree-inline.c (copy_bb): When inlining __builtin_va_arg_pack_len ()
4797         across a va-arg-pack using call adjust its return value accordingly.
4799 2018-08-16  Martin Sebor  <msebor@redhat.com>
4801         PR tree-optimization/86853
4802         * gimple-ssa-sprintf.c (struct format_result): Rename member.
4803         (struct fmtresult): Add member and initialize it in ctors.
4804         (format_character): Handle %C.  Extend range to NUL.  Set MAYFAIL.
4805         (format_string): Handle %S the same as %ls.  Set MAYFAIL.
4806         (format_directive): Set POSUNDER4K when MAYFAIL is set.
4807         (parse_directive): Handle %C same as %c.
4808         (sprintf_dom_walker::compute_format_length): Adjust.
4809         (is_call_safe): Adjust.
4811 2018-08-16  Bernd Edlinger  <bernd.edlinger@hotmail.de>
4813         * builtins.c (c_strlen): Add new parameter eltsize.  Use it
4814         for determining how to count the elements.
4815         * builtins.h (c_strlen): Adjust prototype.
4816         * expr.c (string_constant): Add new parameter mem_size.
4817         Set *mem_size appropriately.
4818         * expr.h (string_constant): Adjust protoype.
4819         * gimple-fold.c (get_range_strlen): Add new parameter eltsize.
4820         * gimple-fold.h (get_range_strlen): Adjust prototype.
4821         * gimple-ssa-sprintf.c (get_string_length): Add new parameter eltsize.
4822         (format_string): Call get_string_length with eltsize.
4824 2018-08-16  David Malcolm  <dmalcolm@redhat.com>
4826         * diagnostic.c (default_diagnostic_start_span_fn): Call pp_string
4827         to emit the span, rather than setting it as the prefix.
4829 2018-08-16  David Malcolm  <dmalcolm@redhat.com>
4831         * diagnostic-show-locus.c (layout::start_annotation_line): Add
4832         "margin_char" parameter, defaulting to space.  Use it in place
4833         of pp_space for the initial part of the margin.
4834         (layout::print_leading_fixits): Use '+' when filling the margin
4835         of line-insertion fix-it hints.
4837 2018-08-16  Segher Boessenkool  <segher@kernel.crashing.org>
4839         * config/rs6000/rs6000.md (two unnamed define_insn and define_split):
4840         Delete.
4842 2018-08-16  Segher Boessenkool  <segher@kernel.crashing.org>
4844         * config/rs6000/altivec.md: Don't set length attribute to the default
4845         value.
4846         * config/rs6000/darwin.md: Ditto.
4847         * config/rs6000/dfp.md: Ditto.
4848         * config/rs6000/htm.md: Ditto.
4849         * config/rs6000/rs6000.md: Ditto.
4850         * config/rs6000/sync.md: Ditto.
4851         * config/rs6000/vsx.md: Ditto.
4853 2018-08-16  Segher Boessenkool  <segher@kernel.crashing.org>
4855         * config/rs6000/altivec.md: Don't set length attribute to the default
4856         value, for branch instructions.
4857         * config/rs6000/darwin.md: Ditto.
4858         * config/rs6000/rs6000.md: Ditto.
4860 2018-08-16  Segher Boessenkool  <segher@kernel.crashing.org>
4862         * config/rs6000/rs6000.md (length): Always define as const_int 4.
4863         (unnamed conditional branch define_insn): Set length to 4 or 8
4864         depending on offset.
4865         (<bd>_<mode>): Similar, for alternative 0.
4866         (<bd>tf_<mode>): Ditto.
4868 2018-08-16  Tamar Christina  <tamar.christina@arm.com>
4870         * expr.c (copy_blkmode_to_reg): Perform larger copies when safe.
4872 2018-08-16  Matthew Malcomson  <matthew.malcomson@arm.com>
4874         * doc/rtl.texi: Replace old RTX class names with new names.
4877 2018-08-16  Vlad Lazar  <vlad.lazar@arm.com>
4879         * expmed.h (canonicalize_comparison): New declaration.
4880         * expmed.c (canonicalize_comparison, equivalent_cmp_code): New function.
4881         * expmed.c (emit_store_flag_1): Add call to canonicalize_comparison.
4882         * optabs.c (prepare_cmp_insn): Likewise.
4883         * rtl.h (unsigned_condition_p): New function which checks if a
4884         comparison operator is unsigned.
4886 2018-08-16  Nathan Sidwell  <nathan@acm.org>
4888         * config/rs6000/rs6000-c.c (rs6000_macro_to_expend): Use cpp_macro_p.
4889         * config/powerpcspc/powerpcspe-c.c (rs6000_macro_to_expend): Likewise.
4891 2018-08-16  Tamar Christina  <tamar.christina@arm.com>
4893         PR target/84711
4894         * config/arm/arm.c (arm_can_change_mode_class): Disallow subreg.
4895         * config/arm/neon.md (movv4hf, movv8hf): Refactored to..
4896         (mov<mov>): ..this and enable unconditionally.
4898 2018-08-16  Tamar Christina  <tamar.christina@arm.com>
4900         * config/arm/neon.md (*neon_mov<mode>): Remove reg-to-reg alternative.
4902 2018-08-16  Sam Tebbs  <sam.tebbs@arm.com>
4904         * config/aarch64/aarch64.opt (mlow-precision-recip-sqrt)
4905         (mlow-precision-sqrt, mlow-precision-div, mverbose-cost-dump): Replace
4906         "Common" with "Target".
4908 2018-08-15  Uros Bizjak  <ubizjak@gmail.com>
4910         * config/i386/i386.opt (mmitigate-rop): Mark as deprecated.
4911         * doc/invoke.texi (mmitigate-rop): Remove.
4912         * config/i386/i386.c: Do not include "regrename.h".
4913         (ix86_rop_should_change_byte_p, reg_encoded_number)
4914         (ix86_get_modrm_for_rop, set_rop_modrm_reg_bits, ix86_mitigate_rop):
4915         Remove.
4916         (ix86_reorg): Remove call to ix86_mitigate_rop.
4917         * config/i386/i386.md (attr "modrm_class"): Remove.
4918         (cmp<mode>_ccno_1, mov<mode>_xor, movstrict<mode>_xor)
4919         (x86_mov<mode>cc_0_m1. x86_mov<mode>cc_0_m1_se)
4920         (x86_mov<mode>cc_0_m1_neg): Remove modrm_class attribute override.
4922 2018-08-15  Will Schmidt  <will_schmidt@vnet.ibm.com>
4924         * config/rs6000/rs600.c (rs6000_gimple_fold_builtin): Add entries to
4925         allow folding of mergeh() and mergel() for the float and double types.
4926         (fold_mergehl_helper): Rework to handle building a permute tree
4927         for float vectors.
4929 2018-08-15  Uros Bizjak  <ubizjak@gmail.com>
4931         * config/i386/i386.c (expand_vec_perm_movs): Enable V4SFmode
4932         for TARGET_SSE.
4934 2018-08-15  David Malcolm  <dmalcolm@redhat.com>
4936         * common.opt (fdiagnostics-show-labels): New option.
4937         * diagnostic-show-locus.c (class layout_range): Add field
4938         "m_label".
4939         (class layout): Add field "m_show_labels_p".
4940         (layout_range::layout_range): Add param "label" and use it to
4941         initialize m_label.
4942         (make_range): Pass in NULL for new "label" param of layout_range's
4943         ctor.
4944         (layout::layout): Initialize m_show_labels_p.
4945         (layout::maybe_add_location_range): Pass in loc_range->m_label
4946         when constructing layout_range instances.
4947         (struct line_label): New struct.
4948         (layout::print_any_labels): New member function.
4949         (layout::print_line): Call it if label-printing is enabled.
4950         (selftest::test_one_liner_labels): New test.
4951         (selftest::test_diagnostic_show_locus_one_liner): Call it.
4952         * diagnostic.c (diagnostic_initialize): Initialize
4953         context->show_labels_p.
4954         * diagnostic.h (struct diagnostic_context): Add field
4955         "show_labels_p".
4956         * doc/invoke.texi (Diagnostic Message Formatting Options): Add
4957         -fno-diagnostics-show-labels.
4958         * dwarf2out.c (gen_producer_string): Add
4959         OPT_fdiagnostics_show_labels to the ignored options.
4960         * gcc-rich-location.c (gcc_rich_location::add_expr): Add "label"
4961         param.
4962         (gcc_rich_location::maybe_add_expr): Likewise.
4963         * gcc-rich-location.h (gcc_rich_location::gcc_rich_location): Add
4964         label" param, defaulting to NULL.
4965         (gcc_rich_location::add_expr): Add "label" param.
4966         (gcc_rich_location::maybe_add_expr): Likewise.
4967         (class text_range_label): New class.
4968         (class range_label_for_type_mismatch): New class.
4969         * gimple-ssa-sprintf.c (fmtwarn): Pass NULL for new label params
4970         of format_warning_va.
4971         (fmtwarn_n): Likewise for new params of format_warning_n_va.
4972         * lto-wrapper.c (merge_and_complain): Add
4973         OPT_fdiagnostics_show_labels to the "pick one setting" options.
4974         (append_compiler_options): Likewise to the dropped options.
4975         (append_diag_options): Likewise to the passed-on options.
4976         * opts.c (common_handle_option): Handle the new option.
4977         * selftest-diagnostic.c
4978         (test_diagnostic_context::test_diagnostic_context): Enable
4979         show_labels_p.
4980         * substring-locations.c: Include "gcc-rich-location.h".
4981         (format_warning_n_va): Add "fmt_label" and "param_label" params
4982         and use them as appropriate.
4983         (format_warning_va): Add "fmt_label" and "param_label" params,
4984         passing them on to format_warning_n_va.
4985         (format_warning_at_substring): Likewise.
4986         (format_warning_at_substring_n): Likewise.
4987         * substring-locations.h (format_warning_va): Add "fmt_label" and
4988         "param_label" params.
4989         (format_warning_n_va): Likewise.
4990         (format_warning_at_substring): Likewise.
4991         (format_warning_at_substring_n): Likewise.
4992         * toplev.c (general_init): Initialize global_dc->show_labels_p.
4994 2018-08-15  Qing Zhao  <qing.zhao@oracle.com>
4996         PR testsuite/86519
4997         * builtins.c (expand_builtin_memcmp): Do not expand the call
4998         when overflow is detected.
5000 2018-08-15  Martin Sebor  <msebor@redhat.com>
5002         PR tree-optimization/71625
5003         * config/aarch64/aarch64-builtins.c
5004         (aarch64_init_simd_builtin_types): Clear Poly8_t's TYPE_STRING_FLAG.
5006 2018-08-15  Ilya Leoshkevich  <iii@linux.ibm.com>
5008         * config/s390/s390.c (s390_reorg): Remove loop.
5010 2018-08-15  Iain Sandoe  <iain@sandoe.co.uk>
5012         * config/darwin.c
5013          (darwin_function_switched_text_sections): Delete.
5014         * gcc/config/darwin.h
5015          (TARGET_ASM_FUNCTION_SWITCHED_TEXT_SECTIONS): Likewise.
5017 2018-08-15  Iain Sandoe  <iain@sandoe.co.uk>
5019         PR target/81685
5020         * config/darwin.h: (DEBUG_STR_OFFSETS_SECTION, DEBUG_LOCLISTS_SECTION,
5021         DEBUG_RNGLISTS_SECTION) new macros.  (DEBUG_PUBNAMES_SECTION,
5022         DEBUG_PUBTYPES_SECTION) update to include GNU variant.
5024 2018-08-15  Martin Liska  <mliska@suse.cz>
5026         PR tree-optimization/86925
5027         * predict.c (expr_expected_value_1): When taking
5028         later predictor, assign also probability.
5029         Use fold_build2_initializer_loc in order to fold
5030         the expression in -frounding-math.
5032 2018-08-14  Allan Sandfeld Jensen <allan.jensen@qt.io>
5034         * config/i386/i386.c (expand_vec_perm_movs): New method matching movs
5035         patterns.
5036         (expand_vec_perm_1): Try the new method.
5038 2018-08-14  Ilya Leoshkevich  <iii@linux.ibm.com>
5040         PR target/86547
5041         * lra-lives.c (remove_some_program_points_and_update_live_ranges):
5042         Check whether lra_live_max_point is 0 before dividing.
5044 2018-08-14  Martin Sebor  <msebor@redhat.com>
5046         PR tree-optimization/86650
5047         * tree-vrp.c (vrp_prop::check_array_ref): Print an inform message.
5048         (vrp_prop::check_mem_ref): Same.
5050 2018-08-13  Liu Hao <lh_mouse@126.com>
5052         * pretty-print.c (eat_esc_sequence): Swap the foreground and
5053         background colors if the COMMON_LVB_REVERSE_VIDEO flag is set,
5054         and clear it thereafter, as it only works for DBCS.
5056 2018-08-13  Liu Hao <lh_mouse@126.com>
5058         * pretty-print.c (mingw_ansi_fputs): Do not call _close() on the
5059         handle returned by _get_osf_handle().
5061 2018-08-13  Will Schmidt  <will_schmidt@vnet.ibm.com>
5063         * gcc/config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support
5064         for folding vec_perm.
5066 2018-08-13  Will Schmidt  <will_schmidt@vnet.ibm.com>
5068         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin):
5069         Add support for gimple-folding of vec_pack() and vec_unpack()
5070         intrinsics.
5072 2018-08-13  Will Schmidt <will_schmidt@vnet.ibm.com>
5074         * config/rs6000/rs6000.c (rs6000_builtin_valid_without_lhs): Add
5075         vec_xst variants to the list.
5076         (rs6000_gimple_fold_builtin): Add support for folding unaligned
5077         vector loads and stores.
5079 2018-08-13  David Edelsohn  <dje.gcc@gmail.com>
5081         * config.gcc (rs6000-ibm-aix4.x): Delete.
5082         (rs6000-ibm-aix5.1): Delete.
5083         (rs6000-ibm-aix5.2): Delete.
5084         (rs6000-ibm-aix5.3): Delete.
5085         * config/rs6000/aix43.h: Delete.
5086         * config/rs6000/aix51.h: Delete.
5087         * config/rs6000/aix52.h: Delete.
5088         * config/rs6000/t-aix43: Delete.
5090 2018-08-13  Ilya Leoshkevich  <iii@linux.ibm.com>
5092         * config/s390/s390.c (s390_decompose_constant_pool_ref):
5093         New function.
5094         (s390_decompose_address): Factor out constant pool ref
5095         decomposition.
5097 2018-08-12  Chung-Ju Wu  <jasonwucj@gmail.com>
5099         * config/nds32/nds32-predicates.c
5100         (nds32_can_use_bclr_p): Change return type as bool.
5101         (nds32_can_use_bset_p): Ditto.
5102         (nds32_can_use_btgl_p): Ditto.
5103         (nds32_can_use_bitci_p): Ditto.
5104         * config/nds32/nds32-protos.h
5105         (nds32_can_use_bclr_p): Change declaration.
5106         (nds32_can_use_bset_p): Ditto.
5107         (nds32_can_use_btgl_p): Ditto.
5108         (nds32_can_use_bitci_p): Ditto.
5110 2018-08-12  Chung-Ju Wu  <jasonwucj@gmail.com>
5112         * config/nds32/nds32.c (nds32_expand_prologue, nds32_expand_epilogue):
5113         Support -msched-prolog-epilog option.
5114         * config/nds32/nds32.opt (msched-prolog-epilog): New option.
5116 2018-08-12  Chung-Ju Wu  <jasonwucj@gmail.com>
5118         * common/config/nds32/nds32-common.c
5119         (nds32_option_optimization_table): Enalbe -malways-align.
5121 2018-08-12  Chung-Ju Wu  <jasonwucj@gmail.com>
5123         * config.gcc (nds32*): Add nds32_isr.h and nds32_init.inc in
5124         extra_headers.
5125         * common/config/nds32/nds32-common.c (nds32_handle_option): Handle
5126         OPT_misr_secure_ case.
5127         * config/nds32/nds32-isr.c: Implementation of backward compatibility.
5128         * config/nds32/nds32-protos.h (nds32_isr_function_critical_p): New.
5129         * config/nds32/nds32.c (nds32_attribute_table): Add critical and
5130         secure attribute.
5131         * config/nds32/nds32.h (nds32_isr_nested_type): Add NDS32_CRITICAL.
5132         (nds32_isr_info): New field security_level.
5133         (TARGET_ISR_VECTOR_SIZE_4_BYTE): New macro.
5134         * config/nds32/nds32.md (return_internal): Consider critical attribute.
5135         * config/nds32/nds32.opt (misr-secure): New option.
5136         * config/nds32/nds32_init.inc: New file.
5137         * config/nds32/nds32_isr.h: New file.
5139 2018-08-11  John David Anglin  <danglin@gcc.gnu.org>
5141         * config/pa/pa.md (UNSPEC_MEMORY_BARRIER): New unspec enum.
5142         Update comment for atomic instructions.
5143         (atomic_storeqi, atomic_storehi, atomic_storesi, atomic_storesf,
5144         atomic_loaddf, atomic_loaddf_1, atomic_storedf, atomic_storedf_1):
5145         Remove.
5146         (atomic_loaddi): Revise fence expansion to only emit fence prior to
5147         load for __ATOMIC_SEQ_CST model.
5148         (atomic_loaddi_1): Remove float register target.
5149         (atomic_storedi): Handle CONST_INT values.
5150         (atomic_storedi_1): Remove float register source.  Add special case
5151         for zero value.
5152         (memory_barrier): New expander and insn.
5154 2018-08-11  Jakub Jelinek  <jakub@redhat.com>
5156         PR tree-optimization/86835
5157         * tree-ssa-math-opts.c (insert_reciprocals): Even when inserting
5158         new_stmt after def_gsi, make sure to insert new_square_stmt after
5159         that stmt, not 2 stmts before it.
5161 2018-08-10  Alexander Monakov  <amonakov@ispras.ru>
5163         PR target/82418
5164         * config/i386/i386.md (<s>mul<mode>3_highpart): Use DWIH mode iterator
5165         instead of SWI48.
5167 2018-08-10  Martin Liska  <mliska@suse.cz>
5169         PR target/83610
5170         * builtin-types.def (BT_FN_LONG_LONG_LONG_DOUBLE): Add new
5171         function type.
5172         * builtins.c (expand_builtin_expect_with_probability):
5173         New function.
5174         (expand_builtin_expect_with_probability): New function.
5175         (build_builtin_expect_predicate): Add new argumnet probability
5176         for BUILT_IN_EXPECT_WITH_PROBABILITY.
5177         (fold_builtin_expect):
5178         (fold_builtin_2):
5179         (fold_builtin_3):
5180         * builtins.def (BUILT_IN_EXPECT_WITH_PROBABILITY):
5181         * builtins.h (fold_builtin_expect): Set new argument.
5182         * doc/extend.texi: Document __builtin_expect_with_probability.
5183         * doc/invoke.texi: Likewise.
5184         * gimple-fold.c (gimple_fold_call): Pass new argument.
5185         * ipa-fnsummary.c (find_foldable_builtin_expect): Handle
5186         also BUILT_IN_EXPECT_WITH_PROBABILITY.
5187         * predict.c (get_predictor_value): New function.
5188         (expr_expected_value): Add new argument probability. Assume
5189         that predictor and probability are always non-null.
5190         (expr_expected_value_1): Likewise.  For __builtin_expect and
5191         __builtin_expect_with_probability set probability.  Handle
5192         combination in binary expressions.
5193         (tree_predict_by_opcode): Simplify code by simply calling
5194         get_predictor_value.
5195         (pass_strip_predict_hints::execute): Add handling of
5196         BUILT_IN_EXPECT_WITH_PROBABILITY.
5197         * predict.def (PRED_BUILTIN_EXPECT_WITH_PROBABILITY): Add
5198         new predictor.
5199         * tree.h (DECL_BUILT_IN_P): New function.
5201 2018-08-10  Martin Liska  <mliska@suse.cz>
5203         PR tree-optimization/85799
5204         * passes.def: Add argument for pass_strip_predict_hints.
5205         * predict.c (class pass_strip_predict_hints): Add new argument
5206         early_p.
5207         (strip_predictor_early): New function.
5208         (pass_strip_predict_hints::execute): Call the function to
5209         strip predictors.
5210         (strip_predict_hints): New function.
5211         * predict.def: Fix comment.
5213 2018-08-10  Thomas Preud'homme  <thomas.preudhomme@linaro.org>
5215         * Makefile.in: Clarify which tm.texi to copy over to assert the
5216         right to grant a GFDL license for all.
5218 2018-08-09  Jeff Law  <law@redhat.com>
5220         * config/m68k/m68k.c (m68k_adjust_decorated_operand): Remove
5221         unused variable.
5223 2018-08-09  Andreas Schwab  <schwab@linux-m68k.org>
5225         * config/m68k/m68k-protos.h (m68k_final_prescan_insn): Remove
5226         prototype.
5228 2018-08-09  Richard Sandiford  <richard.sandiford@arm.com>
5230         * tree-vect-loop.c (vectorizable_reduction): Allow inner-loop
5231         reductions for variable-length vectors.
5233 2018-08-09  David Malcolm  <dmalcolm@redhat.com>
5235         PR other/84889
5236         * common.opt (fdiagnostics-show-line-numbers): New option.
5237         * diagnostic-show-locus.c (class layout): Add fields
5238         "m_show_line_numbers_p" and "m_linenum_width";
5239         (num_digits): New function.
5240         (test_num_digits): New function.
5241         (layout::layout): Initialize new fields.  Update m_x_offset
5242         logic to handle any left margin.
5243         (layout::print_source_line): Print line number when requested.
5244         (layout::start_annotation_line): New member function.
5245         (layout::print_annotation_line): Call it.
5246         (layout::print_leading_fixits): Likewise.
5247         (layout::print_trailing_fixits): Likewise.  Update calls to
5248         move_to_column for new parameter.
5249         (layout::get_x_bound_for_row): Add "add_left_margin" param and use
5250         it to potentially call start_annotation_line.
5251         (layout::show_ruler): Call start_annotation_line.
5252         (selftest::test_line_numbers_multiline_range): New selftest.
5253         (selftest::diagnostic_show_locus_c_tests): Call test_num_digits
5254         and selftest::test_line_numbers_multiline_range.
5255         * diagnostic.c (diagnostic_initialize): Initialize
5256         show_line_numbers_p.
5257         * diagnostic.h (struct diagnostic_context): Add field
5258         "show_line_numbers_p".
5259         * doc/invoke.texi (Diagnostic Message Formatting Options): Add
5260         -fno-diagnostics-show-line-numbers.
5261         * dwarf2out.c (gen_producer_string): Add
5262         OPT_fdiagnostics_show_line_numbers to the ignored options.
5263         * lto-wrapper.c (merge_and_complain): Likewise to the "pick
5264         one setting" options.
5265         (append_compiler_options): Likewise to the dropped options.
5266         (append_diag_options): Likewise to the passed-on options.
5267         * opts.c (common_handle_option): Handle the new option.
5268         * toplev.c (general_init): Set up global_dc->show_line_numbers_p.
5270 2018-08-09  Kelvin Nilsen  <kelvin@gcc.gnu.org>
5272         * doc/extend.texi (PowerPC AltiVec Built-in Functions Available on
5273         ISA 2.07): Correct spelling of bcdsub to be __builtin_bcdsub.  Add
5274         third argument of type "const signed char" to descriptions of
5275         __builtin_bcdadd, __builtin_bcdadd_lt, __builtin_bcdadd_eq,
5276         __builtin_bcdadd_gt, __builtin_bcdadd_ov, __builtin_bcdsub,
5277         __builtin_bcdsub_lt, __builtin_bcdsub_eq, __builtin_bcdsub_gt,
5278         __builtin_bcdsub_ov functions.
5280 2018-08-09  Richard Sandiford  <richard.sandiford@arm.com>
5282         PR tree-optimization/86858
5283         * tree-vect-loop.c (vect_is_simple_reduction): Restore
5284         flow_bb_inside_loop_p calls.
5286 2018-08-09  Richard Sandiford  <richard.sandiford@arm.com>
5288         PR tree-optimization/86871
5289         * tree-vect-stmts.c (vect_transform_stmt): Use gimple_get_lhs
5290         instead of gimple_assign_lhs.
5292 2018-08-09  Richard Earnshaw  <rearnsha@arm.com>
5294         PR target/86887
5295         * config/aarch64/aarch64.md (add<mode>3_carryinC_zero): Add missing
5296         register constraint to operand 0.
5297         (add<mode>3_carryinC): Likewise.
5298         (add<mode>3_carryinV_zero, add<mode>3_carryinV): Likewise.
5300 2018-08-09  Martin Liska  <mliska@suse.cz>
5302         PR c/86895
5303         * common.opt: Remove extra line.
5305 2018-08-09  Martin Liska  <mliska@suse.cz>
5307         * params.def (PARAM_ALIGN_LOOP_ITERATIONS): Remove double dots
5308         at the end of a line, make first letter capital and end up
5309         a sentence with a dot.
5310         (PARAM_LOOP_INTERCHANGE_STRIDE_RATIO): Likewise.
5311         (PARAM_LOOP_BLOCK_TILE_SIZE): Likewise.
5312         (PARAM_GRAPHITE_MAX_NB_SCOP_PARAMS): Likewise.
5313         (PARAM_GRAPHITE_MAX_ARRAYS_PER_SCOP): Likewise.
5314         (PARAM_MAX_ISL_OPERATIONS): Likewise.
5315         (PARAM_GRAPHITE_ALLOW_CODEGEN_ERRORS): Likewise.
5316         (PARAM_PROFILE_FUNC_INTERNAL_ID): Likewise.
5317         (PARAM_INDIR_CALL_TOPN_PROFILE): Likewise.
5318         (PARAM_SLP_MAX_INSNS_IN_BB): Likewise.
5319         (PARAM_IPA_CP_EVAL_THRESHOLD): Likewise.
5320         (PARAM_IPA_CP_RECURSION_PENALTY): Likewise.
5321         (PARAM_IPA_CP_SINGLE_CALL_PENALTY): Likewise.
5322         (PARAM_IPA_CP_LOOP_HINT_BONUS): Likewise.
5323         (PARAM_IPA_CP_ARRAY_INDEX_HINT_BONUS): Likewise.
5324         (PARAM_TREE_REASSOC_WIDTH): Likewise.
5325         (PARAM_HSA_GEN_DEBUG_STORES): Likewise.
5326         (PARAM_MAX_SPECULATIVE_DEVIRT_MAYDEFS): Likewise.
5327         (PARAM_MAX_VRP_SWITCH_ASSERTIONS): Likewise.
5329 2018-08-09  Andreas Krebbel  <krebbel@linux.ibm.com>
5331         PR target/84332
5332         * config/s390/s390.c (s390_option_override_internal): Reduce the
5333         stack-clash-protection-probe-interval param if it would be too big
5334         for z900.
5336 2018-08-08  Andreas Schwab  <schwab@linux-m68k.org>
5338         PR target/46179
5339         * config/m68k/m68k.h (FINAL_PRESCAN_INSN): Don't define.
5340         * config/m68k/m68k.c (handle_move_double): Don't call
5341         m68k_final_prescan_insn.
5342         (m68k_adjust_decorated_operand): Renamed from
5343         m68k_final_prescan_insn, remove first and third operand and
5344         simplify.
5345         (print_operand): Call it.
5346         (print_operand_address): Call it.
5348 2018-08-08  Nathan Sidwell  <nathan@acm.org>
5350         * diagnostic.c (diagnostic_report_current_module): Use
5351         linemap_included_from & linemap_included_from_linemap.
5353 2018-08-08  Hongbo Zhang  <hongbo.zhang@linaro.org>
5355         * config/aarch64/aarch64-cores.def: Add phecda core.
5356         * config/aarch64/aarch64-tune.md: Regenerate.
5357         * doc/invoke.texi: Add phecda core.
5359 2018-08-08  Andreas Krebbel  <krebbel@linux.ibm.com>
5361         PR target/85295
5362         * config/s390/constraints.md ("NxHD0", "NxSD0"): New constraint
5363         definitions.
5364         * config/s390/s390.md ("movti"): Add more alternatives for
5365         constant to GPR copies.
5367 2018-08-08  Andreas Krebbel  <krebbel@linux.ibm.com>
5369         * config/s390/s390.c: Fix whitespace damage throughout the file.
5370         * config/s390/s390.h: Likewise.
5371         * config/s390/tpf.h: Likewise.
5373 2018-08-08  Ilya Leoshkevich  <iii@linux.ibm.com>
5375         * config/s390/s390.c (s390_loadrelative_operand_p):
5376         Remove TARGET_CPU_ZARCH usages.
5377         (s390_rtx_costs): Likewise.
5378         (s390_legitimate_constant_p): Likewise.
5379         (s390_cannot_force_const_mem): Likewise.
5380         (legitimate_reload_constant_p): Likewise.
5381         (s390_preferred_reload_class): Likewise.
5382         (legitimize_pic_address): Likewise.
5383         (legitimize_tls_address): Likewise.
5384         (s390_split_branches): Removed.
5385         (s390_add_execute): Removed.
5386         (s390_dump_pool): Remove TARGET_CPU_ZARCH usages.
5387         (s390_mainpool_start): Likewise.
5388         (s390_mainpool_finish): Likewise.
5389         (s390_mainpool_cancel): Removed.
5390         (s390_chunkify_start): Remove TARGET_CPU_ZARCH usages.
5391         (s390_chunkify_cancel): Likewise.
5392         (s390_return_addr_rtx): Likewise.
5393         (s390_register_info): Remove split_branches_pending_p uages.
5394         (s390_optimize_register_info): Likewise.
5395         (s390_init_frame_layout): Remove TARGET_CPU_ZARCH and
5396         split_branches_pending_p usages.
5397         (s390_can_eliminate): Remove TARGET_CPU_ZARCH usages.
5398         (s390_load_got): Likewise.
5399         (s390_expand_split_stack_prologue): Likewise.
5400         (output_asm_nops): Likewise.
5401         (s390_function_profiler): Likewise.
5402         (s390_emit_call): Likewise.
5403         (s390_conditional_register_usage): Likewise.
5404         (s390_optimize_prologue): Likewise.
5405         (s390_reorg): Remove TARGET_CPU_ZARCH and
5406         split_branches_pending_p usages.
5407         (s390_option_override_internal): Remove TARGET_CPU_ZARCH
5408         usages.
5409         (s390_output_indirect_thunk_function): Likewise.
5410         * config/s390/s390.h (TARGET_CPU_ZARCH): Removed.
5411         (TARGET_CPU_ZARCH_P): Removed.
5412         (struct machine_function): Remove split_branches_pending_p.
5413         * config/s390/s390.md: Remove TARGET_CPU_ZARCH usages.
5415 2018-08-08  Ilya Leoshkevich  <iii@linux.ibm.com>
5417         * common/config/s390/s390-common.c (processor_flags_table):
5418         Remove flags.
5419         * config.gcc: Remove with_arch/with_tune support.
5420         * config/s390/2064.md: Remove cpu attribute comparisons.
5421         * config/s390/driver-native.c (s390_host_detect_local_cpu):
5422         Remove MTN.
5423         * config/s390/linux.h (ASM_SPEC):
5424         Remove -march support.
5425         * config/s390/s390-c.c (s390_cpu_cpp_builtins_internal):
5426         Use a table to get an arch level.
5427         * config/s390/s390-opts.h (enum processor_type):
5428         Remove enum values.
5429         * config/s390/s390.c
5430         (processor_table): Remove entries, add arch_level values.
5431         (s390_issue_rate): Remove cases.
5432         (s390_option_override): Adjust
5433         s390_option_override_internal() call.
5434         (s390_option_override_internal): Remove deprecation warning.
5435         (s390_valid_target_attribute_tree): Adjust
5436         s390_option_override_internal() call.
5437         * config/s390/s390.h (struct s390_processor):
5438         Share with s390-c.c, add arch_level field.
5439         * config/s390/s390.md:
5440         Remove occurrences in cpu attribute.
5441         * config/s390/s390.opt: Remove -march/-mtune support.
5442         * config/s390/tpf.h (ASM_SPEC): Remove -march support.
5443         * doc/invoke.texi: Remove deprecation warning.
5445 2018-08-08  Luis Machado  <luis.machado@linaro.org>
5447         * config/aarch64/aarch64.c (qdf24xx_vector_cost): New static
5448         global.
5449         (qdf24xx_tunings): Set vector cost structure to
5450         qdf24xx_vector_cost.
5452         * config/aarch64/aarch64.c (qdf24xx_addrcost_table)
5453         <register_sextend>: Set to 3.
5455 2018-08-07  Richard Sandiford  <richard.sandiford@arm.com>
5457         PR target/86838
5458         * config/aarch64/iterators.md (FRECP, frecp_suffix): Delete.
5459         * config/aarch64/aarch64-simd.md
5460         (aarch64_frecp<FRECP:frecp_suffix><mode>): Fold FRECPE into...
5461         (@aarch64_frecpe<mode>): ...here and the move FRECPX to...
5462         (aarch64_frecpx<mode>): ...this new pattern.
5463         * config/aarch64/aarch64-simd-builtins.def: Remove comment
5464         about aarch64_frecp<FRECP:frecp_suffix><mode>.
5466 2018-08-07  Martin Liska  <mliska@suse.cz>
5468         PR middle-end/83023
5469         * predict.c (expr_expected_value_1): Handle DECL_IS_MALLOC,
5470         BUILT_IN_REALLOC and DECL_IS_OPERATOR_NEW.
5471         * predict.def (PRED_MALLOC_NONNULL): New predictor.
5472         * doc/extend.texi: Document that malloc attribute adds
5473         hit to compiler.
5475 2018-08-06  John David Anglin  <danglin@gcc.gnu.org>
5477         PR target/86785
5478         * config/pa/pa.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
5479         Define to speculation_safe_value_not_needed.
5481 2018-08-06  Jeff Law  <law@redhat.com>
5483         * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Pass down
5484         the vr_values instance to cprop_into_stmt.
5485         (cprop_into_stmt): Pass vr_values instance down to cprop_operand.
5486         (cprop_operand): Also query EVRP to determine if OP is a constant.
5488 2018-08-06  Nathan Sidwell  <nathan@acm.org>
5490         * diagnostic.c (diagnostic_report_current_module): Reroll
5491         included-at loop.  Translate text.
5493 2018-08-06  David Malcolm  <dmalcolm@redhat.com>
5495         * function-tests.c (selftest::test_expansion_to_rtl): Call
5496         free_after_compilation.
5498 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
5500         * config/aarch64/aarch64.md: Add clobber highs to tls_desc.
5502 2018-08-06  Andreas Krebbel  <krebbel@linux.ibm.com>
5504         * config/s390/s390.c (s390_loop_unroll_adjust): Prevent small
5505         loops with memory block operations from getting unrolled.
5507 2018-08-06  Ulrich Weigand  <uweigand@de.ibm.com>
5509         PR target/86807
5510         * config/spu/spu.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
5511         Define to speculation_safe_value_not_needed.
5513 2018-08-06  Jeff Law  <law@redhat.com>
5515         * reload1.c (forget_old_reloads_1): Adjust CLOBBER_HIGH
5516         assert.
5518 2018-08-06  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
5520         PR target/86662
5521         * gcc/tree.c (build_common_tree_nodes): Initialize integer_types array
5522         with all enabled __intN types.
5524         * gcc/testsuite/gcc.target/msp430/pr86662.c: New test.
5526 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
5528         * alias.c (record_set): Check for clobber high.
5529         * cfgexpand.c (expand_gimple_stmt): Likewise.
5530         * combine-stack-adj.c (single_set_for_csa): Likewise.
5531         * combine.c (find_single_use_1): Likewise.
5532         (set_nonzero_bits_and_sign_copies): Likewise.
5533         (get_combine_src_dest): Likewise.
5534         (is_parallel_of_n_reg_sets): Likewise.
5535         (try_combine): Likewise.
5536         (record_dead_and_set_regs_1): Likewise.
5537         (reg_dead_at_p_1): Likewise.
5538         (reg_dead_at_p): Likewise.
5539         * dce.c (deletable_insn_p): Likewise.
5540         (mark_nonreg_stores_1): Likewise.
5541         (mark_nonreg_stores_2): Likewise.
5542         * df-scan.c (df_find_hard_reg_defs): Likewise.
5543         (df_uses_record): Likewise.
5544         (df_get_call_refs): Likewise.
5545         * dwarf2out.c (mem_loc_descriptor): Likewise.
5546         * haifa-sched.c (haifa_classify_rtx): Likewise.
5547         * ira-build.c (create_insn_allocnos): Likewise.
5548         * ira-costs.c (scan_one_insn): Likewise.
5549         * ira.c (equiv_init_movable_p): Likewise.
5550         (rtx_moveable_p): Likewise.
5551         (interesting_dest_for_shprep): Likewise.
5552         * jump.c (mark_jump_label_1): Likewise.
5553         * postreload-gcse.c (record_opr_changes): Likewise.
5554         * postreload.c (reload_cse_simplify): Likewise.
5555         (struct reg_use): Add source expr.
5556         (reload_combine): Check for clobber high.
5557         (reload_combine_note_use): Likewise.
5558         (reload_cse_move2add): Likewise.
5559         (move2add_note_store): Likewise.
5560         * print-rtl.c (print_pattern): Likewise.
5561         * recog.c (decode_asm_operands): Likewise.
5562         (store_data_bypass_p): Likewise.
5563         (if_test_bypass_p): Likewise.
5564         * regcprop.c (kill_clobbered_value): Likewise.
5565         (kill_set_value): Likewise.
5566         * reginfo.c (reg_scan_mark_refs): Likewise.
5567         * reload1.c (maybe_fix_stack_asms): Likewise.
5568         (eliminate_regs_1): Likewise.
5569         (elimination_effects): Likewise.
5570         (mark_not_eliminable): Likewise.
5571         (scan_paradoxical_subregs): Likewise.
5572         (forget_old_reloads_1): Likewise.
5573         * reorg.c (find_end_label): Likewise.
5574         (try_merge_delay_insns): Likewise.
5575         (redundant_insn): Likewise.
5576         (own_thread_p): Likewise.
5577         (fill_simple_delay_slots): Likewise.
5578         (fill_slots_from_thread): Likewise.
5579         (dbr_schedule): Likewise.
5580         * resource.c (update_live_status): Likewise.
5581         (mark_referenced_resources): Likewise.
5582         (mark_set_resources): Likewise.
5583         * rtl.c (copy_rtx): Likewise.
5584         * rtlanal.c (reg_referenced_p): Likewise.
5585         (single_set_2): Likewise.
5586         (noop_move_p): Likewise.
5587         (note_stores): Likewise.
5588         * sched-deps.c (sched_analyze_reg): Likewise.
5589         (sched_analyze_insn): Likewise.
5591 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
5593         * cse.c (invalidate_reg): New function extracted from...
5594         (invalidate): ...here.
5595         (canonicalize_insn): Check for clobber high.
5596         (invalidate_from_clobbers): invalidate clobber highs.
5597         (invalidate_from_sets_and_clobbers): Likewise.
5598         (count_reg_usage): Check for clobber high.
5599         (insn_live_p): Likewise.
5600         * cselib.c (cselib_expand_value_rtx_1):Likewise.
5601         (cselib_invalidate_regno): Check for clobber in setter.
5602         (cselib_invalidate_rtx): Pass through setter.
5603         (cselib_invalidate_rtx_note_stores):
5604         (cselib_process_insn): Check for clobber high.
5605         * cselib.h (cselib_invalidate_rtx): Add operand.
5607 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
5609         * lra-eliminations.c (lra_eliminate_regs_1): Check for clobber high.
5610         (mark_not_eliminable): Likewise.
5611         * lra-int.h (struct lra_insn_reg): Add clobber high marker.
5612         * lra-lives.c (process_bb_lives): Check for clobber high.
5613         * lra.c (new_insn_reg): Remember clobber highs.
5614         (collect_non_operand_hard_regs): Check for clobber high.
5615         (lra_set_insn_recog_data): Likewise.
5616         (add_regs_to_insn_regno_info): Likewise.
5617         (lra_update_insn_regno_info): Likewise.
5619 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
5621         * rtl.h (reg_is_clobbered_by_clobber_high): Add declarations.
5622         * rtlanal.c (reg_is_clobbered_by_clobber_high): Add function.
5624 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
5626         * emit-rtl.c (verify_rtx_sharing): Check for CLOBBER_HIGH.
5627         (copy_insn_1): Likewise.
5628         (gen_hard_reg_clobber_high): New gen function.
5629         * genconfig.c (walk_insn_part): Check for CLOBBER_HIGH.
5630         * genemit.c (gen_exp): Likewise.
5631         (gen_emit_seq): Pass through info.
5632         (gen_insn): Check for CLOBBER_HIGH.
5633         (gen_expand): Pass through info.
5634         (gen_split): Likewise.
5635         (output_add_clobbers): Likewise.
5636         * genrecog.c (validate_pattern): Check for CLOBBER_HIGH.
5637         (remove_clobbers): Likewise.
5638         * rtl.h (gen_hard_reg_clobber_high): New declaration.
5640 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
5642         * doc/rtl.texi (clobber_high): Add.
5643         (parallel): Add in clobber high
5644         * rtl.c (rtl_check_failed_code3): Add function.
5645         * rtl.def (CLOBBER_HIGH): Add expression.
5646         * rtl.h (RTL_CHECKC3): Add macro.
5647         (rtl_check_failed_code3): Add declaration.
5648         (XC3EXP): Add macro.
5650 2018-08-05  H.J. Lu  <hongjiu.lu@intel.com>
5652         PR target/86386
5653         * config/i386/i386.c (ix86_finalize_stack_frame_flags): Set
5654         cfun->machine->max_used_stack_alignment if needed.
5656 2018-08-04  Martin Sebor  <msebor@redhat.com>
5658         PR tree-optimization/86571
5659         * gimple-ssa-sprintf.c (format_floating): Extend upper bound of
5660         NaN output to 4.
5662 2018-08-03  Sandra Loosemore  <sandra@codesourcery.com>
5664         PR target/86799
5665         * config/nios2/nios2.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
5666         Define.
5668 2018-08-03  Jeff Law  <law@redhat.com>
5670         PR target/86795
5671         * config/mn10300/mn10300.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
5672         Define to speculation_safe_value_not_needed.
5674 2018-08-03  David Malcolm  <dmalcolm@redhat.com>
5676         * doc/gcov.texi (-x): Remove duplicate "to".
5677         * doc/invoke.texi (-Wnoexcept-type): Remove duplicate "calls".
5678         (-Wif-not-aligned): Remove duplicate "is".
5679         (-flto): Remove duplicate "the".
5680         (MicroBlaze Options): In examples of "-mcpu=cpu-type", remove
5681         duplicate "v5.00.b".
5682         (MSP430 Options): Remove duplicate "and" from the description
5683         of "-mgprel-sec=regexp".
5684         (x86 Options): Remove duplicate copies of "vmldLog102" and
5685         vmlsLog104 from description of "-mveclibabi=type".
5687 2018-08-03  Richard Sandiford  <richard.sandiford@arm.com>
5689         * internal-fn.h (first_commutative_argument): Declare.
5690         * internal-fn.c (first_commutative_argument): New function.
5691         * tree-vect-slp.c (vect_get_and_check_slp_defs): Remove extra
5692         restrictions for pattern statements.  Use first_commutative_argument
5693         to look for commutative operands in calls to internal functions.
5695 2018-08-03  Aldy Hernandez  <aldyh@redhat.com>
5697         * Makefile.in (wide-int-range.o): New.
5698         * tree-vrp.c: Move all the wide_int_* functions to...
5699         * wide-int-range.cc: ...here.
5700         * tree-vrp.h: Move all the wide_int_* prototypes to...
5701         * wide-int-range.h: ...here.
5703 2018-08-03  Tom de Vries  <tdevries@suse.de>
5705         * common/config/nvptx/nvptx-common.c (nvptx_except_unwind_info): Return
5706         UI_NONE.
5707         * config/nvptx/nvptx.c (TARGET_ASM_BYTE_OP): Remove define.
5708         * except.c (output_function_exception_table): Do early exit if
5709         targetm_common.except_unwind_info (&global_options) == UI_NONE.
5711 2018-08-03  Martin Liska  <mliska@suse.cz>
5713         * predict.c (dump_prediction): Change to 2 digits
5714         in fraction part.
5716 2018-08-03  Siddhesh Poyarekar  <siddhesh@sourceware.org>
5718         * config/aarch64/falkor.md (falkor_am_1_vxvy_vxvy): Move
5719         neon_dup_q to...
5720         (falkor_am_1_gtov_gtov): ... a new insn reservation.
5722 2018-07-19  Ilya Leoshkevich  <iii@linux.ibm.com>
5724         * config/nds32/nds32.c (nds32_hard_regno_mode_ok): Replace > with >=.
5725         * df-problems.c (df_remove_dead_eq_notes): Replace > with >=.
5726         * dwarf2out.c (mem_loc_descriptor): Replace > with >=.
5727         * lra-constraints.c (spill_hard_reg_in_range): Replace <= with <.
5728         * lra-remat.c (call_used_input_regno_present_p): Replace <= with <.
5730 2018-08-02  David Malcolm  <dmalcolm@redhat.com>
5732         * diagnostic-show-locus.c (diagnostic_show_locus): Use
5733         pp_take_prefix when saving the existing prefix.
5734         * diagnostic.c (diagnostic_append_note): Likewise.
5735         * langhooks.c (lhd_print_error_function): Likewise.
5736         * pretty-print.c (pp_set_prefix): Drop the "const" from "prefix"
5737         param's type.  Free the existing prefix.
5738         (pp_take_prefix): New function.
5739         (pretty_printer::pretty_printer): Drop the prefix parameter.
5740         Rename the length parameter to match the comment.
5741         (pretty_printer::~pretty_printer): Free the prefix.
5742         * pretty-print.h (pretty_printer::pretty_printer): Drop the prefix
5743         parameter.
5744         (struct pretty_printer): Drop the "const" from "prefix" field's
5745         type and clarify memory management.
5746         (pp_set_prefix): Drop the "const" from the 2nd param.
5747         (pp_take_prefix): New decl.
5749 2018-08-02  Aaron Sawdey  <acsawdey@linux.ibm.com>
5751         * config/rs6000/rs6000-string.c (select_block_compare_mode): Move test
5752         for word_mode_ok here instead of passing as argument.
5753         (expand_block_compare): Change select_block_compare_mode() call.
5754         (expand_strncmp_gpr_sequence): New function.
5755         (expand_strn_compare): Make use of expand_strncmp_gpr_sequence.
5757 2018-08-02  Jeff Law  <law@redhat.com>
5759         PR target/86790
5760         * config/m68k/m68k.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
5761         Define to speculation_safe_value_not_needed.
5763         PR target/86784
5764         * config/h8300/h8300.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
5765         Define to speculation_safe_value_not_needed.
5767 2018-08-02  Tom de Vries  <tdevries@suse.de>
5769         PR target/86660
5770         * common/config/nvptx/nvptx-common.c (nvptx_except_unwind_info): New
5771         function.  Return UI_TARGET unconditionally.
5772         (TARGET_EXCEPT_UNWIND_INFO): Redefine to nvptx_except_unwind_info.
5773         * config/nvptx/nvptx.c (TARGET_ASM_BYTE_OP): Emit commented out '.byte'.
5775 2018-08-02  Richard Sandiford  <richard.sandiford@arm.com>
5777         * genemit.c (print_overload_test): Fix typo.
5779 2018-08-02  Richard Biener  <rguenther@suse.de>
5781         PR tree-optimization/86816
5782         * tree-ssa-tail-merge.c (tail_merge_valueize): New function
5783         which checks for value availability before querying it.
5784         (gvn_uses_equal): Use it.
5785         (same_succ_hash): Likewise.
5786         (gimple_equal_p): Likewise.
5788 2018-08-02  Nick Clifton  <nickc@redhat.com>
5790         PR target/86813
5791         * config/stormy16/stormy16.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
5792         Define to speculation_safe_value_not_needed.
5794         PR target/86810
5795         * config/v850/v850.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
5796         Define to speculation_safe_value_not_needed.
5798         PR target/86810
5799         * config/v850/v850.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
5800         Define to speculation_safe_value_not_needed.
5802         PR target/86803
5803         * config/rx/rx.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
5804         Define to speculation_safe_value_not_needed.
5806         PR target/86797
5807         * config/msp430/msp430.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
5808         Define to speculation_safe_value_not_needed.
5810         PR target/86791
5811         * config/mcore/mcore.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
5812         Define to speculation_safe_value_not_needed.
5814         PR target/86789
5815         * config/m32r/m32r.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
5816         Define to speculation_safe_value_not_needed.
5818         PR target/86787
5819         * config/iq2000/iq2000.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
5820         Define to speculation_safe_value_not_needed.
5822         PR target/86782
5823         * config/frv/frv.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define to
5824         speculation_safe_value_not_needed.
5826         PR target/86781
5827         * config/fr30/fr30.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define
5828         to speculation_safe_value_not_needed.
5830 2018-08-02  Richard Sandiford  <richard.sandiford@arm.com>
5832         * doc/md.texi: Expand the documentation of instruction names
5833         to mention port-local uses.  Document '@' in pattern names.
5834         * read-md.h (overloaded_instance, overloaded_name): New structs.
5835         (mapping): Declare.
5836         (md_reader::handle_overloaded_name): New member function.
5837         (md_reader::get_overloads): Likewise.
5838         (md_reader::m_first_overload): New member variable.
5839         (md_reader::m_next_overload_ptr): Likewise.
5840         (md_reader::m_overloads_htab): Likewise.
5841         * read-md.c (md_reader::md_reader): Initialize m_first_overload,
5842         m_next_overload_ptr and m_overloads_htab.
5843         * read-rtl.c (iterator_group): Add "type" and "get_c_token" fields.
5844         (get_mode_token, get_code_token, get_int_token): New functions.
5845         (map_attr_string): Add an optional argument that passes back
5846         the associated iterator.
5847         (overloaded_name_hash, overloaded_name_eq_p, named_rtx_p):
5848         (md_reader::handle_overloaded_name, add_overload_instance): New
5849         functions.
5850         (apply_iterators): Handle '@' names.  Report an error if '@'
5851         is used without iterators.
5852         (initialize_iterators): Initialize the new iterator_group fields.
5853         * genopinit.c (handle_overloaded_code_for)
5854         (handle_overloaded_gen): New functions.
5855         (main): Use them to print declarations of maybe_code_for_* and
5856         maybe_gen_* functions, and inline definitions of code_for_* and gen_*.
5857         * genemit.c (print_overload_arguments, print_overload_test)
5858         (handle_overloaded_code_for, handle_overloaded_gen): New functions.
5859         (main): Use it to print definitions of maybe_code_for_* and
5860         maybe_gen_* functions.
5861         * config/aarch64/aarch64.c (aarch64_split_128bit_move): Use
5862         gen_aarch64_mov{low,high}_di and gen_aarch64_movdi_{low,high}
5863         instead of explicit mode checks.
5864         (aarch64_split_simd_combine): Likewise gen_aarch64_simd_combine.
5865         (aarch64_split_simd_move): Likewise gen_aarch64_split_simd_mov.
5866         (aarch64_emit_load_exclusive): Likewise gen_aarch64_load_exclusive.
5867         (aarch64_emit_store_exclusive): Likewise gen_aarch64_store_exclusive.
5868         (aarch64_expand_compare_and_swap): Likewise
5869         gen_aarch64_compare_and_swap and gen_aarch64_compare_and_swap_lse
5870         (aarch64_gen_atomic_cas): Likewise gen_aarch64_atomic_cas.
5871         (aarch64_emit_atomic_swap): Likewise gen_aarch64_atomic_swp.
5872         (aarch64_constant_pool_reload_icode): Delete.
5873         (aarch64_secondary_reload): Use code_for_aarch64_reload_movcp
5874         instead of aarch64_constant_pool_reload_icode.  Use
5875         code_for_aarch64_reload_mov instead of explicit mode checks.
5876         (rsqrte_type, get_rsqrte_type, rsqrts_type, get_rsqrts_type): Delete.
5877         (aarch64_emit_approx_sqrt): Use gen_aarch64_rsqrte instead of
5878         get_rsqrte_type and gen_aarch64_rsqrts instead of gen_rqrts_type.
5879         (recpe_type, get_recpe_type, recps_type, get_recps_type): Delete.
5880         (aarch64_emit_approx_div): Use gen_aarch64_frecpe instead of
5881         get_recpe_type and gen_aarch64_frecps instead of get_recps_type.
5882         (aarch64_atomic_load_op_code): Delete.
5883         (aarch64_emit_atomic_load_op): Likewise.
5884         (aarch64_gen_atomic_ldop): Use UNSPECV_ATOMIC_* instead of
5885         aarch64_atomic_load_op_code.  Use gen_aarch64_atomic_load
5886         instead of aarch64_emit_atomic_load_op.
5887         * config/aarch64/aarch64.md (aarch64_reload_movcp<GPF_TF:mode><P:mode>)
5888         (aarch64_reload_movcp<VALL:mode><P:mode>, aarch64_reload_mov<mode>)
5889         (aarch64_movdi_<mode>low, aarch64_movdi_<mode>high)
5890         (aarch64_mov<mode>high_di, aarch64_mov<mode>low_di): Add a '@'
5891         character before the pattern name.
5892         * config/aarch64/aarch64-simd.md (aarch64_split_simd_mov<mode>)
5893         (aarch64_rsqrte<mode>, aarch64_rsqrts<mode>)
5894         (aarch64_simd_combine<mode>, aarch64_frecpe<mode>)
5895         (aarch64_frecps<mode>): Likewise.
5896         * config/aarch64/atomics.md (atomic_compare_and_swap<mode>)
5897         (aarch64_compare_and_swap<mode>, aarch64_compare_and_swap<mode>_lse)
5898         (aarch64_load_exclusive<mode>, aarch64_store_exclusive<mode>)
5899         (aarch64_atomic_swp<mode>, aarch64_atomic_cas<mode>)
5900         (aarch64_atomic_load<atomic_ldop><mode>): Likewise.
5902 2018-08-02  Richard Sandiford  <richard.sandiford@arm.com>
5904         * config/aarch64/aarch64.c (aarch64_float_const_representable_p):
5905         Allow HFmode constants if TARGET_FP_F16INST.
5907 2018-08-02  Jackson Woodruff  <jackson.woodruff@arm.com>
5909         PR target/86014
5910         * config/aarch64/aarch64.c (aarch64_operands_adjust_ok_for_ldpstp):
5911         No longer check last store for clobber of address register.
5913 2018-08-02  Martin Liska  <mliska@suse.cz>
5915         PR gcov-profile/86817
5916         * gcov.c (process_all_functions): New function.
5917         (main): Call it.
5918         (process_file): Move functions processing to
5919         process_all_functions.
5921 2018-08-02  David Malcolm  <dmalcolm@redhat.com>
5923         * dumpfile.c (dump_user_location_t::dump_user_location_t): Add
5924         "const" to the "gimple *" and "rtx_insn *" parameters.
5925         * dumpfile.h (dump_user_location_t::dump_user_location_t):
5926         Likewise.
5927         (dump_location_t::dump_location_t): Likewise.
5929 2018-08-01  Martin Sebor  <msebor@redhat.com>
5931         PR tree-optimization/86650
5932         * gimple-pretty-print.c (percent_G_format): Accept a "gimple *"
5933         rather than a "gcall *".  Directly pass the data of interest
5934         to percent_K_format, rather than building a temporary CALL_EXPR
5935         to hold it.
5936         * gimple-fold.c (gimple_fold_builtin_strncpy): Adjust.
5937         (gimple_fold_builtin_strncat): Adjust.
5938         * gimple-ssa-warn-restrict.h (check_bounds_or_overlap): Replace
5939         gcall* argument with gimple*.
5940         * gimple-ssa-warn-restrict.c (check_call): Same.
5941         (wrestrict_dom_walker::before_dom_children): Same.
5942         (builtin_access::builtin_access): Same.
5943         (check_bounds_or_overlap): Same
5944         (maybe_diag_overlap): Same.
5945         (maybe_diag_offset_bounds): Same.
5946         * tree-diagnostic.c (default_tree_printer): Move usage of
5947         EXPR_LOCATION (t) and TREE_BLOCK (t) from within percent_K_format
5948         to this callsite.
5949         * tree-pretty-print.c (percent_K_format): Add argument.
5950         * tree-pretty-print.h: Add argument.
5951         * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Adjust.
5952         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Adjust.
5953         (maybe_diag_stxncpy_trunc): Same.
5954         (handle_builtin_stxncpy): Same.
5955         (handle_builtin_strcat): Same.
5957 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
5959         * match.pd: Optimise pointer range checks.
5961 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
5963         PR tree-optimization/86758
5964         * tree-vect-stmts.c (vectorizable_simd_clone_call): Don't try
5965         to remove pattern statements.
5967 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
5969         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Use the
5970         result of dfs_enumerate_from when constructing stmt_vec_infos,
5971         instead of additionally calling get_loop_body.
5973 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
5975         * tree-vectorizer.h (vect_create_data_ref_ptr): Remove inv_p
5976         parameter.
5977         * tree-vect-data-refs.c (vect_create_data_ref_ptr): Likewise.
5978         When creating an iv, assert that the step is not known to be zero.
5979         (vect_setup_realignment): Update call accordingly.
5980         * tree-vect-stmts.c (vectorizable_store): Likewise.
5981         (vectorizable_load): Likewise.  Handle VMAT_INVARIANT separately.
5983 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
5985         * tree-vectorizer.h (vect_stmt_to_vectorize): New function.
5986         * tree-vect-loop.c (vect_update_vf_for_slp): Use it.
5987         (vectorizable_reduction): Likewise.
5988         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
5989         (vect_detect_hybrid_slp_stmts): Likewise.
5990         * tree-vect-stmts.c (vect_is_simple_use): Likewise.
5992 2018-08-01  Aldy Hernandez  <aldyh@redhat.com>
5994         * tree-vrp (zero_nonzero_bits_from_bounds): Rename to...
5995         (wide_int_set_zero_nonzero_bits): ...this.
5996         (zero_nonzero_bits_from_vr): Rename to...
5997         (vrp_set_zero_nonzero_bits): ...this.
5998         (extract_range_from_multiplicative_op_1): Abstract wide int
5999         code...
6000         (wide_int_range_multiplicative_op): ...here.
6001         (extract_range_from_binary_expr_1): Extract wide int binary
6002         operations into their own functions.
6003         (wide_int_range_lshift): New.
6004         (wide_int_range_can_optimize_bit_op): New.
6005         (wide_int_range_shift_undefined_p): New.
6006         (wide_int_range_bit_xor): New.
6007         (wide_int_range_bit_ior): New.
6008         (wide_int_range_bit_and): New.
6009         (wide_int_range_trunc_mod): New.
6010         (extract_range_into_wide_ints): New.
6011         (vrp_shift_undefined_p): New.
6012         (extract_range_from_multiplicative_op): New.
6013         (vrp_can_optimize_bit_op): New.
6014         * tree-vrp.h (value_range::dump): New.
6015         (wide_int_range_multiplicative_op): New.
6016         (wide_int_range_lshift):New.
6017         (wide_int_range_shift_undefined_p): New.
6018         (wide_int_range_bit_xor): New.
6019         (wide_int_range_bit_ior): New.
6020         (wide_int_range_bit_and): New.
6021         (wide_int_range_trunc_mod): New.
6022         (zero_nonzero_bits_from_bounds): Rename to...
6023         (wide_int_set_zero_nonzero_bits): ...this.
6024         (zero_nonzero_bits_from_vr): Rename to...
6025         (vrp_set_zero_nonzero_bits): ...this.
6026         (range_easy_mask_min_max): Rename to...
6027         (wide_int_range_can_optimize_bit_op): this.
6028         * vr-values.c (simplify_bit_ops_using_ranges): Rename
6029         zero_nonzero_bits_from_vr into vrp_set_zero_nonzero_bits.
6031 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
6033         * tree-vectorizer.h (vect_orig_stmt): New function.
6034         * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Use it.
6035         * tree-vect-loop.c (vect_model_reduction_cost): Likewise.
6036         (vect_create_epilog_for_reduction): Likewise.
6037         (vectorizable_live_operation): Likewise.
6038         * tree-vect-slp.c (vect_find_last_scalar_stmt_in_slp): Likewise.
6039         (vect_detect_hybrid_slp_stmts, vect_schedule_slp): Likewise.
6040         * tree-vect-stmts.c (vectorizable_call): Likewise.
6041         (vectorizable_simd_clone_call, vect_remove_stores): Likewise.
6043 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
6045         * tree-vectorizer.h (vect_transform_stmt): Remove grouped_store
6046         argument.
6047         * tree-vect-stmts.c (vect_transform_stmt): Likewise.
6048         * tree-vect-loop.c (vect_transform_loop_stmt): Update call accordingly.
6049         (vect_transform_loop): Likewise.
6050         * tree-vect-slp.c (vect_schedule_slp_instance): Likewise.
6052 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
6054         * tree-vectorizer.h (vect_schedule_slp): Return void.
6055         * tree-vect-slp.c (vect_schedule_slp_instance): Likewise.
6056         (vect_schedule_slp): Likewise.
6058 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
6060         * tree-vect-loop.c (vect_transform_loop_stmt): Remove slp_scheduled
6061         argument.
6062         (vect_transform_loop): Update calls accordingly.  Schedule SLP
6063         instances before the main loop, if any exist.
6065 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
6067         PR tree-optimization/86749
6068         * tree-vect-patterns.c (vect_determine_min_output_precision_1):
6069         If the lhs is used in a COND_EXPR, check that it is being used
6070         as the "then" or "else" value.
6072 2018-08-01  Tom de Vries  <tdevries@suse.de>
6074         PR target/86800
6075         * config/nvptx/nvptx.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define to
6076         speculation_safe_value_not_needed.
6078 2018-08-01  Richard Biener  <rguenther@suse.de>
6080         * tree-ssa-sccvn.c (visit_phi): Compare invariant addresses
6081         as base and offset.
6083 2018-08-01  Martin Liska  <mliska@suse.cz>
6085         * value-prof.c (gimple_divmod_fixed_value_transform): Unify
6086         format how successful transformation is dumped.
6087         (gimple_mod_pow2_value_transform): Likewise.
6088         (gimple_mod_subtract_transform): Likewise.
6089         (gimple_stringops_transform): Likewise.
6091 2018-08-01  Martin Liska  <mliska@suse.cz>
6093         PR value-prof/35543
6094         * value-prof.c (interesting_stringop_to_profile_p):
6095         Simplify the code and add BUILT_IN_MEMMOVE.
6096         (gimple_stringops_transform): Likewise.
6098 2018-08-01  Sam Tebbs  <sam.tebbs@arm.com>
6100         * config/aarch64/aarch64-simd.md
6101         (*aarch64_get_lane_zero_extendsi<mode>): Rename to...
6102         (*aarch64_get_lane_zero_extend<GPI:mode><VDQQH:mode>): ... This and
6103         use GPI iterator instead of SI mode.
6105 2018-08-01  Richard Earnshaw  <rearnsha@arm.com>
6107         * config/rs6000/rs6000.md (speculation_barrier): Renamed from
6108         rs6000_speculation_barrier.
6109         * config/rs6000/rs6000.c (rs6000_expand_builtin): Adjust for
6110         new barrier pattern name.
6112 2018-08-01  Richard Earnshaw  <rearnsha@arm.com>
6114         * config/i386/i386.md (unspecv): Add UNSPECV_SPECULATION_BARRIER.
6115         (speculation_barrier): New insn.
6117 2018-08-01  Richard Biener  <rguenther@suse.de>
6119         PR bootstrap/86724
6120         * graphite.h: Include isl/id.h and isl/space.h to allow build
6121         with ISL 0.20.
6123 2018-08-01  Jan Willem Jagersma  <jwjagersma@gmail.com>
6125         PR target/86651
6126         * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
6127         mode for COFF targets.
6128         * defaults.h (TARGET_COFF): Define.
6129         * config/i386/djgpp.h (TARGET_ASM_LTO_START, TARGET_ASM_LTO_END,
6130         TARGET_COFF): Define.
6131         (i386_djgpp_asm_lto_start, i386_djgpp_asm_lto_end): Declare.
6132         * config/i386/djgpp.c (saved_debug_info_level): New static variable.
6133         (i386_djgpp_asm_lto_start, i386_djgpp_asm_lto_end): New functions.
6135 2018-07-31  Alexandre Oliva <oliva@adacore.com>
6137         * gimple-streamer-in.c (input_bb): Restore BB discriminator.
6138         * gimple-streamer-out.c (output_bb): Save it.
6139         * lto-streamer-in.c (input_struct_function_base): Restore
6140         instance discriminator if available.  Create map on demand.
6141         * lto-streamer-out.c (output_struct_function_base): Save it if
6142         available.
6143         * final.c (decl_to_instance_map): Document LTO strategy.
6145 2018-07-31  Alexandre Oliva  <oliva@adacore.com>
6146             Olivier Hainque  <hainque@adacore.com>
6148         * debug.h (decl_to_instance_map_t): New type.
6149         (decl_to_instance_map): Declare.
6150         (maybe_create_decl_to_instance_map): New inline function.
6151         * final.c (bb_discriminator, last_bb_discriminator): New statics,
6152         to track basic block discriminators.
6153         (final_start_function_1): Initialize them.
6154         (final_scan_insn_1): On NOTE_INSN_BASIC_BLOCK, track
6155         bb_discriminator.
6156         (decl_to_instance_map): New variable.
6157         (map_decl_to_instance, maybe_set_discriminator): New functions.
6158         (notice_source_line): Set discriminator.
6160 2018-07-31  Ian Lance Taylor  <iant@golang.org>
6162         * targhooks.c (default_have_speculation_safe_value): Add
6163         ATTRIBUTE_UNUSED.
6165 2018-07-31  David Malcolm  <dmalcolm@redhat.com>
6167         * dump-context.h: Include "pretty-print.h".
6168         (dump_context::refresh_dumps_are_enabled): New decl.
6169         (dump_context::emit_item): New decl.
6170         (class dump_context): Add fields "m_test_pp" and
6171         "m_test_pp_flags".
6172         (temp_dump_context::temp_dump_context): Add param "test_pp_flags".
6173         (temp_dump_context::get_dumped_text): New decl.
6174         (class temp_dump_context): Add field "m_pp".
6175         * dumpfile.c (refresh_dumps_are_enabled): Convert to...
6176         (dump_context::refresh_dumps_are_enabled): ...and add a test for
6177         m_test_pp.
6178         (set_dump_file): Update for above change.
6179         (set_alt_dump_file): Likewise.
6180         (dump_loc): New overload, taking a pretty_printer *.
6181         (dump_context::dump_loc): Call end_any_optinfo.  Dump the location
6182         to any test pretty-printer.
6183         (make_item_for_dump_gimple_stmt): New function, adapted from
6184         optinfo::add_gimple_stmt.
6185         (dump_context::dump_gimple_stmt): 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         (make_item_for_dump_gimple_expr): New function, adapted from
6189         optinfo::add_gimple_expr.
6190         (dump_context::dump_gimple_expr): Call it, and use the result,
6191         eliminating the direct usage of dump_file and alt_dump_file in
6192         favor of indirectly using them via emit_item.
6193         (make_item_for_dump_generic_expr): New function, adapted from
6194         optinfo::add_tree.
6195         (dump_context::dump_generic_expr): Call it, and use the result,
6196         eliminating the direct usage of dump_file and alt_dump_file in
6197         favor of indirectly using them via emit_item.
6198         (make_item_for_dump_printf_va): New function, adapted from
6199         optinfo::add_printf_va.
6200         (make_item_for_dump_printf): New function.
6201         (dump_context::dump_printf_va): Call make_item_for_dump_printf_va,
6202         and use the result, eliminating the direct usage of dump_file and
6203         alt_dump_file in favor of indirectly using them via emit_item.
6204         (make_item_for_dump_dec): New function.
6205         (dump_context::dump_dec): Call it, and use the result,
6206         eliminating the direct usage of dump_file and alt_dump_file in
6207         favor of indirectly using them via emit_item.
6208         (make_item_for_dump_symtab_node): New function, adapted from
6209         optinfo::add_symtab_node.
6210         (dump_context::dump_symtab_node): Call it, and use the result,
6211         eliminating the direct usage of dump_file and alt_dump_file in
6212         favor of indirectly using them via emit_item.
6213         (dump_context::begin_scope): Reimplement, avoiding direct usage
6214         of dump_file and alt_dump_file in favor of indirectly using them
6215         via emit_item.
6216         (dump_context::emit_item): New member function.
6217         (temp_dump_context::temp_dump_context): Add param "test_pp_flags".
6218         Set up test pretty-printer on the underlying context.  Call
6219         refresh_dumps_are_enabled.
6220         (temp_dump_context::~temp_dump_context): Call
6221         refresh_dumps_are_enabled.
6222         (temp_dump_context::get_dumped_text): New member function.
6223         (selftest::verify_dumped_text): New function.
6224         (ASSERT_DUMPED_TEXT_EQ): New macro.
6225         (selftest::test_capture_of_dump_calls): Run all tests twice, with
6226         and then without optinfo enabled.  Add uses of
6227         ASSERT_DUMPED_TEXT_EQ to all tests.  Add test of nested scopes.
6228         * dumpfile.h: Update comment for the dump_* API.
6229         * optinfo-emit-json.cc
6230         (selftest::test_building_json_from_dump_calls): Update for new
6231         param for temp_dump_context ctor.
6232         * optinfo.cc (optinfo_item::optinfo_item): Remove "owned" param
6233         and "m_owned" field.
6234         (optinfo_item::~optinfo_item): Likewise.
6235         (optinfo::add_item): New member function.
6236         (optinfo::emit): Update comment.
6237         (optinfo::add_string): Delete.
6238         (optinfo::add_printf): Delete.
6239         (optinfo::add_printf_va): Delete.
6240         (optinfo::add_gimple_stmt): Delete.
6241         (optinfo::add_gimple_expr): Delete.
6242         (optinfo::add_tree): Delete.
6243         (optinfo::add_symtab_node): Delete.
6244         (optinfo::add_dec): Delete.
6245         * optinfo.h (class dump_context): New forward decl.
6246         (optinfo::add_item): New decl.
6247         (optinfo::add_string): Delete.
6248         (optinfo::add_printf): Delete.
6249         (optinfo::add_printf_va): Delete.
6250         (optinfo::add_gimple_stmt): Delete.
6251         (optinfo::add_gimple_expr): Delete.
6252         (optinfo::add_tree): Delete.
6253         (optinfo::add_symtab_node): Delete.
6254         (optinfo::add_dec): Delete.
6255         (optinfo::add_poly_int): Delete.
6256         (optinfo_item::optinfo_item): Remove "owned" param.
6257         (class optinfo_item): Remove field "m_owned".
6259 2018-07-31  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
6261         PR middle-end/86705
6262         * gcc/cfgexpand.c (set_parm_rtl): Use the alignment of Pmode when
6263         MAX_SUPPORTED_STACK_ALIGNMENT would otherwise be exceeded by the
6264         requested variable alignment.
6265         (expand_one_ssa_partition): Likewise.
6266         (expand_one_var): Likewise.
6268 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
6270         * config/pdp11/pdp11.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
6271         to speculation_safe_value_not_needed.
6273 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
6275         * targhooks.h (speculation_safe_value_not_needed): New prototype.
6276         * targhooks.c (speculation_safe_value_not_needed): New function.
6277         * target.def (have_speculation_safe_value): Update documentation.
6278         * doc/tm.texi: Regenerated.
6280 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
6282         * config/aarch64/iterators.md (ALLI_TI): New iterator.
6283         * config/aarch64/aarch64.md (despeculate_copy<ALLI_TI:mode>): New
6284         expand.
6285         (despeculate_copy<ALLI:mode>_insn): New insn.
6286         (despeculate_copyti_insn): New insn.
6287         (despeculate_simple<ALLI:mode>): New insn
6288         (despeculate_simpleti): New insn.
6289         * config/aarch64/aarch64.c (aarch64_speculation_safe_value): New
6290         function.
6291         (TARGET_SPECULATION_SAFE_VALUE): Redefine to
6292         aarch64_speculation_safe_value.
6293         (aarch64_print_operand): Handle const0_rtx in modifier 'H'.
6295 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
6297         * config/aarch64/aarch64-speculation.cc: New file.
6298         * config/aarch64/aarch64-passes.def (pass_track_speculation): Add
6299         before pass_reorder_blocks.
6300         * config/aarch64/aarch64-protos.h (make_pass_track_speculation): Add
6301         prototype.
6302         * config/aarch64/aarch64.c (aarch64_conditional_register_usage): Fix
6303         X14 and X15 when tracking speculation.
6304         * config/aarch64/aarch64.md (register name constants): Add
6305         SPECULATION_TRACKER_REGNUM and SPECULATION_SCRATCH_REGNUM.
6306         (unspec): Add UNSPEC_SPECULATION_TRACKER.
6307         (speculation_barrier): New insn attribute.
6308         (cmp<mode>): Allow SP in comparisons.
6309         (speculation_tracker): New insn.
6310         (speculation_barrier): Add speculation_barrier attribute.
6311         * config/aarch64/t-aarch64: Add make rule for aarch64-speculation.o.
6312         * config.gcc (aarch64*-*-*): Add aarch64-speculation.o to extra_objs.
6313         * doc/invoke.texi (AArch64 Options): Document -mtrack-speculation.
6315 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
6317         * config/aarch64/aarch64.md (cb<optab><mode>1): Disable when
6318         aarch64_track_speculation is true.
6319         (tb<optab><mode>1): Likewise.
6320         * config/aarch64/aarch64.c (aarch64_split_compare_regs): Do not
6321         generate CB[N]Z when tracking speculation.
6322         (aarch64_split_compare_and_swap): Likewise.
6323         (aarch64_split_atomic_op): Likewise.
6325 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
6327         * config/aarch64/aarch64.opt (mtrack-speculation): New target option.
6329 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
6331         * config/aarch64.md (unspecv): Add UNSPECV_SPECULAION_BARRIER.
6332         (speculation_barrier): New insn.
6334 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
6336         * config/arm/unspecs.md (unspecv): Add VUNSPEC_SPECULATION_BARRIER.
6337         * config/arm/arm.md (speculation_barrier): New expand.
6338         (speculation_barrier_insn): New pattern.
6340 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
6342         * builtin-types.def (BT_FN_PTR_PTR_VAR): New function type.
6343         (BT_FN_I1_I1_VAR, BT_FN_I2_I2_VAR, BT_FN_I4_I4_VAR): Likewise.
6344         (BT_FN_I8_I8_VAR, BT_FN_I16_I16_VAR): Likewise.
6345         * builtin-attrs.def (ATTR_NOVOPS_NOTHROW_LEAF_LIST): New attribute
6346         list.
6347         * builtins.def (BUILT_IN_SPECULATION_SAFE_VALUE_N): New builtin.
6348         (BUILT_IN_SPECULATION_SAFE_VALUE_PTR): New internal builtin.
6349         (BUILT_IN_SPECULATION_SAFE_VALUE_1): Likewise.
6350         (BUILT_IN_SPECULATION_SAFE_VALUE_2): Likewise.
6351         (BUILT_IN_SPECULATION_SAFE_VALUE_4): Likewise.
6352         (BUILT_IN_SPECULATION_SAFE_VALUE_8): Likewise.
6353         (BUILT_IN_SPECULATION_SAFE_VALUE_16): Likewise.
6354         * builtins.c (expand_speculation_safe_value): New function.
6355         (expand_builtin): Call it.
6356         * doc/cpp.texi: Document predefine __HAVE_SPECULATION_SAFE_VALUE.
6357         * doc/extend.texi: Document __builtin_speculation_safe_value.
6358         * doc/md.texi: Document "speculation_barrier" pattern.
6359         * doc/tm.texi.in: Pull in TARGET_SPECULATION_SAFE_VALUE and
6360         TARGET_HAVE_SPECULATION_SAFE_VALUE.
6361         * doc/tm.texi: Regenerated.
6362         * target.def (have_speculation_safe_value, speculation_safe_value): New
6363         hooks.
6364         * targhooks.c (default_have_speculation_safe_value): New function.
6365         (default_speculation_safe_value): New function.
6366         * targhooks.h (default_have_speculation_safe_value): Add prototype.
6367         (default_speculation_safe_value): Add prototype.
6369 2018-07-31  David Malcolm  <dmalcolm@redhat.com>
6371         * dump-context.h (dump_context::dump_loc): New decl.
6372         * dumpfile.c (dump_context::dump_loc): New member function.
6373         (dump_context::dump_gimple_stmt_loc): Reimplement using dump_loc
6374         and dump_gimple_stmt.
6375         (dump_context::dump_gimple_expr_loc): Likewise, using
6376         dump_gimple_expr.
6377         (dump_context::dump_generic_expr_loc): Likewise, using
6378         dump_generic_expr.
6379         (dump_context::dump_printf_loc_va): Likewise, using
6380         dump_printf_va.
6381         (dump_context::begin_scope): Explicitly using the global function
6382         "dump_loc", rather than the member function.
6384 2018-07-31  Martin Sebor  <msebor@redhat.com>
6386         PR tree-optimization/86741
6387         * tree-vrp.c (vrp_prop::check_mem_ref): Avoid incomplete types.
6389 2018-07-31  Andreas Krebbel  <krebbel@linux.ibm.com>
6391         * config/s390/s390.c (s390_expand_setmem): Make the unrolling to
6392         depend on whether prefetch instructions will be emitted or not.
6393         Use TARGET_SETMEM_PFD for checking whether prefetch instructions
6394         will be emitted or not.
6395         * config/s390/s390.h (TARGET_SETMEM_PREFETCH_DISTANCE)
6396         (TARGET_SETMEM_PFD): New macros.
6398 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
6400         * tree-vectorizer.h (stmt_vec_info): Turn back into a typedef.
6401         (NULL_STMT_VEC_INFO): Delete.
6402         (stmt_vec_info::operator*): Likewise.
6403         (stmt_vec_info::operator gimple *): Likewise.
6404         * tree-vect-loop.c (vectorizable_reduction): Use NULL instead
6405         of NULL_STMT_VEC_INFO.
6406         * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
6407         (vect_reassociating_reduction_p): Likewise.
6408         * tree-vect-stmts.c (vect_build_gather_load_calls): Likewise.
6409         (vectorizable_store): Likewise.
6410         * tree-vectorizer.c (vec_info::set_vinfo_for_stmt): Likewise.
6411         (vec_info::free_stmt_vec_infos): Likewise.
6413 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
6415         * tree-vectorizer.h (vect_stmt_in_region_p): Delete.
6416         * tree-vectorizer.c (vect_stmt_in_region_p): Likewise.
6418 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
6420         * tree-vectorizer.h (vec_info::new_vinfo_for_stmt)
6421         (vec_info::set_vinfo_for_stmt, vec_info::free_stmt_vec_infos)
6422         (vec_info::free_stmt_vec_info): New private member functions.
6423         (set_stmt_vec_info_vec, free_stmt_vec_infos, vinfo_for_stmt)
6424         (set_vinfo_for_stmt, new_stmt_vec_info, free_stmt_vec_info): Delete.
6425         * tree-parloops.c (gather_scalar_reductions): Remove calls to
6426         set_stmt_vec_info_vec and free_stmt_vec_infos.
6427         * tree-vect-loop.c (_loop_vec_info): Remove call to
6428         set_stmt_vec_info_vec.
6429         * tree-vect-stmts.c (new_stmt_vec_info, set_stmt_vec_info_vec)
6430         (free_stmt_vec_infos, free_stmt_vec_info): Delete in favor of...
6431         * tree-vectorizer.c (vec_info::new_stmt_vec_info)
6432         (vec_info::set_vinfo_for_stmt, vec_info::free_stmt_vec_infos)
6433         (vec_info::free_stmt_vec_info): ...these new functions.  Remove
6434         assignments in {vec_info::,}new_stmt_vec_info that are redundant
6435         with the clearing in the xcalloc.
6436         (stmt_vec_info_vec): Delete.
6437         (vec_info::vec_info): Don't call set_stmt_vec_info_vec.
6438         (vectorize_loops): Likewise.
6439         (vec_info::~vec_info): Remove argument from call to
6440         free_stmt_vec_infos.
6441         (vec_info::add_stmt): Remove vinfo argument from call to
6442         new_stmt_vec_info.
6444 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
6446         * tree-vectorizer.h (free_stmt_vec_info): Take a stmt_vec_info
6447         rather than a gimple stmt.
6448         * tree-vect-stmts.c (free_stmt_vec_info): Likewise.  Don't free
6449         information for pattern statements when passed the original
6450         statement; instead wait to be passed the pattern statement itself.
6451         Don't call set_vinfo_for_stmt here.
6452         (free_stmt_vec_infos): Update call to free_stmt_vec_info.
6453         * tree-vect-loop.c (_loop_vec_info::~loop_vec_info): Don't free
6454         stmt_vec_infos here.
6455         * tree-vect-slp.c (_bb_vec_info::~bb_vec_info): Likewise.
6456         * tree-vectorizer.c (vec_info::remove_stmt): Nullify the statement's
6457         stmt_vec_infos entry.
6459 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
6461         * tree-vectorizer.h (vec_info::replace_stmt): Declare.
6462         * tree-vectorizer.c (vec_info::replace_stmt): New function.
6463         * tree-vect-slp.c (vect_remove_slp_scalar_calls): Use it.
6464         * tree-vect-stmts.c (vectorizable_call): Likewise.
6465         (vectorizable_simd_clone_call): Likewise.
6467 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
6469         * tree-vectorizer.h (vec_info::remove_stmt): Declare.
6470         * tree-vectorizer.c (vec_info::remove_stmt): New function.
6471         * tree-vect-loop-manip.c (vect_set_loop_condition): Use it.
6472         * tree-vect-loop.c (vect_transform_loop): Likewise.
6473         * tree-vect-slp.c (vect_schedule_slp): Likewise.
6474         * tree-vect-stmts.c (vect_remove_stores): Likewise.
6476 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
6478         * tree-vectorizer.h (vec_info::lookup_dr): New member function.
6479         (vect_dr_stmt): Delete.
6480         * tree-vectorizer.c (vec_info::lookup_dr): New function.
6481         * tree-vect-loop-manip.c (vect_update_inits_of_drs): Use it instead
6482         of DR_VECT_AUX.
6483         * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
6484         (vect_analyze_data_ref_dependence, vect_record_base_alignments)
6485         (vect_verify_datarefs_alignment, vect_peeling_supportable)
6486         (vect_analyze_data_ref_accesses, vect_prune_runtime_alias_test_list)
6487         (vect_analyze_data_refs): Likewise.
6488         (vect_slp_analyze_data_ref_dependence): Likewise.  Take a vec_info
6489         argument.
6490         (vect_find_same_alignment_drs): Likewise.
6491         (vect_slp_analyze_node_dependences): Update calls accordingly.
6492         (vect_analyze_data_refs_alignment): Likewise.  Use vec_info::lookup_dr
6493         instead of DR_VECT_AUX.
6494         (vect_get_peeling_costs_all_drs): Take a loop_vec_info instead
6495         of a vector data references.  Use vec_info::lookup_dr instead of
6496         DR_VECT_AUX.
6497         (vect_peeling_hash_get_lowest_cost): Update calls accordingly.
6498         (vect_enhance_data_refs_alignment): Likewise.  Use vec_info::lookup_dr
6499         instead of DR_VECT_AUX.
6501 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
6503         * tree-vectorizer.h (_loop_vec_info::unaligned_dr): Change to
6504         dr_vec_info.
6505         * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Update
6506         accordingly.
6507         * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
6508         * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
6509         (vect_gen_prolog_loop_niters): Likewise.
6511 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
6513         * tree-vectorizer.h (set_dr_misalignment, dr_misalignment)
6514         (DR_TARGET_ALIGNMENT, aligned_access_p, known_alignment_for_access_p)
6515         (vect_known_alignment_in_bytes, vect_dr_behavior)
6516         (vect_get_scalar_dr_size): Take references as dr_vec_infos
6517         instead of data_references.  Update calls to other routines for
6518         which the same change has been made.
6519         * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Take
6520         dr_vec_infos instead of stmt_vec_infos.
6521         (vect_analyze_data_ref_dependence): Update call accordingly.
6522         (vect_slp_analyze_data_ref_dependence)
6523         (vect_record_base_alignments): Use DR_VECT_AUX.
6524         (vect_calculate_target_alignment, vect_compute_data_ref_alignment)
6525         (vect_update_misalignment_for_peel, verify_data_ref_alignment)
6526         (vector_alignment_reachable_p, vect_get_data_access_cost)
6527         (vect_peeling_supportable, vect_analyze_group_access_1)
6528         (vect_analyze_group_access, vect_analyze_data_ref_access)
6529         (vect_vfa_segment_size, vect_vfa_access_size, vect_vfa_align)
6530         (vect_compile_time_alias, vect_small_gap_p)
6531         (vectorizable_with_step_bound_p, vect_duplicate_ssa_name_ptr_info):
6532         (vect_supportable_dr_alignment): Take references as dr_vec_infos
6533         instead of data_references.  Update calls to other routines for
6534         which the same change has been made.
6535         (vect_verify_datarefs_alignment, vect_get_peeling_costs_all_drs)
6536         (vect_find_same_alignment_drs, vect_analyze_data_refs_alignment)
6537         (vect_slp_analyze_and_verify_node_alignment)
6538         (vect_analyze_data_ref_accesses, vect_prune_runtime_alias_test_list)
6539         (vect_create_addr_base_for_vector_ref, vect_create_data_ref_ptr)
6540         (vect_setup_realignment): Use dr_vec_infos.  Update calls after
6541         above changes.
6542         (_vect_peel_info::dr): Replace with...
6543         (_vect_peel_info::dr_info): ...this new field.
6544         (vect_peeling_hash_get_most_frequent)
6545         (vect_peeling_hash_choose_best_peeling): Update accordingly.
6546         (vect_peeling_hash_get_lowest_cost):
6547         (vect_enhance_data_refs_alignment): Likewise.  Update calls to other
6548         routines for which the same change has been made.
6549         (vect_peeling_hash_insert): Likewise.  Take a dr_vec_info instead of a
6550         data_reference.
6551         * tree-vect-loop-manip.c (get_misalign_in_elems)
6552         (vect_gen_prolog_loop_niters): Use dr_vec_infos.  Update calls after
6553         above changes.
6554         * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
6555         * tree-vect-stmts.c (vect_get_store_cost, vect_get_load_cost)
6556         (vect_truncate_gather_scatter_offset, compare_step_with_zero)
6557         (get_group_load_store_type, get_negative_load_store_type)
6558         (vect_get_data_ptr_increment, vectorizable_store)
6559         (vectorizable_load): Likewise.
6560         (ensure_base_align): Take a dr_vec_info instead of a data_reference.
6561         Update calls to other routines for which the same change has been made.
6563 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
6565         * tree-vectorizer.h (vec_info::move_dr): New member function.
6566         (dataref_aux): Rename to...
6567         (dr_vec_info): ...this and add "dr" and "stmt" fields.
6568         (_stmt_vec_info::dr_aux): Update accordingly.
6569         (_stmt_vec_info::data_ref_info): Delete.
6570         (STMT_VINFO_GROUPED_ACCESS, DR_GROUP_FIRST_ELEMENT)
6571         (DR_GROUP_NEXT_ELEMENT, DR_GROUP_SIZE, DR_GROUP_STORE_COUNT)
6572         (DR_GROUP_GAP, DR_GROUP_SAME_DR_STMT, REDUC_GROUP_FIRST_ELEMENT):
6573         (REDUC_GROUP_NEXT_ELEMENT, REDUC_GROUP_SIZE): Use dr_aux.dr instead
6574         of data_ref.
6575         (STMT_VINFO_DATA_REF): Likewise.  Turn into an lvalue.
6576         (STMT_VINFO_DR_INFO): New macro.
6577         (DR_VECT_AUX): Use STMT_VINFO_DR_INKFO and vect_dr_stmt.
6578         (set_dr_misalignment): Update after rename of dataref_aux.
6579         (vect_dr_stmt): Move earlier in file.  Return dr_aux.stmt.
6580         * tree-vect-stmts.c (new_stmt_vec_info): Remove redundant
6581         initialization of STMT_VINFO_DATA_REF.
6582         * tree-vectorizer.c (vec_info::move_dr): New function.
6583         * tree-vect-patterns.c (vect_recog_bool_pattern)
6584         (vect_recog_mask_conversion_pattern)
6585         (vect_recog_gather_scatter_pattern): Use it.
6586         * tree-vect-data-refs.c (vect_analyze_data_refs): Initialize
6587         the "dr" and "stmt" fields of dr_vec_info instead of
6588         STMT_VINFO_DATA_REF.
6590 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
6592         * tree-vectorizer.h (_stmt_vec_info::pattern_stmt_p): New field.
6593         (is_pattern_stmt_p): Use it.
6594         * tree-vect-patterns.c (vect_init_pattern_stmt): Set pattern_stmt_p
6595         on pattern statements.
6597 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
6599         * tree-vect-patterns.c (vect_mark_pattern_stmts): Take the
6600         original stmt as a stmt_vec_info rather than a gimple stmt.
6601         (vect_pattern_recog_1): Take the statement directly as a
6602         stmt_vec_info, rather than via a gimple_stmt_iterator.
6603         Update call to vect_mark_pattern_stmts.
6604         (vect_pattern_recog): Update calls accordingly.
6606 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
6608         * tree-vectorizer.h (vect_get_vec_defs_for_stmt_copy)
6609         (vect_get_vec_def_for_stmt_copy): Take a vec_info rather than
6610         a vect_def_type for the first argument.
6611         * tree-vect-stmts.c (vect_get_vec_defs_for_stmt_copy): Likewise.
6612         (vect_get_vec_def_for_stmt_copy): Likewise.  Return the original
6613         operand if it isn't defined by a vectorized statement.
6614         (vect_build_gather_load_calls): Remove the mask_dt argument and
6615         update calls to vect_get_vec_def_for_stmt_copy.
6616         (vectorizable_bswap): Likewise the dt argument.
6617         (vectorizable_call): Update calls to vectorizable_bswap and
6618         vect_get_vec_def_for_stmt_copy.
6619         (vectorizable_simd_clone_call, vectorizable_assignment)
6620         (vectorizable_shift, vectorizable_operation, vectorizable_condition)
6621         (vectorizable_comparison): Update calls to
6622         vect_get_vec_def_for_stmt_copy.
6623         (vectorizable_store): Likewise.  Remove now-unnecessary calls to
6624         vect_is_simple_use.
6625         (vect_get_loop_based_defs): Remove dt argument and update call
6626         to vect_get_vec_def_for_stmt_copy.
6627         (vectorizable_conversion): Update calls to vect_get_loop_based_defs
6628         and vect_get_vec_def_for_stmt_copy.
6629         (vectorizable_load): Update calls to vect_build_gather_load_calls
6630         and vect_get_vec_def_for_stmt_copy.
6631         * tree-vect-loop.c (vect_create_epilog_for_reduction)
6632         (vectorizable_reduction, vectorizable_live_operation): Update calls
6633         to vect_get_vec_def_for_stmt_copy.
6635 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
6637         * tree-vect-data-refs.c (vect_record_base_alignment): Replace vec_info
6638         and gimple stmt arguments with a stmt_vec_info.
6639         (vect_record_base_alignments): Update calls accordingly.
6640         * tree-vect-slp.c (vect_record_max_nunits): Replace vec_info
6641         and gimple stmt arguments with a stmt_vec_info.
6642         (vect_build_slp_tree_1): Remove vinfo argument and update call
6643         to vect_record_max_nunits.
6644         (vect_build_slp_tree_2): Update calls to vect_build_slp_tree_1
6645         and vect_record_max_nunits.
6647 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
6649         * tree-vectorizer.h (nested_in_vect_loop_p): Move further down
6650         file and take a stmt_vec_info instead of a gimple stmt.
6651         (supportable_widening_operation, vect_finish_replace_stmt)
6652         (vect_finish_stmt_generation, vect_get_store_rhs)
6653         (vect_get_vec_def_for_operand_1, vect_get_vec_def_for_operand)
6654         (vect_get_vec_defs, vect_init_vector, vect_transform_stmt)
6655         (vect_remove_stores, vect_analyze_stmt, vectorizable_condition)
6656         (vect_get_smallest_scalar_type, vect_check_gather_scatter)
6657         (vect_create_data_ref_ptr, bump_vector_ptr)
6658         (vect_permute_store_chain, vect_setup_realignment)
6659         (vect_transform_grouped_load, vect_record_grouped_load_vectors)
6660         (vect_create_addr_base_for_vector_ref, vectorizable_live_operation)
6661         (vectorizable_reduction, vectorizable_induction)
6662         (get_initial_def_for_reduction, is_simple_and_all_uses_invariant)
6663         (vect_get_place_in_interleaving_chain): Take stmt_vec_infos rather
6664         than gimple stmts as arguments.
6665         * tree-vect-data-refs.c (vect_get_smallest_scalar_type)
6666         (vect_preserves_scalar_order_p, vect_slp_analyze_node_dependences)
6667         (can_group_stmts_p, vect_check_gather_scatter)
6668         (vect_create_addr_base_for_vector_ref, vect_create_data_ref_ptr)
6669         (bump_vector_ptr, vect_permute_store_chain, vect_setup_realignment)
6670         (vect_permute_load_chain, vect_shift_permute_load_chain)
6671         (vect_transform_grouped_load)
6672         (vect_record_grouped_load_vectors): Likewise.
6673         * tree-vect-loop.c (vect_fixup_reduc_chain)
6674         (get_initial_def_for_reduction, vect_create_epilog_for_reduction)
6675         (vectorize_fold_left_reduction, is_nonwrapping_integer_induction)
6676         (vectorizable_reduction, vectorizable_induction)
6677         (vectorizable_live_operation, vect_loop_kill_debug_uses): Likewise.
6678         * tree-vect-patterns.c (type_conversion_p, adjust_bool_stmts)
6679         (vect_get_load_store_mask): Likewise.
6680         * tree-vect-slp.c (vect_get_place_in_interleaving_chain)
6681         (vect_analyze_slp_instance, vect_mask_constant_operand_p): Likewise.
6682         * tree-vect-stmts.c (vect_mark_relevant)
6683         (is_simple_and_all_uses_invariant)
6684         (exist_non_indexing_operands_for_use_p, process_use)
6685         (vect_init_vector_1, vect_init_vector, vect_get_vec_def_for_operand_1)
6686         (vect_get_vec_def_for_operand, vect_get_vec_defs)
6687         (vect_finish_stmt_generation_1, vect_finish_replace_stmt)
6688         (vect_finish_stmt_generation, vect_truncate_gather_scatter_offset)
6689         (compare_step_with_zero, vect_get_store_rhs, get_group_load_store_type)
6690         (get_negative_load_store_type, get_load_store_type)
6691         (vect_check_load_store_mask, vect_check_store_rhs)
6692         (vect_build_gather_load_calls, vect_get_strided_load_store_ops)
6693         (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
6694         (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
6695         (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
6696         (get_group_alias_ptr_type, vectorizable_store, hoist_defs_of_uses)
6697         (vectorizable_load, vectorizable_condition, vectorizable_comparison)
6698         (vect_analyze_stmt, vect_transform_stmt, vect_remove_stores)
6699         (supportable_widening_operation): Likewise.
6701 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
6703         * tree-vect-data-refs.c (vect_describe_gather_scatter_call): Take
6704         a stmt_vec_info instead of a gcall.
6705         (vect_check_gather_scatter): Update call accordingly.
6706         * tree-vect-loop-manip.c (iv_phi_p): Take a stmt_vec_info instead
6707         of a gphi.
6708         (vect_can_advance_ivs_p, vect_update_ivs_after_vectorizer)
6709         (slpeel_update_phi_nodes_for_loops):): Update calls accordingly.
6710         * tree-vect-loop.c (vect_transform_loop_stmt): Take a stmt_vec_info
6711         instead of a gimple stmt.
6712         (vect_transform_loop): Update calls accordingly.
6713         * tree-vect-slp.c (vect_split_slp_store_group): Take and return
6714         stmt_vec_infos instead of gimple stmts.
6715         (vect_analyze_slp_instance): Update use accordingly.
6716         * tree-vect-stmts.c (read_vector_array, write_vector_array)
6717         (vect_clobber_variable, vect_stmt_relevant_p, permute_vec_elements)
6718         (vect_use_strided_gather_scatters_p, vect_build_all_ones_mask)
6719         (vect_build_zero_merge_argument, vect_get_gather_scatter_ops)
6720         (vect_gen_widened_results_half, vect_get_loop_based_defs)
6721         (vect_create_vectorized_promotion_stmts, can_vectorize_live_stmts):
6722         Take a stmt_vec_info instead of a gimple stmt and pass stmt_vec_infos
6723         down to subroutines.
6725 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
6727         * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Change the type
6728         of the worklist from a vector of gimple stmts to a vector of
6729         stmt_vec_infos.
6730         * tree-vect-stmts.c (vect_mark_relevant, process_use)
6731         (vect_mark_stmts_to_be_vectorized): Likewise
6733 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
6735         * tree-vect-loop.c (vect_analyze_loop_operations): Look up the
6736         statement before passing it to vect_analyze_stmt.
6737         (vect_create_epilog_for_reduction): Use a stmt_vec_info to walk
6738         the chain of phi vector definitions.  Track the exit phi via its
6739         stmt_vec_info.
6740         (vectorizable_reduction): Set cond_stmt_vinfo directly from the
6741         STMT_VINFO_REDUC_DEF.
6742         * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Use
6743         stmt_vec_infos to handle the statement chains.
6744         (vect_get_slp_defs): Record the first statement in the node
6745         using a stmt_vec_info.
6746         * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Look up
6747         statements here and pass their stmt_vec_info down to subroutines.
6748         (vect_init_vector_1): Hoist call to vinfo_for_stmt and pass it
6749         down to vect_finish_stmt_generation.
6750         (vect_init_vector, vect_get_vec_defs, vect_finish_replace_stmt)
6751         (vect_finish_stmt_generation): Call vinfo_for_stmt and pass
6752         stmt_vec_infos to subroutines.
6753         (vect_remove_stores): Use stmt_vec_infos to handle the statement
6754         chains.
6756 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
6758         * tree-vect-data-refs.c (vect_slp_analyze_node_dependences):
6759         (vect_check_gather_scatter, vect_create_data_ref_ptr, bump_vector_ptr)
6760         (vect_permute_store_chain, vect_setup_realignment)
6761         (vect_permute_load_chain, vect_shift_permute_load_chain)
6762         (vect_transform_grouped_load): Use stmt_vec_info rather than gimple
6763         stmts internally, and when passing values to other vectorizer routines.
6764         * tree-vect-loop-manip.c (vect_can_advance_ivs_p): Likewise.
6765         * tree-vect-loop.c (vect_analyze_scalar_cycles_1)
6766         (vect_analyze_loop_operations, get_initial_def_for_reduction)
6767         (vect_create_epilog_for_reduction, vectorize_fold_left_reduction)
6768         (vectorizable_reduction, vectorizable_induction)
6769         (vectorizable_live_operation, vect_transform_loop_stmt)
6770         (vect_transform_loop): Likewise.
6771         * tree-vect-patterns.c (vect_reassociating_reduction_p)
6772         (vect_recog_widen_op_pattern, vect_recog_mixed_size_cond_pattern)
6773         (vect_recog_bool_pattern, vect_recog_gather_scatter_pattern): Likewise.
6774         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
6775         (vect_slp_analyze_node_operations_1): Likewise.
6776         * tree-vect-stmts.c (vect_mark_relevant, process_use)
6777         (exist_non_indexing_operands_for_use_p, vect_init_vector_1)
6778         (vect_mark_stmts_to_be_vectorized, vect_get_vec_def_for_operand)
6779         (vect_finish_stmt_generation_1, get_group_load_store_type)
6780         (get_load_store_type, vect_build_gather_load_calls)
6781         (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
6782         (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
6783         (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
6784         (vectorizable_store, vectorizable_load, vectorizable_condition)
6785         (vectorizable_comparison, vect_analyze_stmt, vect_transform_stmt)
6786         (supportable_widening_operation): Likewise.
6787         (vect_get_vector_types_for_stmt): Likewise.
6788         * tree-vectorizer.h (vect_dr_behavior): Likewise.
6790 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
6792         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
6793         (vect_slp_analyze_node_dependences, vect_analyze_data_ref_accesses)
6794         (vect_permute_store_chain, vect_permute_load_chain)
6795         (vect_shift_permute_load_chain, vect_transform_grouped_load): Avoid
6796         repeated stmt_vec_info lookups.
6797         * tree-vect-loop-manip.c (vect_can_advance_ivs_p): Likewise.
6798         (vect_update_ivs_after_vectorizer): Likewise.
6799         * tree-vect-loop.c (vect_is_simple_reduction): Likewise.
6800         (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
6801         * tree-vect-patterns.c (adjust_bool_stmts): Likewise.
6802         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
6803         (vect_bb_slp_scalar_cost): Likewise.
6804         * tree-vect-stmts.c (get_group_alias_ptr_type): Likewise.
6806 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
6808         * tree-vect-data-refs.c (vect_check_gather_scatter): Pass the
6809         gcall rather than the generic gimple stmt to gimple_call_internal_fn.
6810         (vect_get_smallest_scalar_type, can_group_stmts_p): Use dyn_cast
6811         to get gassigns and gcalls, rather than operating on generc gimple
6812         stmts.
6813         * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p)
6814         (vect_mark_stmts_to_be_vectorized, vectorizable_store)
6815         (vectorizable_load, vect_analyze_stmt): Likewise.
6816         * tree-vect-loop.c (vectorizable_reduction): Likewise gphi.
6818 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
6820         * tree-vectorizer.h (get_earlier_stmt, get_later_stmt): Take and
6821         return stmt_vec_infos rather than gimple stmts.  Do not accept
6822         null arguments.
6823         (vect_find_last_scalar_stmt_in_slp): Return a stmt_vec_info instead
6824         of a gimple stmt.
6825         * tree-vect-slp.c (vect_find_last_scalar_stmt_in_slp): Likewise.
6826         Update use of get_later_stmt.
6827         (vect_get_constant_vectors): Update call accordingly.
6828         (vect_schedule_slp_instance): Likewise
6829         * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Likewise.
6830         (vect_slp_analyze_instance_dependence): Likewise.
6831         (vect_preserves_scalar_order_p): Update use of get_earlier_stmt.
6833 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
6835         * tree-vectorizer.h (stmt_info_for_cost::stmt): Replace with...
6836         (stmt_info_for_cost::stmt_info): ...this new field.
6837         (add_stmt_costs): Update accordingly.
6838         * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost)
6839         (vect_get_known_peeling_cost): Likewise.
6840         (vect_estimate_min_profitable_iters): Likewise.
6841         * tree-vect-stmts.c (record_stmt_cost): Likewise.
6843 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
6845         * tree-vectorizer.h (_loop_vec_info::may_misalign_stmts): Change
6846         from an auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
6847         * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Update
6848         accordingly.
6849         * tree-vect-loop-manip.c (vect_create_cond_for_align_checks): Likewise.
6851 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
6853         * tree-vectorizer.h (_stmt_vec_info::same_dr_stmt): Change from
6854         a gimple stmt to a stmt_vec_info.
6855         * tree-vect-stmts.c (vectorizable_load): Update accordingly.
6857 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
6859         * tree-vectorizer.h (vec_info::grouped_stores): Change from
6860         an auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
6861         (_loop_vec_info::reduction_chains): Likewise.
6862         * tree-vect-loop.c (vect_fixup_scalar_cycles_with_patterns): Update
6863         accordingly.
6864         * tree-vect-slp.c (vect_analyze_slp): Likewise.
6866 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
6868         * tree-vectorizer.h (_stmt_vec_info::first_element): Change from
6869         a gimple stmt to a stmt_vec_info.
6870         (_stmt_vec_info::next_element): Likewise.
6871         * tree-vect-data-refs.c (vect_update_misalignment_for_peel)
6872         (vect_slp_analyze_and_verify_node_alignment)
6873         (vect_analyze_group_access_1, vect_analyze_group_access)
6874         (vect_small_gap_p, vect_prune_runtime_alias_test_list)
6875         (vect_create_data_ref_ptr, vect_record_grouped_load_vectors)
6876         (vect_supportable_dr_alignment): Update accordingly.
6877         * tree-vect-loop.c (vect_fixup_reduc_chain): Likewise.
6878         (vect_fixup_scalar_cycles_with_patterns, vect_is_slp_reduction)
6879         (vect_is_simple_reduction, vectorizable_reduction): Likewise.
6880         * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
6881         * tree-vect-slp.c (vect_build_slp_tree_1)
6882         (vect_attempt_slp_rearrange_stmts, vect_supported_load_permutation_p)
6883         (vect_split_slp_store_group, vect_analyze_slp_instance)
6884         (vect_analyze_slp, vect_transform_slp_perm_load): Likewise.
6885         * tree-vect-stmts.c (vect_model_store_cost, vect_model_load_cost)
6886         (get_group_load_store_type, get_load_store_type)
6887         (get_group_alias_ptr_type, vectorizable_store, vectorizable_load)
6888         (vect_transform_stmt, vect_remove_stores): Likewise.
6890 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
6892         * tree-vectorizer.h (vect_dr_stmt): Return a stmt_vec_info rather
6893         than a gimple stmt.
6894         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
6895         (vect_slp_analyze_data_ref_dependence, vect_record_base_alignments)
6896         (vect_calculate_target_alignmentm, vect_compute_data_ref_alignment)
6897         (vect_update_misalignment_for_peel, vect_verify_datarefs_alignment)
6898         (vector_alignment_reachable_p, vect_get_data_access_cost)
6899         (vect_get_peeling_costs_all_drs, vect_peeling_hash_get_lowest_cost)
6900         (vect_peeling_supportable, vect_enhance_data_refs_alignment)
6901         (vect_find_same_alignment_drs, vect_analyze_data_refs_alignment)
6902         (vect_analyze_group_access_1, vect_analyze_group_access)
6903         (vect_analyze_data_ref_access, vect_analyze_data_ref_accesses)
6904         (vect_vfa_access_size, vect_small_gap_p, vect_analyze_data_refs)
6905         (vect_supportable_dr_alignment): Remove vinfo_for_stmt from the
6906         result of vect_dr_stmt and use the stmt_vec_info instead of
6907         the associated gimple stmt.
6908         * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
6909         (vect_gen_prolog_loop_niters): Likewise.
6910         * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
6912 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
6914         * tree-vectorizer.h (_slp_tree::stmts): Change from a vec<gimple *>
6915         to a vec<stmt_vec_info>.
6916         * tree-vect-slp.c (vect_free_slp_tree): Update accordingly.
6917         (vect_create_new_slp_node): Take a vec<gimple *> instead of a
6918         vec<stmt_vec_info>.
6919         (_slp_oprnd_info::def_stmts): Change from a vec<gimple *>
6920         to a vec<stmt_vec_info>.
6921         (bst_traits::value_type, bst_traits::value_type): Likewise.
6922         (bst_traits::hash): Update accordingly.
6923         (vect_get_and_check_slp_defs): Change the stmts parameter from
6924         a vec<gimple *> to a vec<stmt_vec_info>.
6925         (vect_two_operations_perm_ok_p, vect_build_slp_tree_1): Likewise.
6926         (vect_build_slp_tree): Likewise.
6927         (vect_build_slp_tree_2): Likewise.  Update uses of
6928         SLP_TREE_SCALAR_STMTS.
6929         (vect_print_slp_tree): Update uses of SLP_TREE_SCALAR_STMTS.
6930         (vect_mark_slp_stmts, vect_mark_slp_stmts_relevant)
6931         (vect_slp_rearrange_stmts, vect_attempt_slp_rearrange_stmts)
6932         (vect_supported_load_permutation_p, vect_find_last_scalar_stmt_in_slp)
6933         (vect_detect_hybrid_slp_stmts, vect_slp_analyze_node_operations_1)
6934         (vect_slp_analyze_node_operations, vect_slp_analyze_operations)
6935         (vect_bb_slp_scalar_cost, vect_slp_analyze_bb_1)
6936         (vect_get_constant_vectors, vect_get_slp_defs)
6937         (vect_transform_slp_perm_load, vect_schedule_slp_instance)
6938         (vect_remove_slp_scalar_calls, vect_schedule_slp): Likewise.
6939         (vect_analyze_slp_instance): Build up a vec of stmt_vec_infos
6940         instead of gimple stmts.
6941         * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Change
6942         the stores parameter for a vec<gimple *> to a vec<stmt_vec_info>.
6943         (vect_slp_analyze_instance_dependence): Update uses of
6944         SLP_TREE_SCALAR_STMTS.
6945         (vect_slp_analyze_and_verify_node_alignment): Likewise.
6946         (vect_slp_analyze_and_verify_instance_alignment): Likewise.
6947         * tree-vect-loop.c (neutral_op_for_slp_reduction): Likewise.
6948         (get_initial_defs_for_reduction): Likewise.
6949         (vect_create_epilog_for_reduction): Likewise.
6950         (vectorize_fold_left_reduction): Likewise.
6951         * tree-vect-stmts.c (vect_prologue_cost_for_slp_op): Likewise.
6952         (vect_model_simple_cost, vectorizable_shift, vectorizable_load)
6953         (can_vectorize_live_stmts): Likewise.
6955 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
6957         * tree-vectorizer.h (_loop_vec_info::reductions): Change from an
6958         auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
6959         (vect_force_simple_reduction): Take and return stmt_vec_infos rather
6960         than gimple stmts.
6961         * tree-parloops.c (valid_reduction_p): Take a stmt_vec_info instead
6962         of a gimple stmt.
6963         (gather_scalar_reductions): Update after above interface changes.
6964         * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Likewise.
6965         (vect_is_simple_reduction): Take and return stmt_vec_infos rather
6966         than gimple stmts.
6967         (vect_force_simple_reduction): Likewise.
6968         * tree-vect-patterns.c (vect_pattern_recog_1): Update use of
6969         LOOP_VINFO_REDUCTIONS.
6970         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
6972 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
6974         * tree-vectorizer.h (_stmt_vec_info::reduc_def): Change from
6975         a gimple stmt to a stmt_vec_info.
6976         * tree-vect-loop.c (vect_active_double_reduction_p)
6977         (vect_force_simple_reduction, vectorizable_reduction): Update
6978         accordingly.
6980 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
6982         * tree-vectorizer.h (_slp_tree::vec_stmts): Change from a
6983         vec<gimple *> to a vec<stmt_vec_info>.
6984         * tree-vect-loop.c (vect_create_epilog_for_reduction): Change
6985         the reduction_phis argument from a vec<gimple *> to a
6986         vec<stmt_vec_info>.
6987         (vectorizable_reduction): Likewise the phis local variable that
6988         is passed to vect_create_epilog_for_reduction.  Update for new type
6989         of SLP_TREE_VEC_STMTS.
6990         (vectorizable_induction): Update for new type of SLP_TREE_VEC_STMTS.
6991         (vectorizable_live_operation): Likewise.
6992         * tree-vect-slp.c (vect_get_slp_vect_defs): Likewise.
6993         (vect_transform_slp_perm_load, vect_schedule_slp_instance): Likewise.
6995 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
6997         * tree-vectorizer.h (_stmt_vec_info::vectorized_stmt): Change from
6998         a gimple stmt to a stmt_vec_info.
6999         (vectorizable_condition, vectorizable_live_operation)
7000         (vectorizable_reduction, vectorizable_induction): Pass back the
7001         vectorized statement as a stmt_vec_info.
7002         * tree-vect-data-refs.c (vect_record_grouped_load_vectors): Update
7003         use of STMT_VINFO_VEC_STMT.
7004         * tree-vect-loop.c (vect_create_epilog_for_reduction): Likewise,
7005         accumulating the inner phis that feed the STMT_VINFO_VEC_STMT
7006         as stmt_vec_infos rather than gimple stmts.
7007         (vectorize_fold_left_reduction): Change vec_stmt from a gimple stmt
7008         to a stmt_vec_info.
7009         (vectorizable_live_operation): Likewise.
7010         (vectorizable_reduction, vectorizable_induction): Likewise,
7011         updating use of STMT_VINFO_VEC_STMT.
7012         * tree-vect-stmts.c (vect_get_vec_def_for_operand_1): Update use
7013         of STMT_VINFO_VEC_STMT.
7014         (vect_build_gather_load_calls, vectorizable_bswap, vectorizable_call)
7015         (vectorizable_simd_clone_call, vectorizable_conversion)
7016         (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
7017         (vectorizable_store, vectorizable_load, vectorizable_condition)
7018         (vectorizable_comparison, can_vectorize_live_stmts): Change vec_stmt
7019         from a gimple stmt to a stmt_vec_info.
7020         (vect_transform_stmt): Update use of STMT_VINFO_VEC_STMT.  Pass a
7021         pointer to a stmt_vec_info to the vectorizable_* routines.
7023 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
7025         * tree-vectorizer.h (_stmt_vec_info::related_stmt): Change from
7026         a gimple stmt to a stmt_vec_info.
7027         (is_pattern_stmt_p): Update accordingly.
7028         * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Likewise.
7029         (vect_record_grouped_load_vectors): Likewise.
7030         * tree-vect-loop.c (vect_determine_vf_for_stmt): Likewise.
7031         (vect_fixup_reduc_chain, vect_update_vf_for_slp): Likewise.
7032         (vect_model_reduction_cost): Likewise.
7033         (vect_create_epilog_for_reduction): Likewise.
7034         (vectorizable_reduction, vectorizable_induction): Likewise.
7035         * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
7036         Return the stmt_vec_info for the pattern statement.
7037         (vect_set_pattern_stmt): Update use of STMT_VINFO_RELATED_STMT.
7038         (vect_split_statement, vect_mark_pattern_stmts): Likewise.
7039         * tree-vect-slp.c (vect_detect_hybrid_slp_stmts): Likewise.
7040         (vect_detect_hybrid_slp, vect_get_slp_defs): Likewise.
7041         * tree-vect-stmts.c (vect_mark_relevant): Likewise.
7042         (vect_get_vec_def_for_operand_1, vectorizable_call): Likewise.
7043         (vectorizable_simd_clone_call, vect_analyze_stmt, new_stmt_vec_info)
7044         (free_stmt_vec_info, vect_is_simple_use): Likewise.
7046 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
7048         * tree-vectorizer.h (vect_finish_replace_stmt): Return a stmt_vec_info
7049         (vect_finish_stmt_generation): Likewise.
7050         * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
7051         (vect_finish_replace_stmt, vect_finish_stmt_generation): Likewise.
7052         (vect_build_gather_load_calls): Use the return value of the above
7053         functions instead of a separate call to vinfo_for_stmt.  Use narrow
7054         scopes for the input gimple stmt and wider scopes for the associated
7055         stmt_vec_info.  Use vec_info::lookup_def when setting these
7056         stmt_vec_infos from an SSA_NAME definition.
7057         (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
7058         (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
7059         (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
7060         (vectorizable_store, vectorizable_load, vectorizable_condition)
7061         (vectorizable_comparison): Likewise.
7062         * tree-vect-loop.c (vectorize_fold_left_reduction): Likewise.
7063         (vectorizable_reduction): Likewise.
7065 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
7067         * tree-vectorizer.h (vect_is_simple_use): Add an optional
7068         stmt_vec_info * parameter before the optional gimple **.
7069         * tree-vect-stmts.c (vect_is_simple_use): Likewise.
7070         (process_use, vect_get_vec_def_for_operand_1): Update callers.
7071         (vect_get_vec_def_for_operand, vectorizable_shift): Likewise.
7072         * tree-vect-loop.c (vectorizable_reduction): Likewise.
7073         (vectorizable_live_operation): Likewise.
7074         * tree-vect-patterns.c (type_conversion_p): Likewise.
7075         (vect_look_through_possible_promotion): Likewise.
7076         (vect_recog_rotate_pattern): Likewise.
7077         * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
7079 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
7081         * tree-vectorizer.h (stmt_vec_info): Temporarily change from
7082         a typedef to a wrapper class.
7083         (NULL_STMT_VEC_INFO): New macro.
7084         (vec_info::stmt_infos): Change to vec<stmt_vec_info>.
7085         (stmt_vec_info::operator*): New function.
7086         (stmt_vec_info::operator gimple *): Likewise.
7087         (set_vinfo_for_stmt): Use NULL_STMT_VEC_INFO.
7088         (add_stmt_costs): Likewise.
7089         * tree-vect-loop-manip.c (iv_phi_p): Likewise.
7090         * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost)
7091         (vect_get_known_peeling_cost): Likewise.
7092         (vect_estimate_min_profitable_iters): Likewise.
7093         * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
7094         * tree-vect-slp.c (vect_remove_slp_scalar_calls): Likewise.
7095         * tree-vect-stmts.c (vect_build_gather_load_calls): Likewise.
7096         (vectorizable_store, free_stmt_vec_infos): Likewise.
7097         (new_stmt_vec_info): Change return type of xcalloc to
7098         _stmt_vec_info *.
7100 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
7102         * tree-vectorizer.h (vec_info::lookup_single_use): Declare.
7103         * tree-vectorizer.c (vec_info::lookup_single_use): New function.
7104         * tree-vect-loop.c (vectorizable_reduction): Use it instead of
7105         a single_imm_use-based sequence.
7106         * tree-vect-stmts.c (supportable_widening_operation): Likewise.
7108 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
7110         * tree-vectorizer.h (vec_info::lookup_def): Declare.
7111         * tree-vectorizer.c (vec_info::lookup_def): New function.
7112         * tree-vect-patterns.c (vect_get_internal_def): Use it.
7113         (vect_widened_op_tree): Likewise.
7114         * tree-vect-stmts.c (vect_is_simple_use): Likewise.
7115         * tree-vect-loop.c (vect_analyze_loop_operations): Likewise.
7116         (vectorizable_reduction): Likewise.
7117         (vect_valid_reduction_input_p): Take a stmt_vec_info instead
7118         of a gimple *.
7119         (vect_is_slp_reduction): Update calls accordingly.  Use
7120         vec_info::lookup_def.
7121         (vect_is_simple_reduction): Likewise
7122         * tree-vect-slp.c (vect_detect_hybrid_slp_1): Use vec_info::lookup_def.
7124 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
7126         * tree-vectorizer.h (vec_info::lookup_stmt): Declare.
7127         * tree-vectorizer.c (vec_info::lookup_stmt): New function.
7128         * tree-vect-loop.c (vect_determine_vf_for_stmt): Use it instead
7129         of vinfo_for_stmt.
7130         (vect_determine_vectorization_factor, vect_analyze_scalar_cycles_1)
7131         (vect_compute_single_scalar_iteration_cost, vect_analyze_loop_form)
7132         (vect_update_vf_for_slp, vect_analyze_loop_operations)
7133         (vect_is_slp_reduction, vectorizable_induction)
7134         (vect_transform_loop_stmt, vect_transform_loop): Likewise.
7135         * tree-vect-patterns.c (vect_init_pattern_stmt):
7136         (vect_determine_min_output_precision_1, vect_determine_precisions)
7137         (vect_pattern_recog): Likewise.
7138         * tree-vect-stmts.c (vect_analyze_stmt, vect_transform_stmt): Likewise.
7139         * config/powerpcspe/powerpcspe.c (rs6000_density_test): Likewise.
7140         * config/rs6000/rs6000.c (rs6000_density_test): Likewise.
7141         * tree-vect-slp.c (vect_detect_hybrid_slp_stmts): Likewise.
7142         (vect_detect_hybrid_slp_1, vect_detect_hybrid_slp_2)
7143         (vect_detect_hybrid_slp): Likewise.  Change the walk_stmt_info
7144         info field from a loop to a loop_vec_info.
7146 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
7148         * tree-vectorizer.h (stmt_vec_info): Move typedef earlier in file.
7149         (vec_info::add_stmt): Declare.
7150         * tree-vectorizer.c (vec_info::add_stmt): New function.
7151         * tree-vect-data-refs.c (vect_create_data_ref_ptr): Use it.
7152         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Likewise.
7153         (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
7154         (vectorizable_induction): Likewise.
7155         * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Likewise.
7156         * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
7157         (vectorizable_simd_clone_call, vectorizable_store): Likewise.
7158         (vectorizable_load): Likewise.
7159         * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
7160         (vect_recog_bool_pattern, vect_recog_mask_conversion_pattern)
7161         (vect_recog_gather_scatter_pattern): Likewise.
7162         (append_pattern_def_seq): Likewise.  Remove a check that is
7163         performed by add_stmt itself.
7165 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
7167         * tree-vect-loop.c (vectorizable_reduction): Fix an instance in
7168         which make_ssa_name was called with new_stmt before new_stmt
7169         had been created.
7171 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
7173         * tree-vect-loop.c (vect_valid_reduction_input_p): New function,
7174         split out from...
7175         (vect_is_slp_reduction): ...here...
7176         (vect_is_simple_reduction): ...and here.  Remove repetition of tests
7177         that are already known to be false.
7179 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
7181         * tree-vectorizer.h (vect_free_slp_instance): Add a final_p parameter.
7182         * tree-vect-slp.c (vect_free_slp_tree): Likewise.  Don't update
7183         STMT_VINFO_NUM_SLP_USES when it's true.
7184         (vect_free_slp_instance): Add a final_p parameter and pass it to
7185         vect_free_slp_tree.
7186         (vect_build_slp_tree_2): Update call to vect_free_slp_instance.
7187         (vect_analyze_slp_instance): Likewise.
7188         (vect_slp_analyze_operations): Likewise.
7189         (vect_slp_analyze_bb_1): Likewise.
7190         * tree-vectorizer.c (vec_info): Likewise.
7191         * tree-vect-loop.c (vect_transform_loop): Likewise.
7193 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
7195         * tree-vect-loop.c (vectorizable_reduction): Assert that the
7196         function is not called for second and subsequent members of
7197         a reduction group.
7199 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
7201         * tree-vect-loop.c (get_initial_def_for_reduction): Move special
7202         cases for nested loops from here to ...
7203         (vect_create_epilog_for_reduction): ...here.  Only call
7204         vect_is_simple_use for inner-loop reductions.
7206 2018-07-31  Martin Liska  <mliska@suse.cz>
7208         PR gcov-profile/85338
7209         PR gcov-profile/85350
7210         PR gcov-profile/85372
7211         * profile.c (struct location_triplet): New.
7212         (struct location_triplet_hash): Likewise.
7213         (output_location): Do not output a BB that
7214         is already recorded for a line.
7215         (branch_prob): Use streamed_locations.
7217 2018-07-31  Martin Liska  <mliska@suse.cz>
7219         PR gcov-profile/85370
7220         * coverage.c (coverage_begin_function): Do not mark target
7221         clones as artificial functions.
7223 2018-07-31  Martin Liska  <mliska@suse.cz>
7225         PR gcov-profile/83813
7226         PR gcov-profile/84758
7227         PR gcov-profile/85217
7228         PR gcov-profile/85332
7229         * profile.c (branch_prob): Do not record GOTO expressions
7230         for GIMPLE statements which locations are already streamed.
7232 2018-07-31  Olivier Hainque  <hainque@adacore.com>
7234         * gcc.c (handle_spec_function): Accept a soft_matched_part
7235         argument, as do_spec_1.  Pass it down to ...
7236         (eval_spec_function): Accept a soft_matched_part argument,
7237         and pass it down to ...
7238         (do_spec_2): Accept a soft_matched_part argument, and pass
7239         it down to do_spec_1.
7240         (do_spec_1): Pass soft_matched_part to handle_spec_function.
7241         (handle_braces): Update call to handle_spec_function.
7242         (driver::set_up_specs): Update calls to do_spec_2.
7243         (compare_debug_dump_opt_spec_function): Likewise.
7244         (compare_debug_self_opt_spec_function): Likewise.
7246 2018-07-31  Olivier Hainque  <hainque@adacore.com>
7248         * common.opt (nolibc): New option.
7249         * doc/invoke.texi (Link Options): Document it.
7250         * gcc.c (LINK_GCC_C_SEQUENCE_SPEC): Honor nolibc.
7251         * config/alpha/linux.h: Likewise.
7252         * config/arc/elf.h: Likewise.
7253         * config/arm/uclinux-elf.h: Likewise.
7254         * config/arm/unknown-elf.h: Likewise.
7255         * config/avr/avrlibc.h: Likewise.
7256         * config/bfin/bfin.h: Likewise.
7257         * config/bfin/linux.h: Likewise.
7258         * config/bfin/uclinux.h: Likewise.
7259         * config/darwin.h: Likewise.
7260         * config/darwin10.h: Likewise.
7261         * config/darwin12.h: Likewise.
7262         * config/gnu-user.h: Likewise.
7263         * config/lm32/uclinux-elf.h: Likewise.
7264         * config/pa/pa-hpux11.h: Likewise.
7265         * config/pa/pa64-hpux.h: Likewise.
7266         * config/sparc/sparc.h: Likewise.
7268 2018-07-31  Olivier Hainque  <hainque@adacore.com>
7270         * gcc.c (getenv_spec_function): Prepend '/' to value for allowed
7271         undefined variables.
7273 2018-07-30  Segher Boessenkool  <segher@kernel.crashing.org>
7275         PR target/86640
7276         * config/arm/arm.c (arm_block_set_aligned_vect): Use gen_int_mode
7277         instead of GEN_INT.
7279 2018-07-30  Bernd Edlinger  <bernd.edlinger@hotmail.de>
7281         * tree-ssa-forwprop.c (simplify_builtin_call): Don't create a not NUL
7282         terminated string literal.
7284 2018-07-30  Segher Boessenkool  <segher@kernel.crashing.org>
7286         PR rtl-optimization/85160
7287         * combine.c (is_just_move): New function.
7288         (try_combine): Allow combining two instructions into two if neither of
7289         the original instructions was a move.
7291 2018-07-30  Alexander Monakov  <amonakov@ispras.ru>
7293         PR target/86673
7294         * doc/extend.texi (Global Register Variables): Discourage use of type
7295         qualifiers.
7296         (Local Register Variables): Likewise.
7298 2018-07-30  Richard Sandiford  <richard.sandiford@arm.com>
7300         PR tree-optimization/86506
7301         * hwint.h (ceil_log2): Resync with hwint.c implementation.
7303 2018-07-30  Ilya Leoshkevich  <iii@linux.ibm.com>
7305         PR target/86547
7306         * lra-constraints.c (spill_hard_reg_in_range): When selecting the
7307         hard_regno, make sure no insn between `from` and `to` clobbers it.
7309 2018-07-30  Cesar Philippidis  <cesar@codesourcery.com>
7310             Tom de Vries  <tdevries@suse.de>
7312         * config/nvptx/nvptx.c (PTX_GANG_DEFAULT): Rename to ...
7313         (PTX_DEFAULT_RUNTIME_DIM): ... this.
7314         (nvptx_goacc_validate_dims): Set default worker and gang dims to
7315         PTX_DEFAULT_RUNTIME_DIM.
7316         (nvptx_dim_limit): Ignore GOMP_DIM_WORKER.
7318 2018-07-29  John David Anglin  <danglin@gcc.gnu.org>
7320         * config/pa/pa.c (pa_output_addr_vec): Align address table.
7321         * config/pa/pa.h (JUMP_TABLES_IN_TEXT_SECTION): Revise comment.
7322         * config/pa/pa32-linux.h (JUMP_TABLES_IN_TEXT_SECTION): Define.
7324 2018-07-27  Michael Meissner  <meissner@linux.ibm.com>
7326         * config/rs6000/constraints.md (wG constraint): Delete, no longer
7327         used.
7328         * config/rs6000/predicates.md (p9_fusion_reg_operand): Rename
7329         predicate to reflect toc fusion has been deleted.
7330         (toc_fusion_mem_raw): Delete, no longer used.
7331         (toc_fusion_mem_wrapped): Likewise.
7332         * config/rs6000/rs6000-cpus.def (POWERPC_MASKS): Delete toc
7333         fusion mask bit.
7334         * config/rs6000/rs6000-protos.h (fusion_wrap_memory_address):
7335         Delete, no longer used.
7336         * config/rs6000/rs6000.c (struct rs6000_reg_addr): Delete fields
7337         meant to be used for toc fusion.
7338         (rs6000_debug_print_mode): Delete toc fusion debugging.
7339         (rs6000_debug_reg_global): Likewise.
7340         (rs6000_init_hard_regno_mode_ok): Delete setting up fields for toc
7341         fusion and secondary reload support that were never used.
7342         (rs6000_option_override_internal): Delete TOC fusion, that was only
7343         partially defined, and it did not work unless you also used the
7344         -mcmodel= switch.
7345         (rs6000_legitimate_address_p): Delete TOC fusion support.
7346         (rs6000_opt_masks): Likewise.
7347         (fusion_wrap_memory_address): Delete function, no longer used.
7348         (fusion_split_address); Delete TOC fusion support.
7349         * config/rs6000/rs6000.h (TARGET_TOC_FUSION_INT): Delete, no
7350         longer used with toc fusion being deleted.
7351         (TARGET_TOC_FUSION_FP): Likewise.
7352         * config/rs6000/rs6000.md (UNSPEC_FUSION_ADDIS): Delete TOC fusion
7353         UNSPEC.
7354         (toc fusion spliter): Delete TOC fusion support.
7355         (toc_fusionload_<mode>): Likewise.
7356         (toc_fusionload_di): Likewise.
7357         (fusion_gpr_load_<mode>): Delete generator function, this insn no
7358         longer needs to be named.  Rename predicate to delete TOC fusion.
7359         (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_load): Likewise.
7360         (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_store): Likewise.
7361         (fusion_vsx_<P:mode>_<GPR_FUSION:mode>_load): Likewise.
7362         (fusion_vsx_<P:mode>_<GPR_FUSION:mode>_store): Likewise.
7363         (p9 fusion peephole2s): Rename predicate to delete TOC fusion.
7365 2018-07-27  Kelvin Nilsen  <kelvin@gcc.gnu.org>
7367         * doc/extend.texi (Basic PowerPC Built-in Functions Available on
7368         ISA 2.05):  Replace __uint128_t with __uint128 and __int128_t with
7369         __int128 in built-in function prototypes.
7370         (PowerPC AltiVec Built-in Functions on ISA 2.07): Likewise.
7371         (PowerPC AltiVec Built-in Functions on ISA 3.0): Likewise.
7373 2018-07-27  Martin Sebor  <msebor@redhat.com>
7375         PR tree-optimization/86696
7376         * tree-ssa-strlen.c (get_min_string_length): Handle all integer
7377         types, including enums.
7378         (handle_char_store): Be prepared for the above function to fail.
7380 2018-07-26  Qing Zhao  <qing.zhao@oracle.com>
7382         * builtins.c (inline_expand_builtin_string_cmp): Disable inlining
7383         when optimization level is lower than 2 or optimize for size.
7385 2018-07-26  Martin Sebor  <msebor@redhat.com>
7387         PR tree-optimization/86043
7388         PR tree-optimization/86042
7389         * tree-ssa-strlen.c (handle_builtin_memcpy): Handle strict overlaps.
7390         (get_string_cst_length): Rename...
7391         (get_min_string_length): ...to this.  Add argument.
7392         (handle_char_store): Extend to handle multi-character stores by
7393         MEM_REF.
7394         * tree.c (initializer_zerop): Use new argument.  Handle MEM_REF.
7395         * tree.h (initializer_zerop): Add argument.
7397 2018-07-26  Jakub Jelinek  <jakub@redhat.com>
7399         PR middle-end/86660
7400         * omp-low.c (scan_sharing_clauses): Don't ignore map clauses for
7401         declare target to variables if they have always,{to,from,tofrom} map
7402         kinds.
7404 2018-07-26  Martin Liska  <mliska@suse.cz>
7406         PR lto/86548
7407         * lto-wrapper.c: Add linker_output as prefix
7408         for ltrans_output_file.
7410 2018-07-26  Segher Boessenkool  <segher@kernel.crashing.org>
7412         PR rtl-optimization/85805
7413         * combine.c (reg_nonzero_bits_for_combine): Only use the last set
7414         value for hard registers if that was written in the same mode.
7416 2018-07-26  Martin Liska  <mliska@suse.cz>
7418         PR gcov-profile/86536
7419         * gcov.c (format_gcov): Use printf format %.*f directly
7420         and do not handle special values.
7422 2018-07-25  Claudiu Zissulescu  <claziss@synopsys.com>
7424         * common/config/arc/arc-common.c (arc_option_optimization_table):
7425         Update default optimizations for size.
7427 2018-07-25  Claudiu Zissulescu  <claziss@synopsys.com>
7429         * config/arc/arc.md (movsf_insn): Add short instruction selection.
7430         * config/arc/constraints.md (CfZ): New constraint.
7431         * config/arc/fpu.md (addssf3_fpu): Use CfZ constraint.
7432         (subsf3_fpu): Likewise.
7433         (cmpsf_fpu): Likewise.
7434         (cmpsf_fpu_uneq): Likewise.
7436 2018-07-25  Claudiu Zissulescu  <claziss@synopsys.com>
7438         * config/arc/arc.c (compact_memory_operand_p): Check for uncached
7439         accesses as well.
7440         (arc_is_uncached_mem_p): uncached applies to both the variable and
7441         the pointer.
7443 2018-07-25  Claudiu Zissulescu  <claziss@synopsys.com>
7445         * config/arc/arc.h (ADDITIONAL_REGISTER_NAMES): Add additional
7446         register names.
7448 2018-07-25  David Malcolm  <dmalcolm@redhat.com>
7450         * optinfo-emit-json.cc (class optrecord_json_writer): Convert
7451         field "m_scopes" from vec to auto_vec.
7453 2018-07-25  Martin Liska  <mliska@suse.cz>
7455         * config/powerpcspe/powerpcspe-protos.h (rs6000_loop_align): Fix
7456         return type.
7458 2018-07-25  Richard Biener  <rguenther@suse.de>
7460         PR debug/86654
7461         * dwarf2out.c (dwarf2out_decl): Do not handle nested functions
7462         special wrt context_die late.
7463         (gen_subprogram_die): Re-use DIEs in local scope.
7465 2018-07-25  Richard Sandiford  <richard.sandiford@arm.com>
7467         PR tree-optimization/86644
7468         * hwint.c (ceil_log2): Fix comment.  Return 0 for 0.
7470 2018-07-25  Martin Liska  <mliska@suse.cz>
7472         PR middle-end/86645
7473         * dumpfile.c: And excluded values with TDF_ALL_VALUES.
7474         * dumpfile.h (enum dump_flag): Defince TDF_ALL_VALUES.
7476 2018-07-25  Martin Liska  <mliska@suse.cz>
7478         PR sanitizer/79635
7479         * params.def: Explain ASan abbreviation and provide
7480         a documentation link.
7482 2018-07-24  Martin Sebor  <msebor@redhat.com>
7484         PR tree-optimization/86622
7485         PR tree-optimization/86532
7486         * builtins.h (string_length): Declare.
7487         * builtins.c (c_strlen): Correct handling of non-constant offsets.
7488         (check_access): Be prepared for non-constant length ranges.
7489         (string_length): Make extern.
7490         * expr.c (string_constant): Only handle the minor non-constant
7491         array index.  Use string_constant to compute the length of
7492         a generic string constant.
7494 2018-07-24  Richard Sandiford  <richard.sandiford@arm.com>
7496         PR tree-optimization/86618
7497         * tree-vect-stmts.c (vectorizable_call): Don't take the address
7498         of LOOP_VINFO_MASKS (loop_vinfo) when loop_vinfo is null.
7500 2018-07-24  David Malcolm  <dmalcolm@redhat.com>
7502         PR tree-optimization/86636
7503         * json.cc (json::object::set): Fix comment.  Add assertions.
7504         (json::array::append): Move here from json.h.  Add comment and an
7505         assertion.
7506         (json::string::string): Likewise.
7507         * json.h (json::array::append): Move to json.cc.
7508         (json::string::string): Likewise.
7509         * optinfo-emit-json.cc
7510         (optrecord_json_writer::impl_location_to_json): Assert that we
7511         aren't attempting to write out UNKNOWN_LOCATION, or an ad-hoc
7512         wrapper around it.  Expand the location once, rather than three
7513         times.
7514         (optrecord_json_writer::inlining_chain_to_json): Fix the check for
7515         UNKNOWN_LOCATION, to use LOCATION_LOCUS to look through ad-hoc
7516         wrappers.
7517         (optrecord_json_writer::optinfo_to_json): Likewise, in four
7518         places.  Fix some overlong lines.
7520 2018-07-24  Matthew Malcomson  <matthew.malcomson@arm.com>
7522         * config/aarch64/aarch64-simd.md
7523         (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w<mode>): Split into...
7524         (aarch64_<ANY_EXTEND:su>subw<mode>): ... This...
7525         (aarch64_<ANY_EXTEND:su>addw<mode>): ... And this.
7526         (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w<mode>_internal): Split into...
7527         (aarch64_<ANY_EXTEND:su>subw<mode>_internal): ... This...
7528         (aarch64_<ANY_EXTEND:su>addw<mode>_internal): ... And this.
7529         (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w2<mode>_internal): Split into...
7530         (aarch64_<ANY_EXTEND:su>subw2<mode>_internal): ... This...
7531         (aarch64_<ANY_EXTEND:su>addw2<mode>_internal): ... And this.
7533 2018-07-24  Jakub Jelinek  <jakub@redhat.com>
7535         PR middle-end/86627
7536         * expmed.c (expand_divmod): Punt if d == HOST_WIDE_INT_MIN
7537         and size > HOST_BITS_PER_WIDE_INT.  For size > HOST_BITS_PER_WIDE_INT
7538         and abs_d == d, do the power of two handling if profitable.
7540 2018-07-24  Richard Biener  <rguenther@suse.de>
7542         * match.pd: Add BIT_FIELD_REF canonicalizations.
7544 2018-07-23  Bernd Edlinger  <bernd.edlinger@hotmail.de>
7546         PR c/86617
7547         * genmatch.c (dt_operand::gen_match_op): Avoid folding volatile values.
7549 2018-07-23  Bernd Edlinger  <bernd.edlinger@hotmail.de>
7551         * gimple-fold.c (gimple_fold_builtin_printf): Don't create a not NUL
7552         terminated STRING_CST object.
7554 2018-07-23  Bernd Edlinger  <bernd.edlinger@hotmail.de>
7556         hsa-dump.c (dump_hsa_symbol): Avoid out of scope access to buf.
7558 2018-07-23  Segher Boessenkool  <segher@kernel.crashing.org>
7560         * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Adjust.
7561         * config/rs6000/rs6000-protos.h (rs6000_split_v4si_init): Delete.
7562         * config/rs6000/rs6000.c (rs6000_expand_vector_init): Always force
7563         the elements into a register.
7564         (rs6000_split_v4si_init_di_reg): Delete.
7565         (rs6000_split_v4si_init): Delete.
7566         * config/rs6000/vsx.md (unspec): Delete UNSPEC_VSX_VEC_INIT.
7567         (vsx_init_v4si): Rewrite as a define_expand.
7569 2018-07-23  Segher Boessenkool  <segher@kernel.crashing.org>
7571         * config/rs6000/rs6000.md (splitters for rldimi and rlwimi with the
7572         zero_extend argument from memory): New.
7574 2018-07-22  Martin Sebor  <msebor@redhat.com>
7576         PR bootstrap/86621
7577         * gimple-ssa-warn-alloca.c (alloca_call_type_by_arg): Avoid
7578         diagnosing calls with unknown arguments unless -Walloca-larger-than
7579         is restricted to less than PTRDIFF_MAX bytes.
7581 2018-07-22  Gerald Pfeifer  <gerald@pfeifer.com>
7583         * doc/gcov.texi (Invoking Gcov): Editorial changes.
7585 2018-07-20  David Malcolm  <dmalcolm@redhat.com>
7587         * pretty-print.c (text_info::set_location): Remove redundant
7588         "line_table" parameter from call to rich_location::set_range.
7590 2018-07-20  Martin Sebor  <msebor@redhat.com>
7592         PR middle-end/82063
7593         * builtins.c (expand_builtin_alloca): Adjust.
7594         * calls.c (alloc_max_size): Simplify.
7595         * cgraphunit.c (cgraph_node::expand): Adjust.
7596         * common.opt (larger_than_size, warn_frame_larger_than): Remove
7597         variables.
7598         (frame_larger_than_size): Same.
7599         (-Wframe-larger-than, -Wlarger-than, -Wstack-usage): Change options
7600         to take a HOST_WIDE_INT argument and accept a byte-size suffix.
7601         Initialize.
7602         * doc/invoke.texi (GCC Command Options): Document option arguments.
7603         Explain byte-size arguments and suffixes.
7604         (-Wvla-larger-than, -Wno-alloc-size-larger-than): Update.
7605         (-Wno-alloca-larger-than, -Wno-vla-larger-than): Same.
7606         (-Wframe-larger-than, -Wlarger-than, -Wstack-usage): Same.
7607         * doc/options.texi (UInteger): Expand.
7608         (Host_Wide_Int, ByteSize): Document new properties.
7609         * final.c (final_start_function_1): Include sizes in an error message.
7610         * function.c (frame_offset_overflow): Same.
7611         * gimple-ssa-warn-alloca.c (pass_walloca::gate): Adjust.
7612         (alloca_call_type_by_arg): Change function argument to HOST_WIDE_INT.
7613         Diagnose unbounded alloca calls only for limits of less than
7614         PTRDIFF_MAX.
7615         (alloca_call_type): Adjust.  Diagnose possibly out-of-bounds alloca
7616         calls and VLA size only for limits of less than PTRDIFF_MAX.  Same
7617         for alloca(0).
7618         (pass_walloca::execute): Adjust.  Diagnose alloca calls in loops
7619         only for limits of less than PTRDIFF_MAX.
7620         * langhooks-def.h (lhd_handle_option): Change function argument
7621         to HOST_WIDE_INT.
7622         * langhooks.c (lhd_handle_option): Same.
7623         * langhooks.h (handle_option): Same.
7624         * opt-functions.awk (switch_bit_fields): Handle Host_Wide_Int and
7625         ByteSize flags.
7626         (var_type, var_type_struct): Same.
7627         (var_set): Handle ByteSize flag.
7628         * optc-gen.awk: Add comments to output to ease debugging.  Make
7629         use of HOST_WIDE_INT where appropriate.
7630         * opts-gen-save.awk:  Use %lx to format unsigned long.
7631         * opth-gen.awk: Change function argument to HOST_WIDE_INT.
7632         * opts-common.c (integral_argument): Return HOST_WIDE_INT and add
7633         arguments.  Parse bytes-size suffixes.
7634         (enum_arg_to_value): Change function argument to HOST_WIDE_INT.
7635         (enum_value_to_arg): Same.
7636         (decode_cmdline_option): Handle cl_host_wide_int.  Adjust.
7637         (handle_option): Adjust.
7638         (generate_option): Change function argument to HOST_WIDE_INT.
7639         (cmdline_handle_error): Adjust.
7640         (read_cmdline_option): Change function argument to HOST_WIDE_INT.
7641         (set_option): Change function argument to HOST_WIDE_INT.
7642         (option_enabled): Handle cl_host_wide_int.
7643         (get_option_state): Handle CLVC_SIZE.
7644         (control_warning_option): Same.
7645         * opts.c (common_handle_option): Change function argument to
7646         HOST_WIDE_INT.  Remove handling of OPT_Walloca_larger_than_ and
7647         OPT_Wvla_larger_than_.
7648         * opts.h (enum cl_var_type): Add an enumerator.
7649         * stor-layout.c (layout_decl): Print a more meaningful warning.
7650         * toplev.c (output_stack_usage): Adjust.
7652 2018-07-20  Qing Zhao  <qing.zhao@oracle.com>
7654         * builtins.c (expand_builtin_memcmp): Delete the last parameter for
7655         call to inline_expand_builtin_string_cmp.
7656         (expand_builtin_strcmp): Likewise.
7657         (expand_builtin_strncmp): Likewise.
7658         (inline_string_cmp): Delete the last parameter, change char_type_node
7659         to unsigned_char_type_node for strcmp/strncmp, add conversions to the
7660         two operands.
7661         (inline_expand_builtin_string_cmp): Delete the last parameter, give up
7662         the inlining expansion on target where the type of the call has same or
7663         narrower precision than unsigned char.
7665 2018-07-20  David Malcolm  <dmalcolm@redhat.com>
7667         * Makefile.in (OBJS): Add json.o and optinfo-emit-json.o.
7668         (CFLAGS-optinfo-emit-json.o): Define TARGET_NAME.
7669         * common.opt (fsave-optimization-record): New option.
7670         * coretypes.h (struct kv_pair): Move here from dumpfile.c.
7671         * doc/invoke.texi (-fsave-optimization-record): New option.
7672         * dumpfile.c: Include "optinfo-emit-json.h".
7673         (struct kv_pair): Move to coretypes.h.
7674         (optgroup_options): Make non-static.
7675         (dump_context::end_scope): Call
7676         optimization_records_maybe_pop_dump_scope.
7677         * dumpfile.h (optgroup_options): New decl.
7678         * json.cc: New file.
7679         * json.h: New file.
7680         * optinfo-emit-json.cc: New file.
7681         * optinfo-emit-json.h: New file.
7682         * optinfo.cc: Include "optinfo-emit-json.h".
7683         (optinfo::emit): Call optimization_records_maybe_record_optinfo.
7684         (optinfo_enabled_p): Check optimization_records_enabled_p.
7685         (optinfo_wants_inlining_info_p): Likewise.
7686         * optinfo.h: Update comment.
7687         * profile-count.c (profile_quality_as_string): New function.
7688         * profile-count.h (profile_quality_as_string): New decl.
7689         (profile_count::quality): New accessor.
7690         * selftest-run-tests.c (selftest::run_tests): Call json_cc_tests
7691         and optinfo_emit_json_cc_tests.
7692         * selftest.h (selftest::json_cc_tests): New decl.
7693         (selftest::optinfo_emit_json_cc_tests): New decl.
7694         * toplev.c: Include "optinfo-emit-json.h".
7695         (compile_file): Call optimization_records_finish.
7696         (do_compile): Call optimization_records_start.
7697         * tree-ssa-live.c: Include optinfo.h.
7698         (remove_unused_scope_block_p): Retain inlining information if
7699         optinfo_wants_inlining_info_p returns true.
7701 2018-07-20  Richard Biener  <rguenther@suse.de>
7703         PR debug/86585
7704         * dwarf2out.c (dwarf2out_die_ref_for_decl): Test in_lto_p
7705         to cover -flto-partition=none.
7707 2018-07-20  Martin Liska  <mliska@suse.cz>
7709         * tree.h (DECL_LOCATION_RANGE): Remove unused macro.
7710         (get_decl_source_range): Remove unused function.
7712 2018-07-20  Richard Biener  <rguenther@suse.de>
7714         * tree-ssa-sccvn.h (struct vn_nary_op_s): Add next member.
7715         (struct vn_phi_s): Likewise.
7716         (struct vn_reference_s): Likewise.
7717         * tree-ssa-sccvn.c (vn_nary_op_hasher::equal): Add shortcut
7718         for searching the slot of an entry known to be in the hash itself.
7719         (vn_phi_hasher::equal): Likewise.
7720         (vn_reference_hasher::equal): Likewise.
7721         (last_inserted_ref, last_inserted_phi, last_inserted_nary): New
7722         globals.
7723         (optimistic_info, current_info): Remove, keeping only valid_info.
7724         (vn_reference_lookup_1): Remove fallback lookup.
7725         (vn_reference_lookup_2): Likewise.
7726         (vn_nary_op_lookup_1): Likewise.
7727         (vn_phi_lookup): Likewise.
7728         (vn_nary_build_or_lookup_1): Make sure to not chain the built
7729         hash element.
7730         (vn_reference_insert): Adjust, chain the inserted hash element
7731         at last_inserted_ref.
7732         (vn_reference_insert_pieces): Likewise.
7733         (visit_reference_op_call): Likewise.
7734         (vn_nary_op_insert_into): Chain the inserted hash element at
7735         last_inserted_nary.
7736         (vn_nary_op_insert_pieces): Adjust.
7737         (vn_nary_op_insert): Likewise.
7738         (vn_nary_op_insert_stmt): Likewise.
7739         (vn_phi_insert): Adjust, chain the inserted hash element at
7740         last_inserted_phi.
7741         (process_scc): Remove clearing and copying the optimistic
7742         table.  Instead remove elements inserted during an optimistic
7743         iteration from the single table we maintain.
7744         (init_scc_vn): Adjust.
7745         (free_scc_vn): Likewise.
7746         (sccvn_dom_walker::record_cond): Likewise.
7747         (sccvn_dom_walker::after_dom_children): Likewise.
7749 2018-07-19  Martin Sebor  <msebor@redhat.com>
7751         PR tree-optimization/84047
7752         PR tree-optimization/83776
7753         * tree-vrp.c (vrp_prop::check_mem_ref): New function.
7754         (check_array_bounds): Call it.
7756 2018-07-19  Martin Sebor  <msebor@redhat.com>
7758         * align.h (align_flags): Use member initialization.
7760 2018-07-19  David Malcolm  <dmalcolm@redhat.com>
7762         * Makefile.in (OBJS): Add optinfo.o.
7763         * coretypes.h (class symtab_node): New forward decl.
7764         (struct cgraph_node): New forward decl.
7765         (class varpool_node): New forward decl.
7766         * dump-context.h: New file.
7767         * dumpfile.c: Include "optinfo.h", "dump-context.h", "cgraph.h",
7768         "tree-pass.h".
7769         (refresh_dumps_are_enabled): Use optinfo_enabled_p.
7770         (set_dump_file): Call dumpfile_ensure_any_optinfo_are_flushed.
7771         (set_alt_dump_file): Likewise.
7772         (dump_context::~dump_context): New dtor.
7773         (dump_gimple_stmt): Move implementation to...
7774         (dump_context::dump_gimple_stmt): ...this new member function.
7775         Add the stmt to any pending optinfo, creating one if need be.
7776         (dump_gimple_stmt_loc): Move implementation to...
7777         (dump_context::dump_gimple_stmt_loc): ...this new member function.
7778         Start a new optinfo and add the stmt to it.
7779         (dump_gimple_expr): Move implementation to...
7780         (dump_context::dump_gimple_expr): ...this new member function.
7781         Add the stmt to any pending optinfo, creating one if need be.
7782         (dump_gimple_expr_loc): Move implementation to...
7783         (dump_context::dump_gimple_expr_loc): ...this new member function.
7784         Start a new optinfo and add the stmt to it.
7785         (dump_generic_expr): Move implementation to...
7786         (dump_context::dump_generic_expr): ...this new member function.
7787         Add the tree to any pending optinfo, creating one if need be.
7788         (dump_generic_expr_loc): Move implementation to...
7789         (dump_context::dump_generic_expr_loc): ...this new member
7790         function.  Add the tree to any pending optinfo, creating one if
7791         need be.
7792         (dump_printf): Move implementation to...
7793         (dump_context::dump_printf_va): ...this new member function.  Add
7794         the text to any pending optinfo, creating one if need be.
7795         (dump_printf_loc): Move implementation to...
7796         (dump_context::dump_printf_loc_va): ...this new member function.
7797         Start a new optinfo and add the stmt to it.
7798         (dump_dec): Move implementation to...
7799         (dump_context::dump_dec): ...this new member function.  Add the
7800         value to any pending optinfo, creating one if need be.
7801         (dump_context::dump_symtab_node): New member function.
7802         (dump_context::get_scope_depth): New member function.
7803         (dump_context::begin_scope): New member function.
7804         (dump_context::end_scope): New member function.
7805         (dump_context::ensure_pending_optinfo): New member function.
7806         (dump_context::begin_next_optinfo): New member function.
7807         (dump_context::end_any_optinfo): New member function.
7808         (dump_context::s_current): New global.
7809         (dump_context::s_default): New global.
7810         (dump_scope_depth): Delete global.
7811         (dumpfile_ensure_any_optinfo_are_flushed): New function.
7812         (dump_symtab_node): New function.
7813         (get_dump_scope_depth): Reimplement in terms of dump_context.
7814         (dump_begin_scope): Likewise.
7815         (dump_end_scope): Likewise.
7816         (selftest::temp_dump_context::temp_dump_context): New ctor.
7817         (selftest::temp_dump_context::~temp_dump_context): New dtor.
7818         (selftest::verify_item): New function.
7819         (ASSERT_IS_TEXT): New macro.
7820         (ASSERT_IS_TREE): New macro.
7821         (ASSERT_IS_GIMPLE): New macro.
7822         (selftest::test_capture_of_dump_calls): New test.
7823         (selftest::dumpfile_c_tests): Call it.
7824         * dumpfile.h (dump_printf, dump_printf_loc, dump_basic_block)
7825         (dump_generic_expr_loc, dump_generic_expr, dump_gimple_stmt_loc)
7826         (dump_gimple_stmt, dump_dec): Gather these related decls and add a
7827         descriptive comment.
7828         (dump_function, print_combine_total_stats, enable_rtl_dump_file)
7829         (dump_node, dump_bb): Move these unrelated decls.
7830         (class dump_manager): Add leading comment.
7831         * optinfo.cc: New file.
7832         * optinfo.h: New file.
7834 2018-07-19  Michael Collison  <michael.collison@arm.com>
7835             Richard Henderson <rth@redhat.com>
7837         * config/aarch64/aarch64.md (subv<GPI>4, usubv<GPI>4): New patterns.
7838         (subti): Handle op1 zero.
7839         (subvti4, usub4ti4): New.
7840         (*sub<GPI>3_compare1_imm): New.
7841         (sub<GPI>3_carryinCV): New.
7842         (*sub<GPI>3_carryinCV_z1_z2, *sub<GPI>3_carryinCV_z1): New.
7843         (*sub<GPI>3_carryinCV_z2, *sub<GPI>3_carryinCV): New.
7845 2018-07-19  Michael Collison  <michael.collison@arm.com>
7846             Richard Henderson <rth@redhat.com>
7848         * config/aarch64/aarch64.md: (addv<GPI>4, uaddv<GPI>4): New.
7849         (addti3): Create simpler code if low part is already known to be 0.
7850         (addvti4, uaddvti4): New.
7851         (*add<GPI>3_compareC_cconly_imm): New.
7852         (*add<GPI>3_compareC_cconly): New.
7853         (*add<GPI>3_compareC_imm): New.
7854         (*add<GPI>3_compareC): Rename from add<GPI>3_compare1; do not
7855         handle constants within this pattern..
7856         (*add<GPI>3_compareV_cconly_imm): New.
7857         (*add<GPI>3_compareV_cconly): New.
7858         (*add<GPI>3_compareV_imm): New.
7859         (add<GPI>3_compareV): New.
7860         (add<GPI>3_carryinC, add<GPI>3_carryinV): New.
7861         (*add<GPI>3_carryinC_zero, *add<GPI>3_carryinV_zero): New.
7862         (*add<GPI>3_carryinC, *add<GPI>3_carryinV): New.
7863         ((*add<GPI>3_compareC_cconly_imm): Replace 'ne' operator
7864         with 'comparison' operator.
7865         (*add<GPI>3_compareV_cconly_imm): Ditto.
7866         (*add<GPI>3_compareV_cconly): Ditto.
7867         (*add<GPI>3_compareV_imm): Ditto.
7868         (add<GPI>3_compareV): Ditto.
7869         (add<mode>3_carryinC): Ditto.
7870         (*add<mode>3_carryinC_zero): Ditto.
7871         (*add<mode>3_carryinC): Ditto.
7872         (add<mode>3_carryinV): Ditto.
7873         (*add<mode>3_carryinV_zero): Ditto.
7874         (*add<mode>3_carryinV): Ditto.
7876 2018-07-19  Michael Collison  <michael.collison@arm.com>
7877             Richard Henderson <rth@redhat.com>
7879         * config/aarch64/aarch64-modes.def (CC_V): New.
7880         * config/aarch64/aarch64-protos.h
7881         (aarch64_addti_scratch_regs): Declare
7882         (aarch64_subvti_scratch_regs): Declare.
7883         (aarch64_expand_subvti): Declare.
7884         (aarch64_gen_unlikely_cbranch): Declare
7885         * config/aarch64/aarch64.c (aarch64_select_cc_mode): Test
7886         for signed overflow using CC_Vmode.
7887         (aarch64_get_condition_code_1): Handle CC_Vmode.
7888         (aarch64_gen_unlikely_cbranch): New function.
7889         (aarch64_addti_scratch_regs): New function.
7890         (aarch64_subvti_scratch_regs): New function.
7891         (aarch64_expand_subvti): New function.
7893 2018-07-19  Andre Vieira  <andre.simoesdiasvieira@arm.com>
7895         * config/aarch64/aarch64-option-extensions.def: New entry for profile
7896         extension.
7897         * config/aarch64/aarch64.h (AARCH64_FL_PROFILE): New.
7898         * doc/invoke.texi (aarch64-feature-modifiers): New entry for profile
7899         extension.
7901 2018-07-19  Andre Vieira  <andre.simoesdiasvieira@arm.com>
7903         PR target/83009
7904         * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
7905         address check not strict.
7907 2018-07-19  Andre Vieira  <andre.simoesdiasvieira@arm.com>
7909         * config/aarch64/aarch64-simd.md (aarch64_simd_mov<VQ:mode>): Replace
7910         Umq with Umn.
7911         (store_pair_lanes<mode>): Likewise.
7912         * config/aarch64/aarch64-protos.h (aarch64_addr_query_type): Add new
7913         enum value 'ADDR_QUERY_LDP_STP_N'.
7914         * config/aarch64/aarch64.c (aarch64_addr_query_type): Likewise.
7915         (aarch64_print_address_internal): Add declaration.
7916         (aarch64_print_ldpstp_address): Remove.
7917         (aarch64_classify_address): Adapt mode for 'ADDR_QUERY_LDP_STP_N'.
7918         (aarch64_print_operand): Change printing of 'y'.
7919         * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Use
7920         new enum value 'ADDR_QUERY_LDP_STP_N', don't hardcode mode and use
7921         'true' rather than '1'.
7922         * config/aarch64/constraints.md (Uml): Likewise.
7923         (Uml): Rename to Umn.
7924         (Umq): Remove.
7926 2018-07-19  Richard Biener  <rguenther@suse.de>
7928         * tree-ssa-sccvn.h (struct vn_phi_s): Make phiargs member
7929         a trailing array.
7930         * tree-ssa-sccvn.c: Remove alloc-pool.h use.
7931         (vn_phi_hasher): Derive from nofree_ptr_hash and remove remove method.
7932         (vn_reference_hasher): Likewise.
7933         (struct vn_tables_s): Remove obstack and alloc-pool members.
7934         (vn_tables_obstack, vn_tables_insert_obstack): New global obstacks.
7935         (vn_nary_build_or_lookup_1): Manually build in vn_tables_insert_obstack.
7936         (vn_reference_insert): Allocate from obstack instead of from alloc-pool.
7937         (vn_reference_insert_pieces): Likewise.
7938         (alloc_vn_nary_op_noinit): Adjust.
7939         (vn_nary_op_insert_stmt): Allocate phiargs in-place.
7940         (vn_phi_eq): Adjust.
7941         (shared_lookup_phiargs): Remove.
7942         (vn_phi_lookup): Allocate temporary vn_phi_s on the stack.
7943         (vn_phi_insert): Allocate from obstack instead of from alloc-pool.
7944         (visit_reference_op_call): Likewise.
7945         (copy_nary, copy_phi, copy_reference): Remove.
7946         (process_scc): Rewind the obstack when iterating.  Do not
7947         copy the elements to valid_info but just move them from one
7948         hashtable to the other.
7949         (allocate_vn_table): Adjust.
7950         (free_vn_table): Likewise.
7951         (init_scc_vn): Likewise.
7952         (free_scc_vn): Likewise.
7954 2018-07-19  H.J. Lu  <hongjiu.lu@intel.com>
7956         PR target/86560
7957         * config/i386/i386.c (rest_of_insert_endbranch): Lookup
7958         indirect_return as function type attribute.
7959         (ix86_attribute_table): Change indirect_return to function
7960         type attribute.
7961         * doc/extend.texi: Update indirect_return attribute.
7963 2018-07-19  Aldy Hernandez  <aldyh@redhat.com>
7965         * wide-int.h (widest2_int): New.
7966         * gimple-fold.c (arith_overflowed_p): Use it.
7967         * tree.h (widest2_int_cst): New.
7968         * tree-vrp.c (wide_int_binop_overflow): Rename from
7969         vrp_int_const_binop.
7970         Rewrite to work on trees.
7971         (extract_range_from_multiplicative_op_1): Abstract code to...
7972         (wide_int_range_min_max): ...here.
7973         (wide_int_range_cross_product): ...and here.
7974         (extract_range_from_binary_expr_1): Abstract overflow code to...
7975         (wide_int_range_mult_wrapping): ...here.
7976         * tree-vrp.h (wide_int_range_cross_product): New.
7977         (wide_int_range_mult_wrapping): New.
7979 2018-07-19  Andrew Senkevich  <andrew.senkevich@intel.com>
7980             Julia Koval  <julia.koval@intel.com>
7982         * config/i386/x86-tune-costs.h (skylake_memcpy,
7983         skylake_memset): Replace rep_prefix with unrolling for size 512.
7985 2018-07-18  Kugan Vivekanandarajah  <kuganv@linaro.org>
7987         PR middle-end/86544
7988         * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): Handle
7989         comparision with EQ_EXPR in last stmt.
7991 2018-07-18  Kelvin Nilsen  <kelvin@gcc.gnu.org>
7993         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Rename
7994         this subsection to "PowerPC AltiVec/VSX Built-in Functions".
7995         (PowerPC AltiVec/VSX Built-in Functions): New name for subsection
7996         previously known as "PowerPC AltiVec Built-in Functions".  Move
7997         some material to new subsubsections "PowerPC AltiVec Built-in
7998         Functions on ISA 2.06" and "PowerPC AltiVec Built-in Functions on
7999         ISA 2.07".
8000         (PowerPC Altivec Built-in Functions on ISA 2.05): New subsubsection.
8001         (PowerPC Altivec Built-in Functions on ISA 2.06): Likewise.
8002         (PowerPC Altivec Built-in Functions on ISA 2.07): Likewise.
8003         (PowerPC Altivec Built-in Functions on ISA 3.0): Likewise.
8005 2018-07-18  Richard Biener  <rguenther@suse.de>
8007         PR tree-optimization/86557
8008         * tree-vect-patterns.c (vect_recog_divmod_pattern): Also handle
8009         EXACT_DIV_EXPR.
8011 2018-07-18  Ilya Leoshkevich  <iii@linux.ibm.com>
8013         * config/s390/s390.c (s390_function_profiler): Generate CFI.
8015 2018-07-17  Jeff Law  <law@redhat.com>
8017         * config/arm/arm.c (get_label_padding): Update for recent
8018         changes to label_to_alignment.
8020         PR tree-optimization/86010
8021         * tree-ssa-dse.c (compute_trims): Fix typo/thinko.
8023         * config/mips/mips.c (vr4130_align_insns): Update for recent
8024         changes to label_to_alignment.
8026         * config/frv/frv.c (frv_label_align): Update for recent changes
8027         to label_to_alignment.
8029         * config/nios2/nios2.c (nios2_label_align): Update for recent
8030         changes which dropped ALIGN_LABELS_LOG.
8032 2018-07-17  Andreas Schwab  <schwab@linux-m68k.org>
8034         * config/m68k/m68k.md (umulsi3_highpart+1, const_umulsi3_highpart)
8035         (smulsi3_highpart+1, const_smulsi3_highpart): Add CC_STATUS_INIT.
8037 2018-07-17  Claudiu Zissulescu  <claziss@synopsys.com>
8039         * config/arc/arc.c (arc_label_align): Use align_labels instead of
8040         deprecated align_labels_log.
8042 2018-07-17  Richard Biener  <rguenther@suse.de>
8044         PR lto/86456
8045         * dwarf2out.c (init_sections_and_labels): Always generate
8046         a debug_line_str_section for early LTO debug.
8047         (dwarf2out_finish): Reset debug_line_str_hash output early.
8048         Bump counter for extra dwarf5 .debug_loc labels to not conflict
8049         with fat LTO part.
8050         (dwarf2out_early_finish): Output debug_line_str.
8052 2018-07-17  Robin Dapp  <rdapp@linux.ibm.com>
8054         * config/s390/s390.c (preferred_la_operand_p): Do not use LA with
8055         index register on z196 or later.
8057 2018-07-17  Robin Dapp  <rdapp@linux.ibm.com>
8059         * config/s390/s390.c (s390_default_align): Set default function
8060         alignment to 16.
8061         (s390_override_options_after_change): Call s390_default align.
8062         (s390_option_override_internal): Call s390_default_align.
8063         (TARGET_OVERRIDE_OPTIONS_AFTER_CHANGE): Define.
8065 2018-07-17  Jakub Jelinek  <jakub@redhat.com>
8067         PR middle-end/86542
8068         * omp-low.c (create_task_copyfn): Copy over also fields corresponding
8069         to _looptemp_ clauses, other than the first two.
8071 2018-07-17  Martin Liska  <mliska@suse.cz>
8073         * opts.c: Do not enable OPT_falign_* for -Os.
8075 2018-07-17  Martin Liska  <mliska@suse.cz>
8077         * align.h (MAX_CODE_ALIGN): New.
8078         (MAX_CODE_ALIGN_VALUE): New.
8079         * common/config/i386/i386-common.c (ix86_handle_option):
8080         (MAX_CODE_ALIGN): Moved to align.h.
8081         * final.c (MAX_CODE_ALIGN): Likewise.
8082         * opts.c (parse_and_check_align_values):
8083         (MAX_CODE_ALIGN): Likewise.
8084         (MAX_CODE_ALIGN_VALUE): Likewise.
8086 2018-07-17  Martin Liska  <mliska@suse.cz>
8088         * config/i386/att.h (ASM_OUTPUT_ALIGN): Fix spacing
8089         in order to fulfil coding style.
8090         * config/i386/cygming.h (ASM_OUTPUT_ALIGN): Likewise.
8091         * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
8092         * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
8093         * config/iq2000/iq2000.h (ASM_OUTPUT_ALIGN): Likewise.
8094         * config/pa/pa.h (ASM_OUTPUT_ALIGN): Likewise.
8095         * config/sparc/sol2.h (ASM_OUTPUT_ALIGN_WITH_NOP): Likewise.
8096         * config/sparc/sparc.h (ASM_OUTPUT_ALIGN): Likewise.
8097         * config/visium/visium.h (ASM_OUTPUT_ALIGN): Likewise.
8098         (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
8100 2018-07-17  Martin Liska  <mliska@suse.cz>
8102         * align.h: New file.
8103         * config/alpha/alpha.c (alpha_align_insns_1): Use align_functions
8104         directly.
8105         * config/i386/i386.c (ix86_avoid_jump_mispredicts): Use new return type
8106         align_flags of label_to_alignment.
8107         * config/m32r/m32r.h (LOOP_ALIGN): Wrap returned values into
8108         align_flags class.
8109         * config/m68k/m68k.c: Do not use removed align_labels_value and
8110         align_loops_value.
8111         * config/nds32/nds32.h (JUMP_ALIGN): Wrap result into align_flags class.
8112         (LOOP_ALIGN): Likewise.
8113         (LABEL_ALIGN): Likewise.
8114         * config/powerpcspe/powerpcspe.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
8115         Remove not used macro.
8116         (rs6000_loop_align): Change return type to align_flags.
8117         (rs6000_loop_align_max_skip): Remove.
8118         * config/rs6000/rs6000-protos.h (rs6000_loop_align):
8119         Change return type to align_flags.
8120         * config/rs6000/rs6000.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
8121         Remove not used macro.
8122         (rs6000_loop_align):  Change return type to align_flags.
8123         (rs6000_loop_align_max_skip): Remove.
8124         * config/rx/rx.h (JUMP_ALIGN): Wrap integer values
8125         * config/rx/rx-protos.h (rx_align_for_label): Make it
8126         static function.
8127         * config/rx/rx.c (rx_align_for_label): Change return type
8128         to align_flags.
8129         (rx_max_skip_for_label): Remove TARGET_ASM_*_ALIGN_MAX_SKIP
8130         macro definitions.
8131         into align_flags class.
8132         (LABEL_ALIGN): Likewise.
8133         (LOOP_ALIGN): Likewise.
8134         * config/s390/s390.c (s390_label_align): Use align_flags
8135         class member.
8136         (s390_asm_output_function_label): Likewise.
8137         * config/sh/sh.c (sh_override_options_after_change):
8138         Use align_flags class directly without macros.
8139         (find_barrier): Likewise.
8140         (barrier_align): Likewise.
8141         (sh_loop_align): Likewise.
8142         * config/spu/spu.c (spu_option_override):
8143         Use align_flags_tuple::get_value instead of removed macros.
8144         (spu_sched_init): Likewise.
8145         * config/spu/spu.h (GTY): Likewise.
8146         * config/visium/visium.c (visium_option_override):
8147         Set "8" as default secondary alignment.
8148         * config/visium/visium.h (SUBALIGN_LOG): Define to 3
8149         in order to guarantee secondary alignment of 8.
8150         * coretypes.h: Include align.h header file.
8151         * doc/tm.texi: Remove TARGET_ASM_JUMP_ALIGN_MAX_SKIP,
8152         TARGET_ASM_LOOP_ALIGN_MAX_SKIP, TARGET_ASM_LABEL_ALIGN_MAX_SKIP
8153         and TARGET_ASM_LABEL_ALIGN_AFTER_BARRIER_MAX_SKIP macros.
8154         * doc/tm.texi.in: Likewise.
8155         * final.c (struct label_alignment): Remove not used structure.
8156         (LABEL_ALIGN): Change type to align_flags.
8157         (LOOP_ALIGN): Likewise.
8158         (JUMP_ALIGN): Likewise.
8159         (default_loop_align_max_skip): Remove.
8160         (default_label_align_max_skip): Likewise.
8161         (default_jump_align_max_skip): Likewise.
8162         (default_label_align_after_barrier_max_skip):
8163         (LABEL_TO_ALIGNMENT): Change to access label_align vector.
8164         (LABEL_TO_MAX_SKIP): Remove.
8165         (label_to_alignment): Return align_flags type instead of integer.
8166         (label_to_max_skip): Remove.
8167         (align_fuzz): Use align_flags type.
8168         (compute_alignments): Use align_flags type and use align_flags::max
8169         to combine multiple alignments.
8170         (grow_label_align): Grow vec instead of C array.
8171         (update_alignments): Assign just LABEL_TO_ALIGNMENT.
8172         (shorten_branches):  Use align_flags type and use align_flags::max
8173         to combine multiple alignments.
8174         (final_scan_insn_1): Remove usage of secondary alignment that comes
8175         from label alignment, but instead use proper secondary alignment
8176         which is computed in grow_label_align.
8177         * flags.h (struct align_flags_tuple): Move to align.h.
8178         (struct align_flags): Likewise.
8179         (state_align_loops): Rename to align_loops.
8180         (state_align_jumps): Rename to align_jumps.
8181         (state_align_labels): Rename to align_labels.
8182         (state_align_functions): Rename to align_functions.
8183         (align_loops_log): Remove.
8184         (align_jumps_log): Remove.
8185         (align_labels_log): Remove.
8186         (align_functions_log): Remove.
8187         (align_loops_max_skip): Remove.
8188         (align_jumps_max_skip): Remove.
8189         (align_labels_max_skip): Remove.
8190         (align_functions_max_skip): Remove.
8191         (align_loops_value): Remove.
8192         (align_jumps_value): Remove.
8193         (align_labels_value): Remove.
8194         (align_functions_value): Remove.
8195         * output.h (label_to_alignment): Change return type to align_flags.
8196         (label_to_max_skip): Remove.
8197         * target.def: Remove loop_align_max_skip, label_align_max_skip,
8198         jump_align_max_skip macros.
8199         * targhooks.h (default_loop_align_max_skip): Remove.
8200         (default_label_align_max_skip): Likewise.
8201         (default_jump_align_max_skip): Likewise.
8202         (default_label_align_after_barrier_max_skip): Remove.
8203         * toplev.c (read_log_maxskip): Use ::normalize function.
8204         (parse_N_M): Remove not used argument and also call ::normalize.
8205         (parse_alignment_opts): Do not pass unused arguments.
8206         * varasm.c (assemble_start_function): Use directly align_functions
8207         instead of removed macros.
8208         * system.h: Do not poison removed macros.
8210 2018-07-17  Jakub Jelinek  <jakub@redhat.com>
8212         PR middle-end/86539
8213         * gimplify.c (gimplify_omp_for): Ensure taskloop firstprivatized init
8214         and cond temporaries don't have reference type if iterator has
8215         pointer type.  For init use &for_pre_body instead of pre_p if
8216         for_pre_body is non-empty.
8218 2018-07-16  Segher Boessenkool  <segher@kernel.crashing.org>
8220         * config/rs6000/rs6000.md (trunc<mode>sf2): Expand truncates of
8221         double-double modes to SFmode directly directly.
8222         (trunc<mode>sf2_fprs): Delete.
8224 2018-07-16  Segher Boessenkool  <segher@kernel.crashing.org>
8226         * config/rs6000/rs6000.c (init_float128_ibm): Use the correct names
8227         for conversions between IFmode and the decimal floating point modes.
8228         (init_float128_ieee): Use the correct names for conversions between
8229         KFmode and the decimal floating point modes.
8231 2018-07-16  Segher Boessenkool  <segher@kernel.crashing.org>
8233         * config/rs6000/rs6000.c (init_float128_ibm): Use more correct names
8234         for the conversions between TDmode and IFmode.
8235         (init_float128_ieee): Use more correct names for the conversions
8236         between TDmode and KFmode.
8238 2018-07-16  Jakub Jelinek  <jakub@redhat.com>
8240         PR tree-optimization/86526
8241         * builtins.c (expand_builtin_memcmp): Formatting fixes.
8242         (inline_expand_builtin_string_cmp): Likewise.
8243         (inline_string_cmp): Likewise.  Use c_readstr instead of
8244         builtin_memcpy_read_str.  Add unit_mode temporary.
8246 2018-07-16  Bernd Edlinger  <bernd.edlinger@hotmail.de>
8248         PR middle-end/86528
8249         * builtins.c (check_access): Bail out if range[0] is no INTEGER_CST.
8250         * expr.c (string_constant): Fix the element size of ARRAY_TYPE.
8252 2018-07-16  Kelvin Nilsen  <kelvin@gcc.gnu.org>
8254         * doc/extend.texi (PowerPC AltiVec Built-in Functions):
8255         Alphabetize prototypes of built-in functions, separating out
8256         built-in functions that are listed in this section but should be
8257         described elsewhere.
8259 2018-07-16  Uros Bizjak  <ubizjak@gmail.com>
8261         PR target/86511
8262         * expmed.c (emit_store_flag): Do not emit setcc followed by a
8263         conditional move when trapping comparison was split to a
8264         non-trapping one (and vice versa).
8266 2018-07-16  Ilya Leoshkevich  <iii@linux.ibm.com>
8268         * config/s390/s390.c (s390_function_profiler): Generate nops
8269         instead of profiler call sequences.
8270         * config/s390/s390.opt: Add the new option.
8272 2018-07-16  Ilya Leoshkevich  <iii@linux.ibm.com>
8274         * config/s390/s390.c (s390_function_profiler): Generate
8275         __mcount_loc section.
8276         * config/s390/s390.opt: Add the new option.
8278 2018-07-16  Ilya Leoshkevich  <iii@linux.ibm.com>
8280         * common.opt: Add the new warning.
8281         * config/s390/s390.c (s390_function_profiler): Emit "brasl
8282         %r0,__fentry__" when -mfentry is specified.
8283         (s390_option_override_internal): Disallow -mfentry for 31-bit
8284         CPUs.
8285         * config/s390/s390.opt: Add the new option.
8287 2018-07-16  Richard Biener  <rguenther@suse.de>
8289         PR lto/86523
8290         * dwarf2out.c (dwarf2out_register_external_die): Assign DIE parents
8291         for function-local FUNCTION_DECL and RESULT_DECL immediately.
8293 2018-07-16  Martin Liska  <mliska@suse.cz>
8295         PR ipa/86529
8296         * ipa-pure-const.c (malloc_candidate_p): Revert ::get
8297         to ::get_create.
8299 2017-07-16  Claudiu Zissulescu  <claziss@synopsys.com>
8301         * config/arc/arcHS.md: Update ARCHS scheduling rules.
8303 2017-07-16  Claudiu Zissulescu  <claziss@synopsys.com>
8305         * config/arc/arc-arch.h (arc_tune_attr): Add new tune parameters
8306         for ARCHS4x.
8307         * config/arc/arc-cpus.def (hs4x): New cpu.
8308         (hs4xd): Likewise.
8309         * config/arc/arc-tables.opt: Regenerate.
8310         * config/arc/arc.c (arc_sched_issue_rate): New function.
8311         (TARGET_SCHED_ISSUE_RATE): Define.
8312         (TARGET_SCHED_EXPOSED_PIPELINE): Likewise.
8313         * config/arc/arc.md (attr type): Add fpu_fuse, fpu_sdiv, fpu_ddiv,
8314         fpu_cvt.
8315         (attr tune): Add ARCHS4x tune values.
8316         (attr tune_dspmpy): Define.
8317         (*tst): Correct instruction type.
8318         * config/arc/arcHS.md: Don't use this automaton for ARCHS4x cpus.
8319         * config/arc/arcHS4x.md: New file.
8320         * config/arc/fpu.md: Update instruction type attributes.
8321         * config/arc/t-multilib: Regenerate.
8323 2018-07-16  Tom de Vries  <tdevries@suse.de>
8325         PR debug/86455
8326         * var-tracking.c (vt_initialize): Fix pre_dec handling.
8328 2018-07-16  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
8330         * config/aarch64/atomics.md (aarch64_store_execlusive<mode>): Add
8331         early clobber.
8333 2018-07-16  Eric Botcazou  <ebotcazou@adacore.com>
8335         PR tree-optimization/86514
8336         * tree-ssa-reassoc.c (init_range_entry) <CASE_CONVERT>: Return for a
8337         conversion to a boolean type from a type with greater precision.
8339 2018-07-16  Tom de Vries  <tdevries@suse.de>
8341         * var-tracking.c (vt_initialize): Print adjusted insn slim if
8342         dump_flags request TDF_SLIM.
8344 2018-07-16  Aldy Hernandez  <aldyh@redhat.com>
8346         * fold-const.c (int_const_binop_1): Abstract...
8347         (wide_int_binop): ...wide int code here.
8348         (poly_int_binop): ...poly int code here.
8349         Abstract the rest of int_const_binop_1 into int_const_binop.
8350         * fold-const.h (wide_int_binop): New.
8351         * tree-vrp.c (vrp_int_const_binop): Call wide_int_binop.
8352         Remove useless PLUS/MINUS_EXPR case.
8353         (zero_nonzero_bits_from_vr): Move wide int code...
8354         (zero_nonzero_bits_from_bounds): ...here.
8355         (extract_range_from_binary_expr_1): Move mask optimization code...
8356         (range_easy_mask_min_max): ...here.
8357         * tree-vrp.h (zero_nonzero_bits_from_bounds): New.
8358         (range_easy_mask_min_max): New.
8360 2018-07-15  Jeff Law  <law@redhat.com>
8362         PR target/85993
8363         * config/sh/sh.c (output_mi_thunk): Remove dead conditional
8364         block.
8366 2018-07-14  Jim Wilson  <jimw@sifive.com>
8368         * config/riscv/linux.h (TARGET_ASM_FILE_END): New.
8370 2018-07-14  Paul Koning  <ni1d@arrl.net>
8372         * config/pdp11/pdp11.c (pdp11_rtx_costs): Bugfixes.
8374 2018-07-13  Jan Hubicka  <hubicka@ucw.cz>
8376         * lto-streamer-out.c (copy_function_or_variable): Dump info about
8377         copying section.
8379 2018-07-13  Bill Schmidt  <wschmidt@linux.ibm.com>
8380             Steve Munroe  <munroesj52@gmail.com>
8382         * config/rs6000/emmintrin.h (_mm_and_si128): New function.
8383         (_mm_andnot_si128): Likewise.
8384         (_mm_or_si128): Likewise.
8385         (_mm_xor_si128): Likewise.
8387 2018-07-13  Qing Zhao  <qing.zhao@oracle.com>
8389         PR middle-end/78809
8390         * builtins.c (expand_builtin_memcmp): Inline the calls first
8391         when result_eq is false.
8392         (expand_builtin_strcmp): Inline the calls first.
8393         (expand_builtin_strncmp): Likewise.
8394         (inline_string_cmp): New routine. Expand a string compare
8395         call by using a sequence of char comparison.
8396         (inline_expand_builtin_string_cmp): New routine. Inline expansion
8397         a call to str(n)cmp/memcmp.
8398         * doc/invoke.texi (--param builtin-string-cmp-inline-length):
8399         New option.
8400         * params.def (BUILTIN_STRING_CMP_INLINE_LENGTH): New.
8402 2018-07-13  Richard Earnshaw  <rearnsha@arm.com>
8404         * config/arm/driver-arm.c: Include arm-native.h.
8405         (host_detect_local_cpu): Use auto-generated data tables.
8406         (vendors, arm_cpu_table): Delete.  Move part information to ...
8407         * config/arm/arm-cpus.in: ... here.
8408         * config/arm/parsecpu.awk (gen_native): New function.
8409         (vendor, part): New CPU fields.
8410         (END): Add support for building the native CPU detection tables.
8411         * config/arm/t-arm (arm-native.h): Add build rule.
8412         (driver-arm.o): Add dependency on arm-native.h.
8414 2018-07-13  Richard Biener  <rguenther@suse.de>
8416         PR middle-end/85974
8417         * match.pd (addr1 - addr2): Allow either of the operand to
8418         have a conversion.
8420 2018-07-13  Tom de Vries  <tdevries@suse.de>
8422         * tree-inline.c (remap_ssa_name): Save and reuse debug exprs generated
8423         in remap_ssa_name.
8425 2018-07-13  Jackson Woodruff  <jackson.woodruff@arm.com>
8427         * config/aarch64/aarch64.c (aarch64_operands_adjust_ok_for_ldpstp): Use
8428         arrays instead of numbered variables.
8430 2018-07-13  Eric Botcazou  <ebotcazou@adacore.com>
8432         * config/sparc/sparc-protos.h (sparc_compute_frame_size): Delete.
8433         * config/sparc/sparc.c (sparc_compute_frame_size): Make static.
8435 2018-07-13  Richard Biener  <rguenther@suse.de>
8437         PR debug/86452
8438         * dwarf2out.c (gen_type_die_with_usage): Use scope_die_for
8439         instead of get_context_die.
8441 2018-07-13  Kugan Vivekanandarajah  <kuganv@linaro.org>
8442             Richard Biener  <rguenther@suse.de>
8444         PR middle-end/86489
8445         * tree-ssa-loop-niter.c (number_of_iterations_popcount): Check
8446         that the loop latch destination where phi is defined.
8448 2018-07-12  Kito Cheng  <kito.cheng@gmail.com>
8450         * config/riscv/riscv.c (enum riscv_privilege_levels): Add UNKNOWN_MODE.
8451         (riscv_expand_epilogue): Add assertion to check interrupt mode.
8452         (riscv_set_current_function): Extract getting interrupt type to new
8453         function.
8454         (riscv_get_interrupt_type): New function.
8455         (riscv_merge_decl_attributes): New function, checking interrupt type is
8456         same.
8457         (TARGET_MERGE_DECL_ATTRIBUTES): Define.
8459 2018-07-12  Paul Koning  <ni1d@arrl.net>
8461         * config/pdp11/pdp11.c (pdp11_output_def): Fix typo in .set
8462         directive.
8464 2018-07-12  Paul Koning  <ni1d@arrl.net>
8466         * doc/rtl.texi (REG_NONNEG): Remove decrement and branch until
8467         zero reference, add doloop_end instead.
8468         * doc/md.texi (decrement_and_branch_until_zero): Remove.
8469         (Looping patterns): Remove decrement_and_branch_until_zero.  Add
8470         detail for doloop_end.
8472 2018-07-12  Martin Sebor  <msebor@redhat.com>
8474         PR c/86453
8475         * attribs.c (decl_attributes): Reject conflicting attributes before
8476         calling attribute handlers.
8478 2018-07-12  Jan Hubicka  <hubicka@ucw.cz>
8480         * dumpfile.c (gcc::dump_manager::get_dump_file_name): Add PART
8481          parameter.
8482         (gcc::dump_manager::get_dump_file_name): likewise.
8483         (dump_begin): Likewise.
8484         * dumpfile.h (dump_begin): Update prototype.
8485         (gcc::dump_manager::get_dump_file_name,
8486         gcc::dump_manager::get_dump_file_name): Update prototype.
8488 2018-07-12  Richard Sandiford  <richard.sandiford@linaro.org>
8490         * internal-fn.h (vectorizable_internal_fn_p): New function.
8491         * tree-vect-slp.c (compatible_calls_p): Likewise.
8492         (vect_build_slp_tree_1): Remove nops argument.  Handle calls
8493         to internal functions.
8494         (vect_build_slp_tree_2): Update call to vect_build_slp_tree_1.
8496 2018-07-12  Richard Sandiford  <richard.sandiford@linaro.org>
8498         * fold-const.h (inverse_conditions_p): Declare.
8499         * fold-const.c (inverse_conditions_p): New function.
8500         * match.pd: Use inverse_conditions_p.  Add folds of view_converts
8501         that test the inverse condition of a conditional internal function.
8502         * internal-fn.h (vectorized_internal_fn_supported_p): Declare.
8503         * internal-fn.c (internal_fn_mask_index): Handle conditional
8504         internal functions.
8505         (vectorized_internal_fn_supported_p): New function.
8506         * tree-if-conv.c: Include internal-fn.h and fold-const.h.
8507         (any_pred_load_store): Replace with...
8508         (need_to_predicate): ...this new variable.
8509         (redundant_ssa_names): New variable.
8510         (ifcvt_can_use_mask_load_store): Move initial checks to...
8511         (ifcvt_can_predicate): ...this new function.  Handle tree codes
8512         for which a conditional internal function exists.
8513         (if_convertible_gimple_assign_stmt_p): Use ifcvt_can_predicate
8514         instead of ifcvt_can_use_mask_load_store.  Update after variable
8515         name change.
8516         (predicate_load_or_store): New function, split out from
8517         predicate_mem_writes.
8518         (check_redundant_cond_expr): New function.
8519         (value_available_p): Likewise.
8520         (predicate_rhs_code): Likewise.
8521         (predicate_mem_writes): Rename to...
8522         (predicate_statements): ...this.  Use predicate_load_or_store
8523         and predicate_rhs_code.
8524         (combine_blocks, tree_if_conversion): Update after above name changes.
8525         (ifcvt_local_dce): Handle redundant_ssa_names.
8526         * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): Handle
8527         general conditional functions.
8528         * tree-vect-stmts.c (vectorizable_call): Likewise.
8530 2018-07-12  Richard Sandiford  <richard.sandiford@linaro.org>
8531             Alan Hayward  <alan.hayward@arm.com>
8532             David Sherwood  <david.sherwood@arm.com>
8534         * internal-fn.h (can_interpret_as_conditional_op_p): Declare.
8535         * internal-fn.c (can_interpret_as_conditional_op_p): New function.
8536         * tree-ssa-math-opts.c (convert_mult_to_fma_1): Handle conditional
8537         plus and minus and convert them into IFN_COND_FMA-based sequences.
8538         (convert_mult_to_fma): Handle conditional plus and minus.
8540 2018-07-12  Richard Sandiford  <richard.sandiford@linaro.org>
8542         * doc/md.texi (cond_fma, cond_fms, cond_fnma, cond_fnms): Document.
8543         * optabs.def (cond_fma_optab, cond_fms_optab, cond_fnma_optab)
8544         (cond_fnms_optab): New optabs.
8545         * internal-fn.def (COND_FMA, COND_FMS, COND_FNMA, COND_FNMS): New
8546         internal functions.
8547         (FMA): Use DEF_INTERNAL_FLT_FN rather than DEF_INTERNAL_FLT_FLOATN_FN.
8548         * internal-fn.h (get_conditional_internal_fn): Declare.
8549         (get_unconditional_internal_fn): Likewise.
8550         * internal-fn.c (cond_ternary_direct): New macro.
8551         (expand_cond_ternary_optab_fn): Likewise.
8552         (direct_cond_ternary_optab_supported_p): Likewise.
8553         (FOR_EACH_COND_FN_PAIR): Likewise.
8554         (get_conditional_internal_fn): New function.
8555         (get_unconditional_internal_fn): Likewise.
8556         * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 5.
8557         (gimple_match_op::gimple_match_op): Add a new overload for 5
8558         operands.
8559         (gimple_match_op::set_op): Likewise.
8560         (gimple_resimplify5): Declare.
8561         * genmatch.c (decision_tree::gen): Generate simplifications for
8562         5 operands.
8563         * gimple-match-head.c (gimple_simplify): Define an overload for
8564         5 operands.  Handle calls with 5 arguments in the top-level overload.
8565         (convert_conditional_op): Handle conversions from unconditional
8566         internal functions to conditional ones.
8567         (gimple_resimplify5): New function.
8568         (build_call_internal): Pass a fifth operand.
8569         (maybe_push_res_to_seq): Likewise.
8570         (try_conditional_simplification): Try converting conditional
8571         internal functions to unconditional internal functions.
8572         Handle 3-operand unconditional forms.
8573         * match.pd (UNCOND_TERNARY, COND_TERNARY): Operator lists.
8574         Define ternary equivalents of the current rules for binary conditional
8575         internal functions.
8576         * config/aarch64/aarch64.c (aarch64_preferred_else_value): Handle
8577         ternary operations.
8578         * config/aarch64/iterators.md (UNSPEC_COND_FMLA, UNSPEC_COND_FMLS)
8579         (UNSPEC_COND_FNMLA, UNSPEC_COND_FNMLS): New unspecs.
8580         (optab): Handle them.
8581         (SVE_COND_FP_TERNARY): New int iterator.
8582         (sve_fmla_op, sve_fmad_op): New int attributes.
8583         * config/aarch64/aarch64-sve.md (cond_<optab><mode>)
8584         (*cond_<optab><mode>_2, *cond_<optab><mode_4)
8585         (*cond_<optab><mode>_any): New SVE_COND_FP_TERNARY patterns.
8587 2018-07-12  Richard Sandiford  <richard.sandiford@linaro.org>
8589         * target.def (preferred_else_value): New target hook.
8590         * doc/tm.texi.in (TARGET_PREFERRED_ELSE_VALUE): New hook.
8591         * doc/tm.texi: Regenerate.
8592         * targhooks.h (default_preferred_else_value): Declare.
8593         * targhooks.c (default_preferred_else_value): New function.
8594         * internal-fn.h (conditional_internal_fn_code): Declare.
8595         * internal-fn.c (FOR_EACH_CODE_MAPPING): New macro.
8596         (get_conditional_internal_fn): Use it.
8597         (conditional_internal_fn_code): New function.
8598         * gimple-match.h (gimple_match_cond): New struct.
8599         (gimple_match_op): Add a cond member function.
8600         (gimple_match_op::gimple_match_op): Update all forms to take a
8601         gimple_match_cond.
8602         * genmatch.c (expr::gen_transform): Use the same condition as res_op
8603         for the suboperation, but don't specify a particular else_value.
8604         * tree-ssa-sccvn.c (vn_nary_simplify, vn_reference_lookup_3)
8605         (visit_nary_op, visit_reference_op_load): Pass
8606         gimple_match_cond::UNCOND to the gimple_match_op constructor.
8607         * gimple-match-head.c: Include tree-eh.h
8608         (convert_conditional_op): New function.
8609         (maybe_resimplify_conditional_op): Likewise.
8610         (gimple_resimplify1): Call maybe_resimplify_conditional_op.
8611         (gimple_resimplify2): Likewise.
8612         (gimple_resimplify3): Likewise.
8613         (gimple_resimplify4): Likewise.
8614         (maybe_push_res_to_seq): Return null for conditional operations.
8615         (try_conditional_simplification): New function.
8616         (gimple_simplify): Call it.  Pass conditions to the gimple_match_op
8617         constructor.
8618         * match.pd: Fold VEC_COND_EXPRs of an IFN_COND_* call to a new
8619         IFN_COND_* call.
8620         * config/aarch64/aarch64.c (aarch64_preferred_else_value): New
8621         function.
8622         (TARGET_PREFERRED_ELSE_VALUE): Redefine.
8624 2018-07-12  Jan Hubicka  <hubicka@ucw.cz>
8626         * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
8627         DECL_FCONTEXT
8628         (hash_tree): Do not hash DECL_FCONTEXT
8629         * tree-streamer-in.c (lto_input_ts_field_decl_tree_pointers):
8630         Do not stream DECL_FCONTEXT.
8631         * tree-streamer-out.c (write_ts_field_decl_tree_pointers): Likewise.
8632         * tree.c (free_lang_data_in_decl): Free DECL_FCONTEXT.
8634 2018-07-12  Richard Biener  <rguenther@suse.de>
8636         PR debug/86462
8637         * dwarf2out.c (gen_block_die): Only output blocks when they have
8638         at least one !DECL_IGNORED_P variable.
8640 2018-07-12  Richard Biener  <rguenther@suse.de>
8642         PR target/84829
8643         * config/gnu-user.h (GNU_USER_TARGET_NO_PTHREADS_LIB_SPEC):
8644         Remove -mieee-fp handling.
8646 2018-07-12  Richard Biener  <rguenther@suse.de>
8648         * tree-ssa-sccvn.c (vn_lookup_simplify_result): Remove bogus
8649         left-over from last patch.
8651 2018-07-12  Jakub Jelinek  <jakub@redhat.com>
8653         PR tree-optimization/86492
8654         * gimple-ssa-store-merging.c
8655         (imm_store_chain_info::coalesce_immediate_stores): Call
8656         check_no_overlap even for the merge_overlapping case.  Formatting fix.
8658 2018-07-12  Richard Biener  <rguenther@suse.de>
8660         PR middle-end/86479
8661         * fold-const.c (fold_binary_op_with_conditional_arg): Do not
8662         move possibly trapping operations into the conditional.
8664 2018-07-12  Richard Biener  <rguenther@suse.de>
8666         * tree-ssa-sccvn.c (mprts_hook_cnt): Remove.
8667         (vn_lookup_simplify_result): Remove recursion limit applied
8668         here.
8669         (vn_nary_build_or_lookup_1): Adjust.
8670         (try_to_simplify): Likewise.
8671         * gimple-match-head.c (gimple_resimplify1): Instead apply one
8672         here.
8673         (gimple_resimplify2): Likewise.
8674         (gimple_resimplify3): Likewise.
8675         (gimple_resimplify4): Likewise.
8677 2018-07-11  Jakub Jelinek  <jakub@redhat.com>
8679         * config/i386/avx512bitalgintrin.h (_mm512_mask_bitshuffle_epi64_mask):
8680         Use __mmask64 type instead of __mmask8 for __M argument.
8681         * config/i386/avx512fintrin.h (_mm512_mask_xor_epi64,
8682         _mm512_maskz_xor_epi64): Use __mmask8 type instead of __mmask16 for
8683         __U argument.
8684         (_mm512_mask_cmpneq_epi64_mask): Use __mmask8 type instead of
8685         __mmask16 for __M argument.
8686         (_mm512_maskz_insertf32x4, _mm512_maskz_inserti32x4,
8687         _mm512_mask_insertf32x4, _mm512_mask_inserti32x4): Cast last argument
8688         to __mmask16 instead of __mmask8.
8689         * config/i386/avx512vlintrin.h (_mm_mask_add_ps, _mm_maskz_add_ps,
8690         _mm256_mask_add_ps, _mm256_maskz_add_ps, _mm_mask_sub_ps,
8691         _mm_maskz_sub_ps, _mm256_mask_sub_ps, _mm256_maskz_sub_ps,
8692         _mm256_maskz_cvtepi32_ps, _mm_maskz_cvtepi32_ps): Use __mmask8 type
8693         instead of __mmask16 for __U argument.
8694         * config/i386/avx512vlbwintrin.h (_mm_mask_cmp_epi8_mask): Use
8695         __mmask16 instead of __mmask8 for __U argument.
8696         (_mm256_mask_cmp_epi8_mask): Use __mmask32 instead of __mmask16 for
8697         __U argument.
8698         (_mm256_cmp_epi8_mask): Use __mmask32 return type instead of
8699         __mmask16.
8700         (_mm_mask_cmp_epu8_mask): Use __mmask16 instead of __mmask8 for __U
8701         argument.
8702         (_mm256_mask_cmp_epu8_mask): Use __mmask32 instead of __mmask16 for
8703         __U argument.
8704         (_mm256_cmp_epu8_mask): Use __mmask32 return type instead of
8705         __mmask16.
8706         (_mm_mask_cmp_epi16_mask): Cast last argument to __mmask8 instead
8707         of __mmask16.
8708         (_mm256_mask_cvtepi8_epi16): Use __mmask16 instead of __mmask32 for
8709         __U argument.
8710         (_mm_mask_cvtepi8_epi16): Use __mmask8 instead of __mmask32 for
8711         __U argument.
8712         (_mm256_mask_cvtepu8_epi16): Use __mmask16 instead of __mmask32 for
8713         __U argument.
8714         (_mm_mask_cvtepu8_epi16): Use __mmask8 instead of __mmask32 for
8715         __U argument.
8716         (_mm256_mask_cmpneq_epu8_mask, _mm256_mask_cmplt_epu8_mask,
8717         _mm256_mask_cmpge_epu8_mask, _mm256_mask_cmple_epu8_mask): Change
8718         return type as well as __M argument type and all casts from __mmask8
8719         to __mmask32.
8720         (_mm256_mask_cmpneq_epu16_mask, _mm256_mask_cmplt_epu16_mask,
8721         _mm256_mask_cmpge_epu16_mask, _mm256_mask_cmple_epu16_mask): Change
8722         return type as well as __M argument type and all casts from __mmask8
8723         to __mmask16.
8724         (_mm256_mask_cmpneq_epi8_mask, _mm256_mask_cmplt_epi8_mask,
8725         _mm256_mask_cmpge_epi8_mask, _mm256_mask_cmple_epi8_mask): Change
8726         return type as well as __M argument type and all casts from __mmask8
8727         to __mmask32.
8728         (_mm256_mask_cmpneq_epi16_mask, _mm256_mask_cmplt_epi16_mask,
8729         _mm256_mask_cmpge_epi16_mask, _mm256_mask_cmple_epi16_mask): Change
8730         return type as well as __M argument type and all casts from __mmask8
8731         to __mmask16.
8732         * config/i386/avx512vbmi2vlintrin.h (_mm_mask_shrdi_epi32,
8733         _mm_mask_shldi_epi32): Cast last argument to __mmask8 instead of
8734         __mmask16.
8736 2018-07-11  Grazvydas Ignotas  <notasas@gmail.com>
8738         * config/i386/avx512bwintrin.h: (_mm512_mask_cmp_epi8_mask,
8739         _mm512_mask_cmp_epu8_mask): Use __mmask64 type instead of __mmask32
8740         for __U argument.
8742 2018-07-11  Paul Koning  <ni1d@arrl.net>
8744         * doc/md.texi (define_subst): Document how multiple occurrences of
8745         the same argument in the replacement pattern are handled.
8747 2018-07-11  Paul Koning  <ni1d@arrl.net>
8749         * doc/extend.texi (Common Variable Attributes): Move "mode" into
8750         alphabetical order.
8751         (Common Type Attributes): Add "mode" attribute.
8753 2018-07-11  Jan Hubicka  <hubicka@ucw.cz>
8755         * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not
8756         stream DECL_ORIGINAL_TYPE.
8757         (DFS::DFS_write_tree_body): Drop hack handling local external decls.
8758         (hash_tree): Do not walk DECL_ORIGINAL_TYPE.
8759         * tree-streamer-in.c (lto_input_ts_decl_non_common_tree_pointers):
8760         Do not walk original type.
8761         * tree-streamer-out.c (streamer_write_chain): Drop hack handling
8762         external decls.
8763         (write_ts_decl_non_common_tree_pointers): Do not stream
8764         DECL_ORIGINAL_TYPE
8765         * tree.c (free_lang_data_in_decl): Clear DECL_ORIGINAL_TYPE.
8766         (find_decls_types_r): Do not walk DEC_ORIGINAL_TYPE.
8768 2018-07-11  Aldy Hernandez  <aldyh@redhat.com>
8770         * tree-ssa-threadupdate.c (thread_through_all_blocks): Do not jump
8771         thread twice from the same starting edge.
8773 2018-07-11  Aldy Hernandez  <aldyh@redhat.com>
8775         * vr-values.c (gimple_stmt_nonzero_p): Abstract common code to...
8776         * gimple.c (gimple_call_nonnull_result_p): ...here...
8777         (gimple_call_nonnull_arg): ...and here.
8778         * gimple.h (gimple_call_nonnull_result_p): New.
8779         (gimple_call_nonnull_arg): New.
8781 2018-07-11  Richard Earnshaw  <rearnsha@arm.com>
8783         * config/arm/arm-cpus.in: Move information from fpu field of each
8784         cpu definition to the isa field.
8785         * config/arm/parsecpu.awk (fpu): Delete match rule.
8786         (gen_comm_data): Don't add bits from the CPU's FPU entry.
8788 2018-07-11  Richard Biener  <rguenther@suse.de>
8790         PR debug/86457
8791         * dwarf2out.c (init_sections_and_labels): Use
8792         output_asm_line_debug_info consistently.
8793         (dwarf2out_early_finish): Likewise.
8794         (dwarf2out_finish): Remove DW_AT_stmt_list from early generated
8795         type units.
8797 2018-07-11  Richard Biener  <rguenther@suse.de>
8799         * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1):
8800         Rework father_bb setting in a way to avoid propagating constants
8801         multiple times on a loop body.
8803 2018-07-10  Mark Wielaard  <mark@klomp.org>
8805         PR debug/86459
8806         * dwarf2out.c (output_macinfo_op): Fix dwarf_FORM typo in gcc_assert.
8808 2018-07-10  Richard Biener  <rguenther@suse.de>
8810         * hash-map.h (hash_map::iterator::operator*): Return
8811         references to key and value.
8813 2018-07-10  Jakub Jelinek  <jakub@redhat.com>
8815         PR c++/86443
8816         * gimplify.c (find_combined_omp_for): Add DATA argument, in addition
8817         to finding the inner OMP_FOR/OMP_SIMD stmt find non-trivial wrappers,
8818         BLOCKs with BLOCK_VARs, OMP_PARALLEL in between, OMP_FOR in between.
8819         (gimplify_omp_for): For composite loops, move outer
8820         OMP_{DISTRIBUTE,TASKLOOP,FOR,PARALLEL} right around innermost
8821         OMP_FOR/OMP_SIMD if there are any non-trivial wrappers.  For class
8822         iterators add any needed clauses.  Allow OMP_FOR_ORIG_DECLS to contain
8823         TREE_LIST for both the original class iterator and the "last" helper
8824         var.  Gimplify OMP_FOR_PRE_BODY before the outermost composite
8825         loop, remember has_decl_expr from outer composite loops for the
8826         innermost OMP_SIMD in TREE_PRIVATE bit on OMP_FOR_INIT.
8828 2018-07-09  Martin Sebor  <msebor@redhat.com>
8830         PR middle-end/77357
8831         PR middle-end/86428
8832         * builtins.c (c_strlen): Avoid out-of-bounds warnings when
8833         accessing implicitly initialized array elements.
8834         * expr.c (string_constant): Handle string initializers of
8835         character arrays within aggregates.
8836         * gimple-fold.c (fold_array_ctor_reference): Add argument.
8837         Store element offset.  As a special case, handle zero size.
8838         (fold_nonarray_ctor_reference): Same.
8839         (fold_ctor_reference): Add argument.  Store subobject offset.
8840         * gimple-fold.h (fold_ctor_reference): Add argument.
8842 2018-07-09  Paul Koning  <ni1d@arrl.net>
8844         * config/pdp11/pdp11.c (pdp11_addr_cost): New function.
8845         (pdp11_insn_cost): New function.
8846         (pdp11_md_asm_adjust): New function.
8847         (TARGET_INVALID_WITHIN_DOLOOP): Define.
8848         (pdp11_rtx_costs): Update to match machine better.
8849         (output_addr_const_pdp11): Correct format mismatch warnings.
8850         * config/pdp11/pdp11.h (SLOW_BYTE_ACCESS): Correct definition.
8851         * config/pdp11/pdp11.md: General change to add base_cost and/or
8852         length attributes for use by new pdp11_insn_cost function.
8853         (MIN_BRANCH): Correct definition.
8854         (MIN_SOB): Ditto.
8855         (doloop_end): Use standard pattern name for looping pattern.
8856         (doloop_end_nocc): New.
8857         (movsf): Add another constraint alternative.
8858         (zero_extendqihi2): Add constraint alternatives for not in place
8859         extend.
8860         (zero_extendhisi2): Remove.
8861         (shift patterns): Add CC handling variants.
8862         (bswaphi2): New.
8863         (bswapsi2): New.
8864         (rothi3): New.
8865         (define_peephole2): New peephole to recognize mov that sets CC for
8866         subsequent test.
8868 2018-07-09  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
8870         * config/sparc/sparc.c (sparc_fold_builtin) <SPARC_BUILTIN_PDIST,
8871         SPARC_BUILTIN_PDISTN>: Adapt for signature change of wi::neg,
8872         wi::add.
8874 2018-07-09  Jakub Jelinek  <jakub@redhat.com>
8876         PR c/86420
8877         * real.c (real_nextafter): Return true if result is denormal.
8879 2018-07-09  Martin Liska  <mliska@suse.cz>
8881         * common.opt: Add back wrongly removed attribute.
8883 2018-07-09  Richard Biener  <rguenther@suse.de>
8885         PR debug/86413
8886         * dwarf2out.c (gen_block_die): For an early generated DIE
8887         always output high/low PC attributes.
8889 2018-07-09  Tom de Vries  <tdevries@suse.de>
8891         * cfgexpand.c (expand_debug_source_expr): Handle VAR_DECL.
8892         * tree-inline.c (remap_ssa_name): Handle default def ssa_name mapping
8893         onto VAR_DECL with abstract origin.
8895 2018-07-07  Jim Wilson  <jimw@sifive.com>
8897         * config/riscv/riscv.c (TARGET_CUSTOM_FUNCTION_DESCRIPTORS): New.
8899 2018-07-07  Tom de Vries  <tdevries@suse.de>
8901         * tree-dfa.c (dump_enumerated_decls): Handle cfun->cfg == NULL.
8903 2018-07-07  Aldy Hernandez  <aldyh@redhat.com>
8905         * tree-vrp.c (vrp_int_const_binop): Change overflow type to
8906         overflow_type.
8907         (combine_bound): Use wide-int overflow calculation instead of
8908         rolling our own.
8909         * calls.c (maybe_warn_alloc_args_overflow): Change overflow type to
8910         overflow_type.
8911         * fold-const.c (int_const_binop_2): Same.
8912         (extract_muldiv_1): Same.
8913         (fold_div_compare): Same.
8914         (fold_abs_const): Same.
8915         * match.pd: Same.
8916         * poly-int.h (add): Same.
8917         (sub): Same.
8918         (neg): Same.
8919         (mul): Same.
8920         * predict.c (predict_iv_comparison): Same.
8921         * profile-count.c (slow_safe_scale_64bit): Same.
8922         * simplify-rtx.c (simplify_const_binary_operation): Same.
8923         * tree-chrec.c (tree_fold_binomial): Same.
8924         * tree-data-ref.c (split_constant_offset_1): Same.
8925         * tree-if-conv.c (idx_within_array_bound): Same.
8926         * tree-scalar-evolution.c (iv_can_overflow_p): Same.
8927         * tree-ssa-phiopt.c (minmax_replacement): Same.
8928         * tree-vect-loop.c (is_nonwrapping_integer_induction): Same.
8929         * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Same.
8930         * vr-values.c (vr_values::adjust_range_with_scev): Same.
8931         * wide-int.cc (wi::add_large): Same.
8932         (wi::mul_internal): Same.
8933         (wi::sub_large): Same.
8934         (wi::divmod_internal): Same.
8935         * wide-int.h: Change overflow type to overflow_type for neg, add,
8936         mul, smul, umul, div_trunc, div_floor, div_ceil, div_round,
8937         mod_trunc, mod_ceil, mod_round, add_large, sub_large,
8938         mul_internal, divmod_internal.
8939         (overflow_type): New enum.
8940         (accumulate_overflow): New.
8942 2018-07-06  Kugan Vivekanandarajah  <kugan.vivekanandarajah@linaro.org>
8944         * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): New.
8945         (tree_ssa_phiopt_worker): Call cond_removal_in_popcount_pattern.
8947 2018-07-06  Kugan Vivekanandarajah  <kuganv@linaro.org>
8949         * tree-ssa-loop-niter.c (number_of_iterations_popcount): If popcount
8950         argument is checked for zero before entering loop, avoid checking again.
8952 2018-07-06  Kugan Vivekanandarajah  <kuganv@linaro.org>
8954         * gimplify.h (generic_expr_could_trap_p): Set as global function.
8955         * gimplify.h (generic_expr_could_trap_p): Likwise.
8956         * tree-scalar-evolution.c (expression_expensive_p): Handle COND_EXPR.
8958 2018-07-06  Jakub Jelinek  <jakub@redhat.com>
8960         PR tree-optimization/86401
8961         * fold-const.c (fold_binary_loc) <case BIT_AND_EXPR>: Move the
8962         ((A & N) + B) & M -> (A + B) & M etc. optimization into ...
8963         (fold_bit_and_mask): ... here.  New helper function for match.pd.
8964         * fold-const.h (fold_bit_and_mask): Declare.
8965         * match.pd (((A & N) + B) & M -> (A + B) & M): New optimization.
8967 2018-07-06  Peter Bergner  <bergner@linux.ibm.com>
8969         PR target/86324
8970         * target.def (translate_mode_attribute): New hook.
8971         * targhooks.h (default_translate_mode_attribute): Declare.
8972         * targhooks.c (default_translate_mode_attribute): New function.
8973         * doc/tm.texi.in (TARGET_TRANSLATE_MODE_ATTRIBUTE): New hook.
8974         * doc/tm.texi: Regenerate.
8975         * config/rs6000/rs6000.c (TARGET_TRANSLATE_MODE_ATTRIBUTE): Define.
8976         (rs6000_translate_mode_attribute): New function.
8978 2018-07-06  Paul Koning  <ni1d@arrl.net>
8980         * doc/md.texi (define_split): Document DONE and FAIL.
8981         (define_peephole2): Ditto.
8983 2018-07-05  Jeff Law  <law@redhat.com>
8985         PR tree-optimization/86010
8986         * tree-ssa-dse.c (compute_trims): More aggressively trim at
8987         both the head and tail of mem* and str* calls.
8989 2018-07-05  Jim Wilson  <jimw@sifive.com>
8991         * config.gcc (riscv*-*-*): When setting xlen, handle riscv-*.
8993 2018-07-05  Indu Bhagat  <indu.bhagat@oracle.com>
8995         * config/aarch64/aarch64-simd.md: correct flags text for
8996         MIN_EXPR replacement.
8998 2018-07-05  James Clarke  <jrtc27@jrtc27.com>
9000         * configure: Regenerated.
9002 2018-07-05  Carl Love  <cel@us.ibm.com>
9004         * config/rs6000/rs6000-c.c: Map ALTIVEC_BUILTIN_VEC_UNPACKH for
9005         float argument to VSX_BUILTIN_DOUBLEH_V4SF.
9006         Map ALTIVEC_BUILTIN_VEC_UNPACKL for float argument to
9007         VSX_BUILTIN_DOUBLEL_V4SF.
9009 2018-07-05  Martin Sebor  <msebor@redhat.com>
9011         PR c++/86400
9012         * tree-ssa-strlen.c (maybe_set_strlen_range): Use type size rather
9013         than its domain to compute its the upper bound of a char array.
9015 2018-07-05  Nathan Sidwell  <nathan@acm.org>
9017         Replace NO_IMPLICIT_EXTERN_C with SYSTEM_IMPLICIT_EXTERN_C.
9018         * doc/cpp.texi: Update comment.
9019         * doc/tm.texi: Rebuilt.
9020         * doc/tm.texi.in (NO_IMPLICIT_EXTERN_C): Replace with ...
9021         (SYSTEM_IMPLICIT_EXTERN_C): ... this, opposite sense.
9022         * doc/extend.texi (Backwards Compatibility): Clarify it is system
9023         headers affected by extern "C".
9024         * system.h: Poison NO_IMPLICIT_EXTERN_C.
9025         * config/alpha/alpha.h, config/arm/uclinux-elf.h,
9026         config/bfin/elf.h, config/cris/cris.h, config/darwin.h,
9027         config/dragonfly.h, config/freebsd.h, config/gnu-user.h,
9028         config/i386/cygming.h, config/i386/djgpp.h, config/i386/nto.h,
9029         config/ia64/hpux.h, config/lm32/lm32.h, config/lm32/uclinux-elf.h,
9030         config/lynx.h, config/mips/elf.h, config/mmix/mmix.h,
9031         config/netbsd.h, config/pa/pa-hpux.h, config/powerpcspe/sysv4.h,
9032         config/riscv/elf.h, config/rs6000/sysv4.h, config/rtems.h,
9033         config/s390/tpf.h, config/sh/newlib.h, config/sol2.h,
9034         config/sparc/openbsd64.h, config/sparc/sp-elf.h,
9035         config/sparc/sp64-elf.h, config/spu/spu.h,
9036         config/stormy16/stormy16.h, config/v850/v850.h,
9037         config/visium/visium.h, config/vx-common.h, config/xtensa/elf.h: Don't
9038         define NO_IMPLICIT_EXTERN_C.
9039         * config/rs6000/aix.h: Set SYSTEM_IMPLICIT_EXTERN_C.
9041 2018-07-05  Tamar Christina  <tamar.christina@arm.com>
9043         PR target/84711
9044         * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
9045         instead of GET_MODE_SIZE when comparing Units.
9047 2018-07-05  Tamar Christina  <tamar.christina@arm.com>
9049         PR target/84711
9050         * rtlanal.c (set_noop_p): Constrain on mode change,
9051         include hard-reg-set.h
9053 2018-07-05  Tamar Christina  <tamar.christina@arm.com>
9055         * config/aarch64/aarch64.c (aarch64_expand_movmem): Fix mode size.
9057 2018-07-05  Jakub Jelinek  <jakub@redhat.com>
9059         Revert
9060         2018-07-04  Maxim Ostapenko  <m.ostapenko@samsung.com>
9062         PR sanitizer/84250
9063         * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
9064         libasan.
9065         * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
9067 2018-07-04  Maxim Ostapenko  <m.ostapenko@samsung.com>
9069         PR sanitizer/84250
9070         * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
9071         libasan.
9072         * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
9074 2018-07-04  Eric Botcazou  <ebotcazou@adacore.com>
9076         PR middle-end/86380
9077         * expmed.c (choose_multiplier): Fix incorrect comparison with mask.
9079 2018-07-04  Aldy Hernandez  <aldyh@redhat.com>
9081         * tree-vrp.c (extract_range_from_binary_expr_1): Initialize
9082         neg_*_op* variables.
9084 2018-07-04  Martin Liska  <mliska@suse.cz>
9086         * tree-switch-conversion.c: Define
9087         max_ratio_for_speed and max_ratio_for_size constants.
9089 2018-07-04  Denys Vlasenko  <dvlasenk@redhat.com>
9090             Martin Liska  <mliska@suse.cz>
9092         PR middle-end/66240
9093         PR target/45996
9094         PR c/84100
9095         * common.opt: Rename align options with 'str_' prefix.
9096         * common/config/i386/i386-common.c (set_malign_value): New
9097         function.
9098         (ix86_handle_option): Use it to set -falign-* options/
9099         * config/aarch64/aarch64-protos.h (struct tune_params): Change
9100         type from int to string.
9101         * config/aarch64/aarch64.c: Update default values from int
9102         to string.
9103         * config/alpha/alpha.c (alpha_override_options_after_change):
9104         Likewise.
9105         * config/arm/arm.c (arm_override_options_after_change_1): Likewise.
9106         * config/i386/dragonfly.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
9107         max skip conditionally.
9108         * config/i386/freebsd.h (SUBALIGN_LOG): New.
9109         (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
9110         max skip conditionally.
9111         * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
9112         max skip conditionally.
9113         * config/i386/gnu-user.h (SUBALIGN_LOG): New.
9114         (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
9115         max skip conditionally.
9116         * config/i386/i386.c (struct ptt): Change type from int to
9117         string.
9118         (ix86_default_align): Set default values.
9119         * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Print
9120         max skip conditionally.
9121         * config/i386/iamcu.h (SUBALIGN_LOG): New.
9122         (ASM_OUTPUT_MAX_SKIP_ALIGN):
9123         * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN):
9124         * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
9125         max skip conditionally.
9126         * config/i386/openbsdelf.h (SUBALIGN_LOG): New.
9127         (ASM_OUTPUT_MAX_SKIP_ALIGN) Print max skip conditionally.:
9128         * config/i386/x86-64.h (SUBALIGN_LOG): New.
9129         (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
9130         max skip conditionally.
9131         (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
9132         * config/ia64/ia64.c (ia64_option_override): Set default values
9133         for alignment options.
9134         * config/m68k/m68k.c: Handle new str_align_* options.
9135         * config/mips/mips.c (mips_set_compression_mode): Change
9136         type of constants.
9137         (mips_option_override): Set default values for options.
9138         * config/powerpcspe/powerpcspe.c (rs6000_option_override_internal):
9139         Likewise.
9140         * config/rs6000/rs6000.c (rs6000_option_override_internal):
9141         Likewise.
9142         * config/rx/rx.c (rx_option_override): Likewise.
9143         * config/rx/rx.h (JUMP_ALIGN): Use align_jumps_log.
9144         (LABEL_ALIGN): Use align_labels_log.
9145         (LOOP_ALIGN): Use align_loops_align.
9146         * config/s390/s390.c (s390_asm_output_function_label): Use new
9147         macros.
9148         * config/sh/sh.c (sh_override_options_after_change):
9149         Change type of constants.
9150         * config/spu/spu.c (spu_sched_init): Likewise.
9151         * config/sparc/sparc.c (sparc_option_override): Set default
9152         values for options.
9153         * config/visium/visium.c (visium_option_override): Likewise.
9154         * config/visium/visium.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Do not
9155         emit p2align format with last argument if it's not needed.
9156         * doc/invoke.texi: Document extended format of -falign-*.
9157         * final.c: Use align_labels alignment.
9158         * flags.h (struct target_flag_state): Change type to use
9159         align_flags.
9160         (struct align_flags_tuple): New.
9161         (struct align_flags): Likewise.
9162         (align_loops_log): Redefine macro to use new types.
9163         (align_loops_max_skip): Redefine macro to use new types.
9164         (align_jumps_log): Redefine macro to use new types.
9165         (align_jumps_max_skip): Redefine macro to use new types.
9166         (align_labels_log): Redefine macro to use new types.
9167         (align_labels_max_skip): Redefine macro to use new types.
9168         (align_functions_log): Redefine macro to use new types.
9169         (align_loops): Redefine macro to use new types.
9170         (align_jumps): Redefine macro to use new types.
9171         (align_labels): Redefine macro to use new types.
9172         (align_functions): Redefine macro to use new types.
9173         (align_functions_max_skip): Redefine macro to use new types.
9174         (align_loops_value): New macro.
9175         (align_jumps_value): New macro.
9176         (align_labels_value): New macro.
9177         (align_functions_value): New macro.
9178         * function.c (invoke_set_current_function_hook): Propagate
9179         alignment values from flags to global variables default in
9180         topleev.h.
9181         * ipa-icf.c (sem_function::equals_wpa): Use
9182         cl_optimization_option_eq instead of memcmp.
9183         * lto-streamer.h (cl_optimization_stream_out): Support streaming
9184         of string types.
9185         (cl_optimization_stream_in): Likewise.
9186         * optc-save-gen.awk: Support strings in cl_optimization.
9187         * opth-gen.awk: Likewise.
9188         * opts.c (finish_options): Remove error checking of invalid
9189         value ranges.
9190         (MAX_CODE_ALIGN): Remove.
9191         (MAX_CODE_ALIGN_VALUE): Likewise.
9192         (parse_and_check_align_values): New function.
9193         (check_alignment_argument): Likewise.
9194         (common_handle_option): Use check_alignment_argument.
9195         * opts.h (parse_and_check_align_values): Declare.
9196         * toplev.c (init_alignments): Remove.
9197         (read_log_maxskip): New.
9198         (parse_N_M): Likewise.
9199         (parse_alignment_opts): Likewise.
9200         (backend_init_target): Remove usage of init_alignments.
9201         * toplev.h (parse_alignment_opts): Declare.
9202         * tree-streamer-in.c (streamer_read_tree_bitfields): Add new
9203         argument.
9204         * tree-streamer-out.c (streamer_write_tree_bitfields): Likewise.
9205         * tree.c (cl_option_hasher::equal): New.
9206         * varasm.c: Use new global macros.
9208 2018-07-04  Denys Vlasenko  <dvlasenk@redhat.com>
9210         * config/i386/dragonfly.h: (ASM_OUTPUT_MAX_SKIP_ALIGN):
9211         Use a simpler align directive also if MAXSKIP = ALIGN-1.
9212         * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
9213         * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
9214         * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
9215         * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
9216         * config/i386/freebsd.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Remove "If N
9217         is large, do at least 8 byte alignment" code. Add SUBALIGN_LOG
9218         define. Use a simpler align directive also if MAXSKIP = ALIGN-1.
9219         * config/i386/gnu-user.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
9220         * config/i386/iamcu.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
9221         * config/i386/openbsdelf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
9222         * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
9224 2018-07-04  Martin Liska  <mliska@suse.cz>
9225             Jonathan Wakely  <jwakely@redhat.com>
9227         * coverage.c: Use correct type.
9228         * doc/invoke.texi: Language correction.
9230 2018-07-03  H.J. Lu  <hongjiu.lu@intel.com>
9232         PR target/85620
9233         * config/i386/i386.c (rest_of_insert_endbranch): Also generate
9234         ENDBRANCH for non-tail call which may return via indirect branch.
9235         * doc/extend.texi: Document indirect_return attribute.
9237 2018-07-03  Martin Sebor  <msebor@redhat.com>
9239         PR tree-optimization/86274
9240         * gimple-ssa-sprintf.c (fmtresult::type_max_digits): Verify
9241         precondition.
9242         (format_floating): Correct handling of infinities and NaNs.
9244 2018-07-03  Martin Sebor  <msebor@redhat.com>
9246         * print-tree.c (print_real_cst): New function.
9247         (print_node_brief): Call it.
9248         (print_node): Ditto.
9250 2018-07-03  Jeff Law  <law@redhat.com>
9252         * config/h8300/h8300.md (logical<mode>3_sn, logical<mode>3): Merge
9253         into a single pattern.
9255         * config/h8300/h8300.md (ors code_iterator): New.
9256         (bsetqi_msx, bnotqi_msx patterns and splitters): Consolidate into
9257         a single pattern and single splitter.
9258         (bsethi_msx, bnothi_msx patterns): Consolidate into a single pattern.
9259         (iorqi3_1, xorqi3_1): Likewise.
9260         (iorqi3, xorqi3 expanders): Similarly.
9262         * config/h8300/h8300.md (movmd_internal_normal): Consolidated with
9263         (movmd_internal) into a single pattern using the P mode iterator.
9264         (movmd splitters): Similarly.
9265         (stpcpy_internal_normal, stpcpy_internal): Similarly for thes patterns.
9266         (movsd splitters): Similarly.
9268         * config/h8300/h8300.c (h8300_insn_length_from_table): Consolidate
9269         ADDB, ADDW and ADDL into a single ADD attribute which selects the
9270         right table based on the size of the operand.
9271         * config/h8300/h8300.md (length_table): Corresponding changes. All
9272         references to "addb", "addw" and "addl" changed to "add".
9273         (btst patterns): Merge two variants into a single pattern.
9274         (tstqi, tsthi): Likewise.
9275         (addhi3_incdec, addsi3_incdec): Likewise.
9276         (subhi3_h8300hs, subsi3_h8300hs): Likewise.
9277         (mulhi3, mulsi3): Likewise.
9278         (udivhi3, udivsi3): Likewise.
9279         (divhi3, divsi3): Likewise.
9280         (andorqi3, andorhi3, andorsi3): Likewise.
9282 2018-07-03  Uros Bizjak  <ubizjak@gmail.com>
9284         PR target/85694
9285         * config/i386/sse.md (uavg<mode>3_ceil): New expander.
9286         (<sse2_avx2>_uavg<mode>3<mask_name>): Simplify expander.
9288 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
9290         PR tree-optimization/85694
9291         * config/aarch64/iterators.md (HADD, RHADD): New int iterators.
9292         (u): Handle UNSPEC_SHADD, UNSPEC_UHADD, UNSPEC_SRHADD and
9293         UNSPEC_URHADD.
9294         * config/aarch64/aarch64-simd.md (<u>avg<mode>3_floor)
9295         (<u>avg<mode>3_ceil): New patterns.
9297 2018-07-03  David Malcolm  <dmalcolm@redhat.com>
9299         * gcc.dg/vect/slp-perm-1.c: Remove "note: " prefix from
9300         scan-tree-dump directive.
9301         * gcc.dg/vect/slp-perm-2.c: Likewise.
9302         * gcc.dg/vect/slp-perm-3.c: Likewise.
9303         * gcc.dg/vect/slp-perm-5.c: Likewise.
9304         * gcc.dg/vect/slp-perm-6.c: Likewise.
9305         * gcc.dg/vect/slp-perm-7.c: Likewise.
9306         * gcc.dg/vect/slp-perm-8.c: Likewise.
9308 2018-07-03  Marek Polacek  <polacek@redhat.com>
9310         PR middle-end/86202
9311         * gimple-fold.c (size_must_be_zero_p): Check the type of the size.
9313 2018-07-03  Richard Biener  <rguenther@suse.de>
9315         PR ipa/86389
9316         * tree-ssa-structalias.c (find_func_clobbers): Properly
9317         handle indirect calls.
9319 2018-07-03  Jeff Law  <law@redhat.com>
9321         * config/h8300/h8300.md (HSI, QHSI, QHSIF): New mode iterators.
9322         (shifts): New code iterator.
9323         (movqi, movhi, movsi, movsf expanders): Consolidate into a single
9324         expander.  Fix HImode handling on H8/SX.
9325         (addqi3, addhi3, addsi3 expanders): Consolidate into a single expander.
9326         (subqi3, subhi3, subsi3 expanders): Likewise.
9327         (andqi3, andhi3, andsi3 expanders): Likewise.
9328         (iorqi3, iorhi3, iorsi3 expanders): Likewise.
9329         (xorqi3, xorhi3, xorsi3 expanders): Likewise.
9330         (negqi2, neghi2, negsi2, negsf2 expanders): Likewise.
9331         (one_cmplqi2, one_cmplhi2, one_cmplsi2): Likewise.
9332         (zero_extendqihi2, zero_extendqisi2): Likewise.
9333         (extendqihi2, extendqisi2): Likewise.
9334         (rotlqi3, rotlhi3, rotlsi3): Likewise.
9335         (neghi2_h8300, negsi2_h8300): Likewise for these patterns.
9336         (rotlqi3_1, rotlhi3_1): Likewise.
9337         (logicalhi3_sn, logicalsi3_sn): Likewise.
9338         (logicalhi3, logicalsi3): Likewise.
9340 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
9342         * tree-vect-patterns.c (vect_recog_rotate_pattern)
9343         (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
9344         (vect_recog_mixed_size_cond_pattern, adjust_bool_pattern_cast)
9345         (adjust_bool_pattern, vect_recog_bool_pattern): Pass the vector
9346         type to append_pattern_def_seq instead of creating a stmt_vec_info
9347         directly.
9348         (build_mask_conversion): Likewise.  Remove vinfo argument.
9349         (vect_add_conversion_to_patterm): Likewise, renaming to...
9350         (vect_add_conversion_to_pattern): ...this.
9351         (vect_recog_mask_conversion_pattern): Update call to
9352         build_mask_conversion.  Pass the vector type to
9353         append_pattern_def_seq here too.
9354         (vect_recog_gather_scatter_pattern): Update call to
9355         vect_add_conversion_to_pattern.
9357 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
9359         * tree-vect-patterns.c (new_pattern_def_seq): Delete.
9360         (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
9361         (vect_recog_widen_op_pattern, vect_recog_over_widening_pattern)
9362         (vect_recog_rotate_pattern, vect_synth_mult_by_constant): Don't set
9363         STMT_VINFO_PATTERN_DEF_SEQ to null here.
9364         (vect_recog_pow_pattern, vect_recog_vector_vector_shift_pattern)
9365         (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern): Use
9366         append_pattern_def_seq instead of new_pattern_def_seq.
9367         (vect_recog_divmod_pattern): Do both of the above.
9368         (vect_pattern_recog_1): Assert that STMT_VINO_PATTERN_DEF_SEQ
9369         is null.
9371 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
9373         * tree-vect-patterns.c (vect_recog_dot_prod_pattern):
9374         (vect_recog_sad_pattern, vect_recog_widen_op_pattern)
9375         (vect_recog_widen_mult_pattern, vect_recog_pow_pattern):
9376         (vect_recog_widen_sum_pattern, vect_recog_over_widening_pattern)
9377         (vect_recog_average_pattern, vect_recog_cast_forwprop_pattern)
9378         (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
9379         (vect_recog_vector_vector_shift_pattern, vect_synth_mult_by_constant)
9380         (vect_recog_mult_pattern, vect_recog_divmod_pattern)
9381         (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
9382         (vect_recog_mask_conversion_pattern): Replace vec<gimple *>
9383         parameter with a single stmt_vec_info.
9384         (vect_recog_func_ptr): Likewise.
9385         (vect_recog_gather_scatter_pattern): Likewise, folding in...
9386         (vect_try_gather_scatter_pattern): ...this.
9387         (vect_pattern_recog_1): Remove stmts_to_replace and just pass
9388         the stmt_vec_info of the statement to be matched.  Don't clear
9389         STMT_VINFO_RELATED_STMT.
9390         (vect_pattern_recog): Update call accordingly.
9392 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
9394         PR tree-optimization/85694
9395         * doc/md.texi (avgM3_floor, uavgM3_floor, avgM3_ceil)
9396         (uavgM3_ceil): Document new optabs.
9397         * doc/sourcebuild.texi (vect_avg_qi): Document new target selector.
9398         * internal-fn.def (IFN_AVG_FLOOR, IFN_AVG_CEIL): New internal
9399         functions.
9400         * optabs.def (savg_floor_optab, uavg_floor_optab, savg_ceil_optab)
9401         (savg_ceil_optab): New optabs.
9402         * tree-vect-patterns.c (vect_recog_average_pattern): New function.
9403         (vect_vect_recog_func_ptrs): Add it.
9404         * tree-vect-stmts.c (vectorizable_call): Get the type of the zero
9405         constant directly from the associated lhs.
9407 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
9409         * tree-vect-patterns.c (vect_split_statement): New function.
9410         (vect_convert_input): Use it to try to split an existing cast.
9412 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
9414         * poly-int.h (print_hex): New function.
9415         * dumpfile.h (dump_dec, dump_hex): Declare.
9416         * dumpfile.c (dump_dec, dump_hex): New poly_wide_int functions.
9417         * tree-vectorizer.h (_stmt_vec_info): Add min_output_precision,
9418         min_input_precision, operation_precision and operation_sign.
9419         * tree-vect-patterns.c (vect_get_range_info): New function.
9420         (vect_same_loop_or_bb_p, vect_single_imm_use)
9421         (vect_operation_fits_smaller_type): Delete.
9422         (vect_look_through_possible_promotion): Add an optional
9423         single_use_p parameter.
9424         (vect_recog_over_widening_pattern): Rewrite to use new
9425         stmt_vec_info infomration.  Handle one operation at a time.
9426         (vect_recog_cast_forwprop_pattern, vect_narrowable_type_p)
9427         (vect_truncatable_operation_p, vect_set_operation_type)
9428         (vect_set_min_input_precision): New functions.
9429         (vect_determine_min_output_precision_1): Likewise.
9430         (vect_determine_min_output_precision): Likewise.
9431         (vect_determine_precisions_from_range): Likewise.
9432         (vect_determine_precisions_from_users): Likewise.
9433         (vect_determine_stmt_precisions, vect_determine_precisions): Likewise.
9434         (vect_vect_recog_func_ptrs): Put over_widening first.
9435         Add cast_forwprop.
9436         (vect_pattern_recog): Call vect_determine_precisions.
9438 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
9440         * tree-vect-patterns.c (vect_mark_pattern_stmts): Remove pattern
9441         statements that have been replaced by further pattern statements.
9442         (vect_pattern_recog_1): Clear STMT_VINFO_PATTERN_DEF_SEQ on failure.
9444 2018-07-03  Richard Biener  <rguenther@suse.de>
9446         * tree-vect-stmts.c (vect_is_simple_use): Consolidate dumping,
9447         always set *dt.  Dump vectype in vectype overload.
9448         * dumpfile.h (dump_gimple_expr): New function.
9449         (dump_gimple_expr_loc): Likewise.
9450         * dumpfile.c (dump_gimple_expr): New function.
9451         (dump_gimple_expr_loc): Likewise.
9453 2018-07-02  Jeff Law  <law@redhat.com>
9455         * config/h8300/h8300.md (movqi_h8300, movqi_h8300hs): Consolidate
9456         the H8/300, H8/300H and H8/S variants into a single pattern.
9457         (movhi_h8300, movqi_h8300hs): Similarly.
9458         (pushqi_h8300hs, pushhi_h8300hs): Consolidate into a single pattern.
9459         (QHI mode iterator): New.
9461         * config/h8300/h8300.md: Remove trailing whitespace.
9463 2018-07-02  Jim Wilson  <jimw@sifive.com>
9465         * config/riscv/riscv.c (riscv_expand_epilogue): Use emit_jump_insn
9466         instead of emit_insn for interrupt returns.
9467         * config/riscv/riscv.md (riscv_met): Add (return) to rtl.
9468         (riscv_sret, riscv_uret): Likewise.
9470 2018-07-02  David Malcolm  <dmalcolm@redhat.com>
9472         * pretty-print.c (selftest::test_pp_format): Move save and restore
9473         of quotes to class auto_fix_quotes, and add an instance.
9474         * selftest.c: Include "intl.h".
9475         (selftest::auto_fix_quotes::auto_fix_quotes): New ctor.
9476         (selftest::auto_fix_quotes::~auto_fix_quotes): New dtor.
9477         * selftest.h (selftest::auto_fix_quotes): New class.
9479 2018-07-02  Richard Henderson  <richard.henderson@linaro.org>
9481         * config/aarch64/aarch64-protos.h, config/aarch64/aarch64.c
9482         (aarch64_sve_prepare_conditional_op): Remove.
9483         * config/aarch64/aarch64-sve.md (cond_<SVE_INT_BINARY><SVE_I>):
9484         Allow aarch64_simd_reg_or_zero as select operand; remove
9485         the aarch64_sve_prepare_conditional_op call.
9486         (cond_<SVE_INT_BINARY_SD><SVE_SDI>): Likewise.
9487         (cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
9488         (*cond_<SVE_INT_BINARY><SVE_I>_z): New pattern.
9489         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_z): New pattern.
9490         (*cond_<SVE_COND_FP_BINARY><SVE_F>_z): New pattern.
9491         (*cond_<SVE_INT_BINARY><SVE_I>_any): New pattern.
9492         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_any): New pattern.
9493         (*cond_<SVE_COND_FP_BINARY><SVE_F>_any): New pattern
9494         and a splitters to match all of the *_any patterns.
9495         * config/aarch64/predicates.md (aarch64_sve_any_binary_operator): New.
9497         * config/aarch64/iterators.md (SVE_INT_BINARY_REV): Remove.
9498         (SVE_COND_FP_BINARY_REV): Remove.
9499         (sve_int_op_rev, sve_fp_op_rev): New.
9500         * config/aarch64/aarch64-sve.md (*cond_<SVE_INT_BINARY><SVE_I>_0): New.
9501         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_0): New.
9502         (*cond_<SVE_COND_FP_BINARY><SVE_F>_0): New.
9503         (*cond_<SVE_INT_BINARY><SVE_I>_2): Rename, add movprfx alternative.
9504         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_2): Similarly.
9505         (*cond_<SVE_COND_FP_BINARY><SVE_F>_2): Similarly.
9506         (*cond_<SVE_INT_BINARY><SVE_I>_3): Similarly; use sve_int_op_rev.
9507         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_3): Similarly.
9508         (*cond_<SVE_COND_FP_BINARY><SVE_F>_3): Similarly; use sve_fp_op_rev.
9510         * config/aarch64/aarch64-sve.md (cond_<SVE_COND_FP_BINARY><SVE_F>):
9511         Remove match_dup 1 from the inner unspec.
9512         (*cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
9514         * config/aarch64/aarch64.md (movprfx): New attr.
9515         (length): Default movprfx to 8.
9516         * config/aarch64/aarch64-sve.md (*mul<SVE_I>3): Add movprfx alt.
9517         (*madd<SVE_I>, *msub<SVE_I): Likewise.
9518         (*<su>mul<SVE_I>3_highpart): Likewise.
9519         (*<SVE_INT_BINARY_SD><SVE_SDI>3): Likewise.
9520         (*v<ASHIFT><SVE_I>3): Likewise.
9521         (*<su><MAXMIN><SVE_I>3): Likewise.
9522         (*<su><MAXMIN><SVE_F>3): Likewise.
9523         (*fma<SVE_F>4, *fnma<SVE_F>4): Likewise.
9524         (*fms<SVE_F>4, *fnms<SVE_F>4): Likewise.
9525         (*div<SVE_F>4): Likewise.
9527 2018-07-02  Richard Sandiford  <richard.sandiford@arm.com>
9529         * tree-vect-patterns.c (vect_recog_widen_shift_pattern): Fix typo
9530         in dump string.
9532 2018-07-02  Richard Biener  <rguenther@suse.de>
9534         PR tree-optimization/86363
9535         * tree-ssa-sccvn.c (vn_reference_lookup_3): Check the
9536         memset argument refers to a non-variable address.
9538 2018-07-02  Aldy Hernandez  <aldyh@redhat.com>
9540         * tree-vrp.c (extract_range_from_binary_expr_1): Abstract a lot of the
9541         {PLUS,MINUS}_EXPR code to...
9542         (adjust_symbolic_bound): ...here,
9543         (combine_bound): ...here,
9544         (set_value_range_with_overflow): ...and here.
9546 2018-07-02  Aldy Hernandez  <aldyh@redhat.com>
9548         * tree-vrp.c (extract_range_from_unary_expr): Abstract ABS_EXPR
9549         code...
9550         (extract_range_from_abs_expr): ...here.
9552 2018-07-02  Eric Botcazou  <ebotcazou@adacore.com>
9554         * config/i386/i386.c (ix86_finalize_stack_frame_flags): Do not overrule
9555         -fno-omit-frame-pointer when not optimizing.
9557 2018-07-02  Martin Liska  <mliska@suse.cz>
9559         PR ipa/86279
9560         * ipa-pure-const.c (malloc_candidate_p): Revert usage of ::get.
9561         (propagate_nothrow): Likewise.
9563 2018-07-02  Martin Liska  <mliska@suse.cz>
9565         PR ipa/86323
9566         * ipa-inline.c (early_inliner): Revert wrongly added ::get call.
9568 2018-07-02  David Malcolm  <dmalcolm@redhat.com>
9570         * dumpfile.c (dump_generic_expr_loc): Undo removal of this
9571         function in r262149, changing "loc" param from source_location to
9572         const dump_location_t &.
9573         * dumpfile.h (dump_generic_expr_loc): Undo removal of this
9574         declaration, as above.
9576 2018-07-01  Paul Koning  <ni1d@arrl.net>
9578         * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
9579         -munit-asm, -mgnu-asm, -mdec-asm.
9580         * config/pdp11/pdp11-protos.h (pdp11_gen_int_label): New.
9581         (pdp11_output_labelref): New.
9582         (pdp11_output_def): New.
9583         (pdp11_output_addr_vec_elt): New.
9584         * config/pdp11/pdp11.c: Use tab between opcode and operands.  Use
9585         %# and %@ format codes.
9586         (pdp11_option_override): New.
9587         (TARGET_ASM_FILE_START_FILE_DIRECTIVE): Define.
9588         (pdp11_output_ident): New.
9589         (pdp11_asm_named_section): New.
9590         (pdp11_asm_init_sections): New.
9591         (pdp11_file_start): New.
9592         (pdp11_file_end): New.
9593         (output_ascii): Use .ascii/.asciz for -mdec-asm.
9594         (pdp11_asm_print_operand): Update %# and %$ for -mdec-asm.  Add
9595         %o, like %c but octal.
9596         (pdp11_option_override): New.
9597         * config/pdp11/pdp11.h (TEXT_SECTION_ASM_OP): Update for
9598         -mdec-asm.
9599         (DATA_SECTION_ASM_OP): Ditto.
9600         (READONLY_DATA_SECTION_ASM_OP): New.
9601         (IS_ASM_LOGICAL_LINE_SEPARATOR): New.
9602         (ASM_GENERATE_INTERNAL_LABEL): Use new function.
9603         (ASM_OUTPUT_LABELREF): Ditto.
9604         (ASM_OUTPUT_DEF): Ditto.
9605         (ASM_OUTPUT_EXTERNAL): New.
9606         (ASM_OUTPUT_SOURCE_FILENAME): New.
9607         (ASM_OUTPUT_ADDR_VEC_ELT): Use new function.
9608         (ASM_OUTPUT_SKIP): Update for -mdec-asm.
9609         * config/pdp11/pdp11.md: Use tab between opcode and operands.  Use
9610         %# and %@ format codes.
9611         * config/pdp11/pdp11.opt (mgnu-asm): New.
9612         (mdec-asm): Conflicts with -mgnu-asm and -munix-asm.
9613         (munix-asm): Conflicts with -mdec-asm and -mgnu-asm.
9614         * doc/invoke.txt (PDP-11 Options): Add -mgnu-asm.
9616 2018-07-01  Aldy Hernandez  <aldyh@redhat.com>
9618         * tree-ssa-threadupdate.c (mark_threaded_blocks): Avoid
9619         dereferencing path[] beyond its length.
9620         (debug_path): New.
9621         (debug_all_paths): New.
9622         (rewire_first_differing_edge): New.
9623         (adjust_paths_after_duplication): New.
9624         (duplicate_thread_path): Call adjust_paths_after_duplication.
9625         Add new argument.
9626         (thread_through_all_blocks): Add new argument to
9627         duplicate_thread_path.
9629 2018-06-30  Jim Wilson  <jimw@sifive.com>
9631         * config/riscv/predicates.md (p2m1_shift_operand): New.
9632         (high_mask_shift_operand): New.
9633         * config/riscv/riscv.md (lshrsi3_zero_extend_3+1): New combiner
9634         pattern using p2m1_shift_operand.
9635         (lshsi3_zero_extend_3+2): New combiner pattern using
9636         high_mask_shift_operand.
9638 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
9640         * tree-vect-patterns.c (vect_get_external_def_edge): New function,
9641         split out from...
9642         (vect_recog_rotate_pattern): ...here.
9643         (vect_convert_input): Try to insert casts of invariants in the
9644         preheader.
9645         * tree-vect-loop-manip.c (vect_loop_versioning): Don't require the
9646         preheader to be empty.
9648 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
9650         * tree-vect-patterns.c (append_pattern_def_seq): Take an optional
9651         vector type.  If given, install it in the new statement's
9652         STMT_VINFO_VECTYPE.
9653         (vect_element_precision): New function.
9654         (vect_unpromoted_value): New struct.
9655         (vect_unpromoted_value::vect_unpromoted_value): New function.
9656         (vect_unpromoted_value::set_op): Likewise.
9657         (vect_look_through_possible_promotion): Likewise.
9658         (vect_joust_widened_integer, vect_joust_widened_type): Likewise.
9659         (vect_widened_op_tree, vect_convert_input): Likewise.
9660         (vect_convert_inputs, vect_convert_output): Likewise.
9661         (vect_recog_dot_prod_pattern): Use vect_look_through_possible_promotion
9662         to handle the optional cast of the multiplication result and
9663         vect_widened_op_tree to detect the widened multiplication itself.
9664         Do not require the input and output of promotion casts to have
9665         the same sign, but base the signedness of the operation on the
9666         input rather than the result.  If the pattern includes two
9667         promotions, check that those promotions have the same sign.
9668         Do not restrict the MULT_EXPR handling to a double-width result;
9669         handle quadruple-width results and wider.  Use vect_convert_inputs
9670         to convert the inputs to the common type.
9671         (vect_recog_sad_pattern):  Use vect_look_through_possible_promotion
9672         to handle the optional cast of the ABS result.  Also allow a sign
9673         change or a sign extension between the ABS and MINUS.
9674         Use vect_widened_op_tree to detect the widened subtraction and use
9675         vect_convert_inputs to convert the inputs to the common type.
9676         (vect_handle_widen_op_by_const): Delete.
9677         (vect_recog_widen_op_pattern): New function.
9678         (vect_recog_widen_mult_pattern): Use it.
9679         (vect_recog_widen_shift_pattern): Likewise.
9680         (vect_recog_widen_sum_pattern): Use
9681         vect_look_through_possible_promotion to handle the promoted
9682         PLUS_EXPR operand.
9684 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
9686         * gimple-iterator.c (gsi_for_stmt): Add a new overload that takes
9687         the containing gimple_seq *.
9688         * gimple-iterator.h (gsi_for_stmt): Declare it.
9689         * tree-vect-patterns.c (vect_recog_dot_prod_pattern)
9690         (vect_recog_sad_pattern, vect_recog_widen_sum_pattern)
9691         (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
9692         (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
9693         (vect_recog_mask_conversion_pattern): Remove STMT_VINFO_IN_PATTERN_P
9694         checks.
9695         (vect_init_pattern_stmt, vect_set_pattern_stmt): New functions,
9696         split out from...
9697         (vect_mark_pattern_stmts): ...here.  Handle cases in which the
9698         statement being replaced is part of an existing pattern
9699         definition sequence, inserting the new pattern statements before
9700         the original one.
9701         (vect_pattern_recog_1): Don't return a bool.  If the statement
9702         is already part of a pattern, instead apply pattern matching
9703         to the pattern definition statements.  Don't clear the
9704         STMT_VINFO_RELATED_STMT if is_pattern_stmt_p.
9705         (vect_pattern_recog): Don't break after the first match;
9706         continue processing the pattern definition statements instead.
9707         Don't bail out for STMT_VINFO_IN_PATTERN_P here.
9709 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
9711         * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
9712         (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
9713         (vect_recog_widen_sum_pattern): Use it.
9715 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
9717         * tree-vect-loop.c (vectorizable_reduction): Assert that the
9718         phi is not a pattern statement and has not been replaced by
9719         a pattern statement.
9720         * tree-vect-patterns.c (type_conversion_p): Don't check
9721         STMT_VINFO_IN_PATTERN_P.
9722         (vect_recog_vector_vector_shift_pattern): Likewise.
9723         (vect_recog_dot_prod_pattern): Expect vect_is_simple_use to return
9724         the pattern statement rather than the original statement; check
9725         directly for a WIDEN_MULT_EXPR here.
9726         * tree-vect-slp.c (vect_get_and_check_slp_defs): Expect
9727         vect_is_simple_use to return the pattern statement rather
9728         than the original statement; use is_pattern_stmt_p to check
9729         for such a pattern statement.
9730         * tree-vect-stmts.c (process_use): Expect vect_is_simple_use
9731         to return the pattern statement rather than the original statement;
9732         don't do the same transformation here.
9733         (vect_is_simple_use): If the defining statement has been replaced
9734         by a pattern statement, return the pattern statement instead.
9735         Remove the corresponding (local) transformation from the vectype
9736         overload.
9738 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
9740         * tree-vectorizer.h (vect_is_simple_use): Move the gimple ** to the
9741         end and default to null.
9742         * tree-vect-loop.c (vect_create_epilog_for_reduction)
9743         (vectorizable_reduction): Update calls accordingly, dropping the
9744         gimple ** argument if the passed-back statement isn't needed.
9745         * tree-vect-patterns.c (vect_get_internal_def, type_conversion_p)
9746         (vect_recog_rotate_pattern): Likewise.
9747         (vect_recog_mask_conversion_pattern): Likewise.
9748         * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
9749         (vect_mask_constant_operand_p): Likewise.
9750         * tree-vect-stmts.c (is_simple_and_all_uses_invariant, process_use):
9751         (vect_model_simple_cost, vect_get_vec_def_for_operand): Likewise.
9752         (get_group_load_store_type, get_load_store_type): Likewise.
9753         (vect_check_load_store_mask, vect_check_store_rhs): Likewise.
9754         (vectorizable_call, vectorizable_simd_clone_call): Likewise.
9755         (vectorizable_conversion, vectorizable_assignment): Likewise.
9756         (vectorizable_shift, vectorizable_operation): Likewise.
9757         (vectorizable_store, vect_is_simple_cond): Likewise.
9758         (vectorizable_condition, vectorizable_comparison): Likewise.
9759         (get_same_sized_vectype, vect_get_mask_type_for_stmt): Likewise.
9760         (vect_is_simple_use): Rename the def_stmt argument to def_stmt_out
9761         and move it to the end.  Cope with null def_stmt_outs.
9763 2018-06-30  Bernd Edlinger  <bernd.edlinger@hotmail.de>
9765         * Makefile.in (FLAGS_TO_PASS): Add RANLIB_FOR_TARGET.
9767 2018-06-29  Jeff Law  <law@redhat.com>
9769         * config/v850/v850.c (v850_legitimate_address_p): Handle large
9770         displacements for TARGET_V850E2V3 and newer.
9771         (TARGET_LRA_P): Remove.  Defaults to LRA now.
9772         * config/v850/v850.md (sign23byte_load): Remove.
9773         (unsign23byte_load, sign23hword_load, unsign23hword_load): Likewise.
9774         (23word_load, 23byte_store, 23hword_store, 23word_store): Likewise.
9776 2018-06-29  Martin Liska  <mliska@suse.cz>
9778         PR lto/85759
9779         * coverage.c (coverage_init): Mangle full path name.
9780         * doc/invoke.texi: Document the change.
9781         * gcov-io.c (mangle_path): New.
9782         * gcov-io.h (mangle_path): Likewise.
9783         * gcov.c (mangle_name): Use mangle_path for path mangling.
9785 2018-06-29  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
9787         * config/arm/arm.c (output_move_double): Don't allow STRD instructions
9788         if starting source register is not even.
9790 2018-06-29  Martin Liska  <mliska@suse.cz>
9792         PR tree-optimization/86263
9793         * tree-switch-conversion.c (switch_decision_tree::try_switch_expansion):
9794         Make edge redirection.
9796 2018-06-29  David Malcolm  <dmalcolm@redhat.com>
9798         * dumpfile.c (dump_loc): Add indentation based on scope depth.
9799         (dump_scope_depth): New variable.
9800         (get_dump_scope_depth): New function.
9801         (dump_begin_scope): New function.
9802         (dump_end_scope): New function.
9803         * dumpfile.h (get_dump_scope_depth): New declaration.
9804         (dump_begin_scope): New declaration.
9805         (dump_end_scope): New declaration.
9806         (class auto_dump_scope): New class.
9807         (AUTO_DUMP_SCOPE): New macro.
9808         * tree-vectorizer.h (DUMP_VECT_SCOPE): Reimplement in terms of
9809         AUTO_DUMP_SCOPE.
9811 2018-06-29  Richard Biener  <rguenther@suse.de>
9813         * tree-vect-data-refs.c (vect_analyze_data_ref_dependences): Assert
9814         compute_all_dependences succeeds.
9815         * tree-vect-loop.c (vect_get_datarefs_in_loop): Fail early if we
9816         exceed --param loop-max-datarefs-for-datadeps.
9818 2018-06-29  Jakub Jelinek  <jakub@redhat.com>
9820         * config/rs6000/t-rs6000: Append rs6000-modes.h to TM_H.
9822 2018-06-28  Uros Bizjak  <ubizjak@gmail.com>
9824         PR target/86348
9825         * config/i386/sse.md (*vec_extractv4si_0_zext_sse4): Use
9826         alternative 0 in preferred_for_speed attribute.
9828 2018-06-28  Paul Koning  <ni1d@arrl.net>
9830         * config/pdp11/pdp11-protos.h (pdp11_shift_length): New function.
9831         * config/pdp11/pdp11.c (pdp11_shift_length): New function.
9832         * config/pdp11/pdp11.h (ADJUST_INSN_LENGTH): Remove.
9833         * config/pdp11/pdp11.md: Correct "length" attribute calculation
9834         for shift insn patterns.
9836 2018-06-28  David Malcolm  <dmalcolm@redhat.com>
9838         * cgraph.c (cgraph_node::get_body): Replace assignments to
9839         "dump_file" with calls to set_dump_file.
9840         * dumpfile.c (alt_dump_file): Make static, and group with...
9841         (alt_flags): ...this definition.
9842         (dumps_are_enabled): New variable.
9843         (refresh_dumps_are_enabled): New function.
9844         (set_dump_file): New function.
9845         (set_alt_dump_file): New function.
9846         (gcc::dump_manager::dump_start): Replace assignments to
9847         "dump_file" and "alt_dump_file" with calls to set_dump_file and
9848         set_alt_dump_file.
9849         (gcc::dump_manager::dump_finish): Likewise.
9850         * dumpfile.h (alt_dump_file): Delete decl.
9851         (dumps_are_enabled): New variable decl.
9852         (set_dump_file): New function decl.
9853         (dump_enabled_p): Rewrite in terms of new "dumps_are_enabled"
9854         global.
9855         * tree-nested.c (lower_nested_functions): Replace assignments to
9856         "dump_file" with calls to set_dump_file.
9858 2018-06-28  Eric Botcazou  <ebotcazou@adacore.com>
9860         * tree-cfg.c (verify_gimple_in_cfg): Call verify_location on the
9861         goto_locus of each outgoing edge of each basic block.
9863 2018-06-28  Richard Biener  <rguenther@suse.de>
9865         * dwarf2out.c (decl_scope_table): Remove.
9866         (push_decl_scope): Likewise.
9867         (pop_decl_scope): Likewise.
9868         (gen_type_die_for_member): Do not call push/pop_decl_scope.
9869         (gen_struct_or_union_type_die): Likewise.
9870         (gen_tagged_type_die): Likewise.
9871         (dwarf2out_init): Do not initialize decl_scope_table.
9872         (dwarf2out_c_finalize): Do not free it.
9874 2018-06-28  Richard Biener  <rguenther@suse.de>
9876         * dwarf2out.c (gen_subprogram_die): Use is_unit_die when
9877         deciding whether to not re-use a DIE.
9879 2018-06-28  Richard Biener  <rguenther@suse.de>
9881         * dwarf2out.c (gen_subprogram_die): Always re-use DIEs with an
9882         DW_AT_abstract_origin attribute.
9884 2018-06-28  Martin Liska  <mliska@suse.cz>
9886         * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
9887         Use newly introduced constants.
9888         * tree-switch-conversion.h (struct jump_table_cluster):
9889         Define max_ratio_for_size and max_ratio_for_speed.
9891 2018-06-28  Martin Liska  <mliska@suse.cz>
9893         * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
9894         Add new checking assert to catch invalid state.
9895         (jump_table_cluster::can_be_handled): Handle single case
9896         clusters.
9897         (jump_table_cluster::is_beneficial): Bail out for such case.
9898         (bit_test_cluster::find_bit_tests):
9899         Add new checking assert to catch invalid state.
9900         (bit_test_cluster::can_be_handled): Handle single case
9901         clusters.
9902         (bit_test_cluster::is_beneficial): Bail out for such case.
9903         (switch_decision_tree::analyze_switch_statement):
9904         Fix comment.
9906 2018-06-28  Martin Liska  <mliska@suse.cz>
9908         * common.opt: Introduce -completion option.
9909         * gcc.c (driver_handle_option): Handle it.
9910         (driver::main): Print completions if completion
9911         is set.
9912         * opt-suggestions.c (option_proposer::get_completions):
9913         New function.
9914         (option_proposer::suggest_completion): Likewise.
9915         (option_proposer::find_param_completions): Likewise.
9916         (verify_autocompletions): Likewise.
9917         (test_completion_valid_options): Likewise.
9918         (test_completion_valid_params): Likewise.
9919         (in_completion_p): Likewise.
9920         (empty_completion_p): Likewise.
9921         (test_completion_partial_match): Likewise.
9922         (test_completion_garbage): Likewise.
9923         (opt_proposer_c_tests): Likewise.
9924         * opt-suggestions.h: Declare new functions.
9925         * opts.c (common_handle_option): Handle OPT__completion_.
9926         * selftest-run-tests.c (selftest::run_tests): Add
9927         opt_proposer_c_tests.
9928         * selftest.c (assert_str_startswith): New.
9929         * selftest.h (assert_str_startswith): Likewise.
9930         (opt_proposer_c_tests): New.
9931         (ASSERT_STR_STARTSWITH): Likewise.
9933 2018-06-28  Martin Liska  <mliska@suse.cz>
9935         * Makefile.in: Add opt-suggestions.o.
9936         * gcc-main.c: Include opt-suggestions.h.
9937         * gcc.c (driver::driver): Likewise.
9938         (driver::~driver): Remove m_option_suggestions.
9939         (driver::build_option_suggestions): Moved to option_proposer.
9940         (driver::suggest_option): Likewise.
9941         (driver::handle_unrecognized_options): Use option_proposer.
9942         * gcc.h (class driver): Add new memver m_option_proposer.
9943         * opt-suggestions.c: New file.
9944         * opt-suggestions.h: New file.
9946 2018-06-28  Martin Liska  <mliska@suse.cz>
9948         * vec.h (class auto_string_vec): New (moved from auto_argvec).
9949         (auto_string_vec::~auto_string_vec): Likewise.
9951 2018-06-28  Eric Botcazou  <ebotcazou@adacore.com>
9953         * tree-inline.h (struct copy_body_data): Move remapping_type_depth and
9954         prevent_decl_creation_for_types fields up and add reset_location field.
9955         * tree-inline.c (remap_gimple_stmt): Force input_location on the new
9956         statement if id->reset_location is true.
9957         (copy_edges_for_bb): Do not set goto_locus on the new edges if
9958         id->reset_location is true.
9959         (copy_phis_for_bb): Force input_location on the arguments if
9960         id->reset_location is true.
9961         (expand_call_inline): Set id->reset_location if DECL_IGNORED_P
9962         is set on the function to be inlined.
9964 2018-06-27  Stephan Bergmann  <sbergman@redhat.com>
9966         * doc/invoke.texi (Debugging Options): Add -gsplit-dwarf.
9968 2018-06-27  Dimitar Dimitrov  <dimitar@dinux.eu>
9970         * lra-eliminations.c (update_reg_eliminate): Mark all spanning hard
9971         registers for Pmode.
9972         * lra-lives.c (check_pseudos_live_through_calls): Mark all spanning
9973         hard registers for the clobbered pseudo.
9975 2018-06-27  Paul Koning  <ni1d@arrl.net>
9977         * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
9978         mutually exclusive options.
9979         * config/pdp11/constraints.md (h): New constraint.
9980         (O): Update definition to match shift code generation.
9981         (D): New constraint.
9982         * config/pdp11/pdp11-modes.def (CCNZ): Define mode.
9983         (CCFP): Remove.
9984         * config/pdp11/pdp11-protos.h (int_no_side_effect_operand): New
9985         function.
9986         (output_jump): Change arguments.
9987         (pdp11_fixed_cc_regs): New function.
9988         (pdp11_cc_mode): Ditto.
9989         (pdp11_expand_shift): Ditto.
9990         (pdp11_assemble_shift): Ditto.
9991         (pdp11_small_shift): Ditto.
9992         (pdp11_branch_cost): Remove.
9993         * config/pdp11/pdp11.c (pdp11_assemble_integer): Remove comments
9994         from output.
9995         (pdp11_register_move_cost): Update for CC registers.
9996         (pdp11_rtx_costs): Add case for LSHIFTRT.
9997         (pdp11_output_jump): Add CCNZ mode conditional branches.
9998         (notice_update_cc_on_set): Remove.
9999         (pdp11_cc_mode): New function.
10000         (simple_memory_operand): Correct pre/post decrement case.
10001         (no_side_effect_operand): New function.
10002         (pdp11_regno_reg_class): Add CC_REGS class.
10003         (pdp11_fixed_cc_regs): New function.
10004         (pdp11_small_shift): New function.
10005         (pdp11_expand_shift): New function to expand shift insns.
10006         (pdp11_assemble_shift): New function to output shifts.
10007         (pdp11_branch_cost): Remove.
10008         (pdp11_modes_tieable_p): Make QI/HI modes tieable.
10009         * config/pdp11/pdp11.h (SIZE_TYPE): Ensure 16-bit type.
10010         (WCHAR_TYPE): Ditto.
10011         (PTRDIFF_TYPE): Ditto.
10012         (ADJUST_INSN_LENGTH): New macro.
10013         (FIXED_REGISTERS): Add CC registers.
10014         (CALL_USED_REGISTERS): Ditto.
10015         (reg_class): Ditto.
10016         (REG_CLASS_NAMES): Ditto.
10017         (REG_CLASS_CONTENTS): Ditto.
10018         (SELECT_CC_MODE): Use new function.
10019         (TARGET_FLAGS_REGNUM): New macro.
10020         (TARGET_FIXED_CONDITION_CODE_REGS): Ditto.
10021         (cc0_reg_rtx): Remove.
10022         (CC_STATUS_MDEP): Remove.
10023         (CC_STATUS_MDEFP_INIT): Remove.
10024         (CC_IN_FPU): Remove.
10025         (NOTICE_UPDATE_CC): Remove.
10026         (REGISTER_NAMES): Add CC registers.
10027         (BRANCH_COST): Change to constant 1.
10028         * config/pdp11/pdp11.md: Rewrite for CCmode condition code
10029         handling.
10030         * config/pdp11/pdp11.opt (mbcopy): Remove.
10031         (mbcopy-builtin): Remove.
10032         (mbranch-cheap): Remove.
10033         (mbranch-expensive): Remove.
10034         * config/pdp11/predicates.md (expand_shift_operand): Update to
10035         match shift code generation.
10036         (ccnz_operator): New predicate.
10037         * doc/invoke.texi (PDP-11 Options): Remove deleted options
10038         -mbcopy, -mbcopy-builtin, -mbranch-cheap, -mbranch-expensive.
10039         Remove non-existent option -mabshi, -mno-abshi.  Document mutually
10040         exclusive options.
10041         * doc/md.texi (PDP-11): Document new D and h constraints.  Update
10042         description of O constraint.
10044 2018-06-27  Jeff Law  <law@redhat.com>
10045             Austin Law  <austinklaw@gmail.com>
10047         * config/v850/v850.md (addsi3_set_flags): New pattern.
10048         (subsi3_set_flags, negsi2_set_flags, andsi3_set_flags): Likewise.
10049         (iorsi3_set_flags, xorsi3_set_flags, one_cmplsi2_set_flags): Likewise.
10050         (zero_extendhisi2_v850_set_flags): Likewise.
10051         (zero_extendqisi2_v850_set_flags): Likewise.
10052         (ashlsi3_set_flags, ashlsi3_v850e2_set_flags): Likewise.
10053         (lshrsi3_set_flags, lshrsi3_v850e2_set_flags): Likewise.
10054         (ashrsi3_set_flags, ashrsi3_v850e2_set_flags): Likewise.
10056         * config/v850/v850-protos.h (notice_update_cc): Remove.
10057         * config/v850/v850.c (v850_compare_op0, v850_compare_op1): Remove.
10058         (v850_print_operand): Handle 'D' and "d".
10059         (v850_select_cc_mode): Remove ATTRIBUTE_UNUSED for last argument.
10060         Add handling of arithmetic/logical operations compared against zero.
10061         (v850_gen_float_compare): Remove ATTRIBUTE_UNUSED for last argument.
10062         Do not look at v850_compare_op, instead get mode from last argument.
10063         (v850_gen_compare): Remove
10064         (increment_stack): Use addsi3_clobber_flags to avoid splitting failure
10065         after reload for prologue insns.
10066         (expand_prologue): Account for CLOBBER of CC_REGNUM in various
10067         patterns.
10068         (construct_save_jarl): Likewise.
10069         (TARGET_FLAGS_REGNUM): Define.
10070         * config/v850/v850.h (v850_compare_op0, v850_compare_op1): Remove.
10071         (NOTICE_UPDATE_CC): Remove.
10072         * config/v850/v850.md (v850_tst1): Use (reg:CCZ CC_REGNUM) rather
10073         than cc0.  Conditionalize on reload_completed.
10074         (cmpsi_insn, setfcc_insn): Likewise.
10075         (tst1 splitter): Turn into define_and_split which sets the flags
10076         after reload.
10077         (cstoresi4, cbranchsf4, cbranchdf4, cbranchsi4_insn): Likewise.
10078         (cbranchsi4, branch_normal, branch_invert): Do not expose cc0 here.
10079         (cstoresf4, cstoredf4): Clobber the flags.
10080         (cmpsi, cmpsf, cmpdf): Remove expanders.
10081         (setf_insn): Remove pattern.
10082         (addsi3): Turn into define_and_split which clobbers the flags after
10083         reload and a suitable pattern (addsi3_clobber_flags) for use after
10084         reload.
10085         (subsi3, negsi2, andsi3, iorsi3, xorsi3, one_cmplsi2) Likewise.
10086         (ashlsi3, ashlsi3_v850e2, lshrsi3, lsh4si3_v850e2): Likewise.
10087         (ashrsi3, ashrsi3_v850e2): Likewise.
10088         (bins): Clobber the flags.
10089         (movsicc_normal_cc, movsicc_normal, movsicc_tst1): Likewise.
10090         (movsicc_tst1_revesed, sasf, swap and rotate patterns): Likewise.
10091         (fix_loop_counter, call_internal_short, call_internal_long): Likewise.
10092         (call_value_internal_short, call_value_internal_long): Likewise.
10093         (callt_save_interrupt, callt_return_interrupt): Likewise.
10094         (save_interrupt, return_interrupt): Likewise.
10095         (callt_save_all_interrupt, save_all_interrupt): Likewise.
10096         (_save_all_interrupt, callt_restore_all_interrupt): Likewise.
10097         (restore_all_interrupt, _restore_all_interrupt): Likewise.
10098         (All FP comparisons): Only allow after reload has completed.
10099         (trfsr): Likewise.
10100         (divh, divhu): Tweak output template.
10101         (branch_z_normal, branch_z_invert): Remove
10102         (branch_nz_normal, branch_nz_invert): Likewise.
10103         (extendhisi_insn, extendqisi_insn): Do not clobber flags.
10105         * config/v850/v850-modes.def (CCZ, CCNZ): Add new modes.
10106         * config/v850/v850.c (notice_update_cc): Remove.
10107         * config/v850/v850.h  (CC_OVERFLOW_UNUSABLE): Remove
10108         (CC_NO_CARRY): Likewise.
10109         (NOTICE_UPDATE_CC): Define to nothing.
10110         * config/v850/v850.md: Remove block comment on cc0 handling
10111         Remove "cc" attribute from all patterns.  Remove cc_status handling
10112         from all patterns.  Minor formatting fixes.
10114 2018-06-27  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
10116         * config/aarch64/aarch64-cores.def (cortex-a76): New entry.
10117         (cortex-a76.cortex-a55): Likewise.
10118         * config/aarch64/aarch64-tune.md: Regenerate.
10119         * doc/invoke.texi (AArch64 Options): Document cortex-a76 and
10120         cortex-a76.cortex-a55.
10122 2018-06-27  Jeff Law  <law@redhat.com>
10124         * config/v850/t-v850 (MULTILIB_OPTIONS): Remove 8byte-align.
10125         (MULTILIB_DIRNAMES): Similarly.
10127 2018-06-27  Eric Botcazou  <ebotcazou@adacore.com>
10129         * gimple.h (gimple_return_retbnd): Delete.
10130         (gimple_return_set_retbnd): Likewise.
10131         * cgraphunit.c (cgraph_node::expand_thunk): Remove call to
10132         gimple_return_set_retbnd.
10133         * gimple-pretty-print.c (dump_gimple_return): Remove call to
10134         gimple_return_retbnd and adjust.
10135         * tree-inline.h (struct copy_body_data): Remove retbnd field.
10136         * tree-inline.c (remap_gimple_stmt): Remove handling of retbnd.
10137         Explicitly return NULL in a couple more cases.  Move assertion
10138         on debug statements and remove unreachable code.
10139         (reset_debug_binding): Do not test id->retbnd.
10140         (expand_call_inline): Do not set it.
10142 2018-06-27  Rasmus Villemoes  <rv@rasmusvillemoes.dk>
10144         * configure.ac: Add --disable-gcov option.
10145         * configure: Regenerate.
10146         * Makefile.in: Honour @enable_gcov@.
10147         * doc/install.texi: Document --disable-gcov.
10149 2018-06-27  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
10151         * config/arm/arm-cpus.in (cortex-a76): New entry.
10152         (cortex-a76.cortex-a55): Likewise.
10153         * config/arm/arm-tables.opt: Regenerate.
10154         * config/arm/arm-tune.md: Likewise.
10155         * config/arm/driver-arm.c (arm_cpu_table): Add Cortex-A76 entry.
10156         * doc/invoke.texi (ARM Options): Document cortex-a76 and
10157         cortex-a76.cortex-a55.
10159 2018-06-27  Tamar Christina  <tamar.christina@arm.com>
10161         PR target/85769
10162         * config/aarch64/aarch64.md (*movhf_aarch64): Add dup v0.4h pattern.
10164 2018-06-27  Siddhesh Poyarekar  <siddhesh@sourceware.org>
10166         * config/aarch64/aarch64.h (CALL_USE_REGISTERS): Fix obsolete
10167         comment.
10168         (EPILOGUE_USES): Likewise.
10170 2018-06-26  Eric Botcazou  <ebotcazou@adacore.com>
10172         * tree-inline.c (remap_location): New function extracted from...
10173         (copy_edges_for_bb): Add ID parameter.  Remap goto_locus.
10174         (copy_phis_for_bb): ...here.  Call remap_location.
10175         (copy_cfg_body): Adjust call to copy_edges_for_bb.
10177 2018-06-26  Aaron Sawdey  <acsawdey@linux.ibm.com>
10179         * config/rs6000/rs6000-string.c (expand_block_clear): Don't use
10180         unaligned vsx for 16B memset.
10182 2018-06-26  Segher Boessenkool  <segher@kernel.crashing.org>
10184         PR target/86285
10185         * config/rs6000/rs6000.c (rs6000_init_builtins): Do not set
10186         ieee128_float_type_node to long_double_type_node unless
10187         TARGET_LONG_DOUBLE_128 is set.
10189 2018-06-26  David Malcolm  <dmalcolm@redhat.com>
10191         * cfgloop.c (get_loop_location): Convert return type from
10192         location_t to dump_user_location_t, replacing INSN_LOCATION lookups
10193         by implicit construction from rtx_insn *, and using
10194         dump_user_location_t::from_function_decl for the fallback case.
10195         * cfgloop.h (get_loop_location): Convert return type from
10196         location_t to dump_user_location_t.
10197         * cgraphunit.c (walk_polymorphic_call_targets): Update call to
10198         dump_printf_loc to pass in a dump_location_t rather than a
10199         location_t, via the gimple stmt.
10200         * coverage.c (get_coverage_counts): Update calls to
10201         dump_printf_loc to pass in dump_location_t rather than a
10202         location_t.
10203         * doc/optinfo.texi (Dump types): Convert example of
10204         dump_printf_loc from taking "locus" to taking "insn".  Update
10205         description of the "_loc" calls to cover dump_location_t.
10206         * dumpfile.c: Include "backend.h", "gimple.h", "rtl.h", and
10207         "selftest.h".
10208         (dump_user_location_t::dump_user_location_t): New constructors,
10209         from gimple *stmt and rtx_insn *.
10210         (dump_user_location_t::from_function_decl): New function.
10211         (dump_loc): Make static.
10212         (dump_gimple_stmt_loc): Convert param "loc" from location_t to
10213         const dump_location_t &.
10214         (dump_generic_expr_loc): Delete.
10215         (dump_printf_loc): Convert param "loc" from location_t to
10216         const dump_location_t &.
10217         (selftest::test_impl_location): New function.
10218         (selftest::dumpfile_c_tests): New function.
10219         * dumpfile.h: Include "profile-count.h".
10220         (class dump_user_location_t): New class.
10221         (struct dump_impl_location_t): New struct.
10222         (class dump_location_t): New class.
10223         (dump_printf_loc): Convert 2nd param from source_location to
10224         const dump_location_t &.
10225         (dump_generic_expr_loc): Delete.
10226         (dump_gimple_stmt_loc): Convert 2nd param from source_location to
10227         const dump_location_t &.
10228         * gimple-fold.c (fold_gimple_assign): Update call to
10229         dump_printf_loc to pass in a dump_location_t rather than a
10230         location_t, via the gimple stmt.
10231         (gimple_fold_call): Likewise.
10232         * gimple-loop-interchange.cc
10233         (loop_cand::analyze_iloop_reduction_var): Update for change to
10234         check_reduction_path.
10235         (tree_loop_interchange::interchange): Update for change to
10236         find_loop_location.
10237         * graphite-isl-ast-to-gimple.c (scop_to_isl_ast): Update for
10238         change in return-type of find_loop_location.
10239         (graphite_regenerate_ast_isl): Likewise.
10240         * graphite-optimize-isl.c (optimize_isl): Likewise.
10241         * graphite.c (graphite_transform_loops): Likewise.
10242         * ipa-devirt.c (ipa_devirt): Update call to dump_printf_loc to
10243         pass in a dump_location_t rather than a location_t, via the
10244         gimple stmt.
10245         * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
10246         * ipa.c (walk_polymorphic_call_targets): Likewise.
10247         * loop-unroll.c (report_unroll): Convert "locus" param from
10248         location_t to dump_location_t.
10249         (decide_unrolling): Update for change to get_loop_location's
10250         return type.
10251         * omp-grid.c (struct grid_prop): Convert field "target_loc" from
10252         location_t to dump_user_location_t.
10253         (grid_find_single_omp_among_assignments_1): Updates calls to
10254         dump_printf_loc to pass in a dump_location_t rather than a
10255         location_t, via the gimple stmt.
10256         (grid_parallel_clauses_gridifiable): Convert "tloc" from
10257         location_t to dump_location_t.  Updates calls to dump_printf_loc
10258         to pass in a dump_location_t rather than a location_t, via the
10259         gimple stmt.
10260         (grid_inner_loop_gridifiable_p): Likewise.
10261         (grid_dist_follows_simple_pattern): Likewise.
10262         (grid_gfor_follows_tiling_pattern): Likewise.
10263         (grid_target_follows_gridifiable_pattern): Likewise.
10264         (grid_attempt_target_gridification): Convert initialization
10265         of local "grid" from memset to zero-initialization; FIXME: does
10266         this require C++11?  Update call to dump_printf_loc to pass in a
10267         optinfo_location rather than a location_t, via the gimple stmt.
10268         * profile.c (read_profile_edge_counts): Updates call to
10269         dump_printf_loc to pass in a dump_location_t rather than a
10270         location_t
10271         (compute_branch_probabilities): Likewise.
10272         * selftest-run-tests.c (selftest::run_tests): Call
10273         dumpfile_c_tests.
10274         * selftest.h (dumpfile_c_tests): New decl.
10275         * tree-loop-distribution.c (pass_loop_distribution::execute):
10276         Update for change in return type of find_loop_location.
10277         * tree-parloops.c (parallelize_loops): Likewise.
10278         * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Convert
10279         "locus" from location_t to dump_user_location_t.
10280         (canonicalize_loop_induction_variables): Likewise.
10281         * tree-ssa-loop-ivopts.c (tree_ssa_iv_optimize_loop): Update
10282         for change in return type of find_loop_location.
10283         * tree-ssa-loop-niter.c (number_of_iterations_exit): Update call
10284         to dump_printf_loc to pass in a dump_location_t rather than a
10285         location_t, via the stmt.
10286         * tree-ssa-sccvn.c (eliminate_dom_walker::before_dom_children):
10287         Likewise.
10288         * tree-vect-loop-manip.c (find_loop_location): Convert return
10289         type from source_location to dump_user_location_t.
10290         (vect_do_peeling): Update for above change.
10291         (vect_loop_versioning): Update for change in type of
10292         vect_location.
10293         * tree-vect-loop.c (check_reduction_path): Convert "loc" param
10294         from location_t to dump_user_location_t.
10295         (vect_estimate_min_profitable_iters): Update for change in type
10296         of vect_location.
10297         * tree-vect-slp.c (vect_print_slp_tree): Convert param "loc" from
10298         location_t to dump_location_t.
10299         (vect_slp_bb): Update for change in type of vect_location.
10300         * tree-vectorizer.c (vect_location): Convert from source_location
10301         to dump_user_location_t.
10302         (try_vectorize_loop_1): Update for change in vect_location's type.
10303         (vectorize_loops): Likewise.
10304         (increase_alignment): Likewise.
10305         * tree-vectorizer.h (vect_location): Convert from source_location
10306         to dump_user_location_t.
10307         (find_loop_location): Convert return type from source_location to
10308         dump_user_location_t.
10309         (check_reduction_path): Convert 1st param from location_t to
10310         dump_user_location_t.
10311         * value-prof.c (check_counter): Update call to dump_printf_loc to
10312         pass in a dump_user_location_t rather than a location_t; update
10313         call to error_at for change in type of "locus".
10314         (check_ic_target): Update call to dump_printf_loc to
10315         pass in a dump_user_location_t rather than a location_t, via the
10316         call_stmt.
10318 2018-06-26  Robin Dapp  <rdapp@linux.vnet.ibm.com>
10320         * config/s390/s390.h (enum processor_flags): Do not use
10321         default tune parameter when -march was specified.
10323 2018-06-26  Jakub Jelinek  <jakub@redhat.com>
10325         PR target/86314
10326         * config/i386/i386.md (setcc + movzbl to xor + setcc peephole2s):
10327         Check reg_overlap_mentioned_p in addition to reg_set_p with the same
10328         operands.
10330 2018-06-26  Richard Biener  <rguenther@suse.de>
10332         PR tree-optimization/86287
10333         PR bootstrap/86316
10334         * tree-vect-loop.c (vect_transform_loop_stmt): Fix read-after-free.
10335         (vect_analyze_loop): Initialize n_stmts.
10337 2018-06-26  Richard Biener  <rguenther@suse.de>
10339         PR middle-end/86271
10340         * fold-const.c (fold_convertible_p): Pointer extension
10341         isn't valid.
10343 2018-06-26  Alexandre Oliva <aoliva@redhat.com>
10345         PR debug/86064
10346         * dwarf2out.c (loc_list_has_views): Adjust comments.
10347         (dw_loc_list): Split single cross-partition range with
10348         nonzero locview.
10350 2018-06-25  Jeff Law  <law@redhat.com>
10352         * common/config/v850/v850-common.c (TARGET_DEFAULT_TARGET_FLAGS): Turn
10353         on -mbig-switch by default.
10355         * config/v850/predicates.md (const_float_1_operand): Fix match_code
10356         test.
10357         (const_float_0_operand): Remove unused predicate.
10358         * config/v850/v850.md (define_constants): Remove UNSPEC_LOOP.
10359         (define_c_enum unspec): Add LOOP, RCP and RSQRT constants.
10360         (recipsf2): New expander.  Original pattern now called
10361         (recipsf2_insn).
10362         (recipdf2, recipdf2_insn): Similarly.
10363         (rsqrtsf2, rsqrtsf2_insn): Similarly
10364         (rsqrtdf2, rsqrtdf2_insn): Similarly
10366 2018-06-26  Gerald Pfeifer  <gerald@pfeifer.com>
10368         * ginclude/stddef.h: Remove an obsolete comment on FreeBSD 5.
10369         Simplify logic for FreeBSD (twice).
10371 2018-06-25  Martin Sebor  <msebor@redhat.com>
10373         PR tree-optimization/86204
10374         * tree-ssa-strlen.c (handle_builtin_strlen): Avoid storing
10375         a strnlen result if it's less than the length of the string.
10377 2018-06-25  Martin Sebor  <msebor@redhat.com>
10379         PR tree-optimization/85700
10380         * gimple-fold.c (gimple_fold_builtin_strncat): Adjust comment.
10381         * tree-ssa-strlen.c (is_strlen_related_p): Handle integer subtraction.
10382         (maybe_diag_stxncpy_trunc): Distinguish strncat from strncpy.
10384 2018-06-25  Martin Sebor  <msebor@redhat.com>
10386         * doc/extend.texi (Zero-length arrays): Update and clarify.
10388 2018-06-25  Michael Meissner  <meissner@linux.ibm.com>
10390         * config.gcc (powerpc64le*): Revert January 16th, 2018 patch that
10391         added IEEE/IBM long double multilib support on PowerPC little
10392         endian Linux systems.
10393         * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Likewise.
10394         (MULTILIB_DEFAULTS): Likewise.
10395         * config/rs6000/rs6000.c (rs6000_option_override_internal):
10396         Likewise.
10397         * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Likewise.
10398         * config/rs6000/t-ldouble-linux64le-ibm: Delete, no longer used.
10399         * config/rs6000/t-ldouble-linux64le-ieee: Delete, no longer used.
10401 2018-06-25  Alexander Monakov  <amonakov@ispras.ru>
10403         PR middle-end/86311
10404         * sort.cc (REORDER_23): Avoid memcpy with same destination and source.
10405         (REORDER_45): Likewise.
10407 2018-06-25  Jeff Law  <law@redhat.com>
10409         * config/v850/v850.md (divmodhi4): Make sure to sign extend the
10410         dividend to 32 bits.  Adjust length.
10411         (udivmodhi4): Cleanup output template.  Fix length.
10413 2018-06-25  Carl Love  <cel@us.ibm.com>
10415         * config/rs6000/vsx.md: Change word selector to prefered location.
10417 2018-06-25  Richard Biener  <rguenther@suse.de>
10419         PR tree-optimization/86304
10420         * tree-vectorizer.c (vectorize_loops): Walk over new possibly
10421         epilogue-if-converted loops as well.
10423 2018-06-25  Jan Hubicka  <hubicka@ucw.cz>
10425         * lto-section-out.c (lto_begin_section): Do not print section
10426         name for noaddr and unnumbered dumps.
10428 2018-06-25  Richard Biener  <rguenther@suse.de>
10430         * tree-vectorizer.h (struct vec_info_shared): New structure
10431         with parts split out from struct vec_info and loop_nest from
10432         struct _loop_vec_info.
10433         (struct vec_info): Adjust accordingly.
10434         (struct _loop_vec_info): Likewise.
10435         (LOOP_VINFO_LOOP_NEST): Adjust.
10436         (LOOP_VINFO_DATAREFS): Likewise.
10437         (LOOP_VINFO_DDRS): Likewise.
10438         (struct _bb_vec_info): Likewise.
10439         (BB_VINFO_DATAREFS): Likewise.
10440         (BB_VINFO_DDRS): Likewise.
10441         (struct _stmt_vec_info): Add dr_aux member.
10442         (DR_VECT_AUX): Adjust to refer to member of DR_STMTs vinfo.
10443         (DR_MISALIGNMENT_UNINITIALIZED): New.
10444         (set_dr_misalignment): Adjust.
10445         (dr_misalignment): Assert misalign isn't DR_MISALIGNMENT_UNINITIALIZED.
10446         (vect_analyze_loop): Adjust prototype.
10447         (vect_analyze_loop_form): Likewise.
10448         * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
10449         Compute dependences lazily.
10450         (vect_record_base_alignments): Use shared datarefs/ddrs.
10451         (vect_verify_datarefs_alignment): Likewise.
10452         (vect_analyze_data_refs_alignment): Likewise.
10453         (vect_analyze_data_ref_accesses): Likewise.
10454         (vect_analyze_data_refs): Likewise.
10455         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Add
10456         constructor parameter for shared part.
10457         (vect_analyze_loop_form): Pass in shared part and adjust.
10458         (vect_analyze_loop_2): Pass in storage for the number of
10459         stmts.  Move loop nest finding to the caller.  Compute
10460         datarefs lazily.
10461         (vect_analyze_loop): Pass in shared part.
10462         (vect_transform_loop): Verify shared datarefs are unchanged.
10463         * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Add
10464         constructor parameter for shared part.
10465         (vect_slp_analyze_bb_1): Pass in shared part and adjust.
10466         (vect_slp_bb): Verify shared datarefs are unchanged before
10467         transform.
10468         * tree-vect-stmts.c (ensure_base_align): Adjust for DR_AUX
10469         change.
10470         (new_stmt_vec_info): Initialize DR_AUX misalignment to
10471         DR_MISALIGNMENT_UNINITIALIZED.
10472         * tree-vectorizer.c (vec_info::vec_info): Add constructor
10473         parameter for shared part.
10474         (vec_info::~vec_info): Adjust.
10475         (vec_info_shared::vec_info_shared): New.
10476         (vec_info_shared::~vec_info_shared): Likewise.
10477         (vec_info_shared::save_datarefs): Likewise.
10478         (vec_info_shared::check_datarefs): Likewise.
10479         (try_vectorize_loop_1): Construct shared part live for analyses
10480         of a single loop for multiple vector sizes.
10481         * tree-parloops.c (gather_scalar_reductions): Adjust.
10483 2018-06-25  Richard Biener  <rguenther@suse.de>
10485         * tree-vect-data-refs.c (vect_find_stmt_data_reference): Modify
10486         DR for SIMD lane accesses here and mark DR with (void *)-1 aux.
10487         (vect_analyze_data_refs): Remove similar code from here and
10488         simplify accordingly.
10490 2018-06-25  Richard Biener  <rguenther@suse.de>
10492         * tree-vect-data-refs.c (vect_check_gather_scatter): Fail
10493         for reverse storage order accesses rather than asserting
10494         they cannot happen here.
10496 2018-06-25  Tom de Vries  <tdevries@suse.de>
10498         PR debug/86257
10499         * config/i386/i386.md (define_insn "*tls_global_dynamic_64_<mode>"):
10500         Use data16 instead of .byte for insn prefix.
10502 2018-06-25  Andreas Krebbel  <krebbel@linux.ibm.com>
10504         PR C++/86082
10505         * parser.c (make_char_string_pack): Pass this literal chars
10506         through cpp_interpret_string.
10507         (cp_parser_userdef_numeric_literal): Check the result of
10508         make_char_string_pack.
10510 2018-06-24  Maya Rashish  <coypu@sdf.org>
10512         * ginclude/stddef.h: Simplify conditions around avoiding
10513         re-definition of __size_t.
10515 2018-06-22  Jan Hubicka  <hubicka@ucw.cz>
10517         * lto-streamer-out.c (tree_is_indexable): Make LABEL_DECL nonindexable
10518         unless it is forced or nonlocal; assert that we stream no IMPORTED_DECL.
10520 2018-06-22  Maya Rashish  <coypu@sdf.org>
10522         * doc/invoke.texi (mno-fancy-math-387): Update for changes
10523         made to OpenBSD and NetBSD through the years.
10525 2018-06-22  Kelvin Nilsen  <kelvin@gcc.gnu.org>
10527         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
10528         behavior of vec_pack (vector double, vector double) to match
10529         behavior of vec_float2 (vector double, vector double).
10531 2018-06-22  Olivier Hainque  <hainque@adacore.com>
10533         * gimplify.c (gimplify_function_tree): Prevent creation
10534         of a trampoline for the address of the current function
10535         passed to entry/exit instrumentation hooks.
10537 2018-06-22  Aaron Sawdey  <acsawdey@linux.ibm.com>
10539         PR target/86222
10540         * config/rs6000/rs6000-string.c (expand_strn_compare): Handle -m32
10541         correctly.
10543 2018-06-22  Martin Liska  <mliska@suse.cz>
10545         PR tree-optimization/86263
10546         * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
10547         Bail out if is_enabled is false.
10548         * tree-switch-conversion.h (jump_table_cluster::is_enabled):
10549         New declaration.
10550         (jump_table_cluster::is_enabled): New function.
10552 2018-06-22  Jan Hubicka  <hubicka@ucw.cz>
10554         * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
10555         BINFO_BASE_ACCESSES and BINFO_VPTR_FIELD.
10556         * tree-streamer-in.c (streamer_read_tree_bitfields): Likewise.
10557         (lto_input_ts_binfo_tree_pointers): Likewise.
10558         * tree-streamer-out.c (streamer_write_tree_bitfields,
10559         write_ts_binfo_tree_pointers): Likewise.
10560         * tree.c (free_lang_data_in_binfo): Clear BINFO_VPTR_FIELD.
10562 2018-06-22  Jan Hubicka  <hubicka@ucw.cz>
10564         * tree.c (free_lang_data_in_type): Free all TYPE_VFIELDs.
10566 2018-06-22  Martin Liska  <mliska@suse.cz>
10568         * symbol-summary.h (get): Make it pure and inline move
10569         functionality from ::get function.
10570         (get): Remove and inline into ::get and ::get_create.
10571         (get_create): Move code from ::get function.
10573 2018-06-22  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
10575         PR target/85994
10576         * config/i386/sol2.h (CPP_SPEC): Don't pass -P for
10577         -x assembler-with-cpp.
10579 2018-06-22  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
10581         * config/sol2.h (TARGET_OS_CPP_BUILTINS): Define
10582         _FILE_OFFSET_BITS=64 for C++.
10584 2018-06-21  Michael Meissner  <meissner@linux.ibm.com>
10586         * config/rs6000/rs6000.md (extendtfif2): Add missing 128-bit
10587         conversion insn that shows up when pr85657-3.c is compiled using
10588         IEEE 128-bit long double.
10589         (neg<mode>2_internal): Use the correct mode to check whether the
10590         mode is IBM extended.
10591         * config/rs6000/rs6000.c (init_float128_ieee): Prevent complex
10592         multiply and divide external functions from being created more
10593         than once.
10595 2018-06-21  Eric Botcazou  <ebotcazou@adacore.com>
10597         * cfgrtl.c (fixup_reorder_chain): Do not emit NOPs in DECL_IGNORED_P
10598         functions.
10599         (rtl_merge_blocks): Likewise.  Do not emit a NOP if the location of
10600         the edge can be forwarded.
10601         (cfg_layout_merge_blocks): Likewise.
10603 2018-06-21  Eric Botcazou  <ebotcazou@adacore.com>
10605         * except.c (finish_eh_generation): Commit edge insertions only after
10606         the EH edges have been redirected from post-landing to landing pads.
10608 2018-06-21  Eric Botcazou  <ebotcazou@adacore.com>
10610         * tree-nested.c (get_frame_type): Use create_tmp_var_raw instead of
10611         create_tmp_var_for to create the FRAME decl.
10612         (finalize_nesting_tree_1): Do not unchain the FRAME decl.
10614 2018-06-21  Eric Botcazou  <ebotcazou@adacore.com>
10616         * tree-inline.c (copy_edges_for_bb): Minor tweak.
10617         (maybe_move_debug_stmts_to_successors): Also reset the locus of the
10618         debug statement when resetting its value.
10619         (expand_call_inline): Copy the locus of the call onto the assignment
10620         of the return value, if any.  Use local variable in more cases.
10622 2018-06-21  Martin Liska  <mliska@suse.cz>
10624         * ipa-pure-const.c (propagate_nothrow): Use
10625         funct_state_summaries->get.
10626         (dump_malloc_lattice): Likewise.
10627         (propagate_malloc): Likewise.
10629 2018-06-21  Richard Biener  <rguenther@suse.de>
10631         * lto-streamer-out.c (DFS::DFS_write_tree_body): Update outdated
10632         comment.  Follow BLOCK_ABSTRACT_ORIGIN unconditionally.
10633         * tree-streamer-in.c (lto_input_ts_block_tree_pointers): Update
10634         comment.
10635         * tree-streamer-out.c (write_ts_block_tree_pointers): Stream
10636         BLOCK_ABSTRACT_ORIGIN unconditionally.
10638 2018-06-21  David Malcolm  <dmalcolm@redhat.com>
10640         * ipa-cp.c (ipcp_driver): Set edge_clone_summaries to NULL after
10641         deleting it.
10642         * ipa-reference.c (ipa_reference_c_finalize): Delete
10643         ipa_ref_opt_sum_summaries and set it to NULL.
10645 2018-06-21  Tom de Vries  <tdevries@suse.de>
10647         PR tree-optimization/85859
10648         * tree-ssa-tail-merge.c (stmt_local_def): Copy gimple_is_call
10649         test with comment from bb_no_side_effects_p.
10651 2018-06-21  Richard Biener  <rguenther@suse.de>
10653         PR tree-optimization/86232
10654         * tree-ssa-loop-niter.c (number_of_iterations_popcount): Adjust
10655         max for constant niter.
10657 2018-06-21  Andre Vieira  <andre.simoesdiasvieira@arm.com>
10659         * config/aarch64/aarch64-simd.md
10660         (*aarch64_crypto_aes<aes_op>v16qi_xor_combine): New.
10662 2018-06-21  Andre Vieira  <andre.simoesdiasvieira@arm.com>
10664         * config/aarch64/aarch64-simd.md (aarch64_crypto_aes<aes_op>v16qi):
10665         Make opernads of the unspec commutative.
10667 2018-06-21  Richard Biener  <rguenther@suse.de>
10669         * tree-data-ref.c (dr_step_indicator): Handle NULL DR_STEP.
10670         * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr):
10671         Avoid calling vect_mark_for_runtime_alias_test with gathers or scatters.
10672         (vect_analyze_data_ref_dependence): Re-order checks to deal with
10673         NULL DR_STEP.
10674         (vect_record_base_alignments): Do not record base alignment
10675         for gathers or scatters.
10676         (vect_compute_data_ref_alignment): Drop return value that is always
10677         true.  Bail out early for gathers or scatters.
10678         (vect_enhance_data_refs_alignment): Bail out early for gathers
10679         or scatters.
10680         (vect_find_same_alignment_drs): Likewise.
10681         (vect_analyze_data_refs_alignment): Remove dead code.
10682         (vect_slp_analyze_and_verify_node_alignment): Likewise.
10683         (vect_analyze_data_refs): For possible gathers or scatters do
10684         not create an alternate DR, just check their possible validity
10685         and mark them.  Adjust DECL_NONALIASED handling to not rely
10686         on DR_BASE_ADDRESS.
10687         * tree-vect-loop-manip.c (vect_update_inits_of_drs): Do not
10688         update inits of gathers or scatters.
10689         * tree-vect-patterns.c (vect_recog_mask_conversion_pattern):
10690         Also copy gather/scatter flag to pattern vinfo.
10692 2018-06-20  Kelvin Nilsen  <kelvin@gcc.gnu.org>
10694         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
10695         behavior of vec_packsu (vector unsigned long long, vector unsigned
10696         long long) to match behavior of vec_packs with same signature.
10698 2018-06-20  Chung-Lin Tang <cltang@codesourcery.com>
10699             Thomas Schwinge <thomas@codesourcery.com>
10700             Cesar Philippidis  <cesar@codesourcery.com>
10702         * gimplify.c (gimplify_scan_omp_clauses): Add support for
10703         OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
10704         (gimplify_adjust_omp_clauses): Likewise.
10705         (gimplify_oacc_declare_1): Add support for GOMP_MAP_RELEASE, remove
10706         support for GOMP_MAP_FORCE_{ALLOC,TO,FROM,TOFROM}.
10707         (gimplify_omp_target_update): Update handling of acc update and
10708         enter/exit data.
10709         * omp-low.c (install_var_field): Remove unused parameter
10710         base_pointers_restrict.
10711         (scan_sharing_clauses): Remove base_pointers_restrict parameter.
10712         Update call to install_var_field. Handle OMP_CLAUSE_{IF_PRESENT,
10713         FINALIZE}
10714         (omp_target_base_pointers_restrict_p): Delete.
10715         (scan_omp_target): Update call to scan_sharing_clauses.
10716         * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_{IF_PRESENT,
10717         FINALIZE}.
10718         * tree-nested.c (convert_nonlocal_omp_clauses): Handle
10719         OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
10720         (convert_local_omp_clauses): Likewise.
10721         * tree-pretty-print.c (dump_omp_clause): Likewise.
10722         * tree.c (omp_clause_num_ops): Add entries for  OMP_CLAUSE_{IF_PRESENT,
10723         FINALIZE}.
10724         (omp_clause_code_name): Likewise.
10726 2018-06-20  Jakub Jelinek  <jakub@redhat.com>
10728         PR debug/86194
10729         * var-tracking.c (use_narrower_mode_test): Check if shift amount can
10730         be narrowed.
10732         PR tree-optimization/86231
10733         * tree-vrp.c (union_ranges): For (  [  )  ] or (   )[   ] range and
10734         anti-range don't overwrite *vr0min before using it to compute *vr0max.
10736 2018-06-20  Tom de Vries  <tdevries@suse.de>
10738         PR tree-optimization/86097
10739         * tree-ssa-loop-manip.c (canonicalize_loop_ivs): Also convert *nit to
10740         iv type if signedness of iv type is not the same as that of *nit.
10742 2018-06-20  Jakub Jelinek  <jakub@redhat.com>
10744         * cfgrtl.c (rtl_verify_edges): Formatting fix.  If bb->preds has any
10745         EDGE_EH edges, verify they are all EDGE_EH.
10747 2018-06-20  Maya Rashish  <coypu@sdf.org>
10749         * ginclude/stddef.h: Limit #include <machine/ansi.h> to NetBSD.
10751 2018-06-20  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
10753         * config/aarch64/aarch64-tuning-flags.def (no_ldp_stp_qregs): New.
10754         * config/aarch64/aarch64.c (xgene1_tunings): Add
10755         AARCH64_EXTRA_TUNE_NO_LDP_STP_QREGS to tune_flags.
10756         (aarch64_mode_valid_for_sched_fusion_p):
10757         Allow 16-byte modes.
10758         (aarch64_classify_address): Allow 16-byte modes for load_store_pair_p.
10759         * config/aarch64/aarch64-ldpstp.md: Add peepholes for LDP STP of
10760         128-bit modes.
10761         * config/aarch64/aarch64-simd.md (load_pair<VQ:mode><VQ2:mode>):
10762         New pattern.
10763         (vec_store_pair<VQ:mode><VQ2:mode>): Likewise.
10764         * config/aarch64/iterators.md (VQ2): New mode iterator.
10766 2018-06-20  Martin Liska  <mliska@suse.cz>
10768         * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
10769         Change default ratio from 10 to 8.
10771 2018-06-20  Martin Liska  <mliska@suse.cz>
10773         * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
10774         New.
10775         (bit_test_cluster::find_bit_tests): Likewise.
10776         (switch_decision_tree::analyze_switch_statement): Find clusters.
10777         * tree-switch-conversion.h (struct jump_table_cluster): Document
10778         hierarchy.
10780 2018-06-20  Martin Liska  <mliska@suse.cz>
10782         * tree-switch-conversion.c (switch_conversion::collect):
10783         Record m_uniq property.
10784         (switch_conversion::expand): Bail out for special conditions.
10785         (group_cluster::~group_cluster): New.
10786         (group_cluster::group_cluster): Likewise.
10787         (group_cluster::dump): Likewise.
10788         (jump_table_cluster::emit): New.
10789         (switch_decision_tree::fix_phi_operands_for_edges): New.
10790         (struct case_node): Remove struct.
10791         (jump_table_cluster::can_be_handled): New.
10792         (case_values_threshold): Moved to header.
10793         (reset_out_edges_aux): Likewise.
10794         (jump_table_cluster::is_beneficial): New.
10795         (bit_test_cluster::can_be_handled): Likewise.
10796         (add_case_node): Remove.
10797         (bit_test_cluster::is_beneficial): New.
10798         (case_bit_test::cmp): New.
10799         (bit_test_cluster::emit): New.
10800         (expand_switch_as_decision_tree_p): Remove.
10801         (bit_test_cluster::hoist_edge_and_branch_if_true): New.
10802         (fix_phi_operands_for_edge): Likewise.
10803         (switch_decision_tree::analyze_switch_statement): New.
10804         (compute_cases_per_edge): Move ...
10805         (switch_decision_tree::compute_cases_per_edge): ... here.
10806         (try_switch_expansion): Likewise.
10807         (switch_decision_tree::try_switch_expansion): Likewise.
10808         (record_phi_operand_mapping): Likewise.
10809         (switch_decision_tree::record_phi_operand_mapping): Likewise.
10810         (emit_case_decision_tree): Likewise.
10811         (switch_decision_tree::emit): Likewise.
10812         (balance_case_nodes): Likewise.
10813         (switch_decision_tree::balance_case_nodes): Likewise.
10814         (dump_case_nodes): Likewise.
10815         (switch_decision_tree::dump_case_nodes): Likewise.
10816         (emit_jump): Likewise.
10817         (switch_decision_tree::emit_jump): Likewise.
10818         (emit_cmp_and_jump_insns): Likewise.
10819         (switch_decision_tree::emit_cmp_and_jump_insns): Likewise.
10820         (emit_case_nodes): Likewise.
10821         (switch_decision_tree::emit_case_nodes): Likewise.
10822         (conditional_probability): Remove.
10823         * tree-switch-conversion.h (enum cluster_type): New.
10824         (PRINT_CASE): New.
10825         (struct cluster): Likewise.
10826         (cluster::cluster): Likewise.
10827         (struct simple_cluster): Likewise.
10828         (simple_cluster::simple_cluster): Likewise.
10829         (struct group_cluster): Likewise.
10830         (struct jump_table_cluster): Likewise.
10831         (struct bit_test_cluster): Likewise.
10832         (struct min_cluster_item): Likewise.
10833         (struct case_tree_node): Likewise.
10834         (case_tree_node::case_tree_node): Likewise.
10835         (jump_table_cluster::case_values_threshold): Likewise.
10836         (struct case_bit_test): Likewise.
10837         (struct switch_decision_tree): Likewise.
10838         (struct switch_conversion): Likewise.
10839         (switch_decision_tree::reset_out_edges_aux): Likewise.
10841 2018-06-20  Martin Liska  <mliska@suse.cz>
10843         * tree-switch-conversion.c (MAX_CASE_BIT_TESTS): Remove.
10844         (hoist_edge_and_branch_if_true): Likewise.
10845         (expand_switch_using_bit_tests_p): Likewise.
10846         (struct case_bit_test): Likewise.
10847         (case_bit_test_cmp): Likewise.
10848         (emit_case_bit_tests): Likewise.
10849         (switch_conversion::switch_conversion): New class.
10850         (struct switch_conv_info): Remove old struct.
10851         (collect_switch_conv_info): More to ...
10852         (switch_conversion::collect): ... this.
10853         (check_range): Likewise.
10854         (switch_conversion::check_range): Likewise.
10855         (check_all_empty_except_final): Likewise.
10856         (switch_conversion::check_all_empty_except_final): Likewise.
10857         (check_final_bb): Likewise.
10858         (switch_conversion::check_final_bb): Likewise.
10859         (create_temp_arrays): Likewise.
10860         (switch_conversion::create_temp_arrays): Likewise.
10861         (free_temp_arrays): Likewise.
10862         (gather_default_values): Likewise.
10863         (switch_conversion::gather_default_values): Likewise.
10864         (build_constructors): Likewise.
10865         (switch_conversion::build_constructors): Likewise.
10866         (constructor_contains_same_values_p): Likewise.
10867         (switch_conversion::contains_same_values_p): Likewise.
10868         (array_value_type): Likewise.
10869         (switch_conversion::array_value_type): Likewise.
10870         (build_one_array): Likewise.
10871         (switch_conversion::build_one_array): Likewise.
10872         (build_arrays): Likewise.
10873         (switch_conversion::build_arrays): Likewise.
10874         (gen_def_assigns): Likewise.
10875         (switch_conversion::gen_def_assigns): Likewise.
10876         (prune_bbs): Likewise.
10877         (switch_conversion::prune_bbs): Likewise.
10878         (fix_phi_nodes): Likewise.
10879         (switch_conversion::fix_phi_nodes): Likewise.
10880         (gen_inbound_check): Likewise.
10881         (switch_conversion::gen_inbound_check): Likewise.
10882         (process_switch): Use the newly created class.
10883         (switch_conversion::expand): New.
10884         (switch_conversion::~switch_conversion): New.
10885         * tree-switch-conversion.h: New file.
10887 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
10889         * tree-vectorizer.h (NUM_PATTERNS, vect_recog_func_ptr): Move to
10890         tree-vect-patterns.c.
10891         * tree-vect-patterns.c (vect_supportable_direct_optab_p): New function.
10892         (vect_recog_dot_prod_pattern): Use it.  Remove the type_in argument.
10893         (vect_recog_sad_pattern): Likewise.
10894         (vect_recog_widen_sum_pattern): Likewise.
10895         (vect_recog_pow_pattern): Likewise.  Check for a null vectype.
10896         (vect_recog_widen_shift_pattern): Remove the type_in argument.
10897         (vect_recog_rotate_pattern): Likewise.
10898         (vect_recog_mult_pattern): Likewise.
10899         (vect_recog_vector_vector_shift_pattern): Likewise.
10900         (vect_recog_divmod_pattern): Likewise.
10901         (vect_recog_mixed_size_cond_pattern): Likewise.
10902         (vect_recog_bool_pattern): Likewise.
10903         (vect_recog_mask_conversion_pattern): Likewise.
10904         (vect_try_gather_scatter_pattern): Likewise.
10905         (vect_recog_widen_mult_pattern): Likewise.  Check for a null vectype.
10906         (vect_recog_over_widening_pattern): Likewise.
10907         (vect_recog_gather_scatter_pattern): Likewise.
10908         (vect_recog_func_ptr): Move from tree-vectorizer.h
10909         (vect_vect_recog_func_ptrs): Move further down the file.
10910         (vect_recog_func): Likewise.  Remove the third argument.
10911         (NUM_PATTERNS): Define based on vect_vect_recog_func_ptrs.
10912         (vect_pattern_recog_1): Expect the pattern function to do any
10913         necessary target tests.  Also expect it to provide a vector type.
10914         Remove the type_in handling.
10916 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
10918         * tree-vect-patterns.c (vect_pattern_detected): New function.
10919         (vect_recog_dot_prod_patternm, vect_recog_sad_pattern)
10920         (vect_recog_widen_mult_pattern, vect_recog_widen_sum_pattern)
10921         (vect_recog_over_widening_pattern, vect_recog_widen_shift_pattern
10922         (vect_recog_rotate_pattern, vect_recog_vector_vector_shift_pattern)
10923         (vect_recog_mult_pattern, vect_recog_divmod_pattern)
10924         (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
10925         (vect_recog_mask_conversion_pattern)
10926         (vect_try_gather_scatter_pattern): Likewise.
10928 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
10930         * tree-vect-patterns.c (vect_get_internal_def): New function.
10931         (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
10932         (vect_recog_vector_vector_shift_pattern, check_bool_pattern)
10933         (search_type_for_mask_1): Use it.
10935 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
10937         * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
10938         redundant WIDEN_SUM_EXPR handling.
10939         (vect_recog_sad_pattern): Likewise.
10941 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
10943         * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
10944         redundant check that the types of a PLUS_EXPR or MULT_EXPR agree.
10945         (vect_recog_sad_pattern): Likewise PLUS_EXPR, ABS_EXPR and MINUS_EXPR.
10946         (vect_recog_widen_mult_pattern): Likewise MULT_EXPR.
10947         (vect_recog_widen_sum_pattern): Likewise PLUS_EXPR.
10949 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
10951         * tree-vect-stmts.c (vectorizable_call): Make sure that we
10952         use the stmt_vec_info of the original bb statement for the
10953         new zero assignment, even if the call is part of a pattern.
10955 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
10957         * tree-vectorizer.h (_stmt_vec_info): Note above pattern_def_seq
10958         that the sequence is attached to the original statement rather
10959         than the pattern statement.
10960         * tree-vect-loop.c (vect_determine_vf_for_stmt): Take the
10961         PATTERN_DEF_SEQ from the original statement rather than
10962         the main pattern statement.
10963         * tree-vect-stmts.c (free_stmt_vec_info): Likewise.
10964         * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Likewise.
10965         (vect_mark_pattern_stmts): Don't copy the PATTERN_DEF_SEQ.
10967 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
10969         * tree-vect-stmts.c (vect_analyze_stmt): Move the handling of pattern
10970         definition statements before the early exit for statements that aren't
10971         live or relevant.
10972         * tree-vect-loop.c (vect_transform_loop_stmt): New function,
10973         split out from...
10974         (vect_transform_loop): ...here.  Process pattern definition
10975         statements without first checking whether the main pattern
10976         statement is live or relevant.
10978 2018-06-19  Eric Botcazou  <ebotcazou@adacore.com>
10980         * tree-cfgcleanup.c (tree_forwarder_block_p): Do not return false at
10981         -O0 if the locus represent UNKNOWN_LOCATION but have different values.
10983 2018-06-19  Aaron Sawdey  <acsawdey@linux.ibm.com>
10985         * config/rs6000/rs6000-string.c (select_block_compare_mode): Check
10986         TARGET_EFFICIENT_OVERLAPPING_UNALIGNED here instead of in caller.
10987         (do_and3, do_and3_mask, do_compb3, do_rotl3): New functions.
10988         (expand_block_compare): Change select_block_compare_mode call.
10989         (expand_strncmp_align_check): Use new functions, fix comment.
10990         (emit_final_str_compare_gpr): New function.
10991         (expand_strn_compare): Refactor and clean up code.
10992         * config/rs6000/vsx.md (vsx_mov<mode>_64bit): Remove *.
10994 2018-06-19  Tony Reix  <tony.reix@atos.com>
10995             Damien Bergamini  <damien.bergamini@atos.com>
10996             David Edelsohn  <dje.gcc@gmail.com>
10998         * collect2.c (static_obj): New variable.
10999         (static_libs): New variable.
11000         (is_in_list): Uncomment declaration.
11001         (main): Track AIX libraries linked statically.
11002         (is_in_list): Uncomment definition.
11003         (scan_prog_file): Don't add AIX shared libraries initializer
11004         to constructor list if linking statically.
11006 2018-06-19  Max Filippov  <jcmvbkbc@gmail.com>
11008         * config/xtensa/xtensa.md (UNSPEC_FRAME_BLOCKAGE): New unspec
11009         constant.
11010         (allocate_stack, frame_blockage, *frame_blockage): New patterns.
11012 2018-06-19  Jan Hubicka  <hubicka@ucw.cz>
11014         * tree.c (find_decls_types_r): Remove all non-VAR_DECLs from
11015         blocks.
11017 2018-06-19  Martin Liska  <mliska@suse.cz>
11019         * config/i386/i386.c (ix86_can_inline_p): Do not use
11020         ipa_fn_summaries::get_create.
11021         * ipa-cp.c (ipcp_cloning_candidate_p): Replace get_create with
11022         get.
11023         (devirtualization_time_bonus): Likewise.
11024         (ipcp_propagate_stage): Likewise.
11025         * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
11026         (edge_set_predicate): Likewise.
11027         (evaluate_conditions_for_known_args): Likewise.
11028         (evaluate_properties_for_edge): Likewise.
11029         (ipa_call_summary::reset): Tranform to ...
11030         (ipa_call_summary::~ipa_call_summary): ... this.
11031         (ipa_fn_summary::reset): Transform to ...
11032         (ipa_fn_summary::~ipa_fn_summary): ... this.
11033         (ipa_fn_summary_t::remove): Rename to ...
11034         (ipa_fn_summary_t::remove_callees): ... this.
11035         (ipa_fn_summary_t::duplicate): Use placement new
11036         instead of memory copy.
11037         (ipa_call_summary_t::duplicate): Likewise.
11038         (ipa_call_summary_t::remove): Remove.
11039         (dump_ipa_call_summary): Change get_create to get.
11040         (ipa_dump_fn_summary): Dump only when summary exists.
11041         (analyze_function_body): Use symbol_summary::get instead
11042         of get_create.
11043         (compute_fn_summary): Likewise.
11044         (estimate_edge_devirt_benefit): Likewise.
11045         (estimate_edge_size_and_time): Likewise.
11046         (inline_update_callee_summaries): Likewise.
11047         (remap_edge_change_prob): Likewise.
11048         (remap_edge_summaries): Likewise.
11049         (ipa_merge_fn_summary_after_inlining): Likewise.
11050         (write_ipa_call_summary): Likewise.
11051         (ipa_fn_summary_write): Likewise.
11052         (ipa_free_fn_summary): Likewise.
11053         * ipa-fnsummary.h (struct GTY): Add new ctor and copy ctor.
11054         (struct ipa_call_summary): Likewise.
11055         * ipa-icf.c (sem_function::merge): Use symbol_summary::get instead
11056         of get_create.
11057         * ipa-inline-analysis.c (do_estimate_edge_time): Likewise.
11058         (estimate_size_after_inlining): Likewise.
11059         (estimate_growth): Likewise.
11060         (growth_likely_positive): Likewise.
11061         * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
11062         (inline_call): Likewise.
11063         * ipa-inline.c (caller_growth_limits): Likewise.
11064         (can_inline_edge_p): Likewise.
11065         (can_inline_edge_by_limits_p): Likewise.
11066         (compute_uninlined_call_time): Likewise.
11067         (compute_inlined_call_time): Likewise.
11068         (want_inline_small_function_p): Likewise.
11069         (edge_badness): Likewise.
11070         (update_caller_keys): Likewise.
11071         (update_callee_keys): Likewise.
11072         (inline_small_functions): Likewise.
11073         (inline_to_all_callers_1): Likewise.
11074         (dump_overall_stats): Likewise.
11075         (early_inline_small_functions): Likewise.
11076         (early_inliner): Likewise.
11077         * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
11078         * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
11079         * ipa-pure-const.c (malloc_candidate_p): Likewise.
11080         * ipa-split.c (execute_split_functions): Likewise.
11081         * symbol-summary.h: Likewise.
11082         * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
11084 2018-06-19  Richard Biener  <rguenther@suse.de>
11086         * tree-vectorizer.c (try_vectorize_loop_1): Split out of ...
11087         (vectorize_loops): ... here.  Fix dbgcnt handling.
11088         (try_vectorize_loop): Wrap try_vectorize_loop_1.
11090 2018-06-19  Segher Boessenkool  <segher@kernel.crashing.org>
11092         PR target/86197
11093         * config/rs6000/rs6000.md (rs6000_discover_homogeneous_aggregate): An
11094         ieee128 argument takes up only one (vector) register, not two (floating
11095         point) registers.
11097 2018-06-19  Eric Botcazou  <ebotcazou@adacore.com>
11099         * gimplify.c (gimplify_init_constructor): Really never clear for an
11100         incomplete constructor if CONSTRUCTOR_NO_CLEARING is set.
11102 2018-06-19  Richard Biener  <rguenther@suse.de>
11104         PR tree-optimization/86179
11105         * tree-vect-patterns.c (vect_pattern_recog_1): Clean up
11106         after failed recognition.
11108 2018-06-18  Martin Sebor  <msebor@redhat.com>
11110         PR middle-end/85602
11111         * calls.c (maybe_warn_nonstring_arg): Handle strncat.
11112         * tree-ssa-strlen.c (is_strlen_related_p): Make extern.
11113         Handle integer subtraction.
11114         (maybe_diag_stxncpy_trunc): Handle nonstring source arguments.
11115         * tree-ssa-strlen.h (is_strlen_related_p): Declare.
11117 2018-06-18  David Malcolm  <dmalcolm@redhat.com>
11119         * config/frv/frv-protos.h (frv_ifcvt_modify_insn): Strengthen 3rd
11120         param from rtx to rtx_insn *.
11121         * config/frv/frv.c (frv_ifcvt_add_insn): Likewise for "insn"
11122         param.
11123         (frv_ifcvt_modify_insn): Likwise.
11124         (frv_ifcvt_modify_final): Likwise for local "existing_insn",
11125         adding an as_a <rtx_insn *> cast.  Likewise for local "insn".
11126         * config/mips/mips.c (r10k_insert_cache_barriers): Add an
11127         as_a <rtx_insn *> cast to local "unprotected_region" once
11128         it's been established that it's not NULL or pc_rtx.
11129         * config/nds32/nds32-relax-opt.c (nds32_group_insns): Strengthen
11130         param "sethi" from rtx to rtx_insn *.
11131         (nds32_group_float_insns): Likewise for param "insn".
11132         * config/vax/vax-protos.h (vax_output_int_add): Likewise for 1st
11133         param.
11134         (vax_output_int_subtract): Likewise.
11135         * config/vax/vax.c (vax_output_int_add): Likewise for param
11136         "insn".
11137         (vax_output_int_subtract): Likewise.
11138         * emit-rtl.c (set_insn_deleted): Likewise, removing cast.
11139         (emit_pattern_after): Likewise for param "after".
11140         (emit_insn_after): Likewise.
11141         (emit_jump_insn_after): Likewise.
11142         (emit_call_insn_after): Likewise.
11143         (emit_debug_insn_after): Likewise.
11144         (emit_pattern_before): Likewise for param "before".
11145         (emit_insn_before): Likewise.
11146         (emit_jump_insn_before): Likewise.
11147         * final.c (get_insn_template): Likewise for param "insn", removing
11148         a cast.
11149         * output.h (get_insn_template): Likewise for 2nd param.
11150         * rtl.h (emit_insn_before): Likewise.
11151         (emit_jump_insn_before): Likewise.
11152         (emit_debug_insn_before_noloc): Likewise.
11153         (emit_insn_after): Likewise.
11154         (emit_jump_insn_after): Likewise.
11155         (emit_call_insn_after): Likewise.
11156         (emit_debug_insn_after): Likewise.
11157         (set_insn_deleted): Likewise for param.
11159 2018-06-18  Michael Meissner  <meissner@linux.ibm.com>
11161         PR target/85358
11162         * config/rs6000/rs6000-modes.def (toplevel): Rework the 128-bit
11163         floating point modes, so that IFmode is numerically greater than
11164         TFmode, which is greater than KFmode using FRACTIONAL_FLOAT_MODE
11165         to declare the ordering.  This prevents IFmode from being
11166         converted to TFmode when long double is IEEE 128-bit on an ISA 3.0
11167         machine.  Include rs6000-modes.h to share the fractional values
11168         between genmodes* and the rest of the compiler.
11169         (IFmode): Likewise.
11170         (KFmode): Likewise.
11171         (TFmode): Likewise.
11172         * config/rs6000/rs6000-modes.h: New file.
11173         * config/rs6000/rs6000.c (rs6000_debug_reg_global): Change the
11174         meaning of rs6000_long_double_size so that 126..128 selects an
11175         appropriate 128-bit floating point type.
11176         (rs6000_option_override_internal): Likewise.
11177         * config/rs6000/rs6000.h (toplevel): Include rs6000-modes.h.
11178         (TARGET_LONG_DOUBLE_128): Change the meaning of
11179         rs6000_long_double_size so that 126..128 selects an appropriate
11180         128-bit floating point type.
11181         (LONG_DOUBLE_TYPE_SIZE): Update comment.
11182         * config/rs6000/rs6000.md (trunciftf2): Correct the modes of the
11183         source and destination to match the standard usage.
11184         (truncifkf2): Likewise.
11185         (copysign<mode>3, IEEE iterator): Rework copysign of float128 on
11186         ISA 2.07 to use an explicit clobber, instead of passing in a
11187         temporary.
11188         (copysign<mode>3_soft): Likewise.
11190 2018-06-18  David Malcolm  <dmalcolm@redhat.com>
11192         * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
11193         Replace dump_printf_loc call with DUMP_VECT_SCOPE.
11194         (vect_slp_analyze_instance_dependence): Likewise.
11195         (vect_enhance_data_refs_alignment): Likewise.
11196         (vect_analyze_data_refs_alignment): Likewise.
11197         (vect_slp_analyze_and_verify_instance_alignment
11198         (vect_analyze_data_ref_accesses): Likewise.
11199         (vect_prune_runtime_alias_test_list): Likewise.
11200         (vect_analyze_data_refs): Likewise.
11201         * tree-vect-loop-manip.c (vect_update_inits_of_drs): Likewise.
11202         * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
11203         (vect_analyze_scalar_cycles_1): Likewise.
11204         (vect_get_loop_niters): Likewise.
11205         (vect_analyze_loop_form_1): Likewise.
11206         (vect_update_vf_for_slp): Likewise.
11207         (vect_analyze_loop_operations): Likewise.
11208         (vect_analyze_loop): Likewise.
11209         (vectorizable_induction): Likewise.
11210         (vect_transform_loop): Likewise.
11211         * tree-vect-patterns.c (vect_pattern_recog): Likewise.
11212         * tree-vect-slp.c (vect_analyze_slp): Likewise.
11213         (vect_make_slp_decision): Likewise.
11214         (vect_detect_hybrid_slp): Likewise.
11215         (vect_slp_analyze_operations): Likewise.
11216         (vect_slp_bb): Likewise.
11217         * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Likewise.
11218         (vectorizable_bswap): Likewise.
11219         (vectorizable_call): Likewise.
11220         (vectorizable_simd_clone_call): Likewise.
11221         (vectorizable_conversion): Likewise.
11222         (vectorizable_assignment): Likewise.
11223         (vectorizable_shift): Likewise.
11224         (vectorizable_operation): Likewise.
11225         * tree-vectorizer.h (DUMP_VECT_SCOPE): New macro.
11227 2018-06-18  Martin Sebor  <msebor@redhat.com>
11229         PR tree-optimization/81384
11230         * builtin-types.def (BT_FN_SIZE_CONST_STRING_SIZE): New.
11231         * builtins.c (expand_builtin_strnlen): New function.
11232         (expand_builtin): Call it.
11233         (fold_builtin_n): Avoid setting TREE_NO_WARNING.
11234         * builtins.def (BUILT_IN_STRNLEN): New.
11235         * calls.c (maybe_warn_nonstring_arg): Handle BUILT_IN_STRNLEN.
11236         Warn for bounds in excess of maximum object size.
11237         * tree-ssa-strlen.c (maybe_set_strlen_range): Return tree representing
11238         single-value ranges.  Handle strnlen.
11239         (handle_builtin_strlen): Handle strnlen.
11240         (strlen_check_and_optimize_stmt): Same.
11241         * doc/extend.texi (Other Builtins): Document strnlen.
11243 2018-06-18  Maya Rashish  <coypu@sdf.org>
11245         * config/alpha/openbsd.h (TARGET_DEFAULT): Define.
11246         (LINK_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): Likewise.
11247         (INTMAX_TYPE, UINTMAX_TYPE, WINT_TYPE): Likewise.
11249         * config/alpha/elf.h (STARTFILE_SPEC, ENDFILE_SPEC): Move from
11250         here to ...
11251         * config/alpha/linux.h (STARTFILE_SPEC, ENDFILE_SPEC): Here.
11253 2018-06-18  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
11255         * tree.c (escaped_string::escape): Replace cast to char * by
11256         const_cast<char *> (unescaped).
11258 2018-06-18  Nick Clifton  <nickc@redhat.com>
11260         PR 84195
11261         * tree.c (escaped_string): New class.  Converts an unescaped
11262         string into its escaped equivalent.
11263         (warn_deprecated_use): Use the new class to convert the
11264         deprecation message, if present.
11265         (test_escaped_strings): New self test.
11266         (test_c_tests): Add test_escaped_strings.
11267         * doc/extend.texi (deprecated): Add a note that the
11268         deprecation message is affected by the -fmessage-length
11269         option, and that control characters will be escaped.
11270         (#pragma GCC error): Document this pragma.
11271         (#pragma GCC warning): Likewise.
11272         * doc/invoke.texi (-fmessage-length): Document this option's
11273         effect on the #warning and #error preprocessor directives and
11274         the deprecated attribute.
11276 2018-06-18  Eric Botcazou  <ebotcazou@adacore.com>
11278         * tree.c (decl_value_expr_lookup): Revert latest change.
11279         (decl_value_expr_insert): Likewise.
11281 2018-06-17  Eric Botcazou  <ebotcazou@adacore.com>
11283         * gimplify.c (nonlocal_vlas): Delete.
11284         (nonlocal_vla_vars): Likewise.
11285         (gimplify_var_or_parm_decl): Do not add debug VAR_DECLs for non-local
11286         referenced VLAs.
11287         (gimplify_body): Do not create and destroy nonlocal_vlas.
11288         * tree-nested.c: Include diagnostic.h.
11289         (use_pointer_in_frame): Tweak.
11290         (lookup_field_for_decl): Add assertion and declare the transformation.
11291         (convert_nonlocal_reference_op) <PARM_DECL>: Rework and issue an
11292         internal error when the reference is in a wrong context.  Do not
11293         create a debug decl by default.
11294         (note_nonlocal_block_vlas): Delete.
11295         (convert_nonlocal_reference_stmt) <GIMPLE_BIND>: Do not call it.
11296         (convert_local_reference_op) <PARM_DECL>: Skip the frame decl.  Do not
11297         create a debug decl by default.
11298         (convert_gimple_call) <GIMPLE_CALL>: Issue an internal error when the
11299         call is in a wrong context.
11300         (fixup_vla_decls): New function.
11301         (finalize_nesting_tree_1): Adjust comment.  Call fixup_vla_decls if no
11302         debug variables were created.
11303         * tree.c (decl_value_expr_lookup): Add checking assertion.
11304         (decl_value_expr_insert): Likewise.
11306 2018-06-16  Kugan Vivekanandarajah  <kuganv@linaro.org>
11308         PR middle-end/82479
11309         * ipa-fnsummary.c (will_be_nonconstant_expr_predicate): Handle CALL_EXPR.
11310         * tree-scalar-evolution.c (interpret_expr): Likewise.
11311         (expression_expensive_p): Likewise.
11312         * tree-ssa-loop-ivopts.c (contains_abnormal_ssa_name_p): Likewise.
11313         * tree-ssa-loop-niter.c (number_of_iterations_popcount): New.
11314         (number_of_iterations_exit_assumptions): Use number_of_iterations_popcount.
11315         (ssa_defined_by_minus_one_stmt_p): New.
11317 2018-06-16  Kugan Vivekanandarajah  <kuganv@linaro.org>
11319         PR middle-end/64946
11320         * cfgexpand.c (expand_debug_expr): Hande ABSU_EXPR.
11321         * config/i386/i386.c (ix86_add_stmt_cost): Likewise.
11322         * dojump.c (do_jump): Likewise.
11323         * expr.c (expand_expr_real_2): Check operand type's sign.
11324         * fold-const.c (const_unop): Handle ABSU_EXPR.
11325         (fold_abs_const): Likewise.
11326         * gimple-pretty-print.c (dump_unary_rhs): Likewise.
11327         * gimple-ssa-backprop.c (backprop::process_assign_use): Likesie.
11328         (strip_sign_op_1): Likesise.
11329         * match.pd: Add new pattern to generate ABSU_EXPR.
11330         * optabs-tree.c (optab_for_tree_code): Handle ABSU_EXPR.
11331         * tree-cfg.c (verify_gimple_assign_unary): Likewise.
11332         * tree-eh.c (operation_could_trap_helper_p): Likewise.
11333         * tree-inline.c (estimate_operator_cost): Likewise.
11334         * tree-pretty-print.c (dump_generic_node): Likewise.
11335         * tree-vect-patterns.c (vect_recog_sad_pattern): Likewise.
11336         * tree.def (ABSU_EXPR): New.
11338 2018-06-16  Jakub Jelinek  <jakub@redhat.com>
11340         PR middle-end/86095
11341         * common.opt (Wunsafe-loop-optimizations): Add Ignore, remove Var,
11342         documented as preserved for backward compatibility only.
11343         * doc/invoke.texi: Remove -Wunsafe-loop-optimizations documentation.
11345         PR rtl-optimization/86108
11346         * bb-reorder.c (create_forwarder_block): Renamed to ...
11347         (create_eh_forwarder_block): ... this.  Split OLD_BB after labels and
11348         jump from new landing pad to the second part.
11349         (sjlj_fix_up_crossing_landing_pad, dw2_fix_up_crossing_landing_pad):
11350         Adjust callers.
11352 2018-06-15  Jakub Jelinek  <jakub@redhat.com>
11354         PR middle-end/85878
11355         * expr.c (expand_assignment): Remove now redundant COMPLEX_MODE_P
11356         check from first store_expr, use to_mode instead of GET_MODE (to_rtx).
11357         Only call store_expr for halves if the mode is the same.
11359         PR middle-end/86123
11360         * match.pd ((X / Y) == 0 -> X < Y): Don't transform complex divisions.
11361         Fix up comment formatting.
11363 2018-06-15  Bernd Edlinger  <bernd.edlinger@hotmail.de>
11365         * typed-splay-tree.h (typed_splay_tree::remove): New function.
11366         (typed_splay_tree::closure,
11367         typed_splay_tree::inner_foreach_fn, typed_splay_tree::m_inner): Deleted.
11368         (typed_splay_tree::typed_splay_tree,
11369         typed_splay_tree::operator =): Declared private.
11370         (typed_splay_tree::splay_tree_key, typed_splay_tree::splay_tree_value,
11371         typed_splay_tree::splay_tree_node_s, typed_splay_tree::KDEL,
11372         typed_splay_tree::VDEL, typed_splay_tree::splay_tree_delete_helper,
11373         typed_splay_tree::rotate_left, typed_splay_tree::rotate_right,
11374         typed_splay_tree::splay_tree_splay,
11375         typed_splay_tree::splay_tree_foreach_helper,
11376         typed_splay_tree::splay_tree_insert,
11377         typed_splay_tree::splay_tree_remove,
11378         typed_splay_tree::splay_tree_lookup,
11379         typed_splay_tree::splay_tree_predecessor,
11380         typed_splay_tree::splay_tree_successor,
11381         typed_splay_tree::splay_tree_min,
11382         typed_splay_tree::splay_tree_max): Took over from splay-tree.c/.h.
11383         (typed_splay_tree::root, typed_splay_tree::comp,
11384         typed_splay_tree::delete_key,
11385         typed_splay_tree::delete_value): New data members.
11386         * typed-splay-tree.c (selftest::test_str_to_int): Add a test for
11387         typed_splay_tree::remove.
11389 2018-06-15  Matthew Fortune  <matthew.fortune@mips.com>
11391         * config/mips/mips.h (ASM_SPEC): Pass through -mcrc, -mno-crc,
11392         -mginv and -mno-ginv to the assembler.
11393         * config/mips/mips.opt (-mcrc): New option.
11394         (-mginv): Likewise.
11395         * doc/invoke.text (-mcrc): Document.
11396         (-mginv): Likewise.
11398 2018-06-15  Nick Clifton  <nickc@redhat.com>
11400         PR 84195
11401         * tree.c (escaped_string): New class.  Converts an unescaped
11402         string into its escaped equivalent.
11403         (warn_deprecated_use): Use the new class to convert the
11404         deprecation message, if present.
11405         (test_escaped_strings): New self test.
11406         (test_c_tests): Add test_escaped_strings.
11407         * doc/extend.texi (deprecated): Add a note that the
11408         deprecation message is affected by the -fmessage-length
11409         option, and that control characters will be escaped.
11410         (#pragma GCC error): Document this pragma.
11411         (#pragma GCC warning): Likewise.
11412         * doc/invoke.texi (-fmessage-length): Document this option's
11413         effect on the #warning and #error preprocessor directives and
11414         the deprecated attribute.
11416 2018-06-15  Richard Biener  <rguenther@suse.de>
11418         * tree-vect-slp.c (vect_slp_bb): Dump MSG_OPTIMIZED_LOCATIONS
11419         here, also noting vector size used.
11420         * tree-vectorizer.c (vectorize_loops): Adjust.  Note vector
11421         size used in MSG_OPTIMIZED_LOCATIONS dump.
11422         (pass_slp_vectorize::execute): Adjust.
11424 2018-06-15  Claudiu Zissulescu  <claziss@synopsys.com>
11426         PR target/85968
11427         * config/arc/arc.c (arc_return_address_register): Fix
11428         if-condition.
11430 2018-06-15  Richard Biener  <rguenther@suse.de>
11432         PR middle-end/86159
11433         * tree-cfg.c (gimplify_build3): Do not strip sign conversions,
11434         leave useless conversion stripping to force_gimple_operand_gsi.
11435         (gimplify_build2): Likewise.
11436         (gimplify_build1): Likewise.
11438 2018-06-15  Richard Biener  <rguenther@suse.de>
11440         PR middle-end/86076
11441         * tree-cfg.c (move_stmt_op): unshare invariant addresses
11442         before adjusting their block.
11444 2018-06-15  Sebastian Huber  <sebastian.huber@embedded-brains.de>
11446         * config.gcc (riscv*-*-elf* | riscv*-*-rtems*): Use custom
11447         multilibs for *-*-rtems*.
11448         * config/riscv/t-rtems: New file.
11450 2018-06-14  Jakub Jelinek  <jakub@redhat.com>
11452         PR middle-end/86122
11453         * match.pd ((A +- CST1) +- CST2): Punt if last resort
11454         unsigned_type_for returns NULL.
11456         PR target/85945
11457         * lower-subreg.c (find_decomposable_subregs): Don't decompose float
11458         subregs of multi-word pseudos unless the float mode has word size.
11460 2018-06-14  Richard Biener  <rguenther@suse.de>
11462         PR middle-end/86139
11463         * tree-vect-generic.c (build_word_mode_vector_type): Remove
11464         duplicate and harmful type_hash_canon.
11465         * tree.c (type_hash_canon): Assert we didn't find ourselves.
11467 2018-06-14  Richard Biener  <rguenther@suse.de>
11469         PR ipa/86124
11470         * tree-ssa-struct-alias.c (create_variable_info_for): Handle
11471         NULL cgraph_node.
11473 2018-06-14  Sebastian Huber  <sebastian.huber@embedded-brains.de>
11475         * config/rtems.h (STDINT_LONG32): Define.
11477 2018-06-13  Matthew Fortune  <matthew.fortune@mips.com>
11478             Prachi Godbole  <prachi.godbole@imgtec.com>
11480         * config/mips/mips-cpus.def: Define P6600.
11481         * config/mips/mips-tables.opt: Regenerate.
11482         * config/mips/mips.c (mips_ucbranch_type): New enum.
11483         (mips_rtx_cost_data): Add support for P6600.
11484         (mips_issue_rate): Likewise.
11485         (mips_multipass_dfa_lookahead): Likewise.
11486         (mips_avoid_hazard): Likewise.
11487         (mips_reorg_process_insns): Likewise.
11488         (mips_classify_branch_p6600): New function.
11489         * config/mips/mips.h (TUNE_P6600): New define.
11490         (MIPS_ISA_LEVEL_SPEC): Infer mips64r6 from p6600.
11491         (ENABLE_LD_ST_PAIRS): Enable load/store bonding for p6600.
11492         * config/mips/mips.md: Include p6600.md.
11493         (processor): Add p6600.
11494         * config/mips/p6600.md: New file.
11495         * doc/invoke.texi: Add p6600 to supported architectures.
11497 2018-06-13  Martin Sebor  <msebor@redhat.com>
11499         PR tree-optimization/86114
11500         * gimple-fold.c (gimple_fold_builtin_strlen): Only handle LHS
11501         of integer types.
11502         * tree-ssa-strlen.c (maybe_set_strlen_range): Same.
11504 2018-06-13  Richard Biener  <rguenther@suse.de>
11506         * tree-vect-patterns.c (vect_recog_vector_vector_shift_pattern):
11507         Properly set vector type of the intermediate stmt.
11508         * tree-vect-stmts.c (vectorizable_operation): The destination
11509         var always has vectype_out type.
11511 2018-06-13  Jeff Law  <law@redhat.com>
11513         * config/rl78/rl78.c (move_elim_pass): Use TDF_NONE rather than
11514         integer 0 for argument to print_rtl_with_bb.
11515         (rl78_reorg): Likewise.
11517 2018-06-13  David Malcolm  <dmalcolm@redhat.com>
11519         * config/arc/arc.c (hwloop_optimize): Strengthen local "end_label"
11520         from rtx to rtx_insn *.
11521         * config/bfin/bfin.c (hwloop_optimize): Likewise for local
11522         "label".
11523         (add_sched_insns_for_speculation): Likewise for local "target",
11524         converting usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
11525         * config/c6x/c6x.c (reorg_split_calls): Strengthen param "call_labels"
11526         from rtx_insn ** to rtx_code_label **.
11527         (reorg_emit_nops): Likewise.
11528         (c6x_reorg): Likewise for local "call_labels".
11529         * config/sh/sh-protos.h (get_dest_uid): Strengthen 1st param from
11530         rtx to rtx_insn *.
11531         * config/sh/sh.c (dump_table): Strengthen local "lab" from rtx to
11532         rtx_code_label *, adding safe_as_a <rtx_code_label *> casts to
11533         the loops over LABEL_REFS.
11534         (fixup_addr_diff_vecs): Add as_a <rtx_insn *> to usage of
11535         braf_label.
11536         (barrier_align): Convert usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
11537         (get_dest_uid): Strengthen param "label" from rtx to rtx_insn *.
11538         (split_branches): Strengthen local "olabel" from rtx to
11539         rtx_insn *, adding a safe_as_a cast.
11540         * emit-rtl.c (next_real_insn): Strengthen param from "rtx"
11541         to "rtx_insn *".
11542         (add_insn_after): Likewise for first two params.
11543         (add_insn_before): Likewise.
11544         (remove_insn): Likewise for param.
11545         (emit_pattern_before_noloc): Likewise for second and third params.
11546         (emit_jump_insn_before_noloc): Convert NULL_RTX to NULL.
11547         (emit_call_insn_before_noloc): Likewise.
11548         (emit_debug_insn_before_noloc): Strengthen "before" param from "rtx"
11549         to "rtx_insn *".
11550         (emit_barrier_before): Likewise.
11551         (emit_label_before): Strengthen "label" param from "rtx" to
11552         "rtx_code_label *".  Strengthen "before" param from "rtx" to
11553         "rtx_insn *".
11554         (emit_insn_after_1): Strengthen "after" param from "rtx" to
11555         "rtx_insn *".
11556         (emit_pattern_after_noloc): Likewise.
11557         (emit_insn_after_noloc): Likewise.
11558         (emit_jump_insn_after_noloc): Likewise.
11559         (emit_call_insn_after_noloc): Likewise.
11560         (emit_debug_insn_after_noloc): Likewise.
11561         (emit_barrier_after): Likewise.
11562         (emit_label_after): Likewise for both params.
11563         (emit_pattern_after_setloc): Likewise for "after" param.  Convert
11564         "loc" param from "int" to "location_t".
11565         (emit_insn_after_setloc): Likewise.
11566         (emit_jump_insn_after_setloc): Likewise.
11567         (emit_call_insn_after_setloc): Likewise.
11568         (emit_debug_insn_after_setloc): Likewise.
11569         (emit_pattern_before_setloc): Likewise for "before" param.  Convert
11570         "loc" param from "int" to "location_t".
11571         (emit_pattern_before): Convert NULL_RTX to NULL.
11572         (emit_insn_before_setloc): Convert "loc" param from "int" to
11573         "location_t".
11574         (emit_jump_insn_before_setloc): Likewise.
11575         (emit_call_insn_before_setloc): Likewise.
11576         (emit_debug_insn_before_setloc): Strengthen "before" param from rtx to
11577         rtx_insn *.  Convert "loc" param from "int" to "location_t".
11578         * rtl.h (emit_insn_before_setloc, emit_jump_insn_before_setloc,
11579         emit_call_insn_before_setloc, emit_debug_insn_before_setloc):
11580         Convert 3rd param from "int" to "location_t".
11581         (emit_barrier_before, emit_barrier_after, next_real_insn):
11582         Strengthen param from rtx to rtx_insn *.
11583         (emit_label_before): Strengthen 1st param from "rtx" to
11584         "rtx_code_label *".  Strengthen 2nd param from "rtx" to
11585         "rtx_insn *".
11586         (emit_insn_after_noloc, emit_jump_insn_after_noloc,
11587         emit_call_insn_after_noloc, emit_debug_insn_after_noloc):
11588         Strengthen 2nd param from "rtx" to "rtx_insn *".
11589         (emit_insn_after_setloc, emit_jump_insn_after_setloc)
11590         emit_call_insn_after_setloc, emit_debug_insn_after_setloc):
11591         Likewise. Convert 3rd param from "int" to "location_t".
11592         (emit_label_after): Strengthen 1st param from "rtx" to
11593         "rtx_code_label *".
11594         (next_real_insn, remove_insn): Strengthen param from "rtx" to
11595         "rtx_insn *".
11596         (add_insn_before, add_insn_after): Strengthen 1st and 2nd params
11597         from "rtx" to "rtx_insn *".
11599 2018-06-13  Jan Hubicka  <hubicka@gcc.gnu.org>
11601         * cgraph.c (cgraph_node::get_untransformed_body): Dump function
11602         bodies streamed in with -Q.
11603         * dumpfile.c (dump_files): Add lto-stream-out dump file.
11604         * dumpfile.h (tree_dump_index): Add lto_stream_out.
11605         * gimple-streamer-out.c: Include gimple-pretty-print.h
11606         (output_bb): Dump stmts streamed.
11607         * lto-section-out.c: Include print-tree.h
11608         (lto_begin_section): Dump sections created.
11609         (lto_output_decl_index): Dump decl encoded.
11610         * lto-streamer-out.c: Include print-tree.h
11611         (create_output_block): Dump output block created.
11612         (DFS::DFS_write_tree_body): Dump DFS SCCs streamed.
11613         (output_function): Dump function output.
11614         (output_constructor): Dump constructor streamed.
11615         (write_global_stream): Output indexes encoded.
11616         (produce_asm_for_decls): Dump streams encoded.
11617         * lto-streamer.c (streamer_dump_file): New global var.
11618         * lto-streamer.h (streamer_dump_file): Declare.
11619         * passes.c (ipa_write_summaries): Initialize streamer dump.
11620         * varpool.c (varpool_node::get_constructor): Dump constructors streamed
11621         in.
11623 2018-06-13  Eric Botcazou  <ebotcazou@adacore.com>
11625         PR target/86048
11626         * config/i386/winnt.c (i386_pe_seh_cold_init): Do not emit negative
11627         offsets for register save directives.  Emit a second batch of save
11628         directives, if need be, when the function accesses prior frames.
11630 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
11632         * config/arc/fpu.md (fmasf4): Force operand to register.
11633         (fnmasf4): Likewise.
11635 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
11637         * config/arc/arc-protos.h (arc_pad_return): Remove.
11638         * config/arc/arc.c (machine_function): Remove force_short_suffix
11639         and size_reason.
11640         (arc_print_operand): Adjust printing of '&'.
11641         (arc_verify_short): Remove conditional printing of short suffix.
11642         (arc_final_prescan_insn): Remove reference to size_reason.
11643         (pad_return): New function.
11644         (arc_reorg): Call pad_return.
11645         (arc_pad_return): Remove.
11646         (arc_init_machine_status): Remove reference to force_short_suffix.
11647         * config/arc/arc.md (vunspec): Add VUNSPEC_ARC_BLOCKAGE.
11648         (attr length): When attribute iscompact is true force to 2
11649         regardless; in the case of maybe check if we want to force the
11650         instruction to have 4 bytes length.
11651         (nopv): Change it to generate 4 byte long nop as well.
11652         (blockage): New pattern.
11653         (simple_return): Remove call to arc_pad_return.
11654         (p_return_i): Likewise.
11656 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
11658         * config/arc/elf.h (LINK_GCC_C_SEQUENCE_SPEC): Define.
11660 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
11662         * config/arc/builtins.def (SYNC): SYNC instruction is valid on all
11663         ARC cores.
11665 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
11667         * config/arc/arc.c (atomic_exchangesi): EX instruction is default
11668         for ARC700 and ARCv2.
11670 2018-06-13  Chenghua Xu <paul.hua.gm@gmail.com>
11672         PR target/86076
11673         * config/mips/loongson.md (vec_setv4hi): Gen_lowpart for
11674         operands[2] instead of operands[1].
11677 2018-06-12  Richard Sandiford  <richard.sandiford@linaro.org>
11679         * lra-constraints.c (simplify_operand_subreg): In the paradoxical
11680         case, check whether the outer register overlaps an unallocatable
11681         register, not just whether it fits the required class.
11683 2018-06-12  Richard Sandiford  <richard.sandiford@linaro.org>
11685         * poly-int.h (can_div_trunc_p): Add new overload in which all values
11686         are poly_ints.
11687         * alias.c (get_addr): Extend CONST_INT handling to poly_int_rtx_p.
11688         (memrefs_conflict_p): Likewise.
11689         (init_alias_analysis): Likewise.
11690         * cfgexpand.c (expand_debug_expr): Likewise.
11691         * combine.c (combine_simplify_rtx, force_int_to_mode): Likewise.
11692         * cse.c (fold_rtx): Likewise.
11693         * explow.c (adjust_stack, anti_adjust_stack): Likewise.
11694         * expr.c (emit_block_move_hints): Likewise.
11695         (clear_storage_hints, push_block, emit_push_insn): Likewise.
11696         (store_expr_with_bounds, reduce_to_bit_field_precision): Likewise.
11697         (emit_group_load_1): Use rtx_to_poly_int64 for group offsets.
11698         (emit_group_store): Likewise.
11699         (find_args_size_adjust): Use strip_offset.  Use rtx_to_poly_int64
11700         to read the PRE/POST_MODIFY increment.
11701         * calls.c (store_one_arg): Use strip_offset.
11702         * rtlanal.c (rtx_addr_can_trap_p_1): Extend CONST_INT handling to
11703         poly_int_rtx_p.
11704         (set_noop_p): Use rtx_to_poly_int64 for the elements selected
11705         by a VEC_SELECT.
11706         * simplify-rtx.c (avoid_constant_pool_reference): Use strip_offset.
11707         (simplify_binary_operation_1): Extend CONST_INT handling to
11708         poly_int_rtx_p.
11709         * var-tracking.c (compute_cfa_pointer): Take a poly_int64 rather
11710         than a HOST_WIDE_INT.
11711         (hard_frame_pointer_adjustment): Change from HOST_WIDE_INT to
11712         poly_int64.
11713         (adjust_mems, add_stores): Update accodingly.
11714         (vt_canonicalize_addr): Track polynomial offsets.
11715         (emit_note_insn_var_location): Likewise.
11716         (vt_add_function_parameter): Likewise.
11717         (vt_initialize): Likewise.
11719 2018-06-12  Jeff Law  <law@redhat.com>
11721         * config.gcc (alpha*-*-freebsd*): Remove.
11722         * config/alpha/freebsd.h: Remove.
11724 2018-06-12  David Malcolm  <dmalcolm@redhat.com>
11726         PR other/69968
11727         * spellcheck-tree.c (levenshtein_distance): Rename to...
11728         (get_edit_distance): ...this, and update for underlying renaming.
11729         * spellcheck-tree.h (levenshtein_distance): Rename to...
11730         (get_edit_distance): ...this.
11731         * spellcheck.c (levenshtein_distance): Rename to...
11732         (get_edit_distance): ...this.  Convert from Levenshtein distance
11733         to Damerau-Levenshtein distance by supporting transpositions of
11734         adjacent characters.  Rename "v1" to "v_next" and "v0" to
11735         "v_one_ago".
11736         (selftest::levenshtein_distance_unit_test_oneway): Rename to...
11737         (selftest::test_edit_distance_unit_test_oneway): ...this, and
11738         update for underlying renaming.
11739         (selftest::levenshtein_distance_unit_test): Rename to...
11740         (selftest::test_get_edit_distance_unit): ...this, and update for
11741         underlying renaming.
11742         (selftest::test_find_closest_string): Add example from PR 69968
11743         where transposition helps
11744         (selftest::test_metric_conditions): Update for renaming.
11745         (selftest::test_metric_conditions): Likewise.
11746         (selftest::spellcheck_c_tests): Likewise.
11747         * spellcheck.h (levenshtein_distance): Rename both overloads to...
11748         (get_edit_distance): ...this.
11749         (best_match::consider): Update for renaming.
11751 2018-06-12  Martin Sebor  <msebor@redhat.com>
11753         PR tree-optimization/85259
11754         * builtins.c (compute_objsize): Handle constant offsets.
11755         * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Return
11756         true iff a warning has been issued.
11757         * gimple.h (gimple_nonartificial_location): New function.
11758         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Call
11759         gimple_nonartificial_location and handle -Wno-system-headers.
11760         (handle_builtin_stxncpy): Same.
11762 2018-06-12  Martin Sebor  <msebor@redhat.com>
11764         PR c/85931
11765         * fold-const.c (operand_equal_p): Handle SAVE_EXPR.
11767 2018-06-12  Will Schmidt  <will_schmidt@vnet.ibm.com>
11769         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
11770         BUILTIN_VEC_XST entries for pointer to double and long long.
11772 2018-06-12  H.J. Lu  <hongjiu.lu@intel.com>
11774         PR target/85990
11775         * config/i386/gnu-user.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
11776         Update comments.
11777         * config/i386/gnu-user64.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
11778         Likewise.
11780 2018-06-12  Martin Liska  <mliska@suse.cz>
11782         * doc/options.texi: Document IntegerRange.
11784 2018-06-12  Martin Liska  <mliska@suse.cz>
11786         * config/i386/i386.opt: Make MPX-related options as Deprecated.
11787         * opt-functions.awk: Handle Deprecated flag.
11788         * opts-common.c (decode_cmdline_option): Handle cl_deprecated
11789         and report error.
11790         (read_cmdline_option): Report warning for a deprecated option.
11791         * opts.h (struct cl_option): Add new field cl_deprecated.
11792         (CL_ERR_DEPRECATED): New.
11794 2018-06-12  Martin Liska  <mliska@suse.cz>
11796         * doc/options.texi: Document Deprecated option flag.
11798 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
11800         * config/arc/arc-arch.h (arc_extras): New enum.
11801         (arc_cpu_t):Add field extra.
11802         (arc_cpu_types): Consider the extras.
11803         * config/arc/arc-cpus.def: Add extras info.
11804         * config/arc/arc-opts.h (processor_type): Consider extra field.
11805         * config/arc/arc.c (arc_override_options): Handle extra field.
11807 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
11809         * config/arc/arc-arch.h: Update ARC_OPTX macro.
11810         * config/arc/arc-options.def (ARC_OPTX): Introduce a new doc
11811         field.
11812         * config/arc/arc.c (arc_init): Update pic warning.
11813         (irq_range): Update irq range parsing warnings.
11814         (arc_override_options): Update various warning messages.
11815         (arc_handle_aux_attribute): Likewise.
11817 2018-06-12  Robert Suchanek  <robert.suchanek@mips.com>
11819         * config/mips/i6400.md (i6400_fpu_fadd): Remove frint.
11821 2018-06-12  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
11823         * doc/sourcebuild.texi: Document usage of line number 0 in verify
11824         compiler messages directives.
11826 2018-06-12  Matthew Fortune  <mfortune@gmail.com>
11828         * config/mips/mips-cpus.def: New MIPS_CPU for i6500.
11829         * config/mips/mips-tables.opt: Regenerate.
11830         * config/mips/mips.h (MIPS_ISA_LEVEL_SPEC): Mark i6500 as
11831         mips64r6.
11832         * doc/invoke.texi: Document -march=i6500.
11834 2018-06-12  Prachi Godbole  <prachi.godbole@imgtec.com>
11836         * config/mips/i6400.md (i6400_gpmuldiv): Remove cpu_unit.
11837         (i6400_gpmul): Add cpu_unit.
11838         (i6400_gpdiv): Likewise.
11839         (i6400_msa_add_d): Update reservations.
11840         (i6400_msa_int_add) Likewise.
11841         (i6400_msa_short_logic3) Likewise.
11842         (i6400_msa_short_logic2) Likewise.
11843         (i6400_msa_short_logic) Likewise.
11844         (i6400_msa_move) Likewise.
11845         (i6400_msa_cmp) Likewise.
11846         (i6400_msa_short_float2) Likewise.
11847         (i6400_msa_div_d) Likewise.
11848         (i6400_msa_long_logic1) Likewise.
11849         (i6400_msa_long_logic2) Likewise.
11850         (i6400_msa_mult) Likewise.
11851         (i6400_msa_long_float2) Likewise.
11852         (i6400_msa_long_float4) Likewise.
11853         (i6400_msa_long_float5) Likewise.
11854         (i6400_msa_long_float8) Likewise.
11855         (i6400_fpu_fadd): Include frint type.
11856         (i6400_fpu_store): New define_insn_reservation.
11857         (i6400_fpu_load): Likewise.
11858         (i6400_fpu_move): Likewise.
11859         (i6400_fpu_fcmp): Likewise.
11860         (i6400_fpu_fmadd): Likewise.
11861         (i6400_int_mult): Include imul3nc type and update reservation.
11862         (i6400_int_div): Include idiv3 type and update reservation.
11863         (i6400_int_load): Update to check type not move_type.
11864         (i6400_int_store): Likewise.
11865         (i6400_int_prefetch): Set zero latency.
11867 2018-06-12  Eric Botcazou  <ebotcazou@adacore.com>
11869         * gcc.c: Document new %@{...} sequence.
11870         (LINK_COMMAND_SPEC): Use it for the -L switches.
11871         (cpp_unique_options): Use it for the -I switches.
11872         (at_file_argbuf): New global variable.
11873         (in_at_file): Likewise.
11874         (alloc_args): Create at_file_argbuf.
11875         (clear_args): Truncate at_file_argbuf.
11876         (store_arg): If in_at_file, push the argument onto at_file_argbuf.
11877         (open_at_file): New function.
11878         (close_at_file): Likewise.
11879         (create_at_file): Delete.
11880         (do_spec_1) <'i'>: Use open_at_file/close_at_file.
11881         <'o'>: Likewise.
11882         <'@'>: New case.
11883         (validate_switches_from_spec): Deal with %@{...} sequence.
11884         (validate_switches): Likewise.
11885         (driver::finalize): Call clear_args.
11887 2018-06-11  Rasmus Villemoes  <rasmus.villemoes@prevas.dk>
11889         * config/vx-common.h (USE_TM_CLONE_REGISTRY): #define to 0.
11891 2018-06-11  Martin Sebor  <msebor@redhat.com>
11893         * doc/invoke.texi (-Wall): List -Wc++17-compat.
11894         (Wno-class-memaccess): Add @opindex.
11895         (Wno-templates, Wno-multiple-inheritance): Same.
11896         (Wno-virtual-inheritance, Wno-namespaces): Same.
11897         (Wno-pedantic, Wno-chkp, Wcoverage-mismatch): Same.
11898         (Wno-format-overflow, Wno-format-truncation): Same.
11899         (Wno-unused-local-typedefs, Walloc-size-larger-than): Same.
11900         (Wno-alloc-size-larger-than, Wframe-larger-than): Same
11901         (Wno-stack-usage, Wno-c++-compat, Wno-c++11-compat): Same.
11902         (Wno-c++14-compat, Wno-c++17-compat, Wno-openmp-simd): Same.
11903         (Wno-unsuffixed-float-constants, Wno-addr-space-convert): Same.
11904         (Wno-misspelled-isr): Same.
11906 2018-06-11  Martin Sebor  <msebor@redhat.com>
11908         * PR tree-optimization/86083
11909         * tree-ssa-strlen.c (handle_char_store): Use tree_expr_nonzero_p.
11911 2018-06-11  Zhouyi Zhou <zhouzhouyi@gmail.com>
11913         * tree-eh.c (lower_eh_constructs_2): Add a comma to comment.
11915 2018-06-11  Segher Boessenkool  <segher@kernel.crashing.org>
11917         PR target/85755
11918         * config/rs6000/rs6000.md (*movdi_internal32): Put constraint modifiers
11919         on the correct operand.
11920         (*movdi_internal64): Ditto.
11922 2018-06-11  Martin Liska  <mliska@suse.cz>
11924         PR tree-optimization/86089
11925         * tree-ssa-strlen.c (get_string_length): Move back removed hunk.
11927 2018-06-11  Julia Koval  <julia.koval@intel.com>
11929         * config/i386/vaesintrin.h (_mm_aesdec_epi128, _mm_aesdeclast_epi128)
11930         _mm_aesenc_epi128, _mm_aesenclast_epi128: Remove.
11931         * config/i386/vpclmulqdqintrin.h (_mm_clmulepi64_epi128): Remove.
11933 2018-06-11  Olivier Hainque  <hainque@adacore.com>
11935         * dwarf2out.c (gen_compile_unit_die): Fallout to DW_LANG_Ada83
11936         for Ada with strict dwarf2.
11938 2018-06-08  Peter Bergner  <bergner@vnet.ibm.com>
11940         PR target/85755
11941         * config/rs6000/rs6000.c (mem_operand_gpr): Enable PRE_INC and PRE_DEC
11942         addresses.
11944 2018-06-08  Jan Hubicka  <hubicka@ucw.cz>
11946         * dumpfile.c (FIRST_ME_AUTO_NUMBERED_DUMP): Bump to 4.
11948 2018-06-08  David Edelsohn  <dje.gcc@gmail.com>
11950         * config/rs6000/rs6000.c (rs6000_passes_ieee128): Protect with #if
11951         TARGET_ELF.
11953 2018-06-08  Martin Liska  <mliska@suse.cz>
11955         * tree-cfg.h (debug_function): Fix argument type to match
11956         implementation.
11958 2018-06-08  Martin Liska  <mliska@suse.cz>
11960         * config/powerpcspe/powerpcspe.c (rs6000_xcoff_visibility):
11961         Remove usage of MPX-related (and removed) fields.
11962         * config/rs6000/rs6000.c (rs6000_xcoff_visibility): Likewise.
11964 2018-06-08  David Malcolm  <dmalcolm@redhat.com>
11966         * cfg.c (debug): Use TDF_NONE rather than 0.
11967         * cfghooks.c (debug): Likewise.
11968         * dumpfile.c (DUMP_FILE_INFO): Likewise; also for OPTGROUP.
11969         (struct dump_option_value_info): Convert to...
11970         (struct kv_pair): ...this template type.
11971         (dump_options): Convert to kv_pair<dump_flags_t>; use TDF_NONE
11972         rather than 0.
11973         (optinfo_verbosity_options): Likewise.
11974         (optgroup_options): Convert to kv_pair<optgroup_flags_t>; use
11975         OPTGROUP_NONE.
11976         (gcc::dump_manager::dump_register): Use optgroup_flags_t rather
11977         than int for "optgroup_flags" param.
11978         (dump_generic_expr_loc): Use dump_flags_t rather than int for
11979         "dump_kind" param.
11980         (dump_dec): Likewise.
11981         (dump_finish): Use TDF_NONE rather than 0.
11982         (gcc::dump_manager::opt_info_enable_passes): Use optgroup_flags_t
11983         rather than int for "optgroup_flags" param.  Use TDF_NONE rather
11984         than 0.  Update for change to option_ptr.
11985         (opt_info_switch_p_1): Convert "optgroup_flags" param from int *
11986         to optgroup_flags_t *.  Use TDF_NONE and OPTGROUP_NONE rather than
11987         0.  Update for changes to optinfo_verbosity_options and
11988         optgroup_options.
11989         (opt_info_switch_p): Convert optgroup_flags from int to
11990         optgroup_flags_t.
11991         (dump_basic_block): Use dump_flags_t rather than int
11992         for "dump_kind" param.
11993         * dumpfile.h (TDF_ADDRESS, TDF_SLIM, TDF_RAW, TDF_DETAILS,
11994         TDF_STATS, TDF_BLOCKS, TDF_VOPS, TDF_LINENO, TDF_UID)
11995         TDF_STMTADDR, TDF_GRAPH, TDF_MEMSYMS, TDF_RHS_ONLY, TDF_ASMNAME,
11996         TDF_EH, TDF_NOUID, TDF_ALIAS, TDF_ENUMERATE_LOCALS, TDF_CSELIB,
11997         TDF_SCEV, TDF_GIMPLE, TDF_FOLDING, MSG_OPTIMIZED_LOCATIONS,
11998         MSG_MISSED_OPTIMIZATION, MSG_NOTE, MSG_ALL, TDF_COMPARE_DEBUG,
11999         TDF_NONE): Convert from macros to...
12000         (enum dump_flag): ...this new enum.
12001         (dump_flags_t): Update to use enum.
12002         (operator|, operator&, operator~, operator|=, operator&=):
12003         Implement for dump_flags_t.
12004         (OPTGROUP_NONE, OPTGROUP_IPA, OPTGROUP_LOOP, OPTGROUP_INLINE,
12005         OPTGROUP_OMP, OPTGROUP_VEC, OPTGROUP_OTHER, OPTGROUP_ALL):
12006         Convert from macros to...
12007         (enum optgroup_flag): ...this new enum.
12008         (optgroup_flags_t): New typedef.
12009         (operator|, operator|=): Implement for optgroup_flags_t.
12010         (struct dump_file_info): Convert field "alt_flags" to
12011         dump_flags_t.  Convert field "optgroup_flags" to
12012         optgroup_flags_t.
12013         (dump_basic_block): Use dump_flags_t rather than int for param.
12014         (dump_generic_expr_loc): Likewise.
12015         (dump_dec): Likewise.
12016         (dump_register): Convert param "optgroup_flags" to
12017         optgroup_flags_t.
12018         (opt_info_enable_passes): Likewise.
12019         * early-remat.c (early_remat::dump_edge_list): Use TDF_NONE rather
12020         than 0.
12021         * gimple-pretty-print.c (debug): Likewise.
12022         * gimple-ssa-store-merging.c (bswap_replace): Likewise.
12023         (merged_store_group::apply_stores): Likewise.
12024         * gimple-ssa-strength-reduction.c (insert_initializers): Likewise.
12025         * gimple.c (verify_gimple_pp): Likewise.
12026         * graphite-poly.c (print_pbb_body): Likewise.
12027         * passes.c (pass_manager::register_one_dump_file): Convert
12028         local "optgroup_flags" to optgroup_flags_t.
12029         * print-tree.c (print_node): Use TDF_NONE rather than 0.
12030         (debug): Likewise.
12031         (debug_body): Likewise.
12032         * tree-pass.h (struct pass_data): Convert field "optgroup_flags"
12033         to optgroup_flags_t.
12034         * tree-pretty-print.c (print_struct_decl): Use TDF_NONE rather
12035         than 0.
12036         * tree-ssa-math-opts.c (convert_mult_to_fma_1): Likewise.
12037         (convert_mult_to_fma): Likewise.
12038         * tree-ssa-reassoc.c (undistribute_ops_list): Likewise.
12039         * tree-ssa-sccvn.c (vn_eliminate): Likewise.
12040         * tree-vect-data-refs.c (dump_lower_bound): Convert param
12041         "dump_kind" to dump_flags_t.
12043 2018-06-08  Segher Boessenkool  <segher@kernel.crashing.org>
12045         * config/rs6000/rs6000.c (min, max): Delete.
12047 2018-06-08  Segher Boessenkool  <segher@kernel.crashing.org>
12049         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete mention of
12050         -mabi=spe and -mabi=no-spe.
12052 2018-06-08  Martin Liska  <mliska@suse.cz>
12054         * ipa-pure-const.c (propagate_pure_const): Use ::get at places
12055         where we expect an existing summary.
12057 2018-06-08  Martin Liska  <mliska@suse.cz>
12059         * ipa-inline-analysis.c (simple_edge_hints): Use ::get method.
12060         * ipa-inline.h (estimate_edge_growth): Likewise.
12062 2018-06-08  Martin Liska  <mliska@suse.cz>
12064         * cgraph.c (function_version_hasher::hash): Use
12065         cgraph_node::get_uid ().
12066         (function_version_hasher::equal):
12067         * cgraph.h (cgraph_node::get_uid): New method.
12068         * ipa-inline.c (update_caller_keys): Use
12069         cgraph_node::get_uid ().
12070         (update_callee_keys): Likewise.
12071         * ipa-utils.c (searchc): Likewise.
12072         (ipa_reduced_postorder): Likewise.
12073         * lto-cgraph.c (input_node): Likewise.
12074         * passes.c (is_pass_explicitly_enabled_or_disabled): Likewise.
12075         * symbol-summary.h (symtab_insertion): Likewise.
12076         (symtab_removal): Likewise.
12077         (symtab_duplication): Likewise.
12078         * tree-pretty-print.c (dump_function_header): Likewise.
12079         * tree-sra.c (convert_callers_for_node): Likewise.
12081 2018-06-08  Martin Liska  <mliska@suse.cz>
12083         * cgraph.c (symbol_table::create_edge): Always assign a new
12084         unique number.
12085         (symbol_table::free_edge): Do not recycle numbers.
12086         * cgraph.h (cgraph_edge::get): New method.
12087         * symbol-summary.h (symtab_removal): Use it.
12088         (symtab_duplication): Likewise.
12089         (call_summary::hashable_uid): Remove.
12091 2018-06-08  Martin Liska  <mliska@suse.cz>
12093         * ipa-inline-analysis.c (inline_edge_removal_hook): Remove.
12094         (initialize_growth_caches): Remove.
12095         (free_growth_caches): Likewise.
12096         (do_estimate_edge_time): Use edge_growth_cache.
12097         (do_estimate_edge_size): Likewise.
12098         (do_estimate_edge_hints): Likewise.
12099         * ipa-inline.c (reset_edge_caches): Likewise.
12100         (recursive_inlining): Likewise.
12101         (inline_small_functions): Likewise.
12102         * ipa-inline.h (initialize_growth_caches): Remove.
12103         (estimate_edge_size): Likewise.
12104         (estimate_edge_time): Likewise.
12105         (estimate_edge_hints): Likewise.
12106         (reset_edge_growth_cache): Likewise.
12107         * symbol-summary.h (call_summary::remove): New method.
12109 2018-06-08  Martin Liska  <mliska@suse.cz>
12111         * ipa-cp.c (class edge_clone_summary): New summary.
12112         (grow_edge_clone_vectors): Remove.
12113         (ipcp_edge_duplication_hook): Remove.
12114         (class edge_clone_summary_t): New call_summary class.
12115         (ipcp_edge_removal_hook): Remove.
12116         (edge_clone_summary_t::duplicate): New function.
12117         (get_next_cgraph_edge_clone): Use edge_clone_summaries.
12118         (create_specialized_node): Likewise.
12119         (ipcp_driver): Initialize edge_clone_summaries and do not
12120         register hooks.
12122 2018-06-08  Martin Liska  <mliska@suse.cz>
12124         * symbol-summary.h (get): New function.
12125         (call_summary::m_initialize_when_cloning): New class member.
12127 2018-06-08  Martin Liska  <mliska@suse.cz>
12129         * cgraph.c (cgraph_node::remove): Do not recycle uid.
12130         * cgraph.h (symbol_table::release_symbol): Do not pass uid.
12131         (symbol_table::allocate_cgraph_symbol): Do not set uid.
12132         * passes.c (uid_hash_t): Record removed_nodes by their uids.
12133         (remove_cgraph_node_from_order): Use the removed_nodes set.
12134         (do_per_function_toporder): Likwise.
12135         * symbol-summary.h (symtab_insertion): Use cgraph_node::uid
12136         instead of summary_uid.
12137         (symtab_removal): Likewise.
12138         (symtab_duplication): Likewise.
12140 2018-06-08  Martin Liska  <mliska@suse.cz>
12142         * ipa-cp.c (ipcp_store_bits_results): Use
12143         ipcp_transformation_sum.
12144         (ipcp_store_vr_results): Likewise.
12145         * ipa-prop.c (ipcp_grow_transformations_if_necessary): Renamed
12146         to ...
12147         (ipcp_transformation_initialize): ... this.
12148         (ipa_set_node_agg_value_chain):
12149         (ipa_node_params_t::duplicate): Use ipcp_transformation_sum.
12150         (write_ipcp_transformation_info): Likewise.
12151         (read_ipcp_transformation_info): Likewise.
12152         (ipcp_update_bits): Likewise.
12153         (ipcp_update_vr): Likewise.
12154         (ipcp_transform_function): Likewise.
12155         * ipa-prop.h: Rename ipcp_transformation_summary to
12156         ipcp_transformation.
12157         (class ipcp_transformation_t): New function summary.
12158         (ipcp_get_transformation_summary): Use ipcp_transformation_sum.
12159         (ipa_get_agg_replacements_for_node): Likewise.
12161 2018-06-08  Martin Liska  <mliska@suse.cz>
12163         * ipa-pure-const.c (struct funct_state_d): Do it class instead
12164         of struct.
12165         (class funct_state_summary_t): New function_summary class.
12166         (has_function_state): Remove.
12167         (get_function_state): Likewise.
12168         (set_function_state): Likewise.
12169         (add_new_function): Likewise.
12170         (funct_state_summary_t::insert): New function.
12171         (duplicate_node_data): Remove.
12172         (remove_node_data): Remove.
12173         (funct_state_summary_t::duplicate): New function.
12174         (register_hooks): Create new funct_state_summaries.
12175         (pure_const_generate_summary): Use it.
12176         (pure_const_write_summary): Likewise.
12177         (pure_const_read_summary): Likewise.
12178         (propagate_pure_const): Likewise.
12179         (propagate_nothrow): Likewise.
12180         (dump_malloc_lattice): Likewise.
12181         (propagate_malloc): Likewise.
12182         (execute): Do not register hooks, just remove summary
12183         instead.
12184         (pass_ipa_pure_const::pass_ipa_pure_const): Simplify
12185         constructor.
12187 2018-06-08  Martin Liska  <mliska@suse.cz>
12189         * ipa-reference.c (remove_node_data): Remove.
12190         (duplicate_node_data): Likewise.
12191         (class ipa_ref_var_info_summary_t): New class.
12192         (class ipa_ref_opt_summary_t): Likewise.
12193         (get_reference_vars_info): Use ipa_ref_var_info_summaries.
12194         (get_reference_optimization_summary): Use
12195         ipa_ref_opt_sum_summaries.
12196         (set_reference_vars_info): Remove.
12197         (set_reference_optimization_summary): Likewise.
12198         (ipa_init): Create summaries.
12199         (init_function_info): Use function summary.
12200         (ipa_ref_opt_summary_t::duplicate): New function.
12201         (ipa_ref_opt_summary_t::remove): New function.
12202         (get_read_write_all_from_node): Fix GNU coding style.
12203         (propagate): Use function summary.
12204         (write_node_summary_p): Fix GNU coding style.
12205         (stream_out_bitmap): Likewise.
12206         (ipa_reference_read_optimization_summary): Use function summary.
12207         (ipa_reference_c_finalize): Do not release hooks.
12209 2018-06-08  Martin Liska  <mliska@suse.cz>
12211         * ipa-fnsummary.c (dump_ipa_call_summary): Use ::get method.
12212         (analyze_function_body): Extract multiple calls of get_create.
12213         * ipa-inline-analysis.c (simple_edge_hints): Likewise.
12214         * ipa-inline.c (recursive_inlining): Use ::get method.
12215         * ipa-inline.h (estimate_edge_growth): Likewise.
12217 2018-06-08  Martin Liska  <mliska@suse.cz>
12219         * hsa-common.h (enum hsa_function_kind): Rename HSA_NONE to
12220         HSA_INVALID.
12221         (hsa_function_summary::hsa_function_summary): Use the new enum
12222         value.
12223         (hsa_gpu_implementation_p): Use hsa_summaries::get.
12224         * hsa-gen.c (hsa_get_host_function): Likewise.
12225         (get_brig_function_name): Likewise.
12226         * ipa-hsa.c (process_hsa_functions): Likewise.
12227         (ipa_hsa_write_summary): Likewise.
12228         * symbol-summary.h (symtab_duplication): Use ::get function/
12229         (get): New function.
12231 2018-06-08  Martin Liska  <mliska@suse.cz>
12233         * config/i386/i386.c (ix86_can_inline_p): Use get_create instead
12234         of get.
12235         * hsa-common.c (hsa_summary_t::link_functions): Likewise.
12236         (hsa_register_kernel): Likewise.
12237         * hsa-common.h (hsa_gpu_implementation_p): Likewise.
12238         * hsa-gen.c (hsa_get_host_function): Likewise.
12239         (get_brig_function_name): Likewise.
12240         (generate_hsa): Likewise.
12241         (pass_gen_hsail::execute): Likewise.
12242         * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
12243         (devirtualization_time_bonus): Likewise.
12244         (ipcp_propagate_stage): Likewise.
12245         * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
12246         (edge_set_predicate): Likewise.
12247         (evaluate_conditions_for_known_args): Likewise.
12248         (evaluate_properties_for_edge): Likewise.
12249         (ipa_fn_summary::reset): Likewise.
12250         (ipa_fn_summary_t::duplicate): Likewise.
12251         (dump_ipa_call_summary): Likewise.
12252         (ipa_dump_fn_summary): Likewise.
12253         (analyze_function_body): Likewise.
12254         (compute_fn_summary): Likewise.
12255         (estimate_edge_devirt_benefit): Likewise.
12256         (estimate_edge_size_and_time): Likewise.
12257         (estimate_calls_size_and_time): Likewise.
12258         (estimate_node_size_and_time): Likewise.
12259         (inline_update_callee_summaries): Likewise.
12260         (remap_edge_change_prob): Likewise.
12261         (remap_edge_summaries): Likewise.
12262         (ipa_merge_fn_summary_after_inlining): Likewise.
12263         (ipa_update_overall_fn_summary): Likewise.
12264         (read_ipa_call_summary): Likewise.
12265         (inline_read_section): Likewise.
12266         (write_ipa_call_summary): Likewise.
12267         (ipa_fn_summary_write): Likewise.
12268         (ipa_free_fn_summary): Likewise.
12269         * ipa-hsa.c (process_hsa_functions): Likewise.
12270         (ipa_hsa_write_summary): Likewise.
12271         (ipa_hsa_read_section): Likewise.
12272         * ipa-icf.c (sem_function::merge): Likewise.
12273         * ipa-inline-analysis.c (simple_edge_hints): Likewise.
12274         (do_estimate_edge_time): Likewise.
12275         (estimate_size_after_inlining): Likewise.
12276         (estimate_growth): Likewise.
12277         (growth_likely_positive): Likewise.
12278         * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
12279         (inline_call): Likewise.
12280         * ipa-inline.c (caller_growth_limits): Likewise.
12281         (can_inline_edge_p): Likewise.
12282         (can_inline_edge_by_limits_p): Likewise.
12283         (compute_uninlined_call_time): Likewise.
12284         (compute_inlined_call_time): Likewise.
12285         (want_inline_small_function_p): Likewise.
12286         (edge_badness): Likewise.
12287         (update_caller_keys): Likewise.
12288         (update_callee_keys): Likewise.
12289         (recursive_inlining): Likewise.
12290         (inline_small_functions): Likewise.
12291         (inline_to_all_callers_1): Likewise.
12292         (dump_overall_stats): Likewise.
12293         (early_inline_small_functions): Likewise.
12294         (early_inliner): Likewise.
12295         * ipa-inline.h (estimate_edge_growth): Likewise.
12296         * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
12297         * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
12298         * ipa-prop.h (IPA_NODE_REF): Likewise.
12299         (IPA_EDGE_REF): Likewise.
12300         * ipa-pure-const.c (malloc_candidate_p): Likewise.
12301         (propagate_malloc): Likewise.
12302         * ipa-split.c (execute_split_functions): Likewise.
12303         * symbol-summary.h: Rename get to get_create.
12304         (get): Likewise.
12305         (get_create): Likewise.
12306         * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
12308 2018-06-08  Martin Liska  <mliska@suse.cz>
12310         * symbol-summary.h (release): Move definition out of class
12311         declaration.
12312         (symtab_removal): Likewise.
12313         (symtab_duplication): Likewise.
12315 2018-06-08  Martin Liska  <mliska@suse.cz>
12317         * symbol-summary.h (function_summary): Move constructor
12318         implementation out of class declaration.
12319         (release): Likewise.
12320         (symtab_insertion): Likewise.
12321         (symtab_removal): Likewise.
12322         (symtab_duplication): Likewise.
12323         (get): Likewise.
12325 2018-06-08  Martin Liska  <mliska@suse.cz>
12327         * Makefile.in: Remove support for MPX (macros, related functions,
12328         fields in cgraph_node, ...).
12329         * builtin-types.def (BT_BND): Likewise.
12330         (BT_FN_BND_CONST_PTR): Likewise.
12331         (BT_FN_CONST_PTR_BND): Likewise.
12332         (BT_FN_VOID_PTR_BND): Likewise.
12333         (BT_FN_BND_CONST_PTR_SIZE): Likewise.
12334         (BT_FN_VOID_CONST_PTR_BND_CONST_PTR): Likewise.
12335         * builtins.c (expand_builtin_memcpy_with_bounds): Likewise.
12336         (expand_builtin_mempcpy_with_bounds): Likewise.
12337         (expand_builtin_memset_with_bounds): Likewise.
12338         (expand_builtin_memset_args): Likewise.
12339         (std_expand_builtin_va_start): Likewise.
12340         (expand_builtin): Likewise.
12341         (expand_builtin_with_bounds): Likewise.
12342         * builtins.def (DEF_BUILTIN_CHKP): Likewise.
12343         (DEF_LIB_BUILTIN_CHKP): Likewise.
12344         (DEF_EXT_LIB_BUILTIN_CHKP): Likewise.
12345         (DEF_CHKP_BUILTIN): Likewise.
12346         (BUILT_IN_MEMCPY): Likewise.
12347         (BUILT_IN_MEMMOVE): Likewise.
12348         (BUILT_IN_MEMPCPY): Likewise.
12349         (BUILT_IN_MEMSET): Likewise.
12350         (BUILT_IN_STPCPY): Likewise.
12351         (BUILT_IN_STRCAT): Likewise.
12352         (BUILT_IN_STRCHR): Likewise.
12353         (BUILT_IN_STRCPY): Likewise.
12354         (BUILT_IN_STRLEN): Likewise.
12355         (BUILT_IN_MEMCPY_CHK): Likewise.
12356         (BUILT_IN_MEMMOVE_CHK): Likewise.
12357         (BUILT_IN_MEMPCPY_CHK): Likewise.
12358         (BUILT_IN_MEMSET_CHK): Likewise.
12359         (BUILT_IN_STPCPY_CHK): Likewise.
12360         (BUILT_IN_STRCAT_CHK): Likewise.
12361         (BUILT_IN_STRCPY_CHK): Likewise.
12362         * calls.c (store_bounds): Likewise.
12363         (emit_call_1): Likewise.
12364         (special_function_p): Likewise.
12365         (maybe_warn_nonstring_arg): Likewise.
12366         (initialize_argument_information): Likewise.
12367         (finalize_must_preallocate): Likewise.
12368         (compute_argument_addresses): Likewise.
12369         (expand_call): Likewise.
12370         * cfgexpand.c (expand_call_stmt): Likewise.
12371         (expand_return): Likewise.
12372         (expand_gimple_stmt_1): Likewise.
12373         (pass_expand::execute): Likewise.
12374         * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
12375         (cgraph_node::remove): Likewise.
12376         (cgraph_node::dump): Likewise.
12377         (cgraph_node::verify_node): Likewise.
12378         * cgraph.h (chkp_function_instrumented_p): Likewise.
12379         (symtab_node::get_alias_target): Likewise.
12380         (cgraph_node::can_remove_if_no_direct_calls_and_refs_p): Likewise.
12381         (cgraph_local_p): Likewise.
12382         * cgraphbuild.c (cgraph_edge::rebuild_edges): Likewise.
12383         (cgraph_edge::rebuild_references): Likewise.
12384         * cgraphunit.c (varpool_node::finalize_decl): Likewise.
12385         (walk_polymorphic_call_targets): Likewise.
12386         (cgraph_node::expand_thunk): Likewise.
12387         (symbol_table::output_weakrefs): Likewise.
12388         * common/config/i386/i386-common.c (OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET): Likewise.
12389         (ix86_handle_option): Likewise.
12390         * config/i386/constraints.md: Likewise.
12391         * config/i386/i386-builtin-types.def (BND): Likewise.
12392         (VOID): Likewise.
12393         (PVOID): Likewise.
12394         (ULONG): Likewise.
12395         * config/i386/i386-builtin.def (BDESC_END): Likewise.
12396         (BDESC_FIRST): Likewise.
12397         (BDESC): Likewise.
12398         * config/i386/i386-c.c (ix86_target_macros_internal): Likewise.
12399         * config/i386/i386-protos.h (ix86_bnd_prefixed_insn_p): Likewise.
12400         * config/i386/i386.c (enum reg_class): Likewise.
12401         (ix86_target_string): Likewise.
12402         (ix86_option_override_internal): Likewise.
12403         (ix86_conditional_register_usage): Likewise.
12404         (ix86_valid_target_attribute_inner_p): Likewise.
12405         (ix86_set_indirect_branch_type): Likewise.
12406         (ix86_set_current_function): Likewise.
12407         (ix86_function_arg_regno_p): Likewise.
12408         (init_cumulative_args): Likewise.
12409         (ix86_function_arg_advance): Likewise.
12410         (ix86_function_arg): Likewise.
12411         (ix86_pass_by_reference): Likewise.
12412         (ix86_function_value_regno_p): Likewise.
12413         (ix86_function_value_1): Likewise.
12414         (ix86_function_value_bounds): Likewise.
12415         (ix86_return_in_memory): Likewise.
12416         (ix86_setup_incoming_vararg_bounds): Likewise.
12417         (ix86_va_start): Likewise.
12418         (indirect_thunk_need_prefix): Likewise.
12419         (print_reg): Likewise.
12420         (ix86_print_operand): Likewise.
12421         (ix86_expand_call): Likewise.
12422         (ix86_output_function_return): Likewise.
12423         (reg_encoded_number): Likewise.
12424         (BDESC_VERIFYS): Likewise.
12425         (ix86_init_mpx_builtins): Likewise.
12426         (ix86_init_builtins): Likewise.
12427         (ix86_emit_cmove): Likewise.
12428         (ix86_emit_move_max): Likewise.
12429         (ix86_expand_builtin): Likewise.
12430         (ix86_builtin_mpx_function): Likewise.
12431         (ix86_get_arg_address_for_bt): Likewise.
12432         (ix86_load_bounds): Likewise.
12433         (ix86_store_bounds): Likewise.
12434         (ix86_load_returned_bounds): Likewise.
12435         (ix86_store_returned_bounds): Likewise.
12436         (ix86_class_likely_spilled_p): Likewise.
12437         (ix86_hard_regno_mode_ok): Likewise.
12438         (x86_order_regs_for_local_alloc): Likewise.
12439         (ix86_mitigate_rop): Likewise.
12440         (ix86_bnd_prefixed_insn_p): Likewise.
12441         (ix86_mpx_bound_mode): Likewise.
12442         (ix86_make_bounds_constant): Likewise.
12443         (ix86_initialize_bounds): Likewise.
12444         (TARGET_LOAD_BOUNDS_FOR_ARG): Likewise.
12445         (TARGET_STORE_BOUNDS_FOR_ARG): Likewise.
12446         (TARGET_LOAD_RETURNED_BOUNDS): Likewise.
12447         (TARGET_STORE_RETURNED_BOUNDS): Likewise.
12448         (TARGET_CHKP_BOUND_MODE): Likewise.
12449         (TARGET_BUILTIN_CHKP_FUNCTION): Likewise.
12450         (TARGET_CHKP_FUNCTION_VALUE_BOUNDS): Likewise.
12451         (TARGET_CHKP_MAKE_BOUNDS_CONSTANT): Likewise.
12452         (TARGET_CHKP_INITIALIZE_BOUNDS): Likewise.
12453         * config/i386/i386.h (TARGET_MPX): Likewise.
12454         (TARGET_MPX_P): Likewise.
12455         (VALID_BND_REG_MODE): Likewise.
12456         (FIRST_BND_REG): Likewise.
12457         (LAST_BND_REG): Likewise.
12458         (enum reg_class): Likewise.
12459         (BND_REG_P): Likewise.
12460         (BND_REGNO_P): Likewise.
12461         (BNDmode): Likewise.
12462         (ADJUST_INSN_LENGTH): Likewise.
12463         * config/i386/i386.md: Likewise.
12464         * config/i386/i386.opt: Likewise.
12465         * config/i386/linux-common.h (LIBMPX_LIBS): Likewise.
12466         (defined): Likewise.
12467         (LINK_MPX): Likewise.
12468         (MPX_SPEC): Likewise.
12469         (LIBMPX_SPEC): Likewise.
12470         (LIBMPXWRAPPERS_SPEC): Likewise.
12471         (CHKP_SPEC): Likewise.
12472         * config/i386/predicates.md: Likewise.
12473         * dbxout.c (dbxout_type): Likewise.
12474         * doc/extend.texi: Likewise.
12475         * doc/invoke.texi: Likewise.
12476         * doc/md.texi: Likewise.
12477         * doc/tm.texi: Likewise.
12478         * doc/tm.texi.in: Likewise.
12479         * dwarf2out.c (is_base_type): Likewise.
12480         (gen_formal_types_die): Likewise.
12481         (gen_subprogram_die): Likewise.
12482         (gen_type_die_with_usage): Likewise.
12483         (gen_decl_die): Likewise.
12484         (dwarf2out_late_global_decl): Likewise.
12485         * expr.c (expand_assignment): Likewise.
12486         (emit_storent_insn): Likewise.
12487         (store_expr_with_bounds): Likewise.
12488         (store_expr): Likewise.
12489         (expand_expr_real_1): Likewise.
12490         * expr.h (store_expr_with_bounds): Likewise.
12491         * function.c (use_register_for_decl): Likewise.
12492         (struct bounds_parm_data): Likewise.
12493         (assign_parms_augmented_arg_list): Likewise.
12494         (assign_parm_find_entry_rtl): Likewise.
12495         (assign_parm_is_stack_parm): Likewise.
12496         (assign_parm_load_bounds): Likewise.
12497         (assign_bounds): Likewise.
12498         (assign_parms): Likewise.
12499         (expand_function_start): Likewise.
12500         * gcc.c (CHKP_SPEC): Likewise.
12501         * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
12502         * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Likewise.
12503         (wrestrict_dom_walker::check_call): Likewise.
12504         * gimple.c (gimple_build_call_from_tree): Likewise.
12505         * gimple.h (enum gf_mask): Likewise.
12506         (gimple_call_with_bounds_p): Likewise.
12507         (gimple_call_set_with_bounds): Likewise.
12508         * gimplify.c (gimplify_init_constructor): Likewise.
12509         * ipa-cp.c (initialize_node_lattices): Likewise.
12510         (propagate_constants_across_call): Likewise.
12511         (find_more_scalar_values_for_callers_subset): Likewise.
12512         * ipa-hsa.c (process_hsa_functions): Likewise.
12513         * ipa-icf-gimple.c (func_checker::compare_gimple_call): Likewise.
12514         * ipa-icf.c (sem_function::merge): Likewise.
12515         * ipa-inline.c (early_inliner): Likewise.
12516         * ipa-pure-const.c (warn_function_noreturn): Likewise.
12517         (warn_function_cold): Likewise.
12518         (propagate_pure_const): Likewise.
12519         * ipa-ref.h (enum GTY): Likewise.
12520         * ipa-split.c (find_retbnd): Likewise.
12521         (consider_split): Likewise.
12522         (split_function): Likewise.
12523         * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
12524         * ipa.c (walk_polymorphic_call_targets): Likewise.
12525         (symbol_table::remove_unreachable_nodes): Likewise.
12526         (process_references): Likewise.
12527         (cgraph_build_static_cdtor_1): Likewise.
12528         * lto-cgraph.c (lto_output_node): Likewise.
12529         (output_refs): Likewise.
12530         (compute_ltrans_boundary): Likewise.
12531         (input_overwrite_node): Likewise.
12532         (input_node): Likewise.
12533         (input_cgraph_1): Likewise.
12534         * params.def (PARAM_CHKP_MAX_CTOR_SIZE): Likewise.
12535         * passes.c (pass_manager::execute_early_local_passes): Likewise.
12536         (class pass_chkp_instrumentation_passes): Likewise.
12537         (make_pass_chkp_instrumentation_passes): Likewise.
12538         * passes.def: Likewise.
12539         * rtl.h (struct GTY): Likewise.
12540         (CALL_EXPR_WITH_BOUNDS_P): Likewise.
12541         * stor-layout.c (layout_type): Likewise.
12542         * symtab.c: Likewise.
12543         * target.def: Likewise.
12544         * targhooks.c (default_chkp_bound_type): Likewise.
12545         (default_chkp_bound_mode): Likewise.
12546         (default_builtin_chkp_function): Likewise.
12547         (default_chkp_function_value_bounds): Likewise.
12548         (default_chkp_make_bounds_constant): Likewise.
12549         (default_chkp_initialize_bounds): Likewise.
12550         * targhooks.h (default_chkp_bound_type): Likewise.
12551         (default_chkp_bound_mode): Likewise.
12552         (default_builtin_chkp_function): Likewise.
12553         (default_chkp_function_value_bounds): Likewise.
12554         (default_chkp_make_bounds_constant): Likewise.
12555         (default_chkp_initialize_bounds): Likewise.
12556         * toplev.c (compile_file): Likewise.
12557         (process_options): Likewise.
12558         * tree-core.h (DEF_BUILTIN): Likewise.
12559         (DEF_BUILTIN_CHKP): Likewise.
12560         * tree-inline.c (declare_return_variable): Likewise.
12561         (remap_gimple_stmt): Likewise.
12562         (copy_bb): Likewise.
12563         (initialize_inlined_parameters): Likewise.
12564         (expand_call_inline): Likewise.
12565         * tree-pass.h (make_pass_ipa_chkp_versioning): Likewise.
12566         (make_pass_ipa_chkp_early_produce_thunks): Likewise.
12567         (make_pass_ipa_chkp_produce_thunks): Likewise.
12568         (make_pass_chkp): Likewise.
12569         (make_pass_chkp_opt): Likewise.
12570         (make_pass_chkp_instrumentation_passes): Likewise.
12571         * tree-pretty-print.c (dump_generic_node): Likewise.
12572         * tree-ssa-ccp.c (insert_clobber_before_stack_restore): Likewise.
12573         * tree-ssa-dce.c (propagate_necessity): Likewise.
12574         (eliminate_unnecessary_stmts): Likewise.
12575         * tree-ssa-pre.c (create_expression_by_pieces): Likewise.
12576         * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
12577         * tree-ssa-sccvn.h: Likewise.
12578         * tree-ssa-strlen.c (get_string_length): Likewise.
12579         (valid_builtin_call): Likewise.
12580         (adjust_last_stmt): Likewise.
12581         (handle_builtin_strchr): Likewise.
12582         (handle_builtin_strcpy): Likewise.
12583         (handle_builtin_stxncpy): Likewise.
12584         (handle_builtin_memcpy): Likewise.
12585         (handle_builtin_strcat): Likewise.
12586         (strlen_check_and_optimize_stmt): Likewise.
12587         * tree-stdarg.c (expand_ifn_va_arg_1): Likewise.
12588         * tree-streamer-in.c: Likewise.
12589         * tree-streamer.c (record_common_node): Likewise.
12590         * tree.c (tree_code_size): Likewise.
12591         (wide_int_to_tree_1): Likewise.
12592         (type_contains_placeholder_1): Likewise.
12593         (build_common_tree_nodes): Likewise.
12594         * tree.def (POINTER_BOUNDS_TYPE): Likewise.
12595         * tree.h (POINTER_BOUNDS_TYPE_P): Likewise.
12596         (POINTER_BOUNDS_P): Likewise.
12597         (BOUNDED_TYPE_P): Likewise.
12598         (BOUNDED_P): Likewise.
12599         (CALL_WITH_BOUNDS_P): Likewise.
12600         (pointer_bounds_type_node): Likewise.
12601         * value-prof.c (gimple_ic): Likewise.
12602         * var-tracking.c (vt_add_function_parameters): Likewise.
12603         * varasm.c (make_decl_rtl): Likewise.
12604         (assemble_start_function): Likewise.
12605         (output_constant): Likewise.
12606         (maybe_assemble_visibility): Likewise.
12607         * varpool.c (ctor_for_folding): Likewise.
12608         * chkp-builtins.def: Remove.
12609         * ipa-chkp.c: Remove.
12610         * ipa-chkp.h: Remove.
12611         * rtl-chkp.c: Remove.
12612         * rtl-chkp.h: Remove.
12613         * tree-chkp-opt.c: Remove.
12614         * tree-chkp.c: Remove.
12615         * tree-chkp.h: Remove.
12617 2018-06-07  Carl Love  <cel@us.ibm.com>
12619         * config/rs6000/vsx.md (vextract_fp_from_shorth,
12620         vextract_fp_from_shortl): Add BE support.
12622 2018-06-07  Paul Koning  <ni1d@arrl.net>
12624         * compare-elim.c (try_merge_compare): Don't merge compare if
12625         address contains a side effect.
12626         (try_eliminate_compare): Likewise.
12628 2018-06-07  Olga Makhotina  <olga.makhotina@intel.com>
12630         * config.gcc: Support "tremont".
12631         * config/i386/driver-i386.c (host_detect_local_cpu): Detect "tremont".
12632         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
12633         PROCESSOR_TREMONT.
12634         * config/i386/i386.c (m_TREMONT): Define.
12635         (processor_target_table): Add "tremont".
12636         (PTA_TREMONT): Define.
12637         (ix86_lea_outperforms): Add TARGET_TREMONT.
12638         (get_builtin_code_for_version): Handle PROCESSOR_TREMONT.
12639         (fold_builtin_cpu): Add M_INTEL_TREMONT, replace M_INTEL_GOLDMONT
12640         and M_INTEL_GOLDMONT_PLUS.
12641         (fold_builtin_cpu): Add "tremont".
12642         (ix86_add_stmt_cost): Add TARGET_TREMONT.
12643         (ix86_option_override_internal): Add "tremont".
12644         * config/i386/i386.h (processor_costs): Define TARGET_TREMONT.
12645         (processor_type): Add PROCESSOR_TREMONT.
12646         * config/i386/x86-tune.def: Add m_TREMONT.
12647         * doc/invoke.texi: Add tremont as x86 -march=/-mtune= CPU type.
12649 2018-06-07  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
12651         * config/msp430/msp430.c (msp430_mcu_name): Set the "i" in the
12652         symbol defined for msp430i* devices to be lower case.
12654 2018-06-07  Richard Biener  <rguenther@suse.de>
12656         * graphite-sese-to-poly.c (extract_affine): Avoid unneded
12657         wrapping.  Properly wrap the result of a BIT_NOT_EXPR.
12658         Properly wrap signed arithmetic if overflow wraps.
12660 2018-06-07  Jakub Jelinek  <jakub@redhat.com>
12662         PR tree-optimization/69615
12663         * tree-ssa-reassoc.c (optimize_range_tests_var_bound): If rhs2 is lhs
12664         of a cast from a same precision integral SSA_NAME in a bb dominated
12665         by first_bb, retry with rhs2 set to the rhs1 of the cast.  Don't emit
12666         cast to utype if rhs2 has already a compatible type.
12668 2018-06-07  Richard Biener  <rguenther@suse.de>
12670         PR tree-optimization/85935
12671         * graphite-scop-detection.c (find_params_in_bb): Analyze
12672         condition operands with respect to the correct loop.  Assert
12673         the analysis doesn't fail.
12675 2018-06-04  Carl Love  <cel@us.ibm.com>
12677         * config/rs6000/vsx.md (first_match_index_<mode>): Calculate index
12678         using natural element order.  Use gen_lshrsi3 instead of gen_ashrsi3
12679         as it is slightly cheaper.
12680         (first_match_or_eos_index_<mode>):
12681         Calculate index using natural element order.
12682         (first_match_index_<mode>):
12683         Calculate index using natural element order.
12684         (first_match_or_eos_index_<mode>):
12685         Calculate index using natural order.
12686         (define_insn vclzlsbb): Change to define_insn vclzlsbb_<mode>.
12687         for BE and LE modes.
12688         * config/rs6000/rs6000-c.c: Rename P9V_BUILTIN_VCLZLSBB,
12689         P9V_BUILTIN_VCLZLSBB_V16QI.
12690         * config/rs6000/rs6000-builtin.def: Make VCLZLSBB mode
12691         specific.
12693 2018-06-06  Kelvin Nilsen  <kelvin@gcc.gnu.org>
12695         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Adjust
12696         indentation and line wrap for many prototypes.  Add missing
12697         @smallexample directives around block of prototypes for vec_xl and
12698         vec_xst.
12700 2018-06-05  Michael Meissner  <meissner@linux.ibm.com>
12702         * config/rs6000/rs6000.c (rs6000_passes_ieee128): New boolean to
12703         track if we pass or return IEEE 128-bit floating point.
12704         (ieee128_mangling_gcc_8_1): New boolean to say whether to generate
12705         C++ mangling that is compatible with GCC 8.1.
12706         (TARGET_ASM_GLOBALIZE_DECL_NAME): Override target hook.
12707         (init_cumulative_args): Note if we pass or return IEEE 128-bit
12708         floating point types.
12709         (rs6000_function_arg_advance_1): Likewise.
12710         (rs6000_mangle_type): Optionally generate mangled names that match
12711         what GCC 8.1 generated for IEEE 128-bit floating point types.
12712         (rs6000_globalize_decl_name): If we have an external function that
12713         passes or returns IEEE 128-bit types, generate a weak reference
12714         from the mangled name used in GCC 8.1 to the current mangled
12715         name.
12716         (rs6000_init_builtins): Make __ibm128 use the long double type if
12717         long double is IBM extended double.  Make __float128 use the long
12718         double type if long double is IEEE 128-bit.
12720         PR target/85657
12721         * config/rs6000/rs6000-builtin.def (BU_IBM128_2): New helper
12722         macro for __ibm128 built-in functions.
12723         (PACK_IF): Add __ibm128 pack/unpack functions.
12724         (UNPACK_IF): Likewise.
12725         * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Do not
12726         enable long double built-in functions if long double is IEEE
12727         128-bit floating point.
12728         (rs6000_invalid_builtin): Update long double built-in function
12729         error message.
12730         (rs6000_expand_builtin): For PACK_IF and UNPACK_IF built-in
12731         functions, adjust the built-in function to use the long double
12732         built-in function if __ibm128 and long double are the same type.
12733         * doc/extend.texi (PowerPC builtins): Update documention for
12734         __builtin_{,un}pack_longdouble.  Add documentation for
12735         __builtin_{,un}pack_ibm128.
12737 2018-06-06  Jim Wilson  <jimw@sifive.com>
12739         * config/riscv/riscv.c (enum riscv_privilege_levels): New.
12740         (struct machine_function): New field interrupt_mode.
12741         (riscv_handle_type_attribute): New function.  Add forward declaration.
12742         (riscv_attribute_table) <interrupt>: Use riscv_handle_type_attribute.
12743         (riscv_expand_epilogue): Check interrupt_mode field.
12744         (riscv_set_current_function): Check interrupt attribute args and
12745         set interrupt_mode field.
12746         * config/riscv/riscv.md (UNSPECV_SRET, UNSPECV_URET): New.
12747         (riscv_sret, riscv_uret): New.
12748         * doc/extend.texi (RISC-V Function Attributes) <interrupt>: Document
12749         new arguments to interrupt attribute.
12751 2018-06-06  Peter Bergner  <bergner@vnet.ibm.com>
12753         PR target/63177
12754         * /config/rs6000/rs6000.h (ASM_CPU_SPEC): Add support for -mpower9.
12755         Don't handle -mcpu=power8 if -mpower9-vector is also used.
12757 2018-06-06  Kelvin Nilsen  <kelvin@gcc.gnu.org>
12759         * config/rs6000/rs6000-builtin.def (VSX_BUILTIN_VEC_LD,
12760         VSX_BUILTIN_VEC_ST): Add comment to explain non-traditional uses.
12761         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
12762         several redundant entries.
12764 2018-06-06  David Malcolm  <dmalcolm@redhat.com>
12766         * config/i386/i386-protos.h (ix86_expand_call): Strengthen return
12767         type from "rtx" to "rtx_insn *".
12768         * config/i386/i386.c (ix86_expand_split_stack_prologue): Likewise
12769         for local "call_insn", removing cast.
12770         (ix86_expand_call): Likewise, introducing a "call_insn" local.
12772 2018-06-06  Eric Botcazou  <ebotcazou@adacore.com>
12774         PR tree-optimization/86066
12775         * gimple-ssa-store-merging.c (process_store): Do not bypass BIT_NOT_EXPR
12776         for BIT_INSERT_EXPR stores.
12778 2018-06-06  Richard Biener  <rguenther@suse.de>
12780         PR tree-optimization/86062
12781         * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
12782         component refs ontop
12783         of to be offsetted base.
12785 2018-06-06  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
12787         * config/msp430/msp430.c (msp430_attr): Allow interrupt handlers
12788         to be static and remove check on interrupt attribute name.
12790 2018-06-05  Kelvin Nilsen  <kelvin@gcc.gnu.org>
12792         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
12793         volatile qualifier from vec_lvsl and vec_lvsr argument prototypes.
12795 2018-06-05  Steve Ellcey  <sellcey@cavium.com>
12797         PR target/79924
12798         * config/aarch64/aarch64-protos.h (aarch64_err_no_fpadvsimd): Remove
12799         second argument.
12800         * config/aarch64/aarch64-protos..c (aarch64_err_no_fpadvsimd):
12801         Remove second argument, change how error is called.
12802         (aarch64_layout_arg): Remove second argument from
12803         aarch64_err_no_fpadvsimd call.
12804         (aarch64_init_cumulative_args): Ditto.
12805         (aarch64_gimplify_va_arg_expr): Ditto.
12806         * config/aarch64/aarch64.md (mov<mode>): Ditto.
12808 2018-06-05  Uros Bizjak  <ubizjak@gmail.com>
12810         * config/i386/i386.md (simple_return_indirect_internal): New expander.
12811         (*simple_return_indirect_internal<mode>): Rename from
12812         simple_return_indirect_internal.  Use W mode iterator.
12813         (rstorssp): New expander.
12814         (*rstorssp<mode>): Rename from rstorssp.  Use P mode iterator.
12815         (clrssbsy): New expander.
12816         (*clrssbsy<mode>): Rename from clrssbsy.  Use P mode iterator.
12818 2018-06-05  Andre Vieira  <andre.simoesdiasvieira@arm.com>
12820         * config/arm/arm_cmse.h (cmse_nsfptr_create): Change typeof to
12821         __typeof__.
12822         (cmse_check_pointed_object): Likewise.
12824 2018-06-05  Martin Liska  <mliska@suse.cz>
12826         PR gcov-profile/47618
12827         * doc/invoke.texi: Document how -fprofile-dir format
12828         is extended.
12830 2018-06-05  Richard Biener  <rguenther@suse.de>
12832         * tree-cfgcleanup.c (cleanup_control_flow_pre): For edge
12833         removal pretend DOM info isn't available so we do not update
12834         it and only remove edges, not dominated blocks.  Actually free
12835         DOM info in case we removed something.  Remove unreachable blocks.
12836         (mfb_keep_latches): Work with either DOM info or marked backedges.
12837         (cleanup_tree_cfg_noloop): Do not remove unreachable blocks
12838         first.  Mark backedges if DOM info isn't available.
12839         (Re-)compute DOM info after cleanup_control_flow_pre.
12841 2018-06-05  Richard Biener  <rguenther@suse.de>
12843         * tree-cfg.c (struct locus_discrim_map): Store line, not location.
12844         (locus_discrim_hasher::hash): Adjust.
12845         (locus_discrim_hasher::equal): Likewise.
12846         (next_discriminator_for_locus): Work on line directly.
12847         (same_line_p): Pass in expanded locus1 as well.
12848         (assign_discriminators): Avoid redundant location expansions.
12850 2018-06-05  Richard Biener  <rguenther@suse.de>
12852         PR tree-optimization/86046
12853         * tree-ssa.c (maybe_optimize_var): Clear DECL_GIMPLE_REG_P
12854         if required after clearing TREE_ADDRESSABLE.
12856 2018-06-05  Richard Biener  <rguenther@suse.de>
12858         PR tree-optimization/86047
12859         * tree-ssa-loop.c (for_each_index): Glob handling of all
12860         decls and constants and really handle all of them.
12862 2018-06-05  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
12864         PR target/81497
12865         * config/arm/arm-builtins.c (arm_type_qualifiers): Add
12866         qualifier_void_pointer and qualifier_const_void_pointer.
12867         (arm_ldc_qualifiers, arm_stc_qualifiers): Use the above.
12868         (arm_init_builtins): Handle the above.
12869         * config/arm/arm_acle.h (__arm_cdp, __arm_ldc, __arm_ldcl, __arm_stc,
12870         __arm_stcl, __arm_mcr, __arm_cdp2, __arm_ldc2, __arm_ldcl2, __arm_stc2,
12871         __arm_stcl2,__arm_mcr2, __arm_mcrr, __arm_mcrr2): Remove return for
12872         void intrinsics.
12874 2018-06-05  Martin Liska  <mliska@suse.cz>
12876         * auto-profile.c (read_autofdo_file): Do not use
12877         gcov_ctr_summary struct.
12878         (afdo_callsite_hot_enough_for_early_inline): Likewise.
12879         * coverage.c (struct counts_entry): Likewise.
12880         (read_counts_file): Read just single summary entry.
12881         (get_coverage_counts): Use gcov_summary struct.
12882         * coverage.h (get_coverage_counts): Likewise.
12883         * gcov-dump.c (dump_working_sets): Likewise.
12884         (tag_summary): Dump just single summary.
12885         * gcov-io.c (gcov_write_summary): Write just histogram
12886         summary.
12887         (gcov_read_summary): Read just single summary.
12888         (compute_working_sets): Use gcov_summary struct.
12889         * gcov-io.h (GCOV_TAG_SUMMARY_LENGTH): Remove usage
12890         of GCOV_COUNTERS_SUMMABLE.
12891         (GCOV_COUNTERS_SUMMABLE): Remove.
12892         (GCOV_FIRST_VALUE_COUNTER): Replace with
12893         GCOV_COUNTER_V_INTERVAL.
12894         (struct gcov_ctr_summary): Remove.
12895         (struct gcov_summary): Directly use fields of former
12896         gcov_ctr_summary.
12897         (compute_working_sets): Use gcov_summary struct.
12898         * gcov.c (read_count_file): Do not use ctrs fields.
12899         * lto-cgraph.c (merge_profile_summaries): Use gcov_summary
12900         struct.
12901         * lto-streamer.h (struct GTY): Make profile_info gcov_summary
12902         struct.
12903         * profile.c: Likewise.
12904         * profile.h: Likewise.
12906 2018-06-05  Martin Liska  <mliska@suse.cz>
12908         PR gcov-profile/84846
12909         * gcov.c (output_lines): Print working directory only
12910         in intermediate format.
12912 2018-06-05  Andreas Krebbel  <krebbel@linux.ibm.com>
12914         * config/s390/s390-builtin-types.def: Add void function type.
12915         * config/s390/s390-builtins.def: Use the function type for the
12916         tbeginc builtin.
12918 2018-06-04  Jim Wilson  <jimw@sifive.com>
12920         * config/riscv/riscv-protos.h (riscv_expand_epilogue): Change bool arg
12921         to int.
12922         * config/riscv/riscv.c (riscv_for_each_saved_reg): New args epilogue
12923         and maybe_eh_return.  Change regno to unsigned int.  Use new args to
12924         handle EH_RETURN_DATA_REGNO registers properly.
12925         (riscv_expand_prologue): Pass new args to riscv_for_each_saved_reg.
12926         (riscv_expand_epilogue): Update comment.  Change argument name and
12927         type.  Update code to use new name and type.  Pass new args to
12928         riscv_for_each_saved_reg.  Only use EH_RETURN_STACKADJ_RTX when
12929         EXCEPTION_RETURN.
12930         * config/riscv/riscv.md (NORMAL_RETURN): New.
12931         (SIBCALL_RETURN, EXCEPTION_RETURN): New.
12932         (epilogue, sibcall_epilogue): Update riscv_expand_epilogue arg.
12933         (eh_return): Call gen_eh_return_internal and emit barrier.
12934         (eh_return_internal): Call riscv_expand_epilogue.
12936 2018-06-04  Eric Botcazou  <ebotcazou@adacore.com>
12938         * gimple-ssa-store-merging.c (struct merged_store_group): Move up
12939         bit_insertion field and declare can_be_merged_into method.
12940         (merged_store_group::can_be_merged_into): New method.
12941         (imm_store_chain_info::coalesce_immediate): Call it to decide whether
12942         consecutive non-overlapping stores can be merged.  Turn MEM_REF stores
12943         into BIT_INSERT_EXPR stores if the group contains a non-MEM_REF store.
12945 2018-06-04  Richard Biener  <rguenther@suse.de>
12947         PR tree-optimization/85955
12948         * builtins.c (fold_builtin_sincos): Convert pointers to
12949         destination to appropriate type before dereferencing.
12951 2018-06-04  Segher Boessenkool  <segher@kernel.crashing.org>
12953         * config/rs6000/rs6000.md (abs<mode>2 for FLOAT128): Handle IFmode.
12955 2018-06-04  Richard Sandiford  <richard.sandiford@linaro.org>
12957         * expr.c (expand_expr_real_1): Force the operand into memory if
12958         its TYPE_MODE is BLKmode and if there is no integer mode for
12959         the number of bits being extracted.
12961 2018-06-04  Jakub Jelinek  <jakub@redhat.com>
12963         PR target/85832
12964         PR target/86036
12965         * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
12966         Use vptestnm rather than vptestm in (=Yc,v,C) variant.
12968 2018-06-04  Richard Biener  <rguenther@suse.de>
12970         * tree-cfgcleanup.c (cleanup_tree_cfg_1): Fold into...
12971         (cleanup_tree_cfg_noloop): ... single caller.  Do
12972         start_recording_case_labels later.
12974 2018-06-04  Sebastian Peryt  <sebastian.peryt@intel.com>
12976         * config/i386/cldemoteintrin.h: Change define from _X86INTRIN_H_INCLUDED
12977         to _IMMINTRIN_H_INCLUDED.
12978         * config/i386/pconfigintrin.h: Ditto.
12979         * config/i386/waitpkgintrin.h: Ditto.
12980         * config/i386/immintrin.h: Add includes for sgxintrin.h,
12981         pconfigintrin.h, waitpkgintrin.h and cldemoteintrin.h.
12982         * config/i386/x86intrin.h: Remove includes for mintrin.h, xmmintrin.h,
12983         emmintrin.h, pmmintrin.h, tmmintrin.h, smmintrin.h, wmmintrin.h,
12984         bmiintrin.h, bmi2intrin.h, lzcntintrin.h, sgxintrin.h, pconfigintrin.h,
12985         waitpkgintrin.h and cldemoteintrin.h.
12987 2018-06-04  Richard Biener  <rguenther@suse.de>
12989         PR tree-optimization/86038
12990         * tracer.c (find_best_successor): Check probability for
12991         being initialized, bail out if not.
12993 2018-06-04  Richard Earnshaw  <rearnsha@arm.com>
12995         PR target/86003
12996         * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
12997         of bits to ignore when comparing architectures.
12999 2018-06-04  Jakub Jelinek  <jakub@redhat.com>
13001         PR tree-optimization/69615
13002         * fold-const.c (merge_ranges): If range1 is - [x, x] and x is the
13003         maximum or minimum of the type, try to merge it also as if
13004         range1 is + [-, x - 1] or + [x + 1, -].
13006         PR c++/86025
13007         * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
13009 2018-06-03  Eric Botcazou  <ebotcazou@adacore.com>
13011         PR tree-optimization/86034
13012         * gimple-ssa-store-merging.c (output_merged_store): Convert the RHS to
13013         the unsigned bitfield type in a bit insertion sequence if it does not
13014         have a larger precision than the bitfield size.
13015         (process_store): Also bypass widening conversions for BIT_INSERT_EXPR.
13017 2018-06-03  Kito Cheng  <kito.cheng@gmail.com>
13019         * config/nds32/nds32-peephole2.md: Add new patterns for code size.
13021 2018-06-03  Chung-Ju Wu  <jasonwucj@gmail.com>
13023         * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3J.
13024         * config/nds32/nds32.c (nds32_option_override): Consider ARCH_V3J.
13025         * config/nds32/nds32.h (TARGET_ISA_V3): Add ARCH_V3J checking.
13026         * config/nds32/nds32.opt (march): Add enum value Value(ARCH_V3J).
13028 2018-06-03  Chung-Ju Wu  <jasonwucj@gmail.com>
13030         * common/config/nds32/nds32-common.c (nds32_option_optimization_table):
13031         Disable -fdelete-null-pointer-checks for ELF toolchain.
13033 2018-06-02  Chung-Ju Wu  <jasonwucj@gmail.com>
13034             Kito Cheng  <kito.cheng@gmail.com>
13036         * config.gcc (nds32*): Use nds32-linux.opt and nds32-elf.opt.
13037         (nds32le-*-*, nds32be-*-*): Integrate checking process.
13038         (nds32*-*-*): Add glibc and uclibc conditions.
13039         * common/config/nds32/nds32-common.c (nds32_except_unwind_info): New.
13040         (TARGET_EXCEPT_UNWIND_INFO): Define.
13041         * config/nds32/elf.h: New file.
13042         * config/nds32/linux.h: New file.
13043         * config/nds32/nds32-elf.opt: New file.
13044         * config/nds32/nds32-linux.opt: New file.
13045         * config/nds32/nds32-fp-as-gp.c
13046         (pass_nds32_fp_as_gp::gate): Consider TARGET_LINUX_ABI.
13047         * config/nds32/nds32.c (nds32_conditional_register_usage): Consider
13048         TARGET_LINUX_ABI.
13049         (nds32_asm_file_end): Ditto.
13050         (nds32_print_operand): Ditto.
13051         (nds32_insert_attributes): Ditto.
13052         (nds32_init_libfuncs): New function.
13053         (TARGET_HAVE_TLS): Define.
13054         (TARGET_INIT_LIBFUNCS): Define.
13055         * config/nds32/nds32.h (TARGET_DEFAULT_RELAX): Apply different relax
13056         spec content.
13057         (TARGET_ELF): Apply different mcmodel setting.
13058         (LINK_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): The content has
13059         been migrated into elf.h and linux.h files.
13060         * config/nds32/nds32.md (add_pc): Consider TARGET_LINUX_ABI.
13061         * config/nds32/nds32.opt (mvh): Consider TARGET_LINUX_ABI.
13062         (mcmodel): The content has been migrated into nds32-elf.opt and
13063         nds32-linux.opt files.
13064         * config/nds32/t-elf: New file.
13065         * config/nds32/t-linux: New file.
13067 2018-06-02  Chung-Ju Wu  <jasonwucj@gmail.com>
13068             Shiva Chen  <shiva0217@gmail.com>
13070         * config/nds32/constants.md (unspec_volatile_element): Add
13071         UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END.
13072         * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp
13073         optimization.
13074         * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare.
13075         (make_pass_nds32_fp_as_gp): Declare.
13076         * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one
13077         optmization pass.
13078         (nds32_asm_function_end_prologue): Remove unused asm output.
13079         (nds32_asm_function_begin_epilogue): Remove unused asm output.
13080         (nds32_asm_file_start): Output necessary fp_as_gp information.
13081         (nds32_option_override): Adjust register usage.
13082         (nds32_expand_prologue): Consider fp_as_gp situation.
13083         (nds32_expand_prologue_v3push): Consider fp_as_gp situation.
13084         * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p.
13085         (epilogue): Ditto.
13086         (return): Ditto.
13087         (simple_return): Ditto.
13088         (omit_fp_begin): Output special directive for fp_as_gp.
13089         (omit_fp_end): Output special directive for fp_as_gp.
13090         * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp,
13091         mforbid-fp-as-gp): New options.
13093 2018-06-01  Mark Wielaard  <mark@klomp.org>
13095         * dwarf2out.c (dwarf2out_finish): Remove generation of
13096         DW_AT_loclists_base.
13098 2018-06-01  Eric Botcazou  <ebotcazou@adacore.com>
13100         * gimple-ssa-store-merging.c: Include gimple-fold.h.
13101         (struct store_immediate_info): Document BIT_INSERT_EXPR stores.
13102         (struct merged_store_group): Add bit_insertion field.
13103         (dump_char_array): Use standard hexadecimal format.
13104         (merged_store_group::merged_store_group): Set bit_insertion to false.
13105         (merged_store_group::apply_stores): Use optimal buffer size.  Deal
13106         with BIT_INSERT_EXPR stores.  Move up code updating the mask and
13107         also print the mask in the dump file.
13108         (pass_store_merging::gate): Minor tweak.
13109         (imm_store_chain_info::coalesce_immediate): Fix wrong association
13110         of stores with groups in dump.  Allow coalescing of BIT_INSERT_EXPR
13111         stores with INTEGER_CST stores.
13112         (count_multiple_uses) <BIT_INSERT_EXPR>: New case.
13113         (imm_store_chain_info::output_merged_store): Add try_bitpos variable
13114         and use it throughout.  Generate bit insertion sequences if need be.
13115         (pass_store_merging::process_store): Remove redundant condition.
13116         Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR.
13118 2018-06-01  Segher Boessenkool  <segher@kernel.crashing.org>
13120         * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
13121         the 128-bit floating point types.  Fix function comment.
13123 2018-06-01  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
13125         * config/aarch64/aarch64-simd.md
13126         (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler
13127         mnemonics.
13128         (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
13129         mnemonics.
13131 2018-06-01  Richard Sandiford  <richard.sandiford@linaro.org>
13133         PR tree-optimization/85989
13134         * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
13135         variable.
13136         (backprop::intersect_uses): Check it when deciding whether this
13137         is a backedge reference.
13138         (backprop::process_block): Add each phi to m_visited_phis
13139         after visiting it, then clear it at the end.
13141 2018-06-01  Richard Biener  <rguenther@suse.de>
13143         * tree-vectorizer.h (vect_dr_stmt): New function.
13144         (vect_get_load_cost): Adjust.
13145         (vect_get_store_cost): Likewise.
13146         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
13147         Use vect_dr_stmt instead of DR_SMTT.
13148         (vect_record_base_alignments): Likewise.
13149         (vect_calculate_target_alignment): Likewise.
13150         (vect_compute_data_ref_alignment): Likewise and make static.
13151         (vect_update_misalignment_for_peel): Likewise.
13152         (vect_verify_datarefs_alignment): Likewise.
13153         (vector_alignment_reachable_p): Likewise.
13154         (vect_get_data_access_cost): Likewise.  Pass down
13155         vinfo to vect_get_load_cost/vect_get_store_cost instead of DR.
13156         (vect_get_peeling_costs_all_drs): Likewise.
13157         (vect_peeling_hash_get_lowest_cost): Likewise.
13158         (vect_enhance_data_refs_alignment): Likewise.
13159         (vect_find_same_alignment_drs): Likewise.
13160         (vect_analyze_data_refs_alignment): Likewise.
13161         (vect_analyze_group_access_1): Likewise.
13162         (vect_analyze_group_access): Likewise.
13163         (vect_analyze_data_ref_access): Likewise.
13164         (vect_analyze_data_ref_accesses): Likewise.
13165         (vect_vfa_segment_size): Likewise.
13166         (vect_small_gap_p): Likewise.
13167         (vectorizable_with_step_bound_p): Likewise.
13168         (vect_prune_runtime_alias_test_list): Likewise.
13169         (vect_analyze_data_refs): Likewise.
13170         (vect_supportable_dr_alignment): Likewise.
13171         * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
13172         (vect_gen_prolog_loop_niters): Likewise.
13173         * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
13174         * tree-vect-patterns.c (vect_recog_bool_pattern): Do not
13175         modify DR_STMT.
13176         (vect_recog_mask_conversion_pattern): Likewise.
13177         (vect_try_gather_scatter_pattern): Likewise.
13178         * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info
13179         to vect_get_store_cost.
13180         (vect_get_store_cost): Get stmt_info instead of DR.
13181         (vect_model_load_cost): Pass stmt_info to vect_get_load_cost.
13182         (vect_get_load_cost): Get stmt_info instead of DR.
13184 2018-06-01  Richard Biener  <rguenther@suse.de>
13186         PR middle-end/86017
13187         * gimple-fold.c (var_decl_component_p): Also allow offsetted
13188         vars wrapped in MEM_REFs.
13190 2018-06-01  Richard Sandiford  <richard.sandiford@linaro.org>
13192         * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
13193         Fix subreg tests so that we only return a choice between
13194         GENERAL_REGS and FP_REGS if the original classes included both.
13196 2018-06-01  Richard Biener  <rguenther@suse.de>
13198         PR ipa/85960
13199         * tree-ssa-structalias.c (get_function_part_constraint):
13200         Handle NULL fi->decl.
13201         (find_func_aliases_for_call): Properly handle indirect
13202         fi from direct call.
13203         (find_func_clobbers): Likewise.
13204         (ipa_pta_execute): Likewise.
13205         (create_variable_info_for): For functions that are ifunc_resolver
13206         resolve to a varinfo that contains the result of the resolver call.
13207         (associate_varinfo_to_alias): Do not treat ifunc resolvers as
13208         aliases.
13210 2018-05-31  Michael Collison  <michael.collison@arm.com>
13212         * config/aarch64/aarch64.md:
13213         (*fix_to_zero_extenddfdi2): New pattern.
13214         * gcc.target/aarch64/fix_extend1.c: New testcase.
13216 2018-05-31  Qing Zhao <qing.zhao@oracle.com>
13218         PR middle-end/78809
13219         PR middle-end/83026
13220         * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
13221         and BUILT_IN_STRNCMP_EQ.
13222         * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
13223         BUILT_IN_STRNCMP_EQ.
13224         * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
13225         handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
13226         (gimple_fold_builtin): Likewise.
13227         * tree-ssa-strlen.c (compute_string_length): New function.
13228         (determine_min_obsize): New function.
13229         (handle_builtin_string_cmp): New function to handle calls to
13230         string compare functions.
13231         (strlen_optimize_stmt): Add handling to builtin string compare
13232         calls.
13233         * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
13234         Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
13235         * tree.c (build_common_builtin_nodes): Add new defines of
13236         BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
13238 2018-05-31  Jakub Jelinek  <jakub@redhat.com>
13240         PR target/85984
13241         * bb-reorder.c (pass_partition_blocks::gate): Return false for
13242         functions with naked attribute.
13244 2018-05-31  Uros Bizjak  <ubizjak@gmail.com>
13246         * config/i386/sse.md (avx_vec_concat<mode>):
13247         Substitute concat_tg_mode mode attribute with xtg_mode.
13248         (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
13249         (concat_tg_mode): Remove mode attribute.
13251 2018-05-31  Martin Sebor  <msebor@redhat.com>
13253         PR c/82063
13254         * calls.c (alloc_max_size): Correct a logic error/typo.
13255         Treat excessive arguments as infinite.  Warn for invalid arguments.
13256         * doc/invoke.texi (-Walloc-size-larger-than): Update.
13258 2018-05-31  H.J. Lu  <hongjiu.lu@intel.com>
13260         PR target/85829
13261         * config/i386/x86-tune.def: Re-enable partial_reg_dependency
13262         and movx for Haswell.
13264 2018-05-31  Chung-Lin Tang  <cltang@codesourcery.com>
13265             Cesar Philippidis  <cesar@codesourcery.com>
13267         PR middle-end/85879
13268         * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
13269         when emitting error on private/firstprivate reductions.
13270         * omp-low.c (lower_omp_target): Avoid reference-type processing
13271         on pointers for firstprivate clause.
13273 2018-05-31  Sameera Deshpande <sameera.deshpande@linaro.org>
13275         * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
13276         (st1x2): Likewise.
13277         (st1x3): Likewise.
13278         * config/aarch64/aarch64-simd.md
13279         (aarch64_ld1x3<VALLDIF:mode>): New pattern.
13280         (aarch64_ld1_x3_<mode>): Likewise
13281         (aarch64_st1x2<VALLDIF:mode>): Likewise
13282         (aarch64_st1_x2_<mode>): Likewise
13283         (aarch64_st1x3<VALLDIF:mode>): Likewise
13284         (aarch64_st1_x3_<mode>): Likewise
13285         * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
13286         (vld1_s8_x3): Likewise.
13287         (vld1_u16_x3): Likewise.
13288         (vld1_s16_x3): Likewise.
13289         (vld1_u32_x3): Likewise.
13290         (vld1_s32_x3): Likewise.
13291         (vld1_u64_x3): Likewise.
13292         (vld1_s64_x3): Likewise.
13293         (vld1_f16_x3): Likewise.
13294         (vld1_f32_x3): Likewise.
13295         (vld1_f64_x3): Likewise.
13296         (vld1_p8_x3): Likewise.
13297         (vld1_p16_x3): Likewise.
13298         (vld1_p64_x3): Likewise.
13299         (vld1q_u8_x3): Likewise.
13300         (vld1q_s8_x3): Likewise.
13301         (vld1q_u16_x3): Likewise.
13302         (vld1q_s16_x3): Likewise.
13303         (vld1q_u32_x3): Likewise.
13304         (vld1q_s32_x3): Likewise.
13305         (vld1q_u64_x3): Likewise.
13306         (vld1q_s64_x3): Likewise.
13307         (vld1q_f16_x3): Likewise.
13308         (vld1q_f32_x3): Likewise.
13309         (vld1q_f64_x3): Likewise.
13310         (vld1q_p8_x3): Likewise.
13311         (vld1q_p16_x3): Likewise.
13312         (vld1q_p64_x3): Likewise.
13313         (vst1_s64_x2): Likewise.
13314         (vst1_u64_x2): Likewise.
13315         (vst1_f64_x2): Likewise.
13316         (vst1_s8_x2): Likewise.
13317         (vst1_p8_x2): Likewise.
13318         (vst1_s16_x2): Likewise.
13319         (vst1_p16_x2): Likewise.
13320         (vst1_s32_x2): Likewise.
13321         (vst1_u8_x2): Likewise.
13322         (vst1_u16_x2): Likewise.
13323         (vst1_u32_x2): Likewise.
13324         (vst1_f16_x2): Likewise.
13325         (vst1_f32_x2): Likewise.
13326         (vst1_p64_x2): Likewise.
13327         (vst1q_s8_x2): Likewise.
13328         (vst1q_p8_x2): Likewise.
13329         (vst1q_s16_x2): Likewise.
13330         (vst1q_p16_x2): Likewise.
13331         (vst1q_s32_x2): Likewise.
13332         (vst1q_s64_x2): Likewise.
13333         (vst1q_u8_x2): Likewise.
13334         (vst1q_u16_x2): Likewise.
13335         (vst1q_u32_x2): Likewise.
13336         (vst1q_u64_x2): Likewise.
13337         (vst1q_f16_x2): Likewise.
13338         (vst1q_f32_x2): Likewise.
13339         (vst1q_f64_x2): Likewise.
13340         (vst1q_p64_x2): Likewise.
13341         (vst1_s64_x3): Likewise.
13342         (vst1_u64_x3): Likewise.
13343         (vst1_f64_x3): Likewise.
13344         (vst1_s8_x3): Likewise.
13345         (vst1_p8_x3): Likewise.
13346         (vst1_s16_x3): Likewise.
13347         (vst1_p16_x3): Likewise.
13348         (vst1_s32_x3): Likewise.
13349         (vst1_u8_x3): Likewise.
13350         (vst1_u16_x3): Likewise.
13351         (vst1_u32_x3): Likewise.
13352         (vst1_f16_x3): Likewise.
13353         (vst1_f32_x3): Likewise.
13354         (vst1_p64_x3): Likewise.
13355         (vst1q_s8_x3): Likewise.
13356         (vst1q_p8_x3): Likewise.
13357         (vst1q_s16_x3): Likewise.
13358         (vst1q_p16_x3): Likewise.
13359         (vst1q_s32_x3): Likewise.
13360         (vst1q_s64_x3): Likewise.
13361         (vst1q_u8_x3): Likewise.
13362         (vst1q_u16_x3): Likewise.
13363         (vst1q_u32_x3): Likewise.
13364         (vst1q_u64_x3): Likewise.
13365         (vst1q_f16_x3): Likewise.
13366         (vst1q_f32_x3): Likewise.
13367         (vst1q_f64_x3): Likewise.
13368         (vst1q_p64_x3): Likewise.
13370 2018-05-30  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
13372         * config/msp430/msp430.c (msp430_output_labelref): Prepend
13373         user_label_prefix to name.
13375         * tree-core.h: Update comment about the format of NAME string
13376         passed to handler in attribute_spec.
13378         * config/msp430/msp430.md: Remove erroneous subreg expression from
13379         zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
13380         zero_extend{q,h}isi2.
13382 2018-05-30  Borislav Petkov  <bp@suse.de>
13384         * doc/extend.texi: Document some architecture specific
13385         constraints and sort entries.
13387 2018-05-30  Martin Sebor  <msebor@redhat.com>
13389         PR middle-end/85369
13390         * builtins.c (expand_builtin_stpcpy_1): New function.
13391         (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
13392         only if the former succeeds.
13394 2018-05-31  Sameera Deshpande <sameera.deshpande@linaro.org>
13396         * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
13397         in saphira.
13399 2018-05-30  Jan Hubicka  <hubicka@ucw.cz>
13401         * doc/invoke.texi (-flinker-output): Document
13403 2018-05-30  Jan Hubicka  <hubicka@ucw.cz>
13405         * passes.c (ipa_write_summaries): Only modify statements if body
13406         is in memory.
13407         * cgraphunit.c (ipa_passes): Also produce intermeidate code when
13408         incrementally linking.
13409         (ipa_passes): Likewise.
13410         * lto-cgraph.c (lto_output_node): When incrementally linking do not
13411         pass down resolution info.
13412         * common.opt (flag_incremental_link): Update info.
13413         * gcc.c (plugin specs): Turn flinker-output=* to
13414         -plugin-opt=-linker-output-known
13415         * toplev.c (compile_file): Also cut compilation when doing incremental
13416         link.
13417         * flag-types. (enum lto_partition_model): Add
13418         LTO_LINKER_OUTPUT_NOLTOREL.
13419         (invoke.texi): Add -flinker-output docs.
13420         * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
13421         link same way as WPA; do not stream in dead initializers.
13423         * dwarf2out.c (dwarf2out_die_ref_for_decl,
13424         darf2out_register_external_decl): Support incremental link.
13426 2018-05-30  Jan Hubicka  <hubicka@ucw.cz>
13428         * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
13430 2018-05-30  Jan Hubicka  <hubicka@ucw.cz>
13432         * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
13433         it down to simple_object_copy_lto_debug_sections.
13434         (run_gcc): Determine incremental LTO link time and configure
13435         lto1 into non-wpa mode, disable renaming of debug sections.
13437 2018-05-30  Kelvin Nilsen  <kelvin@gcc.gnu.org>
13439         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
13440         descriptions of various incorrectly documented functions.
13442 2018-05-30  Andre Vieira  <andre.simoesdiasvieira@arm.com>
13444         Revert:
13445         * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
13446         address check not strict.
13448 2018-05-30  Richard Biener  <rguenther@suse.de>
13450         PR tree-optimization/85964
13451         * tracer.c (better_p): Drop initialized count check, we only
13452         call the function with initialized counts now.
13453         (find_best_successor): Do find a best edge if one
13454         has uninitialized count.
13455         (find_best_predecessor): Likewise.  Do BB frequency check only
13456         if count is initialized.
13458 2017-05-30  Jackson Woodruff  <jackson.woodruff@arm.com>
13460         * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
13461         (aarch64_ldrstr_offset_compare): New.
13462         (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
13463         load/store orderings.
13464         (aarch64_gen_adjusted_ldpstp): Likewise.
13466 2018-05-30  Wilco Dijkstra  <wdijkstr@arm.com>
13468         * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
13469         Check for subset of GENERAL_REGS and FP_REGS.
13470         * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
13471         r=w alternative.
13473 2018-05-30  Richard Sandiford  <richard.sandiford@linaro.org>
13475         * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
13476         and wi::to_poly_offset.  Add the current offset and then check
13477         whether the sum fits, rather than using an unchecked addition of
13478         a checked term.  Check for a shwi rather than a uhwi.
13479         * expr.c (get_bit_range): Use tree_to_poly_uint64.
13480         (store_constructor): Use poly_int_tree_p.
13481         (expand_expr_real_1): Likewise.
13482         * function.c (assign_temp): Likewise.
13483         * fold-const.c (const_binop): Use poly_int_tree_p and
13484         wi::to_poly_offset.
13485         (fold_indirect_ref_1): Likewise.  Use multiple_p to attempt an exact
13486         division.
13487         * ipa-icf-gimple.c (func_checker::compare_operand): Use
13488         to_poly_offset for MEM offsets.
13489         * ipa-icf.c (sem_variable::equals): Likewise.
13490         * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
13491         * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
13492         wi::to_poly_offset for BIT_FIELD_REF offsets.
13493         (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
13494         wi::to_poly_offset.
13495         * var-tracking.c (emit_note_insn_var_location): Use
13496         tree_to_poly_uint64.
13498 2018-05-29  Jim Wilson  <jimw@sifive.com>
13500         * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
13502 2018-05-29  Uros Bizjak  <ubizjak@gmail.com>
13504         PR target/85950
13505         * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
13506         Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
13507         sequence.
13508         (sse4_1_round<mode>2): Use nonimmediate_operand
13509         for operand 1 predicate.
13511 2018-05-29  Martin Sebor  <msebor@redhat.com>
13512             Richard Biener  <rguenther@suse.de>
13514         PR testsuite/85888
13515         * calls.c (get_size_range): Call determine_value_range instead
13516         of get_value_range..
13517         * tree-vrp.h (determine_value_range): Declared new function.
13518         * tree-vrp.c (determine_value_range_1, determine_value_range): New.
13520 2018-05-29  Richard Biener  <rguenther@suse.de>
13522         * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
13523         sure to use non-pattern stmts for get_earlier_stmt arguments.
13524         * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
13525         called on pattern stmts.
13526         (get_later_stmt): Likewise.
13528 2018-05-29  Martin Liska  <mliska@suse.cz>
13530         PR gcov-profile/85759
13531         * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
13532         env variables.
13534 2018-05-29  Jakub Jelinek  <jakub@redhat.com>
13536         * tree-cfg.c (verify_gimple_assign_unary): Add checking for
13537         VEC_UNPACK_*_EXPR.
13538         (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
13539         VEC_PACK_*_EXPR.
13541         PR target/85918
13542         * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
13543         VEC_PACK_FLOAT_EXPR): New tree codes.
13544         * tree-pretty-print.c (op_code_prio): Handle
13545         VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
13546         (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
13547         VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
13548         * tree-inline.c (estimate_operator_cost): Likewise.
13549         * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
13550         * fold-const.c (const_binop): Likewise.
13551         (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
13552         VEC_UNPACK_FIX_TRUNC_LO_EXPR.
13553         * tree-cfg.c (verify_gimple_assign_unary): Likewise.
13554         (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
13555         * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
13556         VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
13557         * expr.c (expand_expr_real_2): Likewise.
13558         * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
13559         vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
13560         vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
13561         optabs.
13562         * optabs.c (expand_widen_pattern_expr): For
13563         VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
13564         sign from result type rather than operand's type.
13565         (expand_binop_directly): For vec_packu_float_optab and
13566         vec_packs_float_optab allow result type to be different from operand's
13567         type.
13568         * optabs-tree.c (optab_for_tree_code): Handle
13569         VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
13570         VEC_PACK_FLOAT_EXPR.  Formatting fixes.
13571         * tree-vect-generic.c (expand_vector_operations_1):  Handle
13572         VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
13573         VEC_PACK_FLOAT_EXPR.
13574         * tree-vect-stmts.c (supportable_widening_operation): Handle
13575         FIX_TRUNC_EXPR.
13576         (supportable_narrowing_operation): Handle FLOAT_EXPR.
13577         * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
13578         * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
13579         (float<floatunssuffix>v2div2sf2): ... this.  Formatting fix.
13580         (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
13581         mode attributes.
13582         (vec_pack<floatprefix>_float_<mode>): New expander.
13583         (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
13584         attributes.
13585         (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
13586         vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
13587         * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
13588         vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
13589         vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
13590         Document.
13591         * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
13592         VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
13593         (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
13594         VEC_PACK_FLOAT_EXPR): Document.
13596 2018-05-29  Richard Biener  <rguenther@suse.de>
13598         * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
13599         member.
13600         (stmt_vec_info_vec): Make pointer.
13601         (init_stmt_vec_info_vec): Remove.
13602         (free_stmt_vec_info_vec): Likewise.
13603         (set_stmt_vec_info_vec): New function.
13604         (free_stmt_vec_infos): Likewise.
13605         (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
13606         (set_vinfo_for_stmt): Likewise.
13607         (get_earlier_stmt): Likewise.
13608         (get_later_stmt): Likewise.
13609         * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
13610         (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
13611         (vec_info::~vec_info): Free stmt_vec_infos.
13612         (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
13613         Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
13614         (pass_slp_vectorize::execute): Likewise.
13615         * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
13616         (free_stmt_vec_info_vec): Likewise.
13617         (set_stmt_vec_info_vec): New function.
13618         (free_stmt_vec_infos): Likewise.
13619         * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
13620         the global stmt_vec_info_vec.
13621         * tree-parloops.c (gather_scalar_reductions): Use
13622         set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
13623         vector.
13625 2018-05-29  Richard Biener  <rguenther@suse.de>
13627         * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
13629 2018-05-29  Martin Liska  <mliska@suse.cz>
13630             David Malcolm  <dmalcolm@redhat.com>
13632         * vec.c (test_reverse): New.
13633         (vec_c_tests): Add new test.
13634         * vec.h (vl_ptr>::reverse): New function.
13636 2018-05-29  Gerald Pfeifer  <gerald@pfeifer.com>
13638         * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
13640         * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
13641         and later.
13643 2018-05-28  Bernd Edlinger  <bernd.edlinger@hotmail.de>
13645         * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
13647 2018-05-28  Richard Biener  <rguenther@suse.de>
13649         PR tree-optimization/85933
13650         * tree-vect-data-refs.c (vect_record_base_alignments): Only
13651         look at stmts marked as vectorizable.
13653 2018-05-28  Richard Biener  <rguenther@suse.de>
13655         PR tree-optimization/85934
13656         * tree-vect-generic.c (expand_vector_operations_1): Hoist
13657         vector boolean check before scalar optimization.
13659 2018-05-28  Jakub Jelinek  <jakub@redhat.com>
13661         * doc/invoke.texi (ARM Options): Use @item instead of @itemx
13662         for armv5te.
13664 2018-05-28  Mark Wielaard  <mark@klomp.org>
13666         * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
13667         if it is an expression containing a minus sign.
13669 2018-05-27  John David Anglin  <danglin@gcc.gnu.org>
13671         * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
13673 2018-05-27  Paul Koning  <ni1d@arrl.net>
13675         * config/pdp11/pdp11.md (truncsihi2): Remove.
13677 2018-05-27  Monk Chiang  <sh.chiang04@gmail.com>
13678             Chung-Ju Wu  <jasonwucj@gmail.com>
13680         * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
13681         implementation.
13682         (unaligned_store_dw): Ditto.
13683         * config/nds32/nds32-memory-manipulation.c
13684         (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
13685         (nds32_gen_dup_4_byte_to_word_value): Rename to ...
13686         (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
13687         (emit_setmem_word_loop): Rename to ...
13688         (emit_setmem_doubleword_loop): ... this.
13689         (nds32_gen_dup_4_byte_to_word_value): New function.
13690         (nds32_gen_dup_8_byte_to_double_word_value): New function.
13691         (nds32_expand_setmem_loop): Refine implementation.
13692         (nds32_expand_setmem_loop_v3m): Ditto.
13693         * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
13694         pattern.
13696 2018-05-27  Chung-Ju Wu  <jasonwucj@gmail.com>
13698         * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
13700 2018-05-27  Chung-Ju Wu  <jasonwucj@gmail.com>
13702         * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
13703         (nds32_init_machine_status): Initialize machine->attr_naked_p and
13704         machine->attr_no_prologue_p.
13705         (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
13706         (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
13707         (nds32_expand_epilogue): Consider attr_naked_p.
13708         (nds32_expand_epilogue_v3pop): Likewise.
13709         (nds32_can_use_return_insn): Likewise.
13710         * config/nds32/nds32.h (machine_function): Add attr_naked_p and
13711         attr_no_prologue_p fields.
13712         * config/nds32/nds32.opt (mret-in-naked-func): New option.
13714 2018-05-27  Jakub Jelinek  <jakub@redhat.com>
13716         PR target/85918
13717         * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
13718         attributes.
13719         * config/i386/sse.md
13720         (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
13721         Rename to ...
13722         (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
13723         ... this.
13724         (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
13725         Rename to ...
13726         (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
13727         ... this.
13728         (*<floatsuffix>floatv2div2sf2): Rename to ...
13729         (*float<floatunssuffix>v2div2sf2): ... this.
13730         (<floatsuffix>floatv2div2sf2_mask): Rename to ...
13731         (float<floatunssuffix>v2div2sf2_mask): ... this.
13732         (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
13733         (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
13734         (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
13735         to ...
13736         (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
13737         ... this.
13738         (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
13739         Rename to ...
13740         (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
13741         ... this.
13742         (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
13743         Rename to ...
13744         (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
13745         ... this.
13746         (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
13747         (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
13748         (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
13749         gen_ufix_truncv8dfv8si2.
13750         * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
13751         __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
13752         __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
13753         __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
13754         __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
13755         __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
13756         __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
13757         Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
13759 2018-05-24  H.J. Lu  <hongjiu.lu@intel.com>
13761         PR target/85900
13762         PR target/85345
13763         * varasm.c (assemble_alias): Lookup ifunc attribute on error.
13765 2018-05-25  Jim Wilson  <jimw@sifive.com>
13767         * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
13768         * config/riscv/riscv.c (struct machine_function): Add
13769         interrupt_handler_p and attribute_checked_p fields.
13770         (riscv_attribute_table): Add interrupt.
13771         (riscv_interrupt_type_p): New.
13772         (riscv_save_reg_p): Save extra regs for interrupt handler.
13773         (riscv_use_save_libcall): Return false  for interrupt handler.
13774         (riscv_first_stack_step): Add forward declaration.
13775         (riscv_compute_frame_info): New local interrupt_save_t1.  Set it
13776         for interrupt handler with large frame.  Use it for saved reg list.
13777         (riscv_expand_prologue): Move flag_stack_usage_info support to
13778         eliminate duplication.
13779         (riscv_expand_epilogue): Generate mret for interrupt handler.
13780         (riscv_epilogue_uses): New.
13781         (riscv_can_use_return_insn): Return false for interrupt handler.
13782         (riscv_function_ok_for_sibcall): Likewise.
13783         (riscv_set_current_function): Add interrupt handler support.
13784         * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
13785         * config/riscv/riscv.md (UNSPECV_MRET): New.
13786         (GP_REGNUM): New.
13787         (riscv_frflags, riscv_fsflags): Use tab after opcode.
13788         (riscv_mret): New.
13789         * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
13791 2018-05-25  Bill Schmidt  <wschmidt@linux.ibm.com>
13793         PR tree-optimization/85712
13794         * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
13795         this candidate has already been replaced in-situ by a copy.
13797 2018-05-25  Jason Merrill  <jason@redhat.com>
13799         PR c++/80485 - inline function non-zero address.
13800         * symtab.c (nonzero_address): Check DECL_COMDAT.
13802 2018-05-25  Uros Bizjak  <ubizjak@gmail.com>
13804         PR target/83628
13805         * config/alpha/alpha.md (ashlsi3): New insn pattern.
13806         (*ashlsi_se): Rename from *ashldi_se.  Define as sign
13807         extension of SImode operation.  Use const123_operand predicate.
13808         (*saddsi_1): Remove.
13809         (*saddl_se_1): Ditto.
13810         (*ssubsi_1): Ditto.
13811         (*ssubl_se_1): Ditto.
13812         * config/alpha/predicates.md (const123_operand): New predicate.
13813         * config/alpha/constraints.md (P): Use IN_RANGE.
13815 2018-05-25  Richard Biener  <rguenther@suse.de>
13817         * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
13818         defaulted to true.
13819         (ref_maybe_used_by_stmt_p): Likewise.
13820         (stmt_may_clobber_ref_p): Likewise.
13821         (stmt_may_clobber_ref_p_1): Likewise.
13822         * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
13823         and pass it along.
13824         (ref_maybe_used_by_stmt_p): Likewise.
13825         (stmt_may_clobber_ref_p): Likewise.
13826         (stmt_may_clobber_ref_p_1): Likewise.
13827         * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
13828         the alias oracle to disambiguate DRs with stmts DR analysis
13829         couldn't handle.
13830         (vect_analyze_data_refs): Do not give up on not analyzable
13831         DRs for BB vectorization.  Remove code truncating the dataref
13832         vector.
13834 2018-05-25  Jakub Jelinek  <jakub@redhat.com>
13836         PR target/85832
13837         * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
13838         Add (=Yk,v,C) variant using vptestm insn.  Use TARGET_AVX512BW
13839         in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
13841 2018-05-25  Richard Biener  <rguenther@suse.de>
13843         * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
13844         function, combining stmt data ref gathering and fatal analysis
13845         parts.
13846         (vect_analyze_data_refs): Remove now redudnant code and simplify.
13847         * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
13848         vect_analyze_loop_2 and use vect_find_stmt_data_reference.
13849         * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
13850         * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
13852 2018-05-25  Bin Cheng  <bin.cheng@arm.com>
13854         PR tree-optimization/85720
13855         * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
13856         SCC if all partitions are builtins.
13857         (version_loop_by_alias_check): New parameter.  Generate cancelable
13858         runtime alias check if all partitions are builtins.
13859         (distribute_loop): Update call to above function.
13861 2018-05-25  Bin Cheng  <bin.cheng@arm.com>
13863         * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
13864         (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
13865         (parm_default_def_partition_arg): Ditto.
13866         (set_parm_default_def_partition): Ditto.
13867         (get_parm_default_def_partitions): Ditto and make it static.
13868         (get_undefined_value_partitions): Ditto and make it static.
13869         (remove_ssa_form): Refactor call to init_var_map here.
13870         * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
13871         computation for loop region.
13872         (coalesce_partitions, compute_optimized_partition_bases): Ditto.
13873         (register_default_def): Delete.
13874         (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
13875         (parm_default_def_partition_arg): Ditto.
13876         (set_parm_default_def_partition): Ditto.
13877         (get_parm_default_def_partitions): Ditto and make it static.
13878         (get_undefined_value_partitions): Ditto and make it static.
13879         (coalesce_with_default, coalesce_with_default): Update comment.
13880         (create_coalesce_list_for_region): New func factored out from
13881         create_outofssa_var_map.
13882         (populate_coalesce_list_for_outofssa): New func factored out from
13883         create_outofssa_var_map and coalesce_ssa_name.
13884         (create_outofssa_var_map): Delete.
13885         (coalesce_ssa_name): Refactor to support live range computation.
13886         * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
13887         (get_parm_default_def_partitions): Delete.
13888         (get_undefined_value_partitions): Ditto.
13889         * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
13890         computation for loop region.
13891         (new_tree_live_info, loe_visit_block): Ditto.
13892         (live_worklist, set_var_live_on_entry): Ditto.
13893         (calculate_live_on_exit, verify_live_on_entry): Ditto.
13894         * tree-ssa-live.h (struct _var_map): New fields.
13895         (init_var_map): Change decl.
13896         (region_contains_p): New.
13898 2018-05-25  Bin Cheng  <bin.cheng@arm.com>
13900         * tree-ssa-live.h (live_merge_and_clear): Delete.
13902 2018-05-25  Richard Biener  <rguenther@suse.de>
13904         PR c++/85912
13905         * tree-dump.c (dequeue_and_dump): Remove access to removed
13906         operand 2 of a SWITCH_EXPR.
13908 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
13910         * doc/sourcebuild.texi (vect_double_cond_arith): Include
13911         multiplication and division.
13912         * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
13913         (cond_udiv@var{m}, cond_umod@var{m}): Document.
13914         * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
13915         (cond_udiv_optab, cond_umod_optab): New optabs.
13916         * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
13917         (IFN_COND_RDIV): New internal functions.
13918         * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
13919         TRUNC_MOD_EXPR and RDIV_EXPR.
13920         * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
13921         * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
13922         New unspecs.
13923         (SVE_INT_BINARY): Include mult.
13924         (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
13925         (optab, sve_int_op): Handle mult.
13926         (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
13927         UNSPEC_COND_DIV.
13928         * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
13929         for SVE_INT_BINARY_SD.
13931 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
13933         * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
13934         (optab, sve_int_op): Handle div and udiv.
13935         * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
13936         for SVE_INT_BINARY_SD.
13937         (*<optab><mode>3): New insn for the same.
13939 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
13941         * tree-vect-patterns.c: Include predict.h.
13942         (vect_recog_divmod_pattern): Restrict check for division support
13943         to when optimizing for size.
13945 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
13947         * doc/sourcebuild.texi (vect_double_cond_arith: Document.
13948         * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
13949         (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
13950         (gimple_match_op::set_op): Likewise.
13951         (gimple_resimplify4): Declare.
13952         * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
13953         (expr::gen_transform): Likewise.
13954         (decision_tree::gen): Generate a simplification routine for 4 operands.
13955         * gimple-match-head.c (gimple_simplify): Add an overload for
13956         4 operands.  In the top-level function, handle up to 4 call
13957         arguments and call gimple_resimplify4.
13958         (gimple_resimplify4): New function.
13959         (build_call_internal): Pass a fourth operand.
13960         (maybe_push_to_seq): Likewise.
13961         * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
13962         Fold VEC_COND_EXPRs of an operation and a default value into
13963         an IFN_COND_* function if possible.
13964         * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
13965         New unspecs.
13966         (SVE_COND_FP_BINARY): Include them.
13967         (optab, sve_fp_op): Handle them.
13968         (SVE_INT_BINARY_REV): New code iterator.
13969         (SVE_COND_FP_BINARY_REV): New int iterator.
13970         (commutative): New int attribute.
13971         * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
13972         Declare.
13973         * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
13974         function.
13975         * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
13976         (*cond_<optab><mode>): New patterns for reversed operands.
13978 2018-05-25  Richard Biener  <rguenther@suse.de>
13980         * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
13981         (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
13982         (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
13983         (STMT_VINFO_GROUPED_ACCESS): Adjust.
13984         * tree-vect-data-refs.c (everywhere): Adjust users.
13985         * tree-vect-loop.c (everywhere): Likewise.
13986         * tree-vect-slp.c (everywhere): Likewise.
13987         * tree-vect-stmts.c (everywhere): Likewise.
13988         * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
13990 2018-05-25  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
13992         * configure.ac (gcc_cv_as_section_has_e): Move to common section.
13993         Rename to...
13994         (gcc_cv_as_section_exclude): ... this.
13995         Try Solaris as #exclude syntax.
13996         * configure: Regenerate.
13997         * config.in: Regenerate.
13998         * config/i386/i386.c (i386_solaris_elf_named_section): Handle
13999         SECTION_EXCLUDE.
14000         * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
14001         [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
14003         * varasm.c (default_elf_asm_named_section): Don't check if
14004         HAVE_GAS_SECTION_EXCLUDE is defined.
14006 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
14008         * doc/md.texi: Update the documentation of the cond_* optabs
14009         to mention the new final operand.  Fix GET_MODE_NUNITS call.
14010         Describe the scalar case too.
14011         * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
14012         * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
14013         instead of 2.
14014         (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
14015         (get_conditional_internal_fn): Update comment.
14016         * tree-vect-loop.c (vectorizable_reduction): Pass the original
14017         accumulator value as a final argument to conditional functions.
14018         * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
14019         a define_expand and add an "else" operand.  Assert for now that
14020         the else operand is equal to operand 2.  Use SVE_INT_BINARY and
14021         SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
14022         (*cond_<optab><mode>): New patterns.
14023         * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
14024         (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
14025         (UNSPEC_COND_EOR): Delete.
14026         (optab): Remove associated mappings.
14027         (SVE_INT_BINARY): New code iterator.
14028         (sve_int_op): Remove int attribute and add "minus" to the code
14029         attribute.
14030         (SVE_COND_INT_OP): Delete.
14031         (SVE_COND_FP_OP): Rename to...
14032         (SVE_COND_FP_BINARY): ...this.
14034 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
14036         * optabs.c (can_reuse_operands_p): New function.
14037         (maybe_legitimize_operands): Try to reuse the results for
14038         earlier operands.
14040 2018-05-24  Uros Bizjak  <ubizjak@gmail.com>
14042         * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
14043         Add {q} suffix to insn mnemonic.
14045 2018-05-23  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
14047         * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
14048         (msp430_warn_func_return): New.
14050 2018-05-24  Roger Sayle  <roger@nextmovesoftware.com>
14052         * fold-const.c (tree_nonzero_bits): New function.
14053         * fold-const.h (tree_nonzero_bits): Likewise.
14054         * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
14055         friends.  POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
14057 2018-05-24  H.J. Lu  <hongjiu.lu@intel.com>
14059         PR target/85900
14060         PR target/85345
14061         * varasm.c (assemble_alias): Check ifunc_resolver only on
14062         FUNCTION_DECL.
14064 2018-05-24  Uros Bizjak  <ubizjak@gmail.com>
14066         PR target/85903
14067         * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
14068         when memory input operand is handled.
14070 2018-05-24  Luis Machado  <luis.machado@linaro.org>
14072         * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
14073         global.
14074         (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
14076 2018-05-24  Richard Sandiford  <richard.sandiford@linaro.org>
14078         * match.pd: Delay FMA folds until after vectorization.
14080 2018-05-24  Andre Vieira  <andre.simoesdiasvieira@arm.com>
14082         PR target/83009
14083         * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
14084         address check not strict.
14086 2018-05-24  Richard Sandiford  <richard.sandiford@linaro.org>
14088         * gimple-match.h (gimple_match_op): New class.
14089         (mprts_hook): Replace parameters with a gimple_match_op *.
14090         (maybe_build_generic_op): Likewise.
14091         (gimple_simplified_result_is_gimple_val): Replace parameters with
14092         a const gimple_match_op *.
14093         (gimple_simplify): Replace code_helper * and tree * parameters with
14094         a gimple_match_op * parameter.
14095         (gimple_resimplify1): Replace code_helper *, tree and tree *
14096         parameters with a gimple_match_op * parameter.
14097         (gimple_resimplify2): Likewise.
14098         (gimple_resimplify3): Likewise.
14099         (maybe_push_res_to_seq): Replace code_helper, tree and tree *
14100         parameters with a gimple_match_op * parameter.
14101         * gimple-match-head.c (gimple_simplify): Change prototypes of
14102         auto-generated functions to take a gimple_match_op * instead of
14103         separate code_helper * and tree * parameters.  Make the same
14104         change in the top-level overload and update calls to the
14105         gimple_resimplify routines.  Update calls to the auto-generated
14106         functions and to maybe_push_res_to_seq in the publicly-facing
14107         operation-specific gimple_simplify overloads.
14108         (gimple_match_op::MAX_NUM_OPS): Define.
14109         (gimple_resimplify1): Replace rcode and ops with a single res_op
14110         parameter.  Update call to gimple_simplify.
14111         (gimple_resimplify2): Likewise.
14112         (gimple_resimplify3): Likewise.
14113         (mprts_hook): Replace parameters with a gimple_match_op *.
14114         (maybe_build_generic_op): Likewise.
14115         (build_call_internal): Replace type, nargs and ops with
14116         a gimple_match_op *.
14117         (maybe_push_res_to_seq): Replace res_code, type and ops parameters
14118         with a single gimple_match_op *.  Update calls to mprts_hook,
14119         build_call_internal and gimple_simplified_result_is_gimple_val.
14120         Factor out code that is common to the tree_code and combined_fn cases.
14121         * genmatch.c (expr::gen_transform): Replace tem_code and
14122         tem_ops with a gimple_match_op called tem_op.  Update calls
14123         to the gimple_resimplify functions and maybe_push_res_to_seq.
14124         (dt_simplify::gen_1): Manipulate res_op instead of res_code and
14125         res_ops.  Update call to the gimple_resimplify functions.
14126         (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
14127         (decision_tree::gen): Make the functions take a gimple_match_op *
14128         called res_op instead of separate res_code and res_ops parameters.
14129         Update call accordingly.
14130         * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
14131         and ops with a single res_op parameter.  Update calls to
14132         maybe_build_generic_op and maybe_push_res_to_seq.
14133         (fold_stmt_1): Update calls to gimple_simplify and
14134         replace_stmt_with_simplification.
14135         (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
14136         and gimple_simplified_result_is_gimple_val.
14137         * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
14138         gimple_simplify.
14139         * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
14140         with a gimple_match_op *.
14141         (vn_nary_build_or_lookup): Likewise.  Update call to
14142         vn_nary_build_or_lookup_1.
14143         (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
14144         gimple_match_op *.  Update calls to the gimple_resimplify routines
14145         and to gimple_simplified_result_is_gimple_val.
14146         (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
14147         Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
14148         (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
14149         (visit_nary_op): Likewise.
14150         (visit_reference_op_load): Likewise.
14152 2018-05-23  Luis Machado  <luis.machado@linaro.org>
14154         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
14155         modifier for printing the step amount.
14157 2018-05-23  Jozef Lawrynowicz  <jozef.l@somniumtech.com>
14159         PR target/78849
14160         * tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
14161         types.
14163 2018-05-23  Segher Boessenkool  <segher@kernel.crashing.org>
14165         * doc/sourcebuild.texi (Endianness): New subsubsection.
14167 2018-05-23  Luis Machado  <luis.machado@linaro.org>
14169         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
14170         <prefetch_dynamic_strides>: New const bool field.
14171         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
14172         prefetch_dynamic_strides.
14173         (exynosm1_prefetch_tune): Likewise.
14174         (thunderxt88_prefetch_tune): Likewise.
14175         (thunderx_prefetch_tune): Likewise.
14176         (thunderx2t99_prefetch_tune): Likewise.
14177         (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
14178         false.
14179         (aarch64_override_options_internal): Update to set
14180         PARAM_PREFETCH_DYNAMIC_STRIDES.
14181         * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
14182         * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
14183         * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
14184         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
14185         prefetch-dynamic-strides setting.
14187 2018-05-23  Luis Machado  <luis.machado@linaro.org>
14189         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
14190         <minimum_stride>: New const int field.
14191         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
14192         minimum_stride field defaulting to -1.
14193         (exynosm1_prefetch_tune): Likewise.
14194         (thunderxt88_prefetch_tune): Likewise.
14195         (thunderx_prefetch_tune): Likewise.
14196         (thunderx2t99_prefetch_tune): Likewise.
14197         (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
14198         <default_opt_level>: Set to 3.
14199         (aarch64_override_options_internal): Update to set
14200         PARAM_PREFETCH_MINIMUM_STRIDE.
14201         * doc/invoke.texi (prefetch-minimum-stride): Document new option.
14202         * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
14203         * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
14204         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
14205         stride is constant and is below the minimum stride threshold.
14207 2018-05-23  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
14209         * config/arm/arm-cpus.in (mode26): Delete.
14210         (armv4): Delete mode26 reference.
14211         * config/arm/arm.c (arm_configure_build_target): Delete use of
14212         isa_bit_mode26.
14214 2018-05-23  Uros Bizjak  <ubizjak@gmail.com>
14216         * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
14217         New insn pattern.
14218         (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
14219         Rewrite expander pattern.  Emit gen_floatunssi<mode>2_i387_with_xmm
14220         for non-SSE modes.
14221         (floatunsdisf2): Rewrite expander pattern.  Hanlde TARGET_AVX512F.
14222         (floatunsdidf2): Ditto.
14224 2018-05-23  Uros Bizjak  <ubizjak@gmail.com>
14226         * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
14227         (fixuns_trunc<mode>si2_avx512f): Ditto.
14228         (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
14229         (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
14230         Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
14232 2018-05-23  Alexander Monakov  <amonakov@ispras.ru>
14234         PR rtl-optimization/79985
14235         * df-scan.c (df_insn_refs_collect): Remove special case for
14236         global registers and asm statements.
14238 2018-05-23  Alexander Monakov  <amonakov@ispras.ru>
14240         * extend.texi (Global Register Variables): Rewrite the bullet list.
14241         Note that the register is available for allocation. Note that access
14242         via inline asm must use constraints. Add note about async-signal
14243         handlers. Remove paragraph about automagic register selection.
14245 2018-05-23  Richard Biener  <rguenther@suse.de>
14247         * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
14248         of fixed offset from memset VN.
14250 2018-05-23  Bill Schmidt  <wschmidt@linux.ibm.com>
14252         * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
14253         first_interp field.
14254         (alloc_cand_and_find_basis): Initialize first_interp field.
14255         (slsr_process_mul): Modify first_interp field.
14256         (slsr_process_add): Likewise.
14257         (slsr_process_cast): Modify first_interp field for each new
14258         interpretation.
14259         (slsr_process_copy): Likewise.
14260         (dump_candidate): Dump first_interp field.
14261         (replace_mult_candidate): Process all interpretations, not just
14262         subsequent ones.
14263         (replace_rhs_if_not_dup): Likewise.
14264         (replace_one_candidate): Likewise.
14266 2018-05-23  Wilco Dijkstra  <wdijkstr@arm.com>
14268         * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
14269         Add new boolean.
14270         (aarch64_needs_frame_chain): New function.
14271         (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
14273 2018-05-23  Sudakshina Das  <sudi.das@arm.com>
14275         PR target/84882
14276         * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
14277         Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
14278         * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
14279         * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
14280         as true for strict-align.
14281         (aarch64_can_inline_p): Perform checks even when callee has no
14282         attributes to check for strict alignment.
14283         * doc/extend.texi (AArch64 Function Attributes): Document
14284         no-strict-align.
14285         * doc/invoke.texi: (AArch64 Options): Likewise.
14287 2018-05-23  Richard Sandiford  <richard.sandiford@linaro.org>
14289         PR tree-optimization/85853
14290         * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
14291         the handling of the root of the node to...
14292         (vect_slp_analyze_node_operations_1): ...this new function,
14293         and run the whole thing with the child nodes' def types
14294         set according to their SLP node's def type.
14296 2018-05-23  Richard Biener  <rguenther@suse.de>
14298         PR middle-end/85874
14299         * tree-data-ref.c (create_runtime_alias_checks): Defer
14300         and ignore overflow warnings.
14302 2018-05-23  Yury Gribov  <tetra2005@gmail.com>
14304         PR tree-optimization/85822
14305         * tree-vrp.c (is_masked_range_test): Fix handling of negative
14306         constants.
14308 2018-05-23  Richard Biener  <rguenther@suse.de>
14310         * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
14311         memset constants via native_interpret_expr.
14313 2018-05-22  H.J. Lu  <hongjiu.lu@intel.com>
14315         PR target/85345
14316         * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
14317         attribute.
14318         (cgraph_node::create_alias): Likewise.
14319         (cgraph_node::get_availability): Check ifunc_resolver instead
14320         of looking up ifunc attribute.
14321         * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
14322         * varasm.c (do_assemble_alias): Likewise.
14323         (assemble_alias): Likewise.
14324         (default_binds_local_p_3): Likewise.
14325         * cgraph.h (cgraph_node): Add ifunc_resolver.
14326         (cgraph_node::only_called_directly_or_aliased_p): Return false
14327         for IFUNC resolver.
14328         * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
14329         attribute.
14330         * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
14331         is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
14332         (symtab_node::binds_to_current_def_p): Check ifunc_resolver
14333         instead of looking up ifunc attribute.
14335 2018-05-22  Luis Machado  <luis.machado@linaro.org>
14337         * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
14339 2018-05-22  Martin Sebor  <msebor@redhat.com>
14341         PR middle-end/85359
14342         * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
14343         only when expasion succeeds.
14344         (expand_builtin_strcmp): Same.
14345         (expand_builtin_strncmp): Same.
14347 2018-05-22  Martin Sebor  <msebor@redhat.com>
14349         * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
14351 2018-05-22  Jackson Woodruff  <jackson.woodruff@arm.com>
14352             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
14354         * config/aarch64/aarch64-ldpstp.md: Replace uses of
14355         aarch64_mem_pair_operand with memory_operand and delete operand swapping
14356         code.
14357         * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
14358         Add check for legitimate_address.
14359         (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
14360         (aarch64_swap_ldrstr_operands): New.
14361         * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
14362         Define prototype.
14364 2018-05-22  Jackson Woodruff  <jackson.woodruff@arm.com>
14365             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
14367         * config/aarch64/aarch64.md: New patterns to generate stp
14368         and ldp.
14369         (store_pair_sw, store_pair_dw): New patterns to generate stp for
14370         single words and double words.
14371         (load_pair_sw, load_pair_dw): Likewise.
14372         (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
14373         Delete.
14374         (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
14375         Delete.
14376         * config/aarch64/aarch64-ldpstp.md: Modify peephole
14377         for different mode ldpstp and add peephole for merged zero stores.
14378         Likewise for loads.
14379         * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
14380         Add size check.
14381         (aarch64_gen_store_pair): Rename calls to match new patterns.
14382         (aarch64_gen_load_pair): Rename calls to match new patterns.
14383         * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
14384         (load_pair<DREG:mode><DREG2:mode>): ... This.
14385         (store_pair<mode>): Rename to...
14386         (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
14387         * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
14388         New mode iterators.
14389         (V_INT_EQUIV): Handle SImode.
14390         * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
14391         New predicate.
14393 2018-05-22  Martin Sebor  <msebor@redhat.com>
14395         PR c/85623
14396         * calls.c (maybe_warn_nonstring_arg): Use string length to set
14397         or ajust the presumed bound on an operation to avoid unnecessary
14398         warnings.
14400 2018-05-22  Martin Sebor  <msebor@redhat.com>
14402         PR tree-optimization/85826
14403         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
14404         assuming that a DECL necesarily has a constant size.
14406 2018-05-22  Richard Sandiford  <richard.sandiford@linaro.org>
14408         PR middle-end/85862
14409         * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
14411 2018-05-22  Richard Biener  <rguenther@suse.de>
14413         PR tree-optimization/85834
14414         * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
14415         non-constant and non-zero memset arguments.
14417 2018-05-22  Martin Liska  <mliska@suse.cz>
14419         PR ipa/85607
14420         * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
14422 2018-05-22  Richard Biener  <rguenther@suse.de>
14424         PR tree-optimization/85863
14425         * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
14426         comparisons when vectype is specified.
14427         (vectorizable_condition): Do not specify vectype for
14428         vect_is_simple_cond when SLP vectorizing.
14430 2018-05-21  Michael Meissner  <meissner@linux.ibm.com>
14432         PR target/85657
14433         * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
14434         define __ibm128 as long double.
14435         * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
14436         as a distinct type when IEEE 128-bit support is enabled.
14437         (init_float128_ieee): Fix up conversions between IFmode and IEEE
14438         128-bit types to use the correct functions.
14439         (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
14440         convert between 128-bit floating point types that have different
14441         modes but the same representation, instead of using gen_lowpart to
14442         makean alias.
14443         * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
14444         KFmode.
14445         (IFKF_reg): New attributes to give the register constraints for
14446         IFmode and KFmode.
14447         (extend<mode>tf2_internal): New insns to mark an explicit
14448         conversion between 128-bit floating point types that have a
14449         different mode but share the same representation.
14451 2018-05-21  Richard Sandiford  <richard.sandiford@linaro.org>
14453         PR tree-optimization/85814
14454         * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
14455         a null return from get_strinfo when unsharing the next
14456         strinfo in the chain.
14458 2018-05-21  Vladimir Mezentsev  <vladimir.mezentsev@oracle.com>
14460         PR gcc/84923
14461         * varasm.c (weak_finish): Clean up weak_decls.
14463 2018-05-21  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
14465         * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
14466         UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
14467         UNSPEC_UADALP values.
14468         * config/aarch64/iterators.md (ABAL): New int iterator.
14469         (ABDL2): Likewise.
14470         (ADALP): Likewise.
14471         (sur): Add mappings for the above.
14472         * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
14473         New define_insn.
14474         (aarch64_<sur>abal<mode>_4): Likewise.
14475         (aarch64_<sur>adalp<mode>_3): Likewise.
14476         (<sur>sadv16qi): New define_expand.
14478 2018-05-21  Alexander Nesterovskiy  <alexander.nesterovskiy@intel.com>
14480         * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
14481         (*movdf_internal): Ditto.
14482         (*rcpsf2_sse): Ditto.
14483         (*rsqrtsf2_sse): Ditto.
14484         (*sqrt<mode>2_sse): Ditto.
14486 2018-05-21  Tamar Christina  <tamar.christina@arm.com>
14488         * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
14489         eor3q<mode>4.
14490         (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
14491         * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
14492         veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
14493         vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
14494         vbcaxq_s64): New.
14495         * config/aarch64/arm_neon.h: Likewise.
14496         * config/aarch64/iterators.md (VQ_I): New.
14498 2018-05-21  Alexey Brodkin <abrodkin@synopsys.com>
14500         * config.gcc: Add arc/t-multilib-linux to tmake_file for
14501         arc*-*-linux*.
14502         * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
14503         MULTILIB_DIRNAMES
14505 2018-05-20  Chung-Ju Wu  <jasonwucj@gmail.com>
14507         * config/nds32/constraints.md (S): New constraint.
14508         * config/nds32/nds32.md (call_internal): Use constraint S.
14509         (call_value_internal): Likewise.
14510         (sibcall_internal): Likewise.
14511         (sibcall_value_internal): Likewise.
14513 2018-05-20  Kito Cheng  <kito.cheng@gmail.com>
14514             Chung-Ju Wu  <jasonwucj@gmail.com>
14516         * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
14517         into consideration.
14519 2018-05-20  Kito Cheng  <kito.cheng@gmail.com>
14520             Chung-Ju Wu  <jasonwucj@gmail.com>
14522         * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
14523         (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
14524         (nds32_rtx_costs_impl): Simplify.
14525         (nds32_address_cost_impl): Simplify.
14526         (nds32_init_rtx_costs): New function.
14527         (nds32_rtx_costs_speed_prefer): Likewise.
14528         (nds32_rtx_costs_size_prefer): Likewise.
14529         (nds32_address_cost_speed_prefer): Likewise.
14530         (nds32_address_cost_speed_fwprop): Likewise.
14531         (nds32_address_cost_size_prefer): Likewise.
14532         * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
14533         * config/nds32/nds32.c (nds32_option_override): Use
14534         nds32_init_rtx_costs function.
14536 2018-05-20  Chung-Ju Wu  <jasonwucj@gmail.com>
14538         * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
14539         * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
14540         (TARGET_PIPELINE_N8): Likewise.
14541         (TARGET_PIPELINE_N10): Likewise.
14542         (TARGET_PIPELINE_N13): Likewise.
14543         (TARGET_PIPELINE_GRAYWOLF): Likewise.
14545 2018-05-19  Monk Chiang  <sh.chiang04@gmail.com>
14547         * config/nds32/nds32-fpu.md: Update copyright year.
14549 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
14551         * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
14553 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
14555         * config/nds32/nds32.c
14556         (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
14557         * config/nds32/nds32.opt (minline-asm-r15): New option.
14559 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
14561         * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
14562         MASK_HW_ABS.
14563         * config/nds32/nds32.md (abssi2): New pattern.
14565 2018-05-19  Uros Bizjak  <ubizjak@gmail.com>
14567         * config/i386/i386.md (rex64namesuffix): New mode attribute.
14568         * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
14569         Merge insn pattern from sse_cvtsi2ss<round_name> and
14570         sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
14571         (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
14572         from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
14573         using SWI48 mode iterator.
14574         (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
14575         sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
14576         (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
14577         pattern from sse_cvttss2si<round_saeonly_name>
14578         and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
14579         (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
14580         from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
14581         using SWI48 mode iterator.
14582         (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
14583         insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
14584         avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
14585         (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
14586         from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
14587         using SWI48 mode iterator.
14588         (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
14589         insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
14590         avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
14591         (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
14592         sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
14593         SWI48 mode iterator.
14594         (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
14595         sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
14596         (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
14597         pattern from sse_cvttsd2si<round_saeonly_name>
14598         and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
14600 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
14602         * config/nds32/nds32-md-auxiliary.c
14603         (nds32_valid_smw_lwm_base_p): Refine.
14604         (nds32_output_smw_single_word): Refine.
14605         (nds32_output_smw_double_word): New.
14606         * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
14608 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
14610         * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
14611         (nds32_output_stack_pop): Refine.
14612         (nds32_expand_unaligned_load): Refine.
14613         (nds32_expand_unaligned_store): Refine.
14615 2018-05-19  Kuan-Lin Chen <kuanlinchentw@gmail.com>
14616             Chung-Ju Wu  <jasonwucj@gmail.com>
14618         * config/nds32/constants.md: Add TP_REGNUM constant.
14619         (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
14620         UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
14621         UNSPEC_ADD32.
14622         * config/nds32/nds32-doubleword.md: Consider flag_pic.
14623         * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
14624         * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
14625         * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
14626         and PIC code generation.
14627         * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
14628         code generation.
14629         * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
14630         optimization.
14631         * config/nds32/nds32.md: Support TLS and PIC.
14632         * config/nds32/nds32.c: Support TLS and PIC.
14633         * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
14634         * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
14635         predicate.
14637 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
14639         * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
14640         mode with E_ prefix.
14642 2018-05-19  Kuan-Lin Chen <kuanlinchentw@gmail.com>
14643             Chung-Ju Wu  <jasonwucj@gmail.com>
14645         * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
14646         * config/nds32/nds32-md-auxiliary.c
14647         (symbolic_reference_mentioned_p): New.
14648         (nds32_legitimize_ict_address): New.
14649         (nds32_expand_ict_move): New.
14650         (nds32_indirect_call_referenced_p): New.
14651         (nds32_symbol_binds_local_p): Delete.
14652         (nds32_long_call_p): Modify.
14653         * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
14654         * config/nds32/nds32-protos.h
14655         (symbolic_reference_mentioned_p): Declare.
14656         (nds32_legitimize_ict_address): Declare.
14657         (nds32_expand_ict_move): Declare.
14658         (nds32_indirect_call_referenced_p): Declare.
14659         * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
14660         (nds32_relax_group): Use nds32_ict_const_p as condition.
14661         * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
14662         (nds32_asm_file_start): Output ict_model directive in asm code.
14663         (nds32_legitimate_address_p): Consider indirect call.
14664         (nds32_print_operand): Consider indirect call.
14665         (nds32_print_operand_address): Consider indirect call.
14666         (nds32_insert_attributes): Handle "indirect_call" attribute.
14667         (TARGET_LEGITIMATE_ADDRESS_P): Define.
14668         (TARGET_LEGITIMATE_CONSTANT_P): Define.
14669         (TARGET_CANNOT_FORCE_CONST_MEM): Define.
14670         (TARGET_DELEGITIMIZE_ADDRESS): Define.
14671         (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
14672         * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
14673         (TARGET_ICT_MODEL_SMALL): Define.
14674         (TARGET_ICT_MODEL_LARGE): Define.
14675         * config/nds32/nds32.md (movsi): Consider ict model.
14676         (call, call_value): Consider ict model.
14677         (sibcall, sibcall_value): Consider ict model.
14678         * config/nds32/nds32.opt (mict-model): New option.
14679         * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
14680         model.
14682 2018-05-18  Kito Cheng  <kito.cheng@gmail.com>
14683             Monk Chiang  <sh.chiang04@gmail.com>
14684             Jim Wilson <jimw@sifive.com>
14686         * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
14687         Add support to parse rv32e*.  Clear MASK_RVE for rv32i and rv64i.
14688         * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
14689         * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
14690         __riscv_32e when TARGET_RVE.  Handle ABI_ILP32E as soft-float ABI.
14691         * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
14692         * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
14693         compute save_libcall_adjustment properly.
14694         (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
14695         (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
14696         * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
14697         (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
14698         (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
14699         (ABI_SPEC): Handle mabi=ilp32e.
14700         * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
14701         (RVE): Add RVE mask.
14702         * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
14703         <-march>: Add rv32e as an example.
14705 2018-05-18  Marc Glisse  <marc.glisse@inria.fr>
14707         PR c++/82899
14708         * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
14709         (intra_create_variable_infos): Handle C++ constructors.
14711 2018-05-18  Martin Liska  <mliska@suse.cz>
14713         * passes.def: Remove a redundant pass.
14715 2018-05-18  Eric Botcazou  <ebotcazou@adacore.com>
14717         PR bootstrap/85838
14718         * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
14720 2018-05-18  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
14722         * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
14723         (ARMv4): Update.
14724         (ARMv2, ARMv3, ARMv3m): Delete fgroups.
14725         (ARMv6m): Update.
14726         (armv2, armv2a, armv3, armv3m): Delete architectures.
14727         (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
14728         arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
14729         arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
14730         Delete cpus.
14731         * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
14732         (*mulsidi3adddi): Likewise.
14733         (mulsidi3): Likewise.
14734         (*mulsidi3_nov6): Likewise.
14735         (umulsidi3): Likewise.
14736         (umulsidi3_nov6): Likewise.
14737         (umaddsidi4): Likewise.
14738         (*umulsidi3adddi): Likewise.
14739         (smulsi3_highpart): Likewise.
14740         (*smulsi3_highpart_nov6): Likewise.
14741         (umulsi3_highpart): Likewise.
14742         (*umulsi3_highpart_nov6): Likewise.
14743         * config/arm/arm.h (arm_arch3m): Delete.
14744         * config/arm/arm.c (arm_arch3m): Delete.
14745         (arm_option_override_internal): Update armv3-related comment.
14746         (arm_configure_build_target): Delete use of isa_bit_mode32.
14747         (arm_option_reconfigure_globals): Delete set of arm_ach3m.
14748         (arm_rtx_costs_internal): Delete check of arm_arch3m.
14749         * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
14750         (mulsa3): Likewise.
14751         (mulusa3): Likewise.
14752         * config/arm/arm-protos.h (arm_arch3m): Delete.
14753         * config/arm/arm-tables.opt: Regenerate.
14754         * config/arm/arm-tune.md: Likewise.
14755         * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
14756         deleted architectures.
14758 2018-05-18  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
14760         * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
14761         (armv5t, armv5te): New features.
14762         (ARMv5, ARMv5e): Delete fgroups.
14763         (ARMv5t, ARMv5te): Adjust for above changes.
14764         (ARMv6m): Likewise.
14765         (armv5, armv5e): Delete arches.
14766         * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
14767         arm_arch5.
14768         (*call_reg_arm): Likewise.
14769         (*call_value_reg_armv5): Likewise.
14770         (*call_value_reg_arm): Likewise.
14771         (*call_symbol): Likewise.
14772         (*call_value_symbol): Likewise.
14773         (*sibcall_insn): Likewise.
14774         (*sibcall_value_insn): Likewise.
14775         (clzsi2): Likewise.
14776         (prefetch): Likewise.
14777         (define_split and define_peephole2 dependent on arm_arch5):
14778         Likewise.
14779         * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
14780         arm_arch5e.
14781         (TARGET_ARM_QBIT): Likewise.
14782         (TARGET_DSP_MULTIPLY): Likewise.
14783         (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
14784         (arm_arch5, arm_arch5e): Delete.
14785         (arm_arch5t, arm_arch5te): Declare.
14786         * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
14787         (arm_arch5t): Declare.
14788         (arm_option_reconfigure_globals): Update for the above.
14789         (arm_options_perform_arch_sanity_checks): Update comment, replace
14790         use of arm_arch5 with arm_arch5t.
14791         (use_return_insn): Likewise.
14792         (arm_emit_call_insn): Likewise.
14793         (output_return_instruction): Likewise.
14794         (arm_final_prescan_insn): Likewise.
14795         (arm_coproc_builtin_available): Likewise.
14796         * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
14797         arm_arch5e with arm_arch5t and arm_arch5te.
14798         * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
14799         (arm_arch5t, arm_arch5te): Declare.
14800         * config/arm/arm-tables.opt: Regenerate.
14801         * config/arm/t-arm-elf: Remove references to armv5, armv5e.
14802         * config/arm/t-multilib: Likewise.
14803         * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
14804         instead of arm_arch5.
14805         (*call_reg_thumb1): Likewise.
14806         (*call_value_reg_thumb1_v5): Likewise.
14807         (*call_value_reg_thumb1): Likewise.
14808         * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
14809         unreachable path.
14810         * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
14812 2018-05-18  Martin Liska  <mliska@suse.cz>
14814         PR gcov-profile/84846
14815         * doc/gcov.texi: Document -t option of gcov tool.
14817 2018-05-18  Martin Liska  <mliska@suse.cz>
14819         PR gcov-profile/84846
14820         * gcov.c (print_usage): Add new -t option.
14821         (process_args): Handle the option.
14822         (generate_results): Use stdout as output when requested by
14823         the option.
14825 2018-05-18  Martin Liska  <mliska@suse.cz>
14827         PR gcov-profile/84846
14828         * coverage.c (coverage_init): Write PWD to .gcno file.
14829         * doc/gcov.texi: Document how working directory is printed.
14830         * gcov-dump.c (dump_gcov_file): Print PWD.
14831         * gcov.c (output_intermediate_file): Likewise.
14832         (read_graph_file): Read PWD string.
14833         (output_lines): Print PWD.
14835 2018-05-18  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
14837         PR middle-end/85817
14838         * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
14839         for retval and return false if all args to phi are zero.
14841 2018-05-18  Richard Biener  <rguenther@suse.de>
14843         * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
14844         method.
14845         (evrp_dom_walker::before_dom_children): Call it.
14847 2018-05-18  Richard Biener  <rguenther@suse.de>
14849         * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
14850         results when processing array refs with variable index.
14852 2018-05-18  Toon Moene  <toon@moene.org>
14854         * doc/invoke.texi: Move -floop-unroll-and-jam documentation
14855         directly after that of -floop-interchange. Indicate that both
14856         options are enabled by default when specifying -O3.
14858 2018-05-18  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
14860         * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
14861         iterator.  Delete separate integer-mode vec_set<mode> expander.
14862         (aarch64_simd_vec_setv2di): Delete.
14863         (vec_setv2di): Delete.
14864         (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
14865         Use VALL_F16 mode iterator.  Add LD1 alternative and use vwcore for
14866         the "w, r" alternative.
14868 2018-05-18  Martin Liska  <mliska@suse.cz>
14870         * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
14871         * tree-pass.h (make_pass_lower_switch_O0): New function.
14872         * tree-switch-conversion.c (node_has_low_bound): Remove.
14873         (node_has_high_bound): Likewise.
14874         (node_is_bounded): Likewise.
14875         (class pass_lower_switch): Make it a template type and create
14876         two instances.
14877         (pass_lower_switch::execute): Add template argument.
14878         (make_pass_lower_switch): New function.
14879         (make_pass_lower_switch_O0): New function.
14880         (do_jump_if_equal): Remove.
14881         (emit_case_nodes): Simplify to just handle all 3 cases and leave
14882         all the hard work to tree optimization passes.
14884 2018-05-18  Martin Liska  <mliska@suse.cz>
14886         * dbgcnt.c (limit_low): Renamed from limit.
14887         (limit_high): New variable.
14888         (dbg_cnt_is_enabled): Check for upper limit.
14889         (dbg_cnt): Adjust dumping.
14890         (dbg_cnt_set_limit_by_index): Add new argument for high
14891         value.
14892         (dbg_cnt_set_limit_by_name): Likewise.
14893         (dbg_cnt_process_single_pair): Parse new format.
14894         (dbg_cnt_process_opt): Use strtok.
14895         (dbg_cnt_list_all_counters): Remove 'value' and add
14896         'limit_high'.
14897         * doc/invoke.texi: Document changes.
14899 2018-05-18  Richard Sandiford  <richard.sandiford@linaro.org>
14901         * doc/sourcebuild.texi (scalar_all_fma): Document.
14902         * tree.def (FMA_EXPR): Delete.
14903         * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
14904         * internal-fn.c (ternary_direct): New macro.
14905         (expand_ternary_optab_fn): Likewise.
14906         (direct_ternary_optab_supported_p): Likewise.
14907         * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
14908         * builtins.c (fold_builtin_fma): Delete.
14909         (fold_builtin_3): Don't call it.
14910         * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
14911         * expr.c (expand_expr_real_2): Likewise.
14912         * fold-const.c (operand_equal_p): Likewise.
14913         (fold_ternary_loc): Likewise.
14914         * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
14915         * gimple.c (DEFTREECODE): Likewise.
14916         * gimplify.c (gimplify_expr): Likewise.
14917         * optabs-tree.c (optab_for_tree_code): Likewise.
14918         * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
14919         * tree-eh.c (operation_could_trap_p): Likewise.
14920         (stmt_could_throw_1_p): Likewise.
14921         * tree-inline.c (estimate_operator_cost): Likewise.
14922         * tree-pretty-print.c (dump_generic_node): Likewise.
14923         (op_code_prio): Likewise.
14924         * tree-ssa-loop-im.c (stmt_cost): Likewise.
14925         * tree-ssa-operands.c (get_expr_operands): Likewise.
14926         * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
14927         * fold-const-call.h (fold_fma): Delete.
14928         * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
14929         CFN_FNMA and CFN_FNMS.
14930         (fold_fma): Delete.
14931         * genmatch.c (combined_fn): New enum.
14932         (commutative_ternary_tree_code): Remove FMA_EXPR handling.
14933         (commutative_op): New function.
14934         (commutate): Use it.  Handle more than 2 operands.
14935         (dt_operand::gen_gimple_expr): Use commutative_op.
14936         (parser::parse_expr): Allow :c to be used with non-binary
14937         operators if the commutative operand is known.
14938         * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
14939         CFN_FMS, CFN_FNMA and CFN_FNMS.
14940         (backprop::process_assign_use): Remove FMA_EXPR handling.
14941         * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
14942         (gen_hsa_fma): New function.
14943         (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
14944         IFN_FNMA and IFN_FNMS.
14945         * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
14946         * gimple-fold.h (follow_all_ssa_edges): Declare.
14947         * gimple-fold.c (follow_all_ssa_edges): New function.
14948         * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
14949         gimple_build interface and use follow_all_ssa_edges to fold the result.
14950         (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
14951         instead of checking for optabs directly.
14952         * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
14953         rather than FMA_EXPRs.
14954         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
14955         call to IFN_FMA instead of an FMA_EXPR.
14957 2018-05-17  Jim Wilson  <jimw@sifive.com>
14959         * expr.c (do_tablejump): When converting index to Pmode, if we have a
14960         sign extended promoted subreg, and the range does not have the sign bit
14961         set, then do a sign extend.
14963         * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
14964         test, check for sign extended subreg and/or constant operands, and
14965         do a sign extend in that case.
14967 2018-05-17  Steve Ellcey  <sellcey@cavium.com>
14969         * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
14970         (thunderx2t99_multiple): Delete psuedo-units from used cpus.
14971         Add untyped.
14972         (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
14973         Change logics_shift_reg to logics_shift_imm.
14974         (thunderx2t99_fp_loadpair_basic): Delete.
14975         (thunderx2t99_fp_storepair_basic): Delete.
14976         (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
14977         (thunderx2t99_asimd_polynomial): Delete.
14978         (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
14979         and neon_fp_mul_d_scalar_q.
14980         (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
14981         (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
14982         (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
14983         (thunderx2t99_asimd_lut): Add missing tbl types.
14984         (thunderx2t99_asimd_ext): Delete.
14985         (thunderx2t99_asimd_load1_1_mult): Delete.
14986         (thunderx2t99_asimd_load1_2_mult): Delete.
14987         (thunderx2t99_asimd_load1_ldp): New.
14988         (thunderx2t99_asimd_load1): New.
14989         (thunderx2t99_asimd_load2): Add missing *load2* types.
14990         (thunderx2t99_asimd_load3): New.
14991         (thunderx2t99_asimd_load4): New.
14992         (thunderx2t99_asimd_store1_1_mult): Delete.
14993         (thunderx2t99_asimd_store1_2_mult): Delete.
14994         (thunderx2t99_asimd_store2_mult): Delete.
14995         (thunderx2t99_asimd_store2_onelane): Delete.
14996         (thunderx2t99_asimd_store_stp): New.
14997         (thunderx2t99_asimd_store1): New.
14998         (thunderx2t99_asimd_store2): New.
14999         (thunderx2t99_asimd_store3): New.
15000         (thunderx2t99_asimd_store4): New.
15002 2018-05-17  Jerome Lambourg  <lambourg@adacore.com>
15004         * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
15005         #include <stdint.h>.  Replace intptr_t with __INTPTR_TYPE__.
15007 2018-05-17  Pat Haugen  <pthaugen@us.ibm.com>
15008             Segher Boessenkool  <segher@kernel.crashing.org>
15010         PR target/85698
15011         * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
15012         operand.
15014 2018-05-17  Richard Biener  <rguenther@suse.de>
15016         * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
15017         for pruning loop and prune defs feeding only already visited PHIs.
15019 2018-05-17  Richard Biener  <rguenther@suse.de>
15021         * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
15023 2018-05-17  Bin Cheng  <bin.cheng@arm.com>
15024             Richard Biener  <rguenther@suse.de>
15026         PR tree-optimization/85793
15027         * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
15028         for VMAT_ELEMENTWISE.
15030 2018-05-17  Richard Sandiford  <richard.sandiford@linaro.org>
15032         * internal-fn.h (lookup_internal_fn): Declare
15033         * internal-fn.c (lookup_internal_fn): New function.
15034         * gimple.c (gimple_build_call_from_tree): Handle calls to
15035         internal functions.
15036         * gimple-pretty-print.c (dump_gimple_call): Print "." before
15037         internal function names.
15038         * tree-pretty-print.c (dump_generic_node): Likewise.
15039         * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
15041 2018-05-17  Richard Sandiford  <richard.sandiford@linaro.org>
15043         * gimple-fold.h (gimple_build): Make the function forms take
15044         combined_fn rather than built_in_function.
15045         (gimple_simplify): Likewise.
15046         * gimple-match-head.c (gimple_simplify): Likewise.
15047         * gimple-fold.c (gimple_build): Likewise.
15048         * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
15049         rather than gimple_build_call_internal.
15050         (get_initial_defs_for_reduction): Likewise.
15051         (vect_create_epilog_for_reduction): Likewise.
15052         (vectorizable_live_operation): Likewise.
15054 2018-05-17  Martin Liska  <mliska@suse.cz>
15056         * gimple-ssa-sprintf.c (format_directive): Do not use
15057         space in between 'G_' and '('.
15059 2018-05-17  Jakub Jelinek  <jakub@redhat.com>
15061         PR target/85323
15062         * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
15063         even if the mask is not all ones.
15065         PR target/85323
15066         * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
15067         vector.
15068         (ix86_gimple_fold_builtin): Likewise.
15070         PR target/85323
15071         * config/i386/i386.c: Include tree-vector-builder.h.
15072         (ix86_vector_shift_count): New function.
15073         (ix86_fold_builtin): Fold shift builtins by scalar count.
15074         (ix86_gimple_fold_builtin): Likewise.
15076         * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
15077         _mm512_setzero): New intrinsics.
15079 2018-05-17  James Greenhalgh  <james.greenhalgh@arm.com>
15080             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
15082         * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
15083         code generation for cases where splatting a value is not useful.
15084         * simplify-rtx.c (simplify_ternary_operation): Simplify
15085         vec_merge across a vec_duplicate and a paradoxical subreg forming
15086         a vector mode to a vec_concat.
15088 2018-05-17  Olga Makhotina  <olga.makhotina@intel.com>
15090         * config.gcc: Support "goldmont-plus".
15091         * config/i386/driver-i386.c (host_detect_local_cpu): Detect
15092         "goldmont-plus".
15093         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
15094         PROCESSOR_GOLDMONT_PLUS.
15095         * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
15096         (processor_target_table): Add "goldmont-plus".
15097         (PTA_GOLDMONT_PLUS): Define.
15098         (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
15099         (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
15100         (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
15101         (fold_builtin_cpu): Add "goldmont-plus".
15102         (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
15103         (ix86_option_override_internal): Add "goldmont-plus".
15104         * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
15105         (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
15106         * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
15107         * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
15109 2018-05-17  Richard Biener  <rguenther@suse.de>
15111         PR tree-optimization/85757
15112         * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
15113         remove defs that only feed that PHI from further processing.
15115 2018-05-16  Jim Wilson  <jimw@sifive.com>
15117         * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
15118         asterisk to name.
15119         (<optab>di3_mask, <optab>di3_mask_1): Likewise.
15121 2018-05-16  Mark Wielaard  <mark@klomp.org>
15123         * dwarf2out.c (count_index_strings): New function.
15124         (output_indirect_strings): Call count_index_strings and generate
15125         header for dwarf_version >= 5.
15127 2018-05-16  Mark Wielaard  <mark@klomp.org>
15129         * dwarf2out.c (dwarf_FORM): New function.
15130         (set_indirect_string): Use dwarf_FORM.
15131         (reset_indirect_string): Likewise.
15132         (size_of_die): Likewise.
15133         (value_format): Likewise.
15134         (output_die): Likewise.
15135         (add_skeleton_AT_string): Likewise.
15136         (output_macinfo_op): Likewise.
15137         (index_string): Likewise.
15138         (output_index_string_offset): Likewise.
15139         (output_index_string): Likewise.
15140         (count_index_strings): Likewise.
15142 2018-05-16  Carl Love  <cel@us.ibm.com>
15144         * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
15145         dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
15147 2018-05-16  Martin Jambor  <mjambor@suse.cz>
15149         * ipa-prop.c (ipa_free_all_edge_args): Remove.
15150         * ipa-prop.h (ipa_free_all_edge_args): Likewise.
15152 2018-05-16  Wilco Dijkstra  <wdijkstr@arm.com>
15154         * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
15155         (fnma<mode>4): Likewise.
15156         (fms<mode>4): Likewise.
15157         (fnms<mode>4): Likewise.
15158         (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
15159         (aarch64_fnma<mode>4): Likewise.
15160         (aarch64_fms<mode>4): Likewise.
15161         (aarch64_fnms<mode>4): Likewise.
15162         (aarch64_fnmadd<mode>4): Likewise.
15164 2018-05-16  Jason Merrill  <jason@redhat.com>
15166         * tree.c (warn_deprecated_use): Return bool.  Simplify logic.
15168 2018-05-16  Richard Biener  <rguenther@suse.de>
15170         * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
15171         (dump_stmt_cost): Declare.
15172         (add_stmt_cost): Dump cost we add.
15173         (add_stmt_costs): New function.
15174         (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
15175         No longer exported.
15176         (vect_analyze_stmt): Adjust prototype.
15177         (vectorizable_condition): Likewise.
15178         (vectorizable_live_operation): Likewise.
15179         (vectorizable_reduction): Likewise.
15180         (vectorizable_induction): Likewise.
15181         * tree-vect-loop.c (vect_analyze_loop_operations): Create local
15182         cost vector to pass to vectorizable_ and record afterwards.
15183         (vect_model_reduction_cost): Take cost vector argument and adjust.
15184         (vect_model_induction_cost): Likewise.
15185         (vectorizable_reduction): Likewise.
15186         (vectorizable_induction): Likewise.
15187         (vectorizable_live_operation): Likewise.
15188         * tree-vect-slp.c (vect_create_new_slp_node): Initialize
15189         SLP_TREE_NUMBER_OF_VEC_STMTS.
15190         (vect_analyze_slp_cost_1): Remove.
15191         (vect_analyze_slp_cost): Likewise.
15192         (vect_slp_analyze_node_operations): Take visited args and
15193         a target cost vector.  Avoid processing already visited stmt sets.
15194         (vect_slp_analyze_operations): Use a local cost vector to gather
15195         costs and register those of non-discarded instances.
15196         (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
15197         (vect_schedule_slp_instance): Remove copying of
15198         SLP_TREE_NUMBER_OF_VEC_STMTS.  Instead assert that it is not
15199         zero.
15200         * tree-vect-stmts.c (record_stmt_cost): Remove path directly
15201         adding cost.  Record cost entry location.
15202         (vect_prologue_cost_for_slp_op): Function to compute cost of
15203         a constant or invariant generated for SLP vect in the prologue,
15204         split out from vect_analyze_slp_cost_1.
15205         (vect_model_simple_cost): Make static.  Adjust for SLP costing.
15206         (vect_model_promotion_demotion_cost): Likewise.
15207         (vect_model_store_cost): Likewise, make static.
15208         (vect_model_load_cost): Likewise.
15209         (vectorizable_bswap): Add cost vector arg and adjust.
15210         (vectorizable_call): Likewise.
15211         (vectorizable_simd_clone_call): Likewise.
15212         (vectorizable_conversion): Likewise.
15213         (vectorizable_assignment): Likewise.
15214         (vectorizable_shift): Likewise.
15215         (vectorizable_operation): Likewise.
15216         (vectorizable_store): Likewise.
15217         (vectorizable_load): Likewise.
15218         (vectorizable_condition): Likewise.
15219         (vectorizable_comparison): Likewise.
15220         (can_vectorize_live_stmts): Likewise.
15221         (vect_analyze_stmt): Likewise.
15222         (vect_transform_stmt): Adjust calls to vectorizable_*.
15223         * tree-vectorizer.c: Include gimple-pretty-print.h.
15224         (dump_stmt_cost): New function.
15226 2018-05-16  Richard Biener  <rguenther@suse.de>
15228         * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
15229         * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
15230         * tree-ssa-dse.c: Include tree-ssa-loop.h.
15231         (check_name): New callback.
15232         (dse_classify_store): Track cycles via a visited bitmap of PHI
15233         defs and simplify handling of in-loop and across loop dead stores
15234         and properly fail for loop-variant refs.  Handle byte-tracking with
15235         multiple defs.  Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
15236         limiting the walk.
15238 2018-05-16  Richard Sandiford  <richard.sandiford@linaro.org>
15240         * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
15241         (vect_get_mask_type_for_stmt): Likewise.
15242         * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
15243         split out from...
15244         (vect_build_slp_tree_1): ...here.  Use vect_get_vector_types_for_stmt
15245         to determine the statement's vector type and the vector type that
15246         should be used for calculating nunits.  Deal with cases in which
15247         the type has to be deferred.
15248         (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
15249         and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
15250         * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
15251         (vect_determine_vf_for_stmt): New functions, split out from...
15252         (vect_determine_vectorization_factor): ...here.
15253         * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
15254         (vect_get_mask_type_for_stmt): New functions, split out from
15255         vect_determine_vectorization_factor.
15257 2018-05-16  Richard Biener  <rguenther@suse.de>
15259         * tree-cfg.c (verify_gimple_assign_ternary): Properly
15260         verify the [VEC_]COND_EXPR embedded comparison.
15262 2018-05-15  Martin Sebor  <msebor@redhat.com>
15264         PR tree-optimization/85753
15265         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
15266         RECORD_TYPE in addition to ARRAY_TYPE.
15268 2018-05-15  Martin Sebor  <msebor@redhat.com>
15270         PR middle-end/85643
15271         * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
15273 2018-05-15  Richard Biener  <rguenther@suse.de>
15275         * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
15276         add by_clobber_p one.  Change algorithm to collect all defs
15277         representing uses we need to walk and try reducing them to
15278         a single one before failing.
15279         (dse_dom_walker::dse_optimize_stmt): Adjust.
15281 2018-05-13  Mark Wielaard  <mark@klomp.org>
15283         * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
15284         (size_of_loc_descr): Likewise.
15285         (output_loc_operands): Likewise.
15286         (output_loc_operands_raw): Likewise.
15287         (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
15288         (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
15289         (hash_loc_operands): Likewise.
15290         (compare_loc_operands): Likewise.
15292 2018-05-14  Mark Wielaard  <mark@klomp.org>
15294         * dwarf2out.c (count_index_addrs): New function.
15295         (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
15297 2018-05-15  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
15299         PR tree-optimization/83648
15300         * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
15301         return value as malloc candidate.
15303 2018-05-15  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
15305         PR ipa/85734
15306         * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite
15307         param as true in call to suggest_attribute.
15309 2018-05-14  Segher Boessenkool  <segher@kernel.crashing.org>
15311         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
15312         -mreadonly-in-sdata.
15314 2018-05-14  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
15316         * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
15317         New pattern.
15318         (aarch64_crypto_aesd_fused): Likewise.
15320 2018-05-14  Wilco Dijkstra  <wdijkstr@arm.com>
15322         * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
15323         (movsi_aarch64): Likewise.
15324         (load_pairsi): Likewise.
15325         (load_pairdi): Likewise.
15326         (store_pairsi): Likewise.
15327         (store_pairdi): Likewise.
15328         (load_pairsf): Likewise.
15329         (load_pairdf): Likewise.
15330         (store_pairsf): Likewise.
15331         (store_pairdf): Likewise.
15332         (zero_extend): Likewise.
15333         (trunc): Swap alternatives.
15334         (fcvt_target): Add '?' to prefer w over r.
15336 2018-05-14  Jakub Jelinek  <jakub@redhat.com>
15338         PR target/85756
15339         * config/i386/i386.md: Disallow non-commutative arithmetics in
15340         last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
15341         optimization.  Use COMMUTATIVE_ARITH_P test rather than != MINUS
15342         in the peephole2 before it.
15344 2018-05-14  Sebastian Peryt  <sebastian.peryt@intel.com>
15346         * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
15347         OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
15348         (ix86_handle_option): Handle -mcldemote.
15349         * config.gcc: New header.
15350         * config/i386/cldemoteintrin.h: New file.
15351         * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
15352         * config/i386/driver-i386.c (host_detect_local_cpu): Detect
15353         -mcldemote.
15354         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
15355         OPTION_MASK_ISA_CLDEMOTE.
15356         * config/i386/i386.c (ix86_target_string): Add -mcldemote.
15357         (ix86_valid_target_attribute_inner_p): Ditto.
15358         (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
15359         (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
15360         (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
15361         * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
15362         * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
15363         (cldemote): New.
15364         * config/i386/i386.opt: Add -mcldemote.
15365         * config/i386/x86intrin.h: New header.
15366         * doc/invoke.texi: Add -mcldemote.
15368 2018-05-14  Richard Biener  <rguenther@suse.de>
15370         * doc/match-and-simplify.texi: Adjust :s documentation.
15372 2018-05-14  Alexander Monakov  <amonakov@ispras.ru>
15374         * sort.cc (REORDER_23): Pass the type for the temporaries instead of
15375         intended memcpy size.
15376         (REORDER_45): Likewise.
15378 2018-05-13  Alexander Monakov  <amonakov@ispras.ru>
15380         * sort.cc: New file.
15381         * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
15382         * vec.c (qsort_chk): Use gcc_qsort.
15383         * Makefile.in (OBJS-libcommon): Add sort.o.
15384         (build/sort.o): New target.  Use it...
15385         (BUILD_RTL): ... here, and...
15386         (build/gencfn-macros): ... here, and...
15387         (build/genmatch): ... here.
15389 2018-05-13  Kito Cheng  <kito.cheng@gmail.com>
15390             Chung-Ju Wu  <jasonwucj@gmail.com>
15392         * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
15393         * config/nds32/nds32-graywolf.md: New file.
15394         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
15395         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
15396         pipeline.
15397         * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
15398         * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
15399         * config/nds32/nds32.md (pipeline_model): Add graywolf.
15400         * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
15401         * config/nds32/pipelines.md: Include n15 settings.
15403 2018-05-13  Kito Cheng  <kito.cheng@gmail.com>
15404             Chung-Ju Wu  <jasonwucj@gmail.com>
15406         * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
15407         * config/nds32/nds32-n13.md: New file.
15408         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
15409         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
15410         pipeline.
15411         * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
15412         * config/nds32/nds32.md (pipeline_model): Add n13.
15413         * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
15414         * config/nds32/pipelines.md: Include n13 settings.
15416 2018-05-13  Kito Cheng  <kito.cheng@gmail.com>
15417             Chung-Ju Wu  <jasonwucj@gmail.com>
15419         * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
15420         * config/nds32/nds32-n10.md: New file.
15421         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
15422         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
15423         pipeline.
15424         * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
15425         * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
15426         * config/nds32/nds32.md (pipeline_model): Add n10.
15427         * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
15428         * config/nds32/pipelines.md: Include n10 settings.
15430 2018-05-13  Monk Chiang  <sh.chiang04@gmail.com>
15431             Kito Cheng  <kito.cheng@gmail.com>
15432             Chung-Ju Wu  <jasonwucj@gmail.com>
15434         * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
15435         * config/nds32/constants.md (unspec_element, unspec_volatile_element):
15436         Add enum values for DSP extension instructions.
15437         * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
15438         New constraints.
15439         * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
15440         sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
15441         New code iterators.
15442         (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
15443         * config/nds32/nds32-dspext.md: New file for DSP implementation.
15444         * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
15445         * config/nds32/nds32-intrinsic.md: Likewise.
15446         * config/nds32/nds32_intrinsic.h: Likewise.
15447         * config/nds32/nds32-md-auxiliary.c: Likewise.
15448         * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
15449         * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
15450         (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
15451         (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
15452         * config/nds32/nds32-protos.h: New declarations for DSP extension.
15453         * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
15454         TYPE_DMAC in switch statement.
15455         * config/nds32/nds32.c: New checking and implementation for DSP
15456         extension instructions.
15457         * config/nds32/nds32.h: Likewise.
15458         * config/nds32/nds32.md: Likewise.
15459         * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
15460         * config/nds32/predicates.md: Implement new predicates for DSP
15461         extension.
15463 2018-05-11  Michael Meissner  <meissner@linux.vnet.ibm.com>
15465         * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
15466         Reformat alternatives and attributes so it is easier to identify
15467         which constraints/attributes go with which instruction.
15468         (mov<mode>_hardfloat32, FMOVE64): Likewise.
15469         (mov<mode>_softfloat32, FMOVE64): Likewise.
15470         (mov<mode>_hardfloat64, FMOVE64): Likewise.
15471         (mov<mode>_softfloat64, FMOVE64): Likewise.
15473 2018-05-11  Kelvin Nilsen  <kelvin@gcc.gnu.org>
15475         * doc/extend.texi (PowerPC Built-in Functions): Rename this
15476         subsection.
15477         (Basic PowerPC Built-in Functions): The new name of the
15478         subsection previously known as "PowerPC Built-in Functions".
15479         (Basic PowerPC Built-in Functions Available on all Configurations):
15480         New subsubsection.
15481         (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
15482         (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
15483         (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
15484         (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
15486 2018-05-11  Martin Jambor  <mjambor@suse.cz>
15488         PR ipa/85655
15489         * ipa-cp.c (intersect_with_plats): Check that the lattice contains
15490         single const.
15492 2018-05-11  Richard Earnshaw  <rearnsha@arm.com>
15494         PR target/85733
15495         * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
15497 2018-05-11  Sebastian Peryt  <sebastian.peryt@intel.com>
15499         * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
15500         OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
15501         (ix86_handle_option): Handle -mwaitpkg.
15502         * config.gcc: New header.
15503         * config/i386/cpuid.h (bit_WAITPKG): New bit.
15504         * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
15505         * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
15506         function type.
15507         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
15508         OPTION_MASK_ISA_WAITPKG.
15509         * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
15510         (ix86_option_override_internal): Add PTA_WAITPKG.
15511         (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
15512         (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
15513         IX86_BUILTIN_TPAUSE.
15514         (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
15515         __builtin_ia32_umwait and __builtin_ia32_tpause.
15516         (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
15517         IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
15518         * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
15519         * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
15520         UNSPECV_TPAUSE): New.
15521         (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
15522         * config/i386/i386.opt: Add -mwaitpkg.
15523         * config/i386/waitpkgintrin.h: New file.
15524         * config/i386/x86intrin.h: New header.
15525         * doc/invoke.texi: Add -mwaitpkg.
15527 2018-05-11  Richard Earnshaw  <rearnsha@arm.com>
15529         PR target/85606
15530         * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
15531         equivalent.
15532         (cortex-m0): Use armv6s-m isa.
15533         (cortex-m0plus): Likewise.
15534         (cortex-m1): Likewise.
15535         (cortex-m0.small-multiply): Likewise.
15536         (cortex-m0plus.small-multiply): Likewise.
15537         (cortex-m1.small-multiply): Likewise.
15539 2018-05-11  Allan Sandfeld Jensen  <allan.jensen@qt.io>
15540             Jakub Jelinek  <jakub@redhat.com>
15542         PR tree-optimization/85692
15543         * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
15544         source permute as well.
15546 2018-05-11  Martin Liska  <mliska@suse.cz>
15548         PR sanitizer/85556
15549         * doc/extend.texi: Document LLVM style format for no_sanitize
15550         attribute.
15552 2018-05-10  Michael Meissner  <meissner@linux.vnet.ibm.com>
15554         * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
15555         mode_supports_vsx_dform_quad to mode_supports_dq_form.
15556         (mode_supports_vsx_dform_quad): Likewise.
15557         (mode_supports_vmx_dform): Move these functions to be next to the
15558         other mode_supports functions.
15559         (mode_supports_dq_form): Likewise.
15560         (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
15561         mode_supports_dq_form.
15562         (reg_offset_addressing_ok_p): Likewise.
15563         (offsettable_ok_by_alignment): Likewise.
15564         (rs6000_legitimate_offset_address_p): Likewise.
15565         (legitimate_lo_sum_address_p): Likewise.
15566         (rs6000_legitimize_address): Likewise.
15567         (rs6000_legitimize_reload_address): Likewise.
15568         (rs6000_secondary_reload_inner): Likewise.
15569         (rs6000_preferred_reload_class): Likewise.
15570         (rs6000_output_move_128bit): Likewise.
15572 2018-05-10  Uros Bizjak  <ubizjak@gmail.com>
15574         * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
15575         Generate SImode target register for null target.
15576         <case IX86_BUILTIN_XGETBV>: Ditto.
15577         <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
15578         * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
15580 2018-05-10  Carl Love  <cel@us.ibm.com>
15582         * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
15583         dcbtt and dcbtstt if operands[2] is 0.
15585 2018-05-10  Uros Bizjak  <ubizjak@gmail.com>
15587         PR target/85693
15588         * config/i386/sse.md (usadv64qi): New expander.
15590 2018-05-10  Segher Boessenkool  <segher@kernel.crashing.org>
15592         * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
15593         altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
15594         -maltivec=be support.
15595         (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
15596         vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
15597         vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
15598         vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
15599         vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
15600         vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
15601         altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
15602         altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
15603         altivec_vsumsws): Adjust.
15604         (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
15605         *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
15606         altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
15607         support.
15608         (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
15609         altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
15610         altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
15611         (altivec_lve<VI_char>x): Delete expand.
15612         (*altivec_lve<VI_char>x_internal): Rename to...
15613         (altivec_lve<VI_char>x): ... this.
15614         (altivec_lvxl_<mode>): Delete expand.
15615         (*altivec_lvxl_<mode>_internal): Rename to ...
15616         (altivec_lvxl_<mode>): ... this.
15617         (altivec_stvxl_<mode>): Delete expand.
15618         (*altivec_stvxl_<mode>_internal): Rename to ...
15619         (altivec_stvxl_<mode>): ... this.
15620         (altivec_stve<VI_char>x): Delete expand.
15621         (*altivec_stve<VI_char>x_internal): Rename to ...
15622         (altivec_stve<VI_char>x): ... this.
15623         (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
15624         doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
15625         reduc_plus_scal_<mode>): Adjust.
15626         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
15627         comment.
15628         (rs6000_cpu_cpp_builtins): Adjust.
15629         (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
15630         * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
15631         altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
15632         * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
15633         -maltivec=be support.
15634         (rs6000_split_vec_extract_var): Adjust.
15635         (rs6000_split_v4si_init): Adjust.
15636         (swap_selector_for_mode): Delete.
15637         (altivec_expand_lvx_be, altivec_expand_stvx_be,
15638         altivec_expand_stvex_be): Delete.
15639         (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
15640         -maltivec=be support.
15641         (rs6000_gimple_fold_builtin): Ditto.
15642         (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
15643         Adjust.
15644         * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
15645         (TARGET_DIRECT_MOVE_64BIT): Adjust.
15646         * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
15647         * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
15648         * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
15649         unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
15650         vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
15651         *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
15652         *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
15653         *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
15654         anonymous split): Adjust.
15655         (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
15656         (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
15658 2018-05-10  Eric Botcazou  <ebotcazou@adacore.com>
15660         * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
15661         when --with-gxx-include-dir is also specified.
15662         * configure: Regenerate.
15664 2018-05-09  Jim Wilson  <jimw@sifive.com>
15666         PR target/84797
15667         * config.gcc (riscv*-*-*): Handle --with-multilib-list.
15668         * config/riscv/t-withmultilib: New.
15669         * config/riscv/withmultilib.h: New.
15670         * doc/install.texi: Document RISC-V --with-multilib-list support.
15672 2018-05-09  Richard Biener  <rguenther@suse.de>
15674         * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
15675         vector.
15676         (vect_bb_vectorization_profitable_p): Adjust.  Compute
15677         actual scalar cost using the cost vector and the add_stmt_cost
15678         machinery.
15680 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
15682         PR rtl-optimization/85645
15683         * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
15684         in the REG_CFA_REGISTER note for LR, don't leave it empty.
15686 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
15688         PR rtl-optimization/85645
15689         * shrink-wrap.c (spread_components): Return a boolean saying if
15690         anything was changed.
15691         (try_shrink_wrapping_separate): Iterate spread_components until
15692         nothing changes anymore.
15694 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
15696         PR rtl-optimization/85645
15697         * regrename.c (build_def_use): Also kill the chains that include the
15698         destination of a REG_CFA_REGISTER note.
15700 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
15702         PR rtl-optimization/85645
15703         *  regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
15704         insn that has a REG_CFA_REGISTER note.
15706 2018-05-09  Richard Sandiford  <richard.sandiford@linaro.org>
15708         * cfgexpand.c (expand_clobber): New function.
15709         (expand_gimple_stmt_1): Use it.
15710         * tree-vect-stmts.c (vect_clobber_variable): New function,
15711         split out from...
15712         (vectorizable_simd_clone_call): ...here.
15713         (vectorizable_store): Emit a clobber either side of an
15714         IFN_STORE_LANES sequence.
15715         (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
15717 2018-05-09  Tom de Vries  <tom@codesourcery.com>
15719         PR target/85626
15720         * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
15721         (define_insn "trap_if_false"): Add exit after trap.
15723 2018-05-09  Eric Botcazou  <ebotcazou@adacore.com>
15725         PR rtl-optimization/85638
15726         * bb-reorder.c: Include common/common-target.h.
15727         (create_forwarder_block): New function extracted from...
15728         (fix_up_crossing_landing_pad): ...here.  Rename into...
15729         (dw2_fix_up_crossing_landing_pad): ...this.
15730         (sjlj_fix_up_crossing_landing_pad): New function.
15731         (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
15732         call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
15733         from both partitions and exit the loop after one iteration.
15735 2018-05-08  Kelvin Nilsen  <kelvin@gcc.gnu.org>
15737         Revert:
15738         * doc/extend.texi (PowerPC Built-in Functions): Rename this
15739         subsection.
15740         (Basic PowerPC Built-in Functions): The new name of the
15741         subsection previously known as "PowerPC Built-in Functions".
15742         (Basic PowerPC Built-in Functions Available on all Configurations):
15743         New subsubsection.
15744         (Basic PowerPC Built-in Functions Available on ISA 2.05): New
15745         subsubsection.
15746         (Basic PowerPC Built-in Functions Available on ISA 2.06): New
15747         subsubsection.
15748         (Basic PowerPC Built-in Functions Available on ISA 2.07): New
15749         subsubsection.
15750         (Basic PowerPC Built-in Functions Available on ISA 3.0): New
15751         subsubsection.
15753 2018-05-08  Jim Wilson  <jimw@sifive.com>
15755         * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
15756         (LD_EMUL_SUFFIX): New.
15757         (LINK_SPEC): Use it.
15759 2018-05-08  Kelvin Nilsen  <kelvin@gcc.gnu.org>
15761         * doc/extend.texi (PowerPC Built-in Functions): Rename this
15762         subsection.
15763         (Basic PowerPC Built-in Functions): The new name of the
15764         subsection previously known as "PowerPC Built-in Functions".
15765         (Basic PowerPC Built-in Functions Available on all Configurations):
15766         New subsubsection.
15767         (Basic PowerPC Built-in Functions Available on ISA 2.05): New
15768         subsubsection.
15769         (Basic PowerPC Built-in Functions Available on ISA 2.06): New
15770         subsubsection.
15771         (Basic PowerPC Built-in Functions Available on ISA 2.07): New
15772         subsubsection.
15773         (Basic PowerPC Built-in Functions Available on ISA 3.0): New
15774         subsubsection.
15776 2018-05-08  Jakub Jelinek  <jakub@redhat.com>
15778         PR target/85683
15779         * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
15780         after cmpelim optimization.
15782 2018-05-08  Olga Makhotina  <olga.makhotina@intel.com>
15784         * config.gcc: Support "goldmont".
15785         * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
15786         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
15787         PROCESSOR_GOLDMONT.
15788         * config/i386/i386.c (m_GOLDMONT): Define.
15789         (processor_target_table): Add "goldmont".
15790         (PTA_GOLDMONT): Define.
15791         (ix86_lea_outperforms): Add TARGET_GOLDMONT.
15792         (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
15793         (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
15794         (fold_builtin_cpu): Add "goldmont".
15795         (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
15796         (ix86_option_override_internal): Add "goldmont".
15797         * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
15798         (processor_type): Add PROCESSOR_GOLDMONT.
15799         * config/i386/i386.md: Add CPU "glm".
15800         * config/i386/glm.md: New file.
15801         * config/i386/x86-tune.def: Add m_GOLDMONT.
15802         * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
15804 2018-05-08  Jakub Jelinek  <jakub@redhat.com>
15806         PR target/85572
15807         * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
15808         E_V4DImode.
15809         * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
15810         VI1248_AVX512VL_AVX512BW.  Handle V2DImode and V4DImode if not
15811         TARGET_AVX512VL using ix86_expand_sse2_abs.  Formatting fixes.
15813         PR target/85317
15814         * config/i386/i386.c (ix86_fold_builtin): Handle
15815         IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
15817         PR target/85480
15818         * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
15819         (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
15821 2018-05-08  Richard Earnshaw  <rearnsha@arm.com>
15823         PR target/85658
15824         * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
15825         (check_arch): Likewise.
15826         (check_fpu): Return the result rather than printing it.
15827         (end arch): Fix operator precedence.
15828         (end cpu): Likewise.
15829         (END): Print the result from check_fpu.
15831 2018-05-08  Richard Sandiford  <richard.sandiford@linaro.org>
15832             Alan Hayward  <alan.hayward@arm.com>
15833             David Sherwood  <david.sherwood@arm.com>
15835         * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
15836         (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
15837         (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
15838         (*fcmuo<mode>_and): New patterns.
15840 2018-05-08  Richard Sandiford  <richard.sandiford@linaro.org>
15842         * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
15843         (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
15844         (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
15845         (cmp_op, sve_imm_con): New code attributes.
15846         (SVE_COND_INT_CMP, imm_con): Delete.
15847         (cmp_op): Remove above unspecs from int attribute.
15848         * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
15849         to...
15850         (*cmp<cmp_op><mode>): ...this.  Use UNSPEC_MERGE_PTRUE instead of
15851         comparison-specific unspecs.
15852         (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
15853         (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
15854         (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
15855         (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
15856         (*vec_fcm<cmp_op><mode>): Rename to...
15857         (*fcm<cmp_op><mode>): ...this and adjust likewise.
15858         (*vec_fcmuo<mode>): Rename to...
15859         (*fcmuo<mode>): ...this and adjust likewise.
15860         (*pred_fcm<cmp_op><mode>): New pattern.
15861         * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
15862         (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
15863         functions.
15864         (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
15865         and UNORDERED.
15866         (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
15867         (aarch64_emit_sve_predicated_cond): New function.
15868         (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
15869         (aarch64_emit_unspec_cond_or): Replace with...
15870         (aarch64_emit_sve_or_conds): ...this new function.  Use
15871         aarch64_emit_sve_ptrue_op for the individual comparisons and
15872         aarch64_emit_binop to OR them together.
15873         (aarch64_emit_inverted_unspec_cond): Replace with...
15874         (aarch64_emit_sve_inverted_cond): ...this new function.  Use
15875         aarch64_emit_sve_ptrue_op for the comparison and
15876         aarch64_emit_unop to invert the result.
15877         (aarch64_expand_sve_vec_cmp_float): Update after the above
15878         changes.  Use aarch64_emit_sve_ptrue_op for native comparisons.
15880 2018-05-07  Nathan Sidwell  <nathan@acm.org>
15882         * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
15883         * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
15884         (Backwards Compatibility): Likewise.
15886 2018-05-07  Luis Machado  <luis.machado@linaro.org>
15888         PR bootstrap/85681
15889         Revert:
15890         2018-05-07  Luis Machado  <luis.machado@linaro.org>
15892         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
15893         <prefetch_dynamic_strides>: New const bool field.
15894         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
15895         prefetch_dynamic_strides.
15896         (exynosm1_prefetch_tune): Likewise.
15897         (thunderxt88_prefetch_tune): Likewise.
15898         (thunderx_prefetch_tune): Likewise.
15899         (thunderx2t99_prefetch_tune): Likewise.
15900         (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
15901         to false.
15902         (aarch64_override_options_internal): Update to set
15903         PARAM_PREFETCH_DYNAMIC_STRIDES.
15904         * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
15905         * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
15906         * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
15907         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
15908         prefetch-dynamic-strides setting.
15910         2018-05-07  Luis Machado  <luis.machado@linaro.org>
15912         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
15913         <minimum_stride>: New const int field.
15914         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
15915         minimum_stride field.
15916         (exynosm1_prefetch_tune): Likewise.
15917         (thunderxt88_prefetch_tune): Likewise.
15918         (thunderx_prefetch_tune): Likewise.
15919         (thunderx2t99_prefetch_tune): Likewise.
15920         (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
15921         (aarch64_override_options_internal): Update to set
15922         PARAM_PREFETCH_MINIMUM_STRIDE.
15923         * doc/invoke.texi (prefetch-minimum-stride): Document new option.
15924         * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
15925         * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
15926         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
15927         stride is constant and is below the minimum stride threshold.
15929 2018-05-07  Luis Machado  <luis.machado@linaro.org>
15931         * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
15932         to 512.
15934 2018-05-07  Luis Machado  <luis.machado@linaro.org>
15936         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
15937         <prefetch_dynamic_strides>: New const bool field.
15938         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
15939         prefetch_dynamic_strides.
15940         (exynosm1_prefetch_tune): Likewise.
15941         (thunderxt88_prefetch_tune): Likewise.
15942         (thunderx_prefetch_tune): Likewise.
15943         (thunderx2t99_prefetch_tune): Likewise.
15944         (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
15945         to false.
15946         (aarch64_override_options_internal): Update to set
15947         PARAM_PREFETCH_DYNAMIC_STRIDES.
15948         * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
15949         * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
15950         * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
15951         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
15952         prefetch-dynamic-strides setting.
15954 2018-05-07  Luis Machado  <luis.machado@linaro.org>
15956         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
15957         <minimum_stride>: New const int field.
15958         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
15959         minimum_stride field.
15960         (exynosm1_prefetch_tune): Likewise.
15961         (thunderxt88_prefetch_tune): Likewise.
15962         (thunderx_prefetch_tune): Likewise.
15963         (thunderx2t99_prefetch_tune): Likewise.
15964         (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
15965         (aarch64_override_options_internal): Update to set
15966         PARAM_PREFETCH_MINIMUM_STRIDE.
15967         * doc/invoke.texi (prefetch-minimum-stride): Document new option.
15968         * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
15969         * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
15970         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
15971         stride is constant and is below the minimum stride threshold.
15973 2018-05-06  Jakub Jelinek  <jakub@redhat.com>
15975         PR c++/85659
15976         * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
15977         the type is addressable.  Don't force op into register if it has
15978         BLKmode.
15980 2018-05-05  Roland McGrath  <mcgrathr@google.com>
15982         PR other/77609
15983         * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
15984         any section for which we don't know a specific type it should have,
15985         regardless of name.  Previously this was done only for the exact
15986         names ".init_array", ".fini_array", and ".preinit_array".
15987         (default_elf_asm_named_section): Add comment about
15988         relationship with default_section_type_flags and SECTION_NOTYPE.
15989         (get_section): Don't consider it a type conflict if one side has
15990         SECTION_NOTYPE and the other doesn't, as long as neither has the
15991         SECTION_BSS et al used in the default_section_type_flags logic.
15993 2018-05-05  Tom de Vries  <tom@codesourcery.com>
15995         PR target/85653
15996         * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
15997         (workaround_barsyncs): New function.
15998         (nvptx_reorg): Use workaround_barsyncs.
15999         * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
16000         (define_expand "nvptx_membar_cta"): New define_expand.
16001         (define_insn "*nvptx_membar_cta"): New insn.
16003 2018-05-04  Pekka Jääskeläinen  <pekka.jaaskelainen@parmance.com>
16005         * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
16006         To improve optimization opportunities.
16007         * builtin-types.def: The new needed builtin types for the above.
16009 2018-05-04  Richard Biener  <rguenther@suse.de>
16011         * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
16012         * gimple-ssa-store-merging.c
16013         (imm_store_chain_info::output_merged_store): Remove redundant create,
16014         release split_store vector contents on failure.
16015         * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
16016         scalar stmt vector on cache hit.
16018 2018-05-04  Segher Boessenkool  <segher@kernel.crashing.org>
16020         * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
16021         Xilinx FP support.
16022         * config.gcc (powerpc-xilinx-eabi*): Remove.
16023         * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
16024         support.
16025         (fusion_addis_mem_combo_load): Ditto.
16026         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
16027         FP support.
16028         (rs6000_cpu_cpp_builtins): Ditto.
16029         * config/rs6000/rs6000-linux.c
16030         (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
16031         * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
16032         * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
16033         support.
16034         (rs6000_setup_reg_addr_masks): Ditto.
16035         (rs6000_init_hard_regno_mode_ok): Ditto.
16036         (rs6000_option_override_internal): Ditto.
16037         (legitimate_lo_sum_address_p): Ditto.
16038         (rs6000_legitimize_address): Ditto.
16039         (rs6000_legitimize_reload_address): Ditto.
16040         (rs6000_legitimate_address_p): Ditto.
16041         (abi_v4_pass_in_fpr): Ditto.
16042         (setup_incoming_varargs): Ditto.
16043         (rs6000_gimplify_va_arg): Ditto.
16044         (rs6000_split_multireg_move): Ditto.
16045         (rs6000_savres_strategy): Ditto.
16046         (rs6000_emit_prologue_components): Ditto.
16047         (rs6000_emit_epilogue_components): Ditto.
16048         (rs6000_emit_prologue): Ditto.
16049         (rs6000_emit_epilogue): Ditto.
16050         (rs6000_elf_file_end): Ditto.
16051         (rs6000_function_value): Ditto.
16052         (rs6000_libcall_value): Ditto.
16053         * config/rs6000/rs6000.h: Ditto.
16054         (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
16055         (TARGET_MINMAX): ... this.  New.
16056         (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
16057         * config/rs6000/rs6000.md: Remove Xilinx FP support.
16058         (*movsi_internal1_single): Delete.
16059         * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
16060         mfpu=, mxilinx-fpu): Delete.
16061         * config/rs6000/singlefp.h: Delete.
16062         * config/rs6000/sysv4.h: Remove Xilinx FP support.
16063         * config/rs6000/t-rs6000: Ditto.
16064         * config/rs6000/t-xilinx: Delete.
16065         * config/rs6000/titan.md: Adjust for fp_type removal.
16066         * config/rs6000/vsx.md: Remove Xilinx FP support.
16067         (VStype_simple): Delete.
16068         (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
16069         * config/rs6000/xfpu.h: Delete.
16070         * config/rs6000/xfpu.md: Delete.
16071         * config/rs6000/xilinx.h: Delete.
16072         * config/rs6000/xilinx.opt: Delete.
16073         * doc/invoke.texi (RS/6000 and PowerPC Options): Remove
16074         -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
16076 2018-05-04  Tom de Vries  <tom@codesourcery.com>
16078         PR libgomp/85639
16079         * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
16080         if ignore == 0.
16082 2018-05-04  Richard Biener  <rguenther@suse.de>
16084         PR middle-end/85627
16085         * tree-complex.c (update_complex_assignment): We are always in SSA form.
16086         (expand_complex_div_wide): Likewise.
16087         (expand_complex_operations_1): Likewise.
16088         (expand_complex_libcall): Preserve EH info of the original stmt.
16089         (tree_lower_complex): Handle removed blocks.
16090         * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
16091         on complex multiplication and division libcall builtins.
16093 2018-05-04  Richard Biener  <rguenther@suse.de>
16095         PR middle-end/85574
16096         * fold-const.c (negate_expr_p): Restrict negation of operand
16097         zero of a division to when we know that can happen without
16098         overflow.
16099         (fold_negate_expr_1): Likewise.
16101 2018-05-04  Jakub Jelinek  <jakub@redhat.com>
16103         PR libstdc++/85466
16104         * real.h (real_nextafter): Declare.
16105         * real.c (real_nextafter): New function.
16106         * fold-const-call.c (fold_const_nextafter): New function.
16107         (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
16108         CASE_CFN_NEXTTOWARD.
16109         (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
16110         even when arg1_mode is different from arg0_mode.
16112 2018-05-03  Nathan Sidwell  <nathan@acm.org>
16114         * doc/extend.texi (Deprecated Features): Remove
16115         -ffriend-injection.
16116         (Backwards Compatibility): Likewise.
16117         * doc/invoke.texi (C++ Language Options): Likewise.
16118         (C++ Dialect Options): Likewise.
16120 2018-05-03  Jakub Jelinek  <jakub@redhat.com>
16122         PR target/85530
16123         * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
16124         _mm512_mask_mullox_epi64): New intrinsics.
16126 2018-05-03  Tom de Vries  <tom@codesourcery.com>
16128         PR testsuite/85106
16129         * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
16130         dump files): Add offload-tree.
16132 2018-05-03  Richard Biener  <rguenther@suse.de>
16134         PR tree-optimization/85615
16135         * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
16136         to loops not nested in BBs loop father to avoid creating multi-entry
16137         loops.
16139 2018-05-03  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
16141         PR tree-optimization/70291
16142         * tree-complex.c (expand_complex_libcall): Add type, inplace_p
16143         arguments.  Change return type to tree.  Emit libcall as a new
16144         statement rather than replacing existing one when inplace_p is true.
16145         (expand_complex_multiplication_components): New function.
16146         (expand_complex_multiplication): Expand floating-point complex
16147         multiplication using the above.
16148         (expand_complex_division): Rename inner_type parameter to type.
16149         Update expand_complex_libcall call-site.
16150         (expand_complex_operations_1): Update expand_complex_multiplication
16151         and expand_complex_division call-sites.
16153 2018-05-02  Jakub Jelinek  <jakub@redhat.com>
16155         PR target/85582
16156         * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
16157         *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
16158         *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
16159         the highest significant bit of the shift count mask is clear.  In
16160         check whether and[sq]i3 is needed verify that all significant bits
16161         of the shift count other than the highest are set.
16163 2018-05-02  Tom de Vries  <tom@codesourcery.com>
16165         PR libgomp/82428
16166         * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
16167         * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
16168         (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
16169         * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
16170         (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
16171         * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
16172         __builtin_goacc_parlevel_size.
16174 2018-05-02  Richard Biener  <rguenther@suse.de>
16176         PR tree-optimization/85597
16177         * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
16178         do not use split vect_get_vec_defs call but call vect_get_slp_defs
16179         directly.
16181 2018-05-02  Tom de Vries  <tom@codesourcery.com>
16183         PR testsuite/85106
16184         * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
16185         dump files): Add ltrans-tree.
16187 2018-05-02  Tom de Vries  <tom@codesourcery.com>
16189         PR testsuite/85106
16190         * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
16191         dump files): Add wpa-ipa.
16193 2018-05-02  Segher Boessenkool  <segher@kernel.crashing.org>
16195         * config.gcc (powerpc*-*-*): Remove paired.h.  Unsupport the
16196         powerpc*-*-linux*paired* target.
16197         * config/rs6000/750cl.h: Delete.
16198         * config/rs6000/paired.h: Delete.
16199         * config/rs6000/paired.md: Delete.
16200         * config/rs6000/predicates.md (easy_vector_constant): Remove paired
16201         float support.
16202         * config/rs6000/rs6000-builtin.def: Remove paired float support.
16203         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
16204         comment.  Remove paired float support.
16205         * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
16206         * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
16207         VECTOR_PAIRED.
16208         * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
16209         paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
16210         declarations.
16211         * config/rs6000/rs6000.c: Remove paired float support.
16212         (paired_expand_vector_init, paired_expand_vector_move,
16213         paired_emit_vector_compare, paired_emit_vector_cond_expr,
16214         (paired_expand_lv_builtin, paired_expand_stv_builtin,
16215         paired_expand_builtin, paired_expand_predicate_builtin,
16216         paired_init_builtins): Delete.
16217         * config/rs6000/rs6000.h: Remove paired float support.
16218         * config/rs6000/rs6000.md: Remove paired float support.
16219         (move_from_CR_ov_bit): Delete.
16220         * config/rs6000/rs6000.opt (mpaired): Delete.
16221         * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
16222         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
16224 2018-05-02  Richard Biener  <rguenther@suse.de>
16226         PR middle-end/85567
16227         * gimplify.c (gimplify_save_expr): When in SSA form allow
16228         SAVE_EXPRs to compute to SSA vars.
16230 2018-05-02  Jakub Jelinek  <jakub@redhat.com>
16232         PR target/85582
16233         * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
16234         *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
16235         *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
16236         clobber operands[2], instead use a new pseudo.  Formatting fixes.
16238 2018-05-02  Richard Sandiford  <richard.sandiford@linaro.org>
16240         PR tree-optimization/85586
16241         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
16242         exit early for statements in the same group if the accesses are
16243         not strided.
16245 2018-05-02  Tom de Vries  <tom@codesourcery.com>
16247         PR lto/85451
16248         * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
16249         error message.
16251 2018-05-01  Marc Glisse  <marc.glisse@inria.fr>
16253         PR tree-optimization/85143
16254         * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
16256 2018-05-01  Tom de Vries  <tom@codesourcery.com>
16258         PR lto/85451
16259         * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
16260         not found" error message.
16262 2018-05-01  Tom de Vries  <tom@codesourcery.com>
16264         PR other/83786
16265         * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
16266         * vec.c (test_ordered_remove_if): New function.
16267         (vec_c_tests): Call test_ordered_remove_if.
16268         * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
16269         * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
16270         * tree-vect-patterns.c (vect_pattern_recog_1): Use
16271         VEC_ORDERED_REMOVE_IF.
16273 2018-05-01  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
16275         PR tree-optimization/82665
16276         * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
16277         pointer subtraction where arguments come from a memchr call.
16279 2018-05-01  Jakub Jelinek  <jakub@redhat.com>
16281         * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
16282         --push-state --as-needed and --pop-state instead of --as-needed and
16283         --no-as-needed if ld supports it.
16284         * configure: Regenerated.
16286         PR web/85578
16287         * doc/install.texi2html: Replace _002d with - and _002a with * in
16288         generated html files using sed.
16290 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
16292         PR c++/85523
16293         * gcc-rich-location.c (blank_line_before_p): New function.
16294         (use_new_line): New function.
16295         (gcc_rich_location::add_fixit_insert_formatted): New function.
16296         * gcc-rich-location.h
16297         (gcc_rich_location::add_fixit_insert_formatted): New function.
16299 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
16301         * selftest.c (assert_streq): Rename "expected" and "actual" to
16302         "val1" and "val2".  Extend NULL-handling to cover both inputs
16303         symmetrically, while still requiring both to be non-NULL for a pass.
16304         * selftest.h (assert_streq): Rename "expected" and "actual" to
16305         "val1" and "val2".
16306         (ASSERT_EQ): Likewise.
16307         (ASSERT_EQ_AT): Likewise.
16308         (ASSERT_KNOWN_EQ): Likewise.
16309         (ASSERT_KNOWN_EQ_AT): Likewise.
16310         (ASSERT_NE): Likewise.
16311         (ASSERT_MAYBE_NE): Likewise.
16312         (ASSERT_MAYBE_NE_AT): Likewise.
16313         (ASSERT_STREQ): Likewise.  Clarify that both must be non-NULL for
16314         the assertion to pass.
16315         (ASSERT_STREQ_AT): Likewise.
16317 2018-04-30  Jonathan Wakely  <jwakely@redhat.com>
16319         * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
16320         interaction with -pie.
16322 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
16324         * selftest.h: Fix alphabetization of per-source-file selftest
16325         declarations.
16327 2018-04-30  Jason Merrill  <jason@redhat.com>
16329         PR c++/61982 - dead stores to destroyed objects.
16330         * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
16331         of clobber.
16333 2018-04-30  Jason Merrill  <jason@redhat.com>
16335         * tree.c (build_clobber): New.
16336         * tree.h: Declare it.
16337         * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
16339 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
16341         * diagnostic-show-locus.c (layout::layout): Update for
16342         location_get_source_line returning a char_span.
16343         (struct char_span): Move to input.h.
16344         (struct correction): Update for fields in char_span becoming
16345         private.
16346         (struct source_line): Update for location_get_source_line
16347         returning a char_span.
16348         (layout::print_line): Likewise.
16349         * edit-context.c (edited_file::print_content): Likewise.
16350         (edited_file::print_diff_hunk): Likewise.
16351         (edited_file::print_run_of_changed_lines): Likewise.
16352         (edited_file::get_num_lines): Likewise.
16353         (edited_line::edited_line): Likewise.
16354         * final.c (asm_show_source): Likewise.
16355         * input.c (location_get_source_line): Convert return type
16356         from const char * to char_span, losing the final "line_len"
16357         param.
16358         (dump_location_info): Update for the above.
16359         (get_substring_ranges_for_loc): Likewise.  Use a char_span
16360         when handling the literal within the line.
16361         (test_reading_source_line): Update for location_get_source_line
16362         returning a char_span.
16363         * input.h (class char_span): Move here from
16364         diagnostic-show-locus.c, converting from a struct to a class.
16365         Make data members private.
16366         (char_span::operator bool): New.
16367         (char_span::length): New.
16368         (char_span::get_buffer): New.
16369         (char_span::operator[]): New.
16370         (char_span::subspan): Make const.
16371         (char_span::xstrdup): New.
16372         (location_get_source_line): Convert return type from const char *
16373         to char_span, losing the final "line_size" param.
16375 2018-04-30  Jan Hubicka  <jh@suse.cz>
16377         * lto-wrapper.c (ltrans_priorities): New static var.
16378         (cmp_priority): New.
16379         (run_gcc): Read priorities and if doing parallel build order
16380         the Makefile by them.
16382 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
16384         * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
16386 2018-04-30  Richard Biener  <rguenther@suse.de>
16388         * tree-cfg.c (verify_address): Remove base argument, add
16389         flag whether to check TREE_ADDRESSABLE and do that.
16390         (verify_expr): Remove.
16391         (verify_types_in_gimple_reference): Add pieces from verify_expr.
16392         (verify_gimple_assign_single): Likewise.
16393         (verify_gimple_switch): Likewise.
16394         (verify_expr_location_1): Dereference tp once.  Add (disabled)
16395         piece from verify_expr.
16396         (verify_gimple_in_cfg): Do not call verify_expr on all ops.
16398 2018-04-30  Claudiu Zissulescu  <claziss@synopsys.com>
16400         * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
16402 2018-04-30  Claudiu Zissulescu  <claziss@synopsys.com>
16404         * config/arc/arc-protos.h (prepare_extend_operands): Remove.
16405         (small_data_pattern): Likewise.
16406         (arc_rewrite_small_data): Likewise.
16407         * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
16408         (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
16409         (get_symbol_alignment): New function.
16410         (legitimate_small_data_address_p): Likewise.
16411         (legitimate_scaled_address): Update, call
16412         legitimate_small_data_address_p.
16413         (output_sdata): New static variable.
16414         (arc_print_operand): Update how we handle small data operands.
16415         (arc_print_operand_address): Likewise.
16416         (arc_legitimate_address_p): Update, use
16417         legitimate_small_data_address_p.
16418         (arc_rewrite_small_data_p): Remove.
16419         (arc_rewrite_small_data_1): Likewise.
16420         (arc_rewrite_small_data): Likewise.
16421         (small_data_pattern): Likewise.
16422         (compact_sda_memory_operand): Update to use
16423         legitimate_small_data_address_p and get_symbol_alignment.
16424         (prepare_move_operands): Don't rewite sdata pattern.
16425         (prepare_extend_operands): Remove.
16426         * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
16427         pattern.
16428         (zero_extendqisi2): Likewise.
16429         (zero_extendhisi2): Likewise.
16430         (extendqihi2): Likewise.
16431         (extendqisi2): Likewise.
16432         (extendhisi2): Likewise.
16433         (addsi3): Likewise.
16434         (subsi3): Likewise.
16435         (andsi3): Likewise.
16436         * config/arc/constraints.md (Usd): Change it to memory constraint.
16438 2018-04-30  Claudiu Zissulescu  <claziss@synopsys.com>
16440         * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
16441         as source of std instructions.
16442         * config/arc/arc.md (movsi_insn): Update pattern predicate to
16443         allow 6-bit constants as source for store instructions.
16444         (movdi_insn): Update instruction pattern to allow 6-bit constants
16445         as source for store instructions.
16447 2018-04-30  Jonathan Wakely  <jwakely@redhat.com>
16449         * doc/invoke.texi (-fdebug-types-section): Fix grammar.
16451 2018-04-30  Nathan Sidwell  <nathan@acm.org>
16452             Sandra Loosemore <sandra@codesourcery.com>
16454         * dumpfile.c (dump_open): Allow '-' for stdout.
16455         * doc/invoke.texi (Developer Options): Document dump filename
16456         determination early.  Document stdin/stdout selection.
16458 2018-04-30  Andrew Sadek  <andrew.sadek.se@gmail.com>
16460         Microblaze Target: PIC data text relative
16462         * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
16463         * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
16464         Add declaration.
16465         * config/microblaze/microblaze.h (microblaze_constant_address_p):
16466         CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
16467         * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
16468         New addressing mode for data-text relative position indepenedent code.
16469         (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
16470         'ADDRESS_SYMBOLIC_TXT_REL'.
16471         (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
16472         (microblaze_legitimate_pic_operand): Exclude function calls from
16473         pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
16474         (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
16475         addresses cases.
16476         (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
16477         (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
16478         (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
16479         for 'address + offset'.
16480         (microblaze_expand_prologue): Add new function prologue call for
16481         'r20' assignation.
16482         (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
16483         'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
16484         table in case of TARGET_PIC_DATA_TEXT_REL.
16485         (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
16486         * config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
16487         Add new macros 'UNSPEC_TEXT',
16488         'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
16489         + exclude function calls from 'UNSPEC_PLT' in case of data text
16490         relative mode.
16491         * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
16492         new target hook for generating address diff vector tables in case of
16493         flag_pic.
16494         * doc/tm.texi : Regenerate.
16495         * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
16496         'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
16497         of addr diff vector generation.
16498         * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
16499         target hook definition.
16500         * targhooks.h, targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
16501         Add default function for generate_pic_addr_diff_vec -> flag_pic.
16502         * doc/invoke.texi (Add new pic option): Add new microblaze pic
16503         option for data text relative.
16505 2018-04-30  Richard Biener  <rguenther@suse.de>
16507         * tree-chrec.h (evolution_function_is_constant_p): Remove
16508         redundant check.
16509         * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
16511 2018-04-30  Richard Biener  <rguenther@suse.de>
16513         PR bootstrap/85571
16514         * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
16516 2018-04-30  Richard Biener  <rguenther@suse.de>
16518         PR tree-optimization/28364
16519         PR tree-optimization/85275
16520         * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
16521         copying first exit test.
16523 2018-04-28  Mark Wielaard  <mark@klomp.org>
16525         * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
16526         dwarf_version >= 5.
16527         (dwarf_AT): Handle DW_AT_addr_base.
16528         (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
16530 2018-04-28  Uros Bizjak  <ubizjak@gmail.com>
16532         PR target/84431
16533         * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
16534         (*ashl<dwi>3_doubleword_mask_1): Ditto.
16535         (*<shift_insn><dwi>3_doubleword_mask): Ditto.
16536         (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
16538 2018-04-28  Richard Biener  <rguenther@suse.de>
16540         * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
16541         (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
16542         to reflect use.  Only add interesting stmts.
16544 2018-04-27  Martin Jambor  <mjambor@suse.cz>
16546         PR ipa/85549
16547         * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
16548         the jump function allows for passing through aggregate values.
16550 2018-04-27  David Malcolm  <dmalcolm@redhat.com>
16552         * input.h (in_system_header_at): Convert from macro to inline
16553         function.
16554         (from_macro_expansion_at): Likewise.
16555         (from_macro_definition_at): Likewise.
16557 2018-04-27  Jeff Law  <law@redhat.com>
16559         * config.gcc: Mark tile* targets as deprecated/obsolete.
16561 2018-04-27  Richard Biener  <rguenther@suse.de>
16563         * config/aarch64/aarch64.c: Simplify ap.__stack advance and
16564         fix for ILP32.
16566 2018-04-27  Richard Biener  <rguenther@suse.de>
16568         * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
16570 2018-04-27  Uros Bizjak  <ubizjak@gmail.com>
16572         * config/i386/i386.md (*movti_internal): Substitute Ye constraint
16573         with Yd constraint. Set "preferred_for_speed" attribute from
16574         TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
16575         with Yd constraint.
16576         (*movdi_internal): Ditto.
16577         (movti_interunit splitters): Remove
16578         TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
16579         (movdi_interunit splitters): Ditto.
16580         * config/i386/constraints.md (Ye): Remove.
16581         (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
16583 2018-04-27  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
16585         PR target/85512
16586         * config/aarch64/constraints.md (Usg): Limit to 31.
16587         (Usj): Limit to 63.
16589 2018-04-27  Jakub Jelinek  <jakub@redhat.com>
16591         PR tree-optimization/85529
16592         * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
16593         argument.  Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
16594         rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
16595         zero extension or masking of the MSB bit.
16596         (optimize_range_tests): Add FIRST_BB argument, pass it through
16597         to optimize_range_tests_var_bound.
16598         (maybe_optimize_range_tests, reassociate_bb): Adjust
16599         optimize_range_tests callers.
16601 2018-04-26  Richard Biener  <rguenther@suse.de>
16602             Jakub Jelinek  <jakub@redhat.com>
16604         * cgraph.h (symbol_table): Just declare debug method here.
16605         * symtab.c (symbol_table::debug): Define.
16607 2018-04-26  Eric Botcazou  <ebotcazou@adacore.com>
16609         * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
16611 2018-04-26  Uros Bizjak  <ubizjak@gmail.com>
16613         * config/i386/i386.md ("isa" attribute): Add x64_sse2.
16614         ("enabled" attribute): Handle x64_sse2 "isa" attribute.
16615         (*movdi_internal): Substitute Yi and Yj constraint with x
16616         and Ym and Yn constraint with y constraint.  Update "isa"
16617         attribute and set "preferred_for_speed" attribute from
16618         TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
16619         (*movsi_internal): Ditto.
16620         (*movdf_internal): Ditto.
16621         (*movsf_internal): Ditto.
16622         (*zero_extendsidi2): Ditto.
16623         * config/i386/sse.md (vec_set<mode>_0): Ditto.
16624         (sse2_loadld): Ditto.
16625         (*vec_extract<ssevecmodelower>_0): Ditto.
16626         (*vec_extractv4si_0_zext_sse4): Ditto.
16627         (vec_concatv2di): Ditto.
16628         (*vec_dup<mode>): Ditto.
16629         * config/i386/mmx.md (*mov<mode>_internal): Ditto.
16630         * config/i386/constraints.md (Yi): Remove.
16631         (Yj): Remove.
16632         (Ym): Remove.
16633         (Yn): Remove.
16635 2018-04-26  Nathan Sidwell  <nathan@acm.org>
16637         * dumpfile.c (dump_open): New.
16638         (dump_open_alternate_stream, dump_start, dump_begin): Call it.
16639         (dump_finish): Detect stdio/stderr by value not name.
16641 2018-04-26  Jonathan Wakely  <jwakely@redhat.com>
16643         * doc/invoke.texi (-Wreturn-type): Document default status for C++.
16645 2018-04-26  Tom de Vries  <tom@codesourcery.com>
16647         PR target/84952
16648         * config/nvptx/nvptx.c (verify_neutering_jumps)
16649         (verify_neutering_labels): New function
16650         (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
16652 2018-04-26  Tom de Vries  <tom@codesourcery.com>
16654         PR target/84025
16655         * config/nvptx/nvptx.c (needs_neutering_p): New function.
16656         (nvptx_single): Use needs_neutering_p to skip over insns that do not
16657         need neutering.
16659 2018-04-26  Richard Biener <rguenther@suse.de>
16660             Tom de Vries  <tom@codesourcery.com>
16662         PR lto/85422
16663         * lto-streamer-out.c (output_function): Fixup loops if required to match
16664         discovery done in the reader.
16666 2018-04-26  Richard Biener  <rguenther@suse.de>
16668         PR tree-optimization/85116
16669         * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
16670         have a loop exit from the single latch predecessor.  Remove
16671         case of header with just condition.
16672         (ch_base::copy_headers): Exclude infinite loops from any
16673         processing.
16674         (pass_ch::execute): Record exits.
16676 2018-04-26  Richard Biener  <rguenther@suse.de>
16678         * tree-vect-data-refs.c (vect_get_data_access_cost): Get
16679         prologue cost vector and pass it to vect_get_load_cost.
16680         (vect_get_peeling_costs_all_drs): Likewise.
16681         (vect_peeling_hash_get_lowest_cost): Likewise.
16682         (vect_enhance_data_refs_alignment): Likewise.
16684 2018-04-26  Richard Biener  <rguenther@suse.de>
16686         PR middle-end/85450
16687         * tree-cfg.c (verify_gimple_assign_unary): Restore proper
16688         checking of integer<->pointer conversions.
16689         * omp-expand.c (expand_omp_for_static_nochunk): Avoid
16690         sign-/zero-extending pointer types.
16691         (expand_omp_for_static_chunk): Likewise.
16693 2018-03-22  Hans-Peter Nilsson  <hp@axis.com>
16694             Jean Lee  <xiaoyur347@gmail.com>
16696         * config/mips/mips.c (mips_asan_shadow_offset): New function.
16697         (TARGET_ASAN_SHADOW_OFFSET): Define.
16698         * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
16699         true for -fsanitize=address.
16701 2018-04-25  Mark Wielaard  <mark@klomp.org>
16703         * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
16704         shorter ones.
16706 2018-04-25  Jakub Jelinek  <jakub@redhat.com>
16708         * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
16709         than "alu", remove explicit "memory" and "imm_disp" attributes.
16710         (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
16712         PR middle-end/85414
16713         * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
16714         case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
16715         gen_lowpart_no_emit.
16717 2018-04-25  Sebastian Peryt  <sebastian.peryt@intel.com>
16719         PR target/85473
16720         * config/i386/i386.c (ix86_expand_builtin): Change memory
16721         operand to XI, extend p0 to Pmode.
16722         * config/i386/i386.md: Change unspec volatile and operand
16723         1 mode to XI, change operand 0 mode to P.
16725 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
16727         * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
16728         GET_MODE_MASK before any checking.
16729         (nds32_can_use_bset_p): Likewise.
16730         (nds32_can_use_btgl_p): Likewise.
16732 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
16734         * config/nds32/nds32-doubleword.md: New define_split pattern for
16735         illegal register number.
16737 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
16739         * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
16741 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
16743         * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
16745 2018-04-25  Richard Biener  <rguenther@suse.de>
16747         * lto-streamer.h (LTO_major_version): Bump to 8.
16749 2018-04-25  Jakub Jelinek  <jakub@redhat.com>
16751         * BASE-VER: Set to 9.0.0.
16753 2018-04-24  Segher Boessenkool  <segher@kernel.crashing.org>
16755         * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
16756         in __abskf2 and __powikf2.
16758 2018-04-24  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
16760         PR target/85512
16761         * config/aarch64/constraints.md (Usg, Usj): New constraints.
16762         * config/aarch64/iterators.md (cmode_simd): New mode attribute.
16763         * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
16764         Use the above on operand 2.  Reindent.
16765         (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
16767 2018-04-24  H.J. Lu  <hongjiu.lu@intel.com>
16769         PR target/85485
16770         * common/config/i386/i386-common.c (ix86_handle_option): Don't
16771         handle OPT_mcet.
16772         * config/i386/i386.opt (mcet): Removed.
16773         * doc/install.texi: Remove -mcet documentation.
16774         * doc/invoke.texi: Likewise.
16776 2018-04-24  H.J. Lu  <hongjiu.lu@intel.com>
16778         PR target/85485
16779         * doc/install.texi: Remove -mcet from bootstrap-cet.
16781 2018-04-24  Jakub Jelinek  <jakub@redhat.com>
16783         PR target/85511
16784         * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
16785         __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
16786         if TARGET_64BIT.
16788         PR target/85503
16789         * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
16790         const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
16791         containing a CONST_VECTOR.
16793 2018-04-24  Cesar Philippidis  <cesar@codesourcery.com>
16795         * doc/install.texi: Update newlib dependency for nvptx.
16797 2018-04-24  Jakub Jelinek  <jakub@redhat.com>
16799         PR target/85508
16800         * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
16801         instead of INTVAL when shifting x left.
16803 2018-04-24  Andreas Krebbel  <krebbel@linux.ibm.com>
16805         PR tree-optimization/85478
16806         * tree-vect-loop.c (vect_analyze_loop_2): Do not call
16807         vect_grouped_store_supported for single element vectors.
16809 2018-04-24  Richard Biener  <rguenther@suse.de>
16811         PR target/85491
16812         * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
16813         load cost increase to the case of non-constant step.
16815 2018-04-24  Jakub Jelinek  <jakub@redhat.com>
16817         PR target/84828
16818         * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
16819         destination if any_malformed_asm.
16821 2018-04-23  Eric Botcazou  <ebotcazou@adacore.com>
16823         PR middle-end/85496
16824         * expr.c (store_field): In the bitfield case, if the value comes from
16825         a function call and is returned in registers by means of a PARALLEL,
16826         do not change the mode of the temporary unless BLKmode and VOIDmode.
16828 2018-04-23  Andrey Belevantsev  <abel@ispras.ru>
16830         PR rtl-optimization/85423
16831         * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
16832         dependencies to debug insns when the previous insn is non-debug.
16834 2018-04-23  Claudiu Zissulescu  <claziss@synopsys.com>
16836         * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
16837         enums into a single definition.
16838         (fls): Fix predicates and printing.
16839         (seti): Likewise.
16841 2018-04-23  Claudiu Zissulescu  <claziss@synopsys.com>
16843         * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
16844         * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
16845         and short u6 immediate.
16846         (check_if_valid_sleep_operand): Remove.
16847         * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
16849 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
16851         * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
16852         flag_always_save_lp condition.
16853         * config/nds32/nds32.opt (malways-save-lp): New option.
16855 2018-04-22  Shiva Chen  <shiva0217@gmail.com>
16857         * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
16858         * config/nds32/nds32.c (nds32_use_load_post_increment): New.
16859         * config/nds32/nds32.h
16860         (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
16861         (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
16863 2018-04-22  Shiva Chen  <shiva0217@gmail.com>
16865         * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
16866         * config/nds32/nds32.c (nds32_ls_333_p): Remove.
16868 2018-04-22  Shiva Chen  <shiva0217@gmail.com>
16869             Chung-Ju Wu  <jasonwucj@gmail.com>
16871         * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
16872         Declare.
16873         * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
16874         * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
16876 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
16878         * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
16880 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
16882         * config/nds32/nds32-protos.h (nds32_data_alignment,
16883         nds32_local_alignment): Declare.
16884         * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
16885         nds32_local_alignment): New functions.
16886         (TARGET_CONSTANT_ALIGNMENT): Define.
16887         * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
16889 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
16891         * config/nds32/nds32.c
16892         (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
16893         (TARGET_MODES_TIEABLE_P): Likewise.
16895 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
16897         * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
16898         level Ofast and Og.
16900 2018-04-22  Monk Chiang  <sh.chiang04@gmail.com>
16901             Chung-Ju Wu  <jasonwucj@gmail.com>
16903         * config/nds32/constants.md (unspec_volatile_element): Add enum values
16904         for unaligned access.
16905         * config/nds32/nds32-intrinsic.c: Implementation of expanding
16906         unaligned access.
16907         * config/nds32/nds32-intrinsic.md: Likewise.
16908         * config/nds32/nds32_intrinsic.h: Likewise.
16909         * config/nds32/nds32.h (nds32_builtins): Likewise.
16910         * config/nds32/nds32.opt (munaligned-access): New option.
16911         * config/nds32/nds32.c (nds32_asm_file_start): Display
16912         flag_unaligned_access status.
16914 2018-04-20  Kito Cheng  <kito.cheng@gmail.com>
16916         * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
16917         -mno-relax is present.
16918         * config/riscv/linux.h (LINK_SPEC): Ditto.
16920 2018-04-20  Martin Sebor  <msebor@redhat.com>
16922         PR c/85365
16923         * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
16924         for null pointers.
16925         (gimple_fold_builtin_stxcpy_chk): Same.
16926         * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
16928 2018-04-20  Michael Meissner  <meissner@linux.ibm.com>
16930         PR target/85456
16931         * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
16932         __powikf2 when long double is IEEE 128-bit.
16934 2018-04-20  Kito Cheng  <kito.cheng@gmail.com>
16936         * config/riscv/riscv.c (riscv_first_stack_step): Round up min
16937         step to make sure stack always aligned.
16939 2018-04-20  Carl Love  <cel@us.ibm.com>
16941         PR target/83402
16942         * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
16943         size check for arg0.
16945 2018-04-20  Nathan Sidwell  <nathan@codesourcery.com>
16946             Tom de Vries  <tom@codesourcery.com>
16948         PR target/85445
16949         * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
16950         Emit insns for calls too.
16951         (nvptx_find_par): Always look for worker-level predecessor insn.
16952         (nvptx_propagate): Add is_call parm, return bool.  Copy frame for
16953         calls.
16954         (nvptx_vpropagate, nvptx_wpropagate): Adjust.
16955         (nvptx_process_pars): Propagate frames for calls.
16957 2018-04-20  H.J. Lu  <hongjiu.lu@intel.com>
16959         PR target/85469
16960         * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
16961         Removed.
16962         (OPTION_MASK_ISA_IBT_UNSET): Likewise.
16963         (ix86_handle_option): Don't handle OPT_mibt.
16964         * config/i386/cet.h: Check __CET__ instead of __IBT__ and
16965         __SHSTK__.
16966         * config/i386/driver-i386.c (host_detect_local_cpu): Remove
16967         has_ibt and ibt.
16968         * config/i386/i386-c.c (ix86_target_macros_internal): Don't
16969         check OPTION_MASK_ISA_IBT nor flag_cf_protection.
16970         (ix86_target_macros): Define __CET__ with flag_cf_protection
16971         for -fcf-protection.
16972         * config/i386/i386.c (isa2_opts): Remove -mibt.
16973         * config/i386/i386.h (TARGET_IBT): Removed.
16974         (TARGET_IBT_P): Likewise.
16975         (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
16976         * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
16977         * config/i386/i386.opt (mcet): Update help message.
16978         (mshstk): Likewise.
16979         (mibt): Removed.
16980         * doc/invoke.texi: Remove -mibt.  Document __CET__.  Document
16981         -mcet as an alias for -mshstk.
16983 2018-04-20  Richard Biener <rguenther@suse.de>
16985         PR middle-end/85475
16986         * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
16987         complexity by forcing a single use of the multiply operand.
16989 2018-04-20  Martin Jambor  <mjambor@suse.cz>
16991         ipa/85449
16992         * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
16993         recursion dependency to only apply to non-clones.
16995 2018-04-20  Martin Jambor  <mjambor@suse.cz>
16997         ipa/85447
16998         * ipa-cp.c (create_specialized_node): Check that clones of
16999         self-recursive edges exist during IPA-CP.
17001 2018-04-19  Toon Moene  <toon@moene.org>
17003         * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
17004         by -O3.
17006 2018-04-19  Jakub Jelinek  <jakub@redhat.com>
17008         PR tree-optimization/85467
17009         * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
17010         VECTOR_TYPE_P macro.  If type is vector type, VIEW_CONVERT_EXPR the
17011         VECTOR_CST element to type.
17013 2018-04-19  H.J. Lu  <hongjiu.lu@intel.com>
17015         PR target/85397
17016         * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
17017         * config/i386/i386.md (builtin_setjmp_setup): Removed.
17018         (builtin_longjmp): Likewise.
17019         (save_stack_nonlocal): New pattern.
17020         (restore_stack_nonlocal): Likewise.
17022 2018-04-19  H.J. Lu  <hongjiu.lu@intel.com>
17024         PR target/85404
17025         * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
17026         Replace ASM_OUTPUT_LABEL with fprintf.
17028 2018-04-19  H.J. Lu  <hongjiu.lu@intel.com>
17030         PR target/85417
17031         * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
17032         Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
17033         * config/i386/i386-c.c (ix86_target_macros_internal): Also
17034         define __IBT__ and __SHSTK__ for -fcf-protection.
17035         * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
17036         TARGET_IBT.
17037         (ix86_trampoline_init): Likewise.
17038         (x86_output_mi_thunk): Likewise.
17039         (ix86_notrack_prefixed_insn_p): Likewise.
17040         (ix86_option_override_internal): Don't disallow -fcf-protection.
17041         * config/i386/i386.md (rdssp<mode>): Also enable for
17042         -fcf-protection.
17043         (incssp<mode>): Likewise.
17044         (nop_endbr): Likewise.
17045         * config/i386/i386.opt (mcet): Change help message to built-in
17046         functions only.
17047         (mibt): Likewise.
17048         (mshstk): Likewise.
17049         * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
17050         on -fcf-protection.  Change -mcet, -mibt and -mshstk to only
17051         enable CET built-in functions.
17053 2018-04-19  Sebastian Peryt  <sebastian.peryt@intel.com>
17055         * common/config/i386/i386-common.c
17056         (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
17057         OPTION_MASK_ISA_MOVDIRI_UNSET,
17058         OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
17059         (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
17060         * config.gcc (movdirintrin.h): New header.
17061         * config/i386/cpuid.h (bit_MOVDIRI,
17062         bit_MOVDIR64B): New bits.
17063         * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
17064         and -mmvodir64b.
17065         * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
17066         (VOID, PVOID, PCVOID)): New function types.
17067         * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
17068         __builtin_ia32_directstoreu_u64,
17069         __builtin_ia32_movdir64b): New builtins.
17070         * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
17071         * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
17072         and -mmovdiri.
17073         (ix86_valid_target_attribute_inner_p): Ditto.
17074         (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
17075         and VOID_FTYPE_PUNSIGNED_UNSIGNED.
17076         (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
17077         * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
17078         TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
17079         * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
17080         (movdiri<mode>, movdir64b_<mode>): New.
17081         * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
17082         * config/i386/immintrin.h: Include movdirintrin.h.
17083         * config/i386/movdirintrin.h: New file.
17084         * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
17086 2018-04-19  Richard Biener  <rguenther@suse.de>
17088         PR middle-end/85455
17089         * cfg.c (clear_bb_flags): When loop state says we have
17090         marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
17092 2018-04-19  Richard Biener  <rguenther@suse.de>
17094         PR tree-optimization/84737
17095         * tree-vect-data-refs.c (vect_copy_ref_info): New function
17096         copying restrict info.
17097         (vect_setup_realignment): Use it.
17098         * tree-vectorizer.h (vect_copy_ref_info): Declare.
17099         * tree-vect-stmts.c (vectorizable_store): Copy ref info from
17100         the first DR to all generated stores.
17101         (vectorizable_load): Likewise for loads.
17103 2018-04-19  Jakub Jelinek  <jakub@redhat.com>
17105         PR tree-optimization/85446
17106         * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
17107         the integral and pointer types to have the same precision.
17109         * doc/install.texi: Document --disable-cet being the default and
17110         --enable-cet=auto.
17112 2018-04-18  Martin Liska  <mliska@suse.cz>
17114         * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
17115         style.
17117 2018-04-18  Martin Liska  <mliska@suse.cz>
17119         Revert
17120         2018-03-02  Eric Botcazou  <ebotcazou@adacore.com>
17122         PR ipa/83983
17123         * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
17124         arguments if they are comparable.
17126 2018-04-18  Martin Liska  <mliska@suse.cz>
17128         Revert
17129         2018-03-13  Eric Botcazou  <ebotcazou@adacore.com>
17131         PR lto/84805
17132         * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
17133         incomplete types.
17135 2018-04-18  H.J. Lu  <hongjiu.lu@intel.com>
17137         PR target/85388
17138         * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
17139         ENDBR after calling __morestack.
17141 2018-04-18  David Malcolm  <dmalcolm@redhat.com>
17143         PR jit/85384
17144         * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
17145         by using gcc_base_ver to generate a gcc_driver_version, and use
17146         it when generating GCC_DRIVER_NAME.
17147         * configure: Regenerate.
17149 2018-04-18  Jakub Jelinek  <jakub@redhat.com>
17151         PR target/81084
17152         * config.gcc: Obsolete powerpc*-*-*spe*.
17154 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
17156         PR debug/84637
17157         * dbxout.c (dbxout_int): Perform negation in unsigned int type.
17158         (stabstr_D): Change type of unum from unsigned int to
17159         unsigned HOST_WIDE_INT.  Perform negation in unsigned HOST_WIDE_INT
17160         type.
17162 2018-04-17  Jim Wilson  <jimw@sifive.com>
17164         PR 84856
17165         * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
17166         RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
17167         Set arg_pointer_offset after using pretend_args_size.
17169 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
17171         PR rtl-optimization/85431
17172         * dse.c (record_store): Ignore zero width stores.
17174         PR sanitizer/85230
17175         * asan.c (handle_builtin_stack_restore): Adjust comment.  Emit
17176         __asan_allocas_unpoison call and last_alloca_addr = new_sp before
17177         __builtin_stack_restore rather than after it.
17178         * builtins.c (expand_asan_emit_allocas_unpoison): Pass
17179         arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
17180         argument instead of virtual_dynamic_stack_rtx.
17182 2018-04-17  Kelvin Nilsen  <kelvin@gcc.gnu.org>
17184         * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
17185         New prototype.
17186         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
17187         Add note to error message to explain internal mapping of overloaded
17188         built-in function name to non-overloaded built-in function name.
17189         * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
17190         function.
17192 2018-04-17  Michael Meissner  <meissner@linux.vnet.ibm.com>
17194         PR target/85424
17195         * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
17196         where the inputs overlap with the output.
17198 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
17200         * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
17201         (=v, v) alternative and explicit "memory" attribute.
17202         (vec_extract_lo_<mode><mask_name>): Likewise.  Also add
17203         "type", "prefix", "prefix_extra", "length_immediate" and "mode"
17204         attributes.
17205         (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
17206         "sselog1" type instead of "sselog".
17207         (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
17208         "sselog".  Remove explicit "memory" attribute.
17209         (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
17210         "type", "prefix", "prefix_extra", "length_immediate" and "mode"
17211         attributes.
17212         (vec_extract_hi_v32hi): Merge all alternatives into one, use
17213         "sselog1" type instead of "sselog".  Remove explicit "memory"
17214         attribute.
17215         (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
17216         use "sselog1" type instead of "sselog".  Remove explicit "memory"
17217         attribute.
17218         (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
17219         "type", "prefix", "prefix_extra", "length_immediate" and "mode"
17220         attributes.
17221         (vec_extract_hi_v64qi): Merge all alternatives into one, use
17222         "sselog1" type instead of "sselog".  Remove explicit "memory"
17223         attribute.
17224         (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
17225         use "sselog1" type instead of "sselog".  Remove explicit "memory"
17226         attribute.
17228         PR target/85430
17229         * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
17231         PR middle-end/85414
17232         * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
17233         on a SUBREG.
17235 2018-04-17  Martin Jambor  <mjambor@suse.cz>
17237         PR ipa/85421
17238         * ipa-cp.c (create_specialized_node): Call
17239         expand_all_artificial_thunks if necessary.
17241 2018-04-17  Martin Liska  <mliska@suse.cz>
17243         PR lto/85405
17244         * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
17245         in message, remote space in between '_G' and '('.
17247 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
17249         PR target/85281
17250         * config/i386/sse.md (reduces<mode><mask_scalar_name>,
17251         avx512f_vmcmp<mode>3<round_saeonly_name>,
17252         avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
17253         avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
17254         avx512f_rndscale<mode><round_saeonly_name>,
17255         avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
17256         avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
17257         Use %<iptr>2 instead of %2 for -masm=intel.
17258         (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
17259         avx512f_vcvttss2usi<round_saeonly_name>,
17260         avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
17261         -masm=intel.
17262         (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
17263         avx512f_vcvttsd2usi<round_saeonly_name>,
17264         avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
17265         Use %q1 instead of %1 for -masm=intel.
17266         (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
17267         avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
17268         of %3 for -masm=intel.
17269         (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
17270         -masm=intel.
17271         (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
17272         -masm=intel.
17273         (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
17274         -masm=intel.
17275         (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
17276         %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
17277         %g1.
17278         (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
17279         -masm=intel.
17280         (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
17281         %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
17282         %g1 and one with %0 and %1.
17283         (avx512er_vmrcp28<mode><round_saeonly_name>,
17284         avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
17285         %1 for -masm=intel.
17286         (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
17287         avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
17288         avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
17289         of %0 and %{%4%} for -masm=intel.
17290         (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
17291         avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
17292         avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
17293         order of %0 and %{%5%}%{z%} for -masm=intel.
17295 2018-04-17  Jan Hubicka  <jh@suse.cz>
17297         PR lto/85405
17298         * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
17300 2018-04-17  Martin Liska  <mliska@suse.cz>
17302         PR ipa/85329
17303         * multiple_target.c (create_dispatcher_calls): Set apostrophes
17304         for target_clone error message.  Make default implementation
17305         clone to be a local declaration.
17306         (separate_attrs): Add new argument and check for an empty
17307         string.
17308         (expand_target_clones): Handle it.
17309         (ipa_target_clone): Make redirection just for target_clones
17310         functions.
17312 2018-04-16  Cesar Philippidis  <cesar@codesourcery.com>
17313             Tom de Vries  <tom@codesourcery.com>
17315         PR middle-end/84955
17316         * omp-expand.c (expand_oacc_for): Add dummy false branch for
17317         tiled basic blocks without omp continue statements.
17319 2018-04-16  Aaron Sawdey  <acsawdey@linux.ibm.com>
17321         PR target/83660
17322         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
17323         vec_extract expression as having side effects to make sure it gets
17324         a cleanup point.
17326 2018-04-16  H.J. Lu  <hongjiu.lu@intel.com>
17328         PR target/85403
17329         * config/i386/i386.c (get_builtin_code_for_version): Check
17330         error_mark_node.
17332 2018-04-16  Olga Makhotina  <olga.makhotina@intel.com>
17334         PR target/84331
17335         * config.gcc: Support "skylake".
17336         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
17337         PROCESSOR_SKYLAKE.
17338         * config/i386/i386.c (m_SKYLAKE): Define.
17339         (processor_target_table): Add "skylake".
17340         (ix86_option_override_internal): Add "skylake".
17341         (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
17342         PROCESSOR_CANNONLAKE.
17343         (get_builtin_code_for_version): Fix priority for
17344         PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
17345         PROCESSOR_SKYLAKE-AVX512.
17346         * config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
17347         (processor_type): Add PROCESSOR_SKYLAKE.
17349 2018-04-16  Paolo Carlini  <paolo.carlini@oracle.com>
17350             Jason Merrill  <jason@redhat.com>
17352         PR c++/85112
17353         * convert.c (convert_to_integer_1): Use direct recursion for
17354         enumeral types and types with a precision less than the number
17355         of bits in their mode.
17357 2018-04-16  Julia Koval  <julia.koval@intel.com>
17359         PR target/84413
17360         * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
17361         X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
17363 2018-04-14  Segher Boessenkool  <segher@kernel.crashing.org>
17365         PR target/85293
17366         * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
17367         * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
17368         and -mno-direct-move.
17370 2018-04-13  Paul A. Clarke  <pc@us.ibm.com>
17372         PR target/83402
17373         * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
17374         Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
17375         Ensure negative shifts result in {0}.
17377 2018-04-13  Vladimir Makarov  <vmakarov@redhat.com>
17379         PR rtl-optimization/79916
17380         * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
17381         regs (if any) to define how to gnerate SD moves when LRA is in
17382         progress.
17384 2018-04-13  Jakub Jelinek  <jakub@redhat.com>
17386         PR rtl-optimization/85393
17387         * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
17388         * except.c (expand_dw2_landing_pad_for_region): Make static.
17389         * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
17390         a label and unconditional jump to old_bb, rather than
17391         expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
17392         basic block.
17394         PR rtl-optimization/85376
17395         * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
17396         zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
17397         instead of a specific value.
17399 2018-04-13  Jan Hubicka  <hubicka@ucw.cz>
17400             Bin Cheng  <bin.cheng@arm.com>
17402         PR tree-optimization/82965
17403         PR tree-optimization/83991
17404         * cfgloopanal.c (expected_loop_iterations_unbounded): Add
17405         by_profile_only parameter.
17406         * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
17407         information if the loop was predicted to iterate too many times.
17408         * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
17410 2018-04-13  Jan Hubicka  <hubicka@ucw.cz>
17412         PR lto/71991
17413         * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
17414         always inline.
17416 2018-04-13  Martin Liska  <mliska@suse.cz>
17417             Jakub Jelinek  <jakub@redhat.com>
17419         PR middle-end/81657
17420         * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
17421         * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
17422         * builtins.c (expand_builtin_memory_copy_args): Use
17423         BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
17424         handle dest_addr == pc_rtx.
17426 2018-04-12  Segher Boessenkool  <segher@kernel.crashing.org>
17428         PR target/85291
17429         * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
17430         asked to not generate direct moves.
17431         (fix_trunc<mode>si2_stfiwx): Similar.
17432         (fix_trunc<mode>si2_internal): Similar.
17434 2018-04-12  Jakub Jelinek  <jakub@redhat.com>
17436         PR debug/83157
17437         * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
17438         * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
17439         lookup if dest in some wider mode is known to be const0_rtx and
17440         if so, record permanent equivalence for it to be ZERO_EXTEND of
17441         the narrower mode destination.
17443 2018-04-12  Cesar Philippidis  <cesar@codesourcery.com>
17445         * lto-streamer-out.c (output_function): Revert 259346.
17446         * omp-expand.c (expand_oacc_for): Likewise.
17448 2018-04-12  Alexander Monakov  <amonakov@ispras.ru>
17450         PR rtl-optimization/85354
17451         * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
17452         * sel-sched.c (sel_global_init): ... here.
17454 2018-04-12  Eric Botcazou  <ebotcazou@adacore.com>
17456         PR target/85238
17457         * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
17458         * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
17459         mode for PE-COFF targets.
17460         * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
17461         (i386_pe_asm_lto_end): Likewise.
17462         * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
17463         (TARGET_ASM_LTO_END): Likewise.
17464         * config/i386/winnt.c (saved_debug_info_level): New static variable.
17465         (i386_pe_asm_lto_start): New function.
17466         (i386_pe_asm_lto_end): Likewise.
17468 2018-04-12  Cesar Philippidis  <cesar@codesourcery.com>
17469             Richard Biener  <rguenther@suse.de>
17471         PR middle-end/84955
17472         * lto-streamer-out.c (output_function): Fix CFG loop state before
17473         streaming out.
17474         * omp-expand.c (expand_oacc_for): Handle calls to internal
17475         functions like regular functions.
17477 2018-04-12  Richard Biener  <rguenther@suse.de>
17479         PR lto/85371
17480         * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
17481         for the early LTO debug to properly generate references to it
17482         during DIE emission.  Do not re-use that for the skeleton for
17483         split-dwarf.
17484         (dwarf2out_early_finish): Likewise.
17486 2018-04-12  Jakub Jelinek  <jakub@redhat.com>
17488         PR target/85328
17489         * config/i386/sse.md
17490         (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
17491         <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
17492         vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
17493         vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
17494         and output is a reg, avoid creating invalid lowpart subreg, but
17495         instead split into a 512-bit move.  Don't split if not AVX512VL,
17496         input is xmm16+ reg and output is a mem.
17497         (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
17498         vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
17499         xmm16+ reg and output is a mem.
17501 2018-04-12  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
17503         * config/s390/s390.c (s390_output_indirect_thunk_function): Check
17504         also for flag_dwarf2_cfi_asm.
17506 2018-04-12  Jakub Jelinek  <jakub@redhat.com>
17508         PR rtl-optimization/85342
17509         * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
17510         a bool scalar var inside of the loop instead.  Don't try to update
17511         recog_data.operand after failed apply_change_group.
17513 2018-04-12  Tom de Vries  <tom@codesourcery.com>
17515         PR target/85296
17516         * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
17517         (nvptx_assemble_decl_begin): Add undefined param.  Declare undefined
17518         array with flexible array member as array without given dimension.
17519         (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
17520         argument for undefined param to true.
17522 2018-04-11  Aaron Sawdey  <acsawdey@linux.ibm.com>
17524         PR target/85321
17525         * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
17526         -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
17527         from PowerPC section.
17528         * config/rs6000/sysv4.opt (mcall-): Improve help text.
17529         * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
17530         help text that is too long.
17531         * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
17532         help text that is too long.
17533         * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
17534         help text that is too long.
17536 2018-04-11  Uros Bizjak  <ubizjak@gmail.com>
17538         * config/alpha/alpha.md (stack_probe_internal): Rename
17539         from "probe_stack".  Update all callers.
17541 2018-04-11  Alexander Monakov  <amonakov@ispras.ru>
17543         PR rtl-optimization/84566
17544         * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
17545         sched_macro_fuse_insns.
17547 2018-04-11  Alexander Monakov  <amonakov@ispras.ru>
17549         PR target/84301
17550         * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
17551         (compute_block_dependences): ... from here.
17553 2018-04-11  Jakub Jelinek  <jakub@redhat.com>
17555         PR tree-optimization/85331
17556         * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
17557         from int to HOST_WIDE_INT.
17559 2018-04-11  Martin Jambor  <mjambor@suse.cz>
17561         PR ipa/84149
17562         * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
17563         (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
17564         not the same as the source val.
17565         (cgraph_edge_brings_value_p): New parameter.
17566         (gather_edges_for_value): Pass destination value to
17567         cgraph_edge_brings_value_p.
17568         (perhaps_add_new_callers): Likewise.
17569         (get_info_about_necessary_edges): Likewise and exclude values brought
17570         only by self-recursive edges.
17571         (create_specialized_node): Redirect only clones of self-calling edges.
17572         (+self_recursive_pass_through_p): New function.
17573         (find_more_scalar_values_for_callers_subset): Use it.
17574         (find_aggregate_values_for_callers_subset): Likewise.
17575         (known_aggs_to_agg_replacement_list): Removed.
17576         (decide_whether_version_node): Re-calculate known constants for all
17577         remaining context clones.
17579 2018-04-11  Richard Biener  <rguenther@suse.de>
17581         PR lto/85339
17582         * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
17583         from early DWARF output.
17584         (dwarf2out_early_finish): Output line info unconditionally into
17585         early DWARF and add reference to it.
17587 2018-04-11  Jakub Jelinek  <jakub@redhat.com>
17589         PR target/85281
17590         * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
17591         (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
17592         other than V2DFmode using iptr mode attribute.
17593         (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
17595 2018-04-11  Alexander Monakov  <amonakov@ispras.ru>
17597         PR rtl-optimization/84659
17598         * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
17600 2018-04-11  Jakub Jelinek  <jakub@redhat.com>
17602         PR debug/85302
17603         * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
17604         SIZEP is NULL.
17605         (output_loc_list): Pass address of a dummy size variable even in the
17606         locview handling loop.
17607         (index_location_lists): Add comment on why skip_loc_list_entry can't
17608         call size_of_locs.
17610 2018-04-11  Thomas Preud'homme  <thomas.preudhomme@arm.com>
17612         PR target/85261
17613         * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
17614         into register.
17616 2018-04-10  Aaron Sawdey  <acsawdey@linux.ibm.com>
17618         PR target/85321
17619         * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
17620         -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
17621         and -mstring-compare-inline-limit.
17623 2018-04-10  Segher Boessenkool  <segher@kernel.crashing.org>
17625         PR target/85287
17626         * config/rs6000/rs6000.md (allocate_stack): Put the residual size
17627         for stack clash protection in a register whenever we need it to be in
17628         a register.
17630 2018-04-10  Segher Boessenkool  <segher@kernel.crashing.org>
17632         * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
17633         Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
17635 2018-04-10  Segher Boessenkool  <segher@kernel.crashing.org>
17637         PR target/85321
17638         * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
17639         the help text.
17640         (mlong-double-): Ditto.
17641         * config/rs6000/sysv4.opt (msdata=): Ditto.
17642         (mtls-size=): Ditto.
17644 2018-04-10  Kelvin Nilsen  <kelvin@gcc.gnu.org>
17646         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
17647         erroneous entries for
17648         "vector int vec_ldl (int, long int *)", and
17649         "vector unsigned int vec_ldl (int, unsigned long int *)".
17650         Add comments and entries for
17651         "vector bool char vec_ldl (int, bool char *)",
17652         "vector bool short vec_ldl (int, bool short *)",
17653         "vector bool int vec_ldl (int, bool int *)",
17654         "vector bool long long vec_ldl (int, bool long long *)",
17655         "vector pixel vec_ldl (int, pixel *)",
17656         "vector long long vec_ldl (int, long long *)",
17657         "vector unsigned long long vec_ldl (int, unsigned long long *)".
17658         * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
17659         type tree bool_long_long_type_node and correct definition of
17660         bool_V2DI_type_node to make reference to this new type tree.
17661         (rs6000_mangle_type): Replace erroneous reference to
17662         bool_long_type_node with bool_long_long_type_node.
17663         * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
17664         comments to emphasize sign distinctions for char and int types and
17665         replace RS6000_BTI_bool_long constant with
17666         RS6000_BTI_bool_long_long constant.  Also add comment to restrict
17667         use of RS6000_BTI_pixel.
17668         (bool_long_type_node): Remove this macro definition.
17669         (bool_long_long_type_node): New macro definition
17671 2018-04-10  Jakub Jelinek  <jakub@redhat.com>
17673         PR rtl-optimization/85300
17674         * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
17675         into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
17676         simplify_unary_operation fails.
17678 2018-04-10  Martin Liska  <mliska@suse.cz>
17680         * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
17681         cgraph_edge and ipa_ref.
17683 2018-04-10  Jakub Jelinek  <jakub@redhat.com>
17685         PR target/85177
17686         PR target/85255
17687         * config/i386/sse.md
17688         (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
17689         computation of the VEC_MERGE selector from mask.
17690         (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
17691         Fix decoding of the VEC_MERGE selector into mask.
17693 2018-04-10  Richard Sandiford  <richard.sandiford@linaro.org>
17695         PR tree-optimization/85286
17696         * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
17698 2018-04-10  Richard Sandiford  <richard.sandiford@linaro.org>
17700         * final.c (final_1): Set insn_last_address as well as
17701         insn_current_address.
17703 2018-04-10  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
17705         PR target/85173
17706         * explow.c (emit_stack_probe): Call validize_mem on memory location
17707         before passing it to gen_probe_stack.  Create address operand and
17708         legitimize it for the probe_stack_address case.
17710 2018-04-09  Jan Hubicka  <jh@suse.cz>
17712         PR lto/85078
17713         * ipa-devirt.c (rebuild_type_inheritance-hash): New.
17714         * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
17715         * tree.c (free_lang_data_in_type): Fix handling of binfos;
17716         walk basetypes.
17717         (free_lang_data): Rebuild type inheritance graph.
17719 2018-04-09  Martin Sebor  <msebor@redhat.com>
17721         * invoke.texi (-finline-small-functions): Mention other optimization
17722         options.
17723         (-findirect-inlining, -fpartial-inlining): Same.
17724         (-finline-functions-called-once): Same.
17725         (-freorder-blocks-and-partition): Same.
17727 2018-04-09  Jan Hubicka  <jh@suse.cz>
17729         PR rtl/84058
17730         * cfgcleanup.c (try_forward_edges): Do not give up on crossing
17731         jumps; choose last target that matches the criteria (i.e.
17732         no partition changes for non-crossing jumps).
17733         * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
17734         support for redirecting crossing jumps to non-crossing.
17736 2018-04-09  Alexey Brodkin  <abrodkin@synopsys.com>
17738         * config/arc/arc.c (arc_expand_prologue): Set stack usage info
17739         also for naked functions.
17741 2018-04-09  Claudiu Zissulescu  <claziss@synopsys.com>
17743         * config/arc/arc.md (add_shift): New pattern.
17744         (add_shift2): Likewise.
17745         (sub_shift): Likewise.
17746         (sub_shift_cmp0_noout): Likewise.
17747         (compare_si_ashiftsi): Likewise.
17748         (xbfu_cmp0_noout): New combine pattern.
17749         (xbfu_cmp0"): Likewise.
17750         (movsi_set_cc_insn): Place the predicable variant first.
17751         (commutative_binary_cmp0_noout): Remove clobber.
17752         (commutative_binary_cmp0): New pattern.
17753         (noncommutative_binary_cmp0): Likewise.
17754         (noncommutative_binary_cmp0_noout): Likewise.
17755         (noncommutative_binary_comparison_result_used): Removed.
17756         (rsub_cmp0): New pattern.
17757         (rsub_cmp0_noout): Likewise.
17758         (extzvsi): Changed, keep only meaningful variants.
17759         (SQH, SEZ): New iterators.
17760         (SQH_postfix): New mode attribute.
17761         (SEZ_prefix): New code attribute.
17762         (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
17763         (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
17764         * config/arc/predicates.md (cc_set_register): Use CC_REG instead
17765         of numerical value.
17766         (noncommutative_operator): Check the availability of barrel
17767         shifter option.
17769 2018-04-09  Richard Biener  <rguenther@suse.de>
17771         PR tree-optimization/85284
17772         * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
17773         Only use the niter constraining form of simple_iv when the exit
17774         is always executed.
17776 2018-04-09  Tom de Vries  <tom@codesourcery.com>
17778         PR target/84041
17779         * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
17780         (define_expand "*memory_barrier"): New define_expand.
17781         (define_insn "memory_barrier"): New insn.
17783 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
17785         PR rtl-optimization/80463
17786         PR rtl-optimization/83972
17787         PR rtl-optimization/83480
17789         * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
17790         correct producer for the insn.
17791         (tidy_control_flow): Fixup seqnos in case of debug insns.
17793 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
17795         PR rtl-optimization/83913
17797         * sel-sched-ir.c (merge_expr_data): Choose the middle between two
17798         different sched-times when merging exprs.
17800 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
17802         PR rtl-optimization/83962
17804         * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
17805         tidy_fallthru_edge and tidy_control_flow.
17807 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
17809         PR rtl-optimization/83530
17811         * sel-sched.c (force_next_insn): New global variable.
17812         (remove_insn_for_debug): When force_next_insn is true, also leave only
17813         next insn in the ready list.
17814         (sel_sched_region): When the region wasn't scheduled, make another pass
17815         over it with force_next_insn set to 1.
17817 2018-04-08  Monk Chiang  <sh.chiang04@gmail.com>
17819         * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
17820         into tm_file.
17821         * config/nds32/constants.md (unspec_volatile_element): Add enum values
17822         for interrupt control.
17823         * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
17824         functions for interrupt control.
17825         * config/nds32/nds32-intrinsic.md: Likewise.
17826         * config/nds32/nds32_intrinsic.h: Likewise.
17827         * config/nds32/nds32.h (nds32_builtins): Likewise.
17829 2018-04-08  Chung-Ju Wu  <jasonwucj@gmail.com>
17831         * config/nds32/nds32.c (nds32_init_machine_status,
17832         nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
17833         strict_aligned_p field.
17834         (nds32_expand_to_rtl_hook): New function.
17835         (TARGET_EXPAND_TO_RTL_HOOK): Define.
17836         * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
17838 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
17839             Chung-Ju Wu  <jasonwucj@gmail.com>
17841         * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
17842         * config/nds32/nds32-n7.md: New file.
17843         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
17844         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
17845         pipeline.
17846         * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
17847         * config/nds32/nds32.md (pipeline_model): Add n7.
17848         * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
17849         * config/nds32/pipelines.md: Include n7 settings.
17851 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
17852             Chung-Ju Wu  <jasonwucj@gmail.com>
17854         * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
17855         * config/nds32/nds32-e8.md: New file.
17856         * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
17857         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
17858         pipeline.
17859         * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
17860         * config/nds32/nds32.md (pipeline_model): Add e8.
17861         * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
17862         * config/nds32/pipelines.md: Include e8 settings.
17864 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
17865             Chung-Ju Wu  <jasonwucj@gmail.com>
17867         * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
17868         * config/nds32/nds32-n8.md: New file.
17869         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
17870         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
17871         pipeline.
17872         * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
17873         * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
17874         * config/nds32/nds32.md (pipeline_model): Add n8.
17875         * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
17876         * config/nds32/pipelines.md: Include n8 settings.
17878 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
17879             Chung-Ju Wu  <jasonwucj@gmail.com>
17881         * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
17882         * config/nds32/nds32-n9-2r1w.md: New file.
17883         * config/nds32/nds32-n9-3r2w.md: New file.
17884         * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
17885         nds32_register_ports): New or modify for cpu n9.
17886         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
17887         pipeline.
17888         * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
17889         * config/nds32/nds32-utils.c: New file.
17890         * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
17891         TARGET_MUL_SLOW): Define.
17892         * config/nds32/nds32.md (pipeline_model): New attribute.
17893         * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
17894         New options that support cpu n9.
17895         * config/nds32/pipelines.md: Include n9 settings.
17896         * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
17898 2018-04-08  Chung-Ju Wu  <jasonwucj@gmail.com>
17900         * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
17901         information if necessary.
17902         (output_cond_branch_compare_zero): Likewise.
17903         * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
17904         (nds32_target_alignment): Refine for alignment.
17905         * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
17906         (FUNCTION_BOUNDARY): Modify.
17907         * config/nds32/nds32.md (call_internal, call_value_internal): Consider
17908         align case.
17909         * config/nds32/nds32.opt (malways-align, malign-functions): New.
17911 2018-04-08  Monk Chiang  <sh.chiang04@gmail.com>
17913         * config/nds32/constants.md (unspec_volatile_element): Add values for
17914         TLB operation and data prefetch.
17915         * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
17916         functions for TLB operation and data prefetch.
17917         * config/nds32/nds32-intrinsic.md: Likewise.
17918         * config/nds32/nds32_intrinsic.h: Likewise.
17919         * config/nds32/nds32.c (nds32_dpref_names): Likewise.
17920         (nds32_print_operand): Likewise.
17921         * config/nds32/nds32.h (nds32_builtins): Likewise.
17923 2018-04-07  Thomas Koenig  <tkoenig@gcc.gnu.org>
17924         Andrew Pinski <pinsika@gcc.gnu.org>
17926         PR middle-end/82976
17927         * match.pd: Use constant_boolean_node of correct type instead of
17928         boolean_true_node or boolean_false_node for simplifying
17929         pointer comparisons to zero.
17931 2018-04-07  Jakub Jelinek  <jakub@redhat.com>
17933         PR tree-optimization/80021
17934         * tree.c (verify_type_variant): Make error call in verify_variant_match
17935         translatable and remove final full stop.
17937 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
17939         * config/nds32/constants.md (unspec_volatile_element): Add
17940         UNSPEC_VOLATILE_EH_RETURN.
17941         * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
17942         nds32_output_stack_pop): Support dwarf exception handling process.
17943         * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
17944         * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
17945         exception handling process.
17946         (nds32_compute_stack_frame): Likewise.
17947         (nds32_return_addr_rtx): Likewise.
17948         (nds32_initial_elimination_offset): Likewise.
17949         (nds32_expand_prologue): Likewise.
17950         (nds32_expand_epilogue): Likewise.
17951         (nds32_dynamic_chain_address): New function.
17952         * config/nds32/nds32.h (machine_function): Add fields for dwarf
17953         exception handling.
17954         (DYNAMIC_CHAIN_ADDRESS): Define.
17955         (EH_RETURN_DATA_REGNO): Define.
17956         (EH_RETURN_STACKADJ_RTX): Define.
17957         * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
17958         patterns for dwarf exception handling.
17960 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
17962         * config/nds32/nds32.h: Clean up obsolete macros.
17964 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
17966         * config/nds32/constants.md (unspec_element, unspec_volatile_element):
17967         Add enum values for particular instructions.
17968         * config/nds32/nds32-intrinsic.c: Implementation of expanding
17969         particular intrinsic functions.
17970         * config/nds32/nds32-intrinsic.md: Likewise.
17971         * config/nds32/nds32_intrinsic.h: Likewise.
17972         * config/nds32/nds32.h (nds32_builtins): Likewise.
17973         * config/nds32/nds32.md (type): Add pbsad and pbsada.
17974         (btst, ave): New patterns for particular instructions.
17976 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
17978         * config/nds32/constants.md (unspec_element, unspec_volatile_element):
17979         Add enum values for atomic load/store and memory sync.
17980         * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
17981         and memory sync.
17982         * config/nds32/nds32-intrinsic.md: Likewise.
17983         * config/nds32/nds32_intrinsic.h: Likewise.
17984         * config/nds32/nds32.h (nds32_builtins): Likewise.
17986 2018-04-07  Jakub Jelinek  <jakub@redhat.com>
17988         PR tree-optimization/85257
17989         * fold-const.c (native_encode_vector): If not all elts could fit
17990         and off is -1, return 0 rather than offset.
17991         * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
17992         (offseti - offset2) / BITS_PER_UNIT as 4th argument to
17993         native_encode_expr.  Verify len * BITS_PER_UNIT >= maxsizei.  Don't
17994         adjust buffer in native_interpret_expr call.
17996 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
17998         * config/nds32/constants.md (unspec_volatile_element): Add cache
17999         control enum values.
18000         * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
18001         * config/nds32/nds32-intrinsic.md: Add cache control patterns.
18002         * config/nds32/nds32.c (nds32_cctl_names): New.
18003         (nds32_print_operand): Handle cache control register names.
18004         * config/nds32/nds32.h (nds32_builtins): New enum values.
18005         * config/nds32/nds32_intrinsic.h: Add cache control enum types and
18006         macros.
18007         * config/nds32/nds32.md (type): Add mmu.
18008         * config/nds32/pipelines.md (simple_insn): Add mmu.
18010 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
18012         * config/nds32/nds32.md (type): Remove call.
18013         * config/nds32/pipelines.md (simple_insn): Likewise.
18015 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
18017         * config/nds32/constants.md (unspec_volatile_element): Add
18018         UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
18019         UNSPEC_VOLATILE_FMFCFG.
18020         * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
18021         description for fmfcfg and fmfcsr.
18022         (bdesc_1arg): Add fmtcsr.
18023         (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
18024         (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
18025         * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
18026         unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
18027         unspec_fmfcfg): New patterns.
18028         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
18029         NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
18030         NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
18031         * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
18032         __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
18033         __nds32__fmfcfg): Define.
18035 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
18037         * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
18038         intrinsic register names.
18039         * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
18040         intrinsic register enum values and macros.
18042 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
18044         * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
18045         for load/store addressing form.
18046         (nds32_print_operand_address): Likewise.
18048 2018-04-06  Eric Botcazou  <ebotcazou@adacore.com>
18050         PR target/85196
18051         * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
18052         based on LABEL_REF.  Remove useless assertion.
18053         (pic_address_needs_scratch): Fix formatting.
18054         (sparc_legitimize_pic_address): Minor tweaks.
18055         (sparc_delegitimize_address): Adjust assertion accordingly.
18056         * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
18057         into symbolic_operand.
18058         (movsi_high_pic_label_ref): Likewise.
18059         (movsi_lo_sum_pic_label_ref): Likewise.
18060         (movdi_pic_label_ref): Likewise.
18061         (movdi_high_pic_label_ref): Likewise.
18062         (movdi_lo_sum_pic_label_ref): Likewise.
18064 2018-04-06  Amaan Cheval  <amaan.cheval@gmail.com>
18066         * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
18067         custom LIB_SPEC setup.
18069 2018-04-06  Ruslan Bukin  <br@bsdpad.com>
18070             Kito Cheng  <kito.cheng@gmail.com>
18072         * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
18073         * config/riscv/freebsd.h: New.
18075 2018-04-06  Chung-Ju Wu  <jasonwucj@gmail.com>
18077         * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
18078         * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
18079         file.
18081 2018-04-06  Chung-Ju Wu  <jasonwucj@gmail.com>
18082             Kito Cheng  <kito.cheng@gmail.com>
18084         * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
18085         nds32_output_call, nds32_symbol_binds_local_p): New functions.
18086         * config/nds32/nds32-protos.h (nds32_output_call,
18087         nds32_output_return): Declare.
18088         * config/nds32/nds32.md: Refine all the call and return patterns.
18090 2018-04-06  Jakub Jelinek  <jakub@redhat.com>
18092         PR debug/85252
18093         * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
18094         build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
18096         PR rtl-optimization/84872
18097         * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
18098         nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
18099         EDGE_CROSSING edge.
18101 2018-04-06  Tamar Christina  <tamar.christina@arm.com>
18103         * expr.c (copy_blkmode_to_reg): Revert 254862.
18104         * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
18106 2018-04-06  Richard Biener  <rguenther@suse.de>
18108         PR middle-end/85244
18109         * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
18110         after seeing a component reference with an adjacent field.  Treat
18111         refs to arrays at struct end of external decls similar to
18112         refs to unconstrained commons.
18114 2018-04-06  Jakub Jelinek  <jakub@redhat.com>
18116         PR sanitizer/85213
18117         * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
18118         look through SAVE_EXPRs with non-side-effects argument.  Adjust
18119         recursive calls.
18120         (fold_comparison): Adjust twoval_comparison_p caller, don't handle
18121         save_p here.
18123 2018-04-06  Richard Biener  <rguenther@suse.de>
18125         PR middle-end/85180
18126         * alias.c (find_base_term): New wrapper around find_base_term
18127         unwinding CSELIB_VAL_PTR changes.
18128         (find_base_term): Do not restore CSELIB_VAL_PTR during the
18129         recursion.
18131 2018-04-06  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
18133         * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
18134         instructions.
18135         * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
18136         constant definitions.
18137         ("nop"): lr 0,0 -> nopr r0
18138         ("nop_lr0", "nop_lr1"): New insn definitions.
18140 2018-04-06  Chung-Ju Wu  <jasonwucj@gmail.com>
18142         * config/nds32/nds32.md (*stack_push, *stack_pop): Use
18143         NDS32_V3PUSH_AVAILABLE_P macro.
18145 2018-04-06  Monk Chiang  <sh.chiang04@gmail.com>
18146             Chung-Ju Wu  <jasonwucj@gmail.com>
18148         * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
18149         (nds32*-*-*): Add float and fpu_config into supported_defaults.
18150         * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
18151         Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
18152         * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
18153         UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
18154         * config/nds32/constraints.md: New constraints and checking for hard
18155         float configuration.
18156         * config/nds32/iterators.md: New mode iterator and attribute for hard
18157         float configuration.
18158         * config/nds32/nds32-doubleword.md: Use hard float alternatives and
18159         patterns.
18160         * config/nds32/nds32-fpu.md: New file.
18161         * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
18162         deal with hard float code generation.
18163         * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
18164         ARCH_V3S.
18165         (abi_type, float_reg_number): New enum type.
18166         * config/nds32/nds32-predicates.c: New predicates for hard float.
18167         * config/nds32/nds32-protos.h: Declare functions for hard float.
18168         * config/nds32/nds32.c: Implementation for hard float configuration.
18169         * config/nds32/nds32.h: Definitions for hard float configuration.
18170         * config/nds32/nds32.md: Include hard float machine description and
18171         modify patterns for hard float configuration.
18172         * config/nds32/nds32.opt: New options for hard float configuration.
18173         * config/nds32/predicates.md: New predicates for hard float
18174         configuration.
18176 2018-04-06  Kuan-Lin Chen  <kuanlinchentw@gmail.com>
18178         * common/config/nds32/nds32-common.c
18179         (nds32_option_optimization_table): Enable -mreleax-hint by default.
18181 2018-04-05  Jakub Jelinek  <jakub@redhat.com>
18183         PR middle-end/85195
18184         * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
18185         CONSTRUCTOR_ELT (ctor, ...)->value.
18187 2018-04-05  Uros Bizjak  <ubizjak@gmail.com>
18189         PR target/85193
18190         * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
18192 2018-04-05  Tom de Vries  <tom@codesourcery.com>
18194         PR target/85204
18195         * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
18196         cond jump.
18198 2018-04-05  Shiva Chen  <shiva0217@gmail.com>
18199             Kito Cheng  <kito.cheng@gmail.com>
18201         * config/nds32/constraints.md (U33): Fine-tune checking condition.
18202         * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
18203         * config/nds32/nds32.h (nds32_16bit_address_type): Add
18204         ADDRESS_POST_MODIFY_LO_REG_IMM3U.
18206 2018-04-05  Shiva Chen  <shiva0217@gmail.com>
18207             Kito Cheng  <kito.cheng@gmail.com>
18209         * config/nds32/constraints.md (Ufe): New memory constraint.
18210         * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
18211         nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
18212         * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
18213         operands.
18214         * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
18215         * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
18217 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
18219         * config/nds32/nds32.md: Use optimize_size in the condition for
18220         alu-shift instructions.
18222 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
18224         * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
18226 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
18228         * config/nds32/nds32.md (negsi2): Refine pattern.
18230 2018-04-05  Kito Cheng  <kito.cheng@gmail.com>
18231             Chung-Ju Wu  <jasonwucj@gmail.com>
18233         * config/nds32/iterators.md (shift_rotate): New code iterator.
18234         (shift): New code attribute.
18235         * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
18236         * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
18237         * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
18238         * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
18239         bit-wise operations.
18240         (andsi3, *andsi3): Ditto.
18241         (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
18242         (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
18243         (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
18244         * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
18245         nds32_ior_operand, nds32_xor_operand): New predicates.
18247 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
18249         * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
18250         (addsi3, subsi3): ... this.
18252 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
18254         * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
18256 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
18258         * config/nds32/nds32.md: Adjust indention.
18260 2018-04-05  Kito Cheng  <kito.cheng@gmail.com>
18262         * config/nds32/nds32.md (feature): New attribute.
18264 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
18266         * config/nds32/nds32.md (subtype): New attribute.
18268 2018-04-04  Thomas Preud'homme  <thomas.preudhomme@arm.com>
18270         PR target/85203
18271         * config/arm/arm-builtins.c (arm_expand_builtin): Change
18272         expansion to perform a bitwise AND of the argument followed by a
18273         boolean negation of the result.
18275 2018-04-04  Peter Bergner  <bergner@vnet.ibm.com>
18277         PR rtl-optimization/84878
18278         * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
18279         the basic block.  Assert the use reference is not artificial and that
18280         it has an associated insn.
18282 2018-04-04  Michael Matz  <matz@suse.de>
18284         * builtins.c (compute_objsize): Pass correct operand
18285         to array_at_struct_end_p.
18287 2018-04-04  Richard Biener  <rguenther@suse.de>
18289         PR lto/85176
18290         * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
18291         from contexts for DINFO_LEVEL_TERSE and below.
18293 2018-04-04  Kito Cheng  <kito.cheng@gmail.com>
18295         * config/nds32/nds32-doubleword.md (move_<mode>): Require
18296         resiter_operand condition.
18297         * config/nds32/nds32.md (*move<mode>): Ditto.
18299 2018-04-04  Kito Cheng  <kito.cheng@gmail.com>
18300             Monk Chiang  <sh.chiang04@gmail.com>
18302         * config/nds32/nds32.md (movmisalign<mode>): New pattern.
18304 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
18306         * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
18308 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
18309             Kito Cheng  <kito.cheng@gmail.com>
18311         * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
18312         nds32_cond_code_str, output_cond_branch,
18313         output_cond_branch_compare_zero, nds32_expand_cbranch,
18314         nds32_expand_cstore, nds32_expand_movcc,
18315         nds32_output_cbranchsi4_equality_zero,
18316         nds32_output_cbranchsi4_equality_reg,
18317         nds32_output_cbranchsi4_equality_reg_or_const_int,
18318         nds32_output_cbranchsi4_greater_less_zero: New functions.
18319         * config/nds32/nds32-protos.h (nds32_expand_cbranch,
18320         nds32_expand_cstore, nds32_expand_movcc,
18321         nds32_output_cbranchsi4_equality_zero,
18322         nds32_output_cbranchsi4_equality_reg,
18323         nds32_output_cbranchsi4_equality_reg_or_const_int,
18324         nds32_output_cbranchsi4_greater_less_zero): Declare.
18325         * config/nds32/predicates.md (nds32_movecc_comparison_operator,
18326         nds32_rimm11s_operand): New predicates.
18327         * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
18328         * config/nds32/nds32.md: Rewrite all the branch and conditional move
18329         patterns.
18331 2018-04-04  Kito Cheng  <kito.cheng@gmail.com>
18333         * config/nds32/nds32-doubleword.md: Refine all the instruction type.
18334         * config/nds32/nds32.md: Ditto.
18335         * config/nds32/pipelines.md: Ditto.
18337 2018-04-04  Richard Biener  <rguenther@suse.de>
18339         PR tree-optimization/85168
18340         * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
18341         propagating abnormals.
18343 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
18345         * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
18347 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
18348             Kito Cheng  <kito.cheng@gmail.com>
18350         * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
18351         * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
18352         * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
18353         (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
18354         * config/nds32/nds32.md (sibcall_internal): New.
18355         (sibcall_register): Remove.
18356         (sibcall_immediate): Remove.
18357         (sibcall_value_internal): New.
18358         (sibcall_value_register): Remove.
18359         (sibcall_value_immediate): Remove.
18360         * config/nds32/predicates.md (nds32_general_register_operand): New.
18361         (nds32_call_address_operand): New.
18363 2018-04-03  Jakub Jelinek  <jakub@redhat.com>
18365         PR rtl-optimization/85167
18366         * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
18367         bb_defs if *split_p, instead preinitialize it to NULL.
18369         PR tree-optimization/85156
18370         * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
18371         evaluating the argument multiple times.
18373 2018-04-03  Bill Schmidt  <wschmidt@linux.ibm.com>
18375         * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
18376         than vector.
18377         (_mm_cvtpd_ps): Likewise.
18378         (_mm_cvttpd_epi32): Likewise.
18379         * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
18380         * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
18381         vector, pixel, and bool following altivec.h include.
18383 2018-04-03  Martin Sebor  <msebor@redhat.com>
18385         * doc/extend.texi (Common Function Attributes): Clarify.
18386         (const attribute): Likewise.
18387         (pure attribute): Likewise.
18389 2018-04-03  Jakub Jelinek  <jakub@redhat.com>
18391         PR target/85169
18392         * config/i386/i386.c (ix86_expand_vector_set): Use
18393         HOST_WIDE_INT_1U << elt instead of 1 << elt.  Formatting fix.
18395 2018-04-03  Uros Bizjak  <ubizjak@gmail.com>
18397         * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
18398         instructions when changing rounding bits to preserve precision bits
18399         in the x87 control word.
18401 2018-04-03  Martin Liska  <mliska@suse.cz>
18403         PR tree-optimization/82491
18404         * rtl.h (strip_offset_and_add): Replace += suboffset with
18405         poly_uint64 () + suboffset.
18407 2018-03-29  Martin Liska  <mliska@suse.cz>
18408             Martin Jambor  <mjambor@suse.cz>
18410         PR ipa/84947
18411         * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
18412         param_type is not an integral or pointer type.
18414 2018-04-03  Richard Biener  <rguenther@suse.de>
18416         * sese.h (recompute_all_dominators): Remove.
18418 2018-04-02  Martin Sebor  <msebor@redhat.com>
18420         * doc/invoke.texi (-Wrestrict): Fix typos.
18422 2018-04-02  Jim Wilson  <jimw@sifive.com>
18424         * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
18425         * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
18426         (<optab>di3, <optab>si3_extend): Likewise.
18427         (<optab>si3_mask, <optab>si3_mask_1): New.
18428         (<optab>di3_mask, <optab>di3_mask_1): New.
18429         (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
18430         (lshrsi3_zero_extend_1): Use VOIDmode shift count.
18431         * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
18433 2018-04-02  Gerald Pfeifer  <gerald@pfeifer.com>
18435         * doc/cpp.texi (Variadic Macros): Fix line continuation in an
18436         example.
18438 2018-04-02  Chung-Ju Wu  <jasonwucj@gmail.com>
18440         * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
18441         (nds32_canonicalize_comparison): New function.
18443 2018-04-01  Chung-Ju Wu  <jasonwucj@gmail.com>
18444             Kito Cheng  <kito.cheng@gmail.com>
18445             Kuan-Lin Chen  <kuanlinchentw@gmail.com>
18447         * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
18448         * config/nds32/constants.md (unspec_volatile_element): Add
18449         UNSPEC_VOLATILE_RELAX_GROUP.
18450         * config/nds32/nds32-relax-opt.c: New file.
18451         * config/nds32/nds32-predicates.c
18452         (nds32_symbol_load_store_p): New function.
18453         * config/nds32/nds32-protos.h
18454         (nds32_symbol_load_store_p): Declare function.
18455         (make_pass_nds32_relax_opt): Declare new rtl pass function.
18456         * config/nds32/nds32.c
18457         (nds32_register_pass): New function to register pass.
18458         (nds32_register_passes): New function to register passes.
18459         * config/nds32/nds32.md (relax_group): New pattern.
18460         * config/nds32/nds32.opt (mrelax-hint): New option.
18461         * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
18463 2018-04-01  Kito Cheng  <kito.cheng@gmail.com>
18465         * config/nds32/t-nds32: Modify files dependency.
18467 2018-04-01  Chung-Ju Wu  <jasonwucj@gmail.com>
18469         * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
18470         (PROFILE_HOOK): Define its implementation.
18472 2018-04-01  Chung-Ju Wu  <jasonwucj@gmail.com>
18474         * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
18475         type and 32-bit size.
18477 2018-04-01  Jakub Jelinek  <jakub@redhat.com>
18479         PR middle-end/85090
18480         * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
18481         (V_128_256): New mode iterator.
18482         (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
18483         (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
18484         (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
18485         of V.
18486         * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
18487         V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
18489 2018-03-31  Segher Boessenkool  <segher@kernel.crashing.org>
18491         PR target/83315
18492         * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
18493         NaN inputs correctly.
18495 2018-03-30  Peter Bergner  <bergner@vnet.ibm.com>
18497         PR target/80546
18498         * config/rs6000/vsx.md (??r): New mode attribute.
18499         (*vsx_mov<mode>_64bit): Use it.
18500         (*vsx_mov<mode>_32bit): Likewise.
18502 2018-03-30  Martin Sebor  <msebor@redhat.com>
18504         PR tree-optimization/84818
18505         * builtins.c (check_access): Use warning_n.
18507 2018-03-30  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
18509         PR target/83822
18510         * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
18511         condition.
18512         * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
18513         condition.
18515 2018-03-30  Julia Koval  <julia.koval@intel.com>
18517         PR target/84413
18518         * x86-tune.def (movx, partial_reg_dependency): Enable for
18519         m_SKYLAKE_AVX512.
18521 2018-03-29  Vladimir Makarov  <vmakarov@redhat.com>
18523         PR inline-asm/84985
18524         * lra-constraints.c (process_alt_operands): Move setting
18525         this_alternative_matches below.
18527 2018-03-29  Martin Liska  <mliska@suse.cz>
18529         PR lto/84995.
18530         * doc/invoke.texi: Document how LTO works with debug info.
18531         Describe auto-load support of binutils.  Mention 'x86-64'
18532         as valid option value of -march option.
18534 2018-03-29  Jakub Jelinek  <jakub@redhat.com>
18536         * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
18538         PR c/85094
18539         * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
18540         For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
18541         OEP_NO_HASH_CHECK for recursive call, to avoid exponential
18542         checking.
18544 2018-03-28  Peter Bergner  <bergner@vnet.ibm.com>
18546         PR target/84912
18547         * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
18548         (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
18549         * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
18550         for RS6000_BTM_POWERPC64.
18551         (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
18552         (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
18553         * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
18554         definition.
18555         (DIVDE): Use it.
18556         (DIVDEU): Likewise.
18558 2018-03-28  Carl Love  <cel@us.ibm.com>
18560         Revert
18561         2017-09-27  Carl Love  <cel@us.ibm.com>
18563         * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
18564         (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
18565         * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
18566         fctiw instruction.
18568 2018-03-28  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
18570         * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
18571         instead of __vector bool.
18572         (_mm_max_pu8): Likewise.
18573         (_mm_min_pi16): Likewise.
18575 2018-03-28  Peter Bergner  <bergner@vnet.ibm.com>
18577         PR target/84912
18578         * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
18579         (DIVWEUO): Likewise.
18580         (DIVDEO): Likewise.
18581         (DIVDEUO): Likewise.
18582         * config/rs6000/rs6000.c (builtin_function_type): Remove support for
18583         DIVWEUO and DIVDEUO.
18584         * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
18585         (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
18586         (div_extend): Likewise.
18587         * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
18588         builtin function.
18589         (__builtin_divweuo): Likewise.
18590         (__builtin_divdeo): Likewise.
18591         (__builtin_divdeuo): Likewise.
18593 2018-03-28  Jakub Jelinek  <jakub@redhat.com>
18595         PR target/85095
18596         * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
18597         *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
18599         PR tree-optimization/82004
18600         * gimple-match-head.c (optimize_pow_to_exp): New function.
18601         * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
18602         Don't fold to exp if optimize_pow_to_exp is false.
18604 2018-03-28  Martin Liska  <mliska@suse.cz>
18606         PR other/84819
18607         * calls.c (initialize_argument_information): Fix trailing space.
18608         * common.opt: Fix typo and provide better explanation for
18609         -fsanitize-coverage option.
18610         * config/i386/i386.opt: Fix typo.
18612 2018-03-28  Jakub Jelinek  <jakub@redhat.com>
18613             Martin Liska  <mliska@suse.cz>
18615         PR sanitizer/85081
18616         * gimplify.c (asan_poison_variable): Don't do the check for
18617         gimplify_omp_ctxp here.
18618         (gimplify_decl_expr): Do it here.
18619         (gimplify_target_expr): Likewise.
18621 2018-03-28  Martin Liska  <mliska@suse.cz>
18623         PR target/84988
18624         * config/i386/i386.c (ix86_function_arg_advance): Do not call
18625         chkp_type_bounds_count if MPX is not enabled.
18627 2018-03-27  Chung-Ju Wu  <jasonwucj@gmail.com>
18629         * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
18631 2018-03-27  Michael Meissner  <meissner@linux.vnet.ibm.com>
18633         PR target/84914
18634         * config/rs6000/rs6000.c (create_complex_muldiv): New helper
18635         function to create the function decl for complex long double
18636         multiply and divide for -mabi=ieeelongdouble.
18637         (init_float128_ieee): Call it.
18639 2018-03-27  H.J. Lu  <hongjiu.lu@intel.com>
18641         PR target/85044
18642         * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
18643         -fcf-protection=branch -mibt.
18644         * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
18646 2018-03-27  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
18648         PR target/81863
18649         * config/arm/arm.c (arm_valid_symbolic_address): Handle
18650         arm_word_relocations.
18652 2018-03-27  Cesar Philippidis  <cesar@codesourcery.com>
18654         PR target/85056
18655         * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
18656         extern array declarations.
18658 2018-03-27  Richard Biener  <rguenther@suse.de>
18660         PR middle-end/84067
18661         * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
18662         explicit single_use checks.
18664 2018-03-27  Richard Biener  <rguenther@suse.de>
18666         PR tree-optimization/85082
18667         * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
18668         Valueize the VUSE.
18670 2018-03-27  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
18672         * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
18673         * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
18674         Turn on fasynchronous-unwind-tables and funwind-tables.
18676 2018-03-26  Uros Bizjak  <ubizjak@gmail.com>
18678         PR target/85073
18679         * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
18680         (*bmi_blsr_<mode>_ccz): Ditto.
18682 2018-03-26  Tom de Vries  <tom@codesourcery.com>
18684         PR tree-optimization/85063
18685         * omp-general.c (offloading_function_p): New function.  Factor out
18686         of ...
18687         * omp-offload.c (pass_omp_target_link::gate): ... here.
18688         * omp-general.h (offloading_function_p): Declare.
18689         * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
18690         with attribute omp declare target for offloading functions.
18692 2018-03-24  Richard Sandiford  <richard.sandiford@linaro.org>
18694         PR tree-optimization/84005
18695         * tree-data-ref.h (get_base_for_alignment): Declare.
18696         * tree-data-ref.c (get_base_for_alignment_1): New function.
18697         (get_base_for_alignment): Likewise.
18698         * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
18699         get_base_for_alignment to find a suitable base object, instead
18700         of always using drb->base_address.
18702 2018-03-23  Jakub Jelinek  <jakub@redhat.com>
18704         PR inline-asm/85022
18705         * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
18706         known size by default.
18708 2018-03-23  Vladimir Makarov  <vmakarov@redhat.com>
18710         PR inline-asm/85030
18711         * lra-constraints.c (process_alt_operands): Don't match BLKmode
18712         and non BLKmode operands.
18714 2018-03-23  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
18716         PR target/85026
18717         * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
18718         Clean up attributes.
18720 2018-03-23  Richard Biener  <rguenther@suse.de>
18722         PR debug/85020
18723         * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
18724         we are going to emit early debug for LTO.
18726 2018-03-23  Jakub Jelinek  <jakub@redhat.com>
18728         PR inline-asm/85034
18729         * function.c (match_asm_constraints_1): Don't optimize if input
18730         doesn't satisfy general_operand predicate for output's mode.
18732         PR inline-asm/85022
18733         * alias.c (write_dependence_p): Don't require for x_canonicalized
18734         non-VOIDmode if x has VOIDmode.
18736         PR sanitizer/85029
18737         * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
18738         just don't try to optimize it rather than assert it never happens.
18740 2018-03-22  Kelvin Nilsen  <kelvin@gcc.gnu.org>
18742         * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
18743         macro expansions for definition of ST_INTERNAL_<mode> and
18744         LD_INTERNAL_<mode> builtins.
18745         * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
18746         Remove prototype.
18747         * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
18748         function.
18749         (altivec_expand_st_builtin): Likewise.
18750         (altivec_expand_builtin): Remove calls to deleted functions.
18751         (rs6000_address_for_altivec): Delete this function.
18752         * config/rs6000/vector.md: Remove expands for
18753         vector_altivec_load_<mode> and vector_altivec_store_<mode>.
18755 2018-03-22  Sudakshina Das  <sudi.das@arm.com>
18757         PR target/84826
18758         * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
18759         * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
18760         re-computing once computed.
18761         (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
18762         (arm_init_machine_status): Initialize
18763         machine->static_chain_stack_bytes.
18765 2018-03-22  Kelvin Nilsen  <kelvin@gcc.gnu.org>
18767         PR target/84760
18768         * doc/extend.texi: Add four new prototypes for vec_ld.
18769         * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
18770         definitions for more logical presentation.
18771         * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
18772         entries for V1TI variants of __builtin_altivec_ld builtin.
18773         * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
18774         handling of V1TI variant of LVX icode pattern.
18775         (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
18776         (rs6000_gimple_fold_builtin): Likewise.
18777         (altivec_init_builtins): Add code to define
18778         __builtin_altivec_lvx_v1ti function.
18780 2018-03-22  Jakub Jelinek  <jakub@redhat.com>
18782         PR inline-asm/84941
18783         * function.c (match_asm_constraints_1): Don't do the optimization
18784         if input isn't a REG, SUBREG, MEM or constant.
18786 2018-03-22  Tom de Vries  <tom@codesourcery.com>
18788         PR tree-optimization/84956
18789         * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
18790         bb_has_abnormal_pred.
18792 2018-03-22  Jakub Jelinek  <jakub@redhat.com>
18794         PR sanitizer/85018
18795         * dwarf2asm.c (dw2_output_indirect_constant_1): Set
18796         DECL_INITIAL (decl) to decl at the end.
18797         * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
18798         adjust the comment.
18800 2018-03-21  Joseph Myers  <joseph@codesourcery.com>
18802         * doc/extend.texi (__builtin_tgmath): Document when complex
18803         integer types are treated as _Complex _Float64.
18805 2018-03-21  Tom de Vries  <tom@codesourcery.com>
18807         * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
18809 2018-03-21  Jakub Jelinek  <jakub@redhat.com>
18811         PR tree-optimization/84960
18812         * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
18813         if it is ENTRY block, move them into single succ of ENTRY in that case.
18815 2018-03-21  Richard Sandiford  <richard.sandiford@linaro.org>
18817         PR tree-optimization/84811
18818         * poly-int.h (poly_span_traits): Remove the T3 parameter and
18819         promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
18820         (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
18821         (known_subrange_p): Update accordingly.  Cast each value involved
18822         in the size comparison, rather than casting the result of the
18823         subtraction.
18825 2018-03-21  Jakub Jelinek  <jakub@redhat.com>
18827         PR tree-optimization/84982
18828         * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
18829         by flipping the least significant bit rather than all bits from
18830         bitpos to bitpos + bitsize - 1.
18832 2018-03-21  Nathan Sidwell  <nathan@acm.org>
18834         * doc/extend.texi (Deprecated Features): Remove mention of
18835         long-deleted deprecations.
18837 2018-03-21  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
18839         PR jit/84288
18840         * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
18841         * configure: Regenerate.
18843 2018-03-21  Tom de Vries  <tom@codesourcery.com>
18845         PR tree-optimization/83126
18846         * tree-parloops.c (num_phis): New function.
18847         (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
18849 2018-03-21  Nathan Sidwell  <nathan@acm.org>
18851         * doc/extend.texi (Deprecated Features): Update deprecated flags,
18852         mention anon-struct/union members and trailing attributes.
18854 2018-03-21  Bin Cheng  <bin.cheng@arm.com>
18856         PR tree-optimization/84969
18857         * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
18858         builtin memset partitions if they set different rhs values.
18860 2018-03-21  Jakub Jelinek  <jakub@redhat.com>
18862         PR rtl-optimization/84989
18863         * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
18864         VEC_DUPLICATE with scalar result mode.
18866 2018-03-21  Martin Liska  <mliska@suse.cz>
18868         PR ipa/84963
18869         * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
18870         not intended return statement.
18872 2018-03-21  Martin Liska  <mliska@suse.cz>
18874         PR target/84988
18875         * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
18876         (chkp_find_bound_slots_1): Limit number of iterations.
18878 2018-03-20  David H. Gutteridge  <dhgutteridge@sympatico.ca>
18880         PR target/84838
18881         * Minor grammar fixes for x86 options.
18883 2018-03-20  Jakub Jelinek  <jakub@redhat.com>
18885         PR debug/84875
18886         * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
18887         holding REG_CFA_RESTORE notes, instead turn them into a USE.
18889 2018-03-20  Peter Bergner  <bergner@vnet.ibm.com>
18891         PR target/83789
18892         * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
18893         (altivec_lvx_<mode>_1op): Likewise.
18894         (altivec_stvx_<mode>_2op): Likewise.
18895         (altivec_stvx_<mode>_1op): Likewise.
18896         (altivec_lvx_<VM2:mode>): New define_expand.
18897         (altivec_stvx_<VM2:mode>): Likewise.
18898         (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
18899         (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
18900         (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
18901         (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
18902         * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
18903         (rs6000_gen_lvx): Likewise.
18904         * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
18905         (altivec_expand_stv_builtin): Likewise.
18906         (altivec_expand_builtin): Likewise.
18907         * config/rs6000/vector.md: Likewise.
18909 2018-03-20  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
18911         PR target/82518
18912         * config/arm/arm.c (arm_array_mode_supported_p): Return false for
18913         BYTES_BIG_ENDIAN.
18915 2018-03-20  Richard Biener  <rguenther@suse.de>
18917         PR target/84986
18918         * config/i386/i386.c (ix86_add_stmt_cost): Only cost
18919         sign-conversions as zero, fall back to standard scalar_stmt
18920         cost for the rest.
18922 2018-03-20  Martin Liska  <mliska@suse.cz>
18924         PR ipa/84825
18925         * predict.c (rebuild_frequencies): Handle case when we have
18926         PROFILE_ABSENT, but flag_guess_branch_prob is false.
18928 2018-03-20  Jakub Jelinek  <jakub@redhat.com>
18930         PR target/84990
18931         * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
18932         flag_section_anchors.
18933         * varasm.c (use_blocks_for_decl_p): Remove hack for
18934         dw2_force_const_mem.
18936         PR target/84845
18937         * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
18938         to ...
18939         (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this.  If pseudos can't
18940         be created, use lowpart_subreg of operands[0] rather than operands[0]
18941         itself.
18942         (*aarch64_reg_<mode>3_minus_mask): Rename to ...
18943         (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
18944         (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
18945         and n constraint instead of aarch64_shift_imm_di and Usd.
18946         (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
18947         (*aarch64_<optab>_reg_minus<mode>3): ... this.
18949 2018-03-20  Sudakshina Das  <sudi.das@arm.com>
18951         PR target/82989
18952         * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
18953         to favor GPR over NEON registers.
18954         (<shift>di3_neon): Likewise.
18956 2018-03-20  Tom de Vries  <tom@codesourcery.com>
18958         PR target/84952
18959         * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
18960         (nvptx_process_pars): Emit bar.sync asap and alap.
18962 2018-03-20  Tom de Vries  <tom@codesourcery.com>
18964         PR target/84954
18965         * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
18966         seen_label if seen_label is already set.
18968 2018-03-20  Jakub Jelinek  <jakub@redhat.com>
18970         PR target/84945
18971         * config/i386/i386.c (fold_builtin_cpu): For features above 31
18972         use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
18973         Use 1U instead of 1.  Formatting fixes.
18975         PR c/84953
18976         * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
18977         instead of TREE_TYPE (s1) for the return value.
18979 2018-03-19  Jakub Jelinek  <jakub@redhat.com>
18981         PR tree-optimization/84946
18982         * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
18983         bitsize + bitsize in poly_uint64 rather than poly_int64.
18985         PR sanitizer/78651
18986         * dwarf2asm.c: Include fold-const.c.
18987         (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
18988         of decl rather than decl itself.
18990         PR rtl-optimization/84643
18991         * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
18993 2018-03-19  Maxim Ostapenko  <m.ostapenko@samsung.com>
18995         PR sanitizer/78651
18996         * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
18997         calling assemble_variable.
18999 2018-03-19  Sudakshina Das  <sudi.das@arm.com>
19001         PR target/81647
19002         * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
19003         instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
19005 2018-03-19  Jim Wilson  <jimw@sifive.com>
19007         PR bootstrap/84856
19008         * config/riscv/riscv.c (riscv_function_arg_boundary): Use
19009         PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
19010         (riscv_first_stack_step): Likewise.
19011         (riscv_option_override): Use STACK_BOUNDARY instead of
19012         MIN_STACK_BOUNDARY.
19013         * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
19014         MIN_STACK_BOUNDARY.
19015         (BIGGEST_ALIGNMENT): Set to 128.
19016         (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
19017         (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
19018         STACK_BOUNDARY.
19020 2018-03-19  Richard Biener  <rguenther@suse.de>
19022         PR tree-optimization/84933
19023         * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
19024         values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
19026 2018-03-19  Richard Biener  <rguenther@suse.de>
19028         PR tree-optimization/84859
19029         * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
19030         (cond_if_else_store_replacement): Perform sinking operation on
19031         single-store BBs regardless of MAX_STORES_TO_SINK setting.
19032         Generalize what a BB with a single eligible store is.
19034 2018-03-19  Richard Biener  <rguenther@suse.de>
19036         PR tree-optimization/84929
19037         * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
19038         chrec_is_positive against non-chrec arg.
19040 2018-03-19  Tamar Christina  <tamar.christina@arm.com>
19042         PR target/84711
19043         * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
19045 2018-03-18  Martin Liska  <mliska@suse.cz>
19047         PR rtl-optimization/84635
19048         * regrename.c (build_def_use): Use matches_mode only when
19049         matches >= 0.
19051 2018-03-18  Richard Sandiford  <richard.sandiford@linaro.org>
19053         PR tree-optimization/84913
19054         * tree-vect-loop.c (vectorizable_reduction): Don't try to
19055         vectorize chains of COND_EXPRs.
19057 2018-03-18  Chung-Ju Wu  <jasonwucj@gmail.com>
19059         * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
19061 2018-03-18  Chung-Ju Wu  <jasonwucj@gmail.com>
19063         * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
19065 2018-03-18  Chung-Ju Wu  <jasonwucj@gmail.com>
19067         * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
19069 2018-03-17  Chung-Ju Wu  <jasonwucj@gmail.com>
19070             Kito Cheng  <kito.cheng@gmail.com>
19072         * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
19073         * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
19074         (nds32_adjust_reg_alloc_order): New function.
19075         * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
19077 2018-03-17  Kito Cheng  <kito.cheng@gmail.com>
19079         * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
19080         nds32_print_operand, nds32_print_operand_address): Use
19081         HOST_WIDE_INT_PRINT_DEC instead.
19083 2018-03-17  Chung-Ju Wu  <jasonwucj@gmail.com>
19085         * config/nds32/nds32.c (nds32_register_priority): Modify cost.
19087 2018-03-17  Jakub Jelinek  <jakub@redhat.com>
19089         PR target/84902
19090         * config/i386/i386.c (initial_ix86_tune_features,
19091         initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
19092         unsigned long long.
19093         (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
19094         to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
19095         rather than 1u << ix86_tune.  Formatting fix.
19096         (ix86_option_override_internal): Change ix86_arch_mask from
19097         unsigned int to unsigned HOST_WIDE_INT, initialize to
19098         HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
19099         (ix86_function_specific_restore): Likewise.
19101 2018-03-16  Jakub Jelinek  <jakub@redhat.com>
19103         PR target/84899
19104         * postreload.c (reload_combine_recognize_pattern): Perform
19105         INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
19106         truncate_int_for_mode the result for the destination's mode.
19108         PR c/84909
19109         * hsa-gen.c (mem_type_for_type): Fix comment typo.
19110         * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
19111         Likewise.
19112         * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
19113         Likewise.
19115 2018-03-16  Vladimir Makarov  <vmakarov@redhat.com>
19117         PR target/84876
19118         * lra-assigns.c (lra_split_hard_reg_for): Don't use
19119         regno_allocno_class_array and sorted_pseudos.
19120         * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
19121         insns where regno is used.
19123 2018-03-16  Martin Liska  <mliska@suse.cz>
19125         PR ipa/84833
19126         * multiple_target.c (create_dispatcher_calls): Redirect
19127         reference in the symbol table.
19129 2018-03-16  Martin Liska  <mliska@suse.cz>
19131         PR ipa/84722
19132         * multiple_target.c (create_dispatcher_calls): Redirect also
19133         an alias.
19135 2018-03-16  Jakub Jelinek  <jakub@redhat.com>
19137         PR c++/79937
19138         PR c++/82410
19139         * tree.h (TARGET_EXPR_NO_ELIDE): Define.
19140         * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
19141         TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
19143 2018-03-16  Julia Koval  <julia.koval@intel.com>
19145         * doc/invoke.texi (Skylake Server): Add CLWB.
19146         Cannonlake): Remove CLWB.
19148 2018-03-16  Jakub Jelinek  <jakub@redhat.com>
19150         PR tree-optimization/84841
19151         * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
19152         1 << 3.
19153         (FLOAT_ONE_CONST_TYPE): Define.
19154         (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
19155         (sort_by_operand_rank): Put entries with higher constant_type last
19156         rather than first to match comments.
19158 2018-03-15  Sandra Loosemore  <sandra@codesourcery.com>
19160         * config/nios2/nios2.md (movsi_internal): Fix thinko in
19161         split predicate.
19163 2018-03-15  Jakub Jelinek  <jakub@redhat.com>
19165         PR c++/79085
19166         * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
19167         check and use address of target always.
19169 2018-03-15  H.J. Lu  <hongjiu.lu@intel.com>
19171         PR target/84574
19172         * config/i386/i386.c (indirect_thunk_needed): Update comments.
19173         (indirect_thunk_bnd_needed): Likewise.
19174         (indirect_thunks_used): Likewise.
19175         (indirect_thunks_bnd_used): Likewise.
19176         (indirect_return_needed): New.
19177         (indirect_return_bnd_needed): Likewise.
19178         (output_indirect_thunk_function): Add a bool argument for
19179         function return.
19180         (output_indirect_thunk_function): Don't generate alias for
19181         function return thunk.
19182         (ix86_code_end): Call output_indirect_thunk_function to generate
19183         function return thunks.
19184         (ix86_output_function_return): Set indirect_return_bnd_needed
19185         and indirect_return_needed instead of indirect_thunk_bnd_needed
19186         and indirect_thunk_needed.
19188 2018-03-15  Olga Makhotina  <olga.makhotina@intel.com>
19190         * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
19191         (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
19192         (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
19194 2018-03-15  David Malcolm  <dmalcolm@redhat.com>
19195             Paul Hua <paul.hua.gm@gmail.com>
19197         PR c/84852
19198         * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
19200 2018-03-15  Segher Boessenkool  <segher@kernel.crashing.org>
19202         * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
19203         TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
19204         resp. SFmode cases.
19206 2018-03-15  Tamar Christina  <tamar.christina@arm.com>
19208         PR target/84711
19209         * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
19210         instead of GET_MODE_SIZE when comparing Units.
19212 2018-03-15  Vladimir Mezentsev  <vladimir.mezentsev@oracle.com>
19214         PR target/68256
19215         * varasm.c (hash_section): Return an unchangeble hash value
19216         * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
19217         Return !aarch64_can_use_per_function_literal_pools_p ().
19219 2018-03-15  Jakub Jelinek  <jakub@redhat.com>
19221         PR target/84860
19222         * optabs.c (emit_conditional_move): Pass address of cmode's copy
19223         rather than address of cmode as last argument to prepare_cmp_insn.
19225 2018-03-15  Julia Koval  <julia.koval@intel.com>
19227         * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
19228         F_AVX512VNNI, F_AVX512BITALG): New.
19230 2018-03-14  John David Anglin  <danglin@gcc.gnu.org>
19232         PR target/83451
19233         * config/pa/pa.c (pa_emit_move_sequence):  Always emit secondary reload
19234         insn for floating-point loads and stores.
19236 2018-03-14  Carl Love  <cel@us.ibm.com>
19238         * config/rs6000/rs6000-c.c: Add macro definitions for
19239         ALTIVEC_BUILTIN_VEC_PERMXOR.
19240         * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
19241         * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
19242         * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
19243         * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
19244         UNSPEC_VPERMXOR.
19245         * config/doc/extend.texi: Add prototypes for vec_permxor.
19247 2018-03-14  David Malcolm  <dmalcolm@redhat.com>
19249         PR c/84852
19250         * diagnostic-show-locus.c (class layout_point): Convert m_line
19251         from int to linenum_type.
19252         (line_span::comparator): Use linenum "compare" function when
19253         comparing line numbers.
19254         (test_line_span): New function.
19255         (layout_range::contains_point): Convert param "row" from int to
19256         linenum_type.
19257         (layout_range::intersects_line_p): Likewise.
19258         (layout::will_show_line_p): Likewise.
19259         (layout::print_source_line): Likewise.
19260         (layout::should_print_annotation_line_p): Likewise.
19261         (layout::print_annotation_line): Likewise.
19262         (layout::print_leading_fixits): Likewise.
19263         (layout::annotation_line_showed_range_p): Likewise.
19264         (struct line_corrections): Likewise for field m_row.
19265         (line_corrections::line_corrections): Likewise for param "row".
19266         (layout::print_trailing_fixits): Likewise.
19267         (layout::get_state_at_point): Likewise.
19268         (layout::get_x_bound_for_row): Likewise.
19269         (layout::print_line): Likewise.
19270         (diagnostic_show_locus): Likewise for locals "last_line" and "row".
19271         (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
19272         * input.c (selftest::test_linenum_comparisons): New function.
19273         (selftest::input_c_tests): Call it.
19274         * selftest.c (selftest::test_assertions): Test ASSERT_GT,
19275         ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
19276         * selftest.h (ASSERT_GT): New macro.
19277         (ASSERT_GT_AT): New macro.
19278         (ASSERT_LT): New macro.
19279         (ASSERT_LT_AT): New macro.
19281 2018-03-14  Segher Boessenkool  <segher@kernel.crashing.org>
19283         PR rtl-optimization/84780
19284         * combine.c (distribute_links): Don't make a link based on pc_rtx.
19286 2018-03-14  Martin Liska  <mliska@suse.cz>
19288         * tree.c (record_node_allocation_statistics): Use
19289         get_stats_node_kind.
19290         (get_stats_node_kind): New function extracted from
19291         record_node_allocation_statistics.
19292         (free_node): Use get_stats_node_kind.
19294 2018-03-14  Richard Biener  <rguenther@suse.de>
19296         * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
19297         that the value-set of ANTIC_IN doesn't grow.
19299         Revert
19300         * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
19301         member.
19302         (BB_VISITED_WITH_VISITED_SUCCS): New define.
19303         (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
19305 2018-03-14  Julia Koval  <julia.koval@intel.com>
19307         * config.gcc (icelake-client, icelake-server): New.
19308         (icelake): Remove.
19309         * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
19310         (initial_ix86_arch_features): Ditto.
19311         (PTA_SKYLAKE): Add SGX.
19312         (PTA_ICELAKE): Remove.
19313         (PTA_ICELAKE_CLIENT): New.
19314         (PTA_ICELAKE_SERVER): New.
19315         (ix86_option_override_internal): Split up icelake on icelake client and
19316         icelake server.
19317         (get_builtin_code_for_version): Ditto.
19318         (fold_builtin_cpu): Ditto.
19319         * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
19320         * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
19321         * config/i386/i386.h (processor_type): Ditto.
19322         * doc/invoke.texi: Ditto.
19324 2018-03-14  Jakub Jelinek  <jakub@redhat.com>
19326         PR sanitizer/83392
19327         * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
19328         INTEGER_CST offset, add it together with bitpos / 8 and
19329         sign extend based on POINTER_SIZE.
19331         PR target/84844
19332         Revert
19333         2017-04-20  Uros Bizjak  <ubizjak@gmail.com>
19335         PR target/78090
19336         * config/i386/constraints.md (Yc): New register constraint.
19337         * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
19338         Use Yc constraint for alternative 2 of operand 0.  Remove
19339         preferred_for_speed attribute.
19341 2018-03-14  Richard Biener  <rguenther@suse.de>
19343         PR tree-optimization/84830
19344         * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
19345         with the old one to avoid oscillations.
19347 2018-03-13  Vladimir Makarov  <vmakarov@redhat.com>
19349         PR target/83712
19350         * lra-assigns.c (find_all_spills_for): Ignore uninteresting
19351         pseudos.
19352         (assign_by_spills): Return a flag of reload assignment failure.
19353         Do not process the reload assignment failures.  Do not spill other
19354         reload pseudos if they has the same reg class.  Update n if
19355         necessary.
19356         (lra_assign): Add a return arg.  Set up from the result of
19357         assign_by_spills call.
19358         (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
19359         * lra-constraints.c (split_reg): Add a new arg.  Use it instead of
19360         usage_insns if it is not NULL.
19361         (spill_hard_reg_in_range): New function.
19362         (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
19363         * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
19364         function prototypes.
19365         (lra_assign): Change prototype.
19366         * lra.c (lra): Add code to deal with fails by splitting hard reg
19367         live ranges.
19369 2018-03-01  Palmer Dabbelt  <palmer@sifive.com>
19371         * config/riscv/riscv.opt (mrelax): New option.
19372         * config/riscv/riscv.c (riscv_file_start): Emit ".option
19373         "norelax" when riscv_mrelax is disabled.
19374         * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
19376 2018-03-13  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
19378         PR target/84743
19379         * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
19380         reassociation for int modes.
19382 2018-03-13  Richard Sandiford  <richard.sandiford@linaro.org>
19384         * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
19385         Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
19386         for big-endian.
19387         * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
19388         * config/aarch64/aarch64-sve.md
19389         (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
19390         (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
19391         (*extend<mode><Vwide>2): Rename to...
19392         (aarch64_sve_extend<mode><Vwide>2): ...this.
19393         (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
19394         renaming the old pattern to...
19395         (aarch64_sve_punpk<perm_hilo>_<mode>): ...this.  Only define
19396         unsigned packs.
19397         (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
19398         define_expand, renaming the old pattern to...
19399         (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
19400         (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
19401         (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
19402         account when deciding which SVE instruction the optab should use.
19403         (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
19405 2018-03-13  Richard Sandiford  <richard.sandiford@linaro.org>
19407         * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
19408         (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
19409         (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
19410         (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
19411         (tlsdesc_small_<mode>): Turn a define_expand and use
19412         tlsdesc_small_sve_<mode> for SVE.  Rename original define_insn to...
19413         (tlsdesc_small_advsimd_<mode>): ...this.
19414         (tlsdesc_small_sve_<mode>): New pattern.
19416 2018-03-13  Richard Sandiford  <richard.sandiford@linaro.org>
19418         * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
19419         (UNSPEC_UMUL_HIGHPART): New constants.
19420         (MUL_HIGHPART): New int iteraor.
19421         (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
19422         * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
19423         define_expand.
19424         (*<su>mul<mode>3_highpart): New define_insn.
19426 2018-03-13  Eric Botcazou  <ebotcazou@adacore.com>
19428         PR lto/84805
19429         * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
19430         incomplete types.
19432 2018-03-13  Martin Liska  <mliska@suse.cz>
19434         PR ipa/84658.
19435         * (sem_item_optimizer::sem_item_optimizer): Initialize new
19436         vector.
19437         (sem_item_optimizer::~sem_item_optimizer): Release it.
19438         (sem_item_optimizer::merge_classes): Register variable aliases.
19439         (sem_item_optimizer::fixup_pt_set): New function.
19440         (sem_item_optimizer::fixup_points_to_sets): Likewise.
19441         * ipa-icf.h: Declare new variables and functions.
19443 2018-03-13  Jakub Jelinek  <jakub@redhat.com>
19445         PR middle-end/84834
19446         * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
19447         integer_pow2p@2 and test integer_pow2p in condition.
19448         (A < 0 ? C : 0): Similarly for @1.
19450         PR middle-end/84831
19451         * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
19452         characters starting at p contain '\0' character, don't look beyond
19453         that.
19455         PR target/84827
19456         * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
19457         pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
19459         PR target/84828
19460         * reg-stack.c (change_stack): Change update_end var from int to
19461         rtx_insn *, if non-NULL don't update just BB_END (current_block), but
19462         also call set_block_for_insn on the newly added insns and rescan.
19464         PR target/84786
19465         * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
19466         on the last operand.
19468         PR c++/84704
19469         * tree.c (stabilize_reference_1): Return save_expr (e) for
19470         STATEMENT_LIST even if it doesn't have side-effects.
19472 2018-03-12  Jonathan Wakely  <jwakely@redhat.com>
19474         * doc/invoke.texi (-mclflushopt): Fix spelling of option.
19476 2018-03-12  Renlin Li  <renlin.li@arm.com>
19478         * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
19479         aarch64_output_scalar_simd_mov_immediate.
19481 2018-03-12  Martin Sebor  <msebor@redhat.com>
19483         PR tree-optimization/83456
19484         * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
19485         for perfectly overlapping calls to memcpy.
19486         (gimple_fold_builtin_memory_chk): Same.
19487         (gimple_fold_builtin_strcpy): Handle no-warning.
19488         (gimple_fold_builtin_stxcpy_chk): Same.
19489         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
19491 2018-03-12  Segher Boessenkool  <segher@kernel.crashing.org>
19493         * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
19494         parameter.  Use it for SFmode.
19495         (rs6000_function_arg_advance_1): Adjust.
19496         (rs6000_function_arg): Adjust.
19497         (rs6000_gimplify_va_arg): Pass false for that new parameter.
19499 2018-03-12  Segher Boessenkool  <segher@kernel.crashing.org>
19501         PR rtl-optimization/84169
19502         PR rtl-optimization/84780
19503         * combine.c (can_combine_p): Check for a 2-insn combination whether
19504         the destination register is used between the two insns, too.
19506 2018-03-12  Richard Biener  <rguenther@suse.de>
19508         PR tree-optimization/84803
19509         * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
19510         for refs DR analysis didn't process.
19512 2018-03-12  Richard Biener  <rguenther@suse.de>
19514         PR tree-optimization/84777
19515         * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
19516         force-vectorize loops ignore whether we are optimizing for size.
19518 2018-03-12  Chung-Ju Wu  <jasonwucj@gmail.com>
19520         * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
19521         (TARGET_MD_ASM_ADJUST): Define.
19523 2018-03-12  Monk Chiang  <sh.chiang04@gmail.com>
19524             Kito Cheng  <kito.cheng@gmail.com>
19525             Chung-Ju Wu  <jasonwucj@gmail.com>
19527         * config/nds32/nds32.c (nds32_compute_stack_frame,
19528         nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
19529         nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
19530         nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
19531         nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
19532         * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
19533         NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
19534         * config/nds32/nds32.md (prologue, epilogue): Use macro
19535         NDS32_V3PUSH_AVAILABLE_P to do checking.
19537 2018-03-11  Jakub Jelinek  <jakub@redhat.com>
19539         PR debug/58150
19540         * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
19541         DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
19542         but on TYPE_SIZE.  Don't do anything for ENUM_IS_OPAQUE if not creating
19543         a new die.  Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE.  Guard
19544         addition of most attributes on !orig_type_die or the attribute not
19545         being present already.  Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
19547 2018-03-11  Kito Cheng  <kito.cheng@gmail.com>
19548             Chung-Ju Wu  <jasonwucj@gmail.com>
19550         * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
19551         __NDS32_VH__ macro.
19552         * config/nds32/nds32.opt (mvh): New option.
19554 2018-03-11  Kito Cheng  <kito.cheng@gmail.com>
19555             Chung-Ju Wu  <jasonwucj@gmail.com>
19557         * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
19558         function.
19559         * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
19560         * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
19561         definition.
19563 2018-03-11  Kito Cheng  <kito.cheng@gmail.com>
19564             Chung-Ju Wu  <jasonwucj@gmail.com>
19566         * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
19567         function.
19568         * config/nds32/nds32-multiple.md (strlensi): New pattern.
19569         * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
19571 2018-03-11  Monk Chiang  <sh.chiang04@gmail.com>
19572             Kito Cheng  <kito.cheng@gmail.com>
19573             Chung-Ju Wu  <jasonwucj@gmail.com>
19575         * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
19576         UNSPEC_FFMISM and UNSPEC_FLMISM.
19577         * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
19578         for ffb, ffmism and flmism.
19579         * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
19580         (unspec_ffmism): Ditto.
19581         (unspec_flmism): Ditto.
19582         (nds32_expand_builtin_impl): Check if string extension is available.
19583         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
19584         NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
19586 2018-03-10  Vladimir Makarov  <vmakarov@redhat.com>
19588         Reverting patch:
19589         2018-03-09  Vladimir Makarov  <vmakarov@redhat.com>
19591         PR target/83712
19592         * lra-assigns.c (assign_by_spills): Return a flag of reload
19593         assignment failure.  Do not process the reload assignment
19594         failures.  Do not spill other reload pseudos if they has the same
19595         reg class.
19596         (lra_assign): Add a return arg.  Set up from the result of
19597         assign_by_spills call.
19598         (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
19599         * lra-constraints.c (split_reg): Add a new arg.  Use it instead of
19600         usage_insns if it is not NULL.
19601         (spill_hard_reg_in_range): New function.
19602         (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
19603         * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
19604         function prototypes.
19605         (lra_assign): Change prototype.
19606         * lra.c (lra): Add code to deal with fails by splitting hard reg
19607         live ranges.
19609 2018-03-10  H.J. Lu  <hongjiu.lu@intel.com>
19611         PR target/84807
19612         * config/i386/i386.opt: Replace Enforcment with Enforcement.
19614 2018-03-10  Alexandre Oliva  <aoliva@redhat.com>
19616         PR debug/84620
19617         * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
19618         (dw_val_node): Add val_symbolic_view.
19619         * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
19620         (symview_upper_bound): New.
19621         (new_line_info_table): Initialize symviews_since_reset.
19622         (dwarf2out_source_line): Count symviews_since_reset and set
19623         symview_upper_bound.
19624         (dw_val_equal_p): Handle symview.
19625         (add_AT_symview): New.
19626         (print_dw_val): Handle symview.
19627         (attr_checksum, attr_checksum_ordered): Likewise.
19628         (same_dw_val_p, size_of_die): Likewise.
19629         (value_format, output_die): Likewise.
19630         (add_high_low_attributes): Use add_AT_symview for entry_view.
19631         (dwarf2out_finish): Reset symview_upper_bound, clear
19632         zero_view_p.
19634 2018-03-09  Peter Bergner  <bergner@vnet.ibm.com>
19636         PR target/83969
19637         * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
19638         Add strict argument and use it.
19639         (rs6000_split_multireg_move): Update for new strict argument.
19640         (mem_operand_gpr): Disallow all non-offsettable addresses.
19641         * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
19643 2018-03-09  Jakub Jelinek  <jakub@redhat.com>
19645         PR target/84772
19646         * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
19647         temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
19648         * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
19650         PR c++/84767
19651         * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
19652         decl, use remap_type if we want to use the type.
19654 2018-03-09  Martin Sebor  <msebor@redhat.com>
19656         PR tree-optimization/84526
19657         * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
19658         Remove dead code.
19659         (builtin_access::generic_overlap): Be prepared to handle non-array
19660         base objects.
19662 2018-03-09  Alexandre Oliva  <aoliva@redhat.com>
19664         PR rtl-optimization/84682
19665         * lra-constraints.c (process_address_1): Check is_address flag
19666         for address constraints.
19667         (process_alt_operands): Likewise.
19668         * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
19669         preprocess_constraints.
19670         * recog.h (preprocess_constraints): Add oploc parameter.
19671         Adjust callers.
19672         * recog.c (preprocess_constraints): Test address_operand for
19673         CT_ADDRESS constraints.
19675 2018-03-09  Vladimir Makarov  <vmakarov@redhat.com>
19677         PR target/83712
19678         * lra-assigns.c (assign_by_spills): Return a flag of reload
19679         assignment failure.  Do not process the reload assignment
19680         failures.  Do not spill other reload pseudos if they has the same
19681         reg class.
19682         (lra_assign): Add a return arg.  Set up from the result of
19683         assign_by_spills call.
19684         (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
19685         * lra-constraints.c (split_reg): Add a new arg.  Use it instead of
19686         usage_insns if it is not NULL.
19687         (spill_hard_reg_in_range): New function.
19688         (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
19689         * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
19690         function prototypes.
19691         (lra_assign): Change prototype.
19692         * lra.c (lra): Add code to deal with fails by splitting hard reg
19693         live ranges.
19695 2018-03-09  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
19697         PR target/83193
19698         * common/config/arm/arm-common.c (arm_parse_arch_option_name):
19699         Accept complain bool parameter.  Only emit errors if it is true.
19700         (arm_parse_cpu_option_name): Likewise.
19701         (arm_target_thumb_only): Adjust callers of the above.
19702         * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
19703         prototype to take a default true bool parameter.
19704         (arm_parse_arch_option_name): Likewise.
19706 2018-03-09  David Malcolm  <dmalcolm@redhat.com>
19707             Francois-Xavier Coudert  <fxcoudert@gcc.gnu.org>
19709         PR jit/64089
19710         PR jit/84288
19711         * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
19712         * configure: Regenerate.
19713         * configure.ac ("linker --version-script option"): New.
19714         ("linker soname option"): New.
19716 2018-03-09  Richard Biener  <rguenther@suse.de>
19718         PR tree-optimization/84775
19719         * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
19720         immediate uses of predicate stmts and mark them modified.
19722         Revert
19723         PR tree-optimization/84178
19724         * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
19725         to caller.
19726         (version_loop_for_if_conversion): Delay update_ssa call.
19727         (tree_if_conversion): Delay update_ssa until after predicate
19728         insertion.
19730 2018-03-09  Eric Botcazou  <ebotcazou@adacore.com>
19732         PR target/84763
19733         * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
19734         when the function accesses prior frames.
19736 2018-03-08  Jakub Jelinek  <jakub@redhat.com>
19738         PR debug/84456
19739         * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
19740         gen_llsym, otherwise call maybe_gen_llsym.
19742         PR inline-asm/84742
19743         * recog.c (asm_operand_ok): Return 0 if multi-character constraint
19744         has ',' character inside of it.
19746 2018-03-08  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
19748         PR target/84748
19749         * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
19750         as clobbering CC_REGNUM.
19752 2018-03-08  Richard Biener  <rguenther@suse.de>
19754         PR middle-end/84552
19755         * tree-scalar-evolution.c: Include tree-into-ssa.h.
19756         (follow_copies_to_constant): Do not follow SSA names registered
19757         for update.
19759 2018-03-08  Richard Biener  <rguenther@suse.de>
19761         PR tree-optimization/84178
19762         * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
19763         to caller.
19764         (version_loop_for_if_conversion): Delay update_ssa call.
19765         (tree_if_conversion): Delay update_ssa until after predicate
19766         insertion.
19768 2018-03-08  David Malcolm  <dmalcolm@redhat.com>
19770         PR tree-optimization/84178
19771         * tree-if-conv.c (release_bb_predicate): Remove the
19772         the assertion that the stmts have NULL use_ops.
19773         Discard the statements, asserting that they haven't
19774         yet been added to a BB.
19776 2018-03-08  Richard Biener  <rguenther@suse.de>
19778         PR tree-optimization/84746
19779         * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
19780         (phi_translate): Pass in destination ANTIC_OUT set.
19781         (phi_translate_1): Likewise.  For a simplified result lookup
19782         a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
19783         (phi_translate_set): Adjust.
19784         (do_pre_regular_insertion): Likewise.
19785         (do_pre_partial_partial_insertion): Likewise.
19787 2018-03-08  Martin Liska  <mliska@suse.cz>
19789         PR gcov-profile/84735
19790         * doc/gcov.texi: Document usage of profile files.
19791         * gcov-io.h: Document changes in the format.
19793 2018-03-08  Alexandre Oliva  <aoliva@redhat.com>
19795         PR debug/84404
19796         PR debug/84408
19797         * dwarf2out.c (struct dw_line_info_table): Update comments for
19798         view == -1.
19799         (FORCE_RESET_NEXT_VIEW): New.
19800         (FORCE_RESETTING_VIEW_P): New.
19801         (RESETTING_VIEW_P): Check for -1 too.
19802         (ZERO_VIEW_P): Likewise.
19803         (new_line_info_table): Force-reset next view.
19804         (dwarf2out_begin_function): Likewise.
19805         (dwarf2out_source_line): Simplify zero_view_p initialization.
19806         Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
19807         view directly.  Omit view when omitting .loc at line 0.
19809 2018-03-08  Jakub Jelinek  <jakub@redhat.com>
19811         PR tree-optimization/84740
19812         * tree-switch-conversion.c (process_switch): Call build_constructors
19813         only if info.phi_count is non-zero.
19815         PR tree-optimization/84739
19816         * tree-tailcall.c (find_tail_calls): Check call arguments against
19817         DECL_ARGUMENTS (current_function_decl) rather than
19818         DECL_ARGUMENTS (func) when checking for tail recursion.
19820 2018-03-07  Jakub Jelinek  <jakub@redhat.com>
19822         * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
19823         Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
19824         Volker Reichelt's entry and add entries for people that perform
19825         GCC fuzzy testing and report numerous bugs.
19827 2018-03-07  Segher Boessenkool  <segher@kernel.crashing.org>
19829         PR target/82411
19830         * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
19831         readonly data in sdata, if that is disabled.
19832         * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
19833         * doc/invoke.texi (RS/6000 and PowerPC Options): Document
19834         -mreadonly-in-sdata option.
19836 2018-03-07  Martin Sebor  <msebor@redhat.com>
19838         PR tree-optimization/84468
19839         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
19840         basic block when looking for nul assignment.
19842 2018-03-07  Eric Botcazou  <ebotcazou@adacore.com>
19844         PR target/84277
19845         * except.h (output_function_exception_table): Adjust prototype.
19846         * except.c (output_function_exception_table): Remove FNNAME parameter
19847         and add SECTION parameter.  Ouput one part of the table at a time.
19848         * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
19849         the first part of the exception table and emit unwind directives.
19850         * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
19851         (i386_pe_seh_cold_init): Likewise.
19852         * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
19853         (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
19854         * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
19855         (ix86_output_call_insn): Emit a nop in one more case for SEH.
19856         * config/i386/winnt.c: Include except.h.
19857         (struct seh_frame_state): Add reg_offset, after_prologue and
19858         in_cold_section fields.
19859         (i386_pe_seh_end_prologue): Set seh->after_prologue.
19860         (i386_pe_seh_cold_init): New function.
19861         (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
19862         to seh->in_cold_section.
19863         (seh_emit_push): Record the offset of the push.
19864         (seh_emit_save): Record the offet of the save.
19865         (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
19866         Test seh->after_prologue to disregard the epilogue.
19867         (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
19868         (i386_pe_end_cold_function): New function.
19870 2018-03-07  Jakub Jelinek  <jakub@redhat.com>
19872         PR fortran/84565
19873         * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
19874         aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
19876         PR c++/84704
19877         * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
19878         on tmp_var.
19879         * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
19880         don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
19882         PR middle-end/84723
19883         * multiple_target.c: Include tree-inline.h and intl.h.
19884         (expand_target_clones): Diagnose and fail if node->definition and
19885         !tree_versionable_function_p (node->decl).
19887 2018-03-06  John David Anglin  <danglin@gcc.gnu.org>
19889         * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
19890         sprint_ul.
19891         (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
19892         (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
19893         * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
19895 2018-03-06  Jakub Jelinek  <jakub@redhat.com>
19897         PR target/84710
19898         * combine.c (try_combine): Use reg_or_subregno instead of handling
19899         just paradoxical SUBREGs and REGs.
19901 2018-03-06  Claudiu Zissulescu  <claziss@synopsys.com>
19903         * config/arc/arc.c (arc_finalize_pic): Remove function.
19904         (arc_must_save_register): We use single base PIC register, remove
19905         checks to save/restore the PIC register.
19906         (arc_expand_prologue): Likewise.
19907         * config/arc/arc-protos.h (arc_set_default_type_attributes):
19908         Remove.
19909         (arc_verify_short): Likewise.
19910         (arc_attr_type): Likewise.
19911         * config/arc/arc.c (arc_set_default_type_attributes): Remove.
19912         (walk_stores): Likewise.
19913         (arc_address_cost): Make it static.
19914         (arc_verify_short): Likewise.
19915         (branch_dest): Likewise.
19916         (arc_attr_type): Likewise.
19917         * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
19918         (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
19919         (arc_final_prescan_insn): Remove inserting the nops due to
19920         hardware hazards.  It is done in reorg step.
19921         (insn_length_variant_t): Remove.
19922         (insn_length_parameters_t): Likewise.
19923         (arc_insn_length_parameters): Likewise.
19924         (arc_get_insn_variants): Likewise.
19925         * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
19927 2018-03-06  Jakub Jelinek  <jakub@redhat.com>
19929         PR inline-asm/84683
19930         * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
19931         assertion failure.
19933         PR tree-optimization/84687
19934         * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
19935         on new_node->decl.
19936         * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
19938 2018-03-05  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
19940         * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
19941         Rename to ppc_speculation_barrier.
19942         * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
19943         __builtin_ppc_speculation_barrier.
19945 2018-03-05  Jakub Jelinek  <jakub@redhat.com>
19947         PR target/84700
19948         * combine.c (combine_simplify_rtx): Don't try to simplify if
19949         if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
19950         are equal to x.
19952 2018-03-05  Segher Boessenkool  <segher@kernel.crashing.org>
19954         * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
19955         to 32 bytes when compiling for POWER9.
19957 2018-03-05  Jakub Jelinek  <jakub@redhat.com>
19959         PR target/84564
19960         * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
19961         regparm >= 3 with no arg reg available also for calls with
19962         flag_force_indirect_call.  Pass decl to ix86_function_regparm.
19964         PR target/84524
19965         * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
19966         orig,vex.
19967         (*<plusminus_insn><mode>3): Likewise.  Remove <mask_operand3> uses.
19969 2018-03-05  Peter Bergner  <bergner@vnet.ibm.com>
19971         PR target/84264
19972         * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
19974 2018-03-05  Richard Biener  <rguenther@suse.de>
19976         PR tree-optimization/84486
19977         * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
19978         When inserting a __builtin_assume_aligned call set the LHS
19979         SSA name alignment info accordingly.
19981 2018-03-05  Wilco Dijkstra  <wdijkstr@arm.com>
19983         PR tree-optimization/84114
19984         * config/aarch64/aarch64.c (aarch64_reassociation_width)
19985         Avoid reassociation of FLOAT_MODE addition.
19987 2018-03-05  Olga Makhotina  <olga.makhotina@intel.com>
19989         * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
19990         OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
19991         OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
19992         (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
19993         * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
19994         * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
19995         * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
19996         and -mwbnoinvd.
19997         * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
19998         __builtin_ia32_wbinvd): New builtins.
19999         (SPECIAL_ARGS2): New.
20000         * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
20001         (SPECIAL_ARGS2): New.
20002         * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
20003         (ix86_valid_target_attribute_inner_p): Ditto.
20004         (ix86_init_mmx_sse_builtins): Add special_args2.
20005         * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
20006         TARGET_WBNOINVD_P): New.
20007         * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
20008         (define_insn "wbinvd", define_insn "wbnoinvd"): New.
20009         * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
20010         * config/i386/immintrin.h (_wbinvd): New intrinsic.
20011         * config/i386/pconfigintrin.h: New file.
20012         * config/i386/wbnoinvdintrin.h: Ditto.
20013         * config/i386/x86intrin.h: Add headers pconfigintrin.h and
20014         wbnoinvdintrin.h.
20015         * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
20017 2018-03-05  Richard Biener  <rguenther@suse.de>
20019         PR tree-optimization/84670
20020         * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
20021         member.
20022         (BB_VISITED_WITH_VISITED_SUCCS): New define.
20023         (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
20024         (compute_antic_aux): Only assert the number of values in ANTIC_IN
20025         doesn't grow if all successors (recursively) were visited at least
20026         once.
20028 2018-03-05  Richard Biener  <rguenther@suse.de>
20030         PR tree-optimization/84650
20031         * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
20032         if executed in the loop pipeline.
20034 2018-03-05  Sandra Loosemore  <sandra@codesourcery.com>
20036         * doc/configfiles.texi (Configuration Files): Move info about
20037         conditionalizing $target-protos.h to...
20038         * doc/sourcebuild.texi (Back End): Here.  Explain how $target.h
20039         differs from $target-protos.h.
20041 2018-03-05  Kito Cheng  <kito.cheng@gmail.com>
20042             Chung-Ju Wu  <jasonwucj@gmail.com>
20044         * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
20045         * config/nds32/nds32-multiple.md (setmemsi): Define.
20046         * config/nds32/nds32-memory-manipulation.c
20047         (nds32_gen_dup_4_byte_to_word_value): New.
20048         (emit_setmem_word_loop): New.
20049         (emit_setmem_byte_loop): New.
20050         (nds32_expand_setmem_loop): New.
20051         (nds32_expand_setmem_loop_v3m): New.
20052         (nds32_expand_setmem_unroll): New.
20053         (nds32_expand_setmem): New.
20055 2018-03-04  Kito Cheng  <kito.cheng@gmail.com>
20056             Chung-Ju Wu  <jasonwucj@gmail.com>
20058         * config/nds32/nds32-memory-manipulation.c
20059         (nds32_emit_load_store): New.
20060         (nds32_emit_post_inc_load_store): New.
20061         (nds32_emit_mem_move): New.
20062         (nds32_emit_mem_move_block): New.
20063         (nds32_expand_movmemsi_loop_unknown_size): New.
20064         (nds32_expand_movmemsi_loop_known_size): New.
20065         (nds32_expand_movmemsi_loop): New.
20066         (nds32_expand_movmemsi_unroll): New.
20067         (nds32_expand_movmemqi): Rename ...
20068         (nds32_expand_movmemsi): ... to this.
20069         * config/nds32/nds32-multiple.md (movmemqi): Rename ...
20070         (movmemsi): ... to this.
20071         * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
20072         (nds32_expand_movmemsi): ... to this.
20074 2018-03-04  Kito Cheng  <kito.cheng@gmail.com>
20075             Monk Chiang  <sh.chiang04@gmail.com>
20076             Chung-Ju Wu  <jasonwucj@gmail.com>
20078         * config/nds32/nds32-protos.h
20079         (nds32_expand_load_multiple): New arguments.
20080         (nds32_expand_store_multiple): Ditto.
20081         (nds32_valid_multiple_load_store): Rename ...
20082         (nds32_valid_multiple_load_store_p): ... to this.
20083         * config/nds32/nds32-memory-manipulation.c
20084         (nds32_expand_load_multiple): Refine implementation.
20085         (nds32_expand_store_multiple): Ditto.
20086         * config/nds32/nds32-multiple.md
20087         (load_multiple): Update nds32_expand_load_multiple interface.
20088         (store_multiple): Update nds32_expand_store_multiple interface.
20089         * config/nds32/nds32-predicates.c
20090         (nds32_valid_multiple_load_store): Rename ...
20091         (nds32_valid_multiple_load_store_p): ... to this and refine
20092         implementation.
20093         * config/nds32/predicates.md
20094         (nds32_load_multiple_and_update_address_operation): New predicate.
20095         (nds32_store_multiple_and_update_address_operation): New predicate.
20097 2018-03-04  Kito Cheng  <kito.cheng@gmail.com>
20098             Chung-Ju Wu  <jasonwucj@gmail.com>
20100         * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
20101         (combo): New attribute.
20102         * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
20104 2018-03-03  Chung-Ju Wu  <jasonwucj@gmail.com>
20106         * config/nds32/nds32.opt: Change -mcmodel= default value.
20108 2018-03-03  Kito Cheng  <kito.cheng@gmail.com>
20109             Monk Chiang  <sh.chiang04@gmail.com>
20110             Chung-Ju Wu  <jasonwucj@gmail.com>
20112         * config/nds32/constants.md (unspec_element): New enum.
20113         * config/nds32/constraints.md (Umw): New constraint.
20114         * config/nds32/nds32-intrinsic.c: Add more builtin functions.
20115         * config/nds32/nds32-intrinsic.md: Likewise.
20116         * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
20117         (nds32_valid_smw_lwm_base_p): New.
20118         (nds32_output_smw_single_word): New.
20119         (nds32_output_lmw_single_word): New.
20120         (nds32_expand_unaligned_load): New.
20121         (nds32_expand_unaligned_store): New.
20122         * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
20123         (nds32_output_smw_single_word): Declare.
20124         (nds32_output_lmw_single_word): Declare.
20125         (nds32_expand_unaligned_load): Declare.
20126         (nds32_expand_unaligned_store): Declare.
20127         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
20128         NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
20129         NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
20130         NDS32_BUILTIN_UASTORE_DW.
20131         * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
20132         predicate.
20134 2018-03-03  Monk Chiang  <sh.chiang04@gmail.com>
20135             Kito Cheng  <kito.cheng@gmail.com>
20136             Chung-Ju Wu  <jasonwucj@gmail.com>
20138         * config/nds32/nds32-intrinsic.c
20139         (nds32_expand_builtin_null_ftype_reg): Delete.
20140         (nds32_expand_builtin_reg_ftype_imm): Ditto.
20141         (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
20142         (nds32_read_argument): New.
20143         (nds32_legitimize_target): Ditto.
20144         (nds32_legitimize_argument): Ditto.
20145         (nds32_check_constant_argument): Ditto.
20146         (nds32_expand_unop_builtin): Ditto.
20147         (nds32_expand_unopimm_builtin): Ditto.
20148         (nds32_expand_binop_builtin): Ditto.
20149         (nds32_builtin_decl_impl): Ditto.
20150         (builtin_description): Ditto.
20151         (nds32_expand_builtin_impl): Rewrite with new infrastructure.
20152         (nds32_init_builtins_impl): Ditto.
20153         * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
20154         (nds32_builtin_decl): New.
20155         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
20156         * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
20158 2018-03-02  Jeff Law  <law@redhat.com>
20160         * reorg.c (stop_search_p): Handle DEBUG_INSN.
20161         (redundant_insn, fill_simple_delay_slots): Likewise.
20162         (fill_slots_from_thread): Likewise.
20163         * resource.c (mark_referenced_resources): Likewise.
20164         (mark_set_resources, find_dead_or_set_registers): Likewise.
20166 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
20168         * substring-locations.h (format_warning_va): Formatting fix for
20169         ATTRIBUTE_GCC_DIAG.
20170         (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
20171         argument.
20172         (format_warning_n_va, format_warning_at_substring_n): New prototypes.
20173         * substring-locations.c: Include intl.h.
20174         (format_warning_va): Turned into small wrapper around
20175         format_warning_n_va, renamed to ...
20176         (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
20177         rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
20178         use ngettext.
20179         (format_warning_at_substring_n): New function.
20180         * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
20181         (fmtwarn): Add ATTRIBUTE_GCC_DIAG.  Turn into a copy of
20182         format_warning_at_substring with just a shorter name instead of
20183         const function pointer.
20184         (fmtwarn_n): New function.
20185         (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
20186         appropriate, get rid of all the fmtstr temporaries, move conditionals
20187         with G_() wrapped string literals directly into fmtwarn arguments,
20188         cast dir.len to (int), formatting fixes.
20190 2018-03-02  Thomas Schwinge  <thomas@codesourcery.com>
20192         * doc/invoke.texi: Remove "Cilk Plus" references.
20194 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
20195             Richard Biener  <rguenther@suse.de>
20197         PR ipa/84628
20198         * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
20199         for error or warning attributes if CALL_FROM_THUNK_P is set.
20200         Formatting fixes.
20202 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
20204         PR target/56540
20205         * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
20206         __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
20208         PR target/56540
20209         * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
20210         __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
20212         * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
20213         instead of -1U in last predictors element's probability member.
20215 2018-03-02  Eric Botcazou  <ebotcazou@adacore.com>
20217         PR ipa/83983
20218         * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
20219         arguments if they are comparable.
20221 2018-03-02  Richard Sandiford  <richard.sandiford@linaro.org>
20223         PR tree-optimization/84634
20224         * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
20225         masks and masked_loop_p with a single loop_masks, making sure it's
20226         null for bb vectorization.
20228 2018-03-02  Richard Sandiford  <richard.sandiford@linaro.org>
20230         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
20231         (vect_analyze_data_ref_access): Use loop->safe_len rather than
20232         loop->force_vectorize to check whether there is no alias.
20234 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
20236         PR target/84614
20237         * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
20238         prototypes.
20239         * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
20240         comments.
20241         (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
20242         * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
20243         instead of a loop around prev_real_insn.
20244         * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
20245         prev_real_insn.
20247         PR inline-asm/84625
20248         * config/i386/i386.c (ix86_print_operand): Use conditional
20249         output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
20250         zero vector.
20252 2018-03-02  Richard Biener  <rguenther@suse.de>
20254         PR tree-optimization/84427
20255         * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
20256         (bitmap_set_subtract_values): Rewrite to handle multiple
20257         exprs per value.
20258         (clean): Likewise.
20259         (prune_clobbered_mems): Likewise.
20260         (phi_translate): Take edge instead of pred/phiblock.
20261         (phi_translate_1): Likewise.
20262         (phi_translate_set): Likewise.  Insert all translated
20263         exprs for a value into the set, keeping possibly multiple
20264         expressions per value.
20265         (compute_antic_aux): Adjust for phi_translate changes.
20266         When intersecting union the expressions and prune those
20267         not in the final value set, keeping possibly multiple
20268         expressions per value.  Do not use value-insertion
20269         for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
20270         all expressions.  Add verification that the value-sets
20271         only shrink during iteration.
20272         (compute_partial_antic_aux): Adjust for the phi_translate changes.
20273         (do_pre_regular_insertion): Likewise.
20274         (do_pre_partial_partial_insertion): Likewise.
20276 2018-03-02  Richard Biener  <rguenther@suse.de>
20278         PR target/82005
20279         * config/darwin.c (saved_debug_info_level): New static global.
20280         (darwin_asm_lto_start): Disable debug info generation for LTO out.
20281         (darwin_asm_lto_end): Restore debug info generation settings.
20283 2018-03-01  Martin Liska  <mliska@suse.cz>
20285         PR sanitizer/82484
20286         * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
20287         volatile arguments.
20289 2018-03-01  Richard Biener  <rguenther@suse.de>
20291         PR debug/84645
20292         * dwarf2out.c (gen_variable_die): Properly handle late VLA
20293         type annotation with LTO when debug was disabled at compile-time.
20295 2018-03-01  Matthew Fortune  <mfortune@gmail.com>
20297         * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
20298         XINT with INTVAL.
20299         (mips_final_postscan_insn): Likewise.
20301 2018-03-01  Richard Sandiford  <richard.sandiford@linaro.org>
20303         PR rtl-optimization/84528
20304         * alias.c (init_alias_target): Add commentary.
20305         (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
20306         a unique base value if the frame pointer is not eliminated
20307         to the stack pointer.
20309 2018-03-01  Tom de Vries  <tom@codesourcery.com>
20311         PR rtl-optimization/83327
20312         * lra-int.h (hard_regs_spilled_into): Declare.
20313         * lra.c (hard_regs_spilled_into): Define.
20314         (init_reg_info): Init hard_regs_spilled_into.
20315         * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
20316         * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
20317         (process_bb_lives): Handle hard_regs_spilled_into.
20318         (lra_create_live_ranges_1): Before doing liveness propagation, clear
20319         regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
20321 2018-02-28  David Edelsohn  <dje.gcc@gmail.com>
20323         * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
20324         (powerpc-ibm-aix[789]*): Default to AIX 7.2.
20325         * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
20326         * config/rs6000/aix72.h: New file.
20328 2018-02-28  Jakub Jelinek  <jakub@redhat.com>
20330         * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
20331         instead of warning_at with conditional singular and plural messages
20332         where possible.
20334         PR target/52991
20335         * stor-layout.c (update_alignment_for_field): For
20336         targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
20337         && !DECL_PACKED (field), do the alignment update, just use
20338         only desired_align instead of MAX (type_align, desired_align)
20339         as the alignment.
20340         (place_field): Don't do known_align < desired_align handling
20341         early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
20342         is non-NULL, instead do it after rli->prev_field handling and
20343         only if not within a bitfield word.  For DECL_PACKED (field)
20344         use type_align of BITS_PER_UNIT.
20346 2018-02-28  Eric Botcazou  <ebotcazou@adacore.com>
20348         * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
20349         superfluous parentheses and trailing spaces.
20351 2018-02-28  Richard Biener  <rguenther@suse.de>
20353         PR tree-optimization/84584
20354         * graphite-scop-detection.c (scop_detection::add_scop): Discard
20355         SCoPs with fake exit edge.
20357 2018-02-28  Martin Liska  <mliska@suse.cz>
20359         PR testsuite/84597
20360         * timevar.c (timer::print): Fix format to properly print 100%
20361         values.
20363 2018-02-28  Richard Biener  <rguenther@suse.de>
20365         PR middle-end/84607
20366         * genmatch.c (capture_info::walk_match): Do not mark
20367         captured expressions without operands as expr_p given
20368         they act more like predicates and should be subject to
20369         "lost tail" side-effect preserving.
20371 2018-02-28  Alexandre Oliva  <aoliva@redhat.com>
20373         PR rtl-optimization/81611
20374         * auto-inc-dec.c (attempt_change): Move dead note from
20375         mem_insn if it's the next use of regno
20376         (find_address): Take address use of reg holding
20377         non-incremented value.  Add parm to limit search to the named
20378         reg only.
20379         (merge_in_block): Attempt to use a mem insn that is the next
20380         use of the original regno.
20382 2018-02-27  Martin Sebor  <msebor@redhat.com>
20384         PR c++/83871
20385         * doc/invoke.texi (-Wmissing-attributes): New option.
20386         * print-tree.c (print_node): Handle DECL_UNINLINABLE.
20388 2018-02-27  Martin Sebor  <msebor@redhat.com>
20390         PR translation/84207
20391         * diagnostic-core.h (warning_n, error_n, inform_n): Change
20392         n argument to unsigned HOST_WIDE_INT.
20393         * diagnostic.c (warning_n, error_n, inform_n): Ditto.
20394         (diagnostic_n_impl): Ditto.  Handle arguments in excess of LONG_MAX.
20395         * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
20396         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
20398 2018-02-27  Richard Biener  <rguenther@suse.de>
20400         PR tree-optimization/84512
20401         * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
20402         Do not use the estimate returned from record_stmt_cost for
20403         the scalar iteration cost but sum properly using add_stmt_cost.
20405 2018-02-27  Richard Biener  <rguenther@suse.de>
20407         PR tree-optimization/84466
20408         * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
20409         Adjust last change to less strictly validate use operands.
20411 2018-02-27  Martin Liska  <mliska@suse.cz>
20413         PR gcov-profile/84548
20414         * gcov.c (process_file): Allow partial overlap and consider it
20415         also as group functions.
20416         (output_lines): Properly calculate range of lines for a group.
20418 2018-02-27  Martin Liska  <mliska@suse.cz>
20420         * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
20421         'ggc' suffixes.  Change first column width.
20422         (timer::print): Fix formatting of the column.
20424 2018-02-27  Alexandre Oliva  <aoliva@redhat.com>
20426         * tree-ssa-live.c (remove_unused_scope_block_p): Do not
20427         preserve inline entry blocks for the sake of debug inline
20428         entry point markers alone.
20429         (remove_unused_locals): Suggest in comments a better place to
20430         force the preservation of inline entry blocks that are
20431         otherwise unused, but do not preserve them.
20433 2018-02-26  H.J. Lu  <hongjiu.lu@intel.com>
20435         * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
20437 2018-02-26  H.J. Lu  <hongjiu.lu@intel.com>
20439         PR target/84039
20440         * config/i386/constraints.md (Bs): Replace
20441         ix86_indirect_branch_register with
20442         TARGET_INDIRECT_BRANCH_REGISTER.
20443         (Bw): Likewise.
20444         * config/i386/i386.md (indirect_jump): Likewise.
20445         (tablejump): Likewise.
20446         (*sibcall_memory): Likewise.
20447         (*sibcall_value_memory): Likewise.
20448         Peepholes of indirect call and jump via memory: Likewise.
20449         (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
20450         (*sibcall_value_GOT_32): Likewise.
20451         * config/i386/predicates.md (indirect_branch_operand): Likewise.
20452         (GOT_memory_operand): Likewise.
20453         (call_insn_operand): Likewise.
20454         (sibcall_insn_operand): Likewise.
20455         (GOT32_symbol_operand): Likewise.
20456         * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
20458 2018-02-26  Eric Botcazou  <ebotcazou@adacore.com>
20460         PR rtl-optimization/83496
20461         * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
20462         booleans to RTXes.  Call fix_reg_dead_note on every non-null element.
20463         (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
20464         redundant insn, if any.
20465         (relax_delay_slots): Likewise.
20466         (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
20468 2018-02-26  Richard Sandiford  <richard.sandiford@linaro.org>
20470         PR tree-optimization/83965
20471         * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
20472         that grouped statements are part of a reduction chain.  Return
20473         true if the statement is not marked as a reduction itself but
20474         is part of a group.
20475         (vect_recog_dot_prod_pattern): Don't check whether the statement
20476         is part of a group here.
20477         (vect_recog_sad_pattern): Likewise.
20478         (vect_recog_widen_sum_pattern): Likewise.
20480 2018-02-26  Eric Botcazou  <ebotcazou@adacore.com>
20482         PR debug/84545
20483         * final.c (rest_of_clean_state): Also look for calls inside sequences.
20485 2018-02-26  H.J. Lu  <hongjiu.lu@intel.com>
20487         PR target/84530
20488         * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
20489         the bool argument.
20490         (ix86_output_indirect_function_return): New prototype.
20491         (ix86_split_simple_return_pop_internal): Likewise.
20492         * config/i386/i386.c (indirect_return_via_cx): New.
20493         (indirect_return_via_cx_bnd): Likewise.
20494         (indirect_thunk_name): Handle return va CX_REG.
20495         (output_indirect_thunk_function): Create alias for
20496         __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
20497         (ix86_output_indirect_jmp): Remove the bool argument.
20498         (ix86_output_indirect_function_return): New function.
20499         (ix86_split_simple_return_pop_internal): Likewise.
20500         * config/i386/i386.md (*indirect_jump): Don't pass false
20501         to ix86_output_indirect_jmp.
20502         (*tablejump_1): Likewise.
20503         (simple_return_pop_internal): Change it to define_insn_and_split.
20504         Call ix86_split_simple_return_pop_internal to split it for
20505         -mfunction-return=.
20506         (simple_return_indirect_internal): Call
20507         ix86_output_indirect_function_return instead of
20508         ix86_output_indirect_jmp.
20510 2018-02-26  Jakub Jelinek  <jakub@redhat.com>
20512         PR bootstrap/84405
20513         * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
20514         memset and value initialization afterwards.
20516 2018-02-26  Christophe Lyon  <christophe.lyon@linaro.org>
20518         * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
20520 2018-02-26  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
20522         PR target/84521
20523         * common/config/aarch64/aarch64-common.c
20524         (aarch_option_optimization_table[]): Switch
20525         off fomit-frame-pointer
20527 2018-02-26  Kito Cheng  <kito.cheng@gmail.com>
20528             Chung-Ju Wu  <jasonwucj@gmail.com>
20530         * config/nds32/nds32-multiple.md (load_multiple): Disallow
20531         volatile memory.
20532         (store_multiple): Ditto.
20534 2018-02-26  Kito Cheng  <kito.cheng@gmail.com>
20536         * config.gcc: Add --with-cpu support for nds32 target.
20537         * config/nds32/nds32-opts.h (nds32_cpu_type): New.
20538         * config/nds32/nds32.opt: Add -mcpu= option.
20540 2018-02-25  Segher Boessenkool  <segher@kernel.crashing.org>
20542         * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
20543         isel=yes): Warn for these deprecated options.
20545 2018-02-23  David Edelsohn  <dje.gcc@gmail.com>
20547         * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
20548         ISA_2_5_MASKS_EMBEDDED.
20550 2018-02-23  Jakub Jelinek  <jakub@redhat.com>
20552         * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
20553         p->max as pointers rather than using iterative_hash_expr.
20555 2018-02-23  Carl Love  <cel@us.ibm.com>
20557         * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
20558         macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
20559         BU_P8V_OVERLOAD_2.
20560         * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
20561         P8V_BUILTIN_VEC_VSIGNED2.  Change VSX_BUILTIN_VEC_VUNSIGNED2 to
20562         P8V_BUILTIN_VEC_VUNSIGNED2.
20564 2018-02-22  Vladimir Makarov  <vmakarov@redhat.com>
20566         PR target/81572
20567         * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
20568         * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
20569         LRA_UNKNOWN_ALT.
20570         * lra-constraints.c (curr_insn_transform): Set up
20571         LRA_NON_CLOBBERED_ALT for moves processed on the fast path.  Use
20572         LRA_UNKNOWN_ALT.
20573         (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
20574         * lra-eliminations.c (spill_pseudos): Ditto.
20575         (process_insn_for_elimination): Ditto.
20576         * lra-lives.c (reg_early_clobber_p): Use the new macros.
20577         * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
20578         LRA_NON_CLOBBERED_ALT.
20580 2018-02-22  Martin Sebor  <msebor@redhat.com>
20582         PR tree-optimization/84480
20583         * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
20584         to maybe_diag_stxncpy_trunc.  Call it.
20585         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
20586         from gimple_fold_builtin_strcpy.  Print inlining stack.
20587         (handle_builtin_stxncpy): Print inlining stack.
20588         * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
20590 2018-02-22  H.J. Lu  <hongjiu.lu@intel.com>
20592         PR target/84176
20593         * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
20594         error when -mindirect-branch=thunk-extern, -fcf-protection=branch
20595         and -fcheck-pointer-bounds are used together.
20596         (indirect_thunk_prefix): New enum.
20597         (indirect_thunk_need_prefix): New function.
20598         (indirect_thunk_name): Replace need_bnd_p with need_prefix.  Use
20599         "_nt" instead of "_bnd" for NOTRACK prefix.
20600         (output_indirect_thunk): Replace need_bnd_p with need_prefix.
20601         (output_indirect_thunk_function): Likewise.
20602         (): Likewise.
20603         (ix86_code_end): Update output_indirect_thunk_function calls.
20604         (ix86_output_indirect_branch_via_reg): Replace
20605         ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
20606         (ix86_output_indirect_branch_via_push): Likewise.
20607         (ix86_output_function_return): Likewise.
20608         * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
20609         incompatible with -fcf-protection=branch and
20610         -fcheck-pointer-bounds.
20612 2018-02-22  Steve Ellcey  <sellcey@cavium.com>
20614         PR target/83335
20615         * config/aarch64/aarch64.c (aarch64_print_address_internal):
20616         Change gcc_assert call to output_operand_lossage.
20618 2018-02-22  Steve Ellcey  <sellcey@cavium.com>
20620         * doc/extend.texi (__builtin_extend_pointer): Document builtin.
20622 2018-02-22  DJ Delorie  <dj@redhat.com>
20623             Sebastian Perta  <sebastian.perta@renesas.com>
20624             Oleg Endo  <olegendo@gcc.gnu.org>
20626         * config/rx/rx.c (rx_rtx_costs): New function.
20627         (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
20629 2018-02-22  Thomas Preud'homme  <thomas.preudhomme@arm.com>
20631         * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
20633 2018-02-22  Martin Liska  <mliska@suse.cz>
20635         PR driver/83193
20636         * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
20637         Add "native" as a possible value.
20639 2018-02-22  Martin Liska  <mliska@suse.cz>
20641         PR driver/83193
20642         * config/i386/i386.c (ix86_option_override_internal):
20643         Add "native" as a possible value for -march and -mtune.
20645 2018-02-22  Jakub Jelinek  <jakub@redhat.com>
20647         PR target/84502
20648         * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
20649         to all type variants.
20651         PR tree-optimization/84503
20652         * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
20653         width as info->bitpos + info->bitsize - start.
20654         (merged_store_group::merge_overlapping): Simplify width computation.
20655         (check_no_overlap): New function.
20656         (imm_store_chain_info::try_coalesce_bswap): Compute expected
20657         start + width and last_order of the group, fail if check_no_overlap
20658         fails.
20659         (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
20660         to group if check_no_overlap fails.
20662 2018-02-21  Segher Boessenkool  <segher@kernel.crashing.org>
20664         * config/rs6000/altivec.md: Delete contraint arguments to
20665         define_expand, define_split, and define_peephole2, and in
20666         define_insn_and_split if always unused.
20667         * config/rs6000/darwin.md: Ditto.
20668         * config/rs6000/dfp.md: Ditto.
20669         * config/rs6000/rs6000.md: Ditto.
20670         * config/rs6000/sync.md: Ditto.
20671         * config/rs6000/vector.md: Ditto.
20672         * config/rs6000/vsx.md: Ditto.
20674 2018-02-21  Segher Boessenkool  <segher@kernel.crashing.org>
20676         * config/rs6000/altivec.md: Write output control strings as braced
20677         blocks instead of double-quoted strings.
20678         * config/rs6000/darwin.md: Ditto.
20679         * config/rs6000/rs6000.md: Ditto.
20680         * config/rs6000/vector.md: Ditto.
20681         * config/rs6000/vsx.md: Ditto.
20683 2018-02-21  Jason Merrill  <jason@redhat.com>
20685         PR c++/84314 - ICE with templates and fastcall attribute.
20686         * attribs.c (build_type_attribute_qual_variant): Remove assert.
20688 2018-02-21  Jan Hubicka  <hubicka@ucw.cz>
20690         * ipa-cp.c (determine_versionability): Fix comment typos.
20692 2018-02-21  Jan Hubicka  <hubicka@ucw.cz>
20694         PR c/84229
20695         * ipa-cp.c (determine_versionability): Do not version functions caling
20696         va_arg_pack.
20698 2018-02-21  Martin Liska  <mliska@suse.cz>
20700         PR driver/83193
20701         * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
20702         Add "native" as a possible value.
20703         * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT):  Define
20704         the macro when native cpu detection is available.
20706 2018-02-21  Martin Liska  <mliska@suse.cz>
20708         PR driver/83193
20709         * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
20710         Add "native" as a possible value.
20711         * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
20712         when native cpu detection is available.
20714 2018-02-21  Jakub Jelinek  <jakub@redhat.com>
20715             Martin Sebor  <msebor@redhat.com>
20717         PR tree-optimization/84478
20718         * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
20719         false.
20720         * gimple-fold.c (get_range_strlen): Make minlen const and assume it
20721         can't be NULL.  Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
20722         support which is conservatively correct, for 2 only stay conservative
20723         for maxlen.  Formatting and comment capitalization fixes.  Add STRICT
20724         argument to the 2 argument get_range_strlen, adjust 6 arg
20725         get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
20726         false.
20727         (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
20728         (gimple_fold_builtin_strlen): Pass true as last argument to
20729         get_range_strlen.
20731 2018-02-20  Martin Sebor  <msebor@redhat.com>
20733         PR middle-end/84095
20734         * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
20735         (builtin_memref::set_base_and_offset): Same.  Handle inner references.
20736         (builtin_memref::builtin_memref): Factor out parts into
20737         set_base_and_offset and call it.
20739 2018-02-20  Richard Sandiford  <richard.sandiford@linaro.org>
20741         PR middle-end/84406
20742         * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
20743         is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
20744         greater precision.  If to_mode is a MODE_PARTIAL_INT, stop the
20745         search at the associated MODE_INT.
20747 2018-02-20  Jeff Law  <law@redhat.com>
20749         PR middle-end/82123
20750         PR tree-optimization/81592
20751         PR middle-end/79257
20752         * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
20753         for range data rather than using global data.
20754         * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
20755         range data rather than using global data.
20756         * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
20757         pass it to children as needed.
20758         (struct directive::fmtresult): Similarly.
20759         (struct directive::set_width): Similarly.
20760         (struct directive::set_precision): Similarly.
20761         (format_integer, format_directive, parse_directive): Similarly.
20762         (format_none): Accept unnamed vr_values parameter.
20763         (format_percent, format_floating, format_character): Similarly.
20764         (format_string, format_plain): Similarly.
20765         * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
20766         the EVRP range analyzer for range data rather than using global data.
20767         * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
20768         gimple-ssa-evrp-analyze.h
20769         (class sprintf_dom_walker): Add after_dom_children member function.
20770         Add evrp_range_analyzer member.
20771         (sprintf_dom_walker::before_dom_children): Call into the EVRP
20772         range analyzer as needed.
20773         (sprintf_dom_walker::after_dom_children): New member function.
20774         * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
20775         if not optimizing.
20776         (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
20777         (evrp_range_analyzer::pop_to_marker): Likewise.
20779 2018-02-20  Richard Sandiford  <richard.sandiford@linaro.org>
20781         PR tree-optimization/84419
20782         * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
20783         with the required type if its current type is compatible but
20784         different.
20786 2018-02-20  Jakub Jelinek  <jakub@redhat.com>
20788         PR middle-end/82004
20789         * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
20790         after vectorization.
20792 2018-02-20  Martin Liska  <mliska@suse.cz>
20794         PR driver/83193
20795         * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
20796         possible values if we don't have a hint.
20798 2018-02-20  Martin Liska  <mliska@suse.cz>
20800         PR c/84310
20801         PR target/79747
20802         * final.c (shorten_branches): Build align_tab array with one
20803         more element.
20804         * opts.c (finish_options): Add alignment option limit check.
20805         (MAX_CODE_ALIGN): Likewise.
20806         (MAX_CODE_ALIGN_VALUE): Likewise.
20807         * doc/invoke.texi: Document maximum allowed option value for
20808         all -falign-* options.
20810 2018-02-19  Jakub Jelinek  <jakub@redhat.com>
20812         PR target/84146
20813         * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
20814         * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
20815         * var-tracking.c (emit_note_insn_var_location): Remove all references
20816         to NOTE_INSN_CALL_ARG_LOCATION.
20817         (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
20818         the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
20819         Use copy_rtx_if_shared.
20820         * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
20821         NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
20822         (dwarf2out_var_location): Remove handling of
20823         NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
20824         on call_insn.
20825         * final.c (final_scan_insn): Remove all references to
20826         NOTE_INSN_CALL_ARG_LOCATION.
20827         (rest_of_clean_state): Likewise.  Remove REG_CALL_ARG_LOCATION notes
20828         before dumping final insns.
20829         * except.c (emit_note_eh_region_end): Remove all references to
20830         NOTE_INSN_CALL_ARG_LOCATION.
20831         * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
20832         * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
20833         * config/arc/arc.c (hwloop_optimize): Likewise.
20834         * config/arm/arm.c (create_fix_barrier): Likewise.
20835         * config/s390/s390.c (s390_chunkify_start): Likewise.
20836         * config/sh/sh.c (find_barrier): Likewise.
20837         * config/i386/i386.c (rest_of_insert_endbranch,
20838         ix86_seh_fixup_eh_fallthru): Likewise.
20839         * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
20840         * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
20841         * config/frv/frv.c (frv_function_prologue): Likewise.
20842         * emit-rtl.c (try_split): Likewise.  Copy over REG_CALL_ARG_LOCATION
20843         reg note.
20844         (note_outside_basic_block_p): Remove all references to
20845         NOTE_INSN_CALL_ARG_LOCATION.
20846         * gengtype.c (adjust_field_rtx_def): Likewise.
20847         * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
20848         Likewise.
20849         * jump.c (cleanup_barriers, delete_related_insns): Likewise.
20850         * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
20852         PR c++/84444
20853         * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
20854         is ADDR_EXPR.
20856         PR tree-optimization/84452
20857         * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
20858         expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
20859         is NULL.
20861 2018-02-19  Martin Liska  <mliska@suse.cz>
20863         PR sanitizer/82183
20864         * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
20866 2018-02-19  Martin Liska  <mliska@suse.cz>
20867             Richard Sandiford  <richard.sandiford@linaro.org>
20869         PR tree-optimization/82491
20870         * gimple-fold.c (get_base_constructor): Make earlier bail out
20871         to prevent ubsan.
20873 2018-02-19  Carl Love  <cel@us.ibm.com>
20875         * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
20876         BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
20877         BU_P8V_OVERLOAD_1.
20878         * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
20879         P8V_BUILTIN_VEC_NEG.
20881 2018-02-19  Sebastian Perta  <sebastian.perta@renesas.com>
20883         * config/rl78/rl78.md (movdf): New define expand.
20885 2018-02-19  Martin Liska  <mliska@suse.cz>
20887         PR other/80589
20888         * doc/invoke.texi: Fix typo.
20889         * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
20891 2018-02-18  Segher Boessenkool  <segher@kernel.crashing.org>
20893         * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
20894         handle rs6000_single_float and rs6000_double_float specially for
20895         e500 family CPUs.
20897 2018-02-16  Jeff Law  <law@redhat.com>
20899         * config/rx/rx.c (add_pop_cfi_notes): New function.;
20900         (pop_regs): Use it.
20902 2018-02-16  Jakub Jelinek  <jakub@redhat.com>
20904         PR ipa/84425
20905         * ipa-inline.c (inline_small_functions): Fix a typo.
20907 2018-02-16  Nathan Sidwell  <nathan@acm.org>
20909         * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
20911 2018-02-16  Carl Love  <cel@us.ibm.com>
20913         * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
20914         Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
20915         from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
20916         * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
20917         expansion to P8V_BUILTIN_VEC_FLOAT2.
20919 2018-02-16  Vladimir Makarov  <vmakarov@redhat.com>
20921         PR rtl-optimization/70023
20922         * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
20923         src_regno into account.
20925 2018-02-16  Carl Love  <cel@us.ibm.com>
20927         * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
20928         * config/rs6000/rs6000-builtin.def: Remove macro expansion for
20929         VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
20930         * config/rs6000/rs6000.c: Remove case statements for
20931         P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
20932         P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
20933         and P9V_BUILTIN_VEC_VINSERT4B.
20934         * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
20935         P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
20936         * config/rs6000/vsx.md:
20937         * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
20938         vec_insert4b.
20940 2018-02-16  Carl Love  <cel@us.ibm.com>
20942         * config/rs6000/altivec.h: Add builtin names vec_extract4b
20943         vec_insert4b.
20944         * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
20945         definitions.
20946         * config/rs6000/rs6000-c.c: Add the definitions for
20947         P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
20948         * config/rs6000/rs6000.c (altivec_expand_builtin): Add
20949         P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
20950         * config/rs6000/vsx.md: Add define_insn extract4b.  Add define_expand
20951         definition for insert4b and define insn *insert3b_internal.
20952         * doc/extend.texi: Add documentation for vec_extract4b.
20954 2018-02-16  Nathan Sidwell  <nathan@acm.org>
20956         * doc/extend.texi (Backwards Compatibility): Mention friend
20957         injection.  Note for-scope is deprecated.
20958         * doc/invoke.texi (-ffriend-injection): Deprecate.
20960 2018-02-16  Segher Boessenkool  <segher@kernel.crashing.org>
20962         * combine.c (try_combine): When adjusting LOG_LINKS for the destination
20963         that moved to I2, also allow destinations that are a paradoxical
20964         subreg (instead of a normal reg).
20966 2018-02-16  Oleg Endo  <olegendo@gcc.gnu.org>
20968         PR target/83831
20969         * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
20970         to QImode.
20972 2018-02-16  Richard Biener  <rguenther@suse.de>
20974         PR tree-optimization/84037
20975         PR tree-optimization/84016
20976         PR target/82862
20977         * config/i386/i386.c (ix86_builtin_vectorization_cost):
20978         Adjust vec_construct for the fact we need additional higher latency
20979         128bit inserts for AVX256 and AVX512 vector builds.
20980         (ix86_add_stmt_cost): Scale vector construction cost for
20981         elementwise loads.
20983 2018-02-16  Richard Biener  <rguenther@suse.de>
20985         PR tree-optimization/84417
20986         * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
20987         the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
20988         (non_rewritable_lvalue_p): Likewise, use poly-ints.
20990 2018-02-16  Martin Liska  <mliska@suse.cz>
20992         PR sanitizer/84307
20993         * internal-fn.def (ASAN_CHECK): Set proper flags.
20994         (ASAN_MARK): Likewise.
20996 2018-02-16  Julia Koval  <julia.koval@intel.com>
20998         * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
20999         from PTA_CANNONLAKE.
21001 2018-02-16  Jakub Jelinek  <jakub@redhat.com>
21003         PR target/84272
21004         * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
21005         Use ++iter rather than iter++ for std::list iterators.
21006         (func_fma_steering::dfs): Likewise.  Don't delete nodes right away,
21007         defer deleting them until all nodes in the forest are processed.  Do
21008         free even leaf nodes.  Change to_process into auto_vec.
21010         PR bootstrap/84405
21011         * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
21012         * vec.h (vec_default_construct): Use memset instead of placement new
21013         if BROKEN_VALUE_INITIALIZATION is defined.
21014         * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
21015         memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
21016         is defined.
21018         PR rtl-optimization/83723
21019         * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
21020         * lra.c (lra_substitute_pseudo): Likewise.  If true, use
21021         gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG.  Pass DEBUG_P to
21022         recursive calls.
21023         (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
21024         callers.
21025         * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
21027 2018-02-16  Eric Botcazou  <ebotcazou@adacore.com>
21029         PR rtl-optimization/81443
21030         * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
21031         from inner REGs to paradoxical SUBREGs.
21033 2018-02-16  Richard Biener  <rguenther@suse.de>
21035         PR tree-optimization/84399
21036         * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
21037         For operands we can analyze at their definition make sure we can
21038         analyze them at each use as well.
21040 2018-02-16  Richard Biener  <rguenther@suse.de>
21042         PR tree-optimization/84190
21043         * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
21044         volatile accesses if the decl isn't volatile.
21046 2018-02-15  Jason Merrill  <jason@redhat.com>
21048         PR c++/84314 - ICE with templates and fastcall attribute.
21049         * attribs.c (build_type_attribute_qual_variant): Don't clobber
21050         TYPE_CANONICAL on an existing type.
21052 2018-02-15  Jakub Jelinek  <jakub@redhat.com>
21054         PR tree-optimization/84383
21055         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
21056         dstoff nor call operand_equal_p if dstbase is NULL.
21058         PR tree-optimization/84334
21059         * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
21060         also a CONSTANT_CLASS_P, punt.
21062 2018-02-14  Jim Wilson  <jimw@sifive.com>
21064         * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
21065         first SMALL_OPERAND check.  New local min_second_step.  Move assert
21066         to where locals are set.  Add TARGET_RVC support.
21067         * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
21069 2018-02-14  Indu Bhagat  <indu.bhagat@oracle.com>
21071         * doc/invoke.texi: Correct -Wformat-overflow code sample.
21073 2018-02-14  Martin Sebor  <msebor@redhat.com>
21075         PR tree-optimization/83698
21076         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
21077         arrays constrain the offset range to their bounds.
21078         (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
21079         (builtin_access::overlap): Avoid setting the size of overlap if it's
21080         already been set.
21081         (maybe_diag_overlap): Also consider arrays when deciding what values
21082         of offsets to include in diagnostics.
21084 2018-02-14  Martin Sebor  <msebor@redhat.com>
21086         PR c/84108
21087         * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
21088         that correspond to the kind of a declaration.
21090 2018-02-14  John David Anglin  <danglin@gcc.gnu.org>
21092         PR target/83984
21093         * config/pa/pa.md: Load address of PIC label using the linkage table
21094         if the label is nonlocal.
21096 2018-02-14  Kelvin Nilsen  <kelvin@gcc.gnu.org>
21098         * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
21099         warning message if user requests -maltivec=be.
21100         * doc/invoke.texi: Document deprecation of -maltivec=be.
21102 2018-02-14  Will Schmidt  <will_schmidt@vnet.ibm.com>
21104         PR target/84220
21105         * config/rs6000/rs6000-c.c: Update definitions for
21106         ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
21107         VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
21109 2018-02-14  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
21111         PR target/84239
21112         * config/i386/cetintrin.h: Remove _rdssp[d|q] and
21113         add _get_ssp intrinsics. Remove argument from
21114         __builtin_ia32_rdssp[d|q].
21115         * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
21116         * config/i386/i386-builtin.def: Remove argument from
21117         __builtin_ia32_rdssp[d|q].
21118         * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
21119         ix86_expand_special_args_builtin for _rdssp[d|q].
21120         * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
21121         Clear register before usage.
21122         * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
21123         Add documentation for new _get_ssp and _inc_ssp intrinsics.
21125 2018-02-14  Richard Sandiford  <richard.sandiford@linaro.org>
21127         PR tree-optimization/84357
21128         * tree-data-ref.c (object_address_invariant_in_loop_p): Check
21129         operand 1 of an ARRAY_REF too.
21131 2018-02-14  Oleg Endo  <olegendo@gcc.gnu.org>
21133         PR target/83831
21134         * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
21135         rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
21136         declarations.
21137         (set_of_reg): New struct.
21138         (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
21139         * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
21140         rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
21141         functions.
21142         * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
21143         Split into bitclr, bitset, bitinvert patterns if appropriate.
21144         (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
21145         use rx_fuse_in_memory_bitop.
21146         (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
21147         to named insn, correct maximum insn length.
21149 2018-02-14  Jozef Lawrynowicz  <jozefl.gcc@gmail.com>
21151         PR target/79242
21152         * machmode.def: Define a complex mode for PARTIAL_INT.
21153         * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
21154         MODE_PARTIAL_INT.
21155         * doc/rtl.texi: Document CSPImode.
21156         * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
21157         handling.
21158         (msp430_hard_regno_nregs_with_padding): Likewise.
21160 2018-02-13  Peter Bergner  <bergner@vnet.ibm.com>
21162         PR target/84279
21163         * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
21165 2018-02-13  Segher Boessenkool  <segher@kernel.crashing.org>
21167         PR rtl-optimization/84169
21168         * combine.c (try_combine): New variable split_i2i3.  Set it to true if
21169         we generated a parallel as new i3 and we split that to new i2 and i3
21170         instructions.  Handle split_i2i3 similar to swap_i2i3: scan the
21171         LOG_LINKs of i3 to see which of those need to link to i2 now.  Link
21172         those to i2, not i1.  Partially rewrite this scan code.
21174 2018-02-13  Jakub Jelinek  <jakub@redhat.com>
21176         PR c/82210
21177         * stor-layout.c (place_field): For variable length fields, adjust
21178         offset_align afterwards not just based on the field's alignment,
21179         but also on the size.
21181         PR middle-end/84309
21182         * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
21183         of exps and logs in the use_exp2 case.
21185 2018-02-13  Jeff Law  <law@redhat.com>
21187         * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
21188         entry for "vector".
21190         * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
21191         ARGS as unused.
21193 2018-02-13  Alexandre Oliva  <aoliva@redhat.com>
21195         PR debug/84342
21196         PR debug/84319
21197         * common.opt (gas-loc-support, gas-locview-support): New.
21198         (ginline-points, ginternal-reset-location-views): New.
21199         * doc/invoke.texi: Document them.  Use @itemx where intended.
21200         (gvariable-location-views): Adjust.
21201         * target.def (reset_location_view): New.
21202         * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
21203         (TARGET_RESET_LOCATION_VIEW): New.
21204         * doc/tm.texi: Rebuilt.
21205         * dwarf2out.c (dwarf2out_default_as_loc_support): New.
21206         (dwarf2out_default_as_locview_support): New.
21207         (output_asm_line_debug_info): Use option variables.
21208         (dwarf2out_maybe_output_loclist_view_pair): Likewise.
21209         (output_loc_list): Likewise.
21210         (add_high_low_attributes): Check option variables.
21211         Don't output entry view attribute in strict mode.
21212         (gen_inlined_subroutine_die): Check option variables.
21213         (dwarf2out_inline_entry): Likewise.
21214         (init_sections_and_labels): Likewise.
21215         (dwarf2out_early_finish): Likewise.
21216         (maybe_reset_location_view): New, from...
21217         (dwarf2out_var_location): ... here.  Call it.
21218         * debug.h (dwarf2out_default_as_loc_support): Declare.
21219         (dwarf2out_default_as_locview_support): Declare.
21220         * hooks.c (hook_int_rtx_insn_0): New.
21221         * hooks.h (hook_int_rtx_insn_0): Declare.
21222         * toplev.c (process_options): Take -gas-loc-support and
21223         -gas-locview-support from dwarf2out.  Enable
21224         -gvariable-location-views by default only with locview
21225         assembler support.  Enable -ginternal-reset-location-views by
21226         default only if the target defines the corresponding hook.
21227         Enable -ginline-points by default if location views are
21228         enabled; force it disabled if statement frontiers are
21229         disabled.
21230         * tree-inline.c (expand_call_inline): Check option variables.
21231         * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
21233 2018-02-13  Richard Sandiford  <richard.sandiford@linaro.org>
21235         PR tree-optimization/84321
21236         * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
21237         handling.  Also check whether the anti-range contains any values
21238         that satisfy the mask; switch to a VR_RANGE if not.
21240 2018-02-13  Paolo Bonzini  <bonzini@gnu.org>
21242         PR sanitizer/84340
21243         * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
21245 2018-02-13  Martin Jambor  <mjambor@suse.cz>
21247         PR c++/83990
21248         * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
21249         of call statements, also set location of a load to a temporary.
21251 2018-02-13  Sebastian Perta  <sebastian.perta@renesas.com>
21253         * config/rl78/rl78.c (add_vector_labels): New function.
21254         * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
21255         * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
21256         * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
21257         which checks that no arguments are passed.
21258         * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
21259         * doc/extend.texi: Documentation for the new attribute.
21261 2018-02-13  Andreas Schwab  <schwab@suse.de>
21263         * config/riscv/linux.h (CPP_SPEC): Define.
21265 2018-02-13  Jakub Jelinek  <jakub@redhat.com>
21267         PR target/84335
21268         * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
21269         OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
21270         OPTION_MASK_ISA_AES as first argument to def_builtin_const
21271         for AES builtins.  Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
21272         instead of OPTION_MASK_ISA_PCLMUL as first argument to
21273         def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
21274         * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
21275         temporarily for AES and PCLMUL builtins.
21277         PR tree-optimization/84339
21278         * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
21279         ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
21280         Formatting fixes.
21282         PR middle-end/84309
21283         * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
21284         exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
21285         * generic-match-head.c (canonicalize_math_after_vectorization_p): New
21286         inline function.
21287         * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
21288         inline function.
21289         * omp-simd-clone.h: New file.
21290         * omp-simd-clone.c: Include omp-simd-clone.h.
21291         (expand_simd_clones): No longer static.
21292         * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
21293         cgraph.h and omp-simd-clone.h.
21294         (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
21295         (vect_recog_widen_shift_pattern): Formatting fix.
21296         (vect_pattern_recog_1): Don't check optab for calls.
21298         PR target/84336
21299         * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
21300         operands[2] into a REG before using gen_lowpart on it.
21302 2018-02-12  Jeff Law  <law@redhat.com>
21304         PR target/83760
21305         * config/sh/sh.c (find_barrier): Consider a sibling call
21306         a barrier as well.
21308         * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
21309         successfully back substituting a reg.
21311 2018-02-12  Richard Biener  <rguenther@suse.de>
21313         PR tree-optimization/84037
21314         * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
21315         parameter, move visited init to caller.
21316         (vect_slp_analyze_operations): Separate cost from validity
21317         check, initialize visited once for all instances.
21318         (vect_schedule_slp): Analyze map to CSE vectorized nodes once
21319         for all instances.
21320         * tree-vect-stmts.c (vect_model_simple_cost): Make early
21321         out an assert.
21322         (vect_model_promotion_demotion_cost): Likewise.
21323         (vectorizable_bswap): Guard cost modeling with !slp_node
21324         instead of !PURE_SLP_STMT to avoid double-counting on hybrid
21325         SLP stmts.
21326         (vectorizable_call): Likewise.
21327         (vectorizable_conversion): Likewise.
21328         (vectorizable_assignment): Likewise.
21329         (vectorizable_shift): Likewise.
21330         (vectorizable_operation): Likewise.
21331         (vectorizable_store): Likewise.
21332         (vectorizable_load): Likewise.
21333         (vectorizable_condition): Likewise.
21334         (vectorizable_comparison): Likewise.
21336 2018-02-12  Paolo Bonzini  <bonzini@gnu.org>
21338         PR sanitizer/84307
21339         * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
21340         (ASAN_MARK): Fix fnspec to account for return value, change pointer
21341         argument from 'R' to 'W' so that the pointed-to datum is clobbered.
21343 2018-02-08  Jan Hubicka  <hubicka@ucw.cz>
21345         PR middle-end/83665
21346         * params.def (inline-min-speedup): Increase from 8 to 15.
21347         (max-inline-insns-auto): Decrease from 40 to 30.
21348         * ipa-split.c (consider_split): Add some buffer for function to
21349         be considered inlining candidate.
21350         * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
21351         default values.
21353 2018-02-12  Richard Biener  <rguenther@suse.de>
21355         PR tree-optimization/84037
21356         * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
21357         matched stmts if we cannot swap the non-matched ones.
21359 2018-02-12  Olga Makhotina  <olga.makhotina@intel.com>
21361         * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
21362         _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
21363         _mm_maskz_scalef_round_ss): New intrinsics.
21364         (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
21365         * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
21366         __builtin_ia32_scalefss_round): Remove.
21367         (__builtin_ia32_scalefsd_mask_round,
21368         __builtin_ia32_scalefss_mask_round): New intrinsics.
21369         * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
21370         (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
21371         ((match_operand:VF_128 2 "<round_nimm_predicate>"
21372         "<round_constraint>")): Changed to ...
21373         ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
21374         "<round_scalar_constraint>")): ... this.
21375         ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
21376         %0, %1, %2<round_op3>}"): Changed to ...
21377         ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
21378         %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
21379         %2<round_scalar_mask_op3>}"): ... this.
21380         * config/i386/subst.md (round_scalar_nimm_predicate): New.
21382 2018-02-12  Olga Makhotina  <olga.makhotina@intel.com>
21384         * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
21385         (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
21386         (_mm_maskz_sqrt_round_ss): New intrinsics.
21387         (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
21388         (__builtin_ia32_sqrtsd_mask_round)
21389         (__builtin_ia32_sqrtss_mask_round): New builtins.
21390         * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
21391         (__builtin_ia32_sqrtss_round): Remove.
21392         (__builtin_ia32_sqrtsd_mask_round)
21393         (__builtin_ia32_sqrtss_mask_round): New builtins.
21394         * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
21395         (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
21396         ((match_operand:VF_128 1 "vector_operand"
21397         "xBm,<round_constraint>")): Changed to ...
21398         ((match_operand:VF_128 1 "vector_operand"
21399         "xBm,<round_scalar_constraint>")): ... this.
21400         (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
21401         %0, %2, %<iptr>1<round_op3>}): Changed to ...
21402         (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
21403         %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
21404         %<iptr>1<round_scalar_mask_op3>}): ... this.
21405         ((set_attr "prefix" "<round_prefix>")): Changed to ...
21406         ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
21408 2018-02-11  Steven Munroe  <munroesj@gcc.gnu.org>
21410         PR target/84266
21411         * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
21412         Cast vec_cmpeq result to correct type.
21413         * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
21414         Cast vec_cmpgt result to correct type.
21416 2018-02-11  Alexandre Oliva  <aoliva@redhat.com>
21418         * final.c (final_scan_insn_1): Renamed from...
21419         (final_scan_insn): ... this.  New wrapper, to recover
21420         seen from the outermost call in recursive ones.
21421         * config/sparc/sparc.c (output_return): Drop seen from call.
21422         (output_sibcall): Likewise.
21423         * config/visium/visium.c (output_branch): Likewise.
21425 2018-02-10  John David Anglin  <danglin@gcc.gnu.org>
21427         * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
21428         function label.
21430 2018-02-10  Alan Modra  <amodra@gmail.com>
21432         PR target/84300
21433         * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
21434         Specify LR as an input.
21436 2018-02-10  Jakub Jelinek  <jakub@redhat.com>
21438         PR sanitizer/83987
21439         * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
21440         remove_member_access_dummy_vars): New functions.
21441         (lower_omp_for, lower_omp_taskreg, lower_omp_target,
21442         lower_omp_1, execute_lower_omp): Use them.
21444         PR rtl-optimization/84308
21445         * shrink-wrap.c (spread_components): Release todo vector.
21447 2018-02-09  Vladimir Makarov  <vmakarov@redhat.com>
21449         PR rtl-optimization/57193
21450         * ira-color.c (struct allocno_color_data): Add member
21451         conflict_allocno_hard_prefs.
21452         (update_conflict_allocno_hard_prefs): New.
21453         (bucket_allocno_compare_func): Add a preference based on
21454         conflict_allocno_hard_prefs.
21455         (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
21456         (color_allocnos): Remove a dead code.  Initiate
21457         conflict_allocno_hard_prefs.  Call update_costs_from_prefs.
21459 2018-02-09  Jakub Jelinek  <jakub@redhat.com>
21461         PR target/84226
21462         * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
21463         constraint from =wa to wa.  Avoid a subreg on the output operand,
21464         instead use a pseudo and subreg it in a move.
21465         (p9_xxbrd_<mode>): Changed to ...
21466         (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
21467         (p9_xxbrd_v2df): New expander.
21468         (p9_xxbrw_<mode>): Changed to ...
21469         (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
21470         (p9_xxbrw_v4sf): New expander.
21472 2018-02-09  Sebastian Perta  <sebastian.perta@renesas.com>
21474         * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
21476 2018-02-09  Peter Bergner  <bergner@vnet.ibm.com>
21478         PR target/83926
21479         * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
21480         multiply in 32-bit mode.
21481         (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
21482         (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
21483         mode.
21485 2018-02-09  Sebastian Perta  <sebastian.perta@renesas.com>
21487         * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
21488         to allow or block "symbol_ref" depending on the value of TARGET_JSR.
21489         * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
21490         * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
21492 2018-02-09  Pierre-Marie de Rodat  <derodat@adacore.com>
21494         PR lto/84213
21495         * dwarf2out.c (is_trivial_indirect_ref): New function.
21496         (dwarf2out_late_global_decl): Do not generate a location
21497         attribute for variables that have a non-trivial DECL_VALUE_EXPR
21498         and that are not defined in the current unit.
21500 2018-02-09  Eric Botcazou  <ebotcazou@adacore.com>
21502         * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
21503         instead of a libcall for UNORDERED.
21505 2018-02-09  Tamar Christina  <tamar.christina@arm.com>
21507         PR target/82641
21508         * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
21509         __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
21511 2018-02-09  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
21513         PR target/PR84295
21514         * config/s390/s390.c (s390_set_current_function): Invoke
21515         s390_indirect_branch_settings also if fndecl didn't change.
21517 2018-02-09  Alexandre Oliva  <aoliva@redhat.com>
21519         * config/rs6000/rs6000.md (blockage): Set length to zero.
21521 2018-02-09  Eric Botcazou  <ebotcazou@adacore.com>
21523         * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
21525 2018-02-09  Jakub Jelinek  <jakub@redhat.com>
21527         PR sanitizer/84285
21528         * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
21529         STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
21530         -static-lib*san.
21532         PR debug/84252
21533         * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
21534         PARALLEL incoming that failed vt_get_decl_and_offset check.
21536         PR middle-end/84237
21537         * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
21538         * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
21539         TREE_READONLY bit.
21540         (get_variable_section): For decls in named .bss* sections pass true as
21541         second argument to bss_initializer_p.
21543 2018-02-09  Marek Polacek  <polacek@redhat.com>
21544             Jakub Jelinek  <jakub@redhat.com>
21546         PR c++/83659
21547         * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
21548         Formatting fixes.  Verify first that tree_fits_poly_int64_p (op01).
21549         Sync some changes from cxx_fold_indirect_ref.
21551 2018-02-09  Alexandre Oliva  <aoliva@redhat.com>
21553         * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
21554         markers.
21555         * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
21556         (BLOCK_INLINE_ENTRY_LABEL): New.
21557         (dwarf2out_var_location): Disregard inline entry markers.
21558         (inline_entry_data): New struct.
21559         (inline_entry_data_hasher): New hashtable type.
21560         (inline_entry_data_hasher::hash): New.
21561         (inline_entry_data_hasher::equal): New.
21562         (inline_entry_data_table): New variable.
21563         (add_high_low_attributes): Add DW_AT_entry_pc and
21564         DW_AT_GNU_entry_view attributes if a pending entry is found
21565         in inline_entry_data_table.  Add old entry_pc attribute only
21566         if debug nonbinding markers are disabled.
21567         (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
21568         markers are enabled.
21569         (block_within_block_p, dwarf2out_inline_entry): New.
21570         (dwarf2out_finish): Check that no entries remained in
21571         inline_entry_data_table.
21572         * final.c (reemit_insn_block_notes): Handle inline entry notes.
21573         (final_scan_insn, notice_source_line): Likewise.
21574         (rest_of_clean_state): Skip inline entry markers.
21575         * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
21576         markers.
21577         * gimple.c (gimple_build_debug_inline_entry): New.
21578         * gimple.h (enum gimple_debug_subcode): Add
21579         GIMPLE_DEBUG_INLINE_ENTRY.
21580         (gimple_build_debug_inline_entry): Declare.
21581         (gimple_debug_inline_entry_p): New.
21582         (gimple_debug_nonbind_marker_p): Adjust.
21583         * insn-notes.def (INLINE_ENTRY): New.
21584         * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
21585         inline entry marker notes.
21586         (print_insn): Likewise.
21587         * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
21588         (INSN_DEBUG_MARKER_KIND): Likewise.
21589         (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
21590         * tree-inline.c (expand_call_inline): Build and insert
21591         debug_inline_entry stmt.
21592         * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
21593         inline entry blocks early, if nonbind markers are enabled.
21594         (dump_scope_block): Dump fragment info.
21595         * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
21596         * doc/gimple.texi (gimple_debug_inline_entry_p): New.
21597         (gimple_build_debug_inline_entry): New.
21598         * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
21599         Enable/disable inline entry points too.
21600         * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
21601         (DEBUG_INSN): Describe inline entry markers.
21603         * common.opt (gvariable-location-views): New.
21604         (gvariable-location-views=incompat5): New.
21605         * config.in: Rebuilt.
21606         * configure: Rebuilt.
21607         * configure.ac: Test assembler for view support.
21608         * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
21609         * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
21610         * dwarf2out.c (var_loc_view): New typedef.
21611         (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
21612         (dwarf2out_locviews_in_attribute): New.
21613         (dwarf2out_locviews_in_loclist): New.
21614         (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
21615         (enum dw_line_info_opcode): Add LI_adv_address.
21616         (struct dw_line_info_table): Add view.
21617         (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
21618         (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
21619         (zero_view_p): New variable.
21620         (ZERO_VIEW_P): New macro.
21621         (output_asm_line_debug_info): New.
21622         (struct var_loc_node): Add view.
21623         (add_AT_view_list, AT_loc_list): New.
21624         (add_var_loc_to_decl): Add view param.  Test it against last.
21625         (new_loc_list): Add view params.  Record them.
21626         (AT_loc_list_ptr): Handle loc and view lists.
21627         (view_list_to_loc_list_val_node): New.
21628         (print_dw_val): Handle dw_val_class_view_list.
21629         (size_of_die): Likewise.
21630         (value_format): Likewise.
21631         (loc_list_has_views): New.
21632         (gen_llsym): Set vl_symbol too.
21633         (maybe_gen_llsym, skip_loc_list_entry): New.
21634         (dwarf2out_maybe_output_loclist_view_pair): New.
21635         (output_loc_list): Output view list or entries too.
21636         (output_view_list_offset): New.
21637         (output_die): Handle dw_val_class_view_list.
21638         (output_dwarf_version): New.
21639         (output_compilation_unit_header): Use it.
21640         (output_skeleton_debug_sections): Likewise.
21641         (output_rnglists, output_line_info): Likewise.
21642         (output_pubnames, output_aranges): Update version comments.
21643         (output_one_line_info_table): Output view numbers in asm comments.
21644         (dw_loc_list): Determine current endview, pass it to new_loc_list.
21645         Call maybe_gen_llsym.
21646         (loc_list_from_tree_1): Adjust.
21647         (add_AT_location_description): Create view list attribute if
21648         needed, check it's absent otherwise.
21649         (convert_cfa_to_fb_loc_list): Adjust.
21650         (maybe_emit_file): Call output_asm_line_debug_info for test.
21651         (dwarf2out_var_location): Reset views as needed.  Precompute
21652         add_var_loc_to_decl args.  Call get_attr_min_length only if we have the
21653         attribute.  Set view.
21654         (new_line_info_table): Reset next view.
21655         (set_cur_line_info_table): Call output_asm_line_debug_info for test.
21656         (dwarf2out_source_line): Likewise.  Output view resets and labels to
21657         the assembler, or select appropriate line info opcodes.
21658         (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
21659         (optimize_string_length): Catch it.  Adjust.
21660         (resolve_addr): Copy vl_symbol along with ll_symbol.  Handle
21661         dw_val_class_view_list, and remove it if no longer needed.
21662         (hash_loc_list): Hash view numbers.
21663         (loc_list_hasher::equal): Compare them.
21664         (optimize_location_lists): Check whether a view list symbol is
21665         needed, and whether the locview attribute is present, and
21666         whether they match.  Remove the locview attribute if no longer
21667         needed.
21668         (index_location_lists): Call skip_loc_list_entry for test.
21669         (dwarf2out_finish): Call output_asm_line_debug_info for test.
21670         Use output_dwarf_version.
21671         * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
21672         (struct dw_val_node): Add val_view_list.
21673         * final.c (SEEN_NEXT_VIEW): New.
21674         (set_next_view_needed): New.
21675         (clear_next_view_needed): New.
21676         (maybe_output_next_view): New.
21677         (final_start_function): Rename to...
21678         (final_start_function_1): ... this.  Take pointer to FIRST,
21679         add SEEN parameter.  Emit param bindings in the initial view.
21680         (final_start_function): Reintroduce SEEN-less interface.
21681         (final): Rename to...
21682         (final_1): ... this.  Take SEEN parameter.  Output final pending
21683         next view at the end.
21684         (final): Reintroduce seen-less interface.
21685         (final_scan_insn): Output pending next view before switching
21686         sections or ending a block.  Mark the next view as needed when
21687         outputting variable locations.  Notify debug backend of section
21688         changes, and of location view changes.
21689         (rest_of_handle_final): Adjust.
21690         * toplev.c (process_options): Autodetect value for debug variable
21691         location views option.  Warn on incompat5 without -gdwarf-5.
21692         * doc/invoke.texi (gvariable-location-views): New.
21693         (gvariable-location-views=incompat5): New.
21694         (gno-variable-location-views): New.
21696 2018-02-08  David Malcolm  <dmalcolm@redhat.com>
21698         PR tree-optimization/84136
21699         * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
21700         that the result of find_edge is non-NULL.
21702 2018-02-08  Sergey Shalnov  <sergey.shalnov@intel.com>
21704         PR target/83008
21705         * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
21706         storing integer register in SImode.  Fix cost of 256 and 512
21707         byte aligned SSE register store.
21709 2018-02-08  Sergey Shalnov  <sergey.shalnov@intel.com>
21711         * config/i386/i386.c (ix86_multiplication_cost): Fix
21712         multiplication cost for TARGET_AVX512DQ.
21714 2018-02-08  Marek Polacek  <polacek@redhat.com>
21716         PR tree-optimization/84238
21717         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
21718         get_range_strlen.
21720 2018-02-08  Richard Sandiford  <richard.sandiford@linaro.org>
21722         PR tree-optimization/84265
21723         * tree-vect-stmts.c (vectorizable_store): Don't treat
21724         VMAT_CONTIGUOUS accesses as grouped.
21725         (vectorizable_load): Likewise.
21727 2018-02-08  Richard Sandiford  <richard.sandiford@linaro.org>
21729         PR tree-optimization/81635
21730         * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
21731         * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
21732         (test_round_for_mask): New functions.
21733         (wide_int_cc_tests): Call test_round_for_mask.
21734         * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
21735         * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
21736         * tree-data-ref.c (split_constant_offset_1): Use it to refine the
21737         range returned by get_range_info.
21739 2018-02-08  Jan Hubicka  <hubicka@ucw.cz>
21741         PR ipa/81360
21742         * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
21743         * symtab.c: Include builtins.h
21744         (symtab_node::output_to_lto_symbol_table_p): Move here
21745         from lto-streamer-out.c:output_symbol_p.
21746         * lto-streamer-out.c (write_symbol): Turn early exit to assert.
21747         (output_symbol_p): Move all logic to symtab.c
21748         (produce_symtab): Update.
21750 2018-02-08  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
21752         * config/s390/s390-opts.h (enum indirect_branch): Define.
21753         * config/s390/s390-protos.h (s390_return_addr_from_memory)
21754         (s390_indirect_branch_via_thunk)
21755         (s390_indirect_branch_via_inline_thunk): Add function prototypes.
21756         (enum s390_indirect_branch_type): Define.
21757         * config/s390/s390.c (struct s390_frame_layout, struct
21758         machine_function): Remove.
21759         (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
21760         (indirect_branch_table_label_no, indirect_branch_table_name):
21761         Define variables.
21762         (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
21763         (enum s390_indirect_branch_option): Define.
21764         (s390_return_addr_from_memory): New function.
21765         (s390_handle_string_attribute): New function.
21766         (s390_attribute_table): Add new attribute handler.
21767         (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
21768         (s390_indirect_branch_via_thunk): New function.
21769         (s390_indirect_branch_via_inline_thunk): New function.
21770         (s390_function_ok_for_sibcall): When jumping via thunk disallow
21771         sibling call optimization for non z10 compiles.
21772         (s390_emit_call): Force indirect branch target to be a single
21773         register.  Add r1 clobber for non-z10 compiles.
21774         (s390_emit_epilogue): Emit return jump via return_use expander.
21775         (s390_reorg): Handle JUMP_INSNs as execute targets.
21776         (s390_option_override_internal): Perform validity checks for the
21777         new command line options.
21778         (s390_indirect_branch_attrvalue): New function.
21779         (s390_indirect_branch_settings): New function.
21780         (s390_set_current_function): Invoke s390_indirect_branch_settings.
21781         (s390_output_indirect_thunk_function):  New function.
21782         (s390_code_end): Implement target hook.
21783         (s390_case_values_threshold): Implement target hook.
21784         (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
21785         macros.
21786         * config/s390/s390.h (struct s390_frame_layout)
21787         (struct machine_function): Move here from s390.c.
21788         (TARGET_INDIRECT_BRANCH_NOBP_RET)
21789         (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
21790         (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
21791         (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
21792         (TARGET_INDIRECT_BRANCH_NOBP_CALL)
21793         (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
21794         (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
21795         (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
21796         (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
21797         * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
21798         (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
21799         (mnemonic attribute): Add values which aren't recognized
21800         automatically.
21801         ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
21802         pattern for branch conversion.  Fix mnemonic attribute.
21803         ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
21804         indirect branch via thunk if requested.
21805         ("indirect_jump", "<code>"): Expand patterns for branch conversion.
21806         ("*indirect_jump"): Disable for branch conversion using out of
21807         line thunks.
21808         ("indirect_jump_via_thunk<mode>_z10")
21809         ("indirect_jump_via_thunk<mode>")
21810         ("indirect_jump_via_inlinethunk<mode>_z10")
21811         ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
21812         ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
21813         ("casesi_jump_via_inlinethunk<mode>_z10")
21814         ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
21815         ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
21816         ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
21817         ("*indirect2_jump"): Disable for branch conversion.
21818         ("casesi_jump"): Turn into expander and expand patterns for branch
21819         conversion.
21820         ("return_use"): New expander.
21821         ("*return"): Emit return via thunk and rename it to ...
21822         ("*return<mode>"): ... this one.
21823         * config/s390/s390.opt: Add new options and and enum for the
21824         option values.
21826 2018-02-08  Richard Sandiford  <richard.sandiford@linaro.org>
21828         * lra-constraints.c (match_reload): Unconditionally use
21829         gen_lowpart_SUBREG, rather than selecting between that
21830         and equivalent gen_rtx_SUBREG code.
21832 2018-02-08  Richard Biener  <rguenther@suse.de>
21834         PR tree-optimization/84233
21835         * tree-ssa-phiprop.c (propagate_with_phi): Use separate
21836         changed flag instead of boguously re-using phi_inserted.
21838 2018-02-08  Martin Jambor  <mjambor@suse.cz>
21840         * hsa-gen.c (get_symbol_for_decl): Set program allocation for
21841         static local variables.
21843 2018-02-08  Richard Biener  <rguenther@suse.de>
21845         PR tree-optimization/84278
21846         * tree-vect-stmts.c (vectorizable_store): When looking for
21847         smaller vector types to perform grouped strided loads/stores
21848         make sure the mode is supported by the target.
21849         (vectorizable_load): Likewise.
21851 2018-02-08  Wilco Dijkstra  <wdijkstr@arm.com>
21853         * config/aarch64/aarch64.c (aarch64_components_for_bb):
21854         Increase LDP/STP opportunities by adding adjacent callee-saves.
21856 2018-02-08  Wilco Dijkstra  <wdijkstr@arm.com>
21858         PR rtl-optimization/84068
21859         PR rtl-optimization/83459
21860         * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
21862 2018-02-08  Aldy Hernandez  <aldyh@redhat.com>
21864         PR tree-optimization/84224
21865         * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
21866         * calls.c (gimple_alloca_call_p): Only return TRUE when we have
21867         non-zero arguments.
21869 2018-02-07  Iain Sandoe  <iain@codesourcery.com>
21871         PR target/84113
21872         * config/rs6000/altivec.md (*restore_world): Remove LR use.
21873         * config/rs6000/predicates.md (restore_world_operation): Adjust op
21874         count, remove one USE.
21876 2018-02-07  Michael Meissner  <meissner@linux.vnet.ibm.com>
21878         * doc/install.texi (Configuration): Document the
21879         --with-long-double-format={ibm,ieee} PowerPC configuration
21880         options.
21882         PR target/84154
21883         * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
21884         Convert from define_expand to be define_insn_and_split.  Rework
21885         float/double/_Float128 conversions to QI/HI/SImode to work with
21886         both ISA 2.07 (power8) or ISA 3.0 (power9).  Fix regression where
21887         conversions to QI/HImode types did a store and then a load to
21888         truncate the value.  For conversions to VSX registers, don't split
21889         the insn, instead emit the code directly.  Use the code iterator
21890         any_fix to combine signed and unsigned conversions.
21891         (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
21892         (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
21893         (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
21894         (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
21895         (fix_<mode>di2_hw): Likewise.
21896         (fixuns_<mode>di2_hw): Likewise.
21897         (fix_<mode>si2_hw): Likewise.
21898         (fixuns_<mode>si2_hw): Likewise.
21899         (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
21900         (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
21901         (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
21902         fix<uns>_trunc<SFDF:mode>si2_p8.
21903         (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer used.
21904         (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
21905         (fix<uns>_<mode>_mem): Likewise.
21906         (fctiw<u>z_<mode>_mem): Likewise.
21907         (fix<uns>_<mode>_mem): Likewise.
21908         (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
21909         the register allocator from doing a direct move to the GPRs to do
21910         a store, and instead use the ISA 3.0 store byte/half-word from
21911         vector register instruction.  For IEEE 128-bit floating point,
21912         also optimize stores of 32-bit ints.
21913         (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
21915 2018-02-07  Alan Hayward  <alan.hayward@arm.com>
21917         * genextract.c (push_pathstr_operand): New function to support
21918         [a-zA-Z].
21919         (walk_rtx): Call push_pathstr_operand.
21920         (print_path): Support [a-zA-Z].
21922 2018-02-07  Richard Biener  <rguenther@suse.de>
21924         PR tree-optimization/84037
21925         * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
21926         (cse_and_gimplify_to_preheader): Declare.
21927         (vect_get_place_in_interleaving_chain): Likewise.
21928         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
21929         ivexpr_map.
21930         (_loop_vec_info::~_loop_vec_info): Delete it.
21931         (cse_and_gimplify_to_preheader): New function.
21932         * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
21933         * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
21934         (vectorizable_load): Likewise.  For grouped stores always base
21935         the IV on the first element.
21936         * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
21937         condition before gimplifying.
21939 2018-02-07  Jakub Jelinek  <jakub@redhat.com>
21941         * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
21942         *DIV_EXPR and *MOD_EXPR.
21944 2018-02-07  H.J. Lu  <hongjiu.lu@intel.com>
21946         PR target/84248
21947         * config/i386/i386.c (ix86_option_override_internal): Mask out
21948         the CF_SET bit when checking -fcf-protection.
21950 2018-02-07  Tom de Vries  <tom@codesourcery.com>
21952         PR libgomp/84217
21953         * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
21954         enough.
21956 2018-02-07  Richard Biener  <rguenther@suse.de>
21958         PR tree-optimization/84204
21959         * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
21960         this place.
21962         PR tree-optimization/84205
21963         * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
21964         special-case isl_ast_op_zdiv_r.
21966         PR tree-optimization/84223
21967         * graphite-scop-detection.c (gather_bbs::before_dom_children):
21968         Only add conditions from within the region.
21969         (gather_bbs::after_dom_children): Adjust.
21971 2018-02-07  Georg-Johann Lay  <avr@gjlay.de>
21973         PR target/84209
21974         * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
21975         * config/avr/avr.md: Only post-reload split REG-REG moves if
21976         either register is GENERAL_REG_P.
21978 2018-02-07  Jakub Jelinek  <jakub@redhat.com>
21980         PR tree-optimization/84235
21981         * tree-ssa-scopedtables.c
21982         (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
21983         if the subtraction is performed in floating point type where NaNs are
21984         honored.  For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
21985         build 1.  Formatting fix.
21987 2018-02-06  Jakub Jelinek  <jakub@redhat.com>
21989         PR target/84146
21990         * config/i386/i386.c (rest_of_insert_endbranch): Only skip
21991         NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
21992         and skip it regardless of bb boundaries.  Use CALL_P macro,
21993         don't test INSN_P (insn) together with CALL_P or JUMP_P check
21994         unnecessarily, formatting fix.
21996 2018-02-06  Michael Collison  <michael.collison@arm.com>
21998         * config/arm/thumb2.md:
21999         (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
22000         (*thumb_mov_notscc): Ditto.
22002 2018-02-06  Michael Meissner  <meissner@linux.vnet.ibm.com>
22004         PR target/84154
22005         * config/rs6000/rs6000.md (su code attribute): Use "u" for
22006         unsigned_fix, not "s".
22008 2018-02-06  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
22010         * configure.ac (gcc_fn_eh_frame_ro): New function.
22011         (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
22012         correct .eh_frame permissions.
22013         * configure: Regenerate.
22015 2018-02-06  Andrew Jenner  <andrew@codeourcery.com>
22017         * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
22018         irrelevant options.
22020 2018-02-06  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
22022         * config/rs6000/rs6000.c (rs6000_option_override_internal):
22023         Display warning message for -mno-speculate-indirect-jumps.
22025 2018-02-06  Andrew Jenner  <andrew@codesourcery.com>
22027         * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
22028         Undocumented.
22029         * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
22031 2018-02-06  Aldy Hernandez  <aldyh@redhat.com>
22033         PR tree-optimization/84225
22034         * tree-eh.c (find_trapping_overflow): Only call
22035         operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
22037 2018-02-06  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
22039         PR target/84145
22040         * config/i386/i386.c: Reimplement the check of possible options
22041         -mibt/-mshstk conbination. Change error messages.
22042         * doc/invoke.texi: Fix a typo: remove extra '='.
22044 2018-02-06  Marek Polacek  <polacek@redhat.com>
22046         PR tree-optimization/84228
22047         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
22049 2018-02-06  Tamar Christina  <tamar.christina@arm.com>
22051         PR target/82641
22052         * config/arm/arm.c (arm_print_asm_arch_directives): Record already
22053         emitted arch directives.
22054         * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
22055         __ARM_FEATURE_COPROC before changing architectures.
22057 2018-02-06  Richard Biener  <rguenther@suse.de>
22059         * config/i386/i386.c (print_reg): Fix typo.
22060         (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
22062 2018-02-06  Eric Botcazou  <ebotcazou@adacore.com>
22064         * configure: Regenerate.
22066 2018-02-05  Martin Sebor  <msebor@redhat.com>
22068         PR tree-optimization/83369
22069         * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
22070         inlining context.
22072 2018-02-05  Martin Liska  <mliska@suse.cz>
22074         * doc/invoke.texi: Cherry-pick upstream r323995.
22076 2018-02-05  Richard Sandiford  <richard.sandiford@linaro.org>
22078         * ira.c (ira_init_register_move_cost): Adjust comment.
22080 2018-02-05  Martin Liska  <mliska@suse.cz>
22082         PR gcov-profile/84137
22083         * doc/gcov.texi: Fix typo in documentation.
22085 2018-02-05  Martin Liska  <mliska@suse.cz>
22087         PR gcov-profile/83879
22088         * doc/gcov.texi: Document necessity of --dynamic-list-data when
22089         using dlopen functionality.
22091 2018-02-05  Olga Makhotina  <olga.makhotina@intel.com>
22093         * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
22094         _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
22095         _mm_maskz_range_ss, _mm_mask_range_round_ss,
22096         _mm_maskz_range_round_ss): New intrinsics.
22097         (__builtin_ia32_rangesd128_round)
22098         (__builtin_ia32_rangess128_round): Remove.
22099         (__builtin_ia32_rangesd128_mask_round,
22100         __builtin_ia32_rangess128_mask_round): New builtins.
22101         * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
22102         __builtin_ia32_rangess128_round): Remove.
22103         (__builtin_ia32_rangesd128_mask_round,
22104         __builtin_ia32_rangess128_mask_round): New builtins.
22105         * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
22106         (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
22107         ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
22108         "<round_saeonly_constraint>")): Changed to ...
22109         ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
22110         "<round_saeonly_scalar_constraint>")): ... this.
22111         ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
22112         %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
22113         ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
22114         %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
22115         %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
22117 2018-02-02  Andrew Jenner  <andrew@codesourcery.com>
22119         * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
22120         options.
22121         * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
22122         Remove all values except native, 8540 and 8548.
22124 2018-02-02  H.J. Lu  <hongjiu.lu@intel.com>
22126         * config/i386/i386.c (ix86_output_function_return): Pass
22127         INVALID_REGNUM, instead of -1, as invalid register number to
22128         indirect_thunk_name and output_indirect_thunk.
22130 2018-02-02  Julia Koval  <julia.koval@intel.com>
22132         * config.gcc: Add -march=icelake.
22133         * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
22134         * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
22135         * config/i386/i386.c (processor_costs): Add m_ICELAKE.
22136         (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
22137         PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
22138         (processor_target_table): Add icelake.
22139         (ix86_option_override_internal): Handle new PTAs.
22140         (get_builtin_code_for_version): Handle icelake.
22141         (M_INTEL_COREI7_ICELAKE): New.
22142         (fold_builtin_cpu): Handle icelake.
22143         * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
22144         * doc/invoke.texi: Add -march=icelake.
22146 2018-02-02  Julia Koval  <julia.koval@intel.com>
22148         * config/i386/i386.c (ix86_option_override_internal): Change flags type
22149         to wide_int_bitmask.
22150         * wide-int-bitmask.h: New.
22152 2018-02-02  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
22154         PR target/84066
22155         * config/i386/i386.md: Replace Pmode with word_mode in
22156         builtin_setjmp_setup and builtin_longjmp to support x32.
22158 2018-02-01  Peter Bergner  <bergner@vnet.ibm.com>
22160         PR target/56010
22161         PR target/83743
22162         * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
22163         #include "opts.h".
22164         (rs6000_supported_cpu_names): New static variable.
22165         (linux_cpu_translation_table): Likewise.
22166         (elf_platform) <cpu>: Define new static variable and use it.
22167         Translate kernel AT_PLATFORM name to canonical name if needed.
22168         Error if platform name is unknown.
22170 2018-02-01  Aldy Hernandez  <aldyh@redhat.com>
22172         PR target/84089
22173         * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
22175 2018-02-01  Jeff Law  <law@redhat.com>
22177         PR target/84128
22178         * config/i386/i386.c (release_scratch_register_on_entry): Add new
22179         OFFSET and RELEASE_VIA_POP arguments.  Use SP+OFFSET to restore
22180         the scratch if RELEASE_VIA_POP is false.
22181         (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
22182         If we have to save a temporary register, decrement SIZE appropriately.
22183         Pass new arguments to release_scratch_register_on_entry.
22184         (ix86_adjust_stack_and_probe): Likewise.
22185         (ix86_emit_probe_stack_range): Pass new arguments to
22186         release_scratch_register_on_entry.
22188 2018-02-01  Uros Bizjak  <ubizjak@gmail.com>
22190         PR rtl-optimization/84157
22191         * combine.c (change_zero_ext): Use REG_P predicate in
22192         front of HARD_REGISTER_P predicate.
22194 2018-02-01  Georg-Johann Lay  <avr@gjlay.de>
22196         * config/avr/avr.c (avr_option_override): Move disabling of
22197         -fdelete-null-pointer-checks to...
22198         * common/config/avr/avr-common.c (avr_option_optimization_table):
22199         ...here.
22201 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
22203         PR tree-optimization/81635
22204         * tree-data-ref.c (split_constant_offset_1): For types that
22205         wrap on overflow, try to use range info to prove that wrapping
22206         cannot occur.
22208 2018-02-01  Renlin Li  <renlin.li@arm.com>
22210         PR target/83370
22211         * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
22212         TAILCALL_ADDR_REGS.
22213         (aarch64_register_move_cost): Likewise.
22214         * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
22215         TAILCALL_ADDR_REGS.
22216         (REG_CLASS_NAMES): Likewise.
22217         (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
22218         TAILCALL_ADDR_REGS. Remove IP registers.
22219         * config/aarch64/aarch64.md (Ucs): Update register constraint.
22221 2018-02-01  Richard Biener  <rguenther@suse.de>
22223         * domwalk.h (dom_walker::dom_walker): Add additional constructor
22224         for specifying RPO order and allow NULL for that.
22225         * domwalk.c (dom_walker::dom_walker): Likewise.
22226         (dom_walker::walk): Handle NULL RPO order.
22227         * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
22228         in RPO order.
22229         (rewrite_update_dom_walker): Likewise.
22230         (mark_def_dom_walker): Likewise.
22232 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
22234         * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
22235         (aarch64_maybe_expand_sve_subreg_move): Declare.
22236         * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
22237         * config/aarch64/predicates.md (aarch64_any_register_operand): New
22238         predicate.
22239         * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
22240         that are semantically a reverse operation.
22241         (*aarch64_sve_mov<mode>_subreg_be): New pattern.
22242         * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
22243         (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
22244         functions.
22245         (aarch64_can_change_mode_class): For big-endian, forbid changes
22246         between two SVE modes if they have different element sizes.
22248 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
22250         * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
22251         the TImode handling for big-endian targets.
22253 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
22255         * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
22256         (*sve_ld1rq<Vesize>): ... this new pattern.  Handle all element sizes,
22257         not just bytes.
22258         * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
22259         Remove BSWAP handing for big-endian targets and use the form of
22260         LD1RQ appropariate for the mode.
22262 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
22264         * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
22265         all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
22266         duplicated element.
22268 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
22270         PR tearget/83845
22271         * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
22272         check for operands that need to go through aarch64_sve_reload_be.
22274 2018-02-01  Jakub Jelinek  <jakub@redhat.com>
22276         PR tree-optimization/81661
22277         PR tree-optimization/84117
22278         * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
22279         * tree-eh.c: Include gimplify.h.
22280         (find_trapping_overflow, replace_trapping_overflow,
22281         rewrite_to_non_trapping_overflow): New functions.
22282         * tree-vect-loop.c: Include tree-eh.h.
22283         (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
22284         * tree-data-ref.c: Include tree-eh.h.
22285         (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
22287 2018-01-31  Uros Bizjak  <ubizjak@gmail.com>
22289         PR rtl-optimization/84123
22290         * combine.c (change_zero_ext): Check if hard register satisfies
22291         can_change_dest_mode before calling gen_lowpart_SUBREG.
22293 2018-01-31  Vladimir Makarov  <vmakarov@redhat.com>
22295         PR target/82444
22296         * ira.c (ira_init_register_move_cost): Remove assert.
22298 2018-01-31  Eric Botcazou  <ebotcazou@adacore.com>
22300         PR rtl-optimization/84071
22301         * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
22302         * doc/tm.texi: Regenerate.
22304 2018-01-31  Richard Biener  <rguenther@suse.de>
22306         PR tree-optimization/84132
22307         * tree-data-ref.c (analyze_miv_subscript): Properly
22308         check whether evolution_function_is_affine_multivariate_p
22309         before calling gcd_of_steps_may_divide_p.
22311 2018-01-31  Julia Koval  <julia.koval@intel.com>
22313         PR target/83618
22314         * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
22315         * config/i386/i386.md (rdpid_rex64) New.
22316         (rdpid): Make 32bit only.
22318 2018-01-29  Aldy Hernandez  <aldyh@redhat.com>
22320         PR lto/84105
22321         * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
22322         an IDENTIFIER_NODE for FUNCTION_TYPE's.
22324 2018-01-31  Eric Botcazou  <ebotcazou@adacore.com>
22326         Revert
22327         2018-01-12  Eric Botcazou  <ebotcazou@adacore.com>
22329         * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
22331 2018-01-31  Eric Botcazou  <ebotcazou@adacore.com>
22333         PR rtl-optimization/84071
22334         * combine.c (record_dead_and_set_regs_1): Record the source unmodified
22335         for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
22337 2018-01-31  Claudiu Zissulescu  <claziss@synopsys.com>
22339         * config/arc/arc.c (arc_handle_aux_attribute): New function.
22340         (arc_attribute_table): Add 'aux' attribute.
22341         (arc_in_small_data_p): Consider aux like variables.
22342         (arc_is_aux_reg_p): New function.
22343         (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
22344         (arc_get_aux_arg): New function.
22345         (prepare_move_operands): Handle aux-register access.
22346         (arc_handle_aux_attribute): New function.
22347         * doc/extend.texi (ARC Variable attributes): Add subsection.
22349 2018-01-31  Claudiu Zissulescu  <claziss@synopsys.com>
22351         * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
22352         * config/arc/arc.c (arc_handle_uncached_attribute): New function.
22353         (arc_attribute_table): Add 'uncached' attribute.
22354         (arc_print_operand): Print '.di' flag for uncached memory
22355         accesses.
22356         (arc_in_small_data_p): Do not consider for small data the uncached
22357         types.
22358         (arc_is_uncached_mem_p): New function.
22359         * config/arc/predicates.md (compact_store_memory_operand): Check
22360         for uncached memory accesses.
22361         (nonvol_nonimm_operand): Likewise.
22362         * doc/extend.texi (ARC Type Attribute): New subsection.
22364 2018-01-31  Jakub Jelinek  <jakub@redhat.com>
22366         PR c/84100
22367         * common.opt (falign-functions=, falign-jumps=, falign-labels=,
22368         falign-loops=): Add Optimization flag.
22370 2018-01-30  Jeff Law  <law@redhat.com>
22372         PR target/84064
22373         * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
22374         INT_REGISTERS_SAVED.  Check it prior to calling
22375         get_scratch_register_on_entry.
22376         (ix86_adjust_stack_and_probe): Similarly.
22377         (ix86_emit_probe_stack_range): Similarly.
22378         (ix86_expand_prologue): Corresponding changes.
22380 2018-01-30  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
22382         PR target/40411
22383         * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
22384         -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
22386 2018-01-30  Vladimir Makarov  <vmakarov@redhat.com>
22388         PR target/84112
22389         * lra-constraints.c (curr_insn_transform): Process AND in the
22390         address.
22392 2018-01-30  Jakub Jelinek  <jakub@redhat.com>
22394         PR rtl-optimization/83986
22395         * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
22396         dependence against last_pending_memory_flush in addition to
22397         pending_jump_insns.
22399 2018-01-30  Alexandre Oliva  <aoliva@redhat.com>
22401         PR tree-optimization/81611
22402         * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
22403         copies.
22405 2018-01-30  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
22407         PR target/83758
22408         * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
22409         a reg rtx.
22411 2018-01-30  Richard Biener  <rguenther@suse.de>
22412             Jakub Jelinek  <jakub@redhat.com>
22414         PR tree-optimization/84111
22415         * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
22416         inner loops added during recursion, as they don't have up-to-date
22417         SSA form.
22419 2018-01-30  Jan Hubicka  <hubicka@ucw.cz>
22421         PR ipa/81360
22422         * ipa-inline.c (can_inline_edge_p): Break out late tests to...
22423         (can_inline_edge_by_limits_p): ... here.
22424         (can_early_inline_edge_p, check_callers,
22425         update_caller_keys, update_callee_keys, recursive_inlining,
22426         add_new_edges_to_heap, speculation_useful_p,
22427         inline_small_functions,
22428         inline_small_functions, flatten_function,
22429         inline_to_all_callers_1): Update.
22431 2018-01-30  Jan Hubicka  <hubicka@ucw.cz>
22433         * profile-count.c (profile_count::combine_with_ipa_count): Handle
22434         zeros correctly.
22436 2018-01-30  Richard Biener  <rguenther@suse.de>
22438         PR tree-optimization/83008
22439         * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
22440         invariant and constant vector uses in stmts when they need
22441         more than one stmt.
22443 2018-01-30  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
22445         PR bootstrap/84017
22446         * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
22447         * configure: Regenerate.
22449 2018-01-30  Richard Sandiford  <richard.sandiford@linaro.org>
22451         * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
22452         pattern.
22453         (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
22454         Use gen_rtx_REG rather than gen_lowpart.
22456 2018-01-30  Richard Sandiford  <richard.sandiford@linaro.org>
22458         * lra-constraints.c (match_reload): Use subreg_lowpart_offset
22459         rather than 0 when creating partial subregs.
22461 2018-01-30  Richard Sandiford  <richard.sandiford@linaro.org>
22463         * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
22464         of usage.
22466 2018-01-29  Michael Meissner  <meissner@linux.vnet.ibm.com>
22468         PR target/81550
22469         * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
22470         and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
22471         -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
22472         flags.  This restores the settings used before the 2017-07-24.
22473         Turning off pre increment/decrement/modify allows IVOPTS to
22474         optimize DF/SF loops where the index is an int.
22476 2018-01-29  Richard Biener  <rguenther@suse.de>
22477             Kelvin Nilsen  <kelvin@gcc.gnu.org>
22479         PR bootstrap/80867
22480         * tree-vect-stmts.c (vectorizable_call): Don't call
22481         targetm.vectorize_builtin_md_vectorized_function if callee is
22482         NULL.
22484 2018-01-22  Carl Love  <cel@us.ibm.com>
22486         * doc/extend.tex: Fix typo in second arg in
22487         __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
22489 2018-01-29  Richard Biener  <rguenther@suse.de>
22491         PR tree-optimization/84086
22492         * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
22493         (flush_ssaname_freelist): When SSA names were released reset
22494         the SCEV hash table.
22496 2018-01-29  Richard Biener  <rguenther@suse.de>
22498         PR tree-optimization/84057
22499         * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
22500         removed paths when removing edges.
22502 2018-01-27  H.J. Lu  <hongjiu.lu@intel.com>
22504         * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
22505         -mfunction-return=@var{choice}.
22507 2018-01-27  Bernd Edlinger  <bernd.edlinger@hotmail.de>
22509         PR diagnostic/84034
22510         * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
22511         Handle CR like TAB.
22512         (layout::print_source_line): Likewise.
22513         (test_get_line_width_without_trailing_whitespace): Add test cases.
22515 2018-01-27  Jakub Jelinek  <jakub@redhat.com>
22517         PR middle-end/84040
22518         * sched-deps.c (sched_macro_fuse_insns): Return immediately for
22519         debug insns.
22521 2018-01-26  Jim Wilson  <jimw@sifive.com>
22523         * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
22525         * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
22526         specified.
22528 2018-01-26  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
22530         * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
22531         and CMP + SUB-immediate -> SUBS.
22533 2018-01-26  Martin Sebor  <msebor@redhat.com>
22535         PR tree-optimization/83896
22536         * tree-ssa-strlen.c (get_string_len): Rename...
22537         (get_string_cst_length): ...to this.  Return HOST_WIDE_INT.
22538         Avoid assuming length is constant.
22539         (handle_char_store): Use HOST_WIDE_INT for string length.
22541 2018-01-26  Uros Bizjak  <ubizjak@gmail.com>
22543         PR target/81763
22544         * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
22545         to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
22547 2018-01-26  Richard Biener  <rguenther@suse.de>
22549         PR rtl-optimization/84003
22550         * dse.c (record_store): Only record redundant stores when
22551         the earlier store aliases at least all accesses the later one does.
22553 2018-01-26  Jakub Jelinek  <jakub@redhat.com>
22555         PR rtl-optimization/83985
22556         * dce.c (deletable_insn_p): Return false for separate shrink wrapping
22557         REG_CFA_RESTORE insns.
22558         (delete_unmarked_insns): Don't ignore separate shrink wrapping
22559         REG_CFA_RESTORE insns here.
22561         PR c/83989
22562         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
22563         use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
22565 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
22567         * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
22568         * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
22569         (arc_init): Likewise.
22570         (arc_override_options): Likewise.
22571         (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
22572         value.
22573         (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
22574         support.
22575         * config/arc/arc.h (TARGET_DBNZ): Define.
22576         * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
22577         properly set the tune attribute.
22578         (dbnz): Use TARGET_DBNZ guard.
22579         * config/arc/arc.opt (mtune): Add core3 option.
22581 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
22583         * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
22584         recognize new pic like addresses.
22585         (arc_delegitimize_address): Clean up.
22587 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
22589         * config/arc/arc-arches.def: Option mrf16 valid for all
22590         architectures.
22591         * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
22592         * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
22593         * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
22594         * config/arc/arc-tables.opt: Regenerate.
22595         * config/arc/arc.c (arc_conditional_register_usage): Handle
22596         reduced register file case.
22597         (arc_file_start): Set must have build attributes.
22598         * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
22599         mrf16 option value.
22600         * config/arc/arc.opt (mrf16): Add new option.
22601         * config/arc/elf.h (ATTRIBUTE_PCS): Define.
22602         * config/arc/genmultilib.awk: Handle new mrf16 option.
22603         * config/arc/linux.h (ATTRIBUTE_PCS): Define.
22604         * config/arc/t-multilib: Regenerate.
22605         * doc/invoke.texi (ARC Options): Document mrf16 option.
22607 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
22609         * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
22610         * config/arc/arc.c (arc_handle_secure_attribute): New function.
22611         (arc_attribute_table): Add 'secure_call' attribute.
22612         (arc_print_operand): Print secure call operand.
22613         (arc_function_ok_for_sibcall): Don't optimize tail calls when
22614         secure.
22615         (arc_is_secure_call_p): New function.  * config/arc/arc.md
22616         (call_i): Add support for sjli instruction.
22617         (call_value_i): Likewise.
22618         * config/arc/constraints.md (Csc): New constraint.
22620 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
22621             John Eric Martin  <John.Martin@emmicro-us.com>
22623         * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
22624         * config/arc/arc.c (_arc_jli_section): New struct.
22625         (arc_jli_section): New type.
22626         (rc_jli_sections): New static variable.
22627         (arc_handle_jli_attribute): New function.
22628         (arc_attribute_table): Add jli_always and jli_fixed attribute.
22629         (arc_file_end): New function.
22630         (TARGET_ASM_FILE_END): Define.
22631         (arc_print_operand): Reuse 'S' letter for JLI output instruction.
22632         (arc_add_jli_section): New function.
22633         (jli_call_scan): Likewise.
22634         (arc_reorg): Call jli_call_scan.
22635         (arc_output_addsi): Remove 'S' from printing asm operand.
22636         (arc_is_jli_call_p): New function.
22637         * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
22638         operand.
22639         (movhi_insn): Likewise.
22640         (movsi_insn): Likewise.
22641         (movsi_set_cc_insn): Likewise.
22642         (loadqi_update): Likewise.
22643         (load_zeroextendqisi_update): Likewise.
22644         (load_signextendqisi_update): Likewise.
22645         (loadhi_update): Likewise.
22646         (load_zeroextendhisi_update): Likewise.
22647         (load_signextendhisi_update): Likewise.
22648         (loadsi_update): Likewise.
22649         (loadsf_update): Likewise.
22650         (movsicc_insn): Likewise.
22651         (bset_insn): Likewise.
22652         (bxor_insn): Likewise.
22653         (bclr_insn): Likewise.
22654         (bmsk_insn): Likewise.
22655         (bicsi3_insn): Likewise.
22656         (cmpsi_cc_c_insn): Likewise.
22657         (movsi_ne): Likewise.
22658         (movsi_cond_exec): Likewise.
22659         (clrsbsi2): Likewise.
22660         (norm_f): Likewise.
22661         (normw): Likewise.
22662         (swap): Likewise.
22663         (divaw): Likewise.
22664         (flag): Likewise.
22665         (sr): Likewise.
22666         (kflag): Likewise.
22667         (ffs): Likewise.
22668         (ffs_f): Likewise.
22669         (fls): Likewise.
22670         (call_i): Remove 'S' asm letter, add jli instruction.
22671         (call_value_i): Likewise.
22672         * config/arc/arc.op (mjli-always): New option.
22673         * config/arc/constraints.md (Cji): New constraint.
22674         * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
22675         operand.
22676         (subsf3_fpx): Likewise.
22677         (mulsf3_fpx): Likewise.
22678         * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
22679         asm operand.
22680         * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
22681         function attrbutes.
22682         * doc/invoke.texi (ARC): Document mjli-always option.
22684 2018-01-26  Sebastian Perta  <sebastian.perta@renesas.com>
22686         * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
22687         avoid addition with 0 and use incw and decw where possible.
22689 2018-01-26  Richard Biener  <rguenther@suse.de>
22691         PR tree-optimization/81082
22692         * fold-const.c (fold_plusminus_mult_expr): Do not perform the
22693         association if it requires casting to unsigned.
22694         * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
22695         from fold_plusminus_mult_expr to catch important cases late when
22696         range info is available.
22698 2018-01-26  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
22700         * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
22701         * configure.ac (hidden_linkonce): New test.
22702         * configure: Regenerate.
22703         * config.in: Regenerate.
22705 2018-01-26  Julia Koval  <julia.koval@intel.com>
22707         * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
22708         _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
22709         _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
22710         _mm_mask_bitshuffle_epi64_mask): Fix type.
22711         * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
22712         USI_FTYPE_V4DI_V4DI_USI): Remove.
22713         * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
22714         __builtin_ia32_vpshufbitqmb256_mask,
22715         __builtin_ia32_vpshufbitqmb128_mask): Fix types.
22716         * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
22717         * config/i386/sse.md (VI1_AVX512VLBW): Change types.
22719 2018-01-26  Alan Modra  <amodra@gmail.com>
22721         PR target/84033
22722         * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
22723         UNSPEC_VBPERMQ.  Sort other unspecs.
22725 2018-01-25  David Edelsohn  <dje.gcc@gmail.com>
22727         * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
22729 2018-01-25  Jan Hubicka  <hubicka@ucw.cz>
22731         PR middle-end/83055
22732         * predict.c (drop_profile): Do not push/pop cfun; update also
22733         node->count.
22734         (handle_missing_profiles): Fix logic looking for zero profiles.
22736 2018-01-25  Jakub Jelinek  <jakub@redhat.com>
22738         PR middle-end/83977
22739         * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
22740         on functions with #pragma omp declare simd or functions with simd
22741         attribute.
22742         * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
22743         * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
22744         Remove trailing \n from warning_at calls.
22746 2018-01-25  Tom de Vries  <tom@codesourcery.com>
22748         PR target/84028
22749         * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
22750         for neutered workers.
22752 2018-01-24  Joseph Myers  <joseph@codesourcery.com>
22754         PR target/68467
22755         * config/m68k/m68k.c (m68k_promote_function_mode): New function.
22756         (TARGET_PROMOTE_FUNCTION_MODE): New macro.
22758 2018-01-24  Jeff Law  <law@redhat.com>
22760         PR target/83994
22761         * i386.c (get_probe_interval): Move to earlier point.
22762         (ix86_compute_frame_layout): If -fstack-clash-protection and
22763         the frame is larger than the probe interval, then use pushes
22764         to save registers rather than reg->mem moves.
22765         (ix86_expand_prologue): Remove conditional for int_registers_saved
22766         assertion.
22768 2018-01-24  Vladimir Makarov  <vmakarov@redhat.com>
22770         PR target/84014
22771         * ira-build.c (setup_min_max_allocno_live_range_point): Set up
22772         min/max for never referenced object.
22774 2018-01-24  Jakub Jelinek  <jakub@redhat.com>
22776         PR middle-end/83977
22777         * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
22778         here.
22779         * omp-low.c (create_omp_child_function): Remove "omp declare simd"
22780         attributes from DECL_ATTRIBUTES (decl) without affecting
22781         DECL_ATTRIBUTES (current_function_decl).
22782         * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
22783         functions with non-NULL DECL_ABSTRACT_ORIGIN.
22785 2018-01-24  Richard Sandiford  <richard.sandiford@linaro.org>
22787         PR tree-optimization/83979
22788         * fold-const.c (fold_comparison): Use constant_boolean_node
22789         instead of boolean_{true,false}_node.
22791 2018-01-24  Jan Hubicka  <hubicka@ucw.cz>
22793         * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
22794         with zero counts.
22796 2018-01-24  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
22798         * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
22799         Simplify the clause that sets the length attribute.
22800         (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
22801         (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
22802         clause that sets the length attribute.
22803         (*sibcall_value_nonlocal_sysv<mode>): Likewise.
22805 2018-01-24  Tom de Vries  <tom@codesourcery.com>
22807         PR target/83589
22808         * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
22809         (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
22810         Add strict parameter.
22811         (prevent_branch_around_nothing): Insert dummy insn between branch to
22812         label and label with no ptx insn inbetween.
22813         * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
22815 2018-01-24  Tom de Vries  <tom@codesourcery.com>
22817         PR target/81352
22818         * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
22819         for neutered threads in warp.
22820         * config/nvptx/nvptx.md (define_insn "exit"): New insn.
22822 2018-01-24  Richard Biener  <rguenther@suse.de>
22824         PR tree-optimization/83176
22825         * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
22826         operands.
22828 2018-01-24  Richard Biener  <rguenther@suse.de>
22830         PR tree-optimization/82819
22831         * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
22832         code generating pluses that are no-ops in the target precision.
22834 2018-01-24  Richard Biener  <rguenther@suse.de>
22836         PR middle-end/84000
22837         * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
22839 2018-01-23  Jan Hubicka  <hubicka@ucw.cz>
22841         * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
22842         to merge probabilities.
22843         * predict.c (probably_never_executed): Also mark as cold functions
22844         with global 0 profile and guessed local profile.
22845         * profile-count.c (profile_probability::combine_with_count): New
22846         member function.
22847         * profile-count.h (profile_probability::operator*,
22848         profile_probability::operator*=, profile_probability::operator/,
22849         profile_probability::operator/=): Reduce precision to adjusted
22850         and set value to guessed on contradictory divisions.
22851         (profile_probability::combine_with_freq): Remove.
22852         (profile_probability::combine_wiht_count): Declare.
22853         (profile_count::force_nonzero):: Set to adjusted.
22854         (profile_count::probability_in):: Set quality to adjusted.
22855         * tree-ssa-tail-merge.c (replace_block_by): Use
22856         combine_with_count.
22858 2018-01-23  Andrew Waterman  <andrew@sifive.com>
22859             Jim Wilson  <jimw@sifive.com>
22861         * config/riscv/riscv.c (riscv_stack_boundary): New.
22862         (riscv_option_override): Set riscv_stack_boundary.  Handle
22863         riscv_preferred_stack_boundary_arg.
22864         * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
22865         (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
22866         (STACK_BOUNDARY): Set to riscv_stack_boundary.
22867         (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
22868         * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
22869         * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
22871 2018-01-23  H.J. Lu  <hongjiu.lu@intel.com>
22873         PR target/83905
22874         * config/i386/i386.c (ix86_expand_prologue): Use cost reference
22875         of struct ix86_frame.
22876         (ix86_expand_epilogue): Likewise.  Add a local variable for
22877         the reg_save_offset field in struct ix86_frame.
22879 2018-01-23  Bin Cheng  <bin.cheng@arm.com>
22881         PR tree-optimization/82604
22882         * tree-loop-distribution.c (enum partition_kind): New enum item
22883         PKIND_PARTIAL_MEMSET.
22884         (partition_builtin_p): Support above new enum item.
22885         (generate_code_for_partition): Ditto.
22886         (compute_access_range): Differentiate cases that equality can be
22887         proven at all loops, the innermost loops or no loops.
22888         (classify_builtin_st, classify_builtin_ldst): Adjust call to above
22889         function.  Set PKIND_PARTIAL_MEMSET for partition appropriately.
22890         (finalize_partitions, distribute_loop): Don't fuse partition of
22891         PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
22892         (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
22893         parloop is enabled.
22895 2018-01-23  Martin Liska  <mliska@suse.cz>
22897         * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
22898         order to ignore the predictor.
22899         (PRED_POLYMORPHIC_CALL): Likewise.
22900         (PRED_RECURSIVE_CALL): Likewise.
22902 2018-01-23  Martin Liska  <mliska@suse.cz>
22904         * tree-profile.c (tree_profiling): Print function header to
22905         aware reader which function we are working on.
22906         * value-prof.c (gimple_find_values_to_profile): Do not print
22907         not interesting value histograms.
22909 2018-01-23  Martin Liska  <mliska@suse.cz>
22911         * profile-count.h (enum profile_quality): Add
22912         profile_uninitialized as the first value. Do not number values
22913         as they are zero based.
22914         (profile_count::verify): Update sanity check.
22915         (profile_probability::verify): Likewise.
22917 2018-01-23  Nathan Sidwell  <nathan@acm.org>
22919         * doc/invoke.texi (ffor-scope): Deprecate.
22921 2018-01-23  David Malcolm  <dmalcolm@redhat.com>
22923         PR tree-optimization/83510
22924         * domwalk.c (set_all_edges_as_executable): New function.
22925         (dom_walker::dom_walker): Convert bool param
22926         "skip_unreachable_blocks" to enum reachability.  Move setup of
22927         edge flags to set_all_edges_as_executable and only do it when
22928         reachability is REACHABLE_BLOCKS.
22929         * domwalk.h (enum dom_walker::reachability): New enum.
22930         (dom_walker::dom_walker): Convert bool param
22931         "skip_unreachable_blocks" to enum reachability.
22932         (set_all_edges_as_executable): New decl.
22933         * graphite-scop-detection.c  (gather_bbs::gather_bbs): Convert
22934         from false for "skip_unreachable_blocks" to ALL_BLOCKS for
22935         "reachability".
22936         * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
22937         but converting true to REACHABLE_BLOCKS.
22938         * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
22939         * tree-vrp.c
22940         (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
22941         Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
22942         (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
22943         REACHABLE_BLOCKS.
22944         (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
22945         if check_all_array_refs will be called.
22947 2018-01-23  David Malcolm  <dmalcolm@redhat.com>
22949         * tree.c (selftest::test_location_wrappers): Add more test
22950         coverage.
22952 2018-01-23  David Malcolm  <dmalcolm@redhat.com>
22954         * sbitmap.c (selftest::test_set_range): Fix memory leaks.
22955         (selftest::test_bit_in_range): Likewise.
22957 2018-01-23  Richard Sandiford  <richard.sandiford@linaro.org>
22959         PR testsuite/83888
22960         * doc/sourcebuild.texi (vect_float): Say that the selector
22961         only describes the situation when -funsafe-math-optimizations is on.
22962         (vect_float_strict): Document.
22964 2018-01-23  Richard Sandiford  <richard.sandiford@linaro.org>
22966         PR tree-optimization/83965
22967         * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
22968         (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
22969         instead of checking only for a reduction.
22970         (vect_recog_widen_sum_pattern): Likewise.
22972 2018-01-23  Jan Hubicka  <hubicka@ucw.cz>
22974         * predict.c (probably_never_executed): Only use precise profile info.
22975         (compute_function_frequency): Skip after inlining hack since we now
22976         have quality checking.
22978 2018-01-23  Jan Hubicka  <hubicka@ucw.cz>
22980         * profile-count.h (profile_probability::very_unlikely,
22981         profile_probability::unlikely, profile_probability::even): Set
22982         precision to guessed.
22984 2018-01-23  Richard Biener  <rguenther@suse.de>
22986         PR tree-optimization/83963
22987         * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
22988         Properly terminate dominator walk when crossing the exit edge not
22989         when visiting its source block.
22991 2018-01-23  Jakub Jelinek  <jakub@redhat.com>
22993         PR c++/83918
22994         * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
22995         VIEW_CONVERT_EXPR to wrap CONST_DECLs.
22997 2018-01-22  Jakub Jelinek  <jakub@redhat.com>
22999         PR tree-optimization/83957
23000         * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs.  Remove
23001         semicolon after for body surrounded by braces.
23003         PR tree-optimization/83081
23004         * profile-count.h (profile_probability::split): New method.
23005         * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
23006         Use profile_probability::split.
23007         (do_compare_rtx_and_jump): Fix adjustment of probabilities
23008         when splitting a single conditional jump into 2.
23010 2018-01-22  David Malcolm  <dmalcolm@redhat.com>
23012         PR tree-optimization/69452
23013         * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
23014         decl.
23016 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
23018         * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
23019         * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
23020         * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
23022 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
23024         * config/rl78/rl78-protos.h (rl78_split_movdi): New function
23025         declaration.
23026         * config/rl78/rl78.md (movdi): New define_expand.
23027         * config/rl78/rl78.c (rl78_split_movdi): New function.
23029 2018-01-22  Michael Meissner  <meissner@linux.vnet.ibm.com>
23031         PR target/83862
23032         * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
23033         no longer used.
23034         * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
23035         * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
23036         128-bit to produce an UNSPEC move to get the double word with the
23037         signbit and then a shift directly to do signbit.
23038         (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
23039         implementation with a new version that just does either a direct
23040         move or a regular move.  Move memory interface to separate insns.
23041         Move insns so they are next to the expander.
23042         (signbit<mode>2_dm_mem_be): New combiner insns to combine load
23043         with signbit move.  Split big and little endian case.
23044         (signbit<mode>2_dm_mem_le): Likewise.
23045         (signbit<mode>2_dm_<su>ext): Delete, no longer used.
23046         (signbit<mode>2_dm2): Likewise.
23048 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
23050         * config/rl78/rl78.md (anddi3): New define_expand.
23052 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
23054         * config/rl78/rl78.md (umindi3): New define_expand.
23056 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
23058         * config/rl78/rl78.md (smindi3): New define_expand.
23060 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
23062         * config/rl78/rl78.md (smaxdi3): New define_expand.
23064 2018-01-22  Carl Love  <cel@us.ibm.com>
23066         * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
23067         LVX_V1TI): Add macro expansion.
23068         * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
23069         definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
23070         VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
23071         * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
23072         Change check to determine if the instruction is a byte reversing
23073         entry.  Fix typo in comment.
23074         * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
23075         for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
23076         Add def_builtin calls for new builtins.
23077         * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
23078         Add define_insn expansion.
23080 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
23082         * config/rl78/rl78.md (umaxdi3): New define_expand.
23084 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
23086         * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
23087         for non-QImode registers.
23089 2018-01-22  Richard Biener  <rguenther@suse.de>
23091         PR tree-optimization/83963
23092         * graphite-scop-detection.c (scop_detection::get_sese): Delay
23093         including the loop exit block.
23094         (scop_detection::merge_sese): Likewise.
23095         (scop_detection::add_scop): Do it here instead.
23097 2018-01-22  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
23099         * doc/sourcebuild.texi (arm_softfloat): Document.
23101 2018-01-21  John David Anglin  <danglin@gcc.gnu.org>
23103         PR gcc/77734
23104         * config/pa/pa.c (pa_function_ok_for_sibcall): Use
23105         targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
23106         Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
23108 2018-01-21  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
23109             David Edelsohn  <dje.gcc@gmail.com>
23111         PR target/83946
23112         * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
23113         Change "crset eq" to "crset 2".
23114         (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
23115         (*call_indirect_aix<mode>_nospec): Likewise.
23116         (*call_value_indirect_aix<mode>_nospec): Likewise.
23117         (*call_indirect_elfv2<mode>_nospec): Likewise.
23118         (*call_value_indirect_elfv2<mode>_nospec): Likewise.
23119         (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
23120         change assembly output from . to $.
23121         (*sibcall_value_nonlocal_sysv<mode>): Likewise.
23122         (indirect_jump<mode>_nospec): Change assembly output from . to $.
23123         (*tablejump<mode>_internal1_nospec): Likewise.
23125 2018-01-21  Oleg Endo  <olegendo@gcc.gnu.org>
23127         PR target/80870
23128         * config/sh/sh_optimize_sett_clrt.cc:
23129         Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
23131 2018-01-20  Richard Sandiford  <richard.sandiford@linaro.org>
23133         PR tree-optimization/83940
23134         * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
23135         offset_dt to vect_constant_def rather than vect_unknown_def_type.
23136         (vect_check_load_store_mask): Add a mask_dt_out parameter and
23137         use it to pass back the definition type.
23138         (vect_check_store_rhs): Likewise rhs_dt_out.
23139         (vect_build_gather_load_calls): Add a mask_dt argument and use
23140         it instead of a call to vect_is_simple_use.
23141         (vectorizable_store): Update calls to vect_check_load_store_mask
23142         and vect_check_store_rhs.  Use the dt returned by the latter instead
23143         of scatter_src_dt.  Use the cached mask_dt and gs_info.offset_dt
23144         instead of calls to vect_is_simple_use.  Pass the scalar rather
23145         than the vector operand to vect_is_simple_use when handling
23146         second and subsequent copies of an rhs value.
23147         (vectorizable_load): Update calls to vect_check_load_store_mask
23148         and vect_build_gather_load_calls.  Use the cached mask_dt and
23149         gs_info.offset_dt instead of calls to vect_is_simple_use.
23151 2018-01-20  Jakub Jelinek  <jakub@redhat.com>
23153         PR middle-end/83945
23154         * tree-emutls.c: Include gimplify.h.
23155         (lower_emutls_2): New function.
23156         (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
23157         with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
23158         it before further processing.
23160         PR target/83930
23161         * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
23162         UINTVAL (trueop1) instead of INTVAL (op1).
23164 2018-01-19  Jakub Jelinek  <jakub@redhat.com>
23166         PR debug/81570
23167         PR debug/83728
23168         * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
23169         INCOMING_FRAME_SP_OFFSET if not defined.
23170         (scan_trace): Add ENTRY argument.  If true and
23171         DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
23172         emit a note to adjust the CFA offset.
23173         (create_cfi_notes): Adjust scan_trace callers.
23174         (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
23175         INCOMING_FRAME_SP_OFFSET in the CIE.
23176         * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
23177         * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
23178         Likewise.
23179         * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
23180         * doc/tm.texi: Regenerated.
23182 2018-01-19  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
23184         PR rtl-optimization/83147
23185         * lra-constraints.c (remove_inheritance_pseudos): Use
23186         lra_substitute_pseudo_within_insn.
23188 2018-01-19  Tom de Vries  <tom@codesourcery.com>
23189             Cesar Philippidis  <cesar@codesourcery.com>
23191         PR target/83920
23192         * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
23194 2018-01-19  Cesar Philippidis  <cesar@codesourcery.com>
23196         PR target/83790
23197         * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
23198         spaces for function labels.
23200 2018-01-19  Martin Liska  <mliska@suse.cz>
23202         * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
23203         (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
23204         (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
23205         (PRED_OPCODE_POSITIVE): Change from 64 to 59.
23206         (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
23207         (PRED_CONST_RETURN): Change from 69 to 65.
23208         (PRED_NULL_RETURN): Change from 91 to 71.
23209         (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
23210         (PRED_LOOP_GUARD): Change from 66 to 73.
23212 2018-01-19  Martin Liska  <mliska@suse.cz>
23214         * predict.c (predict_insn_def): Add new assert.
23215         (struct branch_predictor): Change type to signed integer.
23216         (test_prediction_value_range): Amend test to cover
23217         PROB_UNINITIALIZED.
23218         * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
23219         (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
23220         (PRED_LOOP_ITERATIONS_MAX): Likewise.
23221         (PRED_LOOP_IV_COMPARE): Likewise.
23222         * predict.h (PROB_UNINITIALIZED): Define new constant.
23224 2018-01-19  Martin Liska  <mliska@suse.cz>
23226         * predict.c (dump_prediction): Add new format for
23227         analyze_brprob.py script which is enabled with -details
23228         suboption.
23229         * profile-count.h (precise_p): New function.
23231 2018-01-19  Richard Sandiford  <richard.sandiford@linaro.org>
23233         PR tree-optimization/83922
23234         * tree-vect-loop.c (vect_verify_full_masking): Return false if
23235         there are no statements that need masking.
23236         (vect_active_double_reduction_p): New function.
23237         (vect_analyze_loop_operations): Use it when handling phis that
23238         are not in the loop header.
23240 2018-01-19  Richard Sandiford  <richard.sandiford@linaro.org>
23242         PR tree-optimization/83914
23243         * tree-vect-loop.c (vectorizable_induction): Don't convert
23244         init_expr or apply the peeling adjustment for inductions
23245         that are nested within the vectorized loop.
23247 2018-01-19  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
23249         * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
23250         instead of NEG.
23252 2018-01-18  Jakub Jelinek  <jakub@redhat.com>
23254         PR sanitizer/81715
23255         PR testsuite/83882
23256         * function.h (gimplify_parameters): Add gimple_seq * argument.
23257         * function.c: Include gimple.h and options.h.
23258         (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
23259         for the added local temporaries if needed.
23260         * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
23261         if there are any parameter cleanups, wrap whole body into a
23262         try/finally with the cleanups.
23264 2018-01-18  Wilco Dijkstra  <wdijkstr@arm.com>
23266         PR target/82964
23267         * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
23268         Use GET_MODE_CLASS for scalar floating point.
23270 2018-01-18  Jan Hubicka  <hubicka@ucw.cz>
23272         PR ipa/82256
23273         patch by PaX Team
23274         * cgraphclones.c (cgraph_node::create_version_clone_with_body):
23275         Fix call of call_cgraph_insertion_hooks.
23277 2018-01-18  Martin Sebor  <msebor@redhat.com>
23279         * doc/invoke.texi (-Wclass-memaccess): Tweak text.
23281 2018-01-18  Jan Hubicka  <hubicka@ucw.cz>
23283         PR ipa/83619
23284         * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
23285         frequencies.
23287 2018-01-18  Boris Kolpackov  <boris@codesynthesis.com>
23289         PR other/70268
23290         * common.opt: (-ffile-prefix-map): New option.
23291         * opts.c (common_handle_option): Defer it.
23292         * opts-global.c (handle_common_deferred_options): Handle it.
23293         * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
23294         * file-prefix-map.h: New file.
23295         (remap_debug_filename, add_debug_prefix_map): ...here.
23296         (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
23297         * final.c (debug_prefix_map, add_debug_prefix_map
23298         remap_debug_filename): Move to...
23299         * file-prefix-map.c: New file.
23300         (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
23301         generalize, get rid of alloca(), use strrchr() instead of strchr().
23302         (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
23303         Implement in terms of add_prefix_map().
23304         (remap_macro_filename, remap_debug_filename): Implement in term of
23305         remap_filename().
23306         * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
23307         * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
23308         * dbxout.c: Include file-prefix-map.h.
23309         * varasm.c: Likewise.
23310         * vmsdbgout.c: Likewise.
23311         * xcoffout.c: Likewise.
23312         * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
23313         * doc/cppopts.texi (-fmacro-prefix-map): Document.
23314         * doc/invoke.texi (-ffile-prefix-map): Document.
23315         (-fdebug-prefix-map): Update description.
23317 2018-01-18  Martin Liska  <mliska@suse.cz>
23319         * config/i386/i386.c (indirect_thunk_name): Document that also
23320         lfence is emitted.
23321         (output_indirect_thunk): Document why both instructions
23322         (pause and lfence) are generated.
23324 2018-01-18  Richard Biener  <rguenther@suse.de>
23326         PR tree-optimization/83887
23327         * graphite-scop-detection.c
23328         (scop_detection::get_nearest_dom_with_single_entry): Remove.
23329         (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
23330         (scop_detection::merge_sese): Re-implement with a flood-fill
23331         algorithm that properly finds a SESE region if it exists.
23333 2018-01-18  Jakub Jelinek  <jakub@redhat.com>
23335         PR c/61240
23336         * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
23337         pointer_diff optimizations use view_convert instead of convert.
23339 2018-01-17  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
23341         * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
23342         Generate different code for -mno-speculate-indirect-jumps.
23343         (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
23344         (*call_indirect_aix<mode>): Disable for
23345         -mno-speculate-indirect-jumps.
23346         (*call_indirect_aix<mode>_nospec): New define_insn.
23347         (*call_value_indirect_aix<mode>): Disable for
23348         -mno-speculate-indirect-jumps.
23349         (*call_value_indirect_aix<mode>_nospec): New define_insn.
23350         (*sibcall_nonlocal_sysv<mode>): Generate different code for
23351         -mno-speculate-indirect-jumps.
23352         (*sibcall_value_nonlocal_sysv<mode>): Likewise.
23354 2018-01-17  Michael Meissner  <meissner@linux.vnet.ibm.com>
23356         * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
23357         long double type, set the flags for noting the default long double
23358         type, even if we don't pass or return a long double type.
23360 2018-01-17  Jan Hubicka  <hubicka@ucw.cz>
23362         PR ipa/83051
23363         * ipa-inline.c (flatten_function): Do not overwrite final inlining
23364         failure.
23366 2018-01-17  Will Schmidt  <will_schmidt@vnet.ibm.com>
23368         * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
23369         support for merge[hl].
23370         (fold_mergehl_helper): New helper function.
23371         (tree-vector-builder.h): New #include for tree_vector_builder usage.
23372         * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
23373         (altivec_vmrglw_direct): Add xxmrglw insn.
23375 2018-01-17  Andrew Waterman  <andrew@sifive.com>
23377         * config/riscv/riscv.c (riscv_conditional_register_usage): If
23378         UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
23380 2018-01-17  David Malcolm  <dmalcolm@redhat.com>
23382         PR lto/83121
23383         * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
23384         call the lto_location_cache before reading the
23385         DECL_SOURCE_LOCATION of the types.
23387 2018-01-17  Wilco Dijkstra  <wdijkstr@arm.com>
23388             Richard Sandiford  <richard.sandiford@linaro.org>
23390         * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
23391         * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
23392         (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
23393         SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
23394         * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
23395         Add declaration.
23396         * config/aarch64/constraints.md (aarch64_movti_operand):
23397         Limit immediates.
23398         * config/aarch64/predicates.md (Uti): Add new constraint.
23400 2018-01-17  Carl Love  <cel@us.ibm.com>
23402         * config/rs6000/vsx.md (define_expand xl_len_r,
23403         define_expand stxvl, define_expand *stxvl): Add match_dup argument.
23404         (define_insn): Add, match_dup 1 argument to define_insn stxvll and
23405         lxvll.
23406         (define_expand, define_insn): Move the shift left from  the
23407         define_insn to the define_expand for lxvl and stxvl instructions.
23408         * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
23409         and XL_LEN_R definitions to PURE.
23411 2018-01-17  Uros Bizjak  <ubizjak@gmail.com>
23413         * config/i386/i386.c (indirect_thunk_name): Declare regno
23414         as unsigned int.  Compare regno with INVALID_REGNUM.
23415         (output_indirect_thunk): Ditto.
23416         (output_indirect_thunk_function): Ditto.
23417         (ix86_code_end): Declare regno as unsigned int.  Use INVALID_REGNUM
23418         in the call to output_indirect_thunk_function.
23420 2018-01-17  Richard Sandiford  <richard.sandiford@linaro.org>
23422         PR middle-end/83884
23423         * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
23424         rather than the size of inner_type to determine the stack slot size
23425         when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
23427 2018-01-16  Sebastian Peryt  <sebastian.peryt@intel.com>
23429         PR target/83546
23430         * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
23431         to PTA_SILVERMONT.
23433 2018-01-16  Michael Meissner  <meissner@linux.vnet.ibm.com>
23435         * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
23436         endian Linux systems to optionally enable multilibs for selecting
23437         the long double type if the user configured an explicit type.
23438         * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
23439         have no long double multilibs if not defined.
23440         * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
23441         warn if the user used -mabi={ieee,ibm}longdouble and we built
23442         multilibs for long double.
23443         * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
23444         appropriate multilib option.
23445         (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
23446         multilib options.
23447         * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
23448         for building long double multilibs.
23449         * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
23451 2018-01-16  John David Anglin  <danglin@gcc.gnu.org>
23453         * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
23454         copies.
23456         * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
23457         64 bits.
23458         * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
23459         128 bits.
23461         * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
23462         variables.
23464         * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
23465         return value.
23467 2018-01-16  Eric Botcazou  <ebotcazou@adacore.com>
23469         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
23470         ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
23472 2018-01-16  Kelvin Nilsen  <kelvin@gcc.gnu.org>
23474         * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
23475         different rtl trees depending on TARGET_64BIT.
23476         (rs6000_gen_lvx): Likewise.
23478 2018-01-16  Eric Botcazou  <ebotcazou@adacore.com>
23480         * config/visium/visium.md (nop): Tweak comment.
23481         (hazard_nop): Likewise.
23483 2018-01-16  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
23485         * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
23486         -mspeculate-indirect-jumps.
23487         * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
23488         for -mno-speculate-indirect-jumps.
23489         (*call_indirect_elfv2<mode>_nospec): New define_insn.
23490         (*call_value_indirect_elfv2<mode>): Disable for
23491         -mno-speculate-indirect-jumps.
23492         (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
23493         (indirect_jump): Emit different RTL for
23494         -mno-speculate-indirect-jumps.
23495         (*indirect_jump<mode>): Disable for
23496         -mno-speculate-indirect-jumps.
23497         (*indirect_jump<mode>_nospec): New define_insn.
23498         (tablejump): Emit different RTL for
23499         -mno-speculate-indirect-jumps.
23500         (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
23501         (tablejumpsi_nospec): New define_expand.
23502         (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
23503         (tablejumpdi_nospec): New define_expand.
23504         (*tablejump<mode>_internal1): Disable for
23505         -mno-speculate-indirect-jumps.
23506         (*tablejump<mode>_internal1_nospec): New define_insn.
23507         * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
23508         option.
23510 2018-01-16  Artyom Skrobov tyomitch@gmail.com
23512         * caller-save.c (insert_save): Drop unnecessary parameter.  All
23513         callers updated.
23515 2018-01-16  Jakub Jelinek  <jakub@redhat.com>
23516             Richard Biener  <rguenth@suse.de>
23518         PR libgomp/83590
23519         * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
23520         return early, inline manually is_gimple_sizepos.  Make sure if we
23521         call gimplify_expr we don't end up with a gimple constant.
23522         * tree.c (variably_modified_type_p): Don't return true for
23523         is_gimple_constant (_t).  Inline manually is_gimple_sizepos.
23524         * gimplify.h (is_gimple_sizepos): Remove.
23526 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
23528         PR tree-optimization/83857
23529         * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
23530         vectorizable_live_operation for pure SLP statements.
23531         (vectorizable_live_operation): Handle PHIs.
23533 2018-01-16  Richard Biener  <rguenther@suse.de>
23535         PR tree-optimization/83867
23536         * tree-vect-stmts.c (vect_transform_stmt): Precompute
23537         nested_in_vect_loop_p since the scalar stmt may get invalidated.
23539 2018-01-16  Jakub Jelinek  <jakub@redhat.com>
23541         PR c/83844
23542         * stor-layout.c (handle_warn_if_not_align): Use byte_position and
23543         multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
23544         If off is not INTEGER_CST, issue a may not be aligned warning
23545         rather than isn't aligned.  Use isn%'t rather than isn't.
23546         * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
23547         into MULT_EXPR.
23548         <case MULT_EXPR>: Improve the case when bottom and one of the
23549         MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
23550         operand, in that case check if the other operand is multiple of
23551         bottom divided by the INTEGER_CST operand.
23553 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
23555         PR target/83858
23556         * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
23557         * config/pa/pa-protos.h (pa_function_arg_size): Declare.
23558         * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
23559         pa_function_arg_size instead of FUNCTION_ARG_SIZE.
23560         * config/pa/pa.c (pa_function_arg_advance): Likewise.
23561         (pa_function_arg, pa_arg_partial_bytes): Likewise.
23562         (pa_function_arg_size): New function.
23564 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
23566         * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
23567         in a separate statement.
23569 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
23571         PR tree-optimization/83847
23572         * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
23573         group gathers and scatters.
23575 2018-01-16  Jakub Jelinek  <jakub@redhat.com>
23577         PR rtl-optimization/86620
23578         * params.def (max-sched-ready-insns): Bump minimum value to 1.
23580         PR rtl-optimization/83213
23581         * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
23582         to last if both are JUMP_INSNs.
23584         PR tree-optimization/83843
23585         * gimple-ssa-store-merging.c
23586         (imm_store_chain_info::output_merged_store): Handle bit_not_p on
23587         store_immediate_info for bswap/nop orig_stores.
23589 2018-01-15  Andrew Waterman  <andrew@sifive.com>
23591         * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
23592         !TARGET_MUL.
23593         <UDIV>: Increase cost if !TARGET_DIV.
23595 2018-01-15  Segher Boessenkool  <segher@kernel.crashing.org>
23597         * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
23598         (define_attr "cr_logical_3op"): New.
23599         (cceq_ior_compare): Adjust.
23600         (cceq_ior_compare_complement): Adjust.
23601         (*cceq_rev_compare): Adjust.
23602         * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
23603         (is_cracked_insn): Adjust.
23604         (insn_must_be_first_in_group): Adjust.
23605         * config/rs6000/40x.md: Adjust.
23606         * config/rs6000/440.md: Adjust.
23607         * config/rs6000/476.md: Adjust.
23608         * config/rs6000/601.md: Adjust.
23609         * config/rs6000/603.md: Adjust.
23610         * config/rs6000/6xx.md: Adjust.
23611         * config/rs6000/7450.md: Adjust.
23612         * config/rs6000/7xx.md: Adjust.
23613         * config/rs6000/8540.md: Adjust.
23614         * config/rs6000/cell.md: Adjust.
23615         * config/rs6000/e300c2c3.md: Adjust.
23616         * config/rs6000/e500mc.md: Adjust.
23617         * config/rs6000/e500mc64.md: Adjust.
23618         * config/rs6000/e5500.md: Adjust.
23619         * config/rs6000/e6500.md: Adjust.
23620         * config/rs6000/mpc.md: Adjust.
23621         * config/rs6000/power4.md: Adjust.
23622         * config/rs6000/power5.md: Adjust.
23623         * config/rs6000/power6.md: Adjust.
23624         * config/rs6000/power7.md: Adjust.
23625         * config/rs6000/power8.md: Adjust.
23626         * config/rs6000/power9.md: Adjust.
23627         * config/rs6000/rs64.md: Adjust.
23628         * config/rs6000/titan.md: Adjust.
23630 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
23632         * config/i386/predicates.md (indirect_branch_operand): Rewrite
23633         ix86_indirect_branch_register logic.
23635 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
23637         * config/i386/constraints.md (Bs): Update
23638         ix86_indirect_branch_register check.  Don't check
23639         ix86_indirect_branch_register with GOT_memory_operand.
23640         (Bw): Likewise.
23641         * config/i386/predicates.md (GOT_memory_operand): Don't check
23642         ix86_indirect_branch_register here.
23643         (GOT32_symbol_operand): Likewise.
23645 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
23647         * config/i386/predicates.md (constant_call_address_operand):
23648         Rewrite ix86_indirect_branch_register logic.
23649         (sibcall_insn_operand): Likewise.
23651 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
23653         * config/i386/constraints.md (Bs): Replace
23654         ix86_indirect_branch_thunk_register with
23655         ix86_indirect_branch_register.
23656         (Bw): Likewise.
23657         * config/i386/i386.md (indirect_jump): Likewise.
23658         (tablejump): Likewise.
23659         (*sibcall_memory): Likewise.
23660         (*sibcall_value_memory): Likewise.
23661         Peepholes of indirect call and jump via memory: Likewise.
23662         * config/i386/i386.opt: Likewise.
23663         * config/i386/predicates.md (indirect_branch_operand): Likewise.
23664         (GOT_memory_operand): Likewise.
23665         (call_insn_operand): Likewise.
23666         (sibcall_insn_operand): Likewise.
23667         (GOT32_symbol_operand): Likewise.
23669 2018-01-15  Jakub Jelinek  <jakub@redhat.com>
23671         PR middle-end/83837
23672         * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
23673         type rather than type addr's type points to.
23674         (expand_omp_atomic_mutex): Likewise.
23675         (expand_omp_atomic): Likewise.
23677 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
23679         PR target/83839
23680         * config/i386/i386.c (output_indirect_thunk_function): Use
23681         ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
23682         for  __x86_return_thunk.
23684 2018-01-15  Richard Biener  <rguenther@suse.de>
23686         PR middle-end/83850
23687         * expmed.c (extract_bit_field_1): Fix typo.
23689 2018-01-15  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
23691         PR target/83687
23692         * config/arm/iterators.md (VF): New mode iterator.
23693         * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
23694         Remove integer-related logic from pattern.
23695         (neon_vabd<mode>_3): Likewise.
23697 2018-01-15  Jakub Jelinek  <jakub@redhat.com>
23699         PR middle-end/82694
23700         * common.opt (fstrict-overflow): No longer an alias.
23701         (fwrapv-pointer): New option.
23702         * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
23703         also for pointer types based on flag_wrapv_pointer.
23704         * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
23705         opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
23706         opts->x_flag_wrapv got set.
23707         * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
23708         changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
23709         POINTER_TYPE_OVERFLOW_UNDEFINED.
23710         * match.pd: Likewise in address comparison pattern.
23711         * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
23713 2018-01-15  Richard Biener  <rguenther@suse.de>
23715         PR lto/83804
23716         * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
23717         from TYPE_FIELDS.  Free TYPE_BINFO if not used by devirtualization.
23718         Reset type names to their identifier if their TYPE_DECL doesn't
23719         have linkage (and thus is used for ODR and devirt).
23720         (save_debug_info_for_decl): Remove.
23721         (save_debug_info_for_type): Likewise.
23722         (add_tree_to_fld_list): Adjust.
23723         * tree-pretty-print.c (dump_generic_node): Make dumping of
23724         type names more robust.
23726 2018-01-15  Richard Biener  <rguenther@suse.de>
23728         * BASE-VER: Bump to 8.0.1.
23730 2018-01-14  Martin Sebor  <msebor@redhat.com>
23732         PR other/83508
23733         * builtins.c (check_access): Avoid warning when the no-warning bit
23734         is set.
23736 2018-01-14  Cory Fields  <cory-nospam-@coryfields.com>
23738         * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
23739         * ira-color (allocno_hard_regs_compare): Likewise.
23741 2018-01-14  Nathan Rossi  <nathan@nathanrossi.com>
23743         PR target/83013
23744         * config/microblaze/microblaze.c (microblaze_asm_output_ident):
23745         Use .pushsection/.popsection.
23747 2018-01-14  Martin Sebor  <msebor@redhat.com>
23749         PR c++/81327
23750         * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
23752 2018-01-14  Jakub Jelinek  <jakub@redhat.com>
23754         * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
23755         entry from extra_headers.
23756         (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
23757         extra_headers, make the list bitwise identical to the i?86-*-* one.
23759 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
23761         * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
23762         -mcmodel=large with -mindirect-branch=thunk,
23763         -mindirect-branch=thunk-extern, -mfunction-return=thunk and
23764         -mfunction-return=thunk-extern.
23765         * doc/invoke.texi: Document -mcmodel=large is incompatible with
23766         -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
23767         -mfunction-return=thunk and -mfunction-return=thunk-extern.
23769 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
23771         * config/i386/i386.c (print_reg): Print the name of the full
23772         integer register without '%'.
23773         (ix86_print_operand): Handle 'V'.
23774         * doc/extend.texi: Document 'V' modifier.
23776 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
23778         * config/i386/constraints.md (Bs): Disallow memory operand for
23779         -mindirect-branch-register.
23780         (Bw): Likewise.
23781         * config/i386/predicates.md (indirect_branch_operand): Likewise.
23782         (GOT_memory_operand): Likewise.
23783         (call_insn_operand): Likewise.
23784         (sibcall_insn_operand): Likewise.
23785         (GOT32_symbol_operand): Likewise.
23786         * config/i386/i386.md (indirect_jump): Call convert_memory_address
23787         for -mindirect-branch-register.
23788         (tablejump): Likewise.
23789         (*sibcall_memory): Likewise.
23790         (*sibcall_value_memory): Likewise.
23791         Disallow peepholes of indirect call and jump via memory for
23792         -mindirect-branch-register.
23793         (*call_pop): Replace m with Bw.
23794         (*call_value_pop): Likewise.
23795         (*sibcall_pop_memory): Replace m with Bs.
23796         * config/i386/i386.opt (mindirect-branch-register): New option.
23797         * doc/invoke.texi: Document -mindirect-branch-register option.
23799 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
23801         * config/i386/i386-protos.h (ix86_output_function_return): New.
23802         * config/i386/i386.c (ix86_set_indirect_branch_type): Also
23803         set function_return_type.
23804         (indirect_thunk_name): Add ret_p to indicate thunk for function
23805         return.
23806         (output_indirect_thunk_function): Pass false to
23807         indirect_thunk_name.
23808         (ix86_output_indirect_branch_via_reg): Likewise.
23809         (ix86_output_indirect_branch_via_push): Likewise.
23810         (output_indirect_thunk_function): Create alias for function
23811         return thunk if regno < 0.
23812         (ix86_output_function_return): New function.
23813         (ix86_handle_fndecl_attribute): Handle function_return.
23814         (ix86_attribute_table): Add function_return.
23815         * config/i386/i386.h (machine_function): Add
23816         function_return_type.
23817         * config/i386/i386.md (simple_return_internal): Use
23818         ix86_output_function_return.
23819         (simple_return_internal_long): Likewise.
23820         * config/i386/i386.opt (mfunction-return=): New option.
23821         (indirect_branch): Mention -mfunction-return=.
23822         * doc/extend.texi: Document function_return function attribute.
23823         * doc/invoke.texi: Document -mfunction-return= option.
23825 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
23827         * config/i386/i386-opts.h (indirect_branch): New.
23828         * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
23829         * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
23830         with local indirect jump when converting indirect call and jump.
23831         (ix86_set_indirect_branch_type): New.
23832         (ix86_set_current_function): Call ix86_set_indirect_branch_type.
23833         (indirectlabelno): New.
23834         (indirect_thunk_needed): Likewise.
23835         (indirect_thunk_bnd_needed): Likewise.
23836         (indirect_thunks_used): Likewise.
23837         (indirect_thunks_bnd_used): Likewise.
23838         (INDIRECT_LABEL): Likewise.
23839         (indirect_thunk_name): Likewise.
23840         (output_indirect_thunk): Likewise.
23841         (output_indirect_thunk_function): Likewise.
23842         (ix86_output_indirect_branch_via_reg): Likewise.
23843         (ix86_output_indirect_branch_via_push): Likewise.
23844         (ix86_output_indirect_branch): Likewise.
23845         (ix86_output_indirect_jmp): Likewise.
23846         (ix86_code_end): Call output_indirect_thunk_function if needed.
23847         (ix86_output_call_insn): Call ix86_output_indirect_branch if
23848         needed.
23849         (ix86_handle_fndecl_attribute): Handle indirect_branch.
23850         (ix86_attribute_table): Add indirect_branch.
23851         * config/i386/i386.h (machine_function): Add indirect_branch_type
23852         and has_local_indirect_jump.
23853         * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
23854         to true.
23855         (tablejump): Likewise.
23856         (*indirect_jump): Use ix86_output_indirect_jmp.
23857         (*tablejump_1): Likewise.
23858         (simple_return_indirect_internal): Likewise.
23859         * config/i386/i386.opt (mindirect-branch=): New option.
23860         (indirect_branch): New.
23861         (keep): Likewise.
23862         (thunk): Likewise.
23863         (thunk-inline): Likewise.
23864         (thunk-extern): Likewise.
23865         * doc/extend.texi: Document indirect_branch function attribute.
23866         * doc/invoke.texi: Document -mindirect-branch= option.
23868 2018-01-14  Jan Hubicka  <hubicka@ucw.cz>
23870         PR ipa/83051
23871         * ipa-inline.c (edge_badness): Tolerate roundoff errors.
23873 2018-01-14  Richard Sandiford  <richard.sandiford@linaro.org>
23875         * ipa-inline.c (want_inline_small_function_p): Return false if
23876         inlining has already failed with CIF_FINAL_ERROR.
23877         (update_caller_keys): Call want_inline_small_function_p before
23878         can_inline_edge_p.
23879         (update_callee_keys): Likewise.
23881 2018-01-10  Kelvin Nilsen  <kelvin@gcc.gnu.org>
23883         * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
23884         New function.
23885         (rs6000_quadword_masked_address_p): Likewise.
23886         (quad_aligned_load_p): Likewise.
23887         (quad_aligned_store_p): Likewise.
23888         (const_load_sequence_p): Add comment to describe the outer-most loop.
23889         (mimic_memory_attributes_and_flags): New function.
23890         (rs6000_gen_stvx): Likewise.
23891         (replace_swapped_aligned_store): Likewise.
23892         (rs6000_gen_lvx): Likewise.
23893         (replace_swapped_aligned_load): Likewise.
23894         (replace_swapped_load_constant): Capitalize argument name in
23895         comment describing this function.
23896         (rs6000_analyze_swaps): Add a third pass to search for vector loads
23897         and stores that access quad-word aligned addresses and replace
23898         with stvx or lvx instructions when appropriate.
23899         * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
23900         New function prototype.
23901         (rs6000_quadword_masked_address_p): Likewise.
23902         (rs6000_gen_lvx): Likewise.
23903         (rs6000_gen_stvx): Likewise.
23904         * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
23905         VSX_D (V2DF, V2DI), modify this split to select lvx instruction
23906         when memory address is aligned.
23907         (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
23908         this split to select lvx instruction when memory address is aligned.
23909         (*vsx_le_perm_load_v8hi): Modify this split to select lvx
23910         instruction when memory address is aligned.
23911         (*vsx_le_perm_load_v16qi): Likewise.
23912         (four unnamed splitters): Modify to select the stvx instruction
23913         when memory is aligned.
23915 2018-01-13  Jan Hubicka  <hubicka@ucw.cz>
23917         * predict.c (determine_unlikely_bbs): Handle correctly BBs
23918         which appears in the queue multiple times.
23920 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
23921             Alan Hayward  <alan.hayward@arm.com>
23922             David Sherwood  <david.sherwood@arm.com>
23924         * tree-vectorizer.h (vec_lower_bound): New structure.
23925         (_loop_vec_info): Add check_nonzero and lower_bounds.
23926         (LOOP_VINFO_CHECK_NONZERO): New macro.
23927         (LOOP_VINFO_LOWER_BOUNDS): Likewise.
23928         (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
23929         * tree-data-ref.h (dr_with_seg_len): Add access_size and align
23930         fields.  Make seg_len the distance travelled, not including the
23931         access size.
23932         (dr_direction_indicator): Declare.
23933         (dr_zero_step_indicator): Likewise.
23934         (dr_known_forward_stride_p): Likewise.
23935         * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
23936         tree-ssanames.h.
23937         (runtime_alias_check_p): Allow runtime alias checks with
23938         variable strides.
23939         (operator ==): Compare access_size and align.
23940         (prune_runtime_alias_test_list): Rework for new distinction between
23941         the access_size and seg_len.
23942         (create_intersect_range_checks_index): Likewise.  Cope with polynomial
23943         segment lengths.
23944         (get_segment_min_max): New function.
23945         (create_intersect_range_checks): Use it.
23946         (dr_step_indicator): New function.
23947         (dr_direction_indicator): Likewise.
23948         (dr_zero_step_indicator): Likewise.
23949         (dr_known_forward_stride_p): Likewise.
23950         * tree-loop-distribution.c (data_ref_segment_size): Return
23951         DR_STEP * (niters - 1).
23952         (compute_alias_check_pairs): Update call to the dr_with_seg_len
23953         constructor.
23954         * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
23955         (vect_preserves_scalar_order_p): New function, split out from...
23956         (vect_analyze_data_ref_dependence): ...here.  Check for zero steps.
23957         (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
23958         (vect_vfa_access_size): New function.
23959         (vect_vfa_align): Likewise.
23960         (vect_compile_time_alias): Take access_size_a and access_b arguments.
23961         (dump_lower_bound): New function.
23962         (vect_check_lower_bound): Likewise.
23963         (vect_small_gap_p): Likewise.
23964         (vectorizable_with_step_bound_p): Likewise.
23965         (vect_prune_runtime_alias_test_list): Ignore cross-iteration
23966         depencies if the vectorization factor is 1.  Convert the checks
23967         for nonzero steps into checks on the bounds of DR_STEP.  Try using
23968         a bunds check for variable steps if the minimum required step is
23969         relatively small. Update calls to the dr_with_seg_len
23970         constructor and to vect_compile_time_alias.
23971         * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
23972         function.
23973         (vect_loop_versioning): Call it.
23974         * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
23975         when retrying.
23976         (vect_estimate_min_profitable_iters): Account for any bounds checks.
23978 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
23979             Alan Hayward  <alan.hayward@arm.com>
23980             David Sherwood  <david.sherwood@arm.com>
23982         * doc/sourcebuild.texi (vect_scatter_store): Document.
23983         * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
23984         optabs.
23985         * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
23986         Document.
23987         * genopinit.c (main): Add supports_vec_scatter_store and
23988         supports_vec_scatter_store_cached to target_optabs.
23989         * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
23990         IFN_MASK_SCATTER_STORE.
23991         * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
23992         functions.
23993         * internal-fn.h (internal_store_fn_p): Declare.
23994         (internal_fn_stored_value_index): Likewise.
23995         * internal-fn.c (scatter_store_direct): New macro.
23996         (expand_scatter_store_optab_fn): New function.
23997         (direct_scatter_store_optab_supported_p): New macro.
23998         (internal_store_fn_p): New function.
23999         (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
24000         IFN_MASK_SCATTER_STORE.
24001         (internal_fn_mask_index): Likewise.
24002         (internal_fn_stored_value_index): New function.
24003         (internal_gather_scatter_fn_supported_p): Adjust operand numbers
24004         for scatter stores.
24005         * optabs-query.h (supports_vec_scatter_store_p): Declare.
24006         * optabs-query.c (supports_vec_scatter_store_p): New function.
24007         * tree-vectorizer.h (vect_get_store_rhs): Declare.
24008         * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
24009         true for scatter stores.
24010         (vect_gather_scatter_fn_p): Handle scatter stores too.
24011         (vect_check_gather_scatter): Consider using scatter stores if
24012         supports_vec_scatter_store_p.
24013         * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
24014         scatter stores too.
24015         * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
24016         internal_fn_stored_value_index.
24017         (check_load_store_masking): Handle scatter stores too.
24018         (vect_get_store_rhs): Make public.
24019         (vectorizable_call): Use internal_store_fn_p.
24020         (vectorizable_store): Handle scatter store internal functions.
24021         (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
24022         when deciding whether the end of the group has been reached.
24023         * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
24024         * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
24025         (mask_scatter_store<mode>): New insns.
24027 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
24028             Alan Hayward  <alan.hayward@arm.com>
24029             David Sherwood  <david.sherwood@arm.com>
24031         * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
24032         * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
24033         * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
24034         function.
24035         (vect_use_strided_gather_scatters_p): Take a masked_p argument.
24036         Use vect_truncate_gather_scatter_offset if we can't treat the
24037         operation as a normal gather load or scatter store.
24038         (get_group_load_store_type): Take the gather_scatter_info
24039         as argument.  Try using a gather load or scatter store for
24040         single-element groups.
24041         (get_load_store_type): Update calls to get_group_load_store_type
24042         and vect_use_strided_gather_scatters_p.
24044 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
24045             Alan Hayward  <alan.hayward@arm.com>
24046             David Sherwood  <david.sherwood@arm.com>
24048         * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
24049         optional tree argument.
24050         * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
24051         null target hooks.
24052         (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
24053         but continue to use the current value as a fallback.
24054         (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
24055         to compare the updates.
24056         * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
24057         (get_load_store_type): Use it when handling a strided access.
24058         (vect_get_strided_load_store_ops): New function.
24059         (vect_get_data_ptr_increment): Likewise.
24060         (vectorizable_load): Handle strided gather loads.  Always pass
24061         a step to vect_create_data_ref_ptr and bump_vector_ptr.
24063 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
24064             Alan Hayward  <alan.hayward@arm.com>
24065             David Sherwood  <david.sherwood@arm.com>
24067         * doc/md.texi (gather_load@var{m}): Document.
24068         (mask_gather_load@var{m}): Likewise.
24069         * genopinit.c (main): Add supports_vec_gather_load and
24070         supports_vec_gather_load_cached to target_optabs.
24071         * optabs-tree.c (init_tree_optimization_optabs): Use
24072         ggc_cleared_alloc to allocate target_optabs.
24073         * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
24074         * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
24075         functions.
24076         * internal-fn.h (internal_load_fn_p): Declare.
24077         (internal_gather_scatter_fn_p): Likewise.
24078         (internal_fn_mask_index): Likewise.
24079         (internal_gather_scatter_fn_supported_p): Likewise.
24080         * internal-fn.c (gather_load_direct): New macro.
24081         (expand_gather_load_optab_fn): New function.
24082         (direct_gather_load_optab_supported_p): New macro.
24083         (direct_internal_fn_optab): New function.
24084         (internal_load_fn_p): Likewise.
24085         (internal_gather_scatter_fn_p): Likewise.
24086         (internal_fn_mask_index): Likewise.
24087         (internal_gather_scatter_fn_supported_p): Likewise.
24088         * optabs-query.c (supports_at_least_one_mode_p): New function.
24089         (supports_vec_gather_load_p): Likewise.
24090         * optabs-query.h (supports_vec_gather_load_p): Declare.
24091         * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
24092         and memory_type field.
24093         (NUM_PATTERNS): Bump to 15.
24094         * tree-vect-data-refs.c: Include internal-fn.h.
24095         (vect_gather_scatter_fn_p): New function.
24096         (vect_describe_gather_scatter_call): Likewise.
24097         (vect_check_gather_scatter): Try using internal functions for
24098         gather loads.  Recognize existing calls to a gather load function.
24099         (vect_analyze_data_refs): Consider using gather loads if
24100         supports_vec_gather_load_p.
24101         * tree-vect-patterns.c (vect_get_load_store_mask): New function.
24102         (vect_get_gather_scatter_offset_type): Likewise.
24103         (vect_convert_mask_for_vectype): Likewise.
24104         (vect_add_conversion_to_patterm): Likewise.
24105         (vect_try_gather_scatter_pattern): Likewise.
24106         (vect_recog_gather_scatter_pattern): New pattern recognizer.
24107         (vect_vect_recog_func_ptrs): Add it.
24108         * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
24109         internal_fn_mask_index and internal_gather_scatter_fn_p.
24110         (check_load_store_masking): Take the gather_scatter_info as an
24111         argument and handle gather loads.
24112         (vect_get_gather_scatter_ops): New function.
24113         (vectorizable_call): Check internal_load_fn_p.
24114         (vectorizable_load): Likewise.  Handle gather load internal
24115         functions.
24116         (vectorizable_store): Update call to check_load_store_masking.
24117         * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
24118         * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
24119         * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
24120         (aarch64_gather_scale_operand_d): New predicates.
24121         * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
24122         (mask_gather_load<mode>): New insns.
24124 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
24125             Alan Hayward  <alan.hayward@arm.com>
24126             David Sherwood  <david.sherwood@arm.com>
24128         * optabs.def (fold_left_plus_optab): New optab.
24129         * doc/md.texi (fold_left_plus_@var{m}): Document.
24130         * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
24131         * internal-fn.c (fold_left_direct): Define.
24132         (expand_fold_left_optab_fn): Likewise.
24133         (direct_fold_left_optab_supported_p): Likewise.
24134         * fold-const-call.c (fold_const_fold_left): New function.
24135         (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
24136         * tree-parloops.c (valid_reduction_p): New function.
24137         (gather_scalar_reductions): Use it.
24138         * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
24139         (vect_finish_replace_stmt): Declare.
24140         * tree-vect-loop.c (fold_left_reduction_fn): New function.
24141         (needs_fold_left_reduction_p): New function, split out from...
24142         (vect_is_simple_reduction): ...here.  Accept reductions that
24143         forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
24144         (vect_force_simple_reduction): Also store the reduction type in
24145         the assignment's STMT_VINFO_REDUC_TYPE.
24146         (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
24147         (merge_with_identity): New function.
24148         (vect_expand_fold_left): Likewise.
24149         (vectorize_fold_left_reduction): Likewise.
24150         (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION.  Leave the
24151         scalar phi in place for it.  Check for target support and reject
24152         cases that would reassociate the operation.  Defer the transform
24153         phase to vectorize_fold_left_reduction.
24154         * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
24155         * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
24156         (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
24158 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
24160         * tree-if-conv.c (predicate_mem_writes): Remove redundant
24161         call to ifc_temp_var.
24163 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
24164             Alan Hayward  <alan.hayward@arm.com>
24165             David Sherwood  <david.sherwood@arm.com>
24167         * target.def (legitimize_address_displacement): Take the original
24168         offset as a poly_int.
24169         * targhooks.h (default_legitimize_address_displacement): Update
24170         accordingly.
24171         * targhooks.c (default_legitimize_address_displacement): Likewise.
24172         * doc/tm.texi: Regenerate.
24173         * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
24174         as an argument, moving assert of ad->disp == ad->disp_term to...
24175         (process_address_1): ...here.  Update calls to base_plus_disp_to_reg.
24176         Try calling targetm.legitimize_address_displacement before expanding
24177         the address rather than afterwards, and adjust for the new interface.
24178         * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
24179         Match the new hook interface.  Handle SVE addresses.
24180         * config/sh/sh.c (sh_legitimize_address_displacement): Make the
24181         new hook interface.
24183 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
24185         * Makefile.in (OBJS): Add early-remat.o.
24186         * target.def (select_early_remat_modes): New hook.
24187         * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
24188         * doc/tm.texi: Regenerate.
24189         * targhooks.h (default_select_early_remat_modes): Declare.
24190         * targhooks.c (default_select_early_remat_modes): New function.
24191         * timevar.def (TV_EARLY_REMAT): New timevar.
24192         * passes.def (pass_early_remat): New pass.
24193         * tree-pass.h (make_pass_early_remat): Declare.
24194         * early-remat.c: New file.
24195         * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
24196         function.
24197         (TARGET_SELECT_EARLY_REMAT_MODES): Define.
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         * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
24204         vfm1 with a bound_epilog parameter.
24205         (vect_do_peeling): Update calls accordingly, and move the prologue
24206         call earlier in the function.  Treat the base bound_epilog as 0 for
24207         fully-masked loops and retain vf - 1 for other loops.  Add 1 to
24208         this base when peeling for gaps.
24209         * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
24210         with fully-masked loops.
24211         (vect_estimate_min_profitable_iters): Handle the single peeled
24212         iteration in that case.
24214 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
24215             Alan Hayward  <alan.hayward@arm.com>
24216             David Sherwood  <david.sherwood@arm.com>
24218         * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
24219         single-element interleaving even if the size is not a power of 2.
24220         * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
24221         accesses for single-element interleaving if the group size is
24222         not a power of 2.
24224 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
24225             Alan Hayward  <alan.hayward@arm.com>
24226             David Sherwood  <david.sherwood@arm.com>
24228         * doc/md.texi (fold_extract_last_@var{m}): Document.
24229         * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
24230         * optabs.def (fold_extract_last_optab): New optab.
24231         * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
24232         * internal-fn.c (fold_extract_direct): New macro.
24233         (expand_fold_extract_optab_fn): Likewise.
24234         (direct_fold_extract_optab_supported_p): Likewise.
24235         * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
24236         * tree-vect-loop.c (vect_model_reduction_cost): Handle
24237         EXTRACT_LAST_REDUCTION.
24238         (get_initial_def_for_reduction): Do not create an initial vector
24239         for EXTRACT_LAST_REDUCTION reductions.
24240         (vectorizable_reduction): Leave the scalar phi in place for
24241         EXTRACT_LAST_REDUCTIONs.  Try using EXTRACT_LAST_REDUCTION
24242         ahead of INTEGER_INDUC_COND_REDUCTION.  Do not check for an
24243         epilogue code for EXTRACT_LAST_REDUCTION and defer the
24244         transform phase to vectorizable_condition.
24245         * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
24246         split out from...
24247         (vect_finish_stmt_generation): ...here.
24248         (vect_finish_replace_stmt): New function.
24249         (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
24250         * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
24251         pattern.
24252         * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
24254 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
24255             Alan Hayward  <alan.hayward@arm.com>
24256             David Sherwood  <david.sherwood@arm.com>
24258         * doc/md.texi (extract_last_@var{m}): Document.
24259         * optabs.def (extract_last_optab): New optab.
24260         * internal-fn.def (EXTRACT_LAST): New internal function.
24261         * internal-fn.c (cond_unary_direct): New macro.
24262         (expand_cond_unary_optab_fn): Likewise.
24263         (direct_cond_unary_optab_supported_p): Likewise.
24264         * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
24265         loops using EXTRACT_LAST.
24266         * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
24267         (extract_last_<mode>): ...this optab.
24268         (vec_extract<mode><Vel>): Update accordingly.
24270 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
24271             Alan Hayward  <alan.hayward@arm.com>
24272             David Sherwood  <david.sherwood@arm.com>
24274         * target.def (empty_mask_is_expensive): New hook.
24275         * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
24276         * doc/tm.texi: Regenerate.
24277         * targhooks.h (default_empty_mask_is_expensive): Declare.
24278         * targhooks.c (default_empty_mask_is_expensive): New function.
24279         * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
24280         if the target says that empty masks are expensive.
24281         * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
24282         New function.
24283         (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
24285 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
24286             Alan Hayward  <alan.hayward@arm.com>
24287             David Sherwood  <david.sherwood@arm.com>
24289         * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
24290         (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
24291         (vect_use_loop_mask_for_alignment_p): New function.
24292         (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
24293         * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
24294         niters_skip argument.  Make sure that the first niters_skip elements
24295         of the first iteration are inactive.
24296         (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
24297         Update call to vect_set_loop_masks_directly.
24298         (get_misalign_in_elems): New function, split out from...
24299         (vect_gen_prolog_loop_niters): ...here.
24300         (vect_update_init_of_dr): Take a code argument that specifies whether
24301         the adjustment should be added or subtracted.
24302         (vect_update_init_of_drs): Likewise.
24303         (vect_prepare_for_masked_peels): New function.
24304         (vect_do_peeling): Skip prologue peeling if we're using a mask
24305         instead.  Update call to vect_update_inits_of_drs.
24306         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
24307         mask_skip_niters.
24308         (vect_analyze_loop_2): Allow fully-masked loops with peeling for
24309         alignment.  Do not include the number of peeled iterations in
24310         the minimum threshold in that case.
24311         (vectorizable_induction): Adjust the start value down by
24312         LOOP_VINFO_MASK_SKIP_NITERS iterations.
24313         (vect_transform_loop): Call vect_prepare_for_masked_peels.
24314         Take the number of skipped iterations into account when calculating
24315         the loop bounds.
24316         * tree-vect-stmts.c (vect_gen_while_not): New function.
24318 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
24319             Alan Hayward  <alan.hayward@arm.com>
24320             David Sherwood  <david.sherwood@arm.com>
24322         * doc/sourcebuild.texi (vect_fully_masked): Document.
24323         * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
24324         default value to 0.
24325         * tree-vect-loop.c (vect_analyze_loop_costing): New function,
24326         split out from...
24327         (vect_analyze_loop_2): ...here. Don't check the vectorization
24328         factor against the number of loop iterations if the loop is
24329         fully-masked.
24331 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
24332             Alan Hayward  <alan.hayward@arm.com>
24333             David Sherwood  <david.sherwood@arm.com>
24335         * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
24336         (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
24337         (dump_groups): Update accordingly.
24338         (iv_use::mem_type): New member variable.
24339         (address_p): New function.
24340         (record_use): Add a mem_type argument and initialize the new
24341         mem_type field.
24342         (record_group_use): Add a mem_type argument.  Use address_p.
24343         Remove obsolete null checks of base_object.  Update call to record_use.
24344         (find_interesting_uses_op): Update call to record_group_use.
24345         (find_interesting_uses_cond): Likewise.
24346         (find_interesting_uses_address): Likewise.
24347         (get_mem_type_for_internal_fn): New function.
24348         (find_address_like_use): Likewise.
24349         (find_interesting_uses_stmt): Try find_address_like_use before
24350         calling find_interesting_uses_op.
24351         (addr_offset_valid_p): Use the iv mem_type field as the type
24352         of the addressed memory.
24353         (add_autoinc_candidates): Likewise.
24354         (get_address_cost): Likewise.
24355         (split_small_address_groups_p): Use address_p.
24356         (split_address_groups): Likewise.
24357         (add_iv_candidate_for_use): Likewise.
24358         (autoinc_possible_for_pair): Likewise.
24359         (rewrite_groups): Likewise.
24360         (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
24361         (determine_group_iv_cost): Update after split of USE_ADDRESS.
24362         (get_alias_ptr_type_for_ptr_address): New function.
24363         (rewrite_use_address): Rewrite address uses in calls that were
24364         identified by find_address_like_use.
24366 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
24367             Alan Hayward  <alan.hayward@arm.com>
24368             David Sherwood  <david.sherwood@arm.com>
24370         * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
24371         TARGET_MEM_REFs.
24372         * gimple-expr.h (is_gimple_addressable: Likewise.
24373         * gimple-expr.c (is_gimple_address): Likewise.
24374         * internal-fn.c (expand_call_mem_ref): New function.
24375         (expand_mask_load_optab_fn): Use it.
24376         (expand_mask_store_optab_fn): Likewise.
24378 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
24379             Alan Hayward  <alan.hayward@arm.com>
24380             David Sherwood  <david.sherwood@arm.com>
24382         * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
24383         (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
24384         (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
24385         (cond_umax@var{mode}): Document.
24386         * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
24387         (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
24388         (cond_umin_optab, cond_umax_optab): New optabs.
24389         * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
24390         (COND_IOR, COND_XOR): New internal functions.
24391         * internal-fn.h (get_conditional_internal_fn): Declare.
24392         * internal-fn.c (cond_binary_direct): New macro.
24393         (expand_cond_binary_optab_fn): Likewise.
24394         (direct_cond_binary_optab_supported_p): Likewise.
24395         (get_conditional_internal_fn): New function.
24396         * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
24397         Cope with reduction statements that are vectorized as calls rather
24398         than assignments.
24399         * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
24400         * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
24401         (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
24402         (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
24403         (UNSPEC_COND_EOR): New unspecs.
24404         (optab): Add mappings for them.
24405         (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
24406         (sve_int_op, sve_fp_op): New int attributes.
24408 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
24409             Alan Hayward  <alan.hayward@arm.com>
24410             David Sherwood  <david.sherwood@arm.com>
24412         * optabs.def (while_ult_optab): New optab.
24413         * doc/md.texi (while_ult@var{m}@var{n}): Document.
24414         * internal-fn.def (WHILE_ULT): New internal function.
24415         * internal-fn.h (direct_internal_fn_supported_p): New override
24416         that takes two types as argument.
24417         * internal-fn.c (while_direct): New macro.
24418         (expand_while_optab_fn): New function.
24419         (convert_optab_supported_p): Likewise.
24420         (direct_while_optab_supported_p): New macro.
24421         * wide-int.h (wi::udiv_ceil): New function.
24422         * tree-vectorizer.h (rgroup_masks): New structure.
24423         (vec_loop_masks): New typedef.
24424         (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
24425         and fully_masked_p.
24426         (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
24427         (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
24428         (vect_max_vf): New function.
24429         (slpeel_make_loop_iterate_ntimes): Delete.
24430         (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
24431         (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
24432         (vect_record_loop_mask, vect_get_loop_mask): Likewise.
24433         * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
24434         internal-fn.h, stor-layout.h and optabs-query.h.
24435         (vect_set_loop_mask): New function.
24436         (add_preheader_seq): Likewise.
24437         (add_header_seq): Likewise.
24438         (interleave_supported_p): Likewise.
24439         (vect_maybe_permute_loop_masks): Likewise.
24440         (vect_set_loop_masks_directly): Likewise.
24441         (vect_set_loop_condition_masked): Likewise.
24442         (vect_set_loop_condition_unmasked): New function, split out from
24443         slpeel_make_loop_iterate_ntimes.
24444         (slpeel_make_loop_iterate_ntimes): Rename to..
24445         (vect_set_loop_condition): ...this.  Use vect_set_loop_condition_masked
24446         for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
24447         (vect_do_peeling): Update call accordingly.
24448         (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
24449         loops.
24450         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
24451         mask_compare_type, can_fully_mask_p and fully_masked_p.
24452         (release_vec_loop_masks): New function.
24453         (_loop_vec_info): Use it to free the loop masks.
24454         (can_produce_all_loop_masks_p): New function.
24455         (vect_get_max_nscalars_per_iter): Likewise.
24456         (vect_verify_full_masking): Likewise.
24457         (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
24458         retries, and free the mask rgroups before retrying.  Check loop-wide
24459         reasons for disallowing fully-masked loops.  Make the final decision
24460         about whether use a fully-masked loop or not.
24461         (vect_estimate_min_profitable_iters): Do not assume that peeling
24462         for the number of iterations will be needed for fully-masked loops.
24463         (vectorizable_reduction): Disable fully-masked loops.
24464         (vectorizable_live_operation): Likewise.
24465         (vect_halve_mask_nunits): New function.
24466         (vect_double_mask_nunits): Likewise.
24467         (vect_record_loop_mask): Likewise.
24468         (vect_get_loop_mask): Likewise.
24469         (vect_transform_loop): Handle the case in which the final loop
24470         iteration might handle a partial vector.  Call vect_set_loop_condition
24471         instead of slpeel_make_loop_iterate_ntimes.
24472         * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
24473         (check_load_store_masking): New function.
24474         (prepare_load_store_mask): Likewise.
24475         (vectorizable_store): Handle fully-masked loops.
24476         (vectorizable_load): Likewise.
24477         (supportable_widening_operation): Use vect_halve_mask_nunits for
24478         booleans.
24479         (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
24480         (vect_gen_while): New function.
24481         * config/aarch64/aarch64.md (umax<mode>3): New expander.
24482         (aarch64_uqdec<mode>): New insn.
24484 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
24485             Alan Hayward  <alan.hayward@arm.com>
24486             David Sherwood  <david.sherwood@arm.com>
24488         * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
24489         (reduc_xor_scal_optab): New optabs.
24490         * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
24491         (reduc_xor_scal_@var{m}): Document.
24492         * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
24493         * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
24494         internal functions.
24495         * fold-const-call.c (fold_const_call): Handle them.
24496         * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
24497         internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
24498         * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
24499         (*reduc_<bit_reduc>_scal_<mode>): New patterns.
24500         * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
24501         (UNSPEC_XORV): New unspecs.
24502         (optab): Add entries for them.
24503         (BITWISEV): New int iterator.
24504         (bit_reduc_op): New int attributes.
24506 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
24507             Alan Hayward  <alan.hayward@arm.com>
24508             David Sherwood  <david.sherwood@arm.com>
24510         * doc/md.texi (vec_shl_insert_@var{m}): New optab.
24511         * internal-fn.def (VEC_SHL_INSERT): New internal function.
24512         * optabs.def (vec_shl_insert_optab): New optab.
24513         * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
24514         (duplicate_and_interleave): Likewise.
24515         * tree-vect-loop.c: Include internal-fn.h.
24516         (neutral_op_for_slp_reduction): New function, split out from
24517         get_initial_defs_for_reduction.
24518         (get_initial_def_for_reduction): Handle option 2 for variable-length
24519         vectors by loading the neutral value into a vector and then shifting
24520         the initial value into element 0.
24521         (get_initial_defs_for_reduction): Replace the code argument with
24522         the neutral value calculated by neutral_op_for_slp_reduction.
24523         Use gimple_build_vector for constant-length vectors.
24524         Use IFN_VEC_SHL_INSERT for variable-length vectors if all
24525         but the first group_size elements have a neutral value.
24526         Use duplicate_and_interleave otherwise.
24527         (vect_create_epilog_for_reduction): Take a neutral_op parameter.
24528         Update call to get_initial_defs_for_reduction.  Handle SLP
24529         reductions for variable-length vectors by creating one vector
24530         result for each scalar result, with the elements associated
24531         with other scalar results stubbed out with the neutral value.
24532         (vectorizable_reduction): Call neutral_op_for_slp_reduction.
24533         Require IFN_VEC_SHL_INSERT for double reductions on
24534         variable-length vectors, or SLP reductions that have
24535         a neutral value.  Require can_duplicate_and_interleave_p
24536         support for variable-length unchained SLP reductions if there
24537         is no neutral value, such as for MIN/MAX reductions.  Also require
24538         the number of vector elements to be a multiple of the number of
24539         SLP statements when doing variable-length unchained SLP reductions.
24540         Update call to vect_create_epilog_for_reduction.
24541         * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
24542         and remove initial values.
24543         (duplicate_and_interleave): Make public.
24544         * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
24545         * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
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         * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
24552         (can_duplicate_and_interleave_p): New function.
24553         (vect_get_and_check_slp_defs): Take the vector of statements
24554         rather than just the current one.  Remove excess parentheses.
24555         Restriction rejectinon of vect_constant_def and vect_external_def
24556         for variable-length vectors to boolean types, or types for which
24557         can_duplicate_and_interleave_p is false.
24558         (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
24559         (duplicate_and_interleave): New function.
24560         (vect_get_constant_vectors): Use gimple_build_vector for
24561         constant-length vectors and suitable variable-length constant
24562         vectors.  Use duplicate_and_interleave for other variable-length
24563         vectors.  Don't defer the update when inserting new statements.
24565 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
24566             Alan Hayward  <alan.hayward@arm.com>
24567             David Sherwood  <david.sherwood@arm.com>
24569         * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
24570         min_profitable_iters doesn't go negative.
24572 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
24573             Alan Hayward  <alan.hayward@arm.com>
24574             David Sherwood  <david.sherwood@arm.com>
24576         * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
24577         (vec_mask_store_lanes@var{m}@var{n}): Likewise.
24578         * optabs.def (vec_mask_load_lanes_optab): New optab.
24579         (vec_mask_store_lanes_optab): Likewise.
24580         * internal-fn.def (MASK_LOAD_LANES): New internal function.
24581         (MASK_STORE_LANES): Likewise.
24582         * internal-fn.c (mask_load_lanes_direct): New macro.
24583         (mask_store_lanes_direct): Likewise.
24584         (expand_mask_load_optab_fn): Handle masked operations.
24585         (expand_mask_load_lanes_optab_fn): New macro.
24586         (expand_mask_store_optab_fn): Handle masked operations.
24587         (expand_mask_store_lanes_optab_fn): New macro.
24588         (direct_mask_load_lanes_optab_supported_p): Likewise.
24589         (direct_mask_store_lanes_optab_supported_p): Likewise.
24590         * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
24591         parameter.
24592         (vect_load_lanes_supported): Likewise.
24593         * tree-vect-data-refs.c (strip_conversion): New function.
24594         (can_group_stmts_p): Likewise.
24595         (vect_analyze_data_ref_accesses): Use it instead of checking
24596         for a pair of assignments.
24597         (vect_store_lanes_supported): Take a masked_p parameter.
24598         (vect_load_lanes_supported): Likewise.
24599         * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
24600         vect_store_lanes_supported and vect_load_lanes_supported.
24601         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
24602         * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
24603         parameter.  Don't allow gaps for masked accesses.
24604         Use vect_get_store_rhs.  Update calls to vect_store_lanes_supported
24605         and vect_load_lanes_supported.
24606         (get_load_store_type): Take a masked_p parameter and update
24607         call to get_group_load_store_type.
24608         (vectorizable_store): Update call to get_load_store_type.
24609         Handle IFN_MASK_STORE_LANES.
24610         (vectorizable_load): Update call to get_load_store_type.
24611         Handle IFN_MASK_LOAD_LANES.
24613 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
24614             Alan Hayward  <alan.hayward@arm.com>
24615             David Sherwood  <david.sherwood@arm.com>
24617         * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
24618         modes for SVE.
24619         * config/aarch64/aarch64-protos.h
24620         (aarch64_sve_struct_memory_operand_p): Declare.
24621         * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
24622         (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
24623         (VPRED, vpred): Handle SVE structure modes.
24624         * config/aarch64/constraints.md (Utx): New constraint.
24625         * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
24626         (aarch64_sve_struct_nonimmediate_operand): New predicates.
24627         * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
24628         * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
24629         (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
24630         structure modes.  Split into pieces after RA.
24631         (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
24632         (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
24633         New patterns.
24634         * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
24635         SVE structure modes.
24636         (aarch64_classify_address): Likewise.
24637         (sizetochar): Move earlier in file.
24638         (aarch64_print_operand): Handle SVE register lists.
24639         (aarch64_array_mode): New function.
24640         (aarch64_sve_struct_memory_operand_p): Likewise.
24641         (TARGET_ARRAY_MODE): Redefine.
24643 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
24644             Alan Hayward  <alan.hayward@arm.com>
24645             David Sherwood  <david.sherwood@arm.com>
24647         * target.def (array_mode): New target hook.
24648         * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
24649         * doc/tm.texi: Regenerate.
24650         * hooks.h (hook_optmode_mode_uhwi_none): Declare.
24651         * hooks.c (hook_optmode_mode_uhwi_none): New function.
24652         * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
24653         targetm.array_mode.
24654         * stor-layout.c (mode_for_array): Likewise.  Support polynomial
24655         type sizes.
24657 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
24658             Alan Hayward  <alan.hayward@arm.com>
24659             David Sherwood  <david.sherwood@arm.com>
24661         * fold-const.c (fold_binary_loc): Check the argument types
24662         rather than the result type when testing for a vector operation.
24664 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
24666         * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
24667         * doc/tm.texi: Regenerate.
24669 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
24670             Alan Hayward  <alan.hayward@arm.com>
24671             David Sherwood  <david.sherwood@arm.com>
24673         * doc/invoke.texi (-msve-vector-bits=): Document new option.
24674         (sve): Document new AArch64 extension.
24675         * doc/md.texi (w): Extend the description of the AArch64
24676         constraint to include SVE vectors.
24677         (Upl, Upa): Document new AArch64 predicate constraints.
24678         * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
24679         enum.
24680         * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
24681         (msve-vector-bits=): New option.
24682         * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
24683         SVE when these are disabled.
24684         (sve): New extension.
24685         * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
24686         modes.  Adjust their number of units based on aarch64_sve_vg.
24687         (MAX_BITSIZE_MODE_ANY_MODE): Define.
24688         * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
24689         aarch64_addr_query_type.
24690         (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
24691         (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
24692         (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
24693         (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
24694         (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
24695         (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
24696         (aarch64_simd_imm_zero_p): Delete.
24697         (aarch64_check_zero_based_sve_index_immediate): Declare.
24698         (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
24699         (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
24700         (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
24701         (aarch64_sve_float_mul_immediate_p): Likewise.
24702         (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
24703         rather than an rtx.
24704         (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
24705         (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
24706         (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
24707         (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
24708         (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
24709         (aarch64_regmode_natural_size): Likewise.
24710         * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
24711         (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
24712         left one place.
24713         (AARCH64_ISA_SVE, TARGET_SVE): New macros.
24714         (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
24715         for VG and the SVE predicate registers.
24716         (V_ALIASES): Add a "z"-prefixed alias.
24717         (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
24718         (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
24719         (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
24720         (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
24721         (REG_CLASS_NAMES): Add entries for them.
24722         (REG_CLASS_CONTENTS): Likewise.  Update ALL_REGS to include VG
24723         and the predicate registers.
24724         (aarch64_sve_vg): Declare.
24725         (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
24726         (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
24727         (REGMODE_NATURAL_SIZE): Define.
24728         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
24729         SVE macros.
24730         * config/aarch64/aarch64.c: Include cfgrtl.h.
24731         (simd_immediate_info): Add a constructor for series vectors,
24732         and an associated step field.
24733         (aarch64_sve_vg): New variable.
24734         (aarch64_dbx_register_number): Handle VG and the predicate registers.
24735         (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
24736         (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
24737         (VEC_ANY_DATA, VEC_STRUCT): New constants.
24738         (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
24739         (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
24740         (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
24741         (aarch64_get_mask_mode): New functions.
24742         (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
24743         and FP_LO_REGS.  Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
24744         (aarch64_hard_regno_mode_ok): Handle VG.  Also handle the SVE
24745         predicate modes and predicate registers.  Explicitly restrict
24746         GPRs to modes of 16 bytes or smaller.  Only allow FP registers
24747         to store a vector mode if it is recognized by
24748         aarch64_classify_vector_mode.
24749         (aarch64_regmode_natural_size): New function.
24750         (aarch64_hard_regno_caller_save_mode): Return the original mode
24751         for predicates.
24752         (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
24753         (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
24754         (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
24755         (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
24756         functions.
24757         (aarch64_add_offset): Add a temp2 parameter.  Assert that temp1
24758         does not overlap dest if the function is frame-related.  Handle
24759         SVE constants.
24760         (aarch64_split_add_offset): New function.
24761         (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
24762         them aarch64_add_offset.
24763         (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
24764         and update call to aarch64_sub_sp.
24765         (aarch64_add_cfa_expression): New function.
24766         (aarch64_expand_prologue): Pass extra temporary registers to the
24767         functions above.  Handle the case in which we need to emit new
24768         DW_CFA_expressions for registers that were originally saved
24769         relative to the stack pointer, but now have to be expressed
24770         relative to the frame pointer.
24771         (aarch64_output_mi_thunk): Pass extra temporary registers to the
24772         functions above.
24773         (aarch64_expand_epilogue): Likewise.  Prevent inheritance of
24774         IP0 and IP1 values for SVE frames.
24775         (aarch64_expand_vec_series): New function.
24776         (aarch64_expand_sve_widened_duplicate): Likewise.
24777         (aarch64_expand_sve_const_vector): Likewise.
24778         (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
24779         Handle SVE constants.  Use emit_move_insn to move a force_const_mem
24780         into the register, rather than emitting a SET directly.
24781         (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
24782         (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
24783         (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
24784         (offset_9bit_signed_scaled_p): New functions.
24785         (aarch64_replicate_bitmask_imm): New function.
24786         (aarch64_bitmask_imm): Use it.
24787         (aarch64_cannot_force_const_mem): Reject expressions involving
24788         a CONST_POLY_INT.  Update call to aarch64_classify_symbol.
24789         (aarch64_classify_index): Handle SVE indices, by requiring
24790         a plain register index with a scale that matches the element size.
24791         (aarch64_classify_address): Handle SVE addresses.  Assert that
24792         the mode of the address is VOIDmode or an integer mode.
24793         Update call to aarch64_classify_symbol.
24794         (aarch64_classify_symbolic_expression): Update call to
24795         aarch64_classify_symbol.
24796         (aarch64_const_vec_all_in_range_p): New function.
24797         (aarch64_print_vector_float_operand): Likewise.
24798         (aarch64_print_operand): Handle 'N' and 'C'.  Use "zN" rather than
24799         "vN" for FP registers with SVE modes.  Handle (const ...) vectors
24800         and the FP immediates 1.0 and 0.5.
24801         (aarch64_print_address_internal): Handle SVE addresses.
24802         (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
24803         (aarch64_regno_regclass): Handle predicate registers.
24804         (aarch64_secondary_reload): Handle big-endian reloads of SVE
24805         data modes.
24806         (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
24807         (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
24808         (aarch64_convert_sve_vector_bits): New function.
24809         (aarch64_override_options): Use it to handle -msve-vector-bits=.
24810         (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
24811         rather than an rtx.
24812         (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
24813         Handle SVE vector and predicate modes.  Accept VL-based constants
24814         that need only one temporary register, and VL offsets that require
24815         no temporary registers.
24816         (aarch64_conditional_register_usage): Mark the predicate registers
24817         as fixed if SVE isn't available.
24818         (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
24819         Return true for SVE vector and predicate modes.
24820         (aarch64_simd_container_mode): Take the number of bits as a poly_int64
24821         rather than an unsigned int.  Handle SVE modes.
24822         (aarch64_preferred_simd_mode): Update call accordingly.  Handle
24823         SVE modes.
24824         (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
24825         if SVE is enabled.
24826         (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
24827         (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
24828         (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
24829         (aarch64_sve_float_mul_immediate_p): New functions.
24830         (aarch64_sve_valid_immediate): New function.
24831         (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
24832         Explicitly reject structure modes.  Check for INDEX constants.
24833         Handle PTRUE and PFALSE constants.
24834         (aarch64_check_zero_based_sve_index_immediate): New function.
24835         (aarch64_simd_imm_zero_p): Delete.
24836         (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
24837         vector modes.  Accept constants in the range of CNT[BHWD].
24838         (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
24839         ask for an Advanced SIMD mode.
24840         (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
24841         (aarch64_simd_vector_alignment): Handle SVE predicates.
24842         (aarch64_vectorize_preferred_vector_alignment): New function.
24843         (aarch64_simd_vector_alignment_reachable): Use it instead of
24844         the vector size.
24845         (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
24846         (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
24847         functions.
24848         (MAX_VECT_LEN): Delete.
24849         (expand_vec_perm_d): Add a vec_flags field.
24850         (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
24851         (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
24852         (aarch64_evpc_ext): Don't apply a big-endian lane correction
24853         for SVE modes.
24854         (aarch64_evpc_rev): Rename to...
24855         (aarch64_evpc_rev_local): ...this.  Use a predicated operation for SVE.
24856         (aarch64_evpc_rev_global): New function.
24857         (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
24858         (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
24859         MAX_VECT_LEN.
24860         (aarch64_evpc_sve_tbl): New function.
24861         (aarch64_expand_vec_perm_const_1): Update after rename of
24862         aarch64_evpc_rev.  Handle SVE permutes too, trying
24863         aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
24864         than aarch64_evpc_tbl.
24865         (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
24866         (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
24867         (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
24868         (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
24869         (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
24870         (aarch64_expand_sve_vcond): New functions.
24871         (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
24872         of aarch64_vector_mode_p.
24873         (aarch64_dwarf_poly_indeterminate_value): New function.
24874         (aarch64_compute_pressure_classes): Likewise.
24875         (aarch64_can_change_mode_class): Likewise.
24876         (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
24877         (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
24878         (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
24879         (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
24880         (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
24881         (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
24882         * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
24883         (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
24884         constraints.
24885         (Dn, Dl, Dr): Accept const as well as const_vector.
24886         (Dz): Likewise.  Compare against CONST0_RTX.
24887         * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
24888         of "vector" where appropriate.
24889         (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
24890         (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
24891         (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
24892         (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
24893         (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
24894         (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
24895         (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
24896         (v_int_equiv): Extend to SVE modes.
24897         (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
24898         mode attributes.
24899         (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
24900         (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
24901         (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
24902         (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
24903         (SVE_COND_FP_CMP): New int iterators.
24904         (perm_hilo): Handle the new unpack unspecs.
24905         (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
24906         attributes.
24907         * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
24908         (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
24909         (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
24910         (aarch64_equality_operator, aarch64_constant_vector_operand)
24911         (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
24912         (aarch64_sve_nonimmediate_operand): Likewise.
24913         (aarch64_sve_general_operand): Likewise.
24914         (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
24915         (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
24916         (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
24917         (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
24918         (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
24919         (aarch64_sve_float_arith_immediate): Likewise.
24920         (aarch64_sve_float_arith_with_sub_immediate): Likewise.
24921         (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
24922         (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
24923         (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
24924         (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
24925         (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
24926         (aarch64_sve_float_arith_operand): Likewise.
24927         (aarch64_sve_float_arith_with_sub_operand): Likewise.
24928         (aarch64_sve_float_mul_operand): Likewise.
24929         (aarch64_sve_vec_perm_operand): Likewise.
24930         (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
24931         (aarch64_mov_operand): Accept const_poly_int and const_vector.
24932         (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
24933         as well as const_vector.
24934         (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
24935         in file.  Use CONST0_RTX and CONSTM1_RTX.
24936         (aarch64_simd_or_scalar_imm_zero): Likewise.  Add match_codes.
24937         (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
24938         Use aarch64_simd_imm_zero.
24939         * config/aarch64/aarch64-sve.md: New file.
24940         * config/aarch64/aarch64.md: Include it.
24941         (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
24942         (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
24943         (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
24944         (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
24945         (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
24946         (sve): New attribute.
24947         (enabled): Disable instructions with the sve attribute unless
24948         TARGET_SVE.
24949         (movqi, movhi): Pass CONST_POLY_INT operaneds through
24950         aarch64_expand_mov_immediate.
24951         (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
24952         CNT[BHSD] immediates.
24953         (movti): Split CONST_POLY_INT moves into two halves.
24954         (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
24955         Split additions that need a temporary here if the destination
24956         is the stack pointer.
24957         (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
24958         (*add<mode>3_poly_1): New instruction.
24959         (set_clobber_cc): New expander.
24961 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
24963         * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
24964         parameter and use it instead of GET_MODE_SIZE (innermode).  Use
24965         inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
24966         Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
24967         GET_MODE_NUNITS (innermode).  Also add a first_elem parameter.
24968         Change innermode from fixed_mode_size to machine_mode.
24969         (simplify_subreg): Update call accordingly.  Handle a constant-sized
24970         subreg of a variable-length CONST_VECTOR.
24972 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
24973             Alan Hayward  <alan.hayward@arm.com>
24974             David Sherwood  <david.sherwood@arm.com>
24976         * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
24977         (add_offset_to_base): New function, split out from...
24978         (create_mem_ref): ...here.  When handling a scale other than 1,
24979         check first whether the address is valid without the offset.
24980         Add it into the base if so, leaving the index and scale as-is.
24982 2018-01-12  Jakub Jelinek  <jakub@redhat.com>
24984         PR c++/83778
24985         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
24986         fold_for_warn before checking if arg2 is INTEGER_CST.
24988 2018-01-12  Segher Boessenkool  <segher@kernel.crashing.org>
24990         * config/rs6000/predicates.md (load_multiple_operation): Delete.
24991         (store_multiple_operation): Delete.
24992         * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
24993         * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
24994         * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
24995         guarded by TARGET_STRING.
24996         (rs6000_output_load_multiple): Delete.
24997         * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
24998         OPTION_MASK_STRING / TARGET_STRING handling.
24999         (print_operand) <'N', 'O'>: Add comment that these are unused now.
25000         (const rs6000_opt_masks) <"string">: Change mask to 0.
25001         * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
25002         (MASK_STRING): Delete.
25003         * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
25004         parts.  Simplify.
25005         (load_multiple): Delete.
25006         (*ldmsi8): Delete.
25007         (*ldmsi7): Delete.
25008         (*ldmsi6): Delete.
25009         (*ldmsi5): Delete.
25010         (*ldmsi4): Delete.
25011         (*ldmsi3): Delete.
25012         (store_multiple): Delete.
25013         (*stmsi8): Delete.
25014         (*stmsi7): Delete.
25015         (*stmsi6): Delete.
25016         (*stmsi5): Delete.
25017         (*stmsi4): Delete.
25018         (*stmsi3): Delete.
25019         (movmemsi_8reg): Delete.
25020         (corresponding unnamed define_insn): Delete.
25021         (movmemsi_6reg): Delete.
25022         (corresponding unnamed define_insn): Delete.
25023         (movmemsi_4reg): Delete.
25024         (corresponding unnamed define_insn): Delete.
25025         (movmemsi_2reg): Delete.
25026         (corresponding unnamed define_insn): Delete.
25027         (movmemsi_1reg): Delete.
25028         (corresponding unnamed define_insn): Delete.
25029         * config/rs6000/rs6000.opt (mno-string): New.
25030         (mstring): Replace by deprecation warning stub.
25031         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
25033 2018-01-12  Jakub Jelinek  <jakub@redhat.com>
25035         * regrename.c (regrename_do_replace): If replacing the same
25036         reg multiple times, try to reuse last created gen_raw_REG.
25038         PR debug/81155
25039         * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
25040         main to workaround a bug in GDB.
25042 2018-01-12  Tom de Vries  <tom@codesourcery.com>
25044         PR target/83737
25045         * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
25047 2018-01-12  Vladimir Makarov  <vmakarov@redhat.com>
25049         PR rtl-optimization/80481
25050         * ira-color.c (get_cap_member): New function.
25051         (allocnos_conflict_by_live_ranges_p): Use it.
25052         (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
25053         (setup_slot_coalesced_allocno_live_ranges): Ditto.
25055 2018-01-12  Uros Bizjak  <ubizjak@gmail.com>
25057         PR target/83628
25058         * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
25059         (*saddl_se_1): Ditto.
25060         (*ssubsi_1): Ditto.
25061         (*ssubl_se_1): Ditto.
25063 2018-01-12  Richard Sandiford  <richard.sandiford@linaro.org>
25065         * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
25066         rather than wi::to_widest for DR_INITs.
25067         * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
25068         wi::to_poly_offset rather than wi::to_offset for DR_INIT.
25069         (vect_analyze_data_ref_accesses): Require both DR_INITs to be
25070         INTEGER_CSTs.
25071         (vect_analyze_group_access_1): Note that here.
25073 2018-01-12  Richard Sandiford  <richard.sandiford@linaro.org>
25075         * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
25076         polynomial type sizes.
25078 2018-01-12  Richard Sandiford  <richard.sandiford@linaro.org>
25080         * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
25081         poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
25082         (gimple_add_tmp_var): Likewise.
25084 2018-01-12  Martin Liska  <mliska@suse.cz>
25086         * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
25087         (gimple_alloc_sizes): Likewise.
25088         (dump_gimple_statistics): Use PRIu64 in printf format.
25089         * gimple.h: Change uint64_t to int.
25091 2018-01-12  Martin Liska  <mliska@suse.cz>
25093         * tree-core.h: Use uint64_t instead of int.
25094         * tree.c (tree_node_counts): Likewise.
25095         (tree_node_sizes): Likewise.
25096         (dump_tree_statistics): Use PRIu64 in printf format.
25098 2018-01-12  Martin Liska  <mliska@suse.cz>
25100         * Makefile.in: As qsort_chk is implemented in vec.c, add
25101         vec.o to linkage of gencfn-macros.
25102         * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
25103         passing the info to record_node_allocation_statistics.
25104         (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
25105         and pass the info.
25106         * ggc-common.c (struct ggc_usage): Add operator== and use
25107         it in operator< and compare function.
25108         * mem-stats.h (struct mem_usage): Likewise.
25109         * vec.c (struct vec_usage): Remove operator< and compare
25110         function. Can be simply inherited.
25112 2018-01-12  Martin Jambor  <mjambor@suse.cz>
25114         PR target/81616
25115         * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
25116         * tree-ssa-math-opts.c: Include domwalk.h.
25117         (convert_mult_to_fma_1): New function.
25118         (fma_transformation_info): New type.
25119         (fma_deferring_state): Likewise.
25120         (cancel_fma_deferring): New function.
25121         (result_of_phi): Likewise.
25122         (last_fma_candidate_feeds_initial_phi): Likewise.
25123         (convert_mult_to_fma): Added deferring logic, split actual
25124         transformation to convert_mult_to_fma_1.
25125         (math_opts_dom_walker): New type.
25126         (math_opts_dom_walker::after_dom_children): New method, body moved
25127         here from pass_optimize_widening_mul::execute, added deferring logic
25128         bits.
25129         (pass_optimize_widening_mul::execute): Moved most of code to
25130         math_opts_dom_walker::after_dom_children.
25131         * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
25132         * config/i386/i386.c (ix86_option_override_internal): Added
25133         maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
25135 2018-01-12  Richard Biener  <rguenther@suse.de>
25137         PR debug/83157
25138         * dwarf2out.c (gen_variable_die): Do not reset old_die for
25139         inline instance vars.
25141 2018-01-12  Oleg Endo  <olegendo@gcc.gnu.org>
25143         PR target/81819
25144         * config/rx/rx.c (rx_is_restricted_memory_address):
25145         Handle SUBREG case.
25147 2018-01-12  Richard Biener  <rguenther@suse.de>
25149         PR tree-optimization/80846
25150         * target.def (split_reduction): New target hook.
25151         * targhooks.c (default_split_reduction): New function.
25152         * targhooks.h (default_split_reduction): Declare.
25153         * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
25154         target requests first reduce vectors by combining low and high
25155         parts.
25156         * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
25157         (get_vectype_for_scalar_type_and_size): Export.
25158         * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
25159         * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
25160         * doc/tm.texi: Regenerate.
25161         * config/i386/i386.c (ix86_split_reduction): Implement
25162         TARGET_VECTORIZE_SPLIT_REDUCTION.
25164 2018-01-12  Eric Botcazou  <ebotcazou@adacore.com>
25166         PR target/83368
25167         * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
25168         in PIC mode except for TARGET_VXWORKS_RTP.
25169         * config/sparc/sparc.c: Include cfgrtl.h.
25170         (TARGET_INIT_PIC_REG): Define.
25171         (TARGET_USE_PSEUDO_PIC_REG): Likewise.
25172         (sparc_pic_register_p): New predicate.
25173         (sparc_legitimate_address_p): Use it.
25174         (sparc_legitimize_pic_address): Likewise.
25175         (sparc_delegitimize_address): Likewise.
25176         (sparc_mode_dependent_address_p): Likewise.
25177         (gen_load_pcrel_sym): Remove 4th parameter.
25178         (load_got_register): Adjust call to above.  Remove obsolete stuff.
25179         (sparc_expand_prologue): Do not call load_got_register here.
25180         (sparc_flat_expand_prologue): Likewise.
25181         (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
25182         (sparc_use_pseudo_pic_reg): New function.
25183         (sparc_init_pic_reg): Likewise.
25184         * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
25185         (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
25187 2018-01-12  Christophe Lyon  <christophe.lyon@linaro.org>
25189         * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
25190         Add item for branch_cost.
25192 2018-01-12  Eric Botcazou  <ebotcazou@adacore.com>
25194         PR rtl-optimization/83565
25195         * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
25196         not extend the result to a larger mode for rotate operations.
25197         (num_sign_bit_copies1): Likewise.
25199 2018-01-12  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
25201         PR target/40411
25202         * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
25203         -symbolic.
25204         Use values-Xc.o for -pedantic.
25205         Link with values-xpg4.o for C90, values-xpg6.o otherwise.
25207 2018-01-12  Martin Liska  <mliska@suse.cz>
25209         PR ipa/83054
25210         * ipa-devirt.c (final_warning_record::grow_type_warnings):
25211         New function.
25212         (possible_polymorphic_call_targets): Use it.
25213         (ipa_devirt): Likewise.
25215 2018-01-12  Martin Liska  <mliska@suse.cz>
25217         * profile-count.h (enum profile_quality): Use 0 as invalid
25218         enum value of profile_quality.
25220 2018-01-12  Chung-Ju Wu  <jasonwucj@gmail.com>
25222         * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
25223         -mext-string options.
25225 2018-01-12  Richard Biener  <rguenther@suse.de>
25227         * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
25228         DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
25229         * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
25230         Likewise.
25231         * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
25233 2018-01-11  Michael Meissner  <meissner@linux.vnet.ibm.com>
25235         * configure.ac (--with-long-double-format): Add support for the
25236         configuration option to change the default long double format on
25237         PowerPC systems.
25238         * config.gcc (powerpc*-linux*-*): Likewise.
25239         * configure: Regenerate.
25240         * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
25241         double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
25242         used without modification.
25244 2018-01-11  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
25246         * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
25247         (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
25248         * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
25249         MISC_BUILTIN_SPEC_BARRIER.
25250         (rs6000_init_builtins): Likewise.
25251         * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
25252         enum value.
25253         (speculation_barrier): New define_insn.
25254         * doc/extend.texi: Document __builtin_speculation_barrier.
25256 2018-01-11  Jakub Jelinek  <jakub@redhat.com>
25258         PR target/83203
25259         * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
25260         is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
25261         * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
25262         iterators.
25263         (ssescalarmodesuffix): Add 512-bit vectors.  Use "d" or "q" for
25264         integral modes instead of "ss" and "sd".
25265         (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
25266         vectors with 32-bit and 64-bit elements.
25267         (vecdupssescalarmodesuffix): New mode attribute.
25268         (vec_dup<mode>): Use it.
25270 2018-01-11  H.J. Lu  <hongjiu.lu@intel.com>
25272         PR target/83330
25273         * config/i386/i386.c (ix86_compute_frame_layout): Align stack
25274         frame if argument is passed on stack.
25276 2018-01-11  Jakub Jelinek  <jakub@redhat.com>
25278         PR target/82682
25279         * ree.c (combine_reaching_defs): Optimize also
25280         reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
25281         reg2=any_extend(exp); reg1=reg2;, formatting fix.
25283 2018-01-11  Jan Hubicka  <hubicka@ucw.cz>
25285         PR middle-end/83189
25286         * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
25288 2018-01-11  Jan Hubicka  <hubicka@ucw.cz>
25290         PR middle-end/83718
25291         * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
25292         after they are computed.
25294 2018-01-11  Bin Cheng  <bin.cheng@arm.com>
25296         PR tree-optimization/83695
25297         * gimple-loop-linterchange.cc
25298         (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
25299         reset cached scev information after interchange.
25300         (pass_linterchange::execute): Remove call to scev_reset_htab.
25302 2018-01-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
25304         * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
25305         vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
25306         vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
25307         vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
25308         vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
25309         vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
25310         * config/arm/arm_neon_builtins.def (vfmal_lane_low,
25311         vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
25312         vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
25313         vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
25314         vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
25315         * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
25316         (V_lane_reg): Likewise.
25317         * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
25318         New define_expand.
25319         (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
25320         (vfmal_lane_low<mode>_intrinsic,
25321         vfmal_lane_low<vfmlsel2><mode>_intrinsic,
25322         vfmal_lane_high<vfmlsel2><mode>_intrinsic,
25323         vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
25324         vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
25325         vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
25326         vfmsl_lane_high<mode>_intrinsic): New define_insns.
25328 2018-01-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
25330         * config/arm/arm-cpus.in (fp16fml): New feature.
25331         (ALL_SIMD): Add fp16fml.
25332         (armv8.2-a): Add fp16fml as an option.
25333         (armv8.3-a): Likewise.
25334         (armv8.4-a): Add fp16fml as part of fp16.
25335         * config/arm/arm.h (TARGET_FP16FML): Define.
25336         * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
25337         when appropriate.
25338         * config/arm/arm-modes.def (V2HF): Define.
25339         * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
25340         vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
25341         vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
25342         * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
25343         vfmsl_low, vfmsl_high): New set of builtins.
25344         * config/arm/iterators.md (PLUSMINUS): New code iterator.
25345         (vfml_op): New code attribute.
25346         (VFMLHALVES): New int iterator.
25347         (VFML, VFMLSEL): New mode attributes.
25348         (V_reg): Define mapping for V2HF.
25349         (V_hi, V_lo): New mode attributes.
25350         (VF_constraint): Likewise.
25351         (vfml_half, vfml_half_selector): New int attributes.
25352         * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
25353         define_expand.
25354         (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
25355         vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
25356         New define_insn.
25357         * config/arm/t-arm-elf (v8_fps): Add fp16fml.
25358         * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
25359         * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
25360         * doc/invoke.texi (ARM Options): Document fp16fml.  Update armv8.4-a
25361         documentation.
25362         * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
25363         Document new effective target and option set.
25365 2018-01-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
25367         * config/arm/arm-cpus.in (armv8_4): New feature.
25368         (ARMv8_4a): New fgroup.
25369         (armv8.4-a): New arch.
25370         * config/arm/arm-tables.opt: Regenerate.
25371         * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
25372         * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
25373         * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
25374         Add matching rules for -march=armv8.4-a and extensions.
25375         * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
25377 2018-01-11  Oleg Endo  <olegendo@gcc.gnu.org>
25379         PR target/81821
25380         * config/rx/rx.md (BW): New mode attribute.
25381         (sync_lock_test_and_setsi): Add mode suffix to insn output.
25383 2018-01-11  Richard Biener  <rguenther@suse.de>
25385         PR tree-optimization/83435
25386         * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
25387         * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
25388         * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
25390 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
25391             Alan Hayward  <alan.hayward@arm.com>
25392             David Sherwood  <david.sherwood@arm.com>
25394         * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
25395         field.
25396         (aarch64_classify_address): Initialize it.  Track polynomial offsets.
25397         (aarch64_print_address_internal): Use it to check for a zero offset.
25399 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
25400             Alan Hayward  <alan.hayward@arm.com>
25401             David Sherwood  <david.sherwood@arm.com>
25403         * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
25404         * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
25405         Return a poly_int64 rather than a HOST_WIDE_INT.
25406         (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
25407         rather than a HOST_WIDE_INT.
25408         * config/aarch64/aarch64.h (aarch64_frame): Protect with
25409         HAVE_POLY_INT_H rather than HOST_WIDE_INT.  Change locals_offset,
25410         hard_fp_offset, frame_size, initial_adjust, callee_offset and
25411         final_offset from HOST_WIDE_INT to poly_int64.
25412         * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
25413         to_constant when getting the number of units in an Advanced SIMD
25414         mode.
25415         (aarch64_builtin_vectorized_function): Check for a constant number
25416         of units.
25417         * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
25418         GET_MODE_SIZE.
25419         (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
25420         attribute instead of GET_MODE_NUNITS.
25421         * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
25422         (aarch64_class_max_nregs): Use the constant_lowest_bound of the
25423         GET_MODE_SIZE for fixed-size registers.
25424         (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
25425         (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
25426         (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
25427         (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
25428         (aarch64_print_operand, aarch64_print_address_internal)
25429         (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
25430         (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
25431         (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
25432         Handle polynomial GET_MODE_SIZE.
25433         (aarch64_hard_regno_caller_save_mode): Likewise.  Return modes
25434         wider than SImode without modification.
25435         (tls_symbolic_operand_type): Use strip_offset instead of split_const.
25436         (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
25437         (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
25438         passing and returning SVE modes.
25439         (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
25440         rather than GEN_INT.
25441         (aarch64_emit_probe_stack_range): Take the size as a poly_int64
25442         rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
25443         (aarch64_allocate_and_probe_stack_space): Likewise.
25444         (aarch64_layout_frame): Cope with polynomial offsets.
25445         (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
25446         start_offset as a poly_int64 rather than a HOST_WIDE_INT.  Track
25447         polynomial offsets.
25448         (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
25449         (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
25450         poly_int64 rather than a HOST_WIDE_INT.
25451         (aarch64_get_separate_components, aarch64_process_components)
25452         (aarch64_expand_prologue, aarch64_expand_epilogue)
25453         (aarch64_use_return_insn_p): Handle polynomial frame offsets.
25454         (aarch64_anchor_offset): New function, split out from...
25455         (aarch64_legitimize_address): ...here.
25456         (aarch64_builtin_vectorization_cost): Handle polynomial
25457         TYPE_VECTOR_SUBPARTS.
25458         (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
25459         GET_MODE_NUNITS.
25460         (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
25461         number of elements from the PARALLEL rather than the mode.
25462         (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
25463         rather than GET_MODE_BITSIZE.
25464         (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
25465         (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
25466         (aarch64_expand_vec_perm_const_1): Handle polynomial
25467         d->perm.length () and d->perm elements.
25468         (aarch64_evpc_tbl): Likewise.  Use nelt rather than GET_MODE_NUNITS.
25469         Apply to_constant to d->perm elements.
25470         (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
25471         polynomial CONST_VECTOR_NUNITS.
25472         (aarch64_move_pointer): Take amount as a poly_int64 rather
25473         than an int.
25474         (aarch64_progress_pointer): Avoid temporary variable.
25475         * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
25476         the mode attribute instead of GET_MODE.
25478 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
25479             Alan Hayward  <alan.hayward@arm.com>
25480             David Sherwood  <david.sherwood@arm.com>
25482         * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
25483         x exists before using it.
25484         (aarch64_add_constant_internal): Rename to...
25485         (aarch64_add_offset_1): ...this.  Replace regnum with separate
25486         src and dest rtxes.  Handle the case in which they're different,
25487         including when the offset is zero.  Replace scratchreg with an rtx.
25488         Use 2 additions if there is no spare register into which we can
25489         move a 16-bit constant.
25490         (aarch64_add_constant): Delete.
25491         (aarch64_add_offset): Replace reg with separate src and dest
25492         rtxes.  Take a poly_int64 offset instead of a HOST_WIDE_INT.
25493         Use aarch64_add_offset_1.
25494         (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
25495         an rtx rather than an int.  Take the delta as a poly_int64
25496         rather than a HOST_WIDE_INT.  Use aarch64_add_offset.
25497         (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
25498         (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
25499         aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
25500         (aarch64_expand_epilogue): Update calls to aarch64_add_offset
25501         and aarch64_add_sp.
25502         (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
25503         aarch64_add_constant.
25505 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
25507         * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
25508         Use scalar_float_mode.
25510 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
25512         * config/aarch64/aarch64-simd.md
25513         (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
25514         (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
25515         (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
25516         (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
25517         (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
25518         (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
25519         (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
25520         (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
25521         (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
25522         (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
25524 2018-01-11  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
25526         PR target/83514
25527         * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
25528         targ_options->x_arm_arch_string is non NULL.
25530 2018-01-11  Tamar Christina  <tamar.christina@arm.com>
25532         * config/aarch64/aarch64.h
25533         (AARCH64_FL_FOR_ARCH8_4): Add  AARCH64_FL_DOTPROD.
25535 2018-01-11  Sudakshina Das  <sudi.das@arm.com>
25537         PR target/82096
25538         * expmed.c (emit_store_flag_force): Swap if const op0
25539         and change VOIDmode to mode of op0.
25541 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
25543         PR rtl-optimization/83761
25544         * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
25545         than bytes to mode_for_size.
25547 2018-01-10  Jan Hubicka  <hubicka@ucw.cz>
25549         PR middle-end/83189
25550         * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
25551         * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
25552         profile.
25554 2018-01-10  Jan Hubicka  <hubicka@ucw.cz>
25556         PR middle-end/83575
25557         * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
25558         when in layout mode.
25559         (cfg_layout_finalize): Do not verify cfg before we are out of layout.
25560         * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
25561         partition fixup.
25563 2018-01-10  Michael Collison  <michael.collison@arm.com>
25565         * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
25566         * config/aarch64/aarch64-option-extension.def: Add
25567         AARCH64_OPT_EXTENSION of 'fp16fml'.
25568         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
25569         (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
25570         * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
25571         * config/aarch64/constraints.md (Ui7): New constraint.
25572         * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
25573         (VFMLA_SEL_W): Ditto.
25574         (f16quad): Ditto.
25575         (f16mac1): Ditto.
25576         (VFMLA16_LOW): New int iterator.
25577         (VFMLA16_HIGH): Ditto.
25578         (UNSPEC_FMLAL): New unspec.
25579         (UNSPEC_FMLSL): Ditto.
25580         (UNSPEC_FMLAL2): Ditto.
25581         (UNSPEC_FMLSL2): Ditto.
25582         (f16mac): New code attribute.
25583         * config/aarch64/aarch64-simd-builtins.def
25584         (aarch64_fmlal_lowv2sf): Ditto.
25585         (aarch64_fmlsl_lowv2sf): Ditto.
25586         (aarch64_fmlalq_lowv4sf): Ditto.
25587         (aarch64_fmlslq_lowv4sf): Ditto.
25588         (aarch64_fmlal_highv2sf): Ditto.
25589         (aarch64_fmlsl_highv2sf): Ditto.
25590         (aarch64_fmlalq_highv4sf): Ditto.
25591         (aarch64_fmlslq_highv4sf): Ditto.
25592         (aarch64_fmlal_lane_lowv2sf): Ditto.
25593         (aarch64_fmlsl_lane_lowv2sf): Ditto.
25594         (aarch64_fmlal_laneq_lowv2sf): Ditto.
25595         (aarch64_fmlsl_laneq_lowv2sf): Ditto.
25596         (aarch64_fmlalq_lane_lowv4sf): Ditto.
25597         (aarch64_fmlsl_lane_lowv4sf): Ditto.
25598         (aarch64_fmlalq_laneq_lowv4sf): Ditto.
25599         (aarch64_fmlsl_laneq_lowv4sf): Ditto.
25600         (aarch64_fmlal_lane_highv2sf): Ditto.
25601         (aarch64_fmlsl_lane_highv2sf): Ditto.
25602         (aarch64_fmlal_laneq_highv2sf): Ditto.
25603         (aarch64_fmlsl_laneq_highv2sf): Ditto.
25604         (aarch64_fmlalq_lane_highv4sf): Ditto.
25605         (aarch64_fmlsl_lane_highv4sf): Ditto.
25606         (aarch64_fmlalq_laneq_highv4sf): Ditto.
25607         (aarch64_fmlsl_laneq_highv4sf): Ditto.
25608         * config/aarch64/aarch64-simd.md:
25609         (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
25610         (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
25611         (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
25612         (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
25613         (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
25614         (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
25615         (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
25616         (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
25617         (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
25618         (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
25619         (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
25620         (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
25621         (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
25622         (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
25623         (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
25624         (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
25625         (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
25626         (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
25627         (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
25628         (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
25629         * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
25630         (vfmlsl_low_u32): Ditto.
25631         (vfmlalq_low_u32): Ditto.
25632         (vfmlslq_low_u32): Ditto.
25633         (vfmlal_high_u32): Ditto.
25634         (vfmlsl_high_u32): Ditto.
25635         (vfmlalq_high_u32): Ditto.
25636         (vfmlslq_high_u32): Ditto.
25637         (vfmlal_lane_low_u32): Ditto.
25638         (vfmlsl_lane_low_u32): Ditto.
25639         (vfmlal_laneq_low_u32): Ditto.
25640         (vfmlsl_laneq_low_u32): Ditto.
25641         (vfmlalq_lane_low_u32): Ditto.
25642         (vfmlslq_lane_low_u32): Ditto.
25643         (vfmlalq_laneq_low_u32): Ditto.
25644         (vfmlslq_laneq_low_u32): Ditto.
25645         (vfmlal_lane_high_u32): Ditto.
25646         (vfmlsl_lane_high_u32): Ditto.
25647         (vfmlal_laneq_high_u32): Ditto.
25648         (vfmlsl_laneq_high_u32): Ditto.
25649         (vfmlalq_lane_high_u32): Ditto.
25650         (vfmlslq_lane_high_u32): Ditto.
25651         (vfmlalq_laneq_high_u32): Ditto.
25652         (vfmlslq_laneq_high_u32): Ditto.
25653         * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
25654         (AARCH64_FL_FOR_ARCH8_4): New.
25655         (AARCH64_ISA_F16FML): New ISA flag.
25656         (TARGET_F16FML): New feature flag for fp16fml.
25657         (doc/invoke.texi): Document new fp16fml option.
25659 2018-01-10  Michael Collison  <michael.collison@arm.com>
25661         * config/aarch64/aarch64-builtins.c:
25662         (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
25663         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
25664         (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
25665         * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
25666         (AARCH64_ISA_SHA3): New ISA flag.
25667         (TARGET_SHA3): New feature flag for sha3.
25668         * config/aarch64/iterators.md (sha512_op): New int attribute.
25669         (CRYPTO_SHA512): New int iterator.
25670         (UNSPEC_SHA512H): New unspec.
25671         (UNSPEC_SHA512H2): Ditto.
25672         (UNSPEC_SHA512SU0): Ditto.
25673         (UNSPEC_SHA512SU1): Ditto.
25674         * config/aarch64/aarch64-simd-builtins.def
25675         (aarch64_crypto_sha512hqv2di): New builtin.
25676         (aarch64_crypto_sha512h2qv2di): Ditto.
25677         (aarch64_crypto_sha512su0qv2di): Ditto.
25678         (aarch64_crypto_sha512su1qv2di): Ditto.
25679         (aarch64_eor3qv8hi): Ditto.
25680         (aarch64_rax1qv2di): Ditto.
25681         (aarch64_xarqv2di): Ditto.
25682         (aarch64_bcaxqv8hi): Ditto.
25683         * config/aarch64/aarch64-simd.md:
25684         (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
25685         (aarch64_crypto_sha512su0qv2di): Ditto.
25686         (aarch64_crypto_sha512su1qv2di): Ditto.
25687         (aarch64_eor3qv8hi): Ditto.
25688         (aarch64_rax1qv2di): Ditto.
25689         (aarch64_xarqv2di): Ditto.
25690         (aarch64_bcaxqv8hi): Ditto.
25691         * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
25692         (vsha512h2q_u64): Ditto.
25693         (vsha512su0q_u64): Ditto.
25694         (vsha512su1q_u64): Ditto.
25695         (veor3q_u16): Ditto.
25696         (vrax1q_u64): Ditto.
25697         (vxarq_u64): Ditto.
25698         (vbcaxq_u16): Ditto.
25699         * config/arm/types.md (crypto_sha512): New type attribute.
25700         (crypto_sha3): Ditto.
25701         (doc/invoke.texi): Document new sha3 option.
25703 2018-01-10  Michael Collison  <michael.collison@arm.com>
25705         * config/aarch64/aarch64-builtins.c:
25706         (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
25707         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
25708         (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
25709         (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
25710         * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
25711         (AARCH64_ISA_SM4): New ISA flag.
25712         (TARGET_SM4): New feature flag for sm4.
25713         * config/aarch64/aarch64-simd-builtins.def
25714         (aarch64_sm3ss1qv4si): Ditto.
25715         (aarch64_sm3tt1aq4si): Ditto.
25716         (aarch64_sm3tt1bq4si): Ditto.
25717         (aarch64_sm3tt2aq4si): Ditto.
25718         (aarch64_sm3tt2bq4si): Ditto.
25719         (aarch64_sm3partw1qv4si): Ditto.
25720         (aarch64_sm3partw2qv4si): Ditto.
25721         (aarch64_sm4eqv4si): Ditto.
25722         (aarch64_sm4ekeyqv4si): Ditto.
25723         * config/aarch64/aarch64-simd.md:
25724         (aarch64_sm3ss1qv4si): Ditto.
25725         (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
25726         (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
25727         (aarch64_sm4eqv4si): Ditto.
25728         (aarch64_sm4ekeyqv4si): Ditto.
25729         * config/aarch64/iterators.md (sm3tt_op): New int iterator.
25730         (sm3part_op): Ditto.
25731         (CRYPTO_SM3TT): Ditto.
25732         (CRYPTO_SM3PART): Ditto.
25733         (UNSPEC_SM3SS1): New unspec.
25734         (UNSPEC_SM3TT1A): Ditto.
25735         (UNSPEC_SM3TT1B): Ditto.
25736         (UNSPEC_SM3TT2A): Ditto.
25737         (UNSPEC_SM3TT2B): Ditto.
25738         (UNSPEC_SM3PARTW1): Ditto.
25739         (UNSPEC_SM3PARTW2): Ditto.
25740         (UNSPEC_SM4E): Ditto.
25741         (UNSPEC_SM4EKEY): Ditto.
25742         * config/aarch64/constraints.md (Ui2): New constraint.
25743         * config/aarch64/predicates.md (aarch64_imm2): New predicate.
25744         * config/arm/types.md (crypto_sm3): New type attribute.
25745         (crypto_sm4): Ditto.
25746         * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
25747         (vsm3tt1aq_u32): Ditto.
25748         (vsm3tt1bq_u32): Ditto.
25749         (vsm3tt2aq_u32): Ditto.
25750         (vsm3tt2bq_u32): Ditto.
25751         (vsm3partw1q_u32): Ditto.
25752         (vsm3partw2q_u32): Ditto.
25753         (vsm4eq_u32): Ditto.
25754         (vsm4ekeyq_u32): Ditto.
25755         (doc/invoke.texi): Document new sm4 option.
25757 2018-01-10  Michael Collison  <michael.collison@arm.com>
25759         * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
25760         * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
25761         (AARCH64_FL_FOR_ARCH8_4): New.
25762         (AARCH64_FL_V8_4): New flag.
25763         (doc/invoke.texi): Document new armv8.4-a option.
25765 2018-01-10  Michael Collison  <michael.collison@arm.com>
25767         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
25768         (__ARM_FEATURE_AES): Define if TARGET_AES is true.
25769         (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
25770         * config/aarch64/aarch64-option-extension.def: Add
25771         AARCH64_OPT_EXTENSION of 'sha2'.
25772         (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
25773         (crypto): Disable sha2 and aes if crypto disabled.
25774         (crypto): Enable aes and sha2 if enabled.
25775         (simd): Disable sha2 and aes if simd disabled.
25776         * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
25777         New flags.
25778         (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
25779         (TARGET_SHA2): New feature flag for sha2.
25780         (TARGET_AES): New feature flag for aes.
25781         * config/aarch64/aarch64-simd.md:
25782         (aarch64_crypto_aes<aes_op>v16qi): Make pattern
25783         conditional on TARGET_AES.
25784         (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
25785         (aarch64_crypto_sha1hsi): Make pattern conditional
25786         on TARGET_SHA2.
25787         (aarch64_crypto_sha1hv4si): Ditto.
25788         (aarch64_be_crypto_sha1hv4si): Ditto.
25789         (aarch64_crypto_sha1su1v4si): Ditto.
25790         (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
25791         (aarch64_crypto_sha1su0v4si): Ditto.
25792         (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
25793         (aarch64_crypto_sha256su0v4si): Ditto.
25794         (aarch64_crypto_sha256su1v4si): Ditto.
25795         (doc/invoke.texi): Document new aes and sha2 options.
25797 2018-01-10  Martin Sebor  <msebor@redhat.com>
25799         PR tree-optimization/83781
25800         * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
25801         as string arrays.
25803 2018-01-11  Martin Sebor  <msebor@gmail.com>
25804             Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
25806         PR tree-optimization/83501
25807         PR tree-optimization/81703
25809         * tree-ssa-strlen.c (get_string_cst): Rename...
25810         (get_string_len): ...to this.  Handle global constants.
25811         (handle_char_store): Adjust.
25813 2018-01-10  Kito Cheng  <kito.cheng@gmail.com>
25814             Jim Wilson  <jimw@sifive.com>
25816         * config/riscv/riscv-protos.h (riscv_output_return): New.
25817         * config/riscv/riscv.c (struct machine_function): New naked_p field.
25818         (riscv_attribute_table, riscv_output_return),
25819         (riscv_handle_fndecl_attribute, riscv_naked_function_p),
25820         (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
25821         (riscv_compute_frame_info): Only compute frame->mask if not a naked
25822         function.
25823         (riscv_expand_prologue): Add early return for naked function.
25824         (riscv_expand_epilogue): Likewise.
25825         (riscv_function_ok_for_sibcall): Return false for naked function.
25826         (riscv_set_current_function): New.
25827         (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
25828         (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
25829         * config/riscv/riscv.md (simple_return): Call riscv_output_return.
25830         * doc/extend.texi (RISC-V Function Attributes): New.
25832 2018-01-10  Michael Meissner  <meissner@linux.vnet.ibm.com>
25834         * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
25835         check for 128-bit long double before checking TCmode.
25836         * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
25837         128-bit long doubles before checking TFmode or TCmode.
25838         (FLOAT128_IBM_P): Likewise.
25840 2018-01-10  Martin Sebor  <msebor@redhat.com>
25842         PR tree-optimization/83671
25843         * builtins.c (c_strlen): Unconditionally return zero for the empty
25844         string.
25845         Use -Warray-bounds for warnings.
25846         * gimple-fold.c (get_range_strlen): Handle non-constant lengths
25847         for non-constant array indices with COMPONENT_REF, arrays of
25848         arrays, and pointers to arrays.
25849         (gimple_fold_builtin_strlen): Determine and set length range for
25850         non-constant character arrays.
25852 2018-01-10  Aldy Hernandez  <aldyh@redhat.com>
25854         PR middle-end/81897
25855         * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
25856         empty blocks.
25858 2018-01-10  Eric Botcazou  <ebotcazou@adacore.com>
25860         * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
25862 2018-01-10  Peter Bergner  <bergner@vnet.ibm.com>
25864         PR target/83399
25865         * config/rs6000/rs6000.c (print_operand) <'y'>: Use
25866         VECTOR_MEM_ALTIVEC_OR_VSX_P.
25867         * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
25868         indexed_or_indirect_operand predicate.
25869         (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
25870         (*vsx_le_perm_load_v8hi): Likewise.
25871         (*vsx_le_perm_load_v16qi): Likewise.
25872         (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
25873         (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
25874         (*vsx_le_perm_store_v8hi): Likewise.
25875         (*vsx_le_perm_store_v16qi): Likewise.
25876         (eight unnamed splitters): Likewise.
25878 2018-01-10  Peter Bergner  <bergner@vnet.ibm.com>
25880         * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
25881         * config/rs6000/emmintrin.h: Likewise.
25882         * config/rs6000/mmintrin.h: Likewise.
25883         * config/rs6000/xmmintrin.h: Likewise.
25885 2018-01-10  David Malcolm  <dmalcolm@redhat.com>
25887         PR c++/43486
25888         * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
25889         "public_flag".
25890         * tree.c (tree_nop_conversion): Return true for location wrapper
25891         nodes.
25892         (maybe_wrap_with_location): New function.
25893         (selftest::check_strip_nops): New function.
25894         (selftest::test_location_wrappers): New function.
25895         (selftest::tree_c_tests): Call it.
25896         * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
25897         (maybe_wrap_with_location): New decl.
25898         (EXPR_LOCATION_WRAPPER_P): New macro.
25899         (location_wrapper_p): New inline function.
25900         (tree_strip_any_location_wrapper): New inline function.
25902 2018-01-10  H.J. Lu  <hongjiu.lu@intel.com>
25904         PR target/83735
25905         * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
25906         stack_realign_offset for the largest alignment of stack slot
25907         actually used.
25908         (ix86_find_max_used_stack_alignment): New function.
25909         (ix86_finalize_stack_frame_flags): Use it.  Set
25910         max_used_stack_alignment if we don't realign stack.
25911         * config/i386/i386.h (machine_function): Add
25912         max_used_stack_alignment.
25914 2018-01-10  Christophe Lyon  <christophe.lyon@linaro.org>
25916         * config/arm/arm.opt (-mbranch-cost): New option.
25917         * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
25918         account.
25920 2018-01-10  Segher Boessenkool  <segher@kernel.crashing.org>
25922         PR target/83629
25923         * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
25924         load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
25926 2018-01-10  Richard Biener  <rguenther@suse.de>
25928         PR debug/83765
25929         * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
25930         early out so it also covers the case where we have a non-NULL
25931         origin.
25933 2018-01-10  Richard Sandiford  <richard.sandiford@linaro.org>
25935         PR tree-optimization/83753
25936         * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
25937         for non-strided grouped accesses if the number of elements is 1.
25939 2018-01-10  Jan Hubicka  <hubicka@ucw.cz>
25941         PR target/81616
25942         * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
25943         * i386.h (TARGET_USE_GATHER): Define.
25944         * x86-tune.def (X86_TUNE_USE_GATHER): New.
25946 2018-01-10  Martin Liska  <mliska@suse.cz>
25948         PR bootstrap/82831
25949         * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
25950         * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
25951         partitioning.
25952         * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
25953         CLEANUP_NO_PARTITIONING is not set.
25955 2018-01-10  Richard Sandiford  <richard.sandiford@linaro.org>
25957         * doc/rtl.texi: Remove documentation of (const ...) wrappers
25958         for vectors, as a partial revert of r254296.
25959         * rtl.h (const_vec_p): Delete.
25960         (const_vec_duplicate_p): Don't test for vector CONSTs.
25961         (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
25962         * expmed.c (make_tree): Likewise.
25964         Revert:
25965         * common.md (E, F): Use CONSTANT_P instead of checking for
25966         CONST_VECTOR.
25967         * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
25968         checking for CONST_VECTOR.
25970 2018-01-09  Jan Hubicka  <hubicka@ucw.cz>
25972         PR middle-end/83575
25973         * predict.c (force_edge_cold): Handle in more sane way edges
25974         with no prediction.
25976 2018-01-09  Carl Love  <cel@us.ibm.com>
25978         * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
25979         V4SI, V4SF types.
25980         (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
25981         * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
25982         VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
25983         VMRGOW_V2DI, VMRGOW_V2DF.  Remove definition for VMRGOW.
25984         * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
25985         P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW):  Add definitions.
25986         * config/rs6000/rs6000-protos.h: Add extern defition for
25987         rs6000_generate_float2_double_code.
25988         * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
25989         function.
25990         * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
25991         (float2_v2df): Add define_expand.
25993 2018-01-09  Uros Bizjak  <ubizjak@gmail.com>
25995         PR target/83628
25996         * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
25997         op_mode in the force_to_mode call.
25999 2018-01-09  Richard Sandiford  <richard.sandiford@linaro.org>
26001         * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
26002         instead of checking each element individually.
26003         (aarch64_evpc_uzp): Likewise.
26004         (aarch64_evpc_zip): Likewise.
26005         (aarch64_evpc_ext): Likewise.
26006         (aarch64_evpc_rev): Likewise.
26007         (aarch64_evpc_dup): Test the encoding for a single duplicated element,
26008         instead of checking each element individually.  Return true without
26009         generating rtl if
26010         (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
26011         whether all selected elements come from the same input, instead of
26012         checking each element individually.  Remove calls to gen_rtx_REG,
26013         start_sequence and end_sequence and instead assert that no rtl is
26014         generated.
26016 2018-01-09  Richard Sandiford  <richard.sandiford@linaro.org>
26018         * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
26019         order of HIGH and CONST checks.
26021 2018-01-09  Richard Sandiford  <richard.sandiford@linaro.org>
26023         * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
26024         if the destination isn't an SSA_NAME.
26026 2018-01-09  Richard Biener  <rguenther@suse.de>
26028         PR tree-optimization/83668
26029         * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
26030         move prologue...
26031         (canonicalize_loop_form): ... here, renamed from ...
26032         (canonicalize_loop_closed_ssa_form): ... this and amended to
26033         swap successor edges for loop exit blocks to make us use
26034         the RPO order we need for initial schedule generation.
26036 2018-01-09  Joseph Myers  <joseph@codesourcery.com>
26038         PR tree-optimization/64811
26039         * match.pd: When optimizing comparisons with Inf, avoid
26040         introducing or losing exceptions from comparisons with NaN.
26042 2018-01-09  Martin Liska  <mliska@suse.cz>
26044         PR sanitizer/82517
26045         * asan.c (shadow_mem_size): Add gcc_assert.
26047 2018-01-09  Georg-Johann Lay  <avr@gjlay.de>
26049         Don't save registers in main().
26051         PR target/83738
26052         * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
26053         * config/avr/avr.opt (-mmain-is-OS_task): New target option.
26054         * config/avr/avr.c (avr_set_current_function): Don't error if
26055         naked, OS_task or OS_main are specified at the same time.
26056         (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
26057         OS_main.
26058         (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
26059         attribute.
26060         * common/config/avr/avr-common.c (avr_option_optimization_table):
26061         Switch on -mmain-is-OS_task for optimizing compilations.
26063 2018-01-09  Richard Biener  <rguenther@suse.de>
26065         PR tree-optimization/83572
26066         * graphite.c: Include cfganal.h.
26067         (graphite_transform_loops): Connect infinite loops to exit
26068         and remove fake edges at the end.
26070 2018-01-09  Jan Hubicka  <hubicka@ucw.cz>
26072         * ipa-inline.c (edge_badness): Revert accidental checkin.
26074 2018-01-09  Jan Hubicka  <hubicka@ucw.cz>
26076         PR ipa/80763
26077         * ipa-comdats.c (set_comdat_group): Only set comdat group of real
26078         symbols; not inline clones.
26080 2018-01-09  Jakub Jelinek  <jakub@redhat.com>
26082         PR target/83507
26083         * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
26084         hard registers.  Formatting fixes.
26086         PR preprocessor/83722
26087         * gcc.c (try_generate_repro): Pass
26088         &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
26089         &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
26090         do_report_bug.
26092 2018-01-08  Monk Chiang  <sh.chiang04@gmail.com>
26093             Kito Cheng  <kito.cheng@gmail.com>
26095         * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
26096         (riscv_leaf_function_p): Delete.
26097         (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
26099 2018-01-08  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
26101         * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
26102         function.
26103         (do_ifelse): New function.
26104         (do_isel): New function.
26105         (do_sub3): New function.
26106         (do_add3): New function.
26107         (do_load_mask_compare): New function.
26108         (do_overlap_load_compare): New function.
26109         (expand_compare_loop): New function.
26110         (expand_block_compare): Call expand_compare_loop() when appropriate.
26111         * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
26112         option description.
26113         (-mblock-compare-inline-loop-limit): New option.
26115 2018-01-08  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
26117         PR target/83677
26118         * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
26119         Reverse order of second and third operands in first alternative.
26120         * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
26121         of first and second elements in UNSPEC_VPERMR vector.
26122         (altivec_expand_vec_perm_le): Likewise.
26124 2018-01-08  Jeff Law  <law@redhat.com>
26126         PR rtl-optimizatin/81308
26127         * tree-switch-conversion.c (cfg_altered): New file scoped static.
26128         (process_switch): If group_case_labels makes a change, then set
26129         cfg_altered.
26130         (pass_convert_switch::execute): If a switch is converted, then
26131         set cfg_altered.  Return TODO_cfg_cleanup if cfg_altered is true.
26133         PR rtl-optimization/81308
26134         * recog.c (split_all_insns): Conditionally cleanup the CFG after
26135         splitting insns.
26137 2018-01-08  Vidya Praveen  <vidyapraveen@arm.com>
26139         PR target/83663 - Revert r255946
26140         * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
26141         generation for cases where splatting a value is not useful.
26142         * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
26143         across a vec_duplicate and a paradoxical subreg forming a vector
26144         mode to a vec_concat.
26146 2018-01-08  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
26148         * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
26149         -march=armv8.3-a variants.
26150         * config/arm/t-multilib: Likewise.
26151         * config/arm/t-arm-elf: Likewise.  Handle dotprod extension.
26153 2018-01-08  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
26155         * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
26156         to generate rtl.
26157         (cceq_ior_compare_complement): Give it a name so I can use it, and
26158         change boolean_or_operator predicate to boolean_operator so it can
26159         be used to generate a crand.
26160         (eqne): New code iterator.
26161         (bd/bd_neg): New code_attrs.
26162         (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
26163         a single define_insn.
26164         (<bd>tf_<mode>): A new insn pattern for the conditional form branch
26165         decrement (bdnzt/bdnzf/bdzt/bdzf).
26166         * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
26167         with the new names of the branch decrement patterns, and added the
26168         names of the branch decrement conditional patterns.
26170 2018-01-08  Richard Biener  <rguenther@suse.de>
26172         PR tree-optimization/83563
26173         * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
26174         cache.
26176 2018-01-08  Richard Biener  <rguenther@suse.de>
26178         PR middle-end/83713
26179         * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
26181 2018-01-08  Richard Biener  <rguenther@suse.de>
26183         PR tree-optimization/83685
26184         * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
26185         references to abnormals.
26187 2018-01-08  Richard Biener  <rguenther@suse.de>
26189         PR lto/83719
26190         * dwarf2out.c (output_indirect_strings): Handle empty
26191         skeleton_debug_str_hash.
26192         (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
26194 2018-01-08  Claudiu Zissulescu  <claziss@synopsys.com>
26196         * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
26197         (emit_store_direct): Likewise.
26198         (arc_trampoline_adjust_address): Likewise.
26199         (arc_asm_trampoline_template): New function.
26200         (arc_initialize_trampoline): Use asm_trampoline_template.
26201         (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
26202         * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
26203         * config/arc/arc.md (flush_icache): Delete pattern.
26205 2018-01-08  Claudiu Zissulescu  <claziss@synopsys.com>
26207         * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
26208         * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
26209         munaligned-access.
26211 2018-01-08  Sebastian Huber  <sebastian.huber@embedded-brains.de>
26213         PR target/83681
26214         * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
26215         by not USED_FOR_TARGET.
26216         (make_pass_resolve_sw_modes): Likewise.
26218 2018-01-08  Sebastian Huber  <sebastian.huber@embedded-brains.de>
26220         * config/nios2/nios2.h (nios2_section_threshold): Guard by not
26221         USED_FOR_TARGET.
26223 2018-01-08  Richard Biener  <rguenther@suse.de>
26225         PR middle-end/83580
26226         * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
26228 2018-01-08  Richard Biener  <rguenther@suse.de>
26230         PR middle-end/83517
26231         * match.pd ((t * 2) / 2) -> t): Add missing :c.
26233 2018-01-06  Aldy Hernandez  <aldyh@redhat.com>
26235         PR middle-end/81897
26236         * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
26237         basic blocks with a small number of successors.
26238         (convert_control_dep_chain_into_preds): Improve handling of
26239         forwarder blocks.
26240         (dump_predicates): Split apart into...
26241         (dump_pred_chain): ...here...
26242         (dump_pred_info): ...and here.
26243         (can_one_predicate_be_invalidated_p): Add debugging printfs.
26244         (can_chain_union_be_invalidated_p): Improve check for invalidation
26245         of paths.
26246         (uninit_uses_cannot_happen): Avoid unnecessary if
26247         convert_control_dep_chain_into_preds yielded nothing.
26249 2018-01-06  Martin Sebor  <msebor@redhat.com>
26251         PR tree-optimization/83640
26252         * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
26253         subtracting negative offset from size.
26254         (builtin_access::overlap): Adjust offset bounds of the access to fall
26255         within the size of the object if possible.
26257 2018-01-06  Richard Sandiford  <richard.sandiford@linaro.org>
26259         PR rtl-optimization/83699
26260         * expmed.c (extract_bit_field_1): Restrict the vector usage of
26261         extract_bit_field_as_subreg to cases in which the extracted
26262         value is also a vector.
26264         * lra-constraints.c (process_alt_operands): Test for the equivalence
26265         substitutions when detecting a possible reload cycle.
26267 2018-01-06  Jakub Jelinek  <jakub@redhat.com>
26269         PR debug/83480
26270         * toplev.c (process_options): Don't enable debug_nonbind_markers_p
26271         by default if flag_selective_schedling{,2}.  Formatting fixes.
26273         PR rtl-optimization/83682
26274         * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
26275         if it has non-VECTOR_MODE element mode.
26276         (vec_duplicate_p): Likewise.
26278         PR middle-end/83694
26279         * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
26280         and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
26282 2018-01-05  Jakub Jelinek  <jakub@redhat.com>
26284         PR target/83604
26285         * config/i386/i386-builtin.def
26286         (__builtin_ia32_vgf2p8affineinvqb_v64qi,
26287         __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
26288         Require also OPTION_MASK_ISA_AVX512F in addition to
26289         OPTION_MASK_ISA_GFNI.
26290         (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
26291         __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
26292         OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
26293         to OPTION_MASK_ISA_GFNI.
26294         (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
26295         OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
26296         OPTION_MASK_ISA_AVX512BW.
26297         (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
26298         OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
26299         addition to OPTION_MASK_ISA_GFNI.
26300         (__builtin_ia32_vgf2p8affineinvqb_v16qi,
26301         __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
26302         Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
26303         to OPTION_MASK_ISA_GFNI.
26304         * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
26305         a requirement for all ISAs rather than any of them with a few
26306         exceptions.
26307         (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
26308         processing.
26309         (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
26310         bitmasks to be enabled with 3 exceptions, instead of requiring any
26311         enabled ISA with lots of exceptions.
26312         * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
26313         vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
26314         Change avx512bw in isa attribute to avx512f.
26315         * config/i386/sgxintrin.h: Add license boilerplate.
26316         * config/i386/vaesintrin.h: Likewise.  Fix macro spelling __AVX512F
26317         to __AVX512F__ and __AVX512VL to __AVX512VL__.
26318         (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
26319         _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
26320         defined.
26321         * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
26322         _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
26323         temporarily sse2 rather than sse if not enabled already.
26325         PR target/83604
26326         * config/i386/sse.md (VI248_VLBW): Rename to ...
26327         (VI248_AVX512VL): ... this.  Don't guard V32HI with TARGET_AVX512BW.
26328         (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
26329         vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
26330         vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
26331         vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
26332         mode iterator instead of VI248_VLBW.
26334 2018-01-05  Jan Hubicka  <hubicka@ucw.cz>
26336         * ipa-fnsummary.c (record_modified_bb_info): Add OP.
26337         (record_modified): Skip clobbers; add debug output.
26338         (param_change_prob): Use sreal frequencies.
26340 2018-01-05  Richard Sandiford  <richard.sandiford@linaro.org>
26342         * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
26343         punt for user-aligned variables.
26345 2018-01-05  Richard Sandiford  <richard.sandiford@linaro.org>
26347         * tree-chrec.c (chrec_contains_symbols): Return true for
26348         POLY_INT_CST.
26350 2018-01-05  Sudakshina Das  <sudi.das@arm.com>
26352         PR target/82439
26353         * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
26354         of (x|y) == x for BICS pattern.
26356 2018-01-05  Jakub Jelinek  <jakub@redhat.com>
26358         PR tree-optimization/83605
26359         * gimple-ssa-strength-reduction.c: Include tree-eh.h.
26360         (find_candidates_dom_walker::before_dom_children): Ignore stmts that
26361         can throw.
26363 2018-01-05  Sebastian Huber  <sebastian.huber@embedded-brains.de>
26365         * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
26366         * config/epiphany/rtems.h: New file.
26368 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
26369             Uros Bizjak  <ubizjak@gmail.com>
26371         PR target/83554
26372         * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
26373         QIreg_operand instead of register_operand predicate.
26374         * config/i386/i386.c (ix86_rop_should_change_byte_p,
26375         set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
26376         comments instead of -fmitigate[-_]rop.
26378 2018-01-04  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
26380         PR bootstrap/81926
26381         * cgraphunit.c (symbol_table::compile): Switch to text_section
26382         before calling assembly_start debug hook.
26383         * run-rtl-passes.c (run_rtl_passes): Likewise.
26384         Include output.h.
26386 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
26388         * tree-vrp.c (extract_range_from_binary_expr_1): Check
26389         range_int_cst_p rather than !symbolic_range_p before calling
26390         extract_range_from_multiplicative_op_1.
26392 2018-01-04  Jeff Law  <law@redhat.com>
26394         * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
26395         redundant test in assertion.
26397 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
26399         * doc/rtl.texi: Document machine_mode wrapper classes.
26401 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
26403         * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
26404         using tree_to_uhwi.
26406 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
26408         * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
26409         the VEC_PERM_EXPR fold to fail.
26411 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
26413         PR debug/83585
26414         * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
26415         to switched_sections.
26417 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
26419         PR target/83680
26420         * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
26421         test for d.testing.
26423 2018-01-04  Peter Bergner  <bergner@vnet.ibm.com>
26425         PR target/83387
26426         * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
26427         allow arguments in FP registers if TARGET_HARD_FLOAT is false.
26429 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
26431         PR debug/83666
26432         * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
26433         is BLKmode and bitpos not zero or mode change is needed.
26435 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
26437         PR target/83675
26438         * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
26439         TARGET_VIS2.
26441 2018-01-04  Uros Bizjak  <ubizjak@gmail.com>
26443         PR target/83628
26444         * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
26445         instead of MULT rtx.  Update all corresponding splitters.
26446         (*saddl_se): Ditto.
26447         (*ssub<modesuffix>): Ditto.
26448         (*ssubl_se): Ditto.
26449         (*cmp_sadd_di): Update split patterns.
26450         (*cmp_sadd_si): Ditto.
26451         (*cmp_sadd_sidi): Ditto.
26452         (*cmp_ssub_di): Ditto.
26453         (*cmp_ssub_si): Ditto.
26454         (*cmp_ssub_sidi): Ditto.
26455         * config/alpha/predicates.md (const23_operand): New predicate.
26456         * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
26457         Look for ASHIFT, not MULT inner operand.
26458         (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
26460 2018-01-04  Martin Liska  <mliska@suse.cz>
26462         PR gcov-profile/83669
26463         * gcov.c (output_intermediate_file): Add version to intermediate
26464         gcov file.
26465         * doc/gcov.texi: Document new field 'version' in intermediate
26466         file format. Fix location of '-k' option of gcov command.
26468 2018-01-04  Martin Liska  <mliska@suse.cz>
26470         PR ipa/82352
26471         * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
26473 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
26475         * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
26477 2018-01-03  Martin Sebor  <msebor@redhat.com>
26479         PR tree-optimization/83655
26480         * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
26481         checking calls with invalid arguments.
26483 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
26485         * tree-vect-stmts.c (vect_get_store_rhs): New function.
26486         (vectorizable_mask_load_store): Delete.
26487         (vectorizable_call): Return false for masked loads and stores.
26488         (vectorizable_store): Handle IFN_MASK_STORE.  Use vect_get_store_rhs
26489         instead of gimple_assign_rhs1.
26490         (vectorizable_load): Handle IFN_MASK_LOAD.
26491         (vect_transform_stmt): Don't set is_store for call_vec_info_type.
26493 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
26495         * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
26496         split out from..,
26497         (vectorizable_mask_load_store): ...here.
26498         (vectorizable_load): ...and here.
26500 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
26502         * tree-vect-stmts.c (vect_build_all_ones_mask)
26503         (vect_build_zero_merge_argument): New functions, split out from...
26504         (vectorizable_load): ...here.
26506 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
26508         * tree-vect-stmts.c (vect_check_store_rhs): New function,
26509         split out from...
26510         (vectorizable_mask_load_store): ...here.
26511         (vectorizable_store): ...and here.
26513 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
26515         * tree-vect-stmts.c (vect_check_load_store_mask): New function,
26516         split out from...
26517         (vectorizable_mask_load_store): ...here.
26519 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
26521         * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
26522         (vect_model_store_cost): Take a vec_load_store_type instead of a
26523         vect_def_type.
26524         * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
26525         (vect_model_store_cost): Take a vec_load_store_type instead of a
26526         vect_def_type.
26527         (vectorizable_mask_load_store): Update accordingly.
26528         (vectorizable_store): Likewise.
26529         * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
26531 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
26533         * tree-vect-loop.c (vect_transform_loop): Stub out scalar
26534         IFN_MASK_LOAD calls here rather than...
26535         * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
26537 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
26538             Alan Hayward  <alan.hayward@arm.com>
26539             David Sherwood  <david.sherwood@arm.com>
26541         * expmed.c (extract_bit_field_1): For vector extracts,
26542         fall back to extract_bit_field_as_subreg if vec_extract
26543         isn't available.
26545 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
26546             Alan Hayward  <alan.hayward@arm.com>
26547             David Sherwood  <david.sherwood@arm.com>
26549         * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
26550         they are variable or constant sized.
26551         (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
26552         slots for constant-sized data.
26554 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
26555             Alan Hayward  <alan.hayward@arm.com>
26556             David Sherwood  <david.sherwood@arm.com>
26558         * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
26559         handling COND_EXPRs with boolean comparisons, try to find a better
26560         basis for the mask type than the boolean itself.
26562 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
26564         * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
26565         is calculated and how it can be overridden.
26566         * genmodes.c (max_bitsize_mode_any_mode): New variable.
26567         (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
26568         if defined.
26569         (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
26570         if nonzero.
26572 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
26573             Alan Hayward  <alan.hayward@arm.com>
26574             David Sherwood  <david.sherwood@arm.com>
26576         * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
26577         Remove the mode argument.
26578         (aarch64_simd_valid_immediate): Remove the mode and inverse
26579         arguments.
26580         * config/aarch64/iterators.md (bitsize): New iterator.
26581         * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
26582         (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
26583         * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
26584         aarch64_simd_valid_immediate.
26585         * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
26586         (aarch64_reg_or_bic_imm): Likewise.
26587         * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
26588         with an insn_type enum and msl with a modifier_type enum.
26589         Replace element_width with a scalar_mode.  Change the shift
26590         to unsigned int.  Add constructors for scalar_float_mode and
26591         scalar_int_mode elements.
26592         (aarch64_vect_float_const_representable_p): Delete.
26593         (aarch64_can_const_movi_rtx_p)
26594         (aarch64_simd_scalar_immediate_valid_for_move)
26595         (aarch64_simd_make_constant): Update call to
26596         aarch64_simd_valid_immediate.
26597         (aarch64_advsimd_valid_immediate_hs): New function.
26598         (aarch64_advsimd_valid_immediate): Likewise.
26599         (aarch64_simd_valid_immediate): Remove mode and inverse
26600         arguments.  Rewrite to use the above.  Use const_vec_duplicate_p
26601         to detect duplicated constants and use aarch64_float_const_zero_rtx_p
26602         and aarch64_float_const_representable_p on the result.
26603         (aarch64_output_simd_mov_immediate): Remove mode argument.
26604         Update call to aarch64_simd_valid_immediate and use of
26605         simd_immediate_info.
26606         (aarch64_output_scalar_simd_mov_immediate): Update call
26607         accordingly.
26609 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
26610             Alan Hayward  <alan.hayward@arm.com>
26611             David Sherwood  <david.sherwood@arm.com>
26613         * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
26614         (mode_nunits): Likewise CONST_MODE_NUNITS.
26615         * machmode.def (ADJUST_NUNITS): Document.
26616         * genmodes.c (mode_data::need_nunits_adj): New field.
26617         (blank_mode): Update accordingly.
26618         (adj_nunits): New variable.
26619         (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
26620         parameter.
26621         (emit_mode_size_inline): Set need_bytesize_adj for all modes
26622         listed in adj_nunits.
26623         (emit_mode_nunits_inline): Set need_nunits_adj for all modes
26624         listed in adj_nunits.  Don't emit case statements for such modes.
26625         (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
26626         and CONST_MODE_PRECISION.  Make CONST_MODE_SIZE expand to
26627         nothing if adj_nunits is nonnull.
26628         (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
26629         (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
26630         (emit_mode_fbit): Update use of print_maybe_const_decl.
26631         (emit_move_size): Likewise.  Treat the array as non-const
26632         if adj_nunits.
26633         (emit_mode_adjustments): Handle adj_nunits.
26635 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
26637         * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
26638         * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
26639         (VECTOR_MODES): Use it.
26640         (make_vector_modes): Take the prefix as an argument.
26642 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
26643             Alan Hayward  <alan.hayward@arm.com>
26644             David Sherwood  <david.sherwood@arm.com>
26646         * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
26647         * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
26648         for MODE_VECTOR_BOOL.
26649         * machmode.def (VECTOR_BOOL_MODE): Document.
26650         * genmodes.c (VECTOR_BOOL_MODE): New macro.
26651         (make_vector_bool_mode): New function.
26652         (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
26653         MODE_VECTOR_BOOL.
26654         * lto-streamer-in.c (lto_input_mode_table): Likewise.
26655         * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
26656         Likewise.
26657         * stor-layout.c (int_mode_for_mode): Likewise.
26658         * tree.c (build_vector_type_for_mode): Likewise.
26659         * varasm.c (output_constant_pool_2): Likewise.
26660         * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
26661         CONSTM1_RTX (BImode) are the same thing.  Initialize const_tiny_rtx
26662         for MODE_VECTOR_BOOL.
26663         * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
26664         of mode class checks.
26665         * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
26666         instead of a list of mode class checks.
26667         (expand_vector_scalar_condition): Likewise.
26668         (type_for_widest_vector_mode): Handle BImode as an inner mode.
26670 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
26671             Alan Hayward  <alan.hayward@arm.com>
26672             David Sherwood  <david.sherwood@arm.com>
26674         * machmode.h (mode_size): Change from unsigned short to
26675         poly_uint16_pod.
26676         (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
26677         (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
26678         or if measurement_type is not polynomial.
26679         (fixed_size_mode::includes_p): Check for constant-sized modes.
26680         * genmodes.c (emit_mode_size_inline): Make mode_size_inline
26681         return a poly_uint16 rather than an unsigned short.
26682         (emit_mode_size): Change the type of mode_size from unsigned short
26683         to poly_uint16_pod.  Use ZERO_COEFFS for the initializer.
26684         (emit_mode_adjustments): Cope with polynomial vector sizes.
26685         * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
26686         for GET_MODE_SIZE.
26687         * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
26688         for GET_MODE_SIZE.
26689         * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
26690         * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
26691         * caller-save.c (setup_save_areas): Likewise.
26692         (replace_reg_with_saved_mem): Likewise.
26693         * calls.c (emit_library_call_value_1): Likewise.
26694         * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
26695         * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
26696         (gen_lowpart_for_combine): Likewise.
26697         * convert.c (convert_to_integer_1): Likewise.
26698         * cse.c (equiv_constant, cse_insn): Likewise.
26699         * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
26700         (cselib_subst_to_values): Likewise.
26701         * dce.c (word_dce_process_block): Likewise.
26702         * df-problems.c (df_word_lr_mark_ref): Likewise.
26703         * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
26704         * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
26705         (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
26706         (rtl_for_decl_location): Likewise.
26707         * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
26708         * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
26709         * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
26710         (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
26711         (expand_expr_real_1): Likewise.
26712         * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
26713         (pad_below): Likewise.
26714         * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
26715         * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
26716         * ira.c (get_subreg_tracking_sizes): Likewise.
26717         * ira-build.c (ira_create_allocno_objects): Likewise.
26718         * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
26719         (ira_sort_regnos_for_alter_reg): Likewise.
26720         * ira-costs.c (record_operand_costs): Likewise.
26721         * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
26722         (resolve_simple_move): Likewise.
26723         * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
26724         (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
26725         (lra_constraints): Likewise.
26726         (CONST_POOL_OK_P): Reject variable-sized modes.
26727         * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
26728         (add_pseudo_to_slot, lra_spill): Likewise.
26729         * omp-low.c (omp_clause_aligned_alignment): Likewise.
26730         * optabs-query.c (get_best_extraction_insn): Likewise.
26731         * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
26732         * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
26733         (expand_mult_highpart, valid_multiword_target_p): Likewise.
26734         * recog.c (offsettable_address_addr_space_p): Likewise.
26735         * regcprop.c (maybe_mode_change): Likewise.
26736         * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
26737         * regrename.c (build_def_use): Likewise.
26738         * regstat.c (dump_reg_info): Likewise.
26739         * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
26740         (find_reloads, find_reloads_subreg_address): Likewise.
26741         * reload1.c (eliminate_regs_1): Likewise.
26742         * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
26743         * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
26744         (simplify_binary_operation_1, simplify_subreg): Likewise.
26745         * targhooks.c (default_function_arg_padding): Likewise.
26746         (default_hard_regno_nregs, default_class_max_nregs): Likewise.
26747         * tree-cfg.c (verify_gimple_assign_binary): Likewise.
26748         (verify_gimple_assign_ternary): Likewise.
26749         * tree-inline.c (estimate_move_cost): Likewise.
26750         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
26751         * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
26752         (get_address_cost_ainc): Likewise.
26753         * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
26754         (vect_supportable_dr_alignment): Likewise.
26755         * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
26756         (vectorizable_reduction): Likewise.
26757         * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
26758         (vectorizable_operation, vectorizable_load): Likewise.
26759         * tree.c (build_same_sized_truth_vector_type): Likewise.
26760         * valtrack.c (cleanup_auto_inc_dec): Likewise.
26761         * var-tracking.c (emit_note_insn_var_location): Likewise.
26762         * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
26763         (ADDR_VEC_ALIGN): Likewise.
26765 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
26766             Alan Hayward  <alan.hayward@arm.com>
26767             David Sherwood  <david.sherwood@arm.com>
26769         * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
26770         unsigned short.
26771         (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
26772         or if measurement_type is polynomial.
26773         * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
26774         * combine.c (make_extraction): Likewise.
26775         * dse.c (find_shift_sequence): Likewise.
26776         * dwarf2out.c (mem_loc_descriptor): Likewise.
26777         * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
26778         (extract_bit_field, extract_low_bits): Likewise.
26779         * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
26780         (optimize_bitfield_assignment_op, expand_assignment): Likewise.
26781         (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
26782         * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
26783         * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
26784         * reload.c (find_reloads): Likewise.
26785         * reload1.c (alter_reg): Likewise.
26786         * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
26787         * targhooks.c (default_secondary_memory_needed_mode): Likewise.
26788         * tree-if-conv.c (predicate_mem_writes): Likewise.
26789         * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
26790         * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
26791         * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
26792         * valtrack.c (dead_debug_insert_temp): Likewise.
26793         * varasm.c (mergeable_constant_section): Likewise.
26794         * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
26796 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
26797             Alan Hayward  <alan.hayward@arm.com>
26798             David Sherwood  <david.sherwood@arm.com>
26800         * expr.c (expand_assignment): Cope with polynomial mode sizes
26801         when assigning to a CONCAT.
26803 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
26804             Alan Hayward  <alan.hayward@arm.com>
26805             David Sherwood  <david.sherwood@arm.com>
26807         * machmode.h (mode_precision): Change from unsigned short to
26808         poly_uint16_pod.
26809         (mode_to_precision): Return a poly_uint16 rather than an unsigned
26810         short.
26811         (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
26812         or if measurement_type is not polynomial.
26813         (HWI_COMPUTABLE_MODE_P): Turn into a function.  Optimize the case
26814         in which the mode is already known to be a scalar_int_mode.
26815         * genmodes.c (emit_mode_precision): Change the type of mode_precision
26816         from unsigned short to poly_uint16_pod.  Use ZERO_COEFFS for the
26817         initializer.
26818         * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
26819         for GET_MODE_PRECISION.
26820         * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
26821         for GET_MODE_PRECISION.
26822         * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
26823         as polynomial.
26824         (try_combine, find_split_point, combine_simplify_rtx): Likewise.
26825         (expand_field_assignment, make_extraction): Likewise.
26826         (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
26827         (get_last_value): Likewise.
26828         * convert.c (convert_to_integer_1): Likewise.
26829         * cse.c (cse_insn): Likewise.
26830         * expr.c (expand_expr_real_1): Likewise.
26831         * lra-constraints.c (simplify_operand_subreg): Likewise.
26832         * optabs-query.c (can_atomic_load_p): Likewise.
26833         * optabs.c (expand_atomic_load): Likewise.
26834         (expand_atomic_store): Likewise.
26835         * ree.c (combine_reaching_defs): Likewise.
26836         * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
26837         * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
26838         * tree.h (type_has_mode_precision_p): Likewise.
26839         * ubsan.c (instrument_si_overflow): Likewise.
26841 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
26842             Alan Hayward  <alan.hayward@arm.com>
26843             David Sherwood  <david.sherwood@arm.com>
26845         * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
26846         polynomial numbers of units.
26847         (SET_TYPE_VECTOR_SUBPARTS): Likewise.
26848         (valid_vector_subparts_p): New function.
26849         (build_vector_type): Remove temporary shim and take the number
26850         of units as a poly_uint64 rather than an int.
26851         (build_opaque_vector_type): Take the number of units as a
26852         poly_uint64 rather than an int.
26853         * tree.c (build_vector_from_ctor): Handle polynomial
26854         TYPE_VECTOR_SUBPARTS.
26855         (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
26856         (uniform_vector_p, vector_type_mode, build_vector): Likewise.
26857         (build_vector_from_val): If the number of units is variable,
26858         use build_vec_duplicate_cst for constant operands and
26859         VEC_DUPLICATE_EXPR otherwise.
26860         (make_vector_type): Remove temporary is_constant ().
26861         (build_vector_type, build_opaque_vector_type): Take the number of
26862         units as a poly_uint64 rather than an int.
26863         (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
26864         VECTOR_CST_NELTS.
26865         * cfgexpand.c (expand_debug_expr): Likewise.
26866         * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
26867         (store_constructor, expand_expr_real_1): Likewise.
26868         (const_scalar_mask_from_tree): Likewise.
26869         * fold-const-call.c (fold_const_reduction): Likewise.
26870         * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
26871         (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
26872         (native_encode_vector, vec_cst_ctor_to_array): Likewise.
26873         (fold_relational_const): Likewise.
26874         (native_interpret_vector): Likewise.  Change the size from an
26875         int to an unsigned int.
26876         * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
26877         TYPE_VECTOR_SUBPARTS.
26878         (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
26879         (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
26880         duplicating a non-constant operand into a variable-length vector.
26881         * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
26882         TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
26883         * ipa-icf.c (sem_variable::equals): Likewise.
26884         * match.pd: Likewise.
26885         * omp-simd-clone.c (simd_clone_subparts): Likewise.
26886         * print-tree.c (print_node): Likewise.
26887         * stor-layout.c (layout_type): Likewise.
26888         * targhooks.c (default_builtin_vectorization_cost): Likewise.
26889         * tree-cfg.c (verify_gimple_comparison): Likewise.
26890         (verify_gimple_assign_binary): Likewise.
26891         (verify_gimple_assign_ternary): Likewise.
26892         (verify_gimple_assign_single): Likewise.
26893         * tree-pretty-print.c (dump_generic_node): Likewise.
26894         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
26895         (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
26896         * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
26897         (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
26898         (vect_shift_permute_load_chain): Likewise.
26899         * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
26900         (expand_vector_condition, optimize_vector_constructor): Likewise.
26901         (lower_vec_perm, get_compute_type): Likewise.
26902         * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
26903         (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
26904         * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
26905         (vect_recog_mask_conversion_pattern): Likewise.
26906         * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
26907         (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
26908         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
26909         (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
26910         (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
26911         (vectorizable_shift, vectorizable_operation, vectorizable_store)
26912         (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
26913         (supportable_widening_operation): Likewise.
26914         (supportable_narrowing_operation): Likewise.
26915         * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
26916         Likewise.
26917         * varasm.c (output_constant): Likewise.
26919 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
26920             Alan Hayward  <alan.hayward@arm.com>
26921             David Sherwood  <david.sherwood@arm.com>
26923         * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
26924         so that both the length == 3 and length != 3 cases set up their
26925         own permute vectors.  Add comments explaining why we know the
26926         number of elements is constant.
26927         (vect_permute_load_chain): Likewise.
26929 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
26930             Alan Hayward  <alan.hayward@arm.com>
26931             David Sherwood  <david.sherwood@arm.com>
26933         * machmode.h (mode_nunits): Change from unsigned char to
26934         poly_uint16_pod.
26935         (ONLY_FIXED_SIZE_MODES): New macro.
26936         (pod_mode::measurement_type, scalar_int_mode::measurement_type)
26937         (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
26938         (complex_mode::measurement_type, fixed_size_mode::measurement_type):
26939         New typedefs.
26940         (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
26941         (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
26942         or if measurement_type is not polynomial.
26943         * genmodes.c (ZERO_COEFFS): New macro.
26944         (emit_mode_nunits_inline): Make mode_nunits_inline return a
26945         poly_uint16.
26946         (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
26947         Use ZERO_COEFFS when emitting initializers.
26948         * data-streamer.h (bp_pack_poly_value): New function.
26949         (bp_unpack_poly_value): Likewise.
26950         * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
26951         for GET_MODE_NUNITS.
26952         * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
26953         for GET_MODE_NUNITS.
26954         * tree.c (make_vector_type): Remove temporary shim and make
26955         the real function take the number of units as a poly_uint64
26956         rather than an int.
26957         (build_vector_type_for_mode): Handle polynomial nunits.
26958         * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
26959         * emit-rtl.c (const_vec_series_p_1): Likewise.
26960         (gen_rtx_CONST_VECTOR): Likewise.
26961         * fold-const.c (test_vec_duplicate_folding): Likewise.
26962         * genrecog.c (validate_pattern): Likewise.
26963         * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
26964         * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
26965         * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
26966         (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
26967         (expand_vec_cond_expr, expand_mult_highpart): Likewise.
26968         * rtlanal.c (subreg_get_info): Likewise.
26969         * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
26970         (vect_grouped_load_supported): Likewise.
26971         * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
26972         * tree-vect-loop.c (have_whole_vector_shift): Likewise.
26973         * simplify-rtx.c (simplify_unary_operation_1): Likewise.
26974         (simplify_const_unary_operation, simplify_binary_operation_1)
26975         (simplify_const_binary_operation, simplify_ternary_operation)
26976         (test_vector_ops_duplicate, test_vector_ops): Likewise.
26977         (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
26978         instead of CONST_VECTOR_NUNITS.
26979         * varasm.c (output_constant_pool_2): Likewise.
26980         * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
26981         explicit-encoded elements in the XVEC for variable-length vectors.
26983 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
26985         * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
26987 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
26988             Alan Hayward  <alan.hayward@arm.com>
26989             David Sherwood  <david.sherwood@arm.com>
26991         * coretypes.h (fixed_size_mode): Declare.
26992         (fixed_size_mode_pod): New typedef.
26993         * builtins.h (target_builtins::x_apply_args_mode)
26994         (target_builtins::x_apply_result_mode): Change type to
26995         fixed_size_mode_pod.
26996         * builtins.c (apply_args_size, apply_result_size, result_vector)
26997         (expand_builtin_apply_args_1, expand_builtin_apply)
26998         (expand_builtin_return): Update accordingly.
27000 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27002         * cse.c (hash_rtx_cb): Hash only the encoded elements.
27003         * cselib.c (cselib_hash_rtx): Likewise.
27004         * expmed.c (make_tree): Build VECTOR_CSTs directly from the
27005         CONST_VECTOR encoding.
27007 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
27008             Jeff Law  <law@redhat.com>
27010         PR target/83641
27011         * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
27012         noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
27013         only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
27014         and add REG_CFA_ADJUST_CFA notes in that case to both insns.
27016         PR target/83641
27017         * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
27018         explicitly probe *sp in a noreturn function if there were any callee
27019         register saves or frame pointer is needed.
27021 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
27023         PR debug/83621
27024         * cfgexpand.c (expand_debug_expr): Return NULL if mode is
27025         BLKmode for ternary, binary or unary expressions.
27027         PR debug/83645
27028         * var-tracking.c (delete_vta_debug_insn): New inline function.
27029         (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
27030         insns from get_insns () to NULL instead of each bb separately.
27031         Use delete_vta_debug_insn.  No longer static.
27032         (vt_debug_insns_local, variable_tracking_main_1): Adjust
27033         delete_vta_debug_insns callers.
27034         * rtl.h (delete_vta_debug_insns): Declare.
27035         * final.c (rest_of_handle_final): Call delete_vta_debug_insns
27036         instead of variable_tracking_main.
27038 2018-01-03  Martin Sebor  <msebor@redhat.com>
27040         PR tree-optimization/83603
27041         * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
27042         arguments past the endof the argument list in functions declared
27043         without a prototype.
27044         * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
27045         Avoid checking when arguments are null.
27047 2018-01-03  Martin Sebor  <msebor@redhat.com>
27049         PR c/83559
27050         * doc/extend.texi (attribute const): Fix a typo.
27051         * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
27052         issuing -Wsuggest-attribute for void functions.
27054 2018-01-03  Martin Sebor  <msebor@redhat.com>
27056         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
27057         offset_int::from instead of wide_int::to_shwi.
27058         (maybe_diag_overlap): Remove assertion.
27059         Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
27060         * gimple-ssa-sprintf.c (format_directive): Same.
27061         (parse_directive): Same.
27062         (sprintf_dom_walker::compute_format_length): Same.
27063         (try_substitute_return_value): Same.
27065 2018-01-03  Jeff Law  <law@redhat.com>
27067         PR middle-end/83654
27068         * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
27069         non-constant residual for zero at runtime and avoid probing in
27070         that case.  Reorganize code for trailing problem to mirror handling
27071         of the residual.
27073 2018-01-03  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
27075         PR tree-optimization/83501
27076         * tree-ssa-strlen.c (get_string_cst): New.
27077         (handle_char_store): Call get_string_cst.
27079 2018-01-03  Martin Liska  <mliska@suse.cz>
27081         PR tree-optimization/83593
27082         * tree-ssa-strlen.c: Include tree-cfg.h.
27083         (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
27084         (strlen_dom_walker): Add new member variable m_cleanup_cfg.
27085         (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
27086         to false.
27087         (strlen_dom_walker::before_dom_children): Call
27088         gimple_purge_dead_eh_edges. Dump tranformation with details
27089         dump flags.
27090         (strlen_dom_walker::before_dom_children): Update call by adding
27091         new argument cleanup_eh.
27092         (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
27094 2018-01-03  Martin Liska  <mliska@suse.cz>
27096         PR ipa/83549
27097         * cif-code.def (VARIADIC_THUNK): New enum value.
27098         * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
27099         thunks.
27101 2018-01-03  Jan Beulich  <jbeulich@suse.com>
27103         * sse.md (mov<mode>_internal): Tighten condition for when to use
27104         vmovdqu<ssescalarsize> for TI and OI modes.
27106 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
27108         Update copyright years.
27110 2018-01-03  Martin Liska  <mliska@suse.cz>
27112         PR ipa/83594
27113         * ipa-visibility.c (function_and_variable_visibility): Skip
27114         functions with noipa attribure.
27116 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
27118         * gcc.c (process_command): Update copyright notice dates.
27119         * gcov-dump.c (print_version): Ditto.
27120         * gcov.c (print_version): Ditto.
27121         * gcov-tool.c (print_version): Ditto.
27122         * gengtype.c (create_file): Ditto.
27123         * doc/cpp.texi: Bump @copying's copyright year.
27124         * doc/cppinternals.texi: Ditto.
27125         * doc/gcc.texi: Ditto.
27126         * doc/gccint.texi: Ditto.
27127         * doc/gcov.texi: Ditto.
27128         * doc/install.texi: Ditto.
27129         * doc/invoke.texi: Ditto.
27131 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27133         * vector-builder.h (vector_builder::m_full_nelts): Change from
27134         unsigned int to poly_uint64.
27135         (vector_builder::full_nelts): Update prototype accordingly.
27136         (vector_builder::new_vector): Likewise.
27137         (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
27138         (vector_builder::operator ==): Likewise.
27139         (vector_builder::finalize): Likewise.
27140         * int-vector-builder.h (int_vector_builder::int_vector_builder):
27141         Take the number of elements as a poly_uint64 rather than an
27142         unsigned int.
27143         * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
27144         from unsigned int to poly_uint64.
27145         (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
27146         (vec_perm_indices::new_vector): Likewise.
27147         (vec_perm_indices::length): Likewise.
27148         (vec_perm_indices::nelts_per_input): Likewise.
27149         (vec_perm_indices::input_nelts): Likewise.
27150         * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
27151         number of elements per input as a poly_uint64 rather than an
27152         unsigned int.  Use the original encoding for variable-length
27153         vectors, rather than clamping each individual element.
27154         For the second and subsequent elements in each pattern,
27155         clamp the step and base before clamping their sum.
27156         (vec_perm_indices::series_p): Handle polynomial element counts.
27157         (vec_perm_indices::all_in_range_p): Likewise.
27158         (vec_perm_indices_to_tree): Likewise.
27159         (vec_perm_indices_to_rtx): Likewise.
27160         * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
27161         * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
27162         (tree_vector_builder::new_binary_operation): Handle polynomial
27163         element counts.  Return false if we need to know the number
27164         of elements at compile time.
27165         * fold-const.c (fold_vec_perm): Punt if the number of elements
27166         isn't known at compile time.
27168 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27170         * vec-perm-indices.h (vec_perm_builder): Change element type
27171         from HOST_WIDE_INT to poly_int64.
27172         (vec_perm_indices::element_type): Update accordingly.
27173         (vec_perm_indices::clamp): Handle polynomial element_types.
27174         * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
27175         (vec_perm_indices::all_in_range_p): Likewise.
27176         (tree_to_vec_perm_builder): Check for poly_int64 trees rather
27177         than shwi trees.
27178         * vector-builder.h (vector_builder::stepped_sequence_p): Handle
27179         polynomial vec_perm_indices element types.
27180         * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
27181         * fold-const.c (fold_vec_perm): Likewise.
27182         * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
27183         * tree-vect-generic.c (lower_vec_perm): Likewise.
27184         * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
27185         * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
27186         element type to HOST_WIDE_INT.
27188 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27189             Alan Hayward  <alan.hayward@arm.com>
27190             David Sherwood  <david.sherwood@arm.com>
27192         * alias.c (addr_side_effect_eval): Take the size as a poly_int64
27193         rather than an int.  Use plus_constant.
27194         (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
27195         Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
27197 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27198             Alan Hayward  <alan.hayward@arm.com>
27199             David Sherwood  <david.sherwood@arm.com>
27201         * calls.c (emit_call_1, expand_call): Change struct_value_size from
27202         a HOST_WIDE_INT to a poly_int64.
27204 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27205             Alan Hayward  <alan.hayward@arm.com>
27206             David Sherwood  <david.sherwood@arm.com>
27208         * calls.c (load_register_parameters): Cope with polynomial
27209         mode sizes.  Require a constant size for BLKmode parameters
27210         that aren't described by a PARALLEL.  If BLOCK_REG_PADDING
27211         forces a parameter to be padded at the lsb end in order to
27212         fill a complete number of words, require the parameter size
27213         to be ordered wrt UNITS_PER_WORD.
27215 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27216             Alan Hayward  <alan.hayward@arm.com>
27217             David Sherwood  <david.sherwood@arm.com>
27219         * reload1.c (spill_stack_slot_width): Change element type
27220         from unsigned int to poly_uint64_pod.
27221         (alter_reg): Treat mode sizes as polynomial.
27223 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27224             Alan Hayward  <alan.hayward@arm.com>
27225             David Sherwood  <david.sherwood@arm.com>
27227         * reload.c (complex_word_subreg_p): New function.
27228         (reload_inner_reg_of_subreg, push_reload): Use it.
27230 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27231             Alan Hayward  <alan.hayward@arm.com>
27232             David Sherwood  <david.sherwood@arm.com>
27234         * lra-constraints.c (process_alt_operands): Reject matched
27235         operands whose sizes aren't ordered.
27236         (match_reload): Refer to this check here.
27238 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27239             Alan Hayward  <alan.hayward@arm.com>
27240             David Sherwood  <david.sherwood@arm.com>
27242         * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
27243         that the mode size is in the set {1, 2, 4, 8, 16}.
27245 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27246             Alan Hayward  <alan.hayward@arm.com>
27247             David Sherwood  <david.sherwood@arm.com>
27249         * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
27250         Use plus_constant instead of gen_rtx_PLUS.
27252 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27253             Alan Hayward  <alan.hayward@arm.com>
27254             David Sherwood  <david.sherwood@arm.com>
27256         * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
27257         * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
27258         * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
27259         * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
27260         * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
27261         * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
27262         * config/i386/i386-protos.h (ix86_push_rounding): Declare.
27263         * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
27264         * config/i386/i386.c (ix86_push_rounding): ...this new function.
27265         * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
27266         a poly_int64.
27267         * config/m32c/m32c.c (m32c_push_rounding): Likewise.
27268         * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
27269         * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
27270         * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
27271         * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
27272         * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
27273         * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
27274         * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
27275         * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
27276         * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
27277         function.
27278         * expr.c (emit_move_resolve_push): Treat the input and result
27279         of PUSH_ROUNDING as a poly_int64.
27280         (emit_move_complex_push, emit_single_push_insn_1): Likewise.
27281         (emit_push_insn): Likewise.
27282         * lra-eliminations.c (mark_not_eliminable): Likewise.
27283         * recog.c (push_operand): Likewise.
27284         * reload1.c (elimination_effects): Likewise.
27285         * rtlanal.c (nonzero_bits1): Likewise.
27286         * calls.c (store_one_arg): Likewise.  Require the padding to be
27287         known at compile time.
27289 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27290             Alan Hayward  <alan.hayward@arm.com>
27291             David Sherwood  <david.sherwood@arm.com>
27293         * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
27294         Use plus_constant instead of gen_rtx_PLUS.
27296 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27297             Alan Hayward  <alan.hayward@arm.com>
27298             David Sherwood  <david.sherwood@arm.com>
27300         * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
27301         rather than an int.
27303 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27304             Alan Hayward  <alan.hayward@arm.com>
27305             David Sherwood  <david.sherwood@arm.com>
27307         * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
27308         instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
27309         via stack temporaries.  Treat the mode size as polynomial too.
27311 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27312             Alan Hayward  <alan.hayward@arm.com>
27313             David Sherwood  <david.sherwood@arm.com>
27315         * expr.c (expand_expr_real_2): When handling conversions involving
27316         unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
27317         multiplying int_size_in_bytes by BITS_PER_UNIT.  Treat GET_MODE_BISIZE
27318         as a poly_uint64 too.
27320 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27321             Alan Hayward  <alan.hayward@arm.com>
27322             David Sherwood  <david.sherwood@arm.com>
27324         * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
27326 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27327             Alan Hayward  <alan.hayward@arm.com>
27328             David Sherwood  <david.sherwood@arm.com>
27330         * combine.c (can_change_dest_mode): Handle polynomial
27331         REGMODE_NATURAL_SIZE.
27332         * expmed.c (store_bit_field_1): Likewise.
27333         * expr.c (store_constructor): Likewise.
27334         * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
27335         and polynomial REGMODE_NATURAL_SIZE.
27336         (gen_lowpart_common): Likewise.
27337         * reginfo.c (record_subregs_of_mode): Likewise.
27338         * rtlanal.c (read_modify_subreg_p): Likewise.
27340 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27341             Alan Hayward  <alan.hayward@arm.com>
27342             David Sherwood  <david.sherwood@arm.com>
27344         * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
27345         numbers of elements.
27347 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27348             Alan Hayward  <alan.hayward@arm.com>
27349             David Sherwood  <david.sherwood@arm.com>
27351         * match.pd: Cope with polynomial numbers of vector elements.
27353 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27354             Alan Hayward  <alan.hayward@arm.com>
27355             David Sherwood  <david.sherwood@arm.com>
27357         * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
27358         in a POINTER_PLUS_EXPR.
27360 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27361             Alan Hayward  <alan.hayward@arm.com>
27362             David Sherwood  <david.sherwood@arm.com>
27364         * omp-simd-clone.c (simd_clone_subparts): New function.
27365         (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
27366         (ipa_simd_modify_function_body): Likewise.
27368 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27369             Alan Hayward  <alan.hayward@arm.com>
27370             David Sherwood  <david.sherwood@arm.com>
27372         * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
27373         (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
27374         (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
27375         (expand_vector_condition, vector_element): Likewise.
27376         (subparts_gt): New function.
27377         (get_compute_type): Use subparts_gt.
27378         (count_type_subparts): Delete.
27379         (expand_vector_operations_1): Use subparts_gt instead of
27380         count_type_subparts.
27382 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27383             Alan Hayward  <alan.hayward@arm.com>
27384             David Sherwood  <david.sherwood@arm.com>
27386         * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
27387         (vect_compile_time_alias): ...this new function.  Do the calculation
27388         on poly_ints rather than trees.
27389         (vect_prune_runtime_alias_test_list): Update call accordingly.
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-slp.c (vect_build_slp_tree_1): Handle polynomial
27396         numbers of units.
27397         (vect_schedule_slp_instance): Likewise.
27399 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27400             Alan Hayward  <alan.hayward@arm.com>
27401             David Sherwood  <david.sherwood@arm.com>
27403         * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
27404         constant and extern definitions for variable-length vectors.
27405         (vect_get_constant_vectors): Note that the number of units
27406         is known to be constant.
27408 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27409             Alan Hayward  <alan.hayward@arm.com>
27410             David Sherwood  <david.sherwood@arm.com>
27412         * tree-vect-stmts.c (vectorizable_conversion): Treat the number
27413         of units as polynomial.  Choose between WIDE and NARROW based
27414         on multiple_p.
27416 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27417             Alan Hayward  <alan.hayward@arm.com>
27418             David Sherwood  <david.sherwood@arm.com>
27420         * tree-vect-stmts.c (simd_clone_subparts): New function.
27421         (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
27423 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27424             Alan Hayward  <alan.hayward@arm.com>
27425             David Sherwood  <david.sherwood@arm.com>
27427         * tree-vect-stmts.c (vectorizable_call): Treat the number of
27428         vectors as polynomial.  Use build_index_vector for
27429         IFN_GOMP_SIMD_LANE.
27431 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27432             Alan Hayward  <alan.hayward@arm.com>
27433             David Sherwood  <david.sherwood@arm.com>
27435         * tree-vect-stmts.c (get_load_store_type): Treat the number of
27436         units as polynomial.  Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
27437         for variable-length vectors.
27438         (vectorizable_mask_load_store): Treat the number of units as
27439         polynomial, asserting that it is constant if the condition has
27440         already been enforced.
27441         (vectorizable_store, vectorizable_load): Likewise.
27443 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27444             Alan Hayward  <alan.hayward@arm.com>
27445             David Sherwood  <david.sherwood@arm.com>
27447         * tree-vect-loop.c (vectorizable_live_operation): Treat the number
27448         of units as polynomial.  Punt if we can't tell at compile time
27449         which vector contains the final result.
27451 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27452             Alan Hayward  <alan.hayward@arm.com>
27453             David Sherwood  <david.sherwood@arm.com>
27455         * tree-vect-loop.c (vectorizable_induction): Treat the number
27456         of units as polynomial.  Punt on SLP inductions.  Use an integer
27457         VEC_SERIES_EXPR for variable-length integer reductions.  Use a
27458         cast of such a series for variable-length floating-point
27459         reductions.
27461 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27462             Alan Hayward  <alan.hayward@arm.com>
27463             David Sherwood  <david.sherwood@arm.com>
27465         * tree.h (build_index_vector): Declare.
27466         * tree.c (build_index_vector): New function.
27467         * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
27468         of units as polynomial, forcibly converting it to a constant if
27469         vectorizable_reduction has already enforced the condition.
27470         (vect_create_epilog_for_reduction): Likewise.  Use build_index_vector
27471         to create a {1,2,3,...} vector.
27472         (vectorizable_reduction): Treat the number of units as polynomial.
27473         Choose vectype_in based on the largest scalar element size rather
27474         than the smallest number of units.  Enforce the restrictions
27475         relied on above.
27477 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27478             Alan Hayward  <alan.hayward@arm.com>
27479             David Sherwood  <david.sherwood@arm.com>
27481         * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
27482         number of units as polynomial.
27484 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27485             Alan Hayward  <alan.hayward@arm.com>
27486             David Sherwood  <david.sherwood@arm.com>
27488         * target.h (vector_sizes, auto_vector_sizes): New typedefs.
27489         * target.def (autovectorize_vector_sizes): Return the vector sizes
27490         by pointer, using vector_sizes rather than a bitmask.
27491         * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
27492         * targhooks.c (default_autovectorize_vector_sizes): Likewise.
27493         * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
27494         Likewise.
27495         * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
27496         * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
27497         * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
27498         * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
27499         * omp-general.c (omp_max_vf): Likewise.
27500         * omp-low.c (omp_clause_aligned_alignment): Likewise.
27501         * optabs-query.c (can_vec_mask_load_store_p): Likewise.
27502         * tree-vect-loop.c (vect_analyze_loop): Likewise.
27503         * tree-vect-slp.c (vect_slp_bb): Likewise.
27504         * doc/tm.texi: Regenerate.
27505         * tree-vectorizer.h (current_vector_size): Change from an unsigned int
27506         to a poly_uint64.
27507         * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
27508         the vector size as a poly_uint64 rather than an unsigned int.
27509         (current_vector_size): Change from an unsigned int to a poly_uint64.
27510         (get_vectype_for_scalar_type): Update accordingly.
27511         * tree.h (build_truth_vector_type): Take the size and number of
27512         units as a poly_uint64 rather than an unsigned int.
27513         (build_vector_type): Add a temporary overload that takes
27514         the number of units as a poly_uint64 rather than an unsigned int.
27515         * tree.c (make_vector_type): Likewise.
27516         (build_truth_vector_type): Take the number of units as a poly_uint64
27517         rather than an unsigned int.
27519 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27520             Alan Hayward  <alan.hayward@arm.com>
27521             David Sherwood  <david.sherwood@arm.com>
27523         * target.def (get_mask_mode): Take the number of units and length
27524         as poly_uint64s rather than unsigned ints.
27525         * targhooks.h (default_get_mask_mode): Update accordingly.
27526         * targhooks.c (default_get_mask_mode): Likewise.
27527         * config/i386/i386.c (ix86_get_mask_mode): Likewise.
27528         * doc/tm.texi: Regenerate.
27530 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27531             Alan Hayward  <alan.hayward@arm.com>
27532             David Sherwood  <david.sherwood@arm.com>
27534         * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
27535         * omp-general.c (omp_max_vf): Likewise.
27536         * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
27537         (expand_omp_simd): Handle polynomial safelen.
27538         * omp-low.c (omplow_simd_context): Add a default constructor.
27539         (omplow_simd_context::max_vf): Change from int to poly_uint64.
27540         (lower_rec_simd_input_clauses): Update accordingly.
27541         (lower_rec_input_clauses): Likewise.
27543 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27544             Alan Hayward  <alan.hayward@arm.com>
27545             David Sherwood  <david.sherwood@arm.com>
27547         * tree-vectorizer.h (vect_nunits_for_cost): New function.
27548         * tree-vect-loop.c (vect_model_reduction_cost): Use it.
27549         * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
27550         (vect_analyze_slp_cost): Likewise.
27551         * tree-vect-stmts.c (vect_model_store_cost): Likewise.
27552         (vect_model_load_cost): Likewise.
27554 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27555             Alan Hayward  <alan.hayward@arm.com>
27556             David Sherwood  <david.sherwood@arm.com>
27558         * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
27559         (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
27560         from an unsigned int * to a poly_uint64_pod *.
27561         (calculate_unrolling_factor): New function.
27562         (vect_analyze_slp_instance): Use it.  Track polynomial max_nunits.
27564 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27565             Alan Hayward  <alan.hayward@arm.com>
27566             David Sherwood  <david.sherwood@arm.com>
27568         * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
27569         from an unsigned int to a poly_uint64.
27570         (_loop_vec_info::slp_unrolling_factor): Likewise.
27571         (_loop_vec_info::vectorization_factor): Change from an int
27572         to a poly_uint64.
27573         (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
27574         (vect_get_num_vectors): New function.
27575         (vect_update_max_nunits, vect_vf_for_cost): Likewise.
27576         (vect_get_num_copies): Use vect_get_num_vectors.
27577         (vect_analyze_data_ref_dependences): Change max_vf from an int *
27578         to an unsigned int *.
27579         (vect_analyze_data_refs): Change min_vf from an int * to a
27580         poly_uint64 *.
27581         (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
27582         than an unsigned HOST_WIDE_INT.
27583         * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
27584         (vect_analyze_data_ref_dependence): Change max_vf from an int *
27585         to an unsigned int *.
27586         (vect_analyze_data_ref_dependences): Likewise.
27587         (vect_compute_data_ref_alignment): Handle polynomial vf.
27588         (vect_enhance_data_refs_alignment): Likewise.
27589         (vect_prune_runtime_alias_test_list): Likewise.
27590         (vect_shift_permute_load_chain): Likewise.
27591         (vect_supportable_dr_alignment): Likewise.
27592         (dependence_distance_ge_vf): Take the vectorization factor as a
27593         poly_uint64 rather than an unsigned HOST_WIDE_INT.
27594         (vect_analyze_data_refs): Change min_vf from an int * to a
27595         poly_uint64 *.
27596         * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
27597         vfm1 as a poly_uint64 rather than an int.  Make the same change
27598         for the returned bound_scalar.
27599         (vect_gen_vector_loop_niters): Handle polynomial vf.
27600         (vect_do_peeling): Likewise.  Update call to
27601         vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
27602         (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
27603         be constant.
27604         * tree-vect-loop.c (vect_determine_vectorization_factor)
27605         (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
27606         (vect_get_known_peeling_cost): Likewise.
27607         (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
27608         (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
27609         (vect_transform_loop): Likewise.  Use the lowest possible VF when
27610         updating the upper bounds of the loop.
27611         (vect_min_worthwhile_factor): Make static.  Return an unsigned int
27612         rather than an int.
27613         * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
27614         polynomial unroll factors.
27615         (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
27616         (vect_make_slp_decision): Likewise.
27617         (vect_supported_load_permutation_p): Likewise, and polynomial
27618         vf too.
27619         (vect_analyze_slp_cost): Handle polynomial vf.
27620         (vect_slp_analyze_node_operations): Likewise.
27621         (vect_slp_analyze_bb_1): Likewise.
27622         (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
27623         than an unsigned HOST_WIDE_INT.
27624         * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
27625         (vectorizable_load): Handle polynomial vf.
27626         * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
27627         a poly_uint64.
27628         (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
27630 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27631             Alan Hayward  <alan.hayward@arm.com>
27632             David Sherwood  <david.sherwood@arm.com>
27634         * match.pd: Handle bit operations involving three constants
27635         and try to fold one pair.
27637 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27639         * tree-vect-loop-manip.c: Include gimple-fold.h.
27640         (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
27641         niters_maybe_zero parameters.  Handle other cases besides a step of 1.
27642         (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
27643         Add a path that uses a step of VF instead of 1, but disable it
27644         for now.
27645         (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
27646         and niters_no_overflow parameters.  Update calls to
27647         slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
27648         Create a new SSA name if the latter choses to use a ste other
27649         than zero, and return it via niters_vector_mult_vf_var.
27650         * tree-vect-loop.c (vect_transform_loop): Update calls to
27651         vect_do_peeling, vect_gen_vector_loop_niters and
27652         slpeel_make_loop_iterate_ntimes.
27653         * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
27654         (vect_gen_vector_loop_niters): Update declarations after above changes.
27656 2018-01-02  Michael Meissner  <meissner@linux.vnet.ibm.com>
27658         * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
27659         128-bit round to integer instructions.
27660         (ceil<mode>2): Likewise.
27661         (btrunc<mode>2): Likewise.
27662         (round<mode>2): Likewise.
27664 2018-01-02  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
27666         * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
27667         unaligned VSX load/store on P8/P9.
27668         (expand_block_clear): Allow the use of unaligned VSX
27669         load/store on P8/P9.
27671 2018-01-02  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
27673         * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
27674         New function.
27675         (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
27676         swap associated with both a load and a store.
27678 2018-01-02  Andrew Waterman  <andrew@sifive.com>
27680         * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
27681         * config/riscv/riscv.md (clear_cache): Use it.
27683 2018-01-02  Artyom Skrobov  <tyomitch@gmail.com>
27685         * web.c: Remove out-of-date comment.
27687 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
27689         * expr.c (fixup_args_size_notes): Check that any existing
27690         REG_ARGS_SIZE notes are correct, and don't try to re-add them.
27691         (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
27692         (emit_single_push_insn): ...here.
27694 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
27696         * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
27697         (const_vector_encoded_nelts): New function.
27698         (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
27699         (const_vector_int_elt, const_vector_elt): Declare.
27700         * emit-rtl.c (const_vector_int_elt_1): New function.
27701         (const_vector_elt): Likewise.
27702         * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
27703         of CONST_VECTOR_ELT.
27705 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
27707         * expr.c: Include rtx-vector-builder.h.
27708         (const_vector_mask_from_tree): Use rtx_vector_builder and operate
27709         directly on the tree encoding.
27710         (const_vector_from_tree): Likewise.
27711         * optabs.c: Include rtx-vector-builder.h.
27712         (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
27713         sequence of "u" values.
27714         * vec-perm-indices.c: Include rtx-vector-builder.h.
27715         (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
27716         directly on the vec_perm_indices encoding.
27718 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
27720         * doc/rtl.texi (const_vector): Describe new encoding scheme.
27721         * Makefile.in (OBJS): Add rtx-vector-builder.o.
27722         * rtx-vector-builder.h: New file.
27723         * rtx-vector-builder.c: Likewise.
27724         * rtl.h (rtx_def::u2): Add a const_vector field.
27725         (CONST_VECTOR_NPATTERNS): New macro.
27726         (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
27727         (CONST_VECTOR_DUPLICATE_P): Likewise.
27728         (CONST_VECTOR_STEPPED_P): Likewise.
27729         (CONST_VECTOR_ENCODED_ELT): Likewise.
27730         (const_vec_duplicate_p): Check for a duplicated vector encoding.
27731         (unwrap_const_vec_duplicate): Likewise.
27732         (const_vec_series_p): Check for a non-duplicated vector encoding.
27733         Say that the function only returns true for integer vectors.
27734         * emit-rtl.c: Include rtx-vector-builder.h.
27735         (gen_const_vec_duplicate_1): Delete.
27736         (gen_const_vector): Call gen_const_vec_duplicate instead of
27737         gen_const_vec_duplicate_1.
27738         (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
27739         (gen_const_vec_duplicate): Use rtx_vector_builder.
27740         (gen_const_vec_series): Likewise.
27741         (gen_rtx_CONST_VECTOR): Likewise.
27742         * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
27743         (swap_const_vector_halves): Take an rtx pointer rather than rtx.
27744         Build a new vector rather than modifying a CONST_VECTOR in-place.
27745         (handle_special_swappables): Update call accordingly.
27746         * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
27747         (swap_const_vector_halves): Take an rtx pointer rather than rtx.
27748         Build a new vector rather than modifying a CONST_VECTOR in-place.
27749         (handle_special_swappables): Update call accordingly.
27751 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
27753         * simplify-rtx.c (simplify_const_binary_operation): Use
27754         CONST_VECTOR_ELT instead of XVECEXP.
27756 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
27758         * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
27759         the selector elements to be different from the data elements
27760         if the selector is a VECTOR_CST.
27761         * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
27762         ssizetype for the selector.
27764 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
27766         * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
27767         before testing each element individually.
27768         * tree-vect-generic.c (lower_vec_perm): Likewise.
27770 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
27772         * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
27773         * selftest-run-tests.c (selftest::run_tests): Call it.
27774         * vector-builder.h (vector_builder::operator ==): New function.
27775         (vector_builder::operator !=): Likewise.
27776         * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
27777         (vec_perm_indices::all_from_input_p): New function.
27778         * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
27779         (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
27780         * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
27781         instead of reading the VECTOR_CST directly.  Detect whether both
27782         vector inputs are the same before constructing the vec_perm_indices,
27783         and update the number of inputs argument accordingly.  Use the
27784         utility functions added above.  Only construct sel2 if we need to.
27786 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
27788         * optabs.c (expand_vec_perm_var): Use an explicit encoding for
27789         the broadcast of the low byte.
27790         (expand_mult_highpart): Use an explicit encoding for the permutes.
27791         * optabs-query.c (can_mult_highpart_p): Likewise.
27792         * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
27793         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
27794         (vectorizable_bswap): Likewise.
27795         * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
27796         explicit encoding for the power-of-2 permutes.
27797         (vect_permute_store_chain): Likewise.
27798         (vect_grouped_load_supported): Likewise.
27799         (vect_permute_load_chain): Likewise.
27801 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
27803         * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
27804         * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
27805         * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
27806         * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
27807         * tree-vect-stmts.c (vectorizable_bswap): Likewise.
27808         (vect_gen_perm_mask_any): Likewise.
27810 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
27812         * int-vector-builder.h: New file.
27813         * vec-perm-indices.h: Include int-vector-builder.h.
27814         (vec_perm_indices): Redefine as an int_vector_builder.
27815         (auto_vec_perm_indices): Delete.
27816         (vec_perm_builder): Redefine as a stand-alone class.
27817         (vec_perm_indices::vec_perm_indices): New function.
27818         (vec_perm_indices::clamp): Likewise.
27819         * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
27820         (vec_perm_indices::new_vector): New function.
27821         (vec_perm_indices::new_expanded_vector): Update for new
27822         vec_perm_indices class.
27823         (vec_perm_indices::rotate_inputs): New function.
27824         (vec_perm_indices::all_in_range_p): Operate directly on the
27825         encoded form, without computing elided elements.
27826         (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
27827         encoding.  Update for new vec_perm_indices class.
27828         * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
27829         the given vec_perm_builder.
27830         (expand_vec_perm_var): Update vec_perm_builder constructor.
27831         (expand_mult_highpart): Use vec_perm_builder instead of
27832         auto_vec_perm_indices.
27833         * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
27834         vec_perm_indices instead of auto_vec_perm_indices.  Use a single
27835         or double series encoding as appropriate.
27836         * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
27837         vec_perm_indices instead of auto_vec_perm_indices.
27838         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
27839         * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
27840         (vect_permute_store_chain): Likewise.
27841         (vect_grouped_load_supported): Likewise.
27842         (vect_permute_load_chain): Likewise.
27843         (vect_shift_permute_load_chain): Likewise.
27844         * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
27845         (vect_transform_slp_perm_load): Likewise.
27846         (vect_schedule_slp_instance): Likewise.
27847         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
27848         (vectorizable_mask_load_store): Likewise.
27849         (vectorizable_bswap): Likewise.
27850         (vectorizable_store): Likewise.
27851         (vectorizable_load): Likewise.
27852         * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
27853         vec_perm_indices instead of auto_vec_perm_indices.  Use
27854         tree_to_vec_perm_builder to read the vector from a tree.
27855         * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
27856         vec_perm_builder instead of a vec_perm_indices.
27857         (have_whole_vector_shift): Use vec_perm_builder and
27858         vec_perm_indices instead of auto_vec_perm_indices.  Leave the
27859         truncation to calc_vec_perm_mask_for_shift.
27860         (vect_create_epilog_for_reduction): Likewise.
27861         * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
27862         from auto_vec_perm_indices to vec_perm_indices.
27863         (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
27864         instead of changing individual elements.
27865         (aarch64_vectorize_vec_perm_const): Use new_vector to install
27866         the vector in d.perm.
27867         * config/arm/arm.c (expand_vec_perm_d::perm): Change
27868         from auto_vec_perm_indices to vec_perm_indices.
27869         (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
27870         instead of changing individual elements.
27871         (arm_vectorize_vec_perm_const): Use new_vector to install
27872         the vector in d.perm.
27873         * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
27874         Update vec_perm_builder constructor.
27875         (rs6000_expand_interleave): Likewise.
27876         * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
27877         (rs6000_expand_interleave): Likewise.
27879 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
27881         * optabs-query.c (can_vec_perm_var_p): Check whether lowering
27882         to qimode could truncate the indices.
27883         * optabs.c (expand_vec_perm_var): Likewise.
27885 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
27887         * Makefile.in (OBJS): Add vec-perm-indices.o.
27888         * vec-perm-indices.h: New file.
27889         * vec-perm-indices.c: Likewise.
27890         * target.h (vec_perm_indices): Replace with a forward class
27891         declaration.
27892         (auto_vec_perm_indices): Move to vec-perm-indices.h.
27893         * optabs.h: Include vec-perm-indices.h.
27894         (expand_vec_perm): Delete.
27895         (selector_fits_mode_p, expand_vec_perm_var): Declare.
27896         (expand_vec_perm_const): Declare.
27897         * target.def (vec_perm_const_ok): Replace with...
27898         (vec_perm_const): ...this new hook.
27899         * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
27900         (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
27901         * doc/tm.texi: Regenerate.
27902         * optabs.def (vec_perm_const): Delete.
27903         * doc/md.texi (vec_perm_const): Likewise.
27904         (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
27905         * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
27906         expand_vec_perm for constant permutation vectors.  Assert that
27907         the mode of variable permutation vectors is the integer equivalent
27908         of the mode that is being permuted.
27909         * optabs-query.h (selector_fits_mode_p): Declare.
27910         * optabs-query.c: Include vec-perm-indices.h.
27911         (selector_fits_mode_p): New function.
27912         (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
27913         is defined, instead of checking whether the vec_perm_const_optab
27914         exists.  Use targetm.vectorize.vec_perm_const instead of
27915         targetm.vectorize.vec_perm_const_ok.  Check whether the indices
27916         fit in the vector mode before using a variable permute.
27917         * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
27918         vec_perm_indices instead of an rtx.
27919         (expand_vec_perm): Replace with...
27920         (expand_vec_perm_const): ...this new function.  Take the selector
27921         as a vec_perm_indices rather than an rtx.  Also take the mode of
27922         the selector.  Update call to shift_amt_for_vec_perm_mask.
27923         Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
27924         Use vec_perm_indices::new_expanded_vector to expand the original
27925         selector into bytes.  Check whether the indices fit in the vector
27926         mode before using a variable permute.
27927         (expand_vec_perm_var): Make global.
27928         (expand_mult_highpart): Use expand_vec_perm_const.
27929         * fold-const.c: Includes vec-perm-indices.h.
27930         * tree-ssa-forwprop.c: Likewise.
27931         * tree-vect-data-refs.c: Likewise.
27932         * tree-vect-generic.c: Likewise.
27933         * tree-vect-loop.c: Likewise.
27934         * tree-vect-slp.c: Likewise.
27935         * tree-vect-stmts.c: Likewise.
27936         * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
27937         Delete.
27938         * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
27939         * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
27940         (aarch64_vectorize_vec_perm_const_ok): Fuse into...
27941         (aarch64_vectorize_vec_perm_const): ...this new function.
27942         (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
27943         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
27944         * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
27945         * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
27946         * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
27947         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
27948         (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
27949         into...
27950         (arm_vectorize_vec_perm_const): ...this new function.  Explicitly
27951         check for NEON modes.
27952         * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
27953         * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
27954         * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
27955         (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
27956         into...
27957         (ix86_vectorize_vec_perm_const): ...this new function.  Incorporate
27958         the old VEC_PERM_CONST conditions.
27959         * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
27960         * config/ia64/vect.md (vec_perm_const<mode>): Delete.
27961         * config/ia64/ia64.c (ia64_expand_vec_perm_const)
27962         (ia64_vectorize_vec_perm_const_ok): Merge into...
27963         (ia64_vectorize_vec_perm_const): ...this new function.
27964         * config/mips/loongson.md (vec_perm_const<mode>): Delete.
27965         * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
27966         * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
27967         * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
27968         * config/mips/mips.c (mips_expand_vec_perm_const)
27969         (mips_vectorize_vec_perm_const_ok): Merge into...
27970         (mips_vectorize_vec_perm_const): ...this new function.
27971         * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
27972         * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
27973         * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
27974         * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
27975         * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
27976         (rs6000_expand_vec_perm_const): Delete.
27977         * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
27978         Delete.
27979         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
27980         (altivec_expand_vec_perm_const_le): Take each operand individually.
27981         Operate on constant selectors rather than rtxes.
27982         (altivec_expand_vec_perm_const): Likewise.  Update call to
27983         altivec_expand_vec_perm_const_le.
27984         (rs6000_expand_vec_perm_const): Delete.
27985         (rs6000_vectorize_vec_perm_const_ok): Delete.
27986         (rs6000_vectorize_vec_perm_const): New function.
27987         (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
27988         an element count and rtx array.
27989         (rs6000_expand_extract_even): Update call accordingly.
27990         (rs6000_expand_interleave): Likewise.
27991         * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
27992         * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
27993         * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
27994         * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
27995         (rs6000_expand_vec_perm_const): Delete.
27996         * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
27997         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
27998         (altivec_expand_vec_perm_const_le): Take each operand individually.
27999         Operate on constant selectors rather than rtxes.
28000         (altivec_expand_vec_perm_const): Likewise.  Update call to
28001         altivec_expand_vec_perm_const_le.
28002         (rs6000_expand_vec_perm_const): Delete.
28003         (rs6000_vectorize_vec_perm_const_ok): Delete.
28004         (rs6000_vectorize_vec_perm_const): New function.  Remove stray
28005         reference to the SPE evmerge intructions.
28006         (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
28007         an element count and rtx array.
28008         (rs6000_expand_extract_even): Update call accordingly.
28009         (rs6000_expand_interleave): Likewise.
28010         * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
28011         * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
28012         new function.
28013         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
28015 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
28017         * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
28018         vector mode and that that mode matches the mode of the data
28019         being permuted.
28020         (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
28021         out into expand_vec_perm_var.  Do all CONST_VECTOR handling here,
28022         directly using expand_vec_perm_1 when forcing selectors into
28023         registers.
28024         (expand_vec_perm_var): New function, split out from expand_vec_perm.
28026 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
28028         * optabs-query.h (can_vec_perm_p): Delete.
28029         (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
28030         * optabs-query.c (can_vec_perm_p): Split into...
28031         (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
28032         (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
28033         particular selector is valid.
28034         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
28035         * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
28036         (vect_grouped_load_supported): Likewise.
28037         (vect_shift_permute_load_chain): Likewise.
28038         * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
28039         (vect_transform_slp_perm_load): Likewise.
28040         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
28041         (vectorizable_bswap): Likewise.
28042         (vect_gen_perm_mask_checked): Likewise.
28043         * fold-const.c (fold_ternary_loc): Likewise.  Don't take
28044         implementations of variable permutation vectors into account
28045         when deciding which selector to use.
28046         * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
28047         vec_perm_const_optab is supported; instead use can_vec_perm_const_p
28048         with a false third argument.
28049         * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
28050         to test whether the constant selector is valid and can_vec_perm_var_p
28051         to test whether a variable selector is valid.
28053 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
28055         * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
28056         * optabs-query.c (can_vec_perm_p): Likewise.
28057         * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
28058         instead of vec_perm_indices.
28059         * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
28060         (vect_gen_perm_mask_checked): Likewise,
28061         * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
28062         (vect_gen_perm_mask_checked): Likewise,
28064 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
28066         * optabs-query.h (qimode_for_vec_perm): Declare.
28067         * optabs-query.c (can_vec_perm_p): Split out qimode search to...
28068         (qimode_for_vec_perm): ...this new function.
28069         * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
28071 2018-01-02  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
28073         * rtlanal.c (canonicalize_condition): Return 0 if final rtx
28074         does not have a conditional at the top.
28076 2018-01-02  Richard Biener  <rguenther@suse.de>
28078         * ipa-inline.c (big_speedup_p): Fix expression.
28080 2018-01-02  Jan Hubicka  <hubicka@ucw.cz>
28082         PR target/81616
28083         * config/i386/x86-tune-costs.h: Increase cost of integer load costs
28084         for generic 4->6.
28086 2018-01-02  Jan Hubicka  <hubicka@ucw.cz>
28088         PR target/81616
28089         Generic tuning.
28090         * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
28091         cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
28092         and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
28093         cond_taken_branch_cost 3->4.
28095 2018-01-01  Jakub Jelinek  <jakub@redhat.com>
28097         PR tree-optimization/83581
28098         * tree-loop-distribution.c (pass_loop_distribution::execute): Return
28099         TODO_cleanup_cfg if any changes have been made.
28101         PR middle-end/83608
28102         * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
28103         convert_modes if target mode has the right side, but different mode
28104         class.
28106         PR middle-end/83609
28107         * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
28108         last argument when extracting from CONCAT.  If either from_real or
28109         from_imag is NULL, use expansion through memory.  If result is not
28110         a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
28111         the parts directly to inner mode, if even that fails, use expansion
28112         through memory.
28114         PR middle-end/83623
28115         * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
28116         check for bswap in mode rather than HImode and use that in expand_unop
28117         too.
28119 Copyright (C) 2018 Free Software Foundation, Inc.
28121 Copying and distribution of this file, with or without modification,
28122 are permitted in any medium without royalty provided the copyright
28123 notice and this notice are preserved.