S/390: Fix problem with vec_init expander
[official-gcc.git] / gcc / ChangeLog
blob4f33576c1252dc777030b7f0683a2b7e4fe20ddf
1 2018-10-15  Andreas Krebbel  <krebbel@linux.ibm.com>
3         * config/s390/s390.c (s390_expand_vec_init): Force vector element
4         into reg if it isn't a general operand.
6 2018-10-14  H.J. Lu  <hongjiu.lu@intel.com>
8         PR target/87599
9         * config/i386/sse.md (*vec_dupv2di): Add register source to
10         movddup.
12 2018-10-14  H.J. Lu  <hongjiu.lu@intel.com>
14         PR target/87572
15         * common/config/i386/i386-common.c (OPTION_MASK_ISA_AVX512F_UNSET):
16         Add OPTION_MASK_ISA_AVX512IFMA_UNSET,
17         OPTION_MASK_ISA_AVX5124FMAPS_UNSET and
18         OPTION_MASK_ISA_AVX5124VNNIW_UNSET.
20 2018-10-13  Eric Botcazou  <ebotcazou@adacore.com>
22         * dwarf2cfi.c (struct dw_trace_info): Add args_size_defined_for_eh.
23         (notice_args_size): Set it in the current trace if no insn that can
24         throw internally has been seen yet.
25         (connect_traces): When connecting args_size between traces, allow the
26         incoming values not to match if there is an insn setting it before the
27         first insn that can throw internally; in that case, force the creation
28         of a CFI note on this latter insn.
30 2018-10-13  Jonathan Wakely  <jwakely@redhat.com>
32         * opt-problem.h (opt_wrapper): Use template-argument-list when naming
33         the base class, because using the injected-class-name was not clearly
34         specified until DR 176.
36 2018-10-12  Paul Koning  <ni1d@arrl.net>
38         * config/pdp11/pdp11.md (doloop_end): New expander.
39         (doloop_end_insn): renamed from "doloop_end".
40         (addqi3): New pattern.
41         (subqi3): New pattern.
42         * config/pdp11/predicates.md (incdec_operand): New predicate.
44 2018-10-12  Yury Gribov  <tetra2005@gmail.com>
46         PR middle-end/81376
47         * real.c (format_helper::can_represent_integral_type_p): New function
48         * real.h (format_helper::can_represent_integral_type_p): Ditto.
49         * match.pd: New pattern.
51 2018-10-12  Alexandre Oliva <oliva@adacore.com>
53         * configure.ac: Introduce --enable-large-address-aware
54         to define MINGW_DEFAULT_LARGE_ADDR_AWARE.
55         * doc/install.texi: Document it.
56         * configure, config.in: Rebuilt.
57         * config/i386/mingw32.h (LINK_SPEC_LARGE_ADDR_AWARE): Define,
58         based on MINGW_DEFAULT_LARGE_ADDR_AWARE.
59         (LINK_SPEC): Insert it.
60         * config/i386/mingw-264.h: Likewise.
62         * cygming.h (HAVE_GAS_ALIGNED_COMM): Fallback-define.
64 2018-10-12  Peter Bergner  <bergner@linux.ibm.com>
66         PR rtl-optimization/87600
67         * ira-lives (non_conflicting_reg_copy_p): Disable for non LRA targets.
69 2018-10-12  Paul Koning  <ni1d@arrl.net>
71         * doc/md.texi (doloop_end): Document that the pattern code may
72         need to check operand mode.
74 2018-10-12  Wilco Dijkstra  <wdijkstr@arm.com>
76         * config/aarch64/aarch64.md (zero_extendsidi2_aarch64): Add alternatives
77         to zero-extend between int and floating-point registers.
78         (load_pair_zero_extendsidi2_aarch64): Add alternative for zero-extended
79         ldp into floating-point registers.  Add type and arch attributes.
80         (zero_extend<SHORT:mode><GPI:mode>2_aarch64): Add arch attribute.
81         Use f_loads for type attribute.
83 2018-10-11  Martin Sebor  <msebor@redhat.com>
85         * doc/extend.texi (attribute packed): Correct typos.
87 2018-10-11  Martin Sebor  <msebor@redhat.com>
89         * doc/extend.texi (attribute flatten): Mention interaction with
90         noinline.
92 2018-10-11  Jan Hubicka  <hubicka@ucw.cz>
94         PR target/87156
95         * cgraphclones.c (cgraph_node::create_version_clone_with_body):
96         Set new_decl virtual flag to zero.
98 2018-10-11  Martin Sebor  <msebor@redhat.com>
100         PR middle-end/87593
101         * doc/extend.texi (attribute format_arg): Discuss using multiple
102         attributes on a single function.
104 2018-10-11  Giuliano Belinassi  <giuliano.belinassi@usp.br>
106         PR tree-optimization/86829
107         * match.pd (sin (atan (x))): New simplification rules.
108         (cos (atan (x))): Likewise.
109         * real.c (build_sinatan_real): New function.
110         * real.h (build_sinatan_real): Prototype.
112 2018-10-11  Will Schmidt <will_schmidt@vnet.ibm.com>
114         * config/rs6000/rs6000.c (map_to_integral_tree_type): New helper
115         function.
116         (fold_mergeeo_helper): New helper function.
117         (rs6000_gimple_fold_builtin): Add hooks for vec_mergee and vec_mergeo
118         intrinsics.  Correct some whitespace indentation issues.
120 2018-10-11  Wilco Dijkstra  <wdijkstr@arm.com>
122         PR target/87511
123         * config/aarch64/aarch64.c (aarch64_mask_and_shift_for_ubfiz_p):
124         Use HOST_WIDE_INT_1U for shift.
126 2018-10-11  Doug Rupp  <rupp@adacore.com>
127             Olivier Hainque  <hainque@adacore.com>
129         * config/rs6000/vxworks.h (VXWORKS_RELAX_LINK_SPEC): New macro.
130         Pass --relax to the linker for RTPs.
131         (LINK_SPEC): Append VXWORKS_RELAX_LINK_SPEC.
133 2018-10-11  Andrew Stubbs  <ams@codesourcery.com>
134             Jan Hubicka  <jh@suse.cz>
135             Martin Jambor  <mjambor@suse.cz>
137         * print-rtl.c (print_rtx_operand_codes_E_and_V): Print how many times
138         the same elements are repeated rather than printing all of them.
139         * read-rtl.c (rtx_reader::read_rtx_operand): Recognize and expand
140         "repeated" elements.
141         * read-rtl-function.c (test_loading_repeat): New function.
142         (read_rtl_function_c_tests): Call test_loading_repeat.
143         * rtl-tests.c (test_dumping_repeat): New function.
144         (rtl_tests_c_tests): Call test_dumping_repeat.
146 2018-10-11  Richard Biener  <rguenther@suse.de>
148         * config/i386/x86-tune-costs.h (bdver?_memcpy, bdver?_memset,
149         bdver?_cost): Unify to ...
150         (bdver_memcpy, bdver_memset, bdver_cost): ... this.
151         * config/i386/i386.c (processor_cost_table): Adjust.
153 2018-10-10  Eric Botcazou  <ebotcazou@adacore.com>
155         PR middle-end/87574
156         * cgraphunit.c (cgraph_node::expand_thunk): Force DECL_IGNORED_P on
157         the thunk when expanding to GIMPLE.
159 2018-10-10  Bernd Edlinger  <bernd.edlinger@hotmail.de>
161         * varasm.c (mergeable_string_section): Don't try to move zero-length
162         strings to the merge section.
164 2018-10-10  Uros Bizjak  <ubizjak@gmail.com>
166         PR target/87573
167         * config/i386/mmx.md (const_vector 0 -> mem splitter): New splitter.
169 2018-10-10  Jakub Jelinek  <jakub@redhat.com>
171         PR target/87550
172         * config/i386/i386-builtin.def (IX86_BUILTIN_RDPMC): Move from args set
173         to special_args set.
175 2018-10-10  Richard Biener  <rguenther@suse.de>
177         * config/i386/sse.md (reduc_plus_scal_v8df, reduc_plus_scal_v4df,
178         reduc_plus_scal_v2df, reduc_plus_scal_v16sf, reduc_plus_scal_v8sf,
179         reduc_plus_scal_v4sf): Merge into pattern reducing to half width
180         and recursing and pattern terminating the recursion on SSE
181         vector width using ix86_expand_reduc.
182         (reduc_sminmax_scal_<mode>): Split into part reducing to half
183         width and recursing and SSE2 vector variant doing the final
184         reduction with ix86_expand_reduc.
185         (reduc_uminmax_scal_<mode>): Likewise for the AVX512 variants
186         with terminating the recursion at AVX level, splitting that
187         to SSE there.
189 2018-10-09  David Malcolm  <dmalcolm@redhat.com>
191         * genmatch.c (error_cb): Rename to...
192         (diagnostic_cb): ...this, converting int params to enums.
193         (fatal_at): Update for renaming.
194         (warning_at): Likewise.
195         (main): Likewise.
196         * input.c (selftest::ebcdic_execution_charset::apply):
197         Update for renaming of...
198         (selftest::ebcdic_execution_charset::on_error): ...this, renaming
199         to...
200         (selftest::ebcdic_execution_charset::on_diagnostic): ...this,
201         converting level and reason to enums.
202         (class selftest::lexer_error_sink): Rename to...
203         (class selftest::lexer_test_options): ...this, renaming field
204         "m_errors" to "m_diagnostics".
205         (selftest::lexer_test_options::apply): Update for renaming of...
206         (selftest::lexer_test_options::on_error): ...this, renaming to...
207         (selftest::lexer_test_options::on_diagnostic): ...this
208         converting level and reason to enums.
209         (selftest::test_lexer_string_locations_raw_string_unterminated):
210         Update for renamings.
211         * opth-gen.awk (struct cpp_reason_option_codes_t): Use enum for
212         "reason".
214 2018-10-09  Paul A. Clarke  <pc@us.ibm.com>
216         * config.gcc (powerpc*-*-*): Add pmmintrin.h to extra_headers.
217         * config/rs6000/pmmintrin.h: New file.
219 2018-10-09  Eric Botcazou  <ebotcazou@adacore.com>
221         PR tree-optimization/86659
222         * gimple-match.h (gimple_match_op constructors): Initialize reverse.
224 2018-10-09  Richard Biener  <rguenther@suse.de>
226         PR tree-optimization/63155
227         * tree-ssa-structalias.c: Include tree-ssa.h.
228         (get_constraint_for_ssa_var): For undefs return nothing_id.
229         (find_func_aliases): Cleanup PHI handling.
231 2018-10-09  Richard Biener  <rguenther@suse.de>
233         * tree-vectorizer.c (dump_stmt_cost): Fix cut&paste missing
234         replacements.
236 2018-10-09  Martin Liska  <mliska@suse.cz>
238         * asan.c (asan_emit_stack_protection): If a stack variable
239         is located in a same file as current function, then emit
240         line info into variable definition string.
242 2018-10-08  Eric Botcazou  <ebotcazou@adacore.com>
244         * print-rtl.c (rtx_writer::print_rtx_operand_code_i): Print column
245         information.
247 2018-10-08  Eric Botcazou  <ebotcazou@adacore.com>
249         * cgraphunit.c (cgraph_node::expand_thunk): Do not force DECL_IGNORED_P
250         on the thunk.
252 2018-10-08  Eric Botcazou  <ebotcazou@adacore.com>
254         PR tree-optimization/86659
255         * gimple-match.h (struct gimple_match_op): Add reverse field.
256         (gimple_match_op::set_op): New overloaded method.
257         * gimple-match-head.c (maybe_build_generic_op) <BIT_FIELD_REF>: Set
258         the REF_REVERSE_STORAGE_ORDER flag on the value.
259         (gimple_simplify) <GIMPLE_ASSIGN>: For BIT_FIELD_REF, propagate the
260         REF_REVERSE_STORAGE_ORDER flag and avoid simplifying if it is set.
261         
262 2018-10-08  Richard Sandiford  <richard.sandiford@arm.com>
264         PR middle-end/63155
265         * gimple-ssa-backprop.c (backprop::intersect_uses): Use
266         FOR_EACH_IMM_USE_FAST instead of FOR_EACH_IMM_USE_STMT.
268 2018-10-08  H.J. Lu  <hongjiu.lu@intel.com>
270         PR target/87517
271         * config/i386/avx512fintrin.h (_mm512_mask_fmaddsub_round_pd):
272         Defined with __builtin_ia32_vfmaddsubpd512_mask.
274 2018-10-08  Richard Biener  <rguenther@suse.de>
276         * config/i386/x86-tune-costs.h (znver1_cost): Make AVX256 vector loads
277         cost the same as AVX128 ones.
279 2018-10-08  Paul Koning  <ni1d@arrl.net>
281         * config/pdp11/pdp11-protos.h (output_block_move): Remove.
282         (expand_block_move): New function.
283         * config/pdp11/pdp11.c (output_block_move): Remove.
284         (expand_block_move): New function.
285         * config/pdp11/pdp11.h (MOVE_RATIO): New definition.
286         * config/pdp11/pdp11.md (movmemhi): Use expand_block_move.
287         (*movmemhi1): Remove.
289 2018-10-08  Robin Dapp  <rdapp@linux.ibm.com>
291         * config/s390/2827.md: Increase latencies for some FP instructions.
293 2018-10-08  Richard Biener  <rguenther@suse.de>
295         * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
296         Open a dump scope.
297         * tree-vectorizer.c (dump_stmt_cost): Add cost param and dump it.
298         * tree-vectorizer.h (dump_stmt_cost): Adjust.
299         (add_stmt_cost): Dump return value of the hook.
301 2018-10-08  Richard Biener  <rguenther@suse.de>
303         PR tree-optimization/63155
304         * tree-ssa-propagate.c (add_ssa_edge): Do cheap check first.
305         (ssa_propagation_engine::ssa_propagate): Remove redundant
306         bitmap bit clearing.
308 2018-10-05  Peter Bergner  <bergner@linux.ibm.com>
310         PR rtl-optimization/86939
311         PR rtl-optimization/87479
312         * ira.h (non_conflicting_reg_copy_p): New prototype.
313         * ira-lives.c (ignore_reg_for_conflicts): New static variable.
314         (make_hard_regno_dead): Don't add conflicts for register
315         ignore_reg_for_conflicts.
316         (make_object_dead): Likewise.
317         (non_conflicting_reg_copy_p): New function.
318         (process_bb_node_lives): Set ignore_reg_for_conflicts for copies.
319         Remove special conflict handling of REAL_PIC_OFFSET_TABLE_REGNUM.
320         * lra-lives.c (ignore_reg_for_conflicts): New static variable.
321         (make_hard_regno_dead): Don't add conflicts for register
322         ignore_reg_for_conflicts.  Remove special conflict handling of
323         REAL_PIC_OFFSET_TABLE_REGNUM.  Remove now unused argument
324         check_pic_pseudo_p and update callers.
325         (mark_pseudo_dead): Don't add conflicts for register
326         ignore_reg_for_conflicts.
327         (process_bb_lives): Set ignore_reg_for_conflicts for copies.
329 2018-10-05  Andrew Waterman  <andrew@sifive.com>
330             Jim Wilson  <jimw@sifive.com>
332         * config/riscv/riscv.md (f<quiet_pattern>_quiet<ANYF:mode><X:mode>4):
333         Add define_expand.  Add ! HONOR_SNANS check to current pattern.  Add
334         new pattern using HONOR_SNANS that emits one extra instruction.
336 2018-10-05  Segher Boessenkool  <segher@kernel.crashing.org>
338         * config/rs6000/rs6000.md (unnamed mfcr scc_comparison_operator
339         patterns): Merge SI and DI patterns to a GPR pattern.
340         (unnamed define_insn and define_split for record form of that): Merge
341         to a single define_insn_and_split pattern.
343 2018-10-05  David Malcolm  <dmalcolm@redhat.com>
345         PR c++/56856
346         * input.c (expand_location_to_spelling_point): Add param "aspect"
347         and use rather than hardcoding LOCATION_ASPECT_CARET.
348         (get_substring_ranges_for_loc): Handle the case of a single token
349         within a macro expansion.
350         * input.h (expand_location_to_spelling_point): Add "aspect" param,
351         defaulting to LOCATION_ASPECT_CARET.
353 2018-10-05  Paul Koning  <ni1d@arrl.net>
355         * config/pdp11/pdp11.c (TARGET_CXX_GUARD_TYPE): Define.
356         (TARGET_CXX_CLASS_DATA_ALWAYS_COMDAT): Define.
357         (TARGET_CXX_LIBRARY_RTTI_COMDAT): Define.
358         (pdp11_guard_type): New function.
360 2018-10-05  Paul Koning  <ni1d@arrl.net>
362         * config/pdp11/pdp11.h (FLOAT_TYPE_SIZE): Always 32.
363         * config/pdp11/pdp11.opt (mfloat32): Remove.
364         (mfloat64): Remove.
365         * doc/invoke.texi (pdp11 -mfloat32): Remove:
366         (pdp11 -mfloat64): Remove.
368 2018-10-05  Uros Bizjak  <ubizjak@gmail.com>
370         * config/i386/i386.md (*cmpxf_cc_i387): Remove pattern.
371         (*cmp<mode>_cc_i387): Ditto.
372         (*cmpu<mode>_cc_i387): Ditto.
373         (*cmp<X87MODEF:mode>_<SWI24:mode>_cc_i387): Ditto.
374         * config/i386/i386.c (ix86_expand_fp_compare): Remove
375         "scratch" argument.
376         <case IX86_FPCMP_SAHF>: Do not generate pattern with HImode clobber.
377         Emit x86_sahf_1 pattern.
378         (ix86_expand_compare): Update call to ix86_expand_fp_compare.
379         (ix86_expand_carry_flag_compare): Ditto.
381 2018-10-05  Uros Bizjak  <ubizjak@gmail.com>
383         * config/i386/i386.md (*cmpxf_i387): Change operand 2 predicate
384         to reg_or_0_operand.  Add "C" constraint.
385         (*cmpxf_cc_i387): Ditto.
386         (*cmp<mode>_i387): Change operand 2 predicate
387         to nonimm_or_0_operand.  Add "C" constraint.
388         (*cmp<mode>_cc_i387): Ditto.
389         (*cmp<mode>_0_i387): Remove insn pattern.
390         (*cmp<mode>_0_cc_i387): Ditto.
392 2018-10-05  Uros Bizjak  <ubizjak@gmail.com>
394         * config/i386/constraints.md ("C"): Do not depend on TARGET_SSE.
395         * config/i386/predicates.md (nonimm_or_0_operand): Rename
396         from vector_move_operand.  Update all uses.
398 2018-10-05  Martin Sebor  <msebor@redhat.com>
400         PR tree-optimization/87490
401         * builtins.c (expand_builtin_strnlen): Handle a null data.decl
402         consistently.
404 2018-10-05  Richard Biener  <rguenther@suse.de>
406         PR tree-optimization/63155
407         * tree-ssa-ccp.c (ccp_propagate::visit_phi): Avoid excess
408         vertical space in dumpfiles.
409         * tree-ssa-propagate.h
410         (ssa_propagation_engine::process_ssa_edge_worklist): Remove.
411         * tree-ssa-propagate.c (cfg_blocks_back): New global.
412         (ssa_edge_worklist_back): Likewise.
413         (curr_order): Likewise.
414         (cfg_blocks_get): Remove abstraction.
415         (cfg_blocks_add): Likewise.
416         (cfg_blocks_empty_p): Likewise.
417         (add_ssa_edge): Add to current or next worklist based on
418         RPO index.
419         (add_control_edge): Likewise.
420         (ssa_propagation_engine::process_ssa_edge_worklist): Fold
421         into ...
422         (ssa_propagation_engine::ssa_propagate): ... here.  Unify
423         iteration from CFG and SSA edge worklist so we process
424         everything in RPO order, prioritizing forward progress
425         over iteration.
426         (ssa_prop_init): Allocate new worklists, do not dump
427         immediate uses.
428         (ssa_prop_fini): Free new worklists.
430 2018-10-05  Richard Biener  <rguenther@suse.de>
432         * tree-core.h (tree_block::abstract_flag): Remove.
433         (tree_block::block_num): Make full 32bits.
434         * tree.def (BLOCK): Remove docs about BLOCK_ABSTRACT.
435         * tree.h (BLOCK_ABSTRACT): Remove.
436         * dwarf2out.c (gen_lexical_block_die): Remove dead code
437         resulting from BLOCK_ABSTRACT being always false.
438         (gen_inlined_subroutine_die): Likewise.
439         (gen_block_die): Likewise.
440         * tree.c (block_ultimate_origin): Likewise.
441         * tree-pretty-print.c (dump_block_node): Remove code dealing
442         with BLOCK_ABSTRACT.
443         * tree-ssa-live.c (dump_scope_block): Likewise.
444         * tree-streamer-in.c (unpack_ts_block_value_fields): Likewise.
445         * tree-streamer-out.c (pack_ts_block_value_fields): Likewise.
447 2018-10-05   Richard Biener  <rguenther@suse.de>
449         * config/i386/i386.c (ix86_add_stmt_cost): When scalar cost
450         is asked for initialize mode to the component mode of the
451         vector type.
453 2018-10-05  H.J. Lu  <hongjiu.lu@intel.com>
455         PR target/87522
456         * config/i386/gnu-user.h (ASM_SPEC): Don't pass -msse2avx to
457         assembler for -mavx.
458         * config/i386/gnu-user64.h (ASM_SPEC): Likewise.
460 2018-10-05  Segher Boessenkool  <segher@kernel.crashing.org>
462         PR target/87509
463         * config/rs6000/rs6000-builtin.def (RS6000_BUILTIN_SET_FPSCR_DRN): Use
464         RS6000_BTM_DFP.
465         * config/rs6000/rs6000.md (rs6000_set_fpscr_rn): Require the operand
466         to be DImode.  When using mffscrn, force the operand to a register.
468 2018-10-04  Uros Bizjak  <ubizjak@gmail.com>
470         * config/i386/i386.md (*fop_<X87MODEF:mode>_2_i387): Macroize insn
471         from *fop_<MODEF:mode>_2_i387 and *fop_xf_2_i387 using
472         X87MODEF mode iterator.
473         (*fop_<X87MODEF:mode>_3_i387): Macroize insn from
474         *fop_<MODEF:mode>_3_i387 and *fop_xf_3_i387 using
475         X87MODEF mode iterator.
477 2018-10-04  Vinay Kumar  <vinay.kumar@blackfigtech.com>
479         * doc/invoke.texi (-Wno-prio-ctor-dtor): Document new warning
480         -Wno-prio-ctor-dtor.
482 2018-10-04  David Malcolm  <dmalcolm@redhat.com>
484         * Makefile.in (OBJS): Add opt-problem.o.
485         * dump-context.h: Include "selftest.h.
486         (selftest::temp_dump_context): New forward decl.
487         (class dump_context): Make friend of class
488         selftest::temp_dump_context.
489         (dump_context::dump_loc_immediate): New decl.
490         (class dump_pretty_printer): Move here from dumpfile.c.
491         (class temp_dump_context): Move to namespace selftest.
492         (temp_dump_context::temp_dump_context): Add param
493         "forcibly_enable_dumping".
494         (selftest::verify_dumped_text):
495         (ASSERT_DUMPED_TEXT_EQ): Move here from dumpfile.c.
496         (selftest::verify_item):
497         (ASSERT_IS_TEXT): Move here from dumpfile.c.
498         (ASSERT_IS_TREE): Likewise.
499         (ASSERT_IS_GIMPLE): Likewise.
500         * dumpfile.c (dump_context::dump_loc): Move immediate dumping
501         to...
502         (dump_context::dump_loc_immediate): ...this new function.
503         (class dump_pretty_printer): Move to dump-context.h.
504         (dump_switch_p_1): Don't enable MSG_PRIORITY_REEMITTED.
505         (opt_info_switch_p_1): Enable MSG_PRIORITY_REEMITTED.
506         (temp_dump_context::temp_dump_context): Move to "selftest"
507         namespace.  Add param "forcibly_enable_dumping", and use it to
508         conditionalize the use of m_pp;
509         (selftest::verify_dumped_text): Make non-static.
510         (ASSERT_DUMPED_TEXT_EQ): Move to dump-context.h.
511         (selftest::verify_item): Make non-static.
512         (ASSERT_IS_TEXT): Move to dump-context.h.
513         (ASSERT_IS_TREE): Likewise.
514         (ASSERT_IS_GIMPLE): Likewise.
515         (selftest::test_capture_of_dump_calls): Pass "true" for new
516         param of temp_dump_context.
517         * dumpfile.h (enum dump_flag): Add MSG_PRIORITY_REEMITTED, adding
518         it to MSG_ALL_PRIORITIES.  Update values of TDF_COMPARE_DEBUG and
519         TDF_COMPARE_DEBUG.
520         * opt-problem.cc: New file.
521         * opt-problem.h: New file.
522         * optinfo-emit-json.cc
523         (selftest::test_building_json_from_dump_calls): Pass "true" for
524         new param of temp_dump_context.
525         * optinfo.cc (optinfo_kind_to_dump_flag): New function.
526         (optinfo::emit_for_opt_problem): New function.
527         (optinfo::emit): Clarity which emit_item is used.
528         * optinfo.h (optinfo::get_dump_location): New accessor.
529         (optinfo::emit_for_opt_problem): New decl.
530         (optinfo::emit): Make const.
531         * selftest-run-tests.c (selftest::run_tests): Call
532         selftest::opt_problem_cc_tests.
533         * selftest.h (selftest::opt_problem_cc_tests): New decl.
534         * tree-data-ref.c (dr_analyze_innermost): Convert return type from
535         bool to opt_result, converting fprintf messages to
536         opt_result::failure_at calls.  Add "stmt" param for use by the
537         failure_at calls.
538         (create_data_ref): Pass "stmt" to the dr_analyze_innermost call.
539         (runtime_alias_check_p): Convert return type from bool to
540         opt_result, converting dump_printf calls to
541         opt_result::failure_at, using the statement DDR_A for their
542         location.
543         (find_data_references_in_stmt): Convert return type from bool to
544         opt_result, converting "return false" to opt_result::failure_at
545         with a new message.
546         * tree-data-ref.h: Include "opt-problem.h".
547         (dr_analyze_innermost): Convert return type from bool to opt_result,
548         and add a const gimple * param.
549         (find_data_references_in_stmt): Convert return type from bool to
550         opt_result.
551         (runtime_alias_check_p): Likewise.
552         * tree-predcom.c (find_looparound_phi): Pass "init_stmt" to
553         dr_analyze_innermost.
554         * tree-vect-data-refs.c (vect_mark_for_runtime_alias_test):
555         Convert return type from bool to opt_result, adding a message for
556         the PARAM_VECT_MAX_VERSION_FOR_ALIAS_CHECKS zero case.
557         (vect_analyze_data_ref_dependence): Convert return type from bool
558         to opt_result.  Change sense of return type from "false"
559         effectively meaning "no problems" to "false" meaning a problem,
560         so that "return false" becomes "return opt_result::success".
561         Convert "return true" calls to opt_result::failure_at, using
562         the location of statement A rather than vect_location.
563         (vect_analyze_data_ref_dependences): Convert return type from bool
564         to opt_result.
565         (verify_data_ref_alignment): Likewise, converting dump_printf_loc
566         calls to opt_result::failure_at, using the stmt location rather
567         than vect_location.
568         (vect_verify_datarefs_alignment): Convert return type from bool
569         to opt_result.
570         (vect_enhance_data_refs_alignment): Likewise.  Split local "stat"
571         into multiple more-tightly-scoped copies.
572         (vect_analyze_data_refs_alignment): Convert return type from bool
573         to opt_result.
574         (vect_analyze_data_ref_accesses): Likewise, converting a
575         "return false" to a "return opt_result::failure_at", adding a
576         new message.
577         (vect_prune_runtime_alias_test_list): Convert return type from
578         bool to opt_result, converting dump_printf_loc to
579         opt_result::failure_at.  Add a %G to show the pertinent statement,
580         and use the stmt's location rather than vect_location.
581         (vect_find_stmt_data_reference): Convert return type from
582         bool to opt_result, converting dump_printf_loc to
583         opt_result::failure_at, using stmt's location.
584         (vect_analyze_data_refs):  Convert return type from bool to
585         opt_result.  Convert "return false" to "return
586         opt_result::failure_at", adding messages as needed.
587         * tree-vect-loop.c (vect_determine_vf_for_stmt_1): Convert return
588         type from bool to opt_result.
589         (vect_determine_vf_for_stmt): Likewise.
590         (vect_determine_vectorization_factor): Likewise, converting
591         dump_printf_loc to opt_result::failure_at, using location of phi
592         rather than vect_location.
593         (vect_analyze_loop_form_1): Convert return type from bool to
594         opt_result, converting dump_printf_loc calls, retaining the use of
595         vect_location.
596         (vect_analyze_loop_form): Convert return type from loop_vec_info
597         to opt_loop_vec_info.
598         (vect_analyze_loop_operations): Convert return type from bool to
599         opt_result, converting dump_printf_loc calls, using the location
600         of phi/stmt rather than vect_location where available.  Convert
601         various "return false" to "return opt_result::failure_at" with
602         "unsupported phi" messages.
603         (vect_get_datarefs_in_loop): Convert return type from bool to
604         opt_result.  Add a message for the
605         PARAM_LOOP_MAX_DATAREFS_FOR_DATADEPS failure.
606         (vect_analyze_loop_2): Convert return type from bool to
607         opt_result.  Ensure "ok" is set to a opt_result::failure_at before
608         each "goto again;", adding new messages where needed.
609         Add "unsupported grouped {store|load}" messages.
610         (vect_analyze_loop): Convert return type from loop_vec_info to
611         opt_loop_vec_info.
612         * tree-vect-slp.c (vect_analyze_slp): Convert return type from
613         bool to opt_result.
614         * tree-vect-stmts.c (process_use): Likewise, converting
615         dump_printf_loc call and using stmt location, rather than
616         vect_location.
617         (vect_mark_stmts_to_be_vectorized): Likeise.
618         (vect_analyze_stmt): Likewise, adding a %G.
619         (vect_get_vector_types_for_stmt): Convert return type from bool to
620         opt_result, converting dump_printf_loc calls and using stmt
621         location, rather than vect_location.
622         (vect_get_mask_type_for_stmt): Convert return type from tree to
623         opt_tree, converting dump_printf_loc calls and using stmt location.
624         * tree-vectorizer.c: Include "opt-problem.h.
625         (try_vectorize_loop_1): Flag "Analyzing loop at" dump message as
626         MSG_PRIORITY_INTERNALS.  Convert local "loop_vinfo" from
627         loop_vec_info to opt_loop_vec_info.  If if fails, and dumping is
628         enabled, use it to report at the top level "couldn't vectorize
629         loop" followed by the problem.
630         * tree-vectorizer.h (opt_loop_vec_info): New typedef.
631         (vect_mark_stmts_to_be_vectorized): Convert return type from bool
632         to opt_result.
633         (vect_analyze_stmt): Likewise.
634         (vect_get_vector_types_for_stmt): Likewise.
635         (tree vect_get_mask_type_for_stmt): Likewise.
636         (vect_analyze_data_ref_dependences): Likewise.
637         (vect_enhance_data_refs_alignment): Likewise.
638         (vect_analyze_data_refs_alignment): Likewise.
639         (vect_verify_datarefs_alignment): Likewise.
640         (vect_analyze_data_ref_accesses): Likewise.
641         (vect_prune_runtime_alias_test_list): Likewise.
642         (vect_find_stmt_data_reference): Likewise.
643         (vect_analyze_data_refs): Likewise.
644         (vect_analyze_loop): Convert return type from loop_vec_info to
645         opt_loop_vec_info.
646         (vect_analyze_loop_form): Likewise.
647         (vect_analyze_slp): Convert return type from bool to opt_result.
649 2018-10-04  David Malcolm  <dmalcolm@redhat.com>
651         * doc/invoke.texi (-fopt-info): Document new "internals"
652         sub-option.
653         * dump-context.h (dump_context::apply_dump_filter_p): New decl.
654         * dumpfile.c (dump_options): Update for renaming of MSG_ALL to
655         MSG_ALL_KINDS.
656         (optinfo_verbosity_options): Add "internals".
657         (kind_as_string): Update for renaming of MSG_ALL to MSG_ALL_KINDS.
658         (dump_context::apply_dump_filter_p): New member function.
659         (dump_context::dump_loc): Use apply_dump_filter_p rather than
660         explicitly masking the dump_kind.
661         (dump_context::begin_scope): Increment the scope depth first.  Use
662         apply_dump_filter_p rather than explicitly masking the dump_kind.
663         (dump_context::emit_item): Use apply_dump_filter_p rather than
664         explicitly masking the dump_kind.
665         (dump_dec): Likewise.
666         (dump_hex): Likewise.
667         (dump_switch_p_1): Default to MSG_ALL_PRIORITIES.
668         (opt_info_switch_p_1): Default to MSG_PRIORITY_USER_FACING.
669         (opt_info_switch_p): Update handling of default
670         MSG_OPTIMIZED_LOCATIONS to cope with default of
671         MSG_PRIORITY_USER_FACING.
672         (dump_basic_block): Use apply_dump_filter_p rather than explicitly
673         masking the dump_kind.
674         (selftest::test_capture_of_dump_calls): Update test_dump_context
675         instances to use MSG_ALL_KINDS | MSG_PRIORITY_USER_FACING rather
676         than MSG_ALL.  Generalize scope test to be run at all four
677         combinations of with/without MSG_PRIORITY_USER_FACING and
678         MSG_PRIORITY_INTERNALS, adding examples of explicit priority
679         for each of the two values.
680         * dumpfile.h (enum dump_flag): Add comment about the MSG_* flags.
681         Rename MSG_ALL to MSG_ALL_KINDS.  Add MSG_PRIORITY_USER_FACING,
682         MSG_PRIORITY_INTERNALS, and MSG_ALL_PRIORITIES, updating the
683         values for TDF_COMPARE_DEBUG and TDF_ALL_VALUES.
684         (AUTO_DUMP_SCOPE): Add a note to the comment about the interaction
685         with MSG_PRIORITY_*.
686         * tree-vect-loop-manip.c (vect_loop_versioning): Mark versioning
687         dump messages as MSG_PRIORITY_USER_FACING.
688         * tree-vectorizer.h (DUMP_VECT_SCOPE): Add a note to the comment
689         about the interaction with MSG_PRIORITY_*.
691 2018-10-04  Bernd Edlinger  <bernd.edlinger@hotmail.de>
693         * varasm.c (output_constant): Add new parameter merge_strings.
694         Make strings properly zero terminated in merge string sections.
695         (mergeable_string_section): Don't fail if the last char is non-zero.
696         (assemble_variable_contents): Handle merge string sections.
697         (assemble_variable): Likewise.
698         (assemble_constant_contents): Likewise.
699         (output_constant_def_contents): Likewise.
700         (output_constructor_array_range,
701         output_constructor_regular_field): Adjust call to output_constant.
702         (output_object_block): Adjust call to assemble_constant_contents
703         and assemble_variable_contents.
705 2018-10-04  Martin Liska  <mliska@suse.cz>
707         PR c/87483
708         * cgraphunit.c (process_function_and_variable_attributes):
709         Warn about a function with alias attribute and a body.
711 2018-10-04  Martin Liska  <mliska@suse.cz>
713         PR ipa/82625
714         * multiple_target.c (redirect_to_specific_clone): New function.
715         (ipa_target_clone): Use it.
716         * tree-inline.c: Fix comment.
718 2018-10-04  David Malcolm  <dmalcolm@redhat.com>
720         * dumpfile.c (gcc::dump_manager::dump_manager): Initialize new
721         fields.
722         (gcc::dump_manager::~dump_manager): Free m_optinfo_filename.
723         (gcc::dump_manager::register_pass): New member function, adapted
724         from loop body in gcc::pass_manager::register_pass, adding a
725         call to update_dfi_for_opt_info.
726         (gcc::dump_manager::opt_info_enable_passes): Store the
727         -fopt-info options into the new fields.  Move the loop
728         bodies into...
729         (gcc::dump_manager::update_dfi_for_opt_info): ...this new member
730         function.
731         * dumpfile.h (struct opt_pass): New forward decl.
732         (gcc::dump_manager::register_pass): New decl.
733         (gcc::dump_manager::update_dfi_for_opt_info): New decl.
734         (class gcc::dump_manager): Add fields "m_optgroup_flags",
735         "m_optinfo_flags", and "m_optinfo_filename".
736         * passes.c (gcc::pass_manager::register_pass): Move all of the
737         dump-handling code to gcc::dump_manager::register_pass.
739 2018-10-04  Peter Bergner  <bergner@linux.ibm.com>
741         PR rtl-optimization/87466
742         * target.def (setjmp_preserves_nonvolatile_regs_p): New target hook.
743         * doc/tm.texi.in (TARGET_SETJMP_PRESERVES_NONVOLATILE_REGS_P): New hook.
744         * doc/tm.texi: Regenerate.
745         * ira-lives.c (process_bb_node_lives): Use the new target hook.
746         * lra-lives.c (process_bb_lives): Likewise.
747         * config/rs6000/rs6000.c (TARGET_SETJMP_PRESERVES_NONVOLATILE_REGS_P):
748         Define.
750 2018-10-04  Tamar Christina  <tamar.christina@arm.com>
752         * params.c (add_params): Fix initialization.
754 2018-10-04  Martin Liska  <mliska@suse.cz>
756         PR gcov-profile/84107
757         * tree-profile.c (init_ic_make_global_vars):
758         Remove ic_void_ptr_var and ic_gcov_type_ptr_var.
759         Come up with new ic_tuple* variables.  Emit
760         __gcov_indirect_call{,_topn} variables.
761         (gimple_gen_ic_profiler): Access the variable
762         and emit gimple.
763         (gimple_gen_ic_func_profiler): Access
764         __gcov_indirect_call.callee field.
765         (gimple_init_gcov_profiler): Use ptr_type_node.
766         * value-prof.c (gimple_ic): Use ptr_type_node.
768 2018-10-04  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
770         PR tree-optimization/85787
771         * ipa-pure-const.c (malloc_candidate_p_1): Move most of malloc_candidate_p
772         into this function and add support for detecting multiple phis.
773         (DUMP_AND_RETURN): Move from malloc_candidate_p into top-level macro.
775 2018-10-04  Martin Liska  <mliska@suse.cz>
777         PR ipa/87491
778         * ipa-inline.c (inline_to_all_callers_1):
779         Call ultimate_alias_target for node being inlined.
781 2018-10-03  Jeff Law  <law@redhat.com>
783         * gimple-ssa-sprintf.c (format_string): Do not hardcode size of
784         target's wchar_t.
785         * tree.c (get_typenode_from_name): Moved from fortran/trans-types.c.
786         * tree.h (get_typenode_from_name): Prototype.
788 2018-10-03  Uros Bizjak  <ubizjak@gmail.com>
790         * config/i386/i386.md (*cmp<X87MODEF:mode>_<SWI24:mode>_i387):
791         Change operand 2 predicate to nonimmediate_operand.
792         (*cmp<X87MODEF:mode>_<SWI24:mode>_cc_i387): Ditto.
794 2018-10-03  Martin Sebor  <msebor@redhat.com>
795             Jeff Law  <law@redhat.com>
797         * gimple-ssa-sprintf.c (struct fmtresult): Add new member and
798         initialize it.
799         (get_string_length): Detect unterminated arrays.
800         (format_string): Same.
801         (format_directive): Warn about unterminated arrays.
802         (handle_gimple_call): Mark statements with no_warning as needed.
804 2018-10-03  Jim Wilson  <jimw@sifive.com>
806         * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): For ABI_ILP32E,
807         also define __riscv_abi_rve.  Delete trailing white space.
809 2018-10-03  Paul Koning  <ni1d@arrl.net>
811         Enable LRA register allocator for PDP11.
812         * config/pdp11/constraints.md (Q): Use define_memory_constraint.
813         (R): Likewise.
814         (D): Likewise.
815         * config/pdp11/pdp11.c (pdp11_lra_p): New function.
816         * config/pdp11/pdp11.opt (-mlra): New option.
817         * doc/invoke.texi (PDP-11 Options): Document -mlra.
819 2018-10-03  Uros Bizjak  <ubizjak@gmail.com>
821         * config/i386/i386.md (*<absneg:code>extendsfdf2): Remove.
822         (*<absneg:code>extend<mode>xf2): Ditto.
824 2018-10-03  Aldy Hernandez  <aldyh@redhat.com>
826         PR tree-optimization/87415
827         * tree-vrp.c (set_value_range_with_overflow): Special case one bit
828         precision fields.
830 2018-10-02  Jeff Law  <law@redhat.com>
832         * gimple-fold.c (get_range_strlen): Only set *nonstr when
833         an unterminated string is discovered.  Bubble up range
834         even for unterminated strings.
835         (gimple_fold_builtin_strlen): Do not fold if get_range_strlen
836         indicates the string was not terminated via NONSTR.
838 2018-10-03  Aldy Hernandez  <aldyh@redhat.com>
840         * tree-vrp.c (extract_range_from_unary_expr): Special case all
841         pointer conversions.
842         Do not do anything special for anti-ranges.
844 2018-10-03  Jérôme Lambourg  <lambourg@adacore.com>
846         * config/arm/vxworks.h (ARM_TARGET2_DWARF_FORMAT): Adjust to
847         DW_EH_PE_pcrel | DW_EH_PE_indirect for RTPs.
849 2018-10-03  Martin Liska  <mliska@suse.cz>
851         PR gcov-profile/86109
852         * coverage.c (coverage_begin_function): Do not
853         mark lambdas as artificial.
854         * tree-core.h (struct GTY): Remove tm_clone_flag
855         and introduce new lambda_function.
856         * tree.h (DECL_LAMBDA_FUNCTION): New macro.
858 2018-10-02  Aaron Sawdey  <acsawdey@linux.ibm.com>
860         PR target/87474
861         * config/rs6000/rs6000-string.c (expand_strn_compare): Check that both
862         P8_VECTOR and VSX are enabled.
864 2018-10-02  Andreas Krebbel  <krebbel@linux.ibm.com>
866         * config/s390/driver-native.c (s390_host_detect_local_cpu): Add
867         0x3907 as CPU model number.
869 2018-10-02  Andreas Krebbel  <krebbel@linux.ibm.com>
871         * common/config/s390/s390-common.c: Rename PF_ARCH12 to PF_Z14.
872         * config/s390/s390.h (enum processor_flags): Rename PF_ARCH12 to
873         PF_Z14.  Rename TARGET_CPU_ARCH12 to TARGET_CPU_Z14,
874         TARGET_CPU_ARCH12_P to TARGET_CPU_Z14_P, TARGET_ARCH12 to
875         TARGET_Z14, and TARGET_ARCH12_P to TARGET_Z14_P.
876         * config/s390/s390.md: Likewise. Rename also the cpu attribute
877         value from arch12 to z14.
879 2018-10-02  Uros Bizjak  <ubizjak@gmail.com>
881         * config/i386/i386.md (fxam<mode>2_i387_with_temp): Remove.
882         (isinfxf2): Ditto.
883         (isinf<mode>2): Ditto.
885 2018-10-02  Uros Bizjak  <ubizjak@gmail.com>
887         * config/i386/i386.c (ix86_emit_i387_round): Extend op1 to XFmode
888         before emitting fxam.  Perform calculations in XFmode.
890 2018-10-02  Marc Glisse  <marc.glisse@inria.fr>
892         * match.pd (((X /[ex] A) +- B) * A): New transformation.
894 2018-10-02  Marc Glisse  <marc.glisse@inria.fr>
896         PR middle-end/87319
897         * fold-const.c (fold_plusminus_mult_expr): Handle complex and vectors.
898         * tree.c (signed_or_unsigned_type_for): Handle complex.
900 2018-10-02  Jeff Law  <law@redhat.com>
902         * gimple-fold.c (get_range_strlen): Remove dead code.
904 2018-10-02  Martin Sebor  <msebor@redhat.com>
905             Jeff Law  <law@redhat.com>
907         * builtins.c (unterminated_array): Add new arguments.
908         If argument is not terminated, bubble up size and exact
909         state to callers.
910         (expand_builtin_strnlen): Detect, avoid expanding
911         and diagnose unterminated arrays.
912         (c_strlen): Fill in offset of start of unterminated strings.
913         * builtins.h (unterminated_array): Update prototype.
915 2018-10-02  Richard Biener  <rguenther@suse.de>
917         * config/i386/sse.md (reduc_plus_scal_v4df): Avoid the use
918         of haddv4df, first reduce to SSE width and exploit the fact
919         that we only need element zero with the reduction result.
920         (reduc_plus_scal_v2df): Likewise.
922 2018-10-02  Eric Botcazou  <ebotcazou@adacore.com>
924         * dojump.h (do_jump): Delete.
925         (do_jump_1): Likewise.
926         (split_comparison): Move around.
927         * dojump.c (do_jump): Make static.
928         (do_jump_1): Likewise.
929         (jumpifnot): Move around.
930         (jumpifnot_1): Likewise.
931         (jumpif): Likewise.
932         (jumpif_1): Likewise.
933         * expr.c (expand_expr_real_1): Call jumpif[not] instead of do_jump.
935 2018-10-02  Eric Botcazou  <ebotcazou@adacore.com>
937         * reorg.c (make_return_insns): Use emit_copy_of_insn_after for the
938         insns in the delay slot and add_insn_after for the jump insn.
940 2018-10-02  Richard Biener  <rguenther@suse.de>
942         * tree-inline.c (expand_call_inline): Use the location of
943         the callee declaration for the inline-entry marker.
944         * final.c (notice_source_line): Remove special-casing of
945         NOTE_INSN_INLINE_ENTRY.
947 2018-10-01  Carl Love  <cel@us.ibm.com>
949         PR 69431
950         * config/rs6000/rs6000-builtin.def (__builtin_mffsl): New.
951         (__builtin_mtfsb0): New.
952         (__builtin_mtfsb1): New.
953         ( __builtin_set_fpscr_rn): New.
954         (__builtin_set_fpscr_drn): New.
955         * config/rs6000/rs6000.c (rs6000_expand_mtfsb_builtin): Add.
956         (rs6000_expand_set_fpscr_rn_builtin): Add.
957         (rs6000_expand_set_fpscr_drn_builtin): Add.
958         (rs6000_expand_builtin): Add case statement entries for
959         RS6000_BUILTIN_MTFSB0, RS6000_BUILTIN_MTFSB1,
960         RS6000_BUILTIN_SET_FPSCR_RN, RS6000_BUILTIN_SET_FPSCR_DRN,
961         RS6000_BUILTIN_MFFSL.
962         (rs6000_init_builtins): Add ftype initialization and def_builtin
963         calls for __builtin_mffsl, __builtin_mtfsb0, __builtin_mtfsb1,
964         __builtin_set_fpscr_rn, __builtin_set_fpscr_drn.
965         * config/rs6000.md (rs6000_mtfsb0, rs6000_mtfsb1, rs6000_mffscrn,
966         rs6000_mffscdrn): Add define_insn.
967         (rs6000_set_fpscr_rn, rs6000_set_fpscr_drn): Add define_expand.
968         * doc/extend.texi: Add documentation for the builtins.
970 2018-10-01  Richard Biener  <rguenther@suse.de>
972         PR tree-optimization/87465
973         * tree-ssa-loop-ivcanon.c (tree_estimate_loop_size): Fix typo
974         causing branch miscounts.
976 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
978         * common/config/aarch64/aarch64-common.c (TARGET_OPTION_DEFAULT_PARAM,
979         aarch64_option_default_param):  New.
980         (params.h): Include.
981         (TARGET_OPTION_VALIDATE_PARAM, aarch64_option_validate_param): New.
982         * config/aarch64/aarch64.c (aarch64_override_options_internal): Simplify
983         stack-clash protection validation code.
985 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
987         * params.c (validate_param): New.
988         (add_params): Use it.
989         (set_param_value): Refactor param validation into validate_param.
990         (diagnostic.h): Include.
991         * diagnostic.h (diagnostic_ready_p): New.
993 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
995         * params.c (set_param_value):
996         Add index of parameter being validated.
997         * common/common-target.def (option_validate_param): New.
998         * common/common-targhooks.h (default_option_validate_param): New.
999         * common/common-targhooks.c (default_option_validate_param): New.
1000         * doc/tm.texi.in (TARGET_OPTION_VALIDATE_PARAM): New.
1001         * doc/tm.texi: Regenerate.
1003 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
1005         PR target/86486
1006         * config/aarch64/aarch64.c (aarch64_override_options_internal):
1007         Add validation for stack-clash parameters and set defaults.
1009 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
1011         PR target/86486
1012         * configure.ac: Add stack-clash-protection-guard-size.
1013         * doc/install.texi: Document it.
1014         * config.in (DEFAULT_STK_CLASH_GUARD_SIZE): New.
1015         * params.def: Update comment for guard-size.
1016         (PARAM_STACK_CLASH_PROTECTION_GUARD_SIZE,
1017         PARAM_STACK_CLASH_PROTECTION_PROBE_INTERVAL): Update description.
1018         * configure: Regenerate.
1020 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
1022         PR target/86486
1023         * config/aarch64/aarch64.h (STACK_CLASH_MIN_BYTES_OUTGOING_ARGS,
1024         STACK_DYNAMIC_OFFSET): New.
1025         * config/aarch64/aarch64.c (aarch64_layout_frame):
1026         Update outgoing args size.
1027         (aarch64_stack_clash_protection_alloca_probe_range,
1028         TARGET_STACK_CLASH_PROTECTION_ALLOCA_PROBE_RANGE): New.
1030 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
1032         PR target/86486
1033         * explow.c (anti_adjust_stack_and_probe_stack_clash): Support custom
1034         probe ranges.
1035         * target.def (stack_clash_protection_alloca_probe_range): New.
1036         (stack_clash_protection_final_dynamic_probe): Remove.
1037         * targhooks.h (default_stack_clash_protection_alloca_probe_range) New.
1038         (default_stack_clash_protection_final_dynamic_probe): Remove.
1039         * targhooks.c: Likewise.
1040         * doc/tm.texi.in (TARGET_STACK_CLASH_PROTECTION_ALLOCA_PROBE_RANGE): New.
1041         (TARGET_STACK_CLASH_PROTECTION_FINAL_DYNAMIC_PROBE): Remove.
1042         * doc/tm.texi: Regenerate.
1044 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
1046         PR target/86486
1047         * config/aarch64/aarch64-protos.h (aarch64_output_probe_sve_stack_clash): New.
1048         * config/aarch64/aarch64.c (aarch64_output_probe_sve_stack_clash,
1049         aarch64_clamp_to_uimm12_shift): New.
1050         (aarch64_allocate_and_probe_stack_space): Add SVE specific section.
1051         * config/aarch64/aarch64.md (probe_sve_stack_clash): New.
1053 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
1055         PR target/86486
1056         * config/aarch64/aarch64.c (aarch64_layout_frame): Add assert.
1058 2018-10-01  Jeff Law  <law@redhat.com>
1059             Richard Sandiford <richard.sandiford@linaro.org>
1060             Tamar Christina  <tamar.christina@arm.com>
1062         PR target/86486
1063         * config/aarch64/aarch64.md
1064         (probe_stack_range): Add k (SP) constraint.
1065         * config/aarch64/aarch64.h (STACK_CLASH_CALLER_GUARD,
1066         STACK_CLASH_MAX_UNROLL_PAGES): New.
1067         * config/aarch64/aarch64.c (aarch64_output_probe_stack_range): Emit
1068         stack probes for stack clash.
1069         (aarch64_allocate_and_probe_stack_space): New.
1070         (aarch64_expand_prologue): Use it.
1071         (aarch64_expand_epilogue): Likewise and update IP regs re-use criteria.
1072         (aarch64_sub_sp): Add emit_move_imm optional param.
1074 2018-10-01  MCC CS <deswurstes@users.noreply.github.com>
1076         PR tree-optimization/87261
1077         * match.pd: Remove trailing whitespace.
1078         Add (x & y) | ~(x | y) -> ~(x ^ y),
1079         (~x | y) ^ (x ^ y) -> x | ~y and (x ^ y) | ~(x | y) -> ~(x & y)
1081 2018-10-01  Claudiu Zissulescu  <claziss@synopsys.com>
1083         * config/arc/arc.md (*add_n): Clean up pattern, update instruction
1084         constraints.
1085         (ashlsi3_insn): Update instruction constraints.
1086         (ashrsi3_insn): Likewise.
1087         (rotrsi3): Likewise.
1088         (add_shift): Likewise.
1089         * config/arc/constraints.md (Csz): New 32 bit constraint. It
1090         avoids placing in the limm field small constants which, otherwise,
1091         could end into a small instruction.
1093 2018-10-01  Claudiu Zissulescu  <claziss@synopsys.com>
1095         * config/arc/arc.md (maddsidi4_split): Don't use dmac if the
1096         destination register is not odd-even.
1097         (umaddsidi4_split): Likewise.
1099 2018-10-01  Richard Biener  <rguenther@suse.de>
1101         * tree-inline.c (expand_call_inline): Store origin of fn
1102         in BLOCK_ABSTRACT_ORIGIN for the inline BLOCK.
1103         * tree.c (block_ultimate_origin): Simplify and do some
1104         checking.
1106 2018-09-30  Uros Bizjak  <ubizjak@gmail.com>
1108         * config/i386/mmx.md (EMMS): New int iterator.
1109         (emms): New int attribute.
1110         (mmx_<emms>): Macroize insn from *mmx_emms and *mmx_femms using
1111         EMMS int iterator.  Explicitly declare clobbers.
1112         (mmx_emms): Remove expander.
1113         (mmx_femms): Ditto.
1114         * config/i386/predicates.md (emms_operation): Remove predicate.
1115         (vzeroall_pattern): New predicate.
1116         (vzeroupper_pattern): Rename from vzeroupper_operation.
1117         * config/i386/i386.c (ix86_avx_u128_mode_after): Use
1118         vzeroupper_pattern and vzeroall_pattern predicates.
1120 2018-09-30  Peter Bergner  <bergner@linux.ibm.com>
1122         PR rtl-optimization/86939
1123         * ira-lives.c (make_hard_regno_born): Rename from this...
1124         (make_hard_regno_live): ... to this.  Remove update to conflict
1125         information.  Update function comment.
1126         (make_hard_regno_dead): Add conflict information update.  Update
1127         function comment.
1128         (make_object_born): Rename from this...
1129         (make_object_live): ... to this.  Remove update to conflict information.
1130         Update function comment.
1131         (make_object_dead):  Add conflict information update.  Update function
1132         comment.
1133         (mark_pseudo_regno_live): Call make_object_live.
1134         (mark_pseudo_regno_subword_live): Likewise.
1135         (mark_hard_reg_dead): Update function comment.
1136         (mark_hard_reg_live): Call make_hard_regno_live.
1137         (process_bb_node_lives): Likewise.
1138         * lra-lives.c (make_hard_regno_born): Rename from this...
1139         (make_hard_regno_live): ... to this.  Remove update to conflict
1140         information.  Remove now uneeded check_pic_pseudo_p argument.
1141         Update function comment.
1142         (make_hard_regno_dead): Add check_pic_pseudo_p argument and add update
1143         to conflict information.  Update function comment.
1144         (mark_pseudo_live): Remove update to conflict information.  Update
1145         function comment.
1146         (mark_pseudo_dead): Add conflict information update.
1147         (mark_regno_live): Call make_hard_regno_live.
1148         (mark_regno_dead): Call make_hard_regno_dead with new arguement.
1149         (process_bb_lives): Call make_hard_regno_live and make_hard_regno_dead.
1151 2018-09-29  H.J. Lu  <hongjiu.lu@intel.com>
1153         PR target/87370
1154         * config/i386/i386.c (construct_container): Use TImode for
1155         BLKmode values in 2 integer registers.
1157 2018-09-29  Jeff Law  <law@redhat.com>
1159         * builtins.c (unterminated_array): Pass in c_strlen_data * to
1160         c_strlen rather than just a tree *.
1161         (c_strlen): Change NONSTR argument to a c_strlen_data pointer.
1162         Update recursive calls appropriately.  If caller did not provide a
1163         suitable data pointer, create a local one.  When a non-terminated
1164         string is discovered, bubble up information about the string via the
1165         c_strlen_data object.
1166         * builtins.h (c_strlen): Update prototype.
1167         (c_strlen_data): New structure.
1168         * gimple-fold.c (get_range_strlen): Update calls to c_strlen.
1169         For a type 2 call, if c_strlen indicates a non-terminated string
1170         use the length of the non-terminated string.
1171         (gimple_fold_builtin_stpcpy): Update calls to c_strlen.
1173 2018-09-29  Jakub Jelinek  <jakub@redhat.com>
1175         PR target/87467
1176         * config/i386/avx512fintrin.h (_mm512_abs_pd, _mm512_mask_abs_pd): Use
1177         __m512d type for __A argument rather than __m512.
1179 2018-09-28  John David Anglin  <danglin@gcc.gnu.org>
1181         * match.pd (simple_comparison): Don't optimize if either operand is
1182         a function pointer when target needs function pointer canonicalization.
1184 2018-09-28  Segher Boessenkool  <segher@kernel.crashing.org>
1186         * config/rs6000/driver-rs6000.c (asm_names): Adjust the entries for
1187         power5 .. power9 to remove indirection.
1188         * config/rs6000/rs6000.h (ASM_CPU_POWER5_SPEC, ASM_CPU_POWER6_SPEC,
1189         ASM_CPU_POWER7_SPEC, ASM_CPU_POWER8_SPEC, ASM_CPU_POWER9_SPEC,
1190         ASM_CPU_476_SPEC): Delete.
1191         (ASM_CPU_SPEC): Adjust.
1192         (EXTRA_SPECS): Delete asm_cpu_power5, asm_cpu_power6, asm_cpu_power7,
1193         asm_cpu_power8, asm_cpu_power9, asm_cpu_476.
1195 2018-09-28  Segher Boessenkool  <segher@kernel.crashing.org>
1197         * config.in: Delete HAVE_AS_DCI.
1198         * config/powerpcspe/powerpcspe.h: Treat HAVE_AS_DCI as always true.
1199         * config/rs6000/rs6000.h: Ditto.
1200         * configure.ac: Delete HAVE_AS_DCI.
1201         * configure: Regenerate.
1203 2018-09-28  Segher Boessenkool  <segher@kernel.crashing.org>
1205         * config.in (HAVE_AS_LWSYNC): Delete.
1206         * config/powerpcspe/powerpcspe.h (TARGET_LWSYNC_INSTRUCTION): Delete.
1207         * config/powerpcspe/sync.md (*lwsync): Always generate lwsync, never
1208         do it as a .long .
1209         * config/rs6000/rs6000.h (TARGET_LWSYNC_INSTRUCTION): Delete.
1210         * config/rs6000/sync.md (*lwsync): Always generate lwsync, never do it
1211         as a .long .
1212         * configure.ac: Delete HAVE_AS_LWSYNC.
1213         * configure: Regenerate.
1215 2018-09-28  Eric Botcazou  <ebotcazou@adacore.com>
1216             Pierre-Marie de Rodat  <derodat@adacore.com>
1218         * calls.c (expand_call): Try to do a tail call for thunks at -O0 too.
1219         * cgraph.h (struct cgraph_thunk_info): Add indirect_offset.
1220         (cgraph_node::create_thunk): Add indirect_offset parameter.
1221         (thunk_adjust): Likewise.
1222         * cgraph.c (cgraph_node::create_thunk): Add indirect_offset parameter
1223         and initialize the corresponding field with it.
1224         (cgraph_node::dump): Dump indirect_offset field.
1225         * cgraphclones.c (duplicate_thunk_for_node): Deal with indirect_offset.
1226         * cgraphunit.c (cgraph_node::analyze): Be prepared for external thunks.
1227         (thunk_adjust): Add indirect_offset parameter and deal with it.
1228         (cgraph_node::expand_thunk): Deal with the indirect_offset field and
1229         pass it to thunk_adjust.  Do not call the target hook if it's non-zero
1230         or if the thunk is external or local.  Fix formatting.  Do not chain
1231         the RESULT_DECL to BLOCK_VARS.  Pass the static chain to the target,
1232         if any, in the GIMPLE representation.
1233         * ipa-icf.c (sem_function::equals_wpa): Deal with indirect_offset.
1234         * lto-cgraph.c (lto_output_node): Write indirect_offset field.
1235         (input_node): Read indirect_offset field.
1236         * tree-inline.c (expand_call_inline): Pass indirect_offset field in the
1237         call to thunk_adjust.
1238         * tree-nested.c (struct nesting_info): Add thunk_p field.
1239         (create_nesting_tree): Set it.
1240         (convert_all_function_calls): Copy static chain from targets to thunks.
1241         (finalize_nesting_tree_1): Return early for thunks.
1242         (unnest_nesting_tree_1): Do not finalize thunks.
1243         (gimplify_all_functions): Do not gimplify thunks.
1245 2018-09-28  David Malcolm  <dmalcolm@redhat.com>
1247         * opt-suggestions.c (option_proposer::build_option_suggestions):
1248         Release "option_values".
1250 2018-09-28  David Malcolm  <dmalcolm@redhat.com>
1252         * coverage.c (get_coverage_counts): Convert problem-reporting dump
1253         messages from MSG_OPTIMIZED_LOCATIONS to MSG_MISSED_OPTIMIZATION.
1254         * dumpfile.c (kind_as_string): New function.
1255         (dump_loc): Rather than a hardcoded prefix of "note: ", use
1256         kind_as_string to vary the prefix based on dump_kind.
1257         (selftest::test_capture_of_dump_calls): Update for above.
1259 2018-09-28  Uros Bizjak  <ubizjak@gmail.com>
1261         * config/i386/i386.h (SSE_REGNO): Fix check for FIRST_REX_SSE_REG.
1262         (GET_SSE_REGNO): Rename from SSE_REGNO.  Update all uses for rename.
1264 2018-09-28  Uros Bizjak  <ubizjak@gmail.com>
1266         * config/i386/i386.h (CC_REGNO): Remove FPSR_REG.
1267         * config/i386/i386.c (ix86_fixed_condition_code_regs): Use
1268         INVALID_REGNUM instead of FPSR_REG.
1269         (ix86_md_asm_adjust): Do not clobber FPSR_REG.
1270         * config/i386/i386.md: Update comment of FP compares.
1271         (fldenv): Do not clobber FPSR_REG.
1273 2018-09-28  Richard Biener  <rguenther@suse.de>
1275         * tree.h (BLOCK_ORIGIN): New.
1276         * omp-expand.c (grid_expand_target_grid_body): Assign
1277         BLOCK_ORIGIN to BLOCK_ABSTRACT_ORIGIN.
1278         * tree-inline.c (remap_block): Likewise.
1279         * auto-profile.c (get_function_decl_from_block): Simplify
1280         by eliding the BLOCK_ABSTRACT_ORIGIN chasing.
1281         * langhooks.c (lhd_print_error_function): Likewise.
1282         * optinfo-emit-json.cc (optrecord_json_writer::inlining_chain_to):
1283         Likewise.
1284         * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
1285         * tree.c (block_nonartificial_location): Likewise.
1286         (block_ultimate_origin): Likewise.
1287         * tree-pretty-print.c (percent_K_format): Likewise.  Remove
1288         no longer needed LTO case.
1290 2018-09-28  Andrew Stubbs  <ams@codesourcery.com>
1291             Jan Hubicka  <jh@suse.cz>
1292             Martin Jambor  <mjambor@suse.cz>
1294         * simplify-rtx.c (simplify_merge_mask): New function.
1295         (simplify_ternary_operation): Use it, also see if VEC_MERGEs with the
1296         same masks are used in op1 or op2.
1297         (test_vec_merge): New function.
1298         (test_vector_ops): Call test_vec_merge.
1300 2018-09-28  Eric Botcazou  <ebotcazou@adacore.com>
1302         * config/sparc/sparc-protos.h (sparc_branch_cost): Declare.
1303         * config/sparc/sparc.h (BRANCH_COST): Call sparc_branch_cost.
1304         * config/sparc/sparc.c (struct processor_costs): Add branch_cost field.
1305         (cypress_costs): Set it.
1306         (supersparc_costs): Likewise.
1307         (hypersparc_costs): Likewise.
1308         (leon_cost): Likewise.
1309         (leon3_costs): Likewise.
1310         (sparclet_costs): Likewise.
1311         (ultrasparc_costs): Likewise.
1312         (ultrasparc_costs): Likewise.
1313         (niagara_costs): Likewise.
1314         (niagara2_costs): Likewise.
1315         (niagara3_costs): Likewise.
1316         (niagara4_costs): Likewise.
1317         (niagara7_costs): Likewise.
1318         (m8_costs): Likewise.
1319         (TARGET_CAN_FOLLOW_JUMP): Define.
1320         (pass_work_around_errata::gate): Minor tweak.
1321         (sparc_option_override): Remove MASK_FSMULD mask for V7 processors.
1322         Do not set both MASK_VIS4 and MASK_VIS4B for M8 processor.
1323         Automaitcally clear MASK_FSMULD mask for V7 processors.
1324         (sparc_can_follow_jump): New static function.
1325         (output_ubranch): Deal with CROSSING_JUMP_P.
1326         (sparc_use_sched_lookahead): Rewrite using switch statement.
1327         (sparc_issue_rate): Reorder.
1328         (sparc_branch_cost): New function.
1330 2018-09-27  Martin Sebor  <msebor@redhat.com>
1332         * tree.h (tree_to_shwi): Add attribute nonnull and pure.
1333         (tree_to_poly_int64, tree_to_uhwi, tree_to_poly_uint64): Same.
1334         (int_fits_type_p): Same.
1336 2018-09-27  Uros Bizjak  <ubizjak@gmail.com>
1338         * config/i386/i386.md (FPCR_REG): Remove.
1339         (UNSPEC_FLDCW): Remove.
1340         (x86_fnstcw_1): Use (const_int 0) instead of FPCR_REG.
1341         (x86_fldcw_1): Remove insn pattern.
1342         (fnstenv): Do not clobber FPCR_REG.
1343         (fldenv): Ditto.
1344         * config/i386/i386.h (FIXED_REGISTERS) Remove fpsr register.
1345         (CALL_USED_REGISTERS): Ditto.
1346         (REG_ALLOC_ORDER): Ditto.
1347         (REG_CLASS_CONTENTS): Ditto.
1348         (HI_REGISTER_NAMES): Ditto.
1349         (ADDITIONAL_REGISTER_NAMES): Use defines instead
1350         of numerical constants.
1351         * config/i386/i386.c (regclass_map): Remove fpsr register.
1352         (dbx_register_map): Ditto.
1353         (dbx64_register_map): Ditto.
1354         (svr4_dbx_register_map): Ditto.
1355         (print_reg): Do not handle FPCR_REG.
1357 2018-09-27  Segher Boessenkool  <segher@kernel.crashing.org>
1359         PR target/87149
1360         * config.in (HAVE_AS_CMPB, HAVE_AS_DFP, HAVE_AS_FPRND, HAVE_AS_MFPGPR,
1361         HAVE_AS_POPCNTB, HAVE_AS_POPCNTD, HAVE_AS_POWER8, HAVE_AS_POWER9):
1362         Delete, always treat as true.
1363         * config/powerpcspe/powerpcspe.c (rs6000_option_override_internal):
1364         Ditto.  Simplify remaining code.
1365         * config/powerpcspe/powerpcspe.h: Ditto.
1366         * config/rs6000/rs6000.c (rs6000_option_override_internal): Ditto.
1367         Simplify remaining code.
1368         (rs6000_expand_builtin): Ditto.
1369         * config/rs6000/rs6000.h: Ditto.
1370         * configure.ac: Ditto.
1371         * configure: Regenerate.
1373 2018-09-27  Martin Liska  <mliska@suse.cz>
1375         * coverage.c (get_coverage_counts): Revert the formatting
1376         of missing profile opt info.
1378 2018-09-27  Richard Biener  <rguenther@suse.de>
1380         PR debug/37801
1381         PR debug/87440
1382         * dwarf2out.c (set_block_origin_self): Do not mark outermost
1383         block as we do not output that.
1384         (gen_inlined_subroutine_die): Elide the originally outermost
1385         block, matching what we do for concrete instances.
1386         (decls_for_scope): Add parameter specifying whether to recurse
1387         to subblocks.
1389 2018-09-27  Andrew Stubbs  <ams@codesourcery.com>
1390             Tom de Vries  <tom@codesourcery.com>
1392         PR 82089
1394         * expmed.c (emit_cstore): Fix handling of result_mode == BImode and
1395         STORE_FLAG_VALUE == 1.
1397 2018-09-27  Andreas Krebbel  <krebbel@linux.ibm.com>
1399         * config/s390/s390.md (PPA_TX_ABORT, PPA_OOO_BARRIER): New
1400         constant definitions.
1401         ("tx_assist"): Replace magic number with PPA_TX_ABORT.
1402         ("*ppa"): Enable pattern also for -march=zEC12 -mno-htm.
1403         ("speculation_barrier"): New expander definition.
1405 2018-09-26  Indu Bhagat  <indu.bhagat@oracle.com>
1407         PR gcov-profile/86957
1408         * common.opt: New warning option -Wmissing-profile.
1409         * coverage.c (get_coverage_counts): Add warning for missing .gcda file.
1410         * doc/invoke.texi: Document -Wmissing-profile.
1412 2018-09-26  Jim Wilson  <jimw@sifive.com>
1414         * config/riscv/riscv.md (subsi3_extended2): Add J constraint.
1415         (negdi2, negsi2, negsi2_extended, negsi2_extended2): New.
1417 2018-09-26  Martin Sebor  <msebor@redhat.com>
1419         * tree.c (zerop): Change return type to bool.
1420         (integer_zerop, integer_onep, integer_each_onep): Same.
1421         (integer_all_onesp, integer_minus_onep, integer_pow2p): Same.
1422         (integer_nonzerop, integer_truep, tree_ctz, real_zerop): Same.
1423         (real_onep, real_minus_onep, chain_index): Same.
1424         (print_type_hash_statistics, type_list_equal): Same.
1425         * tree.h (zerop): Same.
1426         (zerop, integer_zerop, integer_onep, integer_each_onep): Same.
1427         (integer_all_onesp, integer_minus_onep, integer_pow2p): Same.
1428         (integer_nonzerop, integer_truep, tree_ctz, real_zerop): Same.
1429         (real_onep, real_minus_onep, chain_index): Same.
1430         (print_type_hash_statistics, type_list_equal): Same.
1432 2018-09-26  Jim Wilson  <jimw@sifive.com>
1434         * config/riscv/riscv.h (FUNCTION_ARG_REGNO_P): Fix comment.
1436 2018-09-26  Jakub Jelinek  <jakub@redhat.com>
1438         PR target/87414
1439         * config/i386/i386.c: Include debug.h and dwarf2out.h.
1440         (output_indirect_thunk): Emit DW_CFA_def_cfa_offset after the
1441         call.
1443 2018-09-25  Andrew Stubbs  <ams@codesourcery.com>
1445         * builtins.c (get_builtin_sync_mem): Force address mode conversion.
1447 2018-09-26  Uros Bizjak  <ubizjak@gmail.com>
1449         * config/i386/i386.h (enum reg_class): Remove FP_TOP_SSE_REGS
1450         and FP_SECOND_SSE_REGS.
1451         (REG_CLASS_NAMES): Ditto.
1452         (REG_CLASS_CONTENTS): Ditto.
1453         * config/i386/i386.c (ix86_preferred_reload_class) Do not handle
1454         FP_TOP_SSE_REGS and FP_SECOND_SSE_REGS classes.
1455         (ix86_preferred_output_reload_class): Ditto.
1456         * config/i386/i386.md (fix_trunc<mode>_i387_fisttp): Change "=&1f"
1457         clobber constraint to "=&f".
1458         (fix_truncdi_i387): Ditto.
1459         (lrintxfdi2): Ditto.
1460         (fistdi2_<rounding>): Ditto.
1461         (fpremxf4_i387): Change "=u" constraint to "=f".
1462         (fprem1xf4_i387): Ditto.
1463         (sincosxf3): Ditto.
1464         (fptanxf4_i387): Ditto.
1465         (fxtractxf3_i387): Ditto.
1466         (fscalexf4_i387): Ditto.
1467         (atan2xf3): Change "u" constraint to "f".
1468         (fyl2xxf3_i387): Ditto.
1469         (fyl2xp1xf3_i387): Ditto.
1471 2018-09-26  Uros Bizjak  <ubizjak@gmail.com>
1473         PR target/87439
1474         * config/i386/i386.h (NUM_MODES_FOR_MODE_SWITCHING): Update
1475         for removed I387_MASK_PM entity.
1478 2018-09-26  Jeff Law  <law@redhat.com>
1479         Revert
1480         2018-09-26  Alexey Neyman  <stilor@att.net>
1482         * graphite.h: Include <isl/id.h> and <isl/space.h>; these
1483         headers are no longer pulled in by <isl/val.h>.
1485 2018-09-26  Richard Biener  <rguenther@suse.de>
1487         PR debug/87443
1488         * dwarf2out.c (gen_lexical_block_die): Do not equate inline
1489         or concrete instance DIE to the tree.  Create abstract origin
1490         attributes also for concrete instances.
1492 2018-09-26  Alexey Neyman  <stilor@att.net>
1494         * graphite.h: Include <isl/id.h> and <isl/space.h>; these
1495         headers are no longer pulled in by <isl/val.h>.
1497 2018-09-26  Matthew Malcomson  <matthew.malcomson@arm.com>
1499         * config/arm/arm.c (arm_split_compare_and_swap, arm_split_atomic_op):
1500         Use new helper functions.
1501         * config/arm/sync.md (atomic_load<mode>, atomic_store<mode>):
1502         Use new helper functions.
1503         * config/arm/aarch-common-protos.h (aarch_mm_needs_acquire,
1504         aarch_mm_needs_release): New declarations.
1505         * config/arm/aarch-common.c (aarch_mm_needs_acquire,
1506         aarch_mm_needs_release): New.
1508 2018-09-26  Eric Botcazou  <ebotcazou@adacore.com>
1510         * config/arm/arm.c (arm_reorg): Skip Thumb reorg pass for thunks.
1511         (arm32_output_mi_thunk): Deal with long calls.
1513 2018-09-26  Richard Biener  <rguenther@suse.de>
1515         PR debug/87428
1516         PR debug/87362
1517         * tree-inline.c (expand_call_inline): When the location
1518         of the call is UNKNOWN_LOCATION use DECL_SOURCE_LOCATION
1519         or BUILTINS_LOCATION for the BLOCK_SOURCE_LOCATION of
1520         the inserted BLOCK to make inlined_function_outer_scope_p
1521         recognize it.
1522         * dwarf2out.c (add_call_src_coords_attributes): Do not add
1523         coords for reserved locations.
1525 2018-09-25  Segher Boessenkool  <segher@kernel.crashing.org>
1527         * config/rs6000/rs6000.md (*movcc_internal1): Use set_attr_alternative.
1528         (*call_indirect_nonlocal_sysv<mode>): Ditto.
1529         (*call_value_indirect_nonlocal_sysv<mode>): Ditto.
1530         (*sibcall_nonlocal_sysv<mode>): Ditto.
1531         (*sibcall_value_nonlocal_sysv<mode>): Ditto.
1532         (<bd>_<mode>): Ditto.
1533         (<bd>tf_<mode>): Ditto.
1535 2018-09-25  Segher Boessenkool  <segher@kernel.crashing.org>
1537         * config/rs6000/altivec.md (*altivec_mov<mode>): Write the output
1538         control string as a list of templates instead of as C code.
1539         (*altivec_movti): Ditto.
1540         * config/rs6000/darwin.md (movdf_low_di): Ditto.
1542 2018-09-25  Jim Wilson  <jimw@sifive.com>
1544         * config/riscv/riscv.c (riscv_split_symbol): Mark auipc label as weak
1545         when target symbol is weak.
1547 2018-09-25  Bernd Edlinger  <bernd.edlinger@hotmail.de>
1549         PR c/87387
1550         * builtins.c (unterminated_array): Simplify.
1551         * expr.c (string_constant): Handle SSA_NAME.  Add more exceptions
1552         where pointer arithmetic is safe.
1554 2018-09-25  Segher Boessenkool  <segher@kernel.crashing.org>
1556         PR target/86987
1557         * config/rs6000/altivec.md (altivec_vspltb): Use
1558         const_0_to_15_operand instead of u5bit_cint_operand.
1559         (*altivec_vspltb_internal): Ditto.
1560         (altivec_vspltb_direct): Ditto.
1561         (altivec_vsplth): Use const_0_to_7_operand instead of
1562         u5bit_cint_operand.
1563         (*altivec_vsplth_internal): Ditto.
1564         (altivec_vsplth_direct): Ditto.
1565         (altivec_vspltw): Use const_0_to_3_operand instead of
1566         u5bit_cint_operand.
1567         (*altivec_vspltw_internal): Ditto.
1568         (altivec_vspltw_direct): Ditto.
1569         (altivec_vspltsf): Ditto.
1570         (*altivec_vspltsf_internal): Ditto.
1571         * config/rs6000/rs6000.c (rs6000_expand_binop_builtin): Handle the
1572         various splats with the proper size immediate.  Reorder the various
1573         cases by ascending size of immediate, and put all such together.
1575 2018-09-25  Richard Biener  <rguenther@suse.de>
1577         PR debug/83941
1578         * dwarf2out.c (add_AT_external_die_ref): Remove now redundant
1579         GC-ification.
1580         (maybe_create_die_with_external_ref): Do not create
1581         DW_TAG_imported_unit here.
1582         (add_abstract_origin_attribute): Handle external BLOCK refs.
1583         (dwarf2out_abstract_function): Simplify LTO case.
1584         (dwarf2out_early_finish): Create DW_TAG_imported_unit explicitely
1585         rather than using maybe_create_die_with_external_ref.
1587 2018-09-25  Uros Bizjak  <ubizjak@gmail.com>
1589         PR target/71278
1590         * config/i386/i386.md (frndintxf2_mask_pm): Remove.
1591         (frndintxf2_mask_pm_i387): Ditto.
1592         (nearbyintxf2): Rewrite expander pattern to match rintxf2.
1593         Enable for !flag_trapping_math.
1594         (nearbyint<mode>2): Enable x87 modes for !flag_trapping_math.
1595         Enable SSE modes for TARGET_SSE4_1 and expand them with round insn.
1596         Change operand 1 predicate to nonimmediate_operand.
1597         (attr "i387_cw"): Remove mask_pm.
1598         * config/i386/i386.h (enum ix86_stack_slot): Remove SLOT_CW_MASK_PM.
1599         (enum ix86_entity): Remove I387_MASK_PM.
1600         * config/i386/i386.c (ix86_i387_mode_needed): Do not
1601         handle I387_MASK_PM.
1602         (ix86_mode_needed): Ditto.
1603         (ix86_mode_after): Ditto.
1604         (ix86_mode_entry): Ditto.
1605         (ix86_mode_exit): Ditto.
1606         (emit_i387_cw_initialization): Do not handle I387_CW_MASK_PM.
1608 2018-09-25  Jakub Jelinek  <jakub@redhat.com>
1610         * vr-values.c (vr_values::vr_values): Initialize to_remove_edges and
1611         to_update_switch_stmts to vNULL instead of calling create on them
1612         immediately.
1614 2018-09-25  Richard Biener  <rguenther@suse.de>
1616         PR tree-optimization/87402
1617         * tree-ssa-sccvn.c (SSA_VISITED): Remove unused function.
1618         (visit_phi): Re-instantiate handling of supposed to be VARYING
1619         but non-VARYING backedge value.
1621 2018-09-25  Richard Biener  <rguenther@suse.de>
1623         PR debug/83941
1624         * dwarf2out.c (struct sym_off_pair): New.
1625         (external_die_map): New global.
1626         (lookup_decl_die): When in LTO create DIEs lazily from the
1627         external_die_map.
1628         (lookup_block_die): New function, create DIEs lazily in LTO.
1629         (equate_block_to_die): New function.
1630         (dwarf2out_die_ref_for_decl): During WPA get the association
1631         from the external DIE map.
1632         (dwarf2out_register_external_die): Record mapping into the
1633         external DIE map.
1634         (maybe_create_die_with_external_ref): New function split out from
1635         DIE generation part of old dwarf2out_register_external_die.
1636         (add_abstract_origin_attribute): Do not return the DIE.  When
1637         in LTO reference externals directly.
1638         (dwarf2out_abstract_function): When in LTO ignore calls for
1639         decls with external DIEs (already present abstract instances).
1640         (gen_call_site_die): Adjust.
1641         (add_high_low_attributes): Likewise.
1642         (gen_lexical_block_die): Likewise.
1643         (gen_inlined_subroutine_die): Likewie.
1644         (gen_block_die): Likewise.
1645         (dwarf2out_inline_entry): Likewise.
1646         (dwarf2out_early_finish): In LTRANS phase create DW_TAG_imported_unit
1647         DIEs.
1649 2018-09-25  Martin Liska  <mliska@suse.cz>
1651         * ipa-fnsummary.c (estimate_node_size_and_time): Scale by two
1652         integers and not by a float value.
1654 2018-09-25  Martin Liska  <mliska@suse.cz>
1656         PR fortran/87394
1657         * dbgcnt.c (dbg_cnt_process_single_pair): Return false
1658         instead of NULL.
1659         * dumpfile.c (dump_enable_all): Remove extra parenthesis.
1660         * gcov-tool.c: Declare the function with ATTRIBUTE_NORETURN.
1661         * godump.c (go_format_type): Remove extra parenthesis.
1663 2018-09-25  Martin Liska  <mliska@suse.cz>
1665         * alias.c (set_dest_equal_p): Remove unused function.
1666         * config/i386/i386.c (def_builtin_pure2): Likewise.
1667         * diagnostic-show-locus.c (class layout): Remove
1668         unused field.
1669         (layout::layout): Likewise here.
1670         * dump-context.h (class temp_dump_context): Likewise.
1671         * dwarf2out.c (add_AT_fde_ref): Remove unused function.
1672         (add_AT_loclistsptr): Likewise.
1673         (add_AT_offset): Likewise.
1674         (get_AT_hi_pc): Likewise.
1675         (is_comdat_die): Likewise.
1676         (type_is_enum): Likewise.
1677         (ceiling): Likewise.
1678         (add_AT_vms_delta): Likewise.
1679         (is_class_die): Likewise.
1680         * edit-context.c (class line_event): Remove unused field.
1681         * graphite-sese-to-poly.c (tree_int_to_gmp): Remove
1682         unused function.
1683         * ipa-cp.c (ipa_get_vr_lat): Likewise.
1684         * lra-constraints.c (ok_for_index_p_nonstrict): Likewise.
1685         (ok_for_base_p_nonstrict): Likewise.
1686         * tree-chrec.c (is_not_constant_evolution): Likewise.
1687         (chrec_fold_poly_cst): Likewise.
1688         * tree-if-conv.c (has_pred_critical_p): Likewise.
1689         * tree-ssa-coalesce.c (print_exprs): Likewise.
1690         * tree-ssa-pre.c (bitmap_set_contains_expr): Likewise.
1691         * tree-ssa-uninit.c (is_and_or_or_p): Likewise.
1692         * tree-vrp.c (value_ranges_intersect_p): Likewise.
1693         (value_range_nonnegative_p): Likewise.
1695 2018-09-25  Martin Liska  <mliska@suse.cz>
1697         * config/powerpcspe/powerpcspe.c (rs6000_output_function_epilogue):
1698         Do not handle "GNU Pascal".
1699         * config/rs6000/rs6000.c (rs6000_output_function_epilogue):
1700         Likewise.
1701         * config/sparc/sparc.c (sparc_pass_by_reference): Remove Pascal
1702         from documentation. Likewise.
1703         * dbxout.c (dbxout_range_type): Likewise.
1704         * doc/cpp.texi: Likewise.
1705         * doc/extend.texi: Likewise.
1706         * doc/frontends.texi: Likewise.
1707         * doc/invoke.texi: Remove Pascal entry.
1708         * tree.def (CLEANUP_POINT_EXPR): Likewise.
1709         * doc/rtl.texi (MODE_FUNCTION): Remove not used entry.
1711 2018-09-25  Martin Liska  <mliska@suse.cz>
1713         PR middle-end/86078
1714         * doc/invoke.texi: Document all parameters and remove default
1715         of the parameters.
1717 2018-09-25  Ilya Leoshkevich  <iii@linux.ibm.com>
1719         PR bootstrap/87417
1720         * rtl.c (rtx_code_size): Take into account that EQ_ATTR_ALT
1721         contains HOST_WIDE_INTs when computing its size.
1723 2018-09-24  Jim Wilson  <jimw@sifive.com>
1725         PR target/87391
1726         * config/riscv/riscv.h (STACK_BOUNDARY): Test riscv_abi == ABI_ILP32E
1727         not TARGET_RVE.
1728         (ABI_STACK_BOUNDARY, MAX_ARGS_IN_REGISTERS): Likewise.
1730 2018-09-24  Andrew Pinski  <apinski@marvell.com>
1732         *  config/aarch64/aarch64.c (aarch_macro_fusion_pair_p): Don't
1733         access prev before checking it for NULLness in the
1734         AARCH64_FUSE_CMP_BRANCH case.
1736 2018-09-24  H.J. Lu  <hongjiu.lu@intel.com>
1738         PR target/82699
1739         * config/i386/i386.c (rest_of_insert_endbranch): Set
1740         endbr_queued_at_entrance to true and don't insert ENDBR if
1741         x86_function_profiler will be called.
1742         (x86_function_profiler): Insert ENDBR if endbr_queued_at_entrance
1743         is true.
1744         * config/i386/i386.h (machine_function): Add
1745         endbr_queued_at_entrance.
1747 2018-09-24  Ilya Leoshkevich  <iii@linux.ibm.com>
1749         * genattrtab.c (mk_attr_alt): Use alternative_mask.
1750         (attr_rtx_1): Adjust caching to match the new EQ_ATTR_ALT field
1751         types.
1752         (check_attr_test): Use alternative_mask.
1753         (get_attr_value): Likewise.
1754         (compute_alternative_mask): Use alternative_mask and XWINT.
1755         (make_alternative_compare): Use alternative_mask.
1756         (attr_alt_subset_p): Use XWINT.
1757         (attr_alt_subset_of_compl_p): Likewise.
1758         (attr_alt_intersection): Use alternative_mask and XWINT.
1759         (attr_alt_union): Likewise.
1760         (attr_alt_complement): Use HOST_WIDE_INT and XWINT.
1761         (mk_attr_alt): Use alternative_mask and HOST_WIDE_INT.
1762         (simplify_test_exp): Use alternative_mask and XWINT.
1763         (write_test_expr): Use alternative_mask and XWINT, adjust bit
1764         number calculation to support 64 bits.  Generate code that
1765         checks 64-bit masks.
1766         (main): Use alternative_mask.
1767         * rtl.def (EQ_ATTR_ALT): Change field types from ii to ww.
1769 2018-09-24  Ilya Leoshkevich  <iii@linux.ibm.com>
1771         PR target/80080
1772         * config/s390/s390.c (s390_emit_epilogue): Do not use PARALLEL
1773         RETURN+USE when returning via %r14.
1775 2018-09-24  Martin Liska  <mliska@suse.cz>
1777         * gcov.c (output_lines): Print colorization legend
1778         for both flag_use_colors and flag_use_hotness_colors.
1779         Reword the help.
1781 2018-09-24  Martin Liska  <mliska@suse.cz>
1783         * coverage.c (get_coverage_counts): Use warning_at
1784         with current_function_decl location. Use %qD in warning
1785         message.
1787 2018-09-24  Martin Liska  <mliska@suse.cz>
1789         * memory-block.h (memory_block_pool::release): Annotate with
1790         valgrind that the memory is not accessible.
1792 2018-09-24  Martin Liska  <mliska@suse.cz>
1794         PR sanitizer/85774
1795         * asan.c: Make asan_handled_variables extern.
1796         * asan.h: Likewise.
1797         * cfgexpand.c (expand_stack_vars): Make sure
1798         a representative is unpoison if another
1799         variable in the partition is handled by
1800         use-after-scope sanitization.
1802 2018-09-24  Richard Biener  <rguenther@suse.de>
1804         PR tree-optimization/63155
1805         * tree-ssa-propagate.c (add_ssa_edge): Avoid adding PHIs to
1806         the worklist when the edge of the respective argument isn't
1807         executable.
1809 2018-09-23   Uros Bizjak  <ubizjak@gmail.com>
1811         * config/i386/i386.h (enum reg_class): Rename MASK_REGS to
1812         ALL_MASK_REGS and MASK_EVEX_REGS to MASK_REGS.
1813         (MASK_CLASS_P): Update for rename.
1814         (MAYBE_MASK_CLASS_P): Ditto.
1815         (REG_CLASS_NAMES): Update.
1816         (REG_CLASS_CONTENT): Update.
1817         * config/i386/i386.c (regclass_map): Update for MASK_REG
1818         and ALL_MASK_REGS rename.
1819         * config/i386/constraints.md (Yk): Update for rename.
1820         (k): Ditto.
1822 2018-09-23   Uros Bizjak  <ubizjak@gmail.com>
1824         * config/i386/i386.h (enum reg_class): Remove
1825         EVEX_SSE_REGS and MOD4_SSE_REGS.
1826         (REG_CLASS_NAMES): Update.
1827         (REG_CLASS_CONTENT): Update.
1828         * config/i386/i386.c (regclass_map): Declare AVX-512 SSE
1829         registers as ALL_SSE_REGS.
1830         (ix86_additional_allocno_class_p): Remove.
1831         (TARGET_ADDITIONAL_ALLOCNO_CLASS_P): Remove.
1832         (ix86_register_priority): Lower priority of EVEX SSE registers.
1833         Use IN_RANGE macro where appropriate.
1834         (ix86_hard_regno_mode_ok): Merge AVX-5124FMAPS and
1835         AVX-5124VNNIW checks.
1836         (ix86_modes_tieable_p): Tie 512-bit SSE modes.
1837         * config/i386/sse.md (avx5124fmaddps_4fmaddps)
1838         (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddps_maskz)
1839         (avx5124fmaddps_4fmaddss, avx5124fmaddps_4fmaddss_mask)
1840         (avx5124fmaddps_4fmaddss_maskz, avx5124fmaddps_4fnmaddps)
1841         (avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddps_maskz)
1842         (avx5124fmaddps_4fnmaddss, avx5124fmaddps_4fnmaddss_mask)
1843         (avx5124fmaddps_4fnmaddss_maskz, avx5124vnniw_vp4dpwssd)
1844         (avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssd_maskz)
1845         (avx5124vnniw_vp4dpwssds, avx5124vnniw_vp4dpwssds_mask)
1846         (avx5124vnniw_vp4dpwssds_maskz): Use "v" instead of "Yh" constraint.
1847         * config/i386/constraints.md (Yh): Remove.
1849 2018-09-23   Uros Bizjak  <ubizjak@gmail.com>
1851         * config/i386/i386.c (regclass_map): Declare integer REX registers
1852         as GENERAL_REGS.
1854 2018-09-23  Gerald Pfeifer  <gerald@pfeifer.com>
1856         * doc/service.texi (Service): Switch the fsf.org link to https.
1858 2018-09-22  Chung-Ju Wu  <jasonwucj@gmail.com>
1860         PR target/86798
1861         * config/nds32/nds32.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
1862         Define to speculation_safe_value_not_needed.
1864 2018-09-21  Florian Weimer  <fweimer@redhat.com>
1866         PR middle-end/81035
1867         * doc/extend.texi (Common Function Attributes): Mention that
1868         noreturn suppresses tail call optimization.
1870 2018-09-21  Jeff Law  <law@redhat.com>
1872         * gimple-ssa-evrp.c (evrp_dom_walker::cleanup): Call
1873         vr_values::cleanup_edges_and_switches.
1874         * tree-vrp.c (to_remove_edges, to_update_switch_stmts): Moved into
1875         vr_values class.
1876         (identify_jump_threads): Remove EDGE_IGNORE handling.
1877         (execute_vrp): Move handling of to_remove_edges and
1878         to_update_switch_stmts into vr_values class member functions.
1879         * tree-vrp.h (switch_update, to_remove_edges): Remove declarations.
1880         (to_update_switch_stmts): Likewise.
1881         * vr-values.c: Include cfghooks.h.
1882         (vr_values::vr_values): Initialize to_remove_edges and
1883         to_update_switch_stmts.
1884         (vr_values::~vr_values): Verify to_remove_edges and
1885         to_update_switch_stmts are empty.
1886         (vr_values::simplify_switch_using_ranges): Set EDGE_IGNORE as needed.
1887         (vr_values::cleanup_edges_and_switches): New member function.
1888         * vr-values.h (vr_values): Add cleanup_edges_and_switches member
1889         function.  Add new data members.
1891 2018-09-21  David Malcolm  <dmalcolm@redhat.com>
1893         PR tree-optimization/87309
1894         * dumpfile.c (dump_context::begin_scope): Filter the dump_loc
1895         calls with pflags and alt_flags.
1896         (selftest::test_capture_of_dump_calls): Add test of interaction of
1897         MSG_OPTIMIZED_LOCATIONS with AUTO_DUMP_SCOPE.
1899 2018-09-21  Olivier Hainque  <hainque@adacore.com>
1901         * config.gcc: Factorize and comment inclusion of vxworks-dummy.h.
1903 2018-09-21  Olivier Hainque  <hainque@adacore.com>
1905         * config/vxworks.h (CLEAR_INSN_CACHE): #define to 1.
1907 2018-09-21  Olivier Hainque  <hainque@adacore.com>
1909         * config/vxworks.h (VXWORKS_LIBGCC_SPEC): Remove -lc_internal.
1910         Merge block comment with the one ahead of VXWORKS_LIBS_RTP. Then:
1911         (VXWORKS_LIBS_RTP): Minor reordering.
1913 2018-09-21  Olivier Hainque  <hainque@adacore.com>
1915         * config/vxworks.h (STARTFILE_PREFIX_SPEC): Define.
1916         (VXWORKS_LIBS_DIR_RTP): Remove definition and use.
1918 2018-09-21  Olivier Hainque  <hainque@adacore.com>
1920         * config/vxworks.h (SIZE_TYPE): Account for TARGET_VXWORKS64.
1921         (PTRDIFF_TYPE): Likewise.
1923 2018-09-21  Olivier Hainque  <hainque@adacore.com>
1925         * config.gcc: Enforce def of TARGET_VXWORKS64 to 1 from
1926         triplet, similar to support for VxWorks7.
1927         * config/vxworks-dummy.h: Provide a default definition
1928         of TARGET_VXWORKS64 to 0.
1930 2018-09-21  Olivier Hainque  <hainque@adacore.com>
1932         * config/vxworks.h (TARGET_VXWORKS7): Move default definition ...
1933         * config/vxworks-dummy.h: here.
1935 2018-09-21  Olivier Hainque  <hainque@adacore.com>
1937         * config.gcc: Prepend vxworks-dummy.h to tm_file for powerpc*
1939 2018-09-21  Shaokun Zhang  <zhangshaokun@hisilicon.com>
1940             Bo Zhou  <zbo.zhou@hisilicon.com>
1942         * config/aarch64/aarch64-cores.def (tsv110): New CPU.
1943         * config/aarch64/aarch64-tune.md: Regenerated.
1944         * doc/invoke.texi (AArch64 Options/-mtune): Add "tsv110".
1945         * config/aarch64/aarch64.c (tsv110_tunings): New tuning table.
1946         * config/aarch64/aarch64-cost-tables.h: Add "tsv110" extra costs.
1948 2018-09-21  Andrew Stubbs  <ams@codesourcery.com>
1949             Julian Brown  <julian@codesourcery.com>
1951         * builtins.c (get_builtin_sync_mem): Handle address spaces.
1953 2018-09-21  Eric Botcazou  <ebotcazou@adacore.com>
1955         * config/rs6000/rs6000.c (rs6000_function_ok_for_sibcall): Return false
1956         if the call takes a static chain.
1958 2018-09-21  Martin Liska  <mliska@suse.cz>
1960         * auto-profile.c (autofdo_source_profile::read): Do not
1961         set sum_all.
1962         (read_profile): Do not add working sets.
1963         (read_autofdo_file): Remove sum_all.
1964         (afdo_callsite_hot_enough_for_early_inline): Remove const
1965         qualifier.
1966         * coverage.c (struct counts_entry): Remove gcov_summary.
1967         (read_counts_file): Read new GCOV_TAG_OBJECT_SUMMARY,
1968         do not support GCOV_TAG_PROGRAM_SUMMARY.
1969         (get_coverage_counts): Remove summary and expected
1970         arguments.
1971         * coverage.h (get_coverage_counts): Likewise.
1972         * doc/gcov-dump.texi: Remove -w option.
1973         * gcov-dump.c (dump_working_sets): Remove.
1974         (main): Do not support '-w' option.
1975         (print_usage): Likewise.
1976         (tag_summary): Likewise.
1977         * gcov-io.c (gcov_write_summary): Do not dump
1978         histogram.
1979         (gcov_read_summary): Likewise.
1980         (gcov_histo_index): Remove.
1981         (gcov_histogram_merge): Likewise.
1982         (compute_working_sets): Likewise.
1983         * gcov-io.h (GCOV_TAG_OBJECT_SUMMARY): Mark
1984         it not obsolete.
1985         (GCOV_TAG_PROGRAM_SUMMARY): Mark it obsolete.
1986         (GCOV_TAG_SUMMARY_LENGTH): Adjust.
1987         (GCOV_HISTOGRAM_SIZE): Remove.
1988         (GCOV_HISTOGRAM_BITVECTOR_SIZE): Likewise.
1989         (struct gcov_summary): Simplify rapidly just
1990         to runs and sum_max fields.
1991         (gcov_histo_index): Remove.
1992         (NUM_GCOV_WORKING_SETS): Likewise.
1993         (compute_working_sets): Likewise.
1994         * gcov-tool.c (print_overlap_usage_message): Remove
1995         trailing empty line.
1996         * gcov.c (read_count_file): Read GCOV_TAG_OBJECT_SUMMARY.
1997         (output_lines): Remove program related line.
1998         * ipa-profile.c (ipa_profile): Do not consider GCOV histogram.
1999         * lto-cgraph.c (output_profile_summary): Do not stream GCOV
2000         histogram.
2001         (input_profile_summary): Do not read it.
2002         (merge_profile_summaries): And do not merge it.
2003         (input_symtab): Do not call removed function.
2004         * modulo-sched.c (sms_schedule): Do not print sum_max.
2005         * params.def (HOT_BB_COUNT_FRACTION): Reincarnate param that was
2006         removed when histogram method was invented.
2007         (HOT_BB_COUNT_WS_PERMILLE): Mention that it's used only in LTO
2008         mode.
2009         * postreload-gcse.c (eliminate_partially_redundant_load): Fix
2010         GCOV coding style.
2011         * predict.c (get_hot_bb_threshold): Use HOT_BB_COUNT_FRACTION
2012         and dump selected value.
2013         * profile.c (add_working_set): Remove.
2014         (get_working_sets): Likewise.
2015         (find_working_set): Likewise.
2016         (get_exec_counts): Do not work with working sets.
2017         (read_profile_edge_counts): Do not inform as sum_max is removed.
2018         (compute_branch_probabilities): Likewise.
2019         (compute_value_histograms): Remove argument for call of
2020         get_coverage_counts.
2021         * profile.h: Do not make gcov_summary const.
2023 2018-09-21  Monk Chiang  <sh.chiang04@gmail.com>
2025         * config.gcc (nds32*-*-*): Set TARGET_DEFAULT_TLSDESC_TRAMPOLINE=0.
2027 2018-09-21  Eric Botcazou  <ebotcazou@adacore.com>
2029         PR tree-optimization/86990
2030         * gimple-ssa-store-merging.c (imm_store_chain_info:coalesce_immediate):
2031         Check that the entire merged store group is made of constants only for
2032         overlapping stores.
2034 2018-09-20  Allan Sandfeld Jensen  <allan.jensen@qt.io>
2036         * gcc.c (LINK_COMMAND_SPEC): Handle -r like -nostdlib.
2037         (VTABLE_VERIFICATION_SPEC): Likewise.
2038         (SANITIZER_EARLY_SPEC): Likewise.
2039         (SANITIZER_SPEC): Likewise.
2040         * config/darwin.h (LINK_COMMAND_SPEC): Likewise.
2041         * doc/invoke.texi (Link Options): Document -r.
2043 2018-09-20  Richard Biener <rguenther@suse.de>
2045         PR middle-end/87054
2046         * gimplify.c (gimplify_expr): Retain alignment of
2047         addressable lvalue in dereference.
2049 2018-09-20  Alexandre Oliva <aoliva@redhat.com>
2051         PR bootstrap/87013
2052         * configure.ac: Check for .loc is_stmt support.
2053         * configure, config.in: Rebuilt.
2054         * dwarf2out.c (dwarf2out_source_line): Skip is_stmt
2055         if not supported.
2057 2018-09-20  Segher Boessenkool  <segher@kernel.crashing.org>
2059         * config/rs6000/rs6000.opt (misel=no, misel=yes): Delete.
2060         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -misel=yes and
2061         -misel=no.
2063 2018-09-20  Segher Boessenkool  <segher@kernel.crashing.org>
2065         * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
2066         VECTOR_OTHER.
2067         * config/rs6000/rs6000.c (rs6000_debug_vector_unit): Delete
2068         case VECTOR_OTHER.
2070 2018-09-20  Marek Polacek  <polacek@redhat.com>
2072         * doc/invoke.texi: Add -Wno-init-list-lifetime to C++ Language Options.
2074 2018-09-20  Richard Sandiford  <richard.sandiford@arm.com>
2076         PR tree-optimization/87288
2077         * tree-vect-loop.c (vect_analyze_loop_2): Take PEELING_FOR_GAPS
2078         into account when determining PEELING_FOR_NITERS.
2080 2018-09-20  Richard Sandiford  <richard.sandiford@arm.com>
2082         PR tree-optimization/86877
2083         * tree-vect-loop.c (vect_analyze_loop_2): Call
2084         vect_verify_datarefs_alignment.
2086 2018-09-19  Marek Polacek  <polacek@redhat.com>
2088         * doc/invoke.texi: Document -Wclass-conversion.
2090 2018-09-19  John David Anglin  <danglin@gcc.gnu.org>
2092         * config/pa/pa.c (pa_adjust_priority): Delete.
2093         (TARGET_SCHED_ADJUST_PRIORITY): Delete define.
2095         * config/pa/pa.md (atomic_storeqi): Restore deleted expander.
2096         (atomic_storehi): Likewise.
2097         (atomic_storesi): Likewise.
2098         (atomic_loaddi): Restore compare and swap exchange loop code.
2100 2018-09-19  Segher Boessenkool  <segher@kernel.crashing.org>
2102         PR rtl-optimization/86902
2103         * combine.c (try_combine): When changing the CC mode used, don't change
2104         an unrelated mode in other_insn to that new CC mode.
2106 2018-09-19  David Malcolm  <dmalcolm@redhat.com>
2108         * tree-data-ref.c (runtime_alias_check_p): Use formatted printing
2109         with %T in place of calls to dump_generic_expr.
2110         (prune_runtime_alias_test_list): Likewise.
2111         (create_runtime_alias_checks): Likewise.
2112         * tree-vect-data-refs.c (vect_check_nonzero_value): Likewise.
2113         (vect_analyze_data_ref_dependence): Likewise.
2114         (vect_slp_analyze_data_ref_dependence): Likewise.
2115         (vect_record_base_alignment): Likewise.  Use %G in place of call
2116         to dump_gimple_stmt.
2117         (vect_compute_data_ref_alignment): Likewise.
2118         (verify_data_ref_alignment): Likewise.
2119         (vect_find_same_alignment_drs): Likewise.
2120         (vect_analyze_group_access_1): Likewise.
2121         (vect_analyze_data_ref_accesses): Likewise.
2122         (dependence_distance_ge_vf): Likewise.
2123         (dump_lower_bound): Likewise.
2124         (vect_prune_runtime_alias_test_list): Likewise.
2125         (vect_find_stmt_data_reference): Likewise.
2126         (vect_analyze_data_refs): Likewise.
2127         (vect_create_addr_base_for_vector_ref): Likewise.
2128         (vect_create_data_ref_ptr): Likewise.
2129         * tree-vect-loop-manip.c (vect_set_loop_condition): Likewise.
2130         (vect_can_advance_ivs_p): Likewise.
2131         (vect_update_ivs_after_vectorizer): Likewise.
2132         (vect_gen_prolog_loop_niters): Likewise.
2133         (vect_prepare_for_masked_peels): Likewise.
2134         * tree-vect-loop.c (vect_determine_vf_for_stmt): Likewise.
2135         (vect_determine_vectorization_factor): Likewise.
2136         (vect_is_simple_iv_evolution): Likewise.
2137         (vect_analyze_scalar_cycles_1): Likewise.
2138         (vect_analyze_loop_operations): Likewise.
2139         (report_vect_op): Likewise.
2140         (vect_is_slp_reduction): Likewise.
2141         (check_reduction_path): Likewise.
2142         (vect_is_simple_reduction): Likewise.
2143         (vect_create_epilog_for_reduction): Likewise.
2144         (vect_finalize_reduction:): Likewise.
2145         (vectorizable_induction): Likewise.
2146         (vect_transform_loop_stmt): Likewise.
2147         (vect_transform_loop): Likewise.
2148         (optimize_mask_stores): Likewise.
2149         * tree-vect-patterns.c (vect_pattern_detected): Likewise.
2150         (vect_split_statement): Likewise.
2151         (vect_recog_over_widening_pattern): Likewise.
2152         (vect_recog_average_pattern): Likewise.
2153         (vect_determine_min_output_precision_1): Likewise.
2154         (vect_determine_precisions_from_range): Likewise.
2155         (vect_determine_precisions_from_users): Likewise.
2156         (vect_mark_pattern_stmts): Likewise.
2157         (vect_pattern_recog_1): Likewise.
2158         * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
2159         (vect_record_max_nunits): Likewise.
2160         (vect_build_slp_tree_1): Likewise.
2161         (vect_build_slp_tree_2): Likewise.
2162         (vect_print_slp_tree): Likewise.
2163         (vect_analyze_slp_instance): Likewise.
2164         (vect_detect_hybrid_slp_stmts): Likewise.
2165         (vect_detect_hybrid_slp_1): Likewise.
2166         (vect_slp_analyze_operations): Likewise.
2167         (vect_slp_analyze_bb_1): Likewise.
2168         (vect_transform_slp_perm_load): Likewise.
2169         (vect_schedule_slp_instance): Likewise.
2170         * tree-vect-stmts.c (vect_mark_relevant): Likewise.
2171         (vect_mark_stmts_to_be_vectorized): Likewise.
2172         (vect_init_vector_1): Likewise.
2173         (vect_get_vec_def_for_operand): Likewise.
2174         (vect_finish_stmt_generation_1): Likewise.
2175         (vect_check_load_store_mask): Likewise.
2176         (vectorizable_call): Likewise.
2177         (vectorizable_conversion): Likewise.
2178         (vectorizable_operation): Likewise.
2179         (vectorizable_load): Likewise.
2180         (vect_analyze_stmt): Likewise.
2181         (vect_is_simple_use): Likewise.
2182         (vect_get_vector_types_for_stmt): Likewise.
2183         (vect_get_mask_type_for_stmt): Likewise.
2184         * tree-vectorizer.c (increase_alignment): Likewise.
2186 2018-09-19  Andrew Stubbs  <ams@codesourcery.com>
2188         * doc/rtl.texi: Adjust vec_select description.
2189         * simplify-rtx.c (simplify_binary_operation_1): Allow VEC_SELECT to use
2190         non-constant selectors.
2192 2018-09-19  Matthew Malcomson  <matthew.malcomson@arm.com>
2194         * config/aarch64/aarch64-protos.h
2195         (aarch64_offset_9bit_signed_unscaled_p): New declaration.
2196         * config/aarch64/aarch64.md (arches): New "rcpc8_4" attribute value.
2197         (arch_enabled): Add check for "rcpc8_4" attribute value of "arch".
2198         * config/aarch64/aarch64.h (AARCH64_FL_RCPC8_4): New bitfield.
2199         (AARCH64_FL_FOR_ARCH8_4): Include AARCH64_FL_RCPC8_4.
2200         (AARCH64_FL_PROFILE): Move index so flags are ordered.
2201         (AARCH64_ISA_RCPC8_4): New flag.
2202         * config/aarch64/aarch64.c (offset_9bit_signed_unscaled_p): Renamed
2203         to aarch64_offset_9bit_signed_unscaled_p.
2204         * config/aarch64/atomics.md (atomic_store<mode>): Allow offset
2205         and use stlur.
2206         * config/aarch64/constraints.md (Ust): New constraint.
2207         * config/aarch64/predicates.md.
2208         (aarch64_9bit_offset_memory_operand): New predicate.
2209         (aarch64_rcpc_memory_operand): New predicate.
2211 2018-09-19  Eric Botcazou  <ebotcazou@adacore.com>
2213         PR rtl-optimization/87361
2214         * rtlanal.c (nonzero_bits1): Revert accidental change.
2216 2018-09-19  Richard Biener  <rguenther@suse.de>
2218         PR tree-optimization/87349
2219         PR tree-optimization/87342
2220         * tree-ssa-sccvn.c (do_rpo_vn): Iterate max_rpo computation.
2222 2018-09-18  Marek Polacek  <polacek@redhat.com>
2224         P1064R0 - Allowing Virtual Function Calls in Constant Expressions
2225         * gimple-fold.c (gimple_get_virt_method_for_vtable): Adjust assert.
2227 2018-09-18  Segher Boessenkool  <segher@kernel.crashing.org>
2229         * config/rs6000/rs6000.md: Remove old "Cygnus sibcall" comment.
2231 2018-09-18  Segher Boessenkool  <segher@kernel.crashing.org>
2233         PR rtl-optimization/86882
2234         * rtlanal.c (reg_overlap_mentioned_p): Handle CLOBBER.
2236 2018-09-18  Uros Bizjak  <ubizjak@gmail.com>
2238         * config/i386/i386.md (*<code>extend<mode>xf2): Macroize insn from
2239         *<code>extendsfxf2 and *<code>extenddfxf2 using MODEF mode iterator.
2241 2018-09-18  Jonathan Wakely  <jwakely@redhat.com>
2243         PR other/87353
2244         * doc/invoke.texi (Link Options): Fix formatting and grammar.
2246 2018-09-18  Richard Biener  <rguenther@suse.de>
2248         PR middle-end/63155
2249         * tree-ssa-coalesce.c (tree_int_map_hasher): Remove.
2250         (compute_samebase_partition_bases): Likewise.
2251         (coalesce_ssa_name): Always use compute_optimized_partition_bases.
2252         (gimple_can_coalesce_p): Simplify.
2254 2018-09-18  Hans-Peter Nilsson  <hp@bitrange.com>
2256         Handle a library implementation of ffs calling __builtin_ffs.
2257         * config/mmix/mmix.c (TARGET_INIT_LIBFUNCS): Override with...
2258         (mmix_init_libfuncs): New function: make __builtin_ffs expand
2259         to __ffsdi2.
2261 2018-09-17  David Malcolm  <dmalcolm@redhat.com>
2263         * diagnostic-show-locus.c (class layout_range): Add field
2264         "m_original_idx".
2265         (layout_range::layout_range): Add "original_idx" param and use it
2266         to initialize new field.
2267         (make_range): Use 0 for original_idx.
2268         (layout::layout): Pass in index to calls to
2269         maybe_add_location_range.
2270         (layout::maybe_add_location_range): Add param "original_idx" and
2271         pass it on to layout_range.
2272         (layout::print_any_labels): Pass on range->m_original_idx to
2273         get_text call.
2274         (gcc_rich_location::add_location_if_nearby): Use 0 for
2275         original_idx.
2276         * gcc-rich-location.h (text_range_label::get_text): Update for new
2277         param.
2278         (range_label_for_type_mismatch::get_text): Likewise.
2280 2018-09-17  Uros Bizjak  <ubizjak@gmail.com>
2282         * config/i386/i386.c (ix86_emit_i387_log1p): Emit fldln2 earlier.
2284 2018-09-17  David Malcolm  <dmalcolm@redhat.com>
2286         * gimple-ssa-sprintf.c (fmtwarn): Update for introduction of
2287         format_string_diagnostic_t.
2288         (fmtwarn_n): Likewise.
2289         * substring-locations.c
2290         (format_string_diagnostic_t::format_string_diagnostic_t) New ctor.
2291         (format_warning_n_va): Convert to...
2292         (format_string_diagnostic_t::emit_warning_n_va): ...this.
2293         (format_warning_va): Convert to...
2294         (format_string_diagnostic_t::emit_warning_va): ...this.
2295         (format_warning_at_substring): Convert to...
2296         (format_string_diagnostic_t::emit_warning): ...this.
2297         (format_warning_at_substring_n): Convert to...
2298         (format_string_diagnostic_t::emit_warning_n): ...this.
2299         * substring-locations.h (class format_string_diagnostic_t): New
2300         class.
2301         (format_warning_va): Convert to
2302         format_string_diagnostic_t::emit_warning_va.
2303         (format_warning_n_va): Convert to
2304         format_string_diagnostic_t::emit_warning_n_va.
2305         (format_warning_at_substring): Convert to
2306         format_string_diagnostic_t::emit_warning.
2307         (format_warning_at_substring_n): Convert to
2308         format_string_diagnostic_t::emit_warning_n.
2310 2018-09-17  Cesar Philippidis  <cesar@codesourcery.com>
2311             Bernd Schmidt <bernds_cb1@t-online.de>
2313         * config/nvptx/nvptx.md (atomic_fetch_<logic><mode>): Enable with
2314         SImode args.
2316 2018-09-17  Uros Bizjak  <ubizjak@gmail.com>
2318         * config/i386/i386.md (truncxf<mode>2_i387_noop_unspec): Change
2319         operand 0 predicate to nonimmediate operand.
2320         (rint<mode>2_frndint): Remove insn pattern.
2321         (rint<mode>2): Change operand 1 predicate to general_operand.
2322         Extend operand 1 to XFmode and generate rintxf2 insn.
2323         (frndintxf2_<rounding>): Rename from frndint<mode>2_<rounding>.
2324         Do not use X87MODEF mode macro.
2325         (frndintxf2_<rounding>_i387): Rename from
2326         frndint<mode>2_<rounding>_i387.  Do not use X87MODEF mode macro.
2327         (<rounding_insn><mode>2): For non-SSE modes, extend operand 1
2328         to XFmode and generate significandxf3 insn.
2330 2018-09-17  Richard Biener  <rguenther@suse.de>
2332         PR tree-optimization/87328
2333         * tree-ssa-sccvn.c (process_bb): Remove assertion about not
2334         visiting unexecutable backedges when not iterating.
2335         (do_rpo_vn): Mark all edges not executable even when not
2336         iterating.
2338 2018-09-17  Martin Jambor  <mjambor@suse.cz>
2340         PR c/63886
2341         * doc/invoke.texi (Warning Options): Likewise.
2343 2018-09-17  Richard Biener  <rguenther@suse.de>
2345         PR tree-optimization/87301
2346         * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_cleanup): Properly
2347         clean EH info from leftover copy assignments.
2349 2018-09-17  Martin Liska  <mliska@suse.cz>
2351         PR gcov-profile/85871
2352         * gcov.c (output_intermediate_file): Fix out of bounds
2353         access.
2355 2018-09-17  Vineet Gupta  <vgupta@synopsys.com>
2357         * config/arc/arc.c: Object attributes for core4 not reflected
2358         correctly.
2359         * config/arc/arc.h: Don't restrict DBNZ to core3 (core4 includes
2360         core3).
2362 2018-09-17  Alexey Brodkin  <abrodkin@synopsys.com>
2364         * config/arc/linux.h (LINK_EH_SPEC): Add missing space.
2366 2018-09-17  Martin Liska  <mliska@suse.cz>
2368         * doc/gcov.texi: Document new option --use-hotness-colors.
2369         * gcov.c (struct source_info): Declare new field.
2370         (source_info::source_info): Set default for maximum_count.
2371         (print_usage): Add new -q option.
2372         (process_args): Process it.
2373         (accumulate_line_info): Save src->maximum_count.
2374         (output_line_beginning): Make color line number if
2375         flag_use_hotness_colors is set.
2376         (output_line_details): Pass default argument value.
2377         (output_lines): Pass src->maximum_count.
2379 2018-09-17  Martin Liska  <mliska@suse.cz>
2381         * common/config/i386/i386-common.c (ix86_get_valid_option_values):
2382         Use processor_names table.
2383         * config/i386/i386.c (ix86_default_align): Use
2384         processor_cost_table for alignment values.
2385         (ix86_option_override_internal): Use processor_names.
2386         (ix86_function_specific_print): Likewise.
2387         * config/i386/i386.h (struct processor_costs):
2388         Add alignment values.
2389         (struct ptt): Remove and replace with const char *.
2390         * config/i386/x86-tune-costs.h (struct processor_costs):
2391         Declare default alignments for all costs.
2393 2018-09-17  Aldy Hernandez  <aldyh@redhat.com>
2395         * tree-vrp.c (extract_range_from_unary_expr): Do not special case
2396         symbolics or VR_VARYING ranges for ABS_EXPR.
2397         * wide-int-range.cc (wide_int_range_abs): Return positive numbers
2398         when range will wrap.
2400 2018-09-15  Eric Botcazou  <ebotcazou@adacore.com>
2402         PR middle-end/86864
2403         * cfgexpand.c (expand_gimple_basic_block): Be prepared for a BARRIER
2404         before and after a JUMP_TABLE_DATA.
2406 2018-09-14  John David Anglin  <danglin@gcc.gnu.org>
2408         PR middle-end/87188
2409         * dojump.c (do_compare_and_jump): Canonicalize function pointers
2410         when one operand is a function pointer.  Use POINTER_TYPE_P and
2411         FUNC_OR_METHOD_TYPE_P.
2412         * expr.c (do_store_flag): Use POINTER_TYPE_P and FUNC_OR_METHOD_TYPE_P.
2413         * fold-const.c (build_range_check): Likewise.
2414         * match.pd (simple_comparison): Likewise.
2416 2018-09-14  David Malcolm  <dmalcolm@redhat.com>
2418         PR c/82967
2419         * spellcheck.c (get_edit_distance_cutoff): New function.
2420         (selftest::test_edit_distance_unit_test_oneway): Rename to...
2421         (selftest::test_get_edit_distance_one_way): ...this.
2422         (selftest::test_get_edit_distance_unit): Rename to...
2423         (selftest::test_get_edit_distance_both_ways): ...this.
2424         (selftest::test_edit_distances): Move tests to this new function,
2425         and test some more pairs of strings.  Update for above renaming.
2426         (selftest::get_old_cutoff): New function.
2427         (selftest::test_get_edit_distance_cutoff): New function.
2428         (selftest::assert_suggested_for): New function.
2429         (ASSERT_SUGGESTED_FOR): New macro.
2430         (selftest::assert_not_suggested_for): New function.
2431         (ASSERT_NOT_SUGGESTED_FOR): New macro.
2432         (selftest::test_suggestions): New function.
2433         (selftest::spellcheck_c_tests): Move test_get_edit_distance_unit
2434         tests to selftest::test_edit_distances and call it.  Add calls to
2435         selftest::test_get_edit_distance_cutoff and
2436         selftest::test_suggestions.
2437         * spellcheck.h (get_edit_distance_cutoff): New function declaration.
2438         (best_match::consider): Replace hard-coded cutoff calculation with
2439         a call to...
2440         (best_match::get_cutoff): New declaration.
2441         (best_match::get_best_meaningful_candidate): Likewise.
2443 2018-09-14  Bernd Edlinger  <bernd.edlinger@hotmail.de>
2445         * builtins.c (fold_builtin_strlen): Remove TODO comment.
2447 2018-09-14  Bernd Edlinger  <bernd.edlinger@hotmail.de>
2449         revert:
2450         2018-07-30  Bernd Edlinger  <bernd.edlinger@hotmail.de>
2452         * tree-ssa-forwprop.c (simplify_builtin_call): Don't create a not NUL
2453         terminated string literal.
2455 2018-09-14  Martin Sebor  <msebor@redhat.com>
2457         * builtins.c (unterminated_array): Handle ARRAY_REF.
2458         (expand_builtin_stpcpy_1): Detect unterminated char arrays.
2459         * builtins.h (unterminated_array): Declare extern.
2460         * gimple-fold.c (gimple_fold_builtin_stpcpy): Detect unterminated
2461         arrays.
2462         (gimple_fold_builtin_sprintf): Propagate NO_WARNING to transformed
2463         calls.
2465 2018-09-14  Martin Sebor  <msebor@redhat.com>
2466             Jeff Law  <law@redhat.com>
2468         * builtins.c (unterminated_array): New.
2469         (expand_builtin_strcpy): Adjust.
2470         (expand_builtin_strcpy_args): Detect unterminated arrays.
2471         * gimple-fold.c (get_maxval_strlen): Add argument.  Detect
2472         unterminated arrays.
2473         * gimple-fold.h (get_maxval_strlen): Add argument.
2474         (gimple_fold_builtin_strcpy): Detec unterminated arrays.
2476         * gimple-fold.c (get_range_strlen): Add argument.
2477         (get_maxval_strlen): Adjust.
2478         * gimple-fold.h (get_range_strlen): Add argument.
2480 2018-09-14  Wei Xiao  <wei3.xiao@intel.com>
2482         * config/i386/movdirintrin.h: Fix copyright year.
2484 2018-09-14  Uros Bizjak  <ubizjak@gmail.com>
2486         * reg-stack.c: Include regs.h.
2487         (replace_reg): Assert that mode is MODE_FLOAT or MODE_COMPLEX_FLOAT.
2488         (emit_pop_insn): Default pop insn mode to the reg_raw_mode of
2489         FIRST_STACK_REG, not DFmode.
2490         (emit_swap_insn): Default swap insn mode to the reg_raw_mode of
2491         FIRST_STACK_REG, not XFmode.  Explicitly construct swap RTX.
2492         (change stack): Default register mode to the reg_raw_mode of
2493         FIRST_STACK_REG, not DFmode.
2494         * config/i386/i386.md (*swap<mode>): Remove insn pattern.
2495         (*swapxf): Rename from swapxf.
2497 2018-09-14  Carl Love  <cel@us.ibm.com>
2499         * config/rs6000/emmintrin.h: Add _MM_SHUFFLE2.
2500         * config/rs6000/xmmintrin.h: Add _MM_SHUFFLE.
2502 2018-09-14  Segher Boessenkool  <segher@kernel.crashing.org>
2504         PR target/87224
2505         * config/rs6000/rs6000.md (*mov<mode>_hardfloat64): Add Z to the Y
2506         alternatives.
2508 2018-09-14  Sam Tebbs  <sam.tebbs@arm.com>
2510         PR target/85628
2511         * config/aarch64/aarch64.md (*aarch64_bfxilsi_uxtw): Define.
2513 2018-09-14  Jason Merrill  <jason@redhat.com>
2515         Fix --enable-gather-detailed-mem-stats.
2516         * hash-table.c (hash_table_usage): Change from variable to function.
2517         * hash-table.h: Adjust.
2518         * Makefile.in: Add missing dependencies on hash-table.h.
2520 2018-09-14  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
2522         PR tree-optimization/87259
2523         PR lto/87283
2524         (pass_cse_reciprocals::execute): Run optimize_recip_sqrt after
2525         execute_cse_reciprocals_1 has tried transforming.
2527 2018-09-14  Aldy Hernandez  <aldyh@redhat.com>
2529         * tree-vrp.c (extract_range_from_binary_expr_1): Normalize
2530         VR_VARYING for PLUS/MINUS_EXPR.
2532 2018-09-14  Ilya Leoshkevich  <iii@linux.ibm.com>
2534         * config/s390/s390-passes.def (INSERT_PASS_BEFORE): Improve
2535         formatting.
2537 2018-09-14  Richard Biener  <rguenther@suse.de>
2539         PR middle-end/63155
2540         * tree-ssa-coalesce.c (ssa_conflicts_merge): Remove conflict
2541         bits for the merged partition.
2543 2018-09-13  Martin Sebor  <msebor@redhat.com>
2544             Bernd Edlinger  <bernd.edlinger@hotmail.de>
2546         * builtins.h (c_srlen): Add argument.
2547         * builtins.c (warn_string_no_nul): New function.
2548         (c_strlen): Add argument and use it.  Update recursive calls.
2549         Pass DECL argument to string_constant to get info on non
2550         terminated strings.  Update *NONSTR as needed.
2551         (fold_builtin_strlen): Add argument to calls to c_strlen.
2552         Warn for unterminated arrays.
2553         (warn_string_no_null): Add prototype.
2554         * expr.c (string_constant): Update arguments.  Update recursive
2555         calls appropriately.  Detect missing NUL terminator and outermost
2556         declaration its missing in.
2557         Improve checks for arrays with nonzero lower bound or elements
2558         that are not a single byte.  Simplify offset computation.
2559         Simplify checks for non-NUL terminated strings.
2560         * gimple-fold.c (get_range_strlen): Add argument to c_strlen call.
2561         * gimple-ssa-sprintf.c (get_string_length): Remove unnecessary code.
2563 2018-09-13  Bernd Edlinger  <bernd.edlinger@hotmail.de>
2565         * builtins.c (c_strlen): Handle not zero terminated STRING_CSTs
2566         correctly.
2567         * fold-const.c (c_getstr): Fix function comment.  Remove unused third
2568         argument.  Fix range checks.
2569         * fold-const.h (c_getstr): Adjust protoype.
2570         * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid folding when
2571         string is constant but contains no NUL byte.
2573         * expr.c (string_constant): Adjust function comment.
2574         Remove bogus check for zero termination.
2576         * fold-const.c (c_getstr): Clamp STRING_LENGTH to STRING_SIZE.
2578         * varasm.c (compare_constant): Compare type size of STRING_CSTs.
2579         (get_constant_size): Don't make STRING_CSTs larger than they are.
2580         (check_string_literal): New check function for STRING_CSTs.
2581         (output_constant): Use it.
2583 2018-09-13  Eric Botcazou  <ebotcazou@adacore.com>
2585         PR target/86812
2586         * config/visium/visium.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define.
2588 2018-09-13  Richard Biener  <rguenther@suse.de>
2590         PR tree-optimization/87263
2591         * tree-ssa-sccvn.c (visit_phi): Revert some earlier changes.
2592         (struct unwind_state): Add max_rpo field.
2593         (do_rpo_vn): Allow up-to-date loop state to be used when not iterating.
2594         Compute max_rpo, the max RPO number a block can be backwards reached
2595         from.  Re-write non-iterating mode to a RPO ordered worklist approach,
2596         separating it from the iterating mode.
2598 2018-09-13  Vlad Lazar  <vlad.lazar@arm.com>
2600         * haifa-sched.c (rank_for_schedule): Schedule by INSN_COST.
2601         (rfs_decision): New scheduling decision.
2603 2018-09-13  Richard Biener  <rguenther@suse.de>
2605         PR bootstrap/87134
2606         * tree-ssa-sccvn.c (vn_nary_op_insert_into): Fix assert.
2607         (vn_nary_op_insert_pieces_predicated): Do not write useless
2608         valid_dominated_by_p entry outside of the allocated storage.
2610 2018-09-13  Omar Sandoval  <osandov@osandov.com>
2611             Tom de Vries  <tdevries@suse.de>
2613         PR debug/86985
2614         * dwarf2out.c (is_c): New function.
2615         (add_subscript_info): Add DW_AT_count of 0 for C zero-length arrays.
2617 2018-09-13  Sam Tebbs  <sam.tebbs@arm.com>
2619         PR target/85628
2620         * config/aarch64/aarch64.md (*aarch64_bfxil):
2621         Define.
2622         * config/aarch64/constraints.md (Ulc): Define.
2623         * config/aarch64/aarch64-protos.h (aarch64_high_bits_all_ones_p):
2624         Define.
2625         * config/aarch64/aarch64.c (aarch64_high_bits_all_ones_p):
2626         New function.
2628 2018-09-13  Vlad Lazar  <vlad.lazar@arm.com>
2630         * config/aarch64/aarch64.h (TARGET_COMPUTE_FRAME_LAYOUT): Define.
2631         * config/aarch64/aarch64.c (aarch64_expand_prologue): Remove
2632         aarch64_layout_frame call.
2633         (aarch64_expand_epilogue): Likewise.
2634         (aarch64_initial_elimination_offset): Likewise.
2635         (aarch64_get_separate_components): Likewise.
2636         (aarch64_use_return_insn_p): Likewise.
2637         (aarch64_layout_frame): Remove unneeded check.
2639 2018-09-13  Jakub Jelinek  <jakub@redhat.com>
2641         * configure.ac: Only append
2642         " : (reconfigured) $TOPLEVEL_CONFIGURE_ARGUMENTS" to
2643         gcc_config_arguments if it was never reconfigured or last reconfigure
2644         was with different arguments.
2645         * configure: Regenerated.
2647 2018-09-13  Jakub Jelinek  <jakub@redhat.com>
2648             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
2650         PR middle-end/87290
2651         * expr.c (maybe_optimize_pow2p_mod_cmp): New function.
2652         (maybe_optimize_mod_cmp): Use it if integer_pow2p treeop1.
2654 2018-09-13  Jakub Jelinek  <jakub@redhat.com>
2656         PR tree-optimization/87287
2657         * fold-const.c (fold_binary_loc) <case EQ_EXPR>: Move signed modulo
2658         X % C == 0 to X % (unsigned) C == 0 optimization to ...
2659         * match.pd (X % C == 0): ... here.  New optimization.
2661 2018-09-12  Jakub Jelinek  <jakub@redhat.com>
2663         PR middle-end/82853
2664         * expr.h (maybe_optimize_mod_cmp): Declare.
2665         * expr.c (mod_inv): New function.
2666         (maybe_optimize_mod_cmp): New function.
2667         (do_store_flag): Use it.
2668         * cfgexpand.c (expand_gimple_cond): Likewise.
2670 2018-09-09  Cesar Philippidis  <cesar@codesourcery.com>
2671             Julian Brown  <julian@codesourcery.com>
2673         PR middle-end/86336
2674         * gimplify.c (gimplify_scan_omp_clauses): Set
2675         target_firstprivatize_array_bases in OpenACC parallel and kernels
2676         region contexts.  Remove GOMP_MAP_FIRSTPRIVATE_REFERENCE clauses from
2677         OpenACC data regions.
2679 2018-09-12  Uros Bizjak  <ubizjak@gmail.com>
2681         * config/i386/i386.md (sqrt_extend<mode>xf3_i387): Remove.
2682         (sqrt<mode>2): Extend operand 1 to XFmode and generate sqrtxf3 insn.
2684 2018-09-12  Richard Biener  <rguenther@suse.de>
2686         PR tree-optimization/87280
2687         * tree-ssa-sccvn.c (process_bb): Handle the case of executable
2688         edge but unreachable target.
2689         (do_rpo_vn): For conservatively handling a PHI only mark
2690         the backedge executable but not the block reachable.
2692 2018-09-12  Richard Biener  <rguenther@suse.de>
2694         PR tree-optimization/87266
2695         * tree-ssa-sccvn.c (do_rpo_vn): Always iterate to not yet
2696         visited blocks.
2698 2018-09-12  Andreas Krebbel  <krebbel@linux.ibm.com>
2700         * config/s390/s390.md (PFPO_RND_MODE_DFP, PFPO_RND_MODE_BFP): New
2701         constants.
2702         ("trunc<BFP:mode><DFP_ALL:mode>2")
2703         ("trunc<DFP_ALL:mode><BFP:mode>2")
2704         ("extend<BFP:mode><DFP_ALL:mode>2")
2705         ("extend<DFP_ALL:mode><BFP:mode>2"): Set proper rounding mode
2706         according to the target operand type.
2708 2018-09-12  Jakub Jelinek  <jakub@redhat.com>
2709             Andreas Krebbel  <krebbel@linux.ibm.com>
2711         PR tree-optimization/86844
2712         * gimple-ssa-store-merging.c
2713         (imm_store_chain_info::coalesce_immediate): For overlapping stores, if
2714         there are any overlapping stores in between them, make sure they are
2715         also coalesced or we give up completely.
2717 2018-09-12  Jakub Jelinek  <jakub@redhat.com>
2719         PR middle-end/87248
2720         * fold-const.c (fold_ternary_loc) <case COND_EXPR>: Verify also that
2721         BIT_AND_EXPR's second operand is a power of two.  Formatting fix.
2723 2018-09-12  Tom de Vries  <tdevries@suse.de>
2725         * common.opt (gdescribe-dies): Add option.
2726         * dwarf2out.c (add_name_and_src_coords_attributes): Add description
2727         attribute for artifical and nameless decls.
2728         (dwarf2out_register_external_die): Add description attribute to
2729         external reference die.
2730         (add_desc_attribute): New functions.
2731         (gen_subprogram_die): Add description attribute to
2732         DW_TAG_call_site_parameter.
2733         * tree-pretty-print.c (print_generic_expr_to_str): New function.
2734         * tree-pretty-print.h (print_generic_expr_to_str): Declare.
2735         * doc/invoke.texi (@item Debugging Options): Add -gdescribe-dies and
2736         -gno-describe-dies.
2737         (@item -gdescribe-dies): Add.
2739 2018-09-12  Aldy Hernandez  <aldyh@redhat.com>
2741         * tree-vrp.c (vrp_shift_undefined_p): Remove.
2742         (extract_range_from_binary_expr_1: Call
2743         wide_int_range_shift_undefined_p instead of vrp_shift_undefined_p.
2744         * wide-int-range.h (wide_int_range_shift_undefined_p): Do not
2745         depend on sign.
2747 2018-09-12  Aldy Hernandez  <aldyh@redhat.com>
2749         * gimple-ssa-warn-alloca.c
2750         (alloca_type_and_limit::alloca_type_and_limit): Initialize limit
2751         field for ALLOCA_BOUND_*_LARGE.
2753 2018-09-11  Nathan Sidwell  <nathan@acm.org>
2755         * gcc.c (load_specs, execute, run_attempt): Use %qs not '%s'.
2757 2018-09-11  Uros Bizjak  <ubizjak@gmail.com>
2759         * reg-stack.c (subst_asm_stack_regs): Call replace_reg also
2760         for clobbers.  Remove obsolete comment.
2762 2018-09-11  Uros Bizjak  <ubizjak@gmail.com>
2764         * config/i386/i386.md (define_attr "type"): Remove mpxmov, mpxmk,
2765         mpxchk, mpxld and mpxst types.
2766         (define_attr length_immediate): Remove all processing of mpx types.
2767         (define_attr prefix_0f): Ditto.
2768         (define_attr memory): Ditto.
2770 2018-09-11  Uros Bizjak  <ubizjak@gmail.com>
2772         * config/i386/i386.md (fyl2x_extend<mode>xf3_i387): Remove.
2773         (log<mode>2): Change operand 1 predicate to general_operand.
2774         Extend operand 1 to XFmode and generate logxf3 insn.
2775         (log10<mode>2): Change operand 1 predicate to general_operand.
2776         Extend operand 1 to XFmode and generate log10xf3 insn.
2777         (log2<mode>2): Change operand 1 predicate to general_operand.
2778         Extend operand 1 to XFmode and generate log2xf3 insn.
2779         (fyl2xp1_extend<mode>xf3_i387): Remove.
2780         (log1p<mode>2): Change operand 1 predicate to general_operand.
2781         Extend operand 1 to XFmode and generate log1pxf3 insn.
2782         (fxtract_extend<mode>xf3_i387): Remove.
2783         (logb<mode>2): Change operand 1 predicate to general_operand.
2784         Extend operand 1 to XFmode and generate logbxf3 insn.
2785         (ilogb<mode>2): Change operand 1 predicate to general_operand.
2786         Extend operand 1 to XFmode and generate fxtractxf3_i387 insn.
2787         (significand<mode>2): Change operand 1 predicate to general_operand.
2788         Extend operand 1 to XFmode and generate significandxf3 insn.
2790 2018-09-11  Nathan Sidwell  <nathan@acm.org>
2792         * gcc.c (perror_with_name, pfatal_with_name): Delete.
2793         (load_specs): Use fatal_error.
2794         (DELETE_IF_ORDINARY, process_command): Use error.
2795         (execute, run_attempt): Use fatal_error.
2797 2018-09-11  Andrew Stubbs  <ams@codesourcery.com>
2799         * diagnostic-core.h (sorry_at): New prototype.
2800         * diagnostic.c (sorry_at): New function.
2802 2018-09-11  Aldy Hernandez  <aldyh@redhat.com>
2804         * tree-vrp.c (extract_range_from_binary_expr_1): Treat all divisions
2805         by zero as VR_UNDEFINED.
2807 2018-09-10  Uros Bizjak  <ubizjak@gmail.com>
2809         * config/i386/i386.md (<sincos>xf2): Rename from *<sincos>xf2_i387.
2810         (*<sincos>_extend<mode>xf2_i387): Remove insn pattern.
2811         (<sincos>mode2): New expander.
2812         (sincos_extend<mode>xf3_i387): Remove insn pattern.
2813         (sincos -> sin, cos splitters): Remove splitter patterns.
2814         (sincos<mode>3): Change operand 2 predicate to general_operand.
2815         Extend operand 2 to XFmode and generate sincosxf3 insn.
2816         (fptanxf4_i387): Change mode of operands 0 and 3 to SFmode.
2817         Change operand 3 predicate to const1_operand.
2818         (fptan_extend<mode>xf4_i387): Remove insn pattern.
2819         (tanxf2): Update operands in the call to fptanxf4_i387.
2820         (tan<mode>2): Change operand 1 predicate to general_operand.
2821         Extend operand 1 to XFmode and generate tanxf3 insn.
2822         (atan2xf3): Rename from *fpatanxf3_i387.
2823         (fpatan_extend<mode>xf3_i387): Remove insn pattern.
2824         (atan2xf3): Remove expander.
2825         (atan2<mode<3):  Change operand 1 and 2 predicates to general_operand.
2826         Extend operands 1 and 2 to XFmode and generate atan2xf3 insn.
2827         (atan<mode>2): Change operand 1 predicate to general_operand.
2828         Extend operand 1 to XFmode and generate atanxf3 insn.
2830 2018-09-10  Uros Bizjak  <ubizjak@gmail.com>
2832         * config/i386/i386.md (x87/SSE constant load splitter): Use
2833         memory_operand instead of nonimmediate_operand for input operand
2834         predicate.
2836 2018-09-09  Uros Bizjak  <ubizjak@gmail.com>
2838         * config/i386/i386.md (float partial SSE register stall splitter): Move
2839         splitter near its instruction pattern.
2840         (float_extend partial SSE register stall splitter): Ditto.
2841         (float_truncate partial SSE register stall splitter): Ditto.
2843 2018-09-09  Hans-Peter Nilsson  <hp@bitrange.com>
2845         PR target/86794
2846         * config/mmix/mmix.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
2847         to speculation_safe_value_not_needed.
2849         PR target/85666
2850         * config/mmix/mmix.c (mmix_assemble_integer): Handle byte-size
2851         non-CONST_INT rtx:es using assemble_integer_with_op ".byte".
2852         (MMIX_CFUN_NEEDS_SAVED_EH_RETURN_ADDRESS): Don't call
2853         leaf_function_p, instead use has_hard_reg_initial_val.
2855 2018-09-09  Nathan Sidwell  <nathan@acm.org>
2857         * gcc.h (pfatal_with_name): Don't declare here.
2858         * gcc.c (pfatal_with_name): Make static.
2860 2018-09-09  Xianmiao Qu  <xianmiao_qu@c-sky.com>
2862         * config/csky/csky.md (*cskyv2_adddi3, *cskyv2_subdi3): Add
2863         earlyclobber.
2865 2018-09-08  John David Anglin  <danglin@gcc.gnu.org>
2867         PR rtl-optimization/85458
2868         * sel-sched.c (sel_target_adjust_priority): Allow backend adjust
2869         priority hook to reduce the priority of EXPR.
2871 2018-09-07  Uros Bizjak  <ubizjak@gmail.com>
2873         * config/i386/i386.md (float<SWI48x:mode><MODEF:mode>2) Enable
2874         DImode for x87 on 32bit targets.  Conditionally disable x87 modes
2875         with X87_ENABLE_FLOAT.  Remove preparation code.
2876         (*float<SWI48:mode><MODEF:mode>2): Rename from
2877         *float<SWI48:mode><MODEF:mode>2_mixed.  Handle x87, SSE and mixed
2878         math using "enabled" attribute.
2879         (*floatdi<MODEF:mode>2_i387): Rename from
2880         *float<SWI48x:mode><MODEF:mode>2_i387.  Handle only DImode and
2881         enable for 32bit targets only.
2882         (floatdi<X87MODEF:mode>2_i387_with_xmm pre-reload splitter): New
2883         splitter.
2884         (floatdi<X87MODEF:mode>2_i387_with_xmm): Use register_operand
2885         as operand 1 predicate.  Rewrite as define_insn_and_split.
2886         (floatdi<X87MODEF:mode>2_i387_with_xmm memory input splitter): Remove.
2888 2018-09-06  Uros Bizjak  <ubizjak@gmail.com>
2890         * reg-stack.c (get_true_reg) <case FLOAT_TRUNCATE>: Reorder
2891         to fallthru to FLOAT case.
2893 2018-09-06  Will Schmidt  <will_schmidt@vnet.ibm.com>
2895         PR target/86731
2896         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Update logic
2897         around folding of vec_sl to handle out of range shift values.
2899 2018-09-06  Uros Bizjak  <ubizjak@gmail.com>
2901         * config/i386/i386.md (fix_trunc<mode>_fisttp_i387_1): Remove.
2902         Update callers to gen_fix_trunc<mode>_i387_fisttp
2903         (fix_trunc<mode>_i387_fisttp): Change operand 0 predicate to
2904         nonimmediate_operand.
2905         (fix_trunc<mode>_i387_fisttp_with_temp): Remove insn pattern
2906         and corresponding splitters.
2907         (*fix_trunc<mode>_i387_1): Always expand with fix_trunc<mode>_i387.
2908         (fix_truncdi_i387): Change operand 0 predicate to nonimmediate_operand.
2909         (fix_truncdi_i387_with_temp): Remove insn pattern
2910         and corresponding splitters.
2911         (fix_trunc<mode>_i387): Change operand 0 predicate to
2912         nonimmediate_operand.
2913         (fix_trunc<mode>_i387_with_temp): Remove insn pattern
2914         and corresponding splitters.
2915         (*fistdi2_1): Remove.
2916         (fistdi2): Ditto.
2917         (fistdi2_with_temp): Remove insn pattern and corresponding splitters.
2918         (lrintxfdi2): Remove expander.  Reimplement as define_insn.
2919         (*fist<mode>2_1): Remove.
2920         (fist<mode>2): Ditto.
2921         (fist<mode>2_with_temp): Remove insn pattern and corresponding
2922         splitters.
2923         (lrintxf<mode>2): Remove expander.  Reimplement as define_insn.
2924         (*fist<mode>2_<rounding>_1): Always expand with fist<mode>2_<rounding>.
2925         (fistdi2_<rounding>): Change operand 0 predicate to
2926         nonimmediate_operand.
2927         (fistdi2_<rounding>_with_temp): Remove insn pattern
2928         and corresponding splitters.
2929         (fist<mode>2_<rounding>): Change operand 0 predicate to
2930         nonimmediate_operand.
2931         (fist<mode>2_<rounding>_with_temp): Remove insn pattern
2932         and corresponding splitters.
2934         (*fixuns_trunc<mode>si2_avx512f_zext): Depend on TARGET_SSE_MATH.
2936 2018-09-06  Bernd Edlinger  <bernd.edlinger@hotmail.de>
2938         * varasm.c (output_constructor_regular_field): Check TYPE_SIZE_UNIT of
2939         the init value.
2941 2018-09-06  Will Schmidt  <will_schmidt@vnet.ibm.com>
2943         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support for
2944         early gimple folding of vec_splat().
2945         * tree-vect-generic.c: Remove static from tree_vec_extract() definition.
2946         * gimple-fold.h: Add an extern define for tree_vec_extract().
2948 2018-09-06  Will Schmidt  <will_schmidt@vnet.ibm.com>
2950         * config/rs6000/rs6000.c (fold_mergehl_helper): Add types_compatible_p
2951         wrappers around TREE_TYPE comparisons.
2953 2018-09-06  Ilya Leoshkevich  <iii@linux.ibm.com>
2955         PR target/80080
2956         * config/s390/predicates.md: Add nonsym_memory_operand.
2957         * config/s390/s390.c (s390_legitimize_cs_operand): If operand
2958         contains a SYMBOL_REF, load it into an intermediate pseudo.
2959         (s390_emit_compare_and_swap): Legitimize operand.
2960         * config/s390/s390.md: Use the new nonsym_memory_operand
2961         with UNSPECV_CAS patterns.
2963 2018-09-06  Ilya Leoshkevich  <iii@linux.ibm.com>
2965         PR target/80080
2966         * config/s390/s390-passes.def: New file.
2967         * config/s390/s390-protos.h (class rtl_opt_pass): Add forward
2968         declaration.
2969         (make_pass_s390_early_mach): Add declaration.
2970         * config/s390/s390.c (make_pass_s390_early_mach):
2971         (s390_option_override): Remove dynamic registration.
2972         * config/s390/t-s390: Add s390-passes.def.
2974 2018-09-06  Ilya Leoshkevich  <iii@linux.ibm.com>
2976         * config/s390/s390.c (s390_decompose_constant_pool_ref):
2977         Remove UNSPEC_LTREL_BASE check.
2978         (annotate_constant_pool_refs): Likewise.
2979         (find_constant_pool_ref): Likewise.
2980         (find_ltrel_base): Removed.
2981         (replace_ltrel_base): Removed.
2982         (s390_mainpool_finish): Remove replace_ltrel_base call.
2983         (s390_chunkify_start): Remove pending LTREL_BASE logic.
2984         (s390_chunkify_finish): Remove replace_ltrel_base call.
2985         * config/s390/s390.md: Remove UNSPEC_LTREL_BASE.
2987 2018-09-06  Hans-Peter Nilsson  <hp@axis.com>
2989         PR target/86779
2990         * config/cris/cris.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
2991         to speculation_safe_value_not_needed.
2993 2018-09-05  Cesar Philippidis  <cesar@codesourcery.com>
2994             Bernd Schmidt  <bernds_cb1@t-online.de>
2996         * config/nvptx/nvptx-opts.h: New file.
2997         * config/nvptx/nvptx.c (nvptx_file_start): Print the correct .target.
2998         * config/nvptx/nvptx.h: Include "nvptx-opts.h".
2999         (ASM_SPEC): Define.
3000         (TARGET_SM35): New macro.
3001         * config/nvptx/nvptx.md (atomic_fetch_<logic><mode>): Enable with the
3002         correct predicate.
3003         * config/nvptx/nvptx.opt (ptx_isa, sm_30, sm_35): New enum and its
3004         values.
3005         (misa=): New option.
3006         * doc/invoke.texi (Nvidia PTX Options): Document -misa.
3008 2018-09-05  Uros Bizjak  <ubizjak@gmail.com>
3010         * config/i386/i386.md (truncdfsf2): Remove expander.
3011         (truncdfsf2_with_temp): Ditto.
3012         (truncxf<mode>2): Ditto.
3013         (*truncdfsf_fast_mixed): Remove insn pattern.
3014         (*truncdfsf_fast_i387): Ditto.
3015         (*truncdfsf_mixed): Ditto.
3016         (*truncdfsf_i387): Ditto.
3017         (*truncdfsf2_i387_1): Ditto.
3018         (*truncxfsf2_mixed): Ditto.
3019         (*truncxfdf2_mixed): Ditto.
3020         (*truncxf<mode>2_i387_noop): Ditto. Update callers
3021         to call gen_truncxf<mode>2 instead.
3022         (*truncxf<mode>2_i387): Remove.
3023         (reg->reg splitters): Remove splitter pattern.
3024         (reg->mem splitters): Ditto.
3026         (truncdfsf2): New insn pattern.
3027         (truncxf<mode>2): Ditto.
3029 2018-09-05  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
3031         * tree-ssa-math-opts.c (is_mult_by): New function.
3032         (is_square_of): Use the above.
3033         (optimize_recip_sqrt): New function.
3034         (pass_cse_reciprocals::execute): Use the above.
3036 2018-09-05  Richard Biener  <rguenther@suse.de>
3038         PR bootstrap/87134
3039         * tree-ssa-sccvn.c (rpo_elim::eliminate_push_avail): Make sure
3040         to zero-init the emplaced vec.
3042 2018-09-05  Martin Liska  <mliska@suse.cz>
3044         PR tree-optimization/87205
3045         * tree-switch-conversion.c (pass_lower_switch::execute):
3046         Group cases for switch statements.
3048 2018-09-05  Richard Biener  <rguenther@suse.de>
3050         PR tree-optimization/87217
3051         * tree-ssa-sccvn.c (vuse_valueize): New.
3052         (vn_reference_lookup_pieces): Use it.
3053         (vn_reference_lookup): Likewise.
3055 2018-09-05  Nathan Sidwell  <nathan@acm.org>
3057         PR c++/87137
3058         * stor-layout.c (place_field): Scan forwards to check last
3059         bitfield when ms_bitfield_placement is in effect.
3061 2018-09-05  Richard Biener  <rguenther@suse.de>
3063         PR bootstrap/87225
3064         * tree-vect-stmts.c (vectorizable_simd_clone_call): Fix bogus
3065         return.
3067 2018-09-05  Siddhesh Poyarekar  <siddhesh@sourceware.org>
3068             Kugan Vivekanandarajah  <kugan.vivekanandarajah@linaro.org>
3070         * config/aarch64/falkor-tag-collision-avoidance.c: New file.
3071         * config.gcc (extra_objs): Build it.
3072         * config/aarch64/t-aarch64 (falkor-tag-collision-avoidance.o):
3073         Likewise.
3074         * config/aarch64/aarch64-passes.def
3075         (pass_tag_collision_avoidance): New pass.
3076         * config/aarch64/aarch64.c (qdf24xx_tunings): Add
3077         AARCH64_EXTRA_TUNE_RENAME_LOAD_REGS to tuning_flags.
3078         (aarch64_classify_address): Remove static qualifier.
3079         (aarch64_address_info, aarch64_address_type): Move to...
3080         * config/aarch64/aarch64-protos.h: ... here.
3081         (make_pass_tag_collision_avoidance): New function.
3082         * config/aarch64/aarch64-tuning-flags.def (rename_load_regs):
3083         New tuning flag.
3085 2018-09-05  Martin Liska  <mliska@suse.cz>
3087         * doc/gcov.texi: Update documentation of humar
3088         readable mode.
3089         * gcov.c (format_count): Print one decimal place, it provides
3090         more fine number of situations like '1G' vs. '1.4G'.
3092 2018-09-05  Martin Liska  <mliska@suse.cz>
3094         PR target/87164
3095         * config/rs6000/rs6000.opt: Mark the option as Deprecated.
3096         * optc-gen.awk: Allow 'Var' for Deprecated options in order
3097         to generate a MASK value.
3099 2018-09-04  H.J. Lu  <hongjiu.lu@intel.com>
3101         PR debug/86593
3102         * dwarf2out.c (based_loc_descr): Allow hard frame pointer even
3103         if frame pointer isn't used.
3104         (compute_frame_pointer_to_fb_displacement): Likewise.
3106 2018-09-04  Jakub Jelinek  <jakub@redhat.com>
3108         PR target/87198
3109         * common/config/i386/i386-common.c (OPTION_MASK_ISA_XSAVEOPT_SET,
3110         OPTION_MASK_ISA_XSAVES_SET, OPTION_MASK_ISA_XSAVEC_SET): Use
3111         OPTION_MASK_ISA_XSAVE_SET instead of OPTION_MASK_ISA_XSAVE.
3112         (OPTION_MASK_ISA_XSAVE_UNSET): Add OPTION_MASK_ISA_XSAVES_UNSET
3113         and OPTION_MASK_ISA_XSAVEC_UNSET.
3115 2018-09-04  Max Filippov  <jcmvbkbc@gmail.com>
3117         * config/xtensa/xtensa.c (xtensa_expand_atomic): Reorder AND and
3118         XOR operations in NAND case.
3120 2018-09-04  Aldy Hernandez  <aldyh@redhat.com>
3122         * wide-int-range.cc (wide_int_range_convert): New.
3123         * wide-int-range.h (wide_int_range_convert): New.
3124         * tree-vrp.c (extract_range_from_unary_expr): Abstract wide int
3125         code into wide_int_range_convert.
3126         (extract_range_into_wide_ints): Do not munge anti range constants
3127         into the entire domain.  Just return the range back.
3129 2018-09-04  Martin Liska  <mliska@suse.cz>
3131         * genmatch.c (output_line_directive): Add new argument
3132         fnargs.
3133         (dt_simplify::gen_1): Encapsulate dump within __builtin_expect.
3135 2018-09-04  Jonathan Wakely  <jwakely@redhat.com>
3137         * doc/invoke.texi (Option Summary): Add whitespace.
3139         * doc/invoke.texi (Option Summary): Add -Waligned-new.
3141 2018-09-04  Richard Biener  <rguenther@suse.de>
3143         PR tree-optimization/87211
3144         * tree-ssa-sccvn.c (visit_phi): When value-numbering to a
3145         backedge value we're supposed to treat as VARYING also number
3146         the PHI to VARYING in case it got a different value-number already.
3148 2018-09-04  Aldy Hernandez  <aldyh@redhat.com>
3150         * tree-vrp.c (vrp_can_optimize_bit_op): Remove.
3151         (extract_range_from_binary_expr_1): Do not call
3152         vrp_can_optimize_bit_op.
3153         * wide-int-range.cc (wide_int_range_can_optimize_bit_op): Make
3154         static.
3155         (wide_int_range_get_mask_and_bounds): New.
3156         (wide_int_range_optimize_bit_op): New.
3157         (wide_int_range_bit_ior): Call wide_int_range_optimize_bit_op.
3158         (wide_int_range_bit_and): Same.
3159         * wide-int-range.h (wide_int_range_can_optimize_bit_op): Remove.
3160         (wide_int_range_optimize_bit_op): New.
3161         (wide_int_range_get_mask_and_bounds): New.
3163 2018-09-04  Richard Biener  <rguenther@suse.de>
3165         PR tree-optimization/87176
3166         * tree-ssa-sccvn.c (visit_phi): Remove redundant allsame
3167         variable.  When value-numbering a virtual PHI node make sure
3168         to not value-number to the backedge value.
3170 2018-09-04  Jonathan Wakely  <jwakely@redhat.com>
3172         * doc/extend.texi (Long Long, Hex Floats): Document support for
3173         long long and hex floats in more recent versions of ISO C++.
3175 2018-09-03  Richard Biener  <rguenther@suse.de>
3177         PR tree-optimization/87177
3178         * tree-ssa-sccvn.c (vuse_ssa_val): Revert previous change, keep
3179         cleanup.
3181 2018-09-03  Alexander Monakov  <amonakov@ispras.ru>
3183         * bb-reorder.c (edge_order): Convert to C-qsort-style
3184         tri-state comparator.
3185         (reorder_basic_blocks_simple): Change std::stable_sort to
3186         gcc_stablesort.
3188 2018-09-03  Alexander Monakov  <amonakov@ispras.ru>
3190         * tree-loop-distribution.c (offset_cmp): Convert to C-qsort-style
3191         tri-state comparator.
3192         (fuse_memset_builtins): Change std::stable_sort to gcc_stablesort.
3194 2018-09-03  Alexander Monakov  <amonakov@ispras.ru>
3196         * sort.cc (struct sort_ctx): New field 'nlim'.  Use it...
3197         (mergesort): ... here as maximum count for using netsort.
3198         (gcc_qsort): Set nlim to 3 if stable sort is requested.
3199         (gcc_stablesort): New.
3200         * system.h (gcc_stablesort): Declare.
3202 2018-09-03  Alexander Monakov  <amonakov@ispras.ru>
3204         * sort.cc (gcc_qsort) [CHECKING_P]: Call qsort_chk.
3205         * system.h (qsort): Always redirect to gcc_qsort.  Update comment.
3206         * vec.c (qsort_chk): Do not call gcc_qsort.  Update comment.
3208 2018-09-03  Segher Boessenkool  <segher@kernel.crashing.org>
3210         * config/rs6000/rs6000.md (*mov<mode>_hardfloat32): Remove %U from the
3211         lxsdx and stxsdx alternatives.
3212         (*mov<mode>_hardfloat64): Ditto.
3213         * config/rs6000/vsx.md (*vsx_extract_<mode>_store): Ditto.
3215 2018-09-03  Richard Biener  <rguenther@suse.de>
3217         PR tree-optimization/87200
3218         * tree-ssa-sccvn.c (vn_nary_build_or_lookup_1): Valueize a
3219         simplify result.
3221 2018-09-03  Martin Liska  <mliska@suse.cz>
3223         PR tree-optimization/87201
3224         * tree-switch-conversion.c (switch_decision_tree::balance_case_nodes):
3225         Fix parenthesis in an expression.
3227 2018-09-03  Richard Biener  <rguenther@suse.de>
3229         PR tree-optimization/87197
3230         * tree-ssa-sccvn.c (vn_nary_build_or_lookup_1): Mark the new def
3231         visited.  CSE the VN_INFO hashtable lookup.
3233         PR tree-optimization/87169
3234         * tree-ssa-sccvn.c (do_rpo_vn): When marking loops for not
3235         iterating make sure there's no extra backedges from irreducible
3236         regions feeding the header.  Mark the destination block
3237         executable.
3239 2018-09-03  Martin Liska  <mliska@suse.cz>
3241         PR driver/83193
3242         * common/common-target.def: Add TARGET_GET_VALID_OPTION_VALUES.
3243         * common/common-targhooks.c (default_get_valid_option_values):
3244         New function.
3245         * common/common-targhooks.h (default_get_valid_option_values):
3246         Likewise.
3247         * common/config/i386/i386-common.c: Move processor_target_table
3248         from i386.c.
3249         (ix86_get_valid_option_values): New function.
3250         (TARGET_GET_VALID_OPTION_VALUES): New macro.
3251         * config/i386/i386.c (struct ptt): Move to i386-common.c.
3252         (PTA_*): Move all defined masks into i386-common.c.
3253         (ix86_function_specific_restore): Use new processor_cost_table.
3254         * config/i386/i386.h (struct ptt): Moved from i386.c.
3255         (struct pta): Likewise.
3256         * doc/tm.texi: Document new TARGET_GET_VALID_OPTION_VALUES.
3257         * doc/tm.texi.in: Likewise.
3258         * opt-suggestions.c (option_proposer::suggest_option):
3259         Pass prefix to build_option_suggestions.
3260         (option_proposer::get_completions): Likewise.
3261         (option_proposer::build_option_suggestions): Use the new target
3262         hook.
3263         * opts.c (struct option_help_tuple): New struct.
3264         (print_filtered_help): Use the new target hook.
3266 2018-09-03  Martin Liska  <mliska@suse.cz>
3268         PR middle-end/59521
3269         * predict.c (set_even_probabilities): Add likely_edges
3270         argument and handle cases where we have precisely one
3271         likely edge.
3272         (combine_predictions_for_bb): Catch also likely_edges.
3273         (tree_predict_by_opcode): Handle gswitch statements.
3274         * tree-cfg.h (find_case_label_for_value): New declaration.
3275         (find_taken_edge_switch_expr): Likewise.
3276         * tree-switch-conversion.c (switch_decision_tree::balance_case_nodes):
3277         Find pivot in decision tree based on probabily, not by number of
3278         nodes.
3280 2018-09-02  Gerald Pfeifer  <gerald@pfeifer.com>
3282         * doc/standards.texi (Standards): Update Objective-C reference.
3284 2018-09-01  Gerald Pfeifer  <gerald@pfeifer.com>
3286         * doc/install.texi (Prerequisites): Update link for MPC.
3288 2018-09-01  Michael Matz  <matz@suse.de>
3290         PR tree-optimization/87074
3291         * gimple-loop-jam.c (unroll_jam_possible_p): Check loop exit
3292         PHIs for outer-loop uses.
3294 2018-09-01  Gerald Pfeifer  <gerald@pfeifer.com>
3296         * doc/generic.texi (OpenMP): Adjust link to openmp.org.
3297         * doc/invoke.texi (C Dialect Options): Ditto.
3299 2018-09-01  Gerald Pfeifer  <gerald@pfeifer.com>
3301         * doc/install.texi (Prerequisites): Adjust link mpfr.org.
3303 2018-08-31  Richard Biener  <rguenther@suse.de>
3305         PR tree-optimization/87168
3306         * tree-ssa-sccvn.c (SSA_VAL): Add visited output parameter.
3307         (rpo_elim::eliminate_avail): When OP was not visited it must
3308         be available.
3310 2018-08-31  David Malcolm  <dmalcolm@redhat.com>
3312         * tree-vrp.c (copy_value_range): Convert param "from" from
3313         "value_range *" to "const value_range *".
3314         (range_is_null): Likewise for param "vr".
3315         (range_int_cst_p): Likewise.
3316         (range_int_cst_singleton_p): Likewise.
3317         (symbolic_range_p): Likewise.
3318         (value_ranges_intersect_p): Likewise for both params.
3319         (value_range_nonnegative_p): Likewise for param "vr".
3320         (value_range_constant_singleton): Likewise.
3321         (vrp_set_zero_nonzero_bits): Likewise for param "ar".
3322         (extract_range_into_wide_ints): Likewise for param "vr".
3323         (extract_range_from_multiplicative_op): Likewise for params "vr0"
3324         and "vr1".
3325         (vrp_can_optimize_bit_op): Likewise.
3326         (extract_range_from_binary_expr_1): Likewise for params "vr0_" and
3327         "vr1_".
3328         (extract_range_from_unary_expr): Likewise.
3329         (debug_value_range): Likewise for param "vr".
3330         (value_range::dump): Add "const" qualifier.
3331         (vrp_prop::check_array_ref): Convert local "vr" from
3332         "value_range *" to "const value_range *".
3333         (vrp_prop::check_mem_ref): Likewise.
3334         (vrp_prop::visit_stmt): Likewise for local "old_vr".
3335         (vrp_intersect_ranges_1): Likewise for param "vr_1".
3336         (vrp_intersect_ranges): Likewise.
3337         (simplify_stmt_for_jump_threading): Likewise for local "vr".
3338         (vrp_prop::vrp_finalize): Likewise.
3339         * tree-vrp.h (value_range::dump): Add "const" qualifier.
3340         (vrp_intersect_ranges): Add "const" qualifier to params as above.
3341         (extract_range_from_unary_expr): Likewise.
3342         (value_range_constant_singleton): Likewise.
3343         (symbolic_range_p): Likewise.
3344         (copy_value_range): Likewise.
3345         (extract_range_from_binary_expr_1): Likewise.
3346         (range_int_cst_p): Likewise.
3347         (vrp_set_zero_nonzero_bits): Likewise.
3348         (range_int_cst_singleton_p): Likewise.
3350 2018-08-31  Vlad Lazar  <vlad.lazar@arm.com>
3352         * config/aarch64/arm_neon.h (vabsd_s64): New.
3353         (vnegd_s64): Likewise.
3355 2018-08-31  Martin Jambor  <mjambor@suse.cz>
3357         * ipa-cp.c (estimate_local_effects): Replace wrong MAX with MIN.
3359 2018-08-31  Martin Liska  <mliska@suse.cz>
3361         * ipa-icf.c (sem_item::add_type): Use
3362         sem_item::m_type_hash_cache.
3363         * ipa-icf.h: Move the cache from sem_item_optimizer
3364         to sem_item.
3366 2018-08-31  Nathan Sidwell  <nathan@acm.org>
3368         * doc/extend.texi (Backwards Compatibility): Remove implicit
3369         extern C leeway of () being (...).
3371 2018-08-31  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
3373         * ipa-inline.c (can_inline_edge_by_limits_p): Fix typos in comment.
3375 2018-08-31  Segher Boessenkool  <segher@kernel.crashing.org>
3377         PR target/86684
3378         PR target/87149
3379         * config/rs6000/rs6000.md (lround<mode>di2): Gate on TARGET_FPRND.
3381 2018-08-31  Jakub Jelinek  <jakub@redhat.com>
3383         PR middle-end/87138
3384         * expmed.c (expand_mult_const): Use immed_wide_int_const instead of
3385         gen_int_mode.  Formatting fixes.
3387 2018-08-30  Sandra Loosemore  <sandra@codesourcery.com>
3389         * target.def (custom_function_descriptors): Improve documentation.
3390         * doc/tm.texi.in (Trampolines): Expand discussion of function
3391         descriptors and move TARGET_CUSTOM_FUNCTION_DESCRIPTORS to the
3392         beginning of the section.
3393         * doc/tm.texi: Regenerated.
3395 2018-08-30  Jose E. Marchesi  <jose.marchesi@oracle.com>
3397         * cfg.h (class auto_edge_flag): Spell out the template-id of the
3398         base class in the initializer list.  This is a workaround for
3399         building with older GCC.
3400         (class auto_bb_flag): Likewise.
3402 2018-08-30  Aaron Sawdey  <acsawdey@linux.ibm.com>
3404         * config/rs6000/altivec.md (altivec_eq<mode>): Remove star.
3405         (altivec_vcmpequ<VI_char>_p): Remove star.
3406         * config/rs6000/rs6000-string.c (do_load_for_compare): Support
3407         vector load modes.
3408         (expand_strncmp_vec_sequence): New function.
3409         (emit_final_str_compare_vec): New function.
3410         (expand_strn_compare): Add support for vector strncmp.
3411         * config/rs6000/rs6000.opt (-mstring-compare-inline-limit): Change
3412         length specification to bytes.
3413         * config/rs6000/vsx.md (vsx_ld_elemrev_v16qi_internal): Remove star.
3414         (vcmpnezb_p): New pattern.
3415         * doc/invoke.texi (RS/6000 and PowerPC Options): Update documentation
3416         for option -mstring-compare-inline-limit.
3418 2018-08-30  Thiago Macieira  <thiago.macieira@intel.com>
3420         * config/i386/i386.c (PTA_WESTMERE): Remove PTA_AES.
3421         (PTA_SKYLAKE): Add PTA_AES.
3422         (PTA_GOLDMONT): Likewise.
3424 2018-08-29  Jan Hubicka  <jh@suse.cz>
3426         PR lto/86517
3427         * lto-opts.c (lto_write_options): Always stream PIC/PIE mode.
3428         * lto-wrapper.c (merge_and_complain): Fix merging of PIC/PIE.
3430 2018-08-29  Jan Hubicka  <jh@suse.cz>
3432         * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not follow
3433         TYPE_STUB_DECL.
3434         (hash_tree): Do not visit TYPE_STUB_DECL.
3435         * tree-streamer-out.c (write_ts_type_common_tree_pointers): Do not
3436         stream TYPE_STUB_DECL.
3437         * tree-streamer-in.c (lto_input_ts_type_common_tree_pointers): Likewise.
3438         * ipa-utils.h (type_with_linkage_p): Do not rely on TYPE_STUB_DECL
3439         after free_lang_data.
3440         (type_in_anonymous_namespace_p): Likewise.
3442 2018-08-29  Jan Hubicka  <jh@suse.cz>
3444         * sreal.h (SREAL_PART_BITS): Change to 31; remove seemingly unnecessary
3445         comment that it has to be even number.
3446         (class sreal): Change m_sig type to int32_t.
3447         * sreal.c (sreal::dump, sreal::to_int, opreator+, operator-): Use
3448         int64_t for temporary calculations.
3449         (sreal_verify_basics): Drop one bit from minimum and maximum.
3451 2018-08-30  Richard Biener  <rguenther@suse.de>
3453         PR tree-optimization/87147
3454         * tree-ssa-sccvn.c (SSA_VISITED): New function.
3455         (visit_phi): When the degenerate result is from the backedge and
3456         we didn't visit its definition yet drop to VARYING.
3457         (do_rpo_vn): Properly mark blocks with incoming backedges as executable.
3459 2018-08-29  Jan Hubicka  <jh@suse.cz>
3461         * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not walk
3462         DECL_VINDEX.
3463         (hash_tree): Likewise.
3465 2018-08-29  Jan Hubicka  <jh@suse.cz>
3467         * tree.c (find_decls_types_r): Walk also TYPE_NEXT_PTR_TO
3468         and TYPE_NEXT_REF_TO.
3470 2018-08-29  Jan Hubicka  <jh@suse.cz>
3472         * sreal.h (SREAL_PART_BITS): Change to 31; remove seemingly unnecessary
3473         comment that it has to be even number.
3474         (class sreal): Change m_sig type to int32_t.
3475         * sreal.c (sreal::dump, sreal::to_int, opreator+, operator-): Use
3476         int64_t for temporary calculations.
3477         (sreal_verify_basics): Drop one bit from minimum and maximum.
3479 2018-08-30  Tamar Christina  <tamar.christina@arm.com>
3481         * config/aarch64/aarch64.c (aarch64_expand_movmem): Set TImode max.
3483 2018-08-30  Vlad Lazar  <vlad.lazar@arm.com>
3485         PR middle-end/86995
3486         * expmed.c (canonicalize_comparison): Use wi::sub instead of wi::add
3487         if to_add is negative.
3489 2018-08-29  Bernd Edlinger  <bernd.edlinger@hotmail.de>
3491         PR middle-end/87053
3492         * builtins.c (c_strlen): Improve range checks.
3494 2018-08-29  Martin Sebor  <msebor@redhat.com>
3495             Jeff Law  <law@redhat.com>
3497         PR tree-optimization/86714
3498         PR tree-optimization/86711
3499         * builtins.c (c_strlen): Add arguments to call to string_constant.
3500         * expr.c (string_constant): Add argument.  Detect missing nul
3501         terminator and outermost declaration it's missing in.
3502         * expr.h (string_constant): Add argument.
3503         * fold-const.c (read_from_constant_string): Add arguments to call to
3504         string_constant.
3505         (c_getstr): Likewise.
3506         * tree-ssa-forwprop.c (simplify_builtin_call): Likewise.
3507         to string_constant.
3508         * tree-ssa-strlen.c (get_stridx): Likewise.
3510 2018-08-29  Jan Hubicka  <jh@suse.cz>
3512         * tree-streamer-in.c (lto_input_ts_function_decl_tree_pointers):
3513         Do not stream DECL_VINDEX.
3514         * tree-streamer-out.c (write_ts_function_decl_tree_pointers): Likewise.
3515         * tree.c (free_lang_data_in_decl): Clear DECL_VINDEX.
3516         (decl_function_context): Use DECL_VIRTUAL_P rather than DECL_VINDEX.
3518 2018-08-29  Richard Biener  <rguenther@suse.de>
3520         * tree-ssa-sccvn.c (vuse_ssa_val): Return NULL for unvisited
3521         virtual operands that are not default defs to honor region
3522         boundaries.
3523         (rpo_vn_valueize): Remove ineffective code here.
3525 2018-08-29  Richard Biener  <rguenther@suse.de>
3527         PR tree-optimization/87132
3528         * tree-ssa-alias.c (get_continuation_for_phi): Do not translate
3529         when skipping defs reachable over backedges.
3531 2018-08-29  Richard Biener  <rguenther@suse.de>
3533         * tree-core.h: Document use of deprecated_flag in SSA_NAME.
3534         * tree.h (SSA_NAME_POINTS_TO_READONLY_MEMORY): Define.
3535         * tree-into-ssa.c (pass_build_ssa::execute): Initialize
3536         function parameters SSA_NAME_POINTS_TO_READONLY_MEMORY from fnspec.
3537         * tree-ssa-sccvn.c (const_parms, init_const_parms): Remove.
3538         (vn_reference_lookup_3): Remove use of const_parms.
3539         (free_rpo_vn): Do not free const_parms.
3540         (do_rpo_vn): Do not call init_const_parms.
3541         * tree-ssa-alias.c (refs_may_alias_p_1): Honor
3542         SSA_NAME_POINTS_TO_READONLY_MEMORY.
3543         (call_may_clobber_ref_p_1): Likewise.
3545 2018-08-29  Alexander Monakov  <amonakov@ispras.ru>
3547         PR other/86726
3548         * invoke.texi (Optimization Options): List -ftree-scev-cprop.
3549         (-O): Ditto.
3550         (-ftree-scev-cprop): Document.
3552 2018-08-29  Jan Hubicka  <jh@suse.cz>
3554         * sreal.h (normalize, normalize_up, normalize_down): Add new_sig/new_exp
3555         parameters.
3556         (sreal constructor): Update.
3557         * sreal.c (sreal:operator+, sreal:operator-, sreal:operator*,
3558         sreal:operator/): Update.
3560 2018-08-29  Martin Liska  <mliska@suse.cz>
3562         * tree-switch-conversion.c (switch_conversion::expand):
3563         Strenghten assumption about gswitch statements.
3565 2018-08-29  Richard Biener  <rguenther@suse.de>
3567         PR tree-optimization/87117
3568         * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_stmt): Only
3569         re-value-number released SSA VDEFs.
3571 2018-08-29  Richard Biener  <rguenther@suse.de>
3573         PR tree-optimization/87126
3574         * tree-ssa-sccvn.c (vn_reference_insert): Remove assert.
3576 2018-08-28  Jim Wilson  <jimw@sifive.com>
3578         * config/riscv/pic.md: Rewrite.
3579         * config/riscv/riscv.c (riscv_address_insns): Return cost of 3 for
3580         invalid address.
3581         * config/riscv/riscv.md (ZERO_EXTEND_LOAD): Delete.
3582         (SOFTF, default_load, softload, softstore): New.
3584 2018-08-28  Jeff Law  <law@redhat.com>
3586         * fold-const.c (fold_binary_loc): Remove recently added assert.
3588 2018-08-28  Joern Rennecke  <joern.rennecke@riscy-ip.com>
3590         * genpreds.c (write_predicate_subfunction): Also add ATTRIBUTE_UNUSED
3591         to OP parmeter of generated function.
3593 2018-08-28  MCC CS  <deswurstes@users.noreply.github.com>
3595         PR tree-optimization/87009
3596         * match.pd: Add boolean optimizations.
3598 2018-08-28  Martin Sebor  <msebor@redhat.com>
3600         PR middle-end/86631
3601         * calls.c (alloc_max_size): Treat HOST_WIDE_INT special.
3602         * gimple-ssa-warn-alloca.c (adjusted_warn_limit): New function.
3603         (pass_walloca::gate): Use it.
3604         (alloca_call_type): Same.
3605         (pass_walloca::execute): Same.
3606         * stor-layout.c (layout_decl): Treat HOST_WIDE_INT special.
3608 2018-08-28  David Malcolm  <dmalcolm@redhat.com>
3610         * dumpfile.h (ATTRIBUTE_GCC_DUMP_PRINTF): Change version check on
3611         GCC_VERSION for usage of "__gcc_dump_printf__" format from
3612         >= 3005 to >= 9000.
3614 2018-08-28  Richard Biener  <rguenther@suse.de>
3616         PR tree-optimization/87124
3617         * tree-ssa-sccvn.c (vn_lookup_simplify_result): Guard against
3618         constants before looking up avail.
3620 2018-08-28  Jakub Jelinek  <jakub@redhat.com>
3622         PR middle-end/87099
3623         * calls.c (maybe_warn_nonstring_arg): Punt early if
3624         warn_stringop_overflow is zero.  Don't call get_range_strlen
3625         on 3rd argument, keep iterating until lenrng[1] is INTEGER_CST.
3626         Swap comparison operands to have constants on rhs.  Only use
3627         lenrng[1] if non-NULL and INTEGER_CST.  Don't uselessly
3628         increment lenrng[0].
3630 2018-08-28  Richard Sandiford  <richard.sandiford@arm.com>
3632         * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Fix unguarded
3633         use of tree_to_shwi.  Remove duplicated test for the size being
3634         a whole number of bytes.
3636 2018-08-28  Richard Biener  <rguenther@suse.de>
3638         PR tree-optimization/87117
3639         * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_cleanup):
3640         Handle removed stmt without LHS (GIMPLE_NOP).
3642 2018-08-28  Richard Biener  <rguenther@suse.de>
3644         PR tree-optimization/87117
3645         * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Exclude
3646         void which is is_gimple_reg_type by checking for COMPLETE_TYPE_P.
3648 2018-08-28  Richard Biener  <rguenther@suse.de>
3650         PR tree-optimization/87117
3651         * tree-ssa-pre.c (compute_avail): Do not make expressions
3652         with predicated values available.
3653         (get_expr_value_id): Assert we do not run into predicated value
3654         expressions.
3656 2018-08-28  Richard Biener  <rguenther@suse.de>
3658         PR tree-optimization/87117
3659         * tree-ssa-operands.c (add_stmt_operand): STRING_CST may
3660         get virtual operands.
3661         (get_expr_operands): Handle STRING_CST like other decls.
3663 2018-08-28  Martin Liska  <mliska@suse.cz>
3665         * tree.h: Update documentation of fndecl_built_in_p
3666         functions.
3669 2018-08-27  Jeff Law  <law@redhat.com>
3670         PR tree-optimization/87110
3671         * tree-ssa-dse.c (compute_trims): Handle non-constant
3672         TYPE_SIZE_UNIT.
3674 2018-08-27  Martin Sebor  <msebor@redhat.com>
3676         PR tree-optimization/86914
3677         * tree-ssa-strlen.c (maybe_set_strlen_range): Avoid MEM_REF.
3679 2018-08-27  Martin Sebor  <msebor@redhat.com>
3681         PR tree-optimization/87112
3682         * builtins.c (expand_builtin_strnlen): Convert c_strlen result to
3683         the type of the bound argument.
3685 2018-08-27  Jeff Law  <law@redhat.com>
3687         * tree-ssa-dse.c (compute_trims): Handle case where the reference's
3688         type does not have a TYPE_SIZE_UNIT.
3690 2018-08-27  Steve Ellcey  <sellcey@cavium.com>
3692         * config/aarch64/aarch64-speculation.cc: Replace include of cfg.h
3693         with include of backend.h.
3695 2018-08-27  Richard Biener  <rguenther@suse.de>
3697         PR tree-optimization/86927
3698         * tree-vect-loop.c (vect_create_epilog_for_reduction): Properly
3699         use const cond reduction code.
3701 2018-08-27  Alexander Monakov  <amonakov@ispras.ru>
3703         PR tree-optimization/85758
3704         * match.pd ((X & Y) ^ Y): Add :s qualifier to inner expression.
3706 2018-08-27  David Malcolm  <dmalcolm@redhat.com>
3708         PR c++/87091
3709         * diagnostic-show-locus.c (class layout_range): Update for
3710         conversion of show_caret_p to a tri-state.
3711         (layout_range::layout_range): Likewise.
3712         (make_range): Likewise.
3713         (layout::maybe_add_location_range): Likewise.
3714         (layout::should_print_annotation_line_p): Don't show annotation
3715         lines for ranges that are SHOW_LINES_WITHOUT_RANGE.
3716         (layout::get_state_at_point): Update for conversion of
3717         show_caret_p to a tri-state.  Bail out early for
3718         SHOW_LINES_WITHOUT_RANGE, so that such ranges don't affect
3719         underlining or source colorization.
3720         (gcc_rich_location::add_location_if_nearby): Update for conversion
3721         of show_caret_p to a tri-state.
3722         (selftest::test_one_liner_multiple_carets_and_ranges): Likewise.
3723         (selftest::test_one_liner_fixit_replace_equal_secondary_range):
3724         Likewise.
3725         (selftest::test_one_liner_labels): Likewise.
3726         * gcc-rich-location.c (gcc_rich_location::add_expr): Update for
3727         conversion of show_caret_p to a tri-state.
3728         * pretty-print.c (text_info::set_location): Likewise.
3729         * pretty-print.h (text_info::set_location): Likewise.
3730         * substring-locations.c (format_warning_n_va): Likewise.
3731         * tree-diagnostic.c (default_tree_printer): Likewise.
3732         * tree-pretty-print.c (newline_and_indent): Likewise.
3734 2018-08-27  David Malcolm  <dmalcolm@redhat.com>
3736         PR c++/87091
3737         * diagnostic-show-locus.c (get_line_span_for_fixit_hint): Show the
3738         line above for line-insertion fix-it hints.
3739         (selftest::test_fixit_insert_containing_newline): Update the
3740         expected results, and add a test with line-numbering enabled.
3742 2018-08-27  Martin Liska  <mliska@suse.cz>
3744         PR sanitizer/86962
3745         * sanopt.c (sanitize_rewrite_addressable_params): Ignore
3746         params with DECL_HAS_VALUE_EXPR_P.
3748 2018-08-27  Martin Liska  <mliska@suse.cz>
3750         * config/i386/i386.c (ix86_expand_set_or_movmem): Dump
3751         selected expansion strategy.
3753 2018-08-27  Martin Liska  <mliska@suse.cz>
3755         * builtins.h (is_builtin_fn): Remove and fndecl_built_in_p.
3756         * builtins.c (is_builtin_fn): Likewise.
3757         * attribs.c (diag_attr_exclusions): Use new function
3758         fndecl_built_in_p and remove check for FUNCTION_DECL if
3759         possible.
3760         (builtin_mathfn_code): Likewise.
3761         (fold_builtin_expect): Likewise.
3762         (fold_call_expr): Likewise.
3763         (fold_builtin_call_array): Likewise.
3764         (fold_call_stmt): Likewise.
3765         (set_builtin_user_assembler_name): Likewise.
3766         (is_simple_builtin): Likewise.
3767         * calls.c (gimple_alloca_call_p): Likewise.
3768         (maybe_warn_nonstring_arg): Likewise.
3769         * cfgexpand.c (expand_call_stmt): Likewise.
3770         * cgraph.c (cgraph_update_edges_for_call_stmt_node): Likewise.
3771         (cgraph_edge::verify_corresponds_to_fndecl): Likewise.
3772         (cgraph_node::verify_node): Likewise.
3773         * cgraphclones.c (build_function_decl_skip_args): Likewise.
3774         (cgraph_node::create_clone): Likewise.
3775         * config/arm/arm.c (arm_insert_attributes): Likewise.
3776         * config/i386/i386.c (ix86_gimple_fold_builtin): Likewise.
3777         * dse.c (scan_insn): Likewise.
3778         * expr.c (expand_expr_real_1): Likewise.
3779         * fold-const.c (operand_equal_p): Likewise.
3780         (fold_binary_loc): Likewise.
3781         * gimple-fold.c (gimple_fold_stmt_to_constant_1): Likewise.
3782         * gimple-low.c (lower_stmt): Likewise.
3783         * gimple-pretty-print.c (dump_gimple_call): Likewise.
3784         * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
3785         Likewise.
3786         * gimple.c (gimple_build_call_from_tree): Likewise.
3787         (gimple_call_builtin_p): Likewise.
3788         (gimple_call_combined_fn): Likewise.
3789         * gimplify.c (gimplify_call_expr): Likewise.
3790         (gimple_boolify): Likewise.
3791         (gimplify_modify_expr): Likewise.
3792         (gimplify_addr_expr): Likewise.
3793         * hsa-gen.c (gen_hsa_insns_for_call): Likewise.
3794         * ipa-cp.c (determine_versionability): Likewise.
3795         * ipa-fnsummary.c (compute_fn_summary): Likewise.
3796         * ipa-param-manipulation.c (ipa_modify_formal_parameters): Likewise.
3797         * ipa-split.c (visit_bb): Likewise.
3798         (split_function): Likewise.
3799         * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
3800         * lto-cgraph.c (input_node): Likewise.
3801         * lto-streamer-out.c (write_symbol): Likewise.
3802         * omp-low.c (setjmp_or_longjmp_p): Likewise.
3803         (lower_omp_1): Likewise.
3804         * predict.c (strip_predict_hints): Likewise.
3805         * print-tree.c (print_node): Likewise.
3806         * symtab.c (symtab_node::output_to_lto_symbol_table_p): Likewise.
3807         * trans-mem.c (is_tm_irrevocable): Likewise.
3808         (is_tm_load): Likewise.
3809         (is_tm_simple_load): Likewise.
3810         (is_tm_store): Likewise.
3811         (is_tm_simple_store): Likewise.
3812         (is_tm_abort): Likewise.
3813         (tm_region_init_1): Likewise.
3814         * tree-call-cdce.c (gen_shrink_wrap_conditions): Likewise.
3815         * tree-cfg.c (verify_gimple_call): Likewise.
3816         (move_stmt_r): Likewise.
3817         (stmt_can_terminate_bb_p): Likewise.
3818         * tree-eh.c (lower_eh_constructs_2): Likewise.
3819         * tree-if-conv.c (if_convertible_stmt_p): Likewise.
3820         * tree-inline.c (remap_gimple_stmt): Likewise.
3821         (copy_bb): Likewise.
3822         (estimate_num_insns): Likewise.
3823         (fold_marked_statements): Likewise.
3824         * tree-sra.c (scan_function): Likewise.
3825         * tree-ssa-ccp.c (surely_varying_stmt_p): Likewise.
3826         (optimize_stack_restore): Likewise.
3827         (pass_fold_builtins::execute): Likewise.
3828         * tree-ssa-dce.c (mark_stmt_if_obviously_necessary): Likewise.
3829         (mark_all_reaching_defs_necessary_1): Likewise.
3830         * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Likewise.
3831         * tree-ssa-forwprop.c (simplify_builtin_call): Likewise.
3832         (pass_forwprop::execute): Likewise.
3833         * tree-ssa-loop-im.c (stmt_cost): Likewise.
3834         * tree-ssa-math-opts.c (pass_cse_reciprocals::execute): Likewise.
3835         * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Likewise.
3836         * tree-ssa-strlen.c (get_string_length): Likewise.
3837         * tree-ssa-structalias.c (handle_lhs_call): Likewise.
3838         (find_func_aliases_for_call): Likewise.
3839         * tree-ssa-ter.c (find_replaceable_in_bb): Likewise.
3840         * tree-stdarg.c (optimize_va_list_gpr_fpr_size): Likewise.
3841         * tree-tailcall.c (find_tail_calls): Likewise.
3842         * tree.c (need_assembler_name_p): Likewise.
3843         (free_lang_data_in_decl): Likewise.
3844         (get_call_combined_fn): Likewise.
3845         * ubsan.c (is_ubsan_builtin_p): Likewise.
3846         * varasm.c (incorporeal_function_p): Likewise.
3847         * tree.h (DECL_BUILT_IN): Remove and replace with
3848         fndecl_built_in_p.
3849         (DECL_BUILT_IN_P): Transfort to fndecl_built_in_p.
3850         (fndecl_built_in_p): New.
3852 2018-08-27  Martin Liska  <mliska@suse.cz>
3854         PR tree-optimization/86847
3855         * tree-switch-conversion.c (switch_decision_tree::dump_case_nodes):
3856         Dump also subtree probability.
3857         (switch_decision_tree::do_jump_if_equal): New function.
3858         (switch_decision_tree::emit_case_nodes): Handle special
3859         situations in balanced tree that can be emitted much simpler.
3860         Fix calculation of probabilities that happen in tree expansion.
3861         * tree-switch-conversion.h (struct cluster): Add
3862         is_single_value_p.
3863         (struct simple_cluster): Likewise.
3864         (struct case_tree_node): Add new function has_child.
3865         (do_jump_if_equal): New.
3867 2018-08-27  Martin Liska  <mliska@suse.cz>
3869         * tree-switch-conversion.c (bit_test_cluster::find_bit_tests):
3870         Add new argument to bit_test_cluster constructor.
3871         (bit_test_cluster::emit): Set bits really number of values
3872         handlel by a test.
3873         (bit_test_cluster::hoist_edge_and_branch_if_true): Add
3874         probability argument.
3875         * tree-switch-conversion.h (struct bit_test_cluster):
3876         Add m_handles_entire_switch.
3878 2018-08-27  Martin Liska  <mliska@suse.cz>
3880         PR tree-optimization/86702
3881         * tree-switch-conversion.c (jump_table_cluster::emit):
3882         Make probabilities even for values in jump table
3883         according to number of cases handled.
3884         (switch_decision_tree::compute_cases_per_edge): Pass
3885         argument to reset_out_edges_aux function.
3886         (switch_decision_tree::analyze_switch_statement): Likewise.
3887         * tree-switch-conversion.h (switch_decision_tree::reset_out_edges_aux):
3888         Make it static.
3890 2018-08-27  Martin Liska  <mliska@suse.cz>
3892         * cfgexpand.c (expand_asm_stmt): Use label_to_block and pass
3893         cfun argument explicitly.
3894         * gimple-pretty-print.c (dump_gimple_switch): Likewise.
3895         * hsa-gen.c (gen_hsa_insns_for_switch_stmt): Use new
3896         function gimple_switch_default_bb.
3897         (convert_switch_statements):
3898         (expand_builtins):
3899         * ipa-fnsummary.c (set_switch_stmt_execution_predicate):
3900         * stmt.c (label_to_block_fn): Use label_to_block and pass
3901         cfun argument explicitly and use gimple_switch_label_bb.
3902         (expand_case): Likewise.
3903         * tree-cfg.c (lower_phi_internal_fn): Use label_to_block and pass
3904         cfun argument explicitly. Likewise.
3905         (make_edges_bb): Likewise.
3906         (make_cond_expr_edges): Likewise.
3907         (get_cases_for_edge): Likewise.
3908         (make_gimple_switch_edges): Likewise.
3909         (label_to_block_fn): Likewise.
3910         (label_to_block): Likewise.
3911         (make_goto_expr_edges): Likewise.
3912         (make_gimple_asm_edges): Likewise.
3913         (main_block_label): Likewise.
3914         (group_case_labels_stmt): Likewise.
3915         (find_taken_edge_computed_goto): Likewise.
3916         (find_taken_edge_switch_expr): Likewise.
3917         (gimple_verify_flow_info): Likewise.
3918         (gimple_redirect_edge_and_branch): Likewise.
3919         (gimple_switch_label_bb): New function.
3920         (gimple_switch_default_bb): Likewise.
3921         (gimple_switch_edge): Likewise.
3922         (gimple_switch_default_edge): Likewise.
3923         * tree-cfg.h (label_to_block_fn): Remove and replace ...
3924         (label_to_block): ... with this.
3925         (gimple_switch_label_bb): New.
3926         (gimple_switch_default_bb): Likewise.
3927         (gimple_switch_edge): Likewise.
3928         (gimple_switch_default_edge): Likewise.
3929         * tree-cfgcleanup.c (convert_single_case_switch): Use
3930         new gimple functions and pass new argument to label_to_block.
3931         (cleanup_control_flow_bb):
3932         * tree-eh.c (make_eh_dispatch_edges): Use label_to_block and pass
3933         cfun argument explicitly.
3934         (make_eh_edges): Likewise.
3935         (redirect_eh_dispatch_edge): Likewise.
3936         (lower_resx): Likewise.
3937         (lower_eh_dispatch): Likewise.
3938         (maybe_remove_unreachable_handlers): Likewise.
3939         (unsplit_eh): Likewise.
3940         (cleanup_empty_eh): Likewise.
3941         (verify_eh_edges): Likewise.
3942         (verify_eh_dispatch_edge): Likewise.
3943         * tree-ssa-dom.c (record_edge_info): Likewise.
3944         * tree-ssa-forwprop.c (simplify_gimple_switch_label_vec): Likewise.
3945         * tree-ssa-threadedge.c (thread_around_empty_blocks): Likewise.
3946         (thread_through_normal_block): Likewise.
3947         * tree-ssa-uncprop.c (associate_equivalences_with_edges): Likewise.
3948         * tree-ssa-uninit.c (convert_control_dep_chain_into_preds):
3949         * tree-switch-conversion.c (switch_conversion::collect): Use new
3950         gimple functions.
3951         (switch_conversion::check_final_bb): Likewise.
3952         (switch_conversion::gather_default_values): Pass new argument
3953         to label_to_block.
3954         (switch_conversion::build_constructors): Likewise.
3955         (switch_decision_tree::compute_cases_per_edge): Use new
3956         gimple_switch_edge function.
3957         (switch_decision_tree::analyze_switch_statement): Pass new argument
3958         to label_to_block.
3959         (switch_decision_tree::try_switch_expansion): Use
3960         gimple_switch_default_edge.
3961         * tree-vrp.c (find_switch_asserts): Pass new argument
3962         to label_to_block.
3963         * vr-values.c (vr_values::vrp_visit_switch_stmt): Likewise.
3964         (vr_values::simplify_switch_using_ranges): Likewise.
3966 2018-08-27  Richard Biener  <rguenther@suse.de>
3968         * cfganal.h (rev_post_order_and_mark_dfs_back_seme): Declare.
3969         * cfganal.c (rev_post_order_and_mark_dfs_back_seme): New function.
3971         * tree-ssa-sccvn.h (struct vn_pval): New structure.
3972         (struct vn_nary_op_s): Add unwind_to member.  Add
3973         predicated_values flag and put result into a union together
3974         with a linked list of vn_pval.
3975         (struct vn_ssa_aux): Add name member to make maintaining
3976         a map of SSA name to vn_ssa_aux possible.  Remove no longer
3977         needed info, dfsnum, low, visited, on_sccstack, use_processed
3978         and range_info_anti_range_p members.
3979         (run_scc_vn, vn_eliminate, free_scc_vn, vn_valueize): Remove.
3980         (do_rpo_vn, run_rpo_vn, eliminate_with_rpo_vn, free_rpo_vn):
3981         New functions.
3982         (vn_valueize): New global.
3983         (vn_context_bb): Likewise.
3984         (VN_INFO_RANGE_INFO, VN_INFO_ANTI_RANGE_P, VN_INFO_RANGE_TYPE,
3985         VN_INFO_PTR_INFO): Remove.
3986         * tree-ssa-sccvn.c: ... (rewrite)
3987         (pass_fre::execute): For -O2+ initialize loops and run
3988         RPO VN in optimistic mode (iterating).  For -O1 and -Og
3989         run RPO VN in non-optimistic mode.
3990         * params.def (PARAM_SCCVN_MAX_SCC_SIZE): Remove.
3991         (PARAM_RPO_VN_MAX_LOOP_DEPTH): Add.
3992         * doc/invoke.texi (sccvn-max-scc-size): Remove.
3993         (rpo-vn-max-loop-depth): Document.
3994         * tree-ssa-alias.c (walk_non_aliased_vuses): Stop walking
3995         when valuezing the VUSE signals we walked out of the region.
3996         * tree-ssa-pre.c (phi_translate_1): Ignore predicated values.
3997         (phi_translate): Set VN context block to use for availability
3998         lookup.
3999         (compute_avail): Likewise.
4000         (pre_valueize): New function.
4001         (pass_pre::execute): Adjust to the RPO VN API.
4003         * tree-ssa-loop-ivcanon.c: Include tree-ssa-sccvn.h.
4004         (propagate_constants_for_unrolling): Remove.
4005         (tree_unroll_loops_completely): Perform value-numbering
4006         on the unrolled bodies loop parent.
4008 2018-08-27  Richard Biener  <rguenther@suse.de>
4010         * tree-ssa-pre.c (compute_antic): Re-use inverted postorder
4011         for partial antic compute.
4013 2018-08-27  Jakub Jelinek  <jakub@redhat.com>
4015         PR rtl-optimization/87065
4016         * combine.c (simplify_if_then_else): Formatting fix.
4017         (if_then_else_cond): Guard MULT optimization with SCALAR_INT_MODE_P
4018         check.
4019         (known_cond): Don't return const_true_rtx for vector modes.  Use
4020         CONST0_RTX instead of const0_rtx.  Formatting fixes.
4022 2018-08-27  Martin Liska  <mliska@suse.cz>
4024         PR gcov-profile/87069
4025         * gcov.c (process_file): Record files already processed
4026         and warn about a file being processed multiple times.
4028 2018-08-27  Martin Liska  <mliska@suse.cz>
4030         PR driver/83193
4031         * config/aarch64/aarch64.c (aarch64_override_options_internal):
4032         Set default values for x_aarch64_*_string strings.
4033         * config/aarch64/aarch64.opt: Remove --{march,mcpu,mtune}==
4034         prefix.  For -mabi do not print '=ABI' in help and use
4035         <option_value> format for -msve-vector-bits and -moverride
4036         options.
4038 2018-08-26  Jeff Law  <law@redhat.com>
4040         * config/mips/frame-header-opt.c: Include "backend.h" rather than
4041         "cfg.h"
4043 2018-08-26  Marek Polacek  <polacek@redhat.com>
4045         PR c++/87029, Implement -Wredundant-move.
4046         * doc/invoke.texi: Document -Wredundant-move.
4048 2018-08-25  Martin Sebor  <msebor@redhat.com>
4050         PR tree-optimization/87059
4051         * builtins.c (expand_builtin_strncmp): Convert MIN_EXPR operand
4052         to the same type as the other.
4053         * fold-const.c (fold_binary_loc): Assert expectation.
4055 2018-08-25  Iain Sandoe  <iain@sandoe.co.uk>
4057         * config/darwin.c (machopic_legitimize_pic_address): Clean up
4058         extraneous parentheses, dead code section and formatting.
4060 2018-08-24  David Malcolm  <dmalcolm@redhat.com>
4062         PR c++/87091
4063         * diagnostic-show-locus.c (layout::layout): Ensure the margin is
4064         wide enough for jumps in the line-numbering to be visible.
4065         (layout::print_gap_in_line_numbering): New member function.
4066         (layout::calculate_line_spans): When using line numbering, merge
4067         line spans that are only 1 line apart.
4068         (diagnostic_show_locus): When printing line numbers, show gaps in
4069         line numbering directly, rather than printing headers.
4070         (selftest::test_diagnostic_show_locus_fixit_lines): Add test of
4071         line-numbering with multiple line spans.
4072         (selftest::test_fixit_insert_containing_newline_2): Add test of
4073         line-numbering, in which the spans are close enough to be merged.
4075 2018-08-24  Aldy Hernandez  <aldyh@redhat.com>
4077         * gimple-ssa-evrp-analyze.c (set_ssa_range_info): Pass value_range
4078         to range_includes_zero_p.  Do not special case VR_ANTI_RANGE.
4079         * tree-vrp.c (range_is_nonnull): Remove.
4080         (range_includes_zero_p): Accept value_range instead of min/max.
4081         (extract_range_from_binary_expr_1): Do not early bail on
4082         POINTER_PLUS_EXPR.
4083         Use range_includes_zero_p instead of range_is_nonnull.
4084         (extract_range_from_unary_expr): Use range_includes_zero_p instead
4085         of range_is_nonnull.
4086         (vrp_meet_1): Pass value_range to range_includes_zero_p.  Do not
4087         special case VR_ANTI_RANGE.
4088         (vrp_finalize): Same.
4089         * tree-vrp.h (range_includes_zero_p): Pass value_range as argument
4090         instead of min/max.
4091         (range_is_nonnull): Remove.
4092         * vr-values.c (vrp_stmt_computes_nonzero): Use
4093         range_includes_zero_p instead of range_is_nonnull.
4094         (extract_range_basic): Pass value_range to range_includes_zero_p
4095         instead of range_is_nonnull.
4097 2018-08-24  Uros Bizjak  <ubizjak@gmail.com>
4099         * emit-rtl.c (init_emit_once): Do not emit MODE_POINTER_BOUNDS RTXes.
4100         * emit-rtl.h (rtl_data): Remove return_bnd.
4101         * explow.c (trunc_int_for_mode): Do not handle POINTER_BOUNDS_MODE_P.
4102         * function.c (diddle_return_value): Do not handle crtl->return_bnd.
4103         * genmodes.c (complete_mode): Do not handle MODE_POINTER_BOUNDS.
4104         (POINTER_BOUNDS_MODE): Remove definition.
4105         (make_pointer_bounds_mode): Remove.
4106         (get_mode_class): Do not handle MODE_POINTER_BOUNDS.
4107         * machmode.h (POINTER_BOUNDS_MODE_P): Remove definition.
4108         (scalare_mode::includes_p): Do not handle MODE_POINTER_BOUNDS.
4109         * mode-classes.def: Do not define MODE_POINTER_BOUNDS.
4110         * stor-layout.c (int_mode_for_mode): Do not handle MODE_POINTER_BOUNDS.
4111         * tree-core.h (enum tree_index): Remove TI_POINTER_BOUNDS_TYPE.
4112         * varasm.c (output_constant_pool_2): Do not handle MODE_POINTER_BOUNDS.
4114         * config/i386/i386-modes.def (BND32, BND64): Remove.
4115         * config/i386/i386.c (dbx_register_map): Remove bound registers.
4116         (dbx64_register_map): Ditto.
4117         (svr4_dbx_register_map): Ditto.
4118         (indirect_thunk_bnd_needed): Remove.
4119         (indirect_thunks_bnd_used): Ditto.
4120         (indirect_return_bnd_needed): Ditto.
4121         (indirect_return_via_cx_bnd): Ditto.
4122         (enum indirect_thunk_prefix): Remove indirect_thunk_prefix_bnd.
4123         (indirect_thunk_name): Remove handling of indirect_thunk_prefix_bnd.
4124         (output_indirect_thunk): Ditto.  Remove need_prefix argument.
4125         (output_indirect_thunk_function): Remove handling of
4126         indirect_return_bnd_needed, indirect_return_via_cx_bnd,
4127         indirect_thunk_bnd_needed and indirect_thunks_bnd_used variables.
4128         (ix86_save_reg): Remove handling of crtl->return_bnd.
4129         (ix86_legitimate_constant_p): Remove handling of POINTER_BOUNDS_MODE_P.
4130         (ix86_print_operand_address_as): Remove handling of UNSPEC_BNDMK_ADDR
4131         and UNSPEC_BNDLX_ADDR.
4132         (ix86_output_indirect_branch_via_reg): Remove handling of
4133         indirect_thunk_prefix_bnd.
4134         (ix86_output_indirect_branch_via_push): Ditto.
4135         (ix86_output_function_return): Ditto.
4136         (ix86_output_indirect_function_return): Ditto.
4137         (avoid_func_arg_motion): Do not handle UNSPEC_BNDSTX.
4138         * config/i386/i386.h (FIXED_REGISTERS): Remove bound registers.
4139         (CALL_USED_REGISTERS): Ditto.
4140         (REG_ALLOC_ORDER): Update for removal of bound registers.
4141         (HI_REGISTER_NAMES): Ditto.
4142         * config/i386/i386.md (UNSPEC_BNDMK, UNSPEC_BNDMK_ADDR, UNSPEC_BNDSTX)
4143         (UNSPEC_BNDLDX, UNSPEC_BNDLDX_ADDR, UNSPEC_BNDCL, UNSPEC_BNDCU)
4144         (UNSPEC_BNDCN, UNSPEC_MPX_FENCE): Remove.
4145         (BND0_REG, BND1_REG, BND2_REG, BND3_REG): Remove
4146         (FIRST_PSEUDO_REG): Update.
4147         (BND): Remove mode iterator.
4148         * config/i386/predicates.md (bnd_mem_operator): Remove.
4150 2018-08-24  Richard Sandiford  <richard.sandiford@arm.com>
4152         * tree-vect-stmts.c (vectorizable_bswap): Handle variable-length
4153         vectors.
4155 2018-08-24  Richard Sandiford  <richard.sandiford@arm.com>
4157         * tree-vect-slp.c (vect_transform_slp_perm_load): Separate out
4158         the case in which the permute needs only a single element and
4159         repeats for every vector of the result.  Extend that case to
4160         handle variable-length vectors.
4161         * tree-vect-stmts.c (vectorizable_load): Update accordingly.
4163 2018-08-24  H.J. Lu  <hongjiu.lu@intel.com>
4165         PR debug/79342
4166         * dwarf2out.c (save_macinfo_strings): Call set_indirect_string
4167         on DW_MACINFO_start_file for -gsplit-dwarf -g3.
4169 2018-08-24  Richard Biener  <rguenther@suse.de>
4171         * cfg.h (struct control_flow_graph): Add edge_flags_allocated and
4172         bb_flags_allocated members.
4173         (auto_flag): New RAII class for allocating flags.
4174         (auto_edge_flag): New RAII class for allocating edge flags.
4175         (auto_bb_flag): New RAII class for allocating bb flags.
4176         * cfgloop.c (verify_loop_structure): Allocate temporary edge
4177         flag dynamically.
4178         * cfganal.c (dfs_enumerate_from): Remove use of visited sbitmap
4179         in favor of temporarily allocated BB flag.
4180         * hsa-brig.c: Re-order includes.
4181         * hsa-dump.c: Likewise.
4182         * hsa-regalloc.c: Likewise.
4183         * print-rtl.c: Likewise.
4184         * profile-count.c: Likewise.
4186 2018-08-24  Segher Boessenkool  <segher@kernel.crashing.org>
4188         PR target/86989
4189         * config/rs6000/rs6000.c (toc_relative_expr_p): Check that the base is
4190         the TOC register.
4192 2018-08-24  Aldy Hernandez  <aldyh@redhat.com>
4194         PR 87073/bootstrap
4195         * wide-int-range.cc (wide_int_range_div): Do not ignore result
4196         from wide_int_range_multiplicative_op.
4198 2018-08-23  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
4200         * tree-vect-data-refs.c (vect_grouped_store_supported): Fix typo
4201         "permutaion".
4203 2018-08-23  Giuliano Belinassi  <giuliano.belinassi@usp.br>
4205         * genmatch.c (parser::parse_operation): Fix typo 'exapnded'
4206         to 'expanded'.
4208 2018-08-23  Alexander Monakov  <amonakov@ispras.ru>
4210         * tree-scalar-evolution.c (final_value_replacement_loop): Dump
4211         full GENERIC expression used for replacement.
4213 2018-08-23  Aldy Hernandez  <aldyh@redhat.com>
4215         * tree-vrp.c (abs_extent_range): Remove.
4216         (extract_range_into_wide_ints): Pass wide ints by reference.
4217         (extract_range_from_binary_expr_1): Rewrite the *DIV_EXPR code.
4218         Pass wide ints by reference in all calls to
4219         extract_range_into_wide_ints.
4220         * wide-int-range.cc (wide_int_range_div): New.
4221         * wide-int-range.h (wide_int_range_div): New.
4222         (wide_int_range_includes_zero_p): New.
4223         (wide_int_range_zero_p): New.
4225 2018-08-23  Matthew Malcomson  <matthew.malcomson@arm.com>
4227         * config/aarch64/aarch64.md (arches): New enum.
4228         (arch): New enum attr.
4229         (arch_enabled): New attr.
4230         (enabled): Now uses arch_enabled only.
4231         (simd, sve, fp16): Removed attribute.
4232         (fp): Attr now defined in terms of 'arch'.
4233         (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64, *movti_aarch64,
4234         *movhf_aarch64, <optab><fcvt_target><GPF:mode>2,
4235         <FCVT_F2FIXED:fcvt_fixed_insn><GPF:mode>3,
4236         <FCVT_FIXED2F:fcvt_fixed_insn><GPI:mode>3): Merge 'fp' and 'simd'
4237         attributes into 'arch'.
4238         (*movsf_aarch64, *movdf_aarch64, *movtf_aarch64, *add<mode>3_aarch64,
4239         subdi3, neg<mode>2, <optab><mode>3, one_cmpl<mode>2,
4240         *<NLOGICAL:optab>_one_cmpl<mode>3, *xor_one_cmpl<mode>3,
4241         *aarch64_ashl_sisd_or_int_<mode>3, *aarch64_lshr_sisd_or_int_<mode>3,
4242         *aarch64_ashr_sisd_or_int_<mode>3, *aarch64_sisd_ushl): Convert use of
4243         'simd' attribute into 'arch'.
4244         (load_pair_sw_<SX:mode><SX2:mode>, load_pair_dw_<DX:mode><DX2:mode>,
4245         store_pair_sw_<SX:mode><SX2:mode>, store_pair_dw_<DX:mode><DX2:mode>):
4246         Convert use of 'fp' attribute to 'arch'.
4247         * config/aarch64/aarch64-simd.md (move_lo_quad_internal_<mode>,
4248         move_lo_quad_internal_<mode>): (different modes) Merge 'fp' and 'simd'
4249         into 'arch'.
4250         (move_lo_quad_internal_be_<mode>, move_lo_quad_internal_be_<mode>):
4251         (different modes) Merge 'fp' and 'simd' into 'arch'.
4252         (*aarch64_combinez<mode>, *aarch64_combinez_be<mode>): Merge 'fp' and
4253         'simd' into 'arch'.
4255 2018-08-23  Segher Boessenkool  <segher@kernel.crashing.org>
4257         PR rtl-optimization/87026
4258         * expmed.c (canonicalize_comparison): If we can no longer create
4259         pseudoregisters, don't.
4261 2018-08-23  Richard Earnshaw  <rearnsha@arm.com>
4263         PR target/86951
4264         * config/arm/arm-protos.h (arm_emit_speculation_barrier): New
4265         prototype.
4266         * config/arm/arm.c (speculation_barrier_libfunc): New static
4267         variable.
4268         (arm_init_libfuncs): Initialize it.
4269         (arm_emit_speculation_barrier): New function.
4270         * config/arm/arm.md (speculation_barrier): Call
4271         arm_emit_speculation_barrier for architectures that do not have
4272         DSB or ISB.
4273         (speculation_barrier_insn): Only match on Armv7 or later.
4275 2018-08-23  Richard Biener  <rguenther@suse.de>
4277         PR middle-end/87024
4278         * tree-inline.c (copy_bb): Drop unused __builtin_va_arg_pack_len
4279         calls.
4281 2018-08-23  Richard Sandiford  <richard.sandiford@arm.com>
4283         * config/aarch64/aarch64.c (aarch64_evpc_sve_tbl): Fix handling
4284         of single-vector TBLs.
4285         (aarch64_vectorize_vec_perm_const): Set one_vector_p when only
4286         one input is given.
4288 2018-08-23  Richard Sandiford  <richard.sandiford@arm.com>
4290         PR target/85910
4291         * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const_1): Fix
4292         aarch64_evpc_tbl guard.
4294 2018-08-22  Bernd Edlinger  <bernd.edlinger@hotmail.de>
4296         * tree-ssa-dse.c (compute_trims): Avoid folding away undefined
4297         behaviour.
4299 2018-08-22  Martin Sebor  <msebor@redhat.com>
4301         PR middle-end/87052
4302         * tree-pretty-print.c (pretty_print_string): Add argument.
4303         (dump_generic_node): Call to pretty_print_string with string size.
4305 2018-08-22  Segher Boessenkool  <segher@kernel.crashing.org>
4307         PR rtl-optimization/86771
4308         * combine.c (try_combine): Do not allow splitting a resulting PARALLEL
4309         of two SETs into those two SETs, one to be placed at i2, if that SETs
4310         destination is modified between i2 and i3.
4312 2018-08-22  Richard Sandiford  <richard.sandiford@arm.com>
4314         PR tree-optimization/86725
4315         * tree-vect-loop.c (vect_inner_phi_in_double_reduction_p): New
4316         function.
4317         (vect_analyze_scalar_cycles_1): Check it.
4319 2018-08-22  Richard Sandiford  <richard.sandiford@arm.com>
4321         PR tree-optimization/86725
4322         * tree-vect-loop.c (vect_is_simple_reduction): When treating
4323         an outer loop phi as a double reduction, make sure that the
4324         single user of the phi result is an inner loop phi.
4326 2018-08-22  Richard Sandiford  <richard.sandiford@arm.com>
4328         * tree-vect-data-refs.c (vect_analyze_group_access_1): Convert
4329         grouped stores with gaps to a strided group.
4331 2018-08-22  Richard Sandiford  <richard.sandiford@arm.com>
4333         * tree-vect-stmts.c (get_group_load_store_type)
4334         (get_load_store_type): Only test STMT_VINFO_STRIDED_P for the
4335         first statement in a group.
4337 2018-08-22  Iain Sandoe  <iain@sandoe.co.uk>
4339         * config/darwin.h (LINK_COMMAND_SPEC_A): Sync LTO options with
4340         the sequence used in gcc/gcc.c.
4342 2018-08-22  Iain Sandoe  <iain@sandoe.co.uk>
4344         PR other/704
4345         * gcc-ar.c (main): Don’t try to invoke the plug-in if we’re not
4346         building it.
4348 2018-08-22  Iain Sandoe  <iain@sandoe.co.uk>
4350         * config/darwin10.h (LINK_GCC_C_SEQUENCE_SPEC): Adjust to use the
4351         Darwin10-specific unwinder-shim.
4352         * config/darwin12.h (LINK_GCC_C_SEQUENCE_SPEC): Remove.
4353         * config/rs6000/darwin.h (DARWIN_CRT1_SPEC, DARWIN_DYLIB1_SPEC):
4354         New to cater for Darwin10 Rosetta.
4356 2018-08-22  Iain Sandoe  <iain@sandoe.co.uk>
4358         * config/i386/i386.c (ix86_output_addr_diff_elt): Move the MACH-O
4359         specific test before the one for HAVE_AS_GOTOFF_IN_DATA.
4361 2018-08-22  Iain Sandoe  <iain@sandoe.co.uk>
4363         PR bootstrap/81033
4364         PR target/81733
4365         PR target/52795
4366         * gcc/dwarf2out.c (FUNC_SECOND_SECT_LABEL): New.
4367         (dwarf2out_switch_text_section): Generate a local label for the
4368         second function sub-section and apply it as the second FDE start
4369         label.
4370         * gcc/final.c (final_scan_insn_1): Emit second FDE label after the
4371         second sub-section start.
4373 2018-08-22  Richard Biener  <rguenther@suse.de>
4375         PR tree-optimization/86988
4376         * tree-vrp.c (vrp_prop::check_mem_ref): Bail out on VLAs.
4378 2018-08-22  Richard Biener  <rguenther@suse.de>
4380         PR tree-optimization/86945
4381         * tree-cfg.c (generate_range_test): Use unsigned arithmetic.
4383 2018-08-22  Alexandre Oliva <oliva@adacore.com>
4385         * config/rs6000/rs6000.c (SMALL_DATA_RELOC, SMALL_DATA_REG): Add
4386         a comment about how uses of r2 for .sdata2 come about.
4388 2018-08-22  Alexandre Oliva <aoliva@redhat.com>
4390         * tree-ssa-reassoc.c (is_reassociable_op): Fix cut&pasto.
4392 2018-08-21  Marek Polacek  <polacek@redhat.com>
4394         PR c++/86981, Implement -Wpessimizing-move.
4395         * doc/invoke.texi: Document -Wpessimizing-move.
4397 2018-08-21  Jan Hubicka  <jh@suse.cz>
4399         * tree.c (find_decls_types_r): Do not check for redundant typedefs.
4400         * tree.h (is_redundant_typedef): Remove.
4401         * dwarf2out.c (is_redundant_typedef): Turn into static function.
4403 2018-08-21  Jan Hubicka  <jh@suse.cz>
4405         * tree.c (free_lang_data_in_decl): Remove types from DECL_CONTEXT
4406         when possible.
4408 2018-08-21  Tamar Christina  <tamar.christina@arm.com>
4410         * expmed.c (extract_low_bits): Reject invalid subregs early.
4412 2018-08-21  Bernd Edlinger  <bernd.edlinger@hotmail.de>
4414         PR middle-end/86121
4415         * tree-ssa-strlen.c (adjust_last_stmt): Avoid folding away undefined
4416         behaviour.
4418 2018-08-21  Rasmus Villemoes  <rv@rasmusvillemoes.dk>
4420         * config/vxworks.h: Guard vxworks_asm_out_constructor and
4421         vxworks_asm_out_destructor by !HAVE_INITFINI_ARRAY_SUPPORT
4422         * config/vxworks.c: Likewise.
4424 2018-08-21  Rasmus Villemoes  <rv@rasmusvillemoes.dk>
4426         * config/vxworks.c: Set targetm.have_ctors_dtors
4427         if HAVE_INITFINI_ARRAY_SUPPORT.
4428         * config/vxworks.h: Set SUPPORTS_INIT_PRIORITY
4429         if HAVE_INITFINI_ARRAY_SUPPORT.
4431 2018-08-21  Rasmus Villemoes  <rv@rasmusvillemoes.dk>
4433         * config/vxworks.h: Add $(WIND_BASE)/target/h/wrn/coreip to
4434         default search path for VxWorks < 7.
4436 2018-08-21  Bernd Edlinger  <bernd.edlinger@hotmail.de>
4438         * gimple-ssa-sprintf.c (decl_constant_value): Remove.
4439         (get_format_string): Refer to c_getstr.
4441 2018-08-21  Tom de Vries  <tdevries@suse.de>
4443         * cgraph.h (debuginfo_early_init, debuginfo_init, debuginfo_fini)
4444         (debuginfo_start, debuginfo_stop, debuginfo_early_start)
4445         (debuginfo_early_stop): Declare.
4446         * cgraphunit.c (debuginfo_early_init, debuginfo_init, debuginfo_fini)
4447         (debuginfo_start, debuginfo_stop, debuginfo_early_start)
4448         (debuginfo_early_stop): New function.
4449         (symbol_table::finalize_compilation_unit): Call debuginfo_early_start
4450         and debuginfo_early_stop.
4451         * dwarf2out.c (dwarf2out_finish, dwarf2out_early_finish): Dump dwarf.
4452         * toplev.c (compile_file): Call debuginfo_start and debuginfo_stop.
4453         (general_init): Call debuginfo_early_init.
4454         (finalize): Call debuginfo_fini.
4455         (do_compile): Call debuginfo_init.
4456         * doc/invoke.texi (@gccoptlist): Add -fdump-debug and
4457         -fdump-early-debug.
4458         (@item -fdump-debug, @item -fdump-earlydebug): Add.
4460 2018-08-21  Tom de Vries  <tdevries@suse.de>
4462         * dwarf2out.c (print_dw_val, print_loc_descr, print_die): Handle
4463         flag_dump_noaddr and flag_dump_unnumbered.
4465 2018-08-21  Aldy Hernandez  <aldyh@redhat.com>
4467         * wide-int-range.cc (wide_int_range_abs): New.
4468         (wide_int_range_order_set): Rename from wide_int_range_min_max.
4469         * wide-int-range.h (wide_int_range_abs): New.
4470         (wide_int_range_min_max): New.
4471         * tree-vrp.c (extract_range_from_unary_expr): Rewrite ABS_EXPR
4472         case to call wide_int_range_abs.
4473         Rewrite MIN/MAX_EXPR to call wide_int_range_min_max.
4474         (extract_range_from_abs_expr): Delete.
4476 2018-08-20  Michael Meissner  <meissner@linux.ibm.com>
4478         PR target/87033
4479         * config/rs6000/rs6000.md (extendsi<mode>2): Change constraints
4480         from 'Y' to 'YZ' to enable the LWAX instruction to be generated
4481         for indexed loads.
4483 2018-08-20  Nathan Sidwell  <nathan@acm.org>
4484             Jeff Law <law@redhat.com>
4486         * config/s390/s390-c (s390_macro_to_expand): Use cpp_macro_p.
4487         * config/spu/spu-c.c (spu_macro_to_expand): Likewise.
4489 2018-08-20  David Malcolm  <dmalcolm@redhat.com>
4491         PR other/84889
4492         * attribs.c (diag_attr_exclusions): Add auto_diagnostic_group instance.
4493         (decl_attributes): Likewise.
4494         * calls.c (maybe_warn_nonstring_arg): Add auto_diagnostic_group
4495         instance.
4496         * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
4497         * diagnostic-core.h (class auto_diagnostic_group): New class.
4498         * diagnostic.c (diagnostic_initialize): Initialize the new fields.
4499         (diagnostic_report_diagnostic): Handle the first diagnostics within
4500         a group.
4501         (emit_diagnostic): Add auto_diagnostic_group instance.
4502         (inform): Likewise.
4503         (inform_n): Likewise.
4504         (warning): Likewise.
4505         (warning_at): Likewise.
4506         (warning_n): Likewise.
4507         (pedwarn): Likewise.
4508         (permerror): Likewise.
4509         (error): Likewise.
4510         (error_n): Likewise.
4511         (error_at): Likewise.
4512         (sorry): Likewise.
4513         (fatal_error): Likewise.
4514         (internal_error): Likewise.
4515         (internal_error_no_backtrace): Likewise.
4516         (auto_diagnostic_group::auto_diagnostic_group): New ctor.
4517         (auto_diagnostic_group::~auto_diagnostic_group): New dtor.
4518         * diagnostic.h (struct diagnostic_context): Add fields
4519         "diagnostic_group_nesting_depth",
4520         "diagnostic_group_emission_count", "begin_group_cb",
4521         "end_group_cb".
4522         * gimple-ssa-isolate-paths.c (find_implicit_erroneous_behavior):
4523         Add auto_diagnostic_group instance(s).
4524         (find_explicit_erroneous_behavior): Likewise.
4525         * gimple-ssa-warn-alloca.c (pass_walloca::execute): Likewise.
4526         * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Likewise.
4527         * gimplify.c (warn_implicit_fallthrough_r): Likewise.
4528         (gimplify_va_arg_expr): Likewise.
4529         * hsa-gen.c (HSA_SORRY_ATV): Likewise.
4530         (HSA_SORRY_AT): Likewise.
4531         * ipa-devirt.c (compare_virtual_tables): Likewise.
4532         (warn_odr): Likewise.
4533         * multiple_target.c (expand_target_clones): Likewise.
4534         * opts-common.c (cmdline_handle_error): Likewise.
4535         * reginfo.c (globalize_reg): Likewise.
4536         * substring-locations.c (format_warning_n_va): Likewise.
4537         * tree-inline.c (expand_call_inline): Likewise.
4538         * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Likewise.
4539         * tree-ssa-loop-niter.c
4540         (do_warn_aggressive_loop_optimizations): Likewise.
4541         * tree-ssa-uninit.c (warn_uninit): Likewise.
4542         * tree.c (warn_deprecated_use): Likewise.
4544 2018-08-20  H.J. Lu  <hongjiu.lu@intel.com>
4546         PR target/87014
4547         * config/i386/i386.md (eh_return): Always update EH return
4548         address in word_mode.
4550 2018-08-20  Chung-Lin Tang  <cltang@codesourcery.com>
4552         * targhooks.c (std_gimplify_va_arg_expr): Properly handle case of when
4553         TARGET_SPLIT_COMPLEX_ARG is defined.
4555 2018-08-20  Bernd Edlinger  <bernd.edlinger@hotmail.de>
4557         * expr.c (store_field): Change gcc_assert to gcc_checking_assert.
4559 2018-08-20  Bernd Edlinger  <bernd.edlinger@hotmail.de>
4561         PR target/86984
4562         * expr.c (expand_assignment): Assert that bitpos is positive.
4563         (store_field): Likewise
4564         (expand_expr_real_1): Make sure that bitpos is positive.
4565         * config/alpha/alpha.h (CONSTANT_ADDRESS_P): Avoid signed
4566         integer overflow.
4568 2018-08-20  Nathan Sidwell  <nathan@acm.org>
4570         * Makefile.in (CPP_ID_DATA_H): Delete.
4571         (CPP_INTERNAL_H): Don't add it.
4572         (GTFILES): Replace CPP_ID_DATA_H with CPPLIB_H.
4573         * gengtype.c (open_base_files): Replace cpp-id-data.h with cpplib.h
4575 2018-08-20  Richard Biener  <rguenther@suse.de>
4577         PR tree-optimization/78655
4578         * tree-vrp.c (extract_range_from_binary_expr_1): Make
4579         pointer + offset nonnull if either operand is nonnull work.
4581 2018-08-20  Tom de Vries  <tdevries@suse.de>
4583         * dwarf2out.c (add_scalar_info): Don't add reference to existing die
4584         unless the referenced die describes the added property using
4585         DW_AT_location or DW_AT_const_value.  Fall back to exprloc case.
4586         Otherwise, add a DW_AT_location to the referenced die.
4588 2018-08-19  Uros Bizjak  <ubizjak@gmail.com>
4590         PR target/86994
4591         * config/i386/i386.c (ix86_rtx_costs) [case SET]: Check source for
4592         register_operand when calling ix86_set_reg_reg_cost.
4593         [case CONST_INT, case CONST, case LABEL_REF, case SYMBOL_REF]:
4594         Set *total to 0 for operands that satisfy x86_64_immediate_operand
4595         predicate and to 1 otherwise.
4597 2018-08-18  Iain Sandoe  <iain@sandoe.co.uk>
4599         * config/darwin.c (darwin_override_options): If -gsplit-dwarf is set,
4600         emit a diagnostic that it is not supported and reset the option.
4601         * config/darwin.h (DRIVER_SELF_SPECS): Note that gsplit-dwarf is not
4602         supported and consume the option.  (ASM_FINAL_SPEC): New.
4604 2018-08-17  Segher Boessenkool  <segher@kernel.crashing.org>
4606         * doc/md.texi (Patterns): Use @ref instead of @xref in the middle of
4607         a sentence.
4609 2018-08-17  Sandra Loosemore  <sandra@codesourcery.com>
4611         C-SKY port: Documentation
4613         * doc/extend.texi (C-SKY Function Attributes): New section.
4614         * doc/invoke.texi (Option Summary): Add C-SKY options.
4615         (C-SKY Options): New section.
4616         * doc/md.texi (Machine Constraints): Document C-SKY constraints.
4618 2018-08-17  Jojo  <jijie_rong@c-sky.com>
4619             Huibin Wang  <huibin_wang@c-sky.com>
4620             Sandra Loosemore  <sandra@codesourcery.com>
4621             Chung-Lin Tang  <cltang@codesourcery.com>
4623         C-SKY port: Backend implementation
4625         * config/csky/*: New.
4626         * common/config/csky/*: New.
4628 2018-08-17  Jojo  <jijie_rong@c-sky.com>
4629             Huibin Wang  <huibin_wang@c-sky.com>
4630             Sandra Loosemore  <sandra@codesourcery.com>
4631             Chung-Lin Tang  <cltang@codesourcery.com>
4632             Andrew Jenner  <andrew@codesourcery.com>
4634         C-SKY port: Configury
4636         * config.gcc (csky-*-*): New.
4637         * configure.ac: Add csky to targets for dwarf2 debug_line support.
4638         * configure: Regenerated.
4640 2018-08-17  David Malcolm  <dmalcolm@redhat.com>
4642         * dump-context.h: Include "dumpfile.h".
4643         (dump_context::dump_printf_va): Convert final param from va_list
4644         to va_list *.  Convert from ATTRIBUTE_PRINTF to
4645         ATTRIBUTE_GCC_DUMP_PRINTF.
4646         (dump_context::dump_printf_loc_va): Likewise.
4647         * dumpfile.c: Include "stringpool.h".
4648         (make_item_for_dump_printf_va): Delete.
4649         (make_item_for_dump_printf): Delete.
4650         (class dump_pretty_printer): New class.
4651         (dump_pretty_printer::dump_pretty_printer): New ctor.
4652         (dump_pretty_printer::emit_items): New member function.
4653         (dump_pretty_printer::emit_any_pending_textual_chunks): New member
4654         function.
4655         (dump_pretty_printer::emit_item): New member function.
4656         (dump_pretty_printer::stash_item): New member function.
4657         (dump_pretty_printer::format_decoder_cb): New member function.
4658         (dump_pretty_printer::decode_format): New member function.
4659         (dump_context::dump_printf_va): Reimplement in terms of
4660         dump_pretty_printer.
4661         (dump_context::dump_printf_loc_va): Convert final param from va_list
4662         to va_list *.
4663         (dump_context::begin_scope): Reimplement call to
4664         make_item_for_dump_printf.
4665         (dump_printf): Update for change to dump_printf_va.
4666         (dump_printf_loc): Likewise.
4667         (selftest::test_capture_of_dump_calls): Convert "stmt" from
4668         greturn * to gimple *.  Add a test_decl.  Add tests of dump_printf
4669         with %T, %E, and %G.
4670         * dumpfile.h (ATTRIBUTE_GCC_DUMP_PRINTF): New macro.
4671         (dump_printf): Replace ATTRIBUTE_PRINTF_2 with
4672         ATTRIBUTE_GCC_DUMP_PRINTF (2, 3).
4673         (dump_printf_loc): Replace ATTRIBUTE_PRINTF_3 with
4674         ATTRIBUTE_GCC_DUMP_PRINTF (3, 0).
4675         * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Convert
4676         use of HOST_WIDE_INT_PRINT_DEC on unsigned HOST_WIDE_INT "count"
4677         within a dump_printf_loc call to "%wu".
4678         (vector_alignment_reachable_p): Merge two dump_printf[_loc] calls,
4679         converting a use of HOST_WIDE_INT_PRINT_DEC to "%wd".  Add a
4680         missing space after "=".
4681         * tree-vect-loop.c (vect_analyze_loop_2) Within a dump_printf
4682         call, convert use of HOST_WIDE_INT_PRINT_DEC to "%wd".
4683         * tree-vect-slp.c (vect_slp_bb): Within a dump_printf_loc call,
4684         convert use of HOST_WIDE_INT_PRINT_UNSIGNED to "%wu".
4685         * tree-vectorizer.c (try_vectorize_loop_1): Likewise.  Remove
4686         duplicate "vectorized" from message.
4688 2018-08-17  Szabolcs Nagy  <szabolcs.nagy@arm.com>
4690         * config/arm/arm-builtins.c (arm_init_simd_builtin_types): Clear
4691         polyNxK_t element's TYPE_STRING_FLAG.
4693 2018-08-17  Segher Boessenkool  <segher@kernel.crashing.org>
4695         * config/rs6000/rs6000.md (*cbranch, *creturn): Name these patterns
4696         (they were unnamed before).  Fix comments.
4698 2018-08-17  Nathan Sidwell  <nathan@acm.org>
4700         * cppbuiltin.c: Include "cpplib.h", not "cpp-id-data.h".
4702 2018-08-17  Richard Biener  <rguenther@suse.de>
4704         PR tree-optimization/86841
4705         * wide-int-range.cc (wide_int_range_lshift): Use to_uhwi.
4707 2018-08-17  Martin Liska  <mliska@suse.cz>
4709         * common.opt: Remove Warn, Init and Report for options with
4710         Ignore/Deprecated flag. Warning is done automatically for
4711         Deprecated flags.
4712         * config/i386/i386.opt: Likewise.
4713         * config/ia64/ia64.opt: Likewise.
4714         * config/rs6000/rs6000.opt: Likewise.
4715         * cppbuiltin.c (define_builtin_macros_for_compilation_flags):
4716         Remove usage of flag_check_pointer_bounds.
4717         * lto-wrapper.c (merge_and_complain): Do not handle
4718         OPT_fcheck_pointer_bounds.
4719         (append_compiler_options): Likewise.
4720         * opt-functions.awk: Do not handle Deprecated.
4721         * optc-gen.awk: Check that Var, Report and Init are not
4722         used for an option with Ignore/Deprecated flag.
4723         * opts-common.c (decode_cmdline_option): Do not report
4724         CL_ERR_DEPRECATED.
4725         (read_cmdline_option): Report warning for OPT_SPECIAL_deprecated
4726         options.
4727         * opts.h (struct cl_option): Remove cl_deprecated flag.
4728         (CL_ERR_DEPRECATED): Remove error enum value.
4730 2018-08-17  Richard Biener  <rguenther@suse.de>
4732         PR middle-end/86505
4733         * tree-inline.c (copy_bb): When inlining __builtin_va_arg_pack_len ()
4734         across a va-arg-pack using call adjust its return value accordingly.
4736 2018-08-16  Martin Sebor  <msebor@redhat.com>
4738         PR tree-optimization/86853
4739         * gimple-ssa-sprintf.c (struct format_result): Rename member.
4740         (struct fmtresult): Add member and initialize it in ctors.
4741         (format_character): Handle %C.  Extend range to NUL.  Set MAYFAIL.
4742         (format_string): Handle %S the same as %ls.  Set MAYFAIL.
4743         (format_directive): Set POSUNDER4K when MAYFAIL is set.
4744         (parse_directive): Handle %C same as %c.
4745         (sprintf_dom_walker::compute_format_length): Adjust.
4746         (is_call_safe): Adjust.
4748 2018-08-16  Bernd Edlinger  <bernd.edlinger@hotmail.de>
4750         * builtins.c (c_strlen): Add new parameter eltsize.  Use it
4751         for determining how to count the elements.
4752         * builtins.h (c_strlen): Adjust prototype.
4753         * expr.c (string_constant): Add new parameter mem_size.
4754         Set *mem_size appropriately.
4755         * expr.h (string_constant): Adjust protoype.
4756         * gimple-fold.c (get_range_strlen): Add new parameter eltsize.
4757         * gimple-fold.h (get_range_strlen): Adjust prototype.
4758         * gimple-ssa-sprintf.c (get_string_length): Add new parameter eltsize.
4759         (format_string): Call get_string_length with eltsize.
4761 2018-08-16  David Malcolm  <dmalcolm@redhat.com>
4763         * diagnostic.c (default_diagnostic_start_span_fn): Call pp_string
4764         to emit the span, rather than setting it as the prefix.
4766 2018-08-16  David Malcolm  <dmalcolm@redhat.com>
4768         * diagnostic-show-locus.c (layout::start_annotation_line): Add
4769         "margin_char" parameter, defaulting to space.  Use it in place
4770         of pp_space for the initial part of the margin.
4771         (layout::print_leading_fixits): Use '+' when filling the margin
4772         of line-insertion fix-it hints.
4774 2018-08-16  Segher Boessenkool  <segher@kernel.crashing.org>
4776         * config/rs6000/rs6000.md (two unnamed define_insn and define_split):
4777         Delete.
4779 2018-08-16  Segher Boessenkool  <segher@kernel.crashing.org>
4781         * config/rs6000/altivec.md: Don't set length attribute to the default
4782         value.
4783         * config/rs6000/darwin.md: Ditto.
4784         * config/rs6000/dfp.md: Ditto.
4785         * config/rs6000/htm.md: Ditto.
4786         * config/rs6000/rs6000.md: Ditto.
4787         * config/rs6000/sync.md: Ditto.
4788         * config/rs6000/vsx.md: Ditto.
4790 2018-08-16  Segher Boessenkool  <segher@kernel.crashing.org>
4792         * config/rs6000/altivec.md: Don't set length attribute to the default
4793         value, for branch instructions.
4794         * config/rs6000/darwin.md: Ditto.
4795         * config/rs6000/rs6000.md: Ditto.
4797 2018-08-16  Segher Boessenkool  <segher@kernel.crashing.org>
4799         * config/rs6000/rs6000.md (length): Always define as const_int 4.
4800         (unnamed conditional branch define_insn): Set length to 4 or 8
4801         depending on offset.
4802         (<bd>_<mode>): Similar, for alternative 0.
4803         (<bd>tf_<mode>): Ditto.
4805 2018-08-16  Tamar Christina  <tamar.christina@arm.com>
4807         * expr.c (copy_blkmode_to_reg): Perform larger copies when safe.
4809 2018-08-16  Matthew Malcomson  <matthew.malcomson@arm.com>
4811         * doc/rtl.texi: Replace old RTX class names with new names.
4814 2018-08-16  Vlad Lazar  <vlad.lazar@arm.com>
4816         * expmed.h (canonicalize_comparison): New declaration.
4817         * expmed.c (canonicalize_comparison, equivalent_cmp_code): New function.
4818         * expmed.c (emit_store_flag_1): Add call to canonicalize_comparison.
4819         * optabs.c (prepare_cmp_insn): Likewise.
4820         * rtl.h (unsigned_condition_p): New function which checks if a
4821         comparison operator is unsigned.
4823 2018-08-16  Nathan Sidwell  <nathan@acm.org>
4825         * config/rs6000/rs6000-c.c (rs6000_macro_to_expend): Use cpp_macro_p.
4826         * config/powerpcspc/powerpcspe-c.c (rs6000_macro_to_expend): Likewise.
4828 2018-08-16  Tamar Christina  <tamar.christina@arm.com>
4830         PR target/84711
4831         * config/arm/arm.c (arm_can_change_mode_class): Disallow subreg.
4832         * config/arm/neon.md (movv4hf, movv8hf): Refactored to..
4833         (mov<mov>): ..this and enable unconditionally.
4835 2018-08-16  Tamar Christina  <tamar.christina@arm.com>
4837         * config/arm/neon.md (*neon_mov<mode>): Remove reg-to-reg alternative.
4839 2018-08-16  Sam Tebbs  <sam.tebbs@arm.com>
4841         * config/aarch64/aarch64.opt (mlow-precision-recip-sqrt)
4842         (mlow-precision-sqrt, mlow-precision-div, mverbose-cost-dump): Replace
4843         "Common" with "Target".
4845 2018-08-15  Uros Bizjak  <ubizjak@gmail.com>
4847         * config/i386/i386.opt (mmitigate-rop): Mark as deprecated.
4848         * doc/invoke.texi (mmitigate-rop): Remove.
4849         * config/i386/i386.c: Do not include "regrename.h".
4850         (ix86_rop_should_change_byte_p, reg_encoded_number)
4851         (ix86_get_modrm_for_rop, set_rop_modrm_reg_bits, ix86_mitigate_rop):
4852         Remove.
4853         (ix86_reorg): Remove call to ix86_mitigate_rop.
4854         * config/i386/i386.md (attr "modrm_class"): Remove.
4855         (cmp<mode>_ccno_1, mov<mode>_xor, movstrict<mode>_xor)
4856         (x86_mov<mode>cc_0_m1. x86_mov<mode>cc_0_m1_se)
4857         (x86_mov<mode>cc_0_m1_neg): Remove modrm_class attribute override.
4859 2018-08-15  Will Schmidt  <will_schmidt@vnet.ibm.com>
4861         * config/rs6000/rs600.c (rs6000_gimple_fold_builtin): Add entries to
4862         allow folding of mergeh() and mergel() for the float and double types.
4863         (fold_mergehl_helper): Rework to handle building a permute tree
4864         for float vectors.
4866 2018-08-15  Uros Bizjak  <ubizjak@gmail.com>
4868         * config/i386/i386.c (expand_vec_perm_movs): Enable V4SFmode
4869         for TARGET_SSE.
4871 2018-08-15  David Malcolm  <dmalcolm@redhat.com>
4873         * common.opt (fdiagnostics-show-labels): New option.
4874         * diagnostic-show-locus.c (class layout_range): Add field
4875         "m_label".
4876         (class layout): Add field "m_show_labels_p".
4877         (layout_range::layout_range): Add param "label" and use it to
4878         initialize m_label.
4879         (make_range): Pass in NULL for new "label" param of layout_range's
4880         ctor.
4881         (layout::layout): Initialize m_show_labels_p.
4882         (layout::maybe_add_location_range): Pass in loc_range->m_label
4883         when constructing layout_range instances.
4884         (struct line_label): New struct.
4885         (layout::print_any_labels): New member function.
4886         (layout::print_line): Call it if label-printing is enabled.
4887         (selftest::test_one_liner_labels): New test.
4888         (selftest::test_diagnostic_show_locus_one_liner): Call it.
4889         * diagnostic.c (diagnostic_initialize): Initialize
4890         context->show_labels_p.
4891         * diagnostic.h (struct diagnostic_context): Add field
4892         "show_labels_p".
4893         * doc/invoke.texi (Diagnostic Message Formatting Options): Add
4894         -fno-diagnostics-show-labels.
4895         * dwarf2out.c (gen_producer_string): Add
4896         OPT_fdiagnostics_show_labels to the ignored options.
4897         * gcc-rich-location.c (gcc_rich_location::add_expr): Add "label"
4898         param.
4899         (gcc_rich_location::maybe_add_expr): Likewise.
4900         * gcc-rich-location.h (gcc_rich_location::gcc_rich_location): Add
4901         label" param, defaulting to NULL.
4902         (gcc_rich_location::add_expr): Add "label" param.
4903         (gcc_rich_location::maybe_add_expr): Likewise.
4904         (class text_range_label): New class.
4905         (class range_label_for_type_mismatch): New class.
4906         * gimple-ssa-sprintf.c (fmtwarn): Pass NULL for new label params
4907         of format_warning_va.
4908         (fmtwarn_n): Likewise for new params of format_warning_n_va.
4909         * lto-wrapper.c (merge_and_complain): Add
4910         OPT_fdiagnostics_show_labels to the "pick one setting" options.
4911         (append_compiler_options): Likewise to the dropped options.
4912         (append_diag_options): Likewise to the passed-on options.
4913         * opts.c (common_handle_option): Handle the new option.
4914         * selftest-diagnostic.c
4915         (test_diagnostic_context::test_diagnostic_context): Enable
4916         show_labels_p.
4917         * substring-locations.c: Include "gcc-rich-location.h".
4918         (format_warning_n_va): Add "fmt_label" and "param_label" params
4919         and use them as appropriate.
4920         (format_warning_va): Add "fmt_label" and "param_label" params,
4921         passing them on to format_warning_n_va.
4922         (format_warning_at_substring): Likewise.
4923         (format_warning_at_substring_n): Likewise.
4924         * substring-locations.h (format_warning_va): Add "fmt_label" and
4925         "param_label" params.
4926         (format_warning_n_va): Likewise.
4927         (format_warning_at_substring): Likewise.
4928         (format_warning_at_substring_n): Likewise.
4929         * toplev.c (general_init): Initialize global_dc->show_labels_p.
4931 2018-08-15  Qing Zhao  <qing.zhao@oracle.com>
4933         PR testsuite/86519
4934         * builtins.c (expand_builtin_memcmp): Do not expand the call
4935         when overflow is detected.
4937 2018-08-15  Martin Sebor  <msebor@redhat.com>
4939         PR tree-optimization/71625
4940         * config/aarch64/aarch64-builtins.c
4941         (aarch64_init_simd_builtin_types): Clear Poly8_t's TYPE_STRING_FLAG.
4943 2018-08-15  Ilya Leoshkevich  <iii@linux.ibm.com>
4945         * config/s390/s390.c (s390_reorg): Remove loop.
4947 2018-08-15  Iain Sandoe  <iain@sandoe.co.uk>
4949         * config/darwin.c
4950          (darwin_function_switched_text_sections): Delete.
4951         * gcc/config/darwin.h
4952          (TARGET_ASM_FUNCTION_SWITCHED_TEXT_SECTIONS): Likewise.
4954 2018-08-15  Iain Sandoe  <iain@sandoe.co.uk>
4956         PR target/81685
4957         * config/darwin.h: (DEBUG_STR_OFFSETS_SECTION, DEBUG_LOCLISTS_SECTION,
4958         DEBUG_RNGLISTS_SECTION) new macros.  (DEBUG_PUBNAMES_SECTION,
4959         DEBUG_PUBTYPES_SECTION) update to include GNU variant.
4961 2018-08-15  Martin Liska  <mliska@suse.cz>
4963         PR tree-optimization/86925
4964         * predict.c (expr_expected_value_1): When taking
4965         later predictor, assign also probability.
4966         Use fold_build2_initializer_loc in order to fold
4967         the expression in -frounding-math.
4969 2018-08-14  Allan Sandfeld Jensen <allan.jensen@qt.io>
4971         * config/i386/i386.c (expand_vec_perm_movs): New method matching movs
4972         patterns.
4973         (expand_vec_perm_1): Try the new method.
4975 2018-08-14  Ilya Leoshkevich  <iii@linux.ibm.com>
4977         PR target/86547
4978         * lra-lives.c (remove_some_program_points_and_update_live_ranges):
4979         Check whether lra_live_max_point is 0 before dividing.
4981 2018-08-14  Martin Sebor  <msebor@redhat.com>
4983         PR tree-optimization/86650
4984         * tree-vrp.c (vrp_prop::check_array_ref): Print an inform message.
4985         (vrp_prop::check_mem_ref): Same.
4987 2018-08-13  Liu Hao <lh_mouse@126.com>
4989         * pretty-print.c (eat_esc_sequence): Swap the foreground and
4990         background colors if the COMMON_LVB_REVERSE_VIDEO flag is set,
4991         and clear it thereafter, as it only works for DBCS.
4993 2018-08-13  Liu Hao <lh_mouse@126.com>
4995         * pretty-print.c (mingw_ansi_fputs): Do not call _close() on the
4996         handle returned by _get_osf_handle().
4998 2018-08-13  Will Schmidt  <will_schmidt@vnet.ibm.com>
5000         * gcc/config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support
5001         for folding vec_perm.
5003 2018-08-13  Will Schmidt  <will_schmidt@vnet.ibm.com>
5005         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin):
5006         Add support for gimple-folding of vec_pack() and vec_unpack()
5007         intrinsics.
5009 2018-08-13  Will Schmidt <will_schmidt@vnet.ibm.com>
5011         * config/rs6000/rs6000.c (rs6000_builtin_valid_without_lhs): Add
5012         vec_xst variants to the list.
5013         (rs6000_gimple_fold_builtin): Add support for folding unaligned
5014         vector loads and stores.
5016 2018-08-13  David Edelsohn  <dje.gcc@gmail.com>
5018         * config.gcc (rs6000-ibm-aix4.x): Delete.
5019         (rs6000-ibm-aix5.1): Delete.
5020         (rs6000-ibm-aix5.2): Delete.
5021         (rs6000-ibm-aix5.3): Delete.
5022         * config/rs6000/aix43.h: Delete.
5023         * config/rs6000/aix51.h: Delete.
5024         * config/rs6000/aix52.h: Delete.
5025         * config/rs6000/t-aix43: Delete.
5027 2018-08-13  Ilya Leoshkevich  <iii@linux.ibm.com>
5029         * config/s390/s390.c (s390_decompose_constant_pool_ref):
5030         New function.
5031         (s390_decompose_address): Factor out constant pool ref
5032         decomposition.
5034 2018-08-12  Chung-Ju Wu  <jasonwucj@gmail.com>
5036         * config/nds32/nds32-predicates.c
5037         (nds32_can_use_bclr_p): Change return type as bool.
5038         (nds32_can_use_bset_p): Ditto.
5039         (nds32_can_use_btgl_p): Ditto.
5040         (nds32_can_use_bitci_p): Ditto.
5041         * config/nds32/nds32-protos.h
5042         (nds32_can_use_bclr_p): Change declaration.
5043         (nds32_can_use_bset_p): Ditto.
5044         (nds32_can_use_btgl_p): Ditto.
5045         (nds32_can_use_bitci_p): Ditto.
5047 2018-08-12  Chung-Ju Wu  <jasonwucj@gmail.com>
5049         * config/nds32/nds32.c (nds32_expand_prologue, nds32_expand_epilogue):
5050         Support -msched-prolog-epilog option.
5051         * config/nds32/nds32.opt (msched-prolog-epilog): New option.
5053 2018-08-12  Chung-Ju Wu  <jasonwucj@gmail.com>
5055         * common/config/nds32/nds32-common.c
5056         (nds32_option_optimization_table): Enalbe -malways-align.
5058 2018-08-12  Chung-Ju Wu  <jasonwucj@gmail.com>
5060         * config.gcc (nds32*): Add nds32_isr.h and nds32_init.inc in
5061         extra_headers.
5062         * common/config/nds32/nds32-common.c (nds32_handle_option): Handle
5063         OPT_misr_secure_ case.
5064         * config/nds32/nds32-isr.c: Implementation of backward compatibility.
5065         * config/nds32/nds32-protos.h (nds32_isr_function_critical_p): New.
5066         * config/nds32/nds32.c (nds32_attribute_table): Add critical and
5067         secure attribute.
5068         * config/nds32/nds32.h (nds32_isr_nested_type): Add NDS32_CRITICAL.
5069         (nds32_isr_info): New field security_level.
5070         (TARGET_ISR_VECTOR_SIZE_4_BYTE): New macro.
5071         * config/nds32/nds32.md (return_internal): Consider critical attribute.
5072         * config/nds32/nds32.opt (misr-secure): New option.
5073         * config/nds32/nds32_init.inc: New file.
5074         * config/nds32/nds32_isr.h: New file.
5076 2018-08-11  John David Anglin  <danglin@gcc.gnu.org>
5078         * config/pa/pa.md (UNSPEC_MEMORY_BARRIER): New unspec enum.
5079         Update comment for atomic instructions.
5080         (atomic_storeqi, atomic_storehi, atomic_storesi, atomic_storesf,
5081         atomic_loaddf, atomic_loaddf_1, atomic_storedf, atomic_storedf_1):
5082         Remove.
5083         (atomic_loaddi): Revise fence expansion to only emit fence prior to
5084         load for __ATOMIC_SEQ_CST model.
5085         (atomic_loaddi_1): Remove float register target.
5086         (atomic_storedi): Handle CONST_INT values.
5087         (atomic_storedi_1): Remove float register source.  Add special case
5088         for zero value.
5089         (memory_barrier): New expander and insn.
5091 2018-08-11  Jakub Jelinek  <jakub@redhat.com>
5093         PR tree-optimization/86835
5094         * tree-ssa-math-opts.c (insert_reciprocals): Even when inserting
5095         new_stmt after def_gsi, make sure to insert new_square_stmt after
5096         that stmt, not 2 stmts before it.
5098 2018-08-10  Alexander Monakov  <amonakov@ispras.ru>
5100         PR target/82418
5101         * config/i386/i386.md (<s>mul<mode>3_highpart): Use DWIH mode iterator
5102         instead of SWI48.
5104 2018-08-10  Martin Liska  <mliska@suse.cz>
5106         PR target/83610
5107         * builtin-types.def (BT_FN_LONG_LONG_LONG_DOUBLE): Add new
5108         function type.
5109         * builtins.c (expand_builtin_expect_with_probability):
5110         New function.
5111         (expand_builtin_expect_with_probability): New function.
5112         (build_builtin_expect_predicate): Add new argumnet probability
5113         for BUILT_IN_EXPECT_WITH_PROBABILITY.
5114         (fold_builtin_expect):
5115         (fold_builtin_2):
5116         (fold_builtin_3):
5117         * builtins.def (BUILT_IN_EXPECT_WITH_PROBABILITY):
5118         * builtins.h (fold_builtin_expect): Set new argument.
5119         * doc/extend.texi: Document __builtin_expect_with_probability.
5120         * doc/invoke.texi: Likewise.
5121         * gimple-fold.c (gimple_fold_call): Pass new argument.
5122         * ipa-fnsummary.c (find_foldable_builtin_expect): Handle
5123         also BUILT_IN_EXPECT_WITH_PROBABILITY.
5124         * predict.c (get_predictor_value): New function.
5125         (expr_expected_value): Add new argument probability. Assume
5126         that predictor and probability are always non-null.
5127         (expr_expected_value_1): Likewise.  For __builtin_expect and
5128         __builtin_expect_with_probability set probability.  Handle
5129         combination in binary expressions.
5130         (tree_predict_by_opcode): Simplify code by simply calling
5131         get_predictor_value.
5132         (pass_strip_predict_hints::execute): Add handling of
5133         BUILT_IN_EXPECT_WITH_PROBABILITY.
5134         * predict.def (PRED_BUILTIN_EXPECT_WITH_PROBABILITY): Add
5135         new predictor.
5136         * tree.h (DECL_BUILT_IN_P): New function.
5138 2018-08-10  Martin Liska  <mliska@suse.cz>
5140         PR tree-optimization/85799
5141         * passes.def: Add argument for pass_strip_predict_hints.
5142         * predict.c (class pass_strip_predict_hints): Add new argument
5143         early_p.
5144         (strip_predictor_early): New function.
5145         (pass_strip_predict_hints::execute): Call the function to
5146         strip predictors.
5147         (strip_predict_hints): New function.
5148         * predict.def: Fix comment.
5150 2018-08-10  Thomas Preud'homme  <thomas.preudhomme@linaro.org>
5152         * Makefile.in: Clarify which tm.texi to copy over to assert the
5153         right to grant a GFDL license for all.
5155 2018-08-09  Jeff Law  <law@redhat.com>
5157         * config/m68k/m68k.c (m68k_adjust_decorated_operand): Remove
5158         unused variable.
5160 2018-08-09  Andreas Schwab  <schwab@linux-m68k.org>
5162         * config/m68k/m68k-protos.h (m68k_final_prescan_insn): Remove
5163         prototype.
5165 2018-08-09  Richard Sandiford  <richard.sandiford@arm.com>
5167         * tree-vect-loop.c (vectorizable_reduction): Allow inner-loop
5168         reductions for variable-length vectors.
5170 2018-08-09  David Malcolm  <dmalcolm@redhat.com>
5172         PR other/84889
5173         * common.opt (fdiagnostics-show-line-numbers): New option.
5174         * diagnostic-show-locus.c (class layout): Add fields
5175         "m_show_line_numbers_p" and "m_linenum_width";
5176         (num_digits): New function.
5177         (test_num_digits): New function.
5178         (layout::layout): Initialize new fields.  Update m_x_offset
5179         logic to handle any left margin.
5180         (layout::print_source_line): Print line number when requested.
5181         (layout::start_annotation_line): New member function.
5182         (layout::print_annotation_line): Call it.
5183         (layout::print_leading_fixits): Likewise.
5184         (layout::print_trailing_fixits): Likewise.  Update calls to
5185         move_to_column for new parameter.
5186         (layout::get_x_bound_for_row): Add "add_left_margin" param and use
5187         it to potentially call start_annotation_line.
5188         (layout::show_ruler): Call start_annotation_line.
5189         (selftest::test_line_numbers_multiline_range): New selftest.
5190         (selftest::diagnostic_show_locus_c_tests): Call test_num_digits
5191         and selftest::test_line_numbers_multiline_range.
5192         * diagnostic.c (diagnostic_initialize): Initialize
5193         show_line_numbers_p.
5194         * diagnostic.h (struct diagnostic_context): Add field
5195         "show_line_numbers_p".
5196         * doc/invoke.texi (Diagnostic Message Formatting Options): Add
5197         -fno-diagnostics-show-line-numbers.
5198         * dwarf2out.c (gen_producer_string): Add
5199         OPT_fdiagnostics_show_line_numbers to the ignored options.
5200         * lto-wrapper.c (merge_and_complain): Likewise to the "pick
5201         one setting" options.
5202         (append_compiler_options): Likewise to the dropped options.
5203         (append_diag_options): Likewise to the passed-on options.
5204         * opts.c (common_handle_option): Handle the new option.
5205         * toplev.c (general_init): Set up global_dc->show_line_numbers_p.
5207 2018-08-09  Kelvin Nilsen  <kelvin@gcc.gnu.org>
5209         * doc/extend.texi (PowerPC AltiVec Built-in Functions Available on
5210         ISA 2.07): Correct spelling of bcdsub to be __builtin_bcdsub.  Add
5211         third argument of type "const signed char" to descriptions of
5212         __builtin_bcdadd, __builtin_bcdadd_lt, __builtin_bcdadd_eq,
5213         __builtin_bcdadd_gt, __builtin_bcdadd_ov, __builtin_bcdsub,
5214         __builtin_bcdsub_lt, __builtin_bcdsub_eq, __builtin_bcdsub_gt,
5215         __builtin_bcdsub_ov functions.
5217 2018-08-09  Richard Sandiford  <richard.sandiford@arm.com>
5219         PR tree-optimization/86858
5220         * tree-vect-loop.c (vect_is_simple_reduction): Restore
5221         flow_bb_inside_loop_p calls.
5223 2018-08-09  Richard Sandiford  <richard.sandiford@arm.com>
5225         PR tree-optimization/86871
5226         * tree-vect-stmts.c (vect_transform_stmt): Use gimple_get_lhs
5227         instead of gimple_assign_lhs.
5229 2018-08-09  Richard Earnshaw  <rearnsha@arm.com>
5231         PR target/86887
5232         * config/aarch64/aarch64.md (add<mode>3_carryinC_zero): Add missing
5233         register constraint to operand 0.
5234         (add<mode>3_carryinC): Likewise.
5235         (add<mode>3_carryinV_zero, add<mode>3_carryinV): Likewise.
5237 2018-08-09  Martin Liska  <mliska@suse.cz>
5239         PR c/86895
5240         * common.opt: Remove extra line.
5242 2018-08-09  Martin Liska  <mliska@suse.cz>
5244         * params.def (PARAM_ALIGN_LOOP_ITERATIONS): Remove double dots
5245         at the end of a line, make first letter capital and end up
5246         a sentence with a dot.
5247         (PARAM_LOOP_INTERCHANGE_STRIDE_RATIO): Likewise.
5248         (PARAM_LOOP_BLOCK_TILE_SIZE): Likewise.
5249         (PARAM_GRAPHITE_MAX_NB_SCOP_PARAMS): Likewise.
5250         (PARAM_GRAPHITE_MAX_ARRAYS_PER_SCOP): Likewise.
5251         (PARAM_MAX_ISL_OPERATIONS): Likewise.
5252         (PARAM_GRAPHITE_ALLOW_CODEGEN_ERRORS): Likewise.
5253         (PARAM_PROFILE_FUNC_INTERNAL_ID): Likewise.
5254         (PARAM_INDIR_CALL_TOPN_PROFILE): Likewise.
5255         (PARAM_SLP_MAX_INSNS_IN_BB): Likewise.
5256         (PARAM_IPA_CP_EVAL_THRESHOLD): Likewise.
5257         (PARAM_IPA_CP_RECURSION_PENALTY): Likewise.
5258         (PARAM_IPA_CP_SINGLE_CALL_PENALTY): Likewise.
5259         (PARAM_IPA_CP_LOOP_HINT_BONUS): Likewise.
5260         (PARAM_IPA_CP_ARRAY_INDEX_HINT_BONUS): Likewise.
5261         (PARAM_TREE_REASSOC_WIDTH): Likewise.
5262         (PARAM_HSA_GEN_DEBUG_STORES): Likewise.
5263         (PARAM_MAX_SPECULATIVE_DEVIRT_MAYDEFS): Likewise.
5264         (PARAM_MAX_VRP_SWITCH_ASSERTIONS): Likewise.
5266 2018-08-09  Andreas Krebbel  <krebbel@linux.ibm.com>
5268         PR target/84332
5269         * config/s390/s390.c (s390_option_override_internal): Reduce the
5270         stack-clash-protection-probe-interval param if it would be too big
5271         for z900.
5273 2018-08-08  Andreas Schwab  <schwab@linux-m68k.org>
5275         PR target/46179
5276         * config/m68k/m68k.h (FINAL_PRESCAN_INSN): Don't define.
5277         * config/m68k/m68k.c (handle_move_double): Don't call
5278         m68k_final_prescan_insn.
5279         (m68k_adjust_decorated_operand): Renamed from
5280         m68k_final_prescan_insn, remove first and third operand and
5281         simplify.
5282         (print_operand): Call it.
5283         (print_operand_address): Call it.
5285 2018-08-08  Nathan Sidwell  <nathan@acm.org>
5287         * diagnostic.c (diagnostic_report_current_module): Use
5288         linemap_included_from & linemap_included_from_linemap.
5290 2018-08-08  Hongbo Zhang  <hongbo.zhang@linaro.org>
5292         * config/aarch64/aarch64-cores.def: Add phecda core.
5293         * config/aarch64/aarch64-tune.md: Regenerate.
5294         * doc/invoke.texi: Add phecda core.
5296 2018-08-08  Andreas Krebbel  <krebbel@linux.ibm.com>
5298         PR target/85295
5299         * config/s390/constraints.md ("NxHD0", "NxSD0"): New constraint
5300         definitions.
5301         * config/s390/s390.md ("movti"): Add more alternatives for
5302         constant to GPR copies.
5304 2018-08-08  Andreas Krebbel  <krebbel@linux.ibm.com>
5306         * config/s390/s390.c: Fix whitespace damage throughout the file.
5307         * config/s390/s390.h: Likewise.
5308         * config/s390/tpf.h: Likewise.
5310 2018-08-08  Ilya Leoshkevich  <iii@linux.ibm.com>
5312         * config/s390/s390.c (s390_loadrelative_operand_p):
5313         Remove TARGET_CPU_ZARCH usages.
5314         (s390_rtx_costs): Likewise.
5315         (s390_legitimate_constant_p): Likewise.
5316         (s390_cannot_force_const_mem): Likewise.
5317         (legitimate_reload_constant_p): Likewise.
5318         (s390_preferred_reload_class): Likewise.
5319         (legitimize_pic_address): Likewise.
5320         (legitimize_tls_address): Likewise.
5321         (s390_split_branches): Removed.
5322         (s390_add_execute): Removed.
5323         (s390_dump_pool): Remove TARGET_CPU_ZARCH usages.
5324         (s390_mainpool_start): Likewise.
5325         (s390_mainpool_finish): Likewise.
5326         (s390_mainpool_cancel): Removed.
5327         (s390_chunkify_start): Remove TARGET_CPU_ZARCH usages.
5328         (s390_chunkify_cancel): Likewise.
5329         (s390_return_addr_rtx): Likewise.
5330         (s390_register_info): Remove split_branches_pending_p uages.
5331         (s390_optimize_register_info): Likewise.
5332         (s390_init_frame_layout): Remove TARGET_CPU_ZARCH and
5333         split_branches_pending_p usages.
5334         (s390_can_eliminate): Remove TARGET_CPU_ZARCH usages.
5335         (s390_load_got): Likewise.
5336         (s390_expand_split_stack_prologue): Likewise.
5337         (output_asm_nops): Likewise.
5338         (s390_function_profiler): Likewise.
5339         (s390_emit_call): Likewise.
5340         (s390_conditional_register_usage): Likewise.
5341         (s390_optimize_prologue): Likewise.
5342         (s390_reorg): Remove TARGET_CPU_ZARCH and
5343         split_branches_pending_p usages.
5344         (s390_option_override_internal): Remove TARGET_CPU_ZARCH
5345         usages.
5346         (s390_output_indirect_thunk_function): Likewise.
5347         * config/s390/s390.h (TARGET_CPU_ZARCH): Removed.
5348         (TARGET_CPU_ZARCH_P): Removed.
5349         (struct machine_function): Remove split_branches_pending_p.
5350         * config/s390/s390.md: Remove TARGET_CPU_ZARCH usages.
5352 2018-08-08  Ilya Leoshkevich  <iii@linux.ibm.com>
5354         * common/config/s390/s390-common.c (processor_flags_table):
5355         Remove flags.
5356         * config.gcc: Remove with_arch/with_tune support.
5357         * config/s390/2064.md: Remove cpu attribute comparisons.
5358         * config/s390/driver-native.c (s390_host_detect_local_cpu):
5359         Remove MTN.
5360         * config/s390/linux.h (ASM_SPEC):
5361         Remove -march support.
5362         * config/s390/s390-c.c (s390_cpu_cpp_builtins_internal):
5363         Use a table to get an arch level.
5364         * config/s390/s390-opts.h (enum processor_type):
5365         Remove enum values.
5366         * config/s390/s390.c
5367         (processor_table): Remove entries, add arch_level values.
5368         (s390_issue_rate): Remove cases.
5369         (s390_option_override): Adjust
5370         s390_option_override_internal() call.
5371         (s390_option_override_internal): Remove deprecation warning.
5372         (s390_valid_target_attribute_tree): Adjust
5373         s390_option_override_internal() call.
5374         * config/s390/s390.h (struct s390_processor):
5375         Share with s390-c.c, add arch_level field.
5376         * config/s390/s390.md:
5377         Remove occurrences in cpu attribute.
5378         * config/s390/s390.opt: Remove -march/-mtune support.
5379         * config/s390/tpf.h (ASM_SPEC): Remove -march support.
5380         * doc/invoke.texi: Remove deprecation warning.
5382 2018-08-08  Luis Machado  <luis.machado@linaro.org>
5384         * config/aarch64/aarch64.c (qdf24xx_vector_cost): New static
5385         global.
5386         (qdf24xx_tunings): Set vector cost structure to
5387         qdf24xx_vector_cost.
5389         * config/aarch64/aarch64.c (qdf24xx_addrcost_table)
5390         <register_sextend>: Set to 3.
5392 2018-08-07  Richard Sandiford  <richard.sandiford@arm.com>
5394         PR target/86838
5395         * config/aarch64/iterators.md (FRECP, frecp_suffix): Delete.
5396         * config/aarch64/aarch64-simd.md
5397         (aarch64_frecp<FRECP:frecp_suffix><mode>): Fold FRECPE into...
5398         (@aarch64_frecpe<mode>): ...here and the move FRECPX to...
5399         (aarch64_frecpx<mode>): ...this new pattern.
5400         * config/aarch64/aarch64-simd-builtins.def: Remove comment
5401         about aarch64_frecp<FRECP:frecp_suffix><mode>.
5403 2018-08-07  Martin Liska  <mliska@suse.cz>
5405         PR middle-end/83023
5406         * predict.c (expr_expected_value_1): Handle DECL_IS_MALLOC,
5407         BUILT_IN_REALLOC and DECL_IS_OPERATOR_NEW.
5408         * predict.def (PRED_MALLOC_NONNULL): New predictor.
5409         * doc/extend.texi: Document that malloc attribute adds
5410         hit to compiler.
5412 2018-08-06  John David Anglin  <danglin@gcc.gnu.org>
5414         PR target/86785
5415         * config/pa/pa.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
5416         Define to speculation_safe_value_not_needed.
5418 2018-08-06  Jeff Law  <law@redhat.com>
5420         * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Pass down
5421         the vr_values instance to cprop_into_stmt.
5422         (cprop_into_stmt): Pass vr_values instance down to cprop_operand.
5423         (cprop_operand): Also query EVRP to determine if OP is a constant.
5425 2018-08-06  Nathan Sidwell  <nathan@acm.org>
5427         * diagnostic.c (diagnostic_report_current_module): Reroll
5428         included-at loop.  Translate text.
5430 2018-08-06  David Malcolm  <dmalcolm@redhat.com>
5432         * function-tests.c (selftest::test_expansion_to_rtl): Call
5433         free_after_compilation.
5435 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
5437         * config/aarch64/aarch64.md: Add clobber highs to tls_desc.
5439 2018-08-06  Andreas Krebbel  <krebbel@linux.ibm.com>
5441         * config/s390/s390.c (s390_loop_unroll_adjust): Prevent small
5442         loops with memory block operations from getting unrolled.
5444 2018-08-06  Ulrich Weigand  <uweigand@de.ibm.com>
5446         PR target/86807
5447         * config/spu/spu.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
5448         Define to speculation_safe_value_not_needed.
5450 2018-08-06  Jeff Law  <law@redhat.com>
5452         * reload1.c (forget_old_reloads_1): Adjust CLOBBER_HIGH
5453         assert.
5455 2018-08-06  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
5457         PR target/86662
5458         * gcc/tree.c (build_common_tree_nodes): Initialize integer_types array
5459         with all enabled __intN types.
5461         * gcc/testsuite/gcc.target/msp430/pr86662.c: New test.
5463 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
5465         * alias.c (record_set): Check for clobber high.
5466         * cfgexpand.c (expand_gimple_stmt): Likewise.
5467         * combine-stack-adj.c (single_set_for_csa): Likewise.
5468         * combine.c (find_single_use_1): Likewise.
5469         (set_nonzero_bits_and_sign_copies): Likewise.
5470         (get_combine_src_dest): Likewise.
5471         (is_parallel_of_n_reg_sets): Likewise.
5472         (try_combine): Likewise.
5473         (record_dead_and_set_regs_1): Likewise.
5474         (reg_dead_at_p_1): Likewise.
5475         (reg_dead_at_p): Likewise.
5476         * dce.c (deletable_insn_p): Likewise.
5477         (mark_nonreg_stores_1): Likewise.
5478         (mark_nonreg_stores_2): Likewise.
5479         * df-scan.c (df_find_hard_reg_defs): Likewise.
5480         (df_uses_record): Likewise.
5481         (df_get_call_refs): Likewise.
5482         * dwarf2out.c (mem_loc_descriptor): Likewise.
5483         * haifa-sched.c (haifa_classify_rtx): Likewise.
5484         * ira-build.c (create_insn_allocnos): Likewise.
5485         * ira-costs.c (scan_one_insn): Likewise.
5486         * ira.c (equiv_init_movable_p): Likewise.
5487         (rtx_moveable_p): Likewise.
5488         (interesting_dest_for_shprep): Likewise.
5489         * jump.c (mark_jump_label_1): Likewise.
5490         * postreload-gcse.c (record_opr_changes): Likewise.
5491         * postreload.c (reload_cse_simplify): Likewise.
5492         (struct reg_use): Add source expr.
5493         (reload_combine): Check for clobber high.
5494         (reload_combine_note_use): Likewise.
5495         (reload_cse_move2add): Likewise.
5496         (move2add_note_store): Likewise.
5497         * print-rtl.c (print_pattern): Likewise.
5498         * recog.c (decode_asm_operands): Likewise.
5499         (store_data_bypass_p): Likewise.
5500         (if_test_bypass_p): Likewise.
5501         * regcprop.c (kill_clobbered_value): Likewise.
5502         (kill_set_value): Likewise.
5503         * reginfo.c (reg_scan_mark_refs): Likewise.
5504         * reload1.c (maybe_fix_stack_asms): Likewise.
5505         (eliminate_regs_1): Likewise.
5506         (elimination_effects): Likewise.
5507         (mark_not_eliminable): Likewise.
5508         (scan_paradoxical_subregs): Likewise.
5509         (forget_old_reloads_1): Likewise.
5510         * reorg.c (find_end_label): Likewise.
5511         (try_merge_delay_insns): Likewise.
5512         (redundant_insn): Likewise.
5513         (own_thread_p): Likewise.
5514         (fill_simple_delay_slots): Likewise.
5515         (fill_slots_from_thread): Likewise.
5516         (dbr_schedule): Likewise.
5517         * resource.c (update_live_status): Likewise.
5518         (mark_referenced_resources): Likewise.
5519         (mark_set_resources): Likewise.
5520         * rtl.c (copy_rtx): Likewise.
5521         * rtlanal.c (reg_referenced_p): Likewise.
5522         (single_set_2): Likewise.
5523         (noop_move_p): Likewise.
5524         (note_stores): Likewise.
5525         * sched-deps.c (sched_analyze_reg): Likewise.
5526         (sched_analyze_insn): Likewise.
5528 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
5530         * cse.c (invalidate_reg): New function extracted from...
5531         (invalidate): ...here.
5532         (canonicalize_insn): Check for clobber high.
5533         (invalidate_from_clobbers): invalidate clobber highs.
5534         (invalidate_from_sets_and_clobbers): Likewise.
5535         (count_reg_usage): Check for clobber high.
5536         (insn_live_p): Likewise.
5537         * cselib.c (cselib_expand_value_rtx_1):Likewise.
5538         (cselib_invalidate_regno): Check for clobber in setter.
5539         (cselib_invalidate_rtx): Pass through setter.
5540         (cselib_invalidate_rtx_note_stores):
5541         (cselib_process_insn): Check for clobber high.
5542         * cselib.h (cselib_invalidate_rtx): Add operand.
5544 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
5546         * lra-eliminations.c (lra_eliminate_regs_1): Check for clobber high.
5547         (mark_not_eliminable): Likewise.
5548         * lra-int.h (struct lra_insn_reg): Add clobber high marker.
5549         * lra-lives.c (process_bb_lives): Check for clobber high.
5550         * lra.c (new_insn_reg): Remember clobber highs.
5551         (collect_non_operand_hard_regs): Check for clobber high.
5552         (lra_set_insn_recog_data): Likewise.
5553         (add_regs_to_insn_regno_info): Likewise.
5554         (lra_update_insn_regno_info): Likewise.
5556 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
5558         * rtl.h (reg_is_clobbered_by_clobber_high): Add declarations.
5559         * rtlanal.c (reg_is_clobbered_by_clobber_high): Add function.
5561 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
5563         * emit-rtl.c (verify_rtx_sharing): Check for CLOBBER_HIGH.
5564         (copy_insn_1): Likewise.
5565         (gen_hard_reg_clobber_high): New gen function.
5566         * genconfig.c (walk_insn_part): Check for CLOBBER_HIGH.
5567         * genemit.c (gen_exp): Likewise.
5568         (gen_emit_seq): Pass through info.
5569         (gen_insn): Check for CLOBBER_HIGH.
5570         (gen_expand): Pass through info.
5571         (gen_split): Likewise.
5572         (output_add_clobbers): Likewise.
5573         * genrecog.c (validate_pattern): Check for CLOBBER_HIGH.
5574         (remove_clobbers): Likewise.
5575         * rtl.h (gen_hard_reg_clobber_high): New declaration.
5577 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
5579         * doc/rtl.texi (clobber_high): Add.
5580         (parallel): Add in clobber high
5581         * rtl.c (rtl_check_failed_code3): Add function.
5582         * rtl.def (CLOBBER_HIGH): Add expression.
5583         * rtl.h (RTL_CHECKC3): Add macro.
5584         (rtl_check_failed_code3): Add declaration.
5585         (XC3EXP): Add macro.
5587 2018-08-05  H.J. Lu  <hongjiu.lu@intel.com>
5589         PR target/86386
5590         * config/i386/i386.c (ix86_finalize_stack_frame_flags): Set
5591         cfun->machine->max_used_stack_alignment if needed.
5593 2018-08-04  Martin Sebor  <msebor@redhat.com>
5595         PR tree-optimization/86571
5596         * gimple-ssa-sprintf.c (format_floating): Extend upper bound of
5597         NaN output to 4.
5599 2018-08-03  Sandra Loosemore  <sandra@codesourcery.com>
5601         PR target/86799
5602         * config/nios2/nios2.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
5603         Define.
5605 2018-08-03  Jeff Law  <law@redhat.com>
5607         PR target/86795
5608         * config/mn10300/mn10300.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
5609         Define to speculation_safe_value_not_needed.
5611 2018-08-03  David Malcolm  <dmalcolm@redhat.com>
5613         * doc/gcov.texi (-x): Remove duplicate "to".
5614         * doc/invoke.texi (-Wnoexcept-type): Remove duplicate "calls".
5615         (-Wif-not-aligned): Remove duplicate "is".
5616         (-flto): Remove duplicate "the".
5617         (MicroBlaze Options): In examples of "-mcpu=cpu-type", remove
5618         duplicate "v5.00.b".
5619         (MSP430 Options): Remove duplicate "and" from the description
5620         of "-mgprel-sec=regexp".
5621         (x86 Options): Remove duplicate copies of "vmldLog102" and
5622         vmlsLog104 from description of "-mveclibabi=type".
5624 2018-08-03  Richard Sandiford  <richard.sandiford@arm.com>
5626         * internal-fn.h (first_commutative_argument): Declare.
5627         * internal-fn.c (first_commutative_argument): New function.
5628         * tree-vect-slp.c (vect_get_and_check_slp_defs): Remove extra
5629         restrictions for pattern statements.  Use first_commutative_argument
5630         to look for commutative operands in calls to internal functions.
5632 2018-08-03  Aldy Hernandez  <aldyh@redhat.com>
5634         * Makefile.in (wide-int-range.o): New.
5635         * tree-vrp.c: Move all the wide_int_* functions to...
5636         * wide-int-range.cc: ...here.
5637         * tree-vrp.h: Move all the wide_int_* prototypes to...
5638         * wide-int-range.h: ...here.
5640 2018-08-03  Tom de Vries  <tdevries@suse.de>
5642         * common/config/nvptx/nvptx-common.c (nvptx_except_unwind_info): Return
5643         UI_NONE.
5644         * config/nvptx/nvptx.c (TARGET_ASM_BYTE_OP): Remove define.
5645         * except.c (output_function_exception_table): Do early exit if
5646         targetm_common.except_unwind_info (&global_options) == UI_NONE.
5648 2018-08-03  Martin Liska  <mliska@suse.cz>
5650         * predict.c (dump_prediction): Change to 2 digits
5651         in fraction part.
5653 2018-08-03  Siddhesh Poyarekar  <siddhesh@sourceware.org>
5655         * config/aarch64/falkor.md (falkor_am_1_vxvy_vxvy): Move
5656         neon_dup_q to...
5657         (falkor_am_1_gtov_gtov): ... a new insn reservation.
5659 2018-07-19  Ilya Leoshkevich  <iii@linux.ibm.com>
5661         * config/nds32/nds32.c (nds32_hard_regno_mode_ok): Replace > with >=.
5662         * df-problems.c (df_remove_dead_eq_notes): Replace > with >=.
5663         * dwarf2out.c (mem_loc_descriptor): Replace > with >=.
5664         * lra-constraints.c (spill_hard_reg_in_range): Replace <= with <.
5665         * lra-remat.c (call_used_input_regno_present_p): Replace <= with <.
5667 2018-08-02  David Malcolm  <dmalcolm@redhat.com>
5669         * diagnostic-show-locus.c (diagnostic_show_locus): Use
5670         pp_take_prefix when saving the existing prefix.
5671         * diagnostic.c (diagnostic_append_note): Likewise.
5672         * langhooks.c (lhd_print_error_function): Likewise.
5673         * pretty-print.c (pp_set_prefix): Drop the "const" from "prefix"
5674         param's type.  Free the existing prefix.
5675         (pp_take_prefix): New function.
5676         (pretty_printer::pretty_printer): Drop the prefix parameter.
5677         Rename the length parameter to match the comment.
5678         (pretty_printer::~pretty_printer): Free the prefix.
5679         * pretty-print.h (pretty_printer::pretty_printer): Drop the prefix
5680         parameter.
5681         (struct pretty_printer): Drop the "const" from "prefix" field's
5682         type and clarify memory management.
5683         (pp_set_prefix): Drop the "const" from the 2nd param.
5684         (pp_take_prefix): New decl.
5686 2018-08-02  Aaron Sawdey  <acsawdey@linux.ibm.com>
5688         * config/rs6000/rs6000-string.c (select_block_compare_mode): Move test
5689         for word_mode_ok here instead of passing as argument.
5690         (expand_block_compare): Change select_block_compare_mode() call.
5691         (expand_strncmp_gpr_sequence): New function.
5692         (expand_strn_compare): Make use of expand_strncmp_gpr_sequence.
5694 2018-08-02  Jeff Law  <law@redhat.com>
5696         PR target/86790
5697         * config/m68k/m68k.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
5698         Define to speculation_safe_value_not_needed.
5700         PR target/86784
5701         * config/h8300/h8300.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
5702         Define to speculation_safe_value_not_needed.
5704 2018-08-02  Tom de Vries  <tdevries@suse.de>
5706         PR target/86660
5707         * common/config/nvptx/nvptx-common.c (nvptx_except_unwind_info): New
5708         function.  Return UI_TARGET unconditionally.
5709         (TARGET_EXCEPT_UNWIND_INFO): Redefine to nvptx_except_unwind_info.
5710         * config/nvptx/nvptx.c (TARGET_ASM_BYTE_OP): Emit commented out '.byte'.
5712 2018-08-02  Richard Sandiford  <richard.sandiford@arm.com>
5714         * genemit.c (print_overload_test): Fix typo.
5716 2018-08-02  Richard Biener  <rguenther@suse.de>
5718         PR tree-optimization/86816
5719         * tree-ssa-tail-merge.c (tail_merge_valueize): New function
5720         which checks for value availability before querying it.
5721         (gvn_uses_equal): Use it.
5722         (same_succ_hash): Likewise.
5723         (gimple_equal_p): Likewise.
5725 2018-08-02  Nick Clifton  <nickc@redhat.com>
5727         PR target/86813
5728         * config/stormy16/stormy16.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
5729         Define to speculation_safe_value_not_needed.
5731         PR target/86810
5732         * config/v850/v850.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
5733         Define to speculation_safe_value_not_needed.
5735         PR target/86810
5736         * config/v850/v850.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
5737         Define to speculation_safe_value_not_needed.
5739         PR target/86803
5740         * config/rx/rx.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
5741         Define to speculation_safe_value_not_needed.
5743         PR target/86797
5744         * config/msp430/msp430.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
5745         Define to speculation_safe_value_not_needed.
5747         PR target/86791
5748         * config/mcore/mcore.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
5749         Define to speculation_safe_value_not_needed.
5751         PR target/86789
5752         * config/m32r/m32r.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
5753         Define to speculation_safe_value_not_needed.
5755         PR target/86787
5756         * config/iq2000/iq2000.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
5757         Define to speculation_safe_value_not_needed.
5759         PR target/86782
5760         * config/frv/frv.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define to
5761         speculation_safe_value_not_needed.
5763         PR target/86781
5764         * config/fr30/fr30.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define
5765         to speculation_safe_value_not_needed.
5767 2018-08-02  Richard Sandiford  <richard.sandiford@arm.com>
5769         * doc/md.texi: Expand the documentation of instruction names
5770         to mention port-local uses.  Document '@' in pattern names.
5771         * read-md.h (overloaded_instance, overloaded_name): New structs.
5772         (mapping): Declare.
5773         (md_reader::handle_overloaded_name): New member function.
5774         (md_reader::get_overloads): Likewise.
5775         (md_reader::m_first_overload): New member variable.
5776         (md_reader::m_next_overload_ptr): Likewise.
5777         (md_reader::m_overloads_htab): Likewise.
5778         * read-md.c (md_reader::md_reader): Initialize m_first_overload,
5779         m_next_overload_ptr and m_overloads_htab.
5780         * read-rtl.c (iterator_group): Add "type" and "get_c_token" fields.
5781         (get_mode_token, get_code_token, get_int_token): New functions.
5782         (map_attr_string): Add an optional argument that passes back
5783         the associated iterator.
5784         (overloaded_name_hash, overloaded_name_eq_p, named_rtx_p):
5785         (md_reader::handle_overloaded_name, add_overload_instance): New
5786         functions.
5787         (apply_iterators): Handle '@' names.  Report an error if '@'
5788         is used without iterators.
5789         (initialize_iterators): Initialize the new iterator_group fields.
5790         * genopinit.c (handle_overloaded_code_for)
5791         (handle_overloaded_gen): New functions.
5792         (main): Use them to print declarations of maybe_code_for_* and
5793         maybe_gen_* functions, and inline definitions of code_for_* and gen_*.
5794         * genemit.c (print_overload_arguments, print_overload_test)
5795         (handle_overloaded_code_for, handle_overloaded_gen): New functions.
5796         (main): Use it to print definitions of maybe_code_for_* and
5797         maybe_gen_* functions.
5798         * config/aarch64/aarch64.c (aarch64_split_128bit_move): Use
5799         gen_aarch64_mov{low,high}_di and gen_aarch64_movdi_{low,high}
5800         instead of explicit mode checks.
5801         (aarch64_split_simd_combine): Likewise gen_aarch64_simd_combine.
5802         (aarch64_split_simd_move): Likewise gen_aarch64_split_simd_mov.
5803         (aarch64_emit_load_exclusive): Likewise gen_aarch64_load_exclusive.
5804         (aarch64_emit_store_exclusive): Likewise gen_aarch64_store_exclusive.
5805         (aarch64_expand_compare_and_swap): Likewise
5806         gen_aarch64_compare_and_swap and gen_aarch64_compare_and_swap_lse
5807         (aarch64_gen_atomic_cas): Likewise gen_aarch64_atomic_cas.
5808         (aarch64_emit_atomic_swap): Likewise gen_aarch64_atomic_swp.
5809         (aarch64_constant_pool_reload_icode): Delete.
5810         (aarch64_secondary_reload): Use code_for_aarch64_reload_movcp
5811         instead of aarch64_constant_pool_reload_icode.  Use
5812         code_for_aarch64_reload_mov instead of explicit mode checks.
5813         (rsqrte_type, get_rsqrte_type, rsqrts_type, get_rsqrts_type): Delete.
5814         (aarch64_emit_approx_sqrt): Use gen_aarch64_rsqrte instead of
5815         get_rsqrte_type and gen_aarch64_rsqrts instead of gen_rqrts_type.
5816         (recpe_type, get_recpe_type, recps_type, get_recps_type): Delete.
5817         (aarch64_emit_approx_div): Use gen_aarch64_frecpe instead of
5818         get_recpe_type and gen_aarch64_frecps instead of get_recps_type.
5819         (aarch64_atomic_load_op_code): Delete.
5820         (aarch64_emit_atomic_load_op): Likewise.
5821         (aarch64_gen_atomic_ldop): Use UNSPECV_ATOMIC_* instead of
5822         aarch64_atomic_load_op_code.  Use gen_aarch64_atomic_load
5823         instead of aarch64_emit_atomic_load_op.
5824         * config/aarch64/aarch64.md (aarch64_reload_movcp<GPF_TF:mode><P:mode>)
5825         (aarch64_reload_movcp<VALL:mode><P:mode>, aarch64_reload_mov<mode>)
5826         (aarch64_movdi_<mode>low, aarch64_movdi_<mode>high)
5827         (aarch64_mov<mode>high_di, aarch64_mov<mode>low_di): Add a '@'
5828         character before the pattern name.
5829         * config/aarch64/aarch64-simd.md (aarch64_split_simd_mov<mode>)
5830         (aarch64_rsqrte<mode>, aarch64_rsqrts<mode>)
5831         (aarch64_simd_combine<mode>, aarch64_frecpe<mode>)
5832         (aarch64_frecps<mode>): Likewise.
5833         * config/aarch64/atomics.md (atomic_compare_and_swap<mode>)
5834         (aarch64_compare_and_swap<mode>, aarch64_compare_and_swap<mode>_lse)
5835         (aarch64_load_exclusive<mode>, aarch64_store_exclusive<mode>)
5836         (aarch64_atomic_swp<mode>, aarch64_atomic_cas<mode>)
5837         (aarch64_atomic_load<atomic_ldop><mode>): Likewise.
5839 2018-08-02  Richard Sandiford  <richard.sandiford@arm.com>
5841         * config/aarch64/aarch64.c (aarch64_float_const_representable_p):
5842         Allow HFmode constants if TARGET_FP_F16INST.
5844 2018-08-02  Jackson Woodruff  <jackson.woodruff@arm.com>
5846         PR target/86014
5847         * config/aarch64/aarch64.c (aarch64_operands_adjust_ok_for_ldpstp):
5848         No longer check last store for clobber of address register.
5850 2018-08-02  Martin Liska  <mliska@suse.cz>
5852         PR gcov-profile/86817
5853         * gcov.c (process_all_functions): New function.
5854         (main): Call it.
5855         (process_file): Move functions processing to
5856         process_all_functions.
5858 2018-08-02  David Malcolm  <dmalcolm@redhat.com>
5860         * dumpfile.c (dump_user_location_t::dump_user_location_t): Add
5861         "const" to the "gimple *" and "rtx_insn *" parameters.
5862         * dumpfile.h (dump_user_location_t::dump_user_location_t):
5863         Likewise.
5864         (dump_location_t::dump_location_t): Likewise.
5866 2018-08-01  Martin Sebor  <msebor@redhat.com>
5868         PR tree-optimization/86650
5869         * gimple-pretty-print.c (percent_G_format): Accept a "gimple *"
5870         rather than a "gcall *".  Directly pass the data of interest
5871         to percent_K_format, rather than building a temporary CALL_EXPR
5872         to hold it.
5873         * gimple-fold.c (gimple_fold_builtin_strncpy): Adjust.
5874         (gimple_fold_builtin_strncat): Adjust.
5875         * gimple-ssa-warn-restrict.h (check_bounds_or_overlap): Replace
5876         gcall* argument with gimple*.
5877         * gimple-ssa-warn-restrict.c (check_call): Same.
5878         (wrestrict_dom_walker::before_dom_children): Same.
5879         (builtin_access::builtin_access): Same.
5880         (check_bounds_or_overlap): Same
5881         (maybe_diag_overlap): Same.
5882         (maybe_diag_offset_bounds): Same.
5883         * tree-diagnostic.c (default_tree_printer): Move usage of
5884         EXPR_LOCATION (t) and TREE_BLOCK (t) from within percent_K_format
5885         to this callsite.
5886         * tree-pretty-print.c (percent_K_format): Add argument.
5887         * tree-pretty-print.h: Add argument.
5888         * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Adjust.
5889         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Adjust.
5890         (maybe_diag_stxncpy_trunc): Same.
5891         (handle_builtin_stxncpy): Same.
5892         (handle_builtin_strcat): Same.
5894 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
5896         * match.pd: Optimise pointer range checks.
5898 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
5900         PR tree-optimization/86758
5901         * tree-vect-stmts.c (vectorizable_simd_clone_call): Don't try
5902         to remove pattern statements.
5904 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
5906         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Use the
5907         result of dfs_enumerate_from when constructing stmt_vec_infos,
5908         instead of additionally calling get_loop_body.
5910 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
5912         * tree-vectorizer.h (vect_create_data_ref_ptr): Remove inv_p
5913         parameter.
5914         * tree-vect-data-refs.c (vect_create_data_ref_ptr): Likewise.
5915         When creating an iv, assert that the step is not known to be zero.
5916         (vect_setup_realignment): Update call accordingly.
5917         * tree-vect-stmts.c (vectorizable_store): Likewise.
5918         (vectorizable_load): Likewise.  Handle VMAT_INVARIANT separately.
5920 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
5922         * tree-vectorizer.h (vect_stmt_to_vectorize): New function.
5923         * tree-vect-loop.c (vect_update_vf_for_slp): Use it.
5924         (vectorizable_reduction): Likewise.
5925         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
5926         (vect_detect_hybrid_slp_stmts): Likewise.
5927         * tree-vect-stmts.c (vect_is_simple_use): Likewise.
5929 2018-08-01  Aldy Hernandez  <aldyh@redhat.com>
5931         * tree-vrp (zero_nonzero_bits_from_bounds): Rename to...
5932         (wide_int_set_zero_nonzero_bits): ...this.
5933         (zero_nonzero_bits_from_vr): Rename to...
5934         (vrp_set_zero_nonzero_bits): ...this.
5935         (extract_range_from_multiplicative_op_1): Abstract wide int
5936         code...
5937         (wide_int_range_multiplicative_op): ...here.
5938         (extract_range_from_binary_expr_1): Extract wide int binary
5939         operations into their own functions.
5940         (wide_int_range_lshift): New.
5941         (wide_int_range_can_optimize_bit_op): New.
5942         (wide_int_range_shift_undefined_p): New.
5943         (wide_int_range_bit_xor): New.
5944         (wide_int_range_bit_ior): New.
5945         (wide_int_range_bit_and): New.
5946         (wide_int_range_trunc_mod): New.
5947         (extract_range_into_wide_ints): New.
5948         (vrp_shift_undefined_p): New.
5949         (extract_range_from_multiplicative_op): New.
5950         (vrp_can_optimize_bit_op): New.
5951         * tree-vrp.h (value_range::dump): New.
5952         (wide_int_range_multiplicative_op): New.
5953         (wide_int_range_lshift):New.
5954         (wide_int_range_shift_undefined_p): New.
5955         (wide_int_range_bit_xor): New.
5956         (wide_int_range_bit_ior): New.
5957         (wide_int_range_bit_and): New.
5958         (wide_int_range_trunc_mod): New.
5959         (zero_nonzero_bits_from_bounds): Rename to...
5960         (wide_int_set_zero_nonzero_bits): ...this.
5961         (zero_nonzero_bits_from_vr): Rename to...
5962         (vrp_set_zero_nonzero_bits): ...this.
5963         (range_easy_mask_min_max): Rename to...
5964         (wide_int_range_can_optimize_bit_op): this.
5965         * vr-values.c (simplify_bit_ops_using_ranges): Rename
5966         zero_nonzero_bits_from_vr into vrp_set_zero_nonzero_bits.
5968 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
5970         * tree-vectorizer.h (vect_orig_stmt): New function.
5971         * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Use it.
5972         * tree-vect-loop.c (vect_model_reduction_cost): Likewise.
5973         (vect_create_epilog_for_reduction): Likewise.
5974         (vectorizable_live_operation): Likewise.
5975         * tree-vect-slp.c (vect_find_last_scalar_stmt_in_slp): Likewise.
5976         (vect_detect_hybrid_slp_stmts, vect_schedule_slp): Likewise.
5977         * tree-vect-stmts.c (vectorizable_call): Likewise.
5978         (vectorizable_simd_clone_call, vect_remove_stores): Likewise.
5980 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
5982         * tree-vectorizer.h (vect_transform_stmt): Remove grouped_store
5983         argument.
5984         * tree-vect-stmts.c (vect_transform_stmt): Likewise.
5985         * tree-vect-loop.c (vect_transform_loop_stmt): Update call accordingly.
5986         (vect_transform_loop): Likewise.
5987         * tree-vect-slp.c (vect_schedule_slp_instance): Likewise.
5989 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
5991         * tree-vectorizer.h (vect_schedule_slp): Return void.
5992         * tree-vect-slp.c (vect_schedule_slp_instance): Likewise.
5993         (vect_schedule_slp): Likewise.
5995 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
5997         * tree-vect-loop.c (vect_transform_loop_stmt): Remove slp_scheduled
5998         argument.
5999         (vect_transform_loop): Update calls accordingly.  Schedule SLP
6000         instances before the main loop, if any exist.
6002 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
6004         PR tree-optimization/86749
6005         * tree-vect-patterns.c (vect_determine_min_output_precision_1):
6006         If the lhs is used in a COND_EXPR, check that it is being used
6007         as the "then" or "else" value.
6009 2018-08-01  Tom de Vries  <tdevries@suse.de>
6011         PR target/86800
6012         * config/nvptx/nvptx.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define to
6013         speculation_safe_value_not_needed.
6015 2018-08-01  Richard Biener  <rguenther@suse.de>
6017         * tree-ssa-sccvn.c (visit_phi): Compare invariant addresses
6018         as base and offset.
6020 2018-08-01  Martin Liska  <mliska@suse.cz>
6022         * value-prof.c (gimple_divmod_fixed_value_transform): Unify
6023         format how successful transformation is dumped.
6024         (gimple_mod_pow2_value_transform): Likewise.
6025         (gimple_mod_subtract_transform): Likewise.
6026         (gimple_stringops_transform): Likewise.
6028 2018-08-01  Martin Liska  <mliska@suse.cz>
6030         PR value-prof/35543
6031         * value-prof.c (interesting_stringop_to_profile_p):
6032         Simplify the code and add BUILT_IN_MEMMOVE.
6033         (gimple_stringops_transform): Likewise.
6035 2018-08-01  Sam Tebbs  <sam.tebbs@arm.com>
6037         * config/aarch64/aarch64-simd.md
6038         (*aarch64_get_lane_zero_extendsi<mode>): Rename to...
6039         (*aarch64_get_lane_zero_extend<GPI:mode><VDQQH:mode>): ... This and
6040         use GPI iterator instead of SI mode.
6042 2018-08-01  Richard Earnshaw  <rearnsha@arm.com>
6044         * config/rs6000/rs6000.md (speculation_barrier): Renamed from
6045         rs6000_speculation_barrier.
6046         * config/rs6000/rs6000.c (rs6000_expand_builtin): Adjust for
6047         new barrier pattern name.
6049 2018-08-01  Richard Earnshaw  <rearnsha@arm.com>
6051         * config/i386/i386.md (unspecv): Add UNSPECV_SPECULATION_BARRIER.
6052         (speculation_barrier): New insn.
6054 2018-08-01  Richard Biener  <rguenther@suse.de>
6056         PR bootstrap/86724
6057         * graphite.h: Include isl/id.h and isl/space.h to allow build
6058         with ISL 0.20.
6060 2018-08-01  Jan Willem Jagersma  <jwjagersma@gmail.com>
6062         PR target/86651
6063         * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
6064         mode for COFF targets.
6065         * defaults.h (TARGET_COFF): Define.
6066         * config/i386/djgpp.h (TARGET_ASM_LTO_START, TARGET_ASM_LTO_END,
6067         TARGET_COFF): Define.
6068         (i386_djgpp_asm_lto_start, i386_djgpp_asm_lto_end): Declare.
6069         * config/i386/djgpp.c (saved_debug_info_level): New static variable.
6070         (i386_djgpp_asm_lto_start, i386_djgpp_asm_lto_end): New functions.
6072 2018-07-31  Alexandre Oliva <oliva@adacore.com>
6074         * gimple-streamer-in.c (input_bb): Restore BB discriminator.
6075         * gimple-streamer-out.c (output_bb): Save it.
6076         * lto-streamer-in.c (input_struct_function_base): Restore
6077         instance discriminator if available.  Create map on demand.
6078         * lto-streamer-out.c (output_struct_function_base): Save it if
6079         available.
6080         * final.c (decl_to_instance_map): Document LTO strategy.
6082 2018-07-31  Alexandre Oliva  <oliva@adacore.com>
6083             Olivier Hainque  <hainque@adacore.com>
6085         * debug.h (decl_to_instance_map_t): New type.
6086         (decl_to_instance_map): Declare.
6087         (maybe_create_decl_to_instance_map): New inline function.
6088         * final.c (bb_discriminator, last_bb_discriminator): New statics,
6089         to track basic block discriminators.
6090         (final_start_function_1): Initialize them.
6091         (final_scan_insn_1): On NOTE_INSN_BASIC_BLOCK, track
6092         bb_discriminator.
6093         (decl_to_instance_map): New variable.
6094         (map_decl_to_instance, maybe_set_discriminator): New functions.
6095         (notice_source_line): Set discriminator.
6097 2018-07-31  Ian Lance Taylor  <iant@golang.org>
6099         * targhooks.c (default_have_speculation_safe_value): Add
6100         ATTRIBUTE_UNUSED.
6102 2018-07-31  David Malcolm  <dmalcolm@redhat.com>
6104         * dump-context.h: Include "pretty-print.h".
6105         (dump_context::refresh_dumps_are_enabled): New decl.
6106         (dump_context::emit_item): New decl.
6107         (class dump_context): Add fields "m_test_pp" and
6108         "m_test_pp_flags".
6109         (temp_dump_context::temp_dump_context): Add param "test_pp_flags".
6110         (temp_dump_context::get_dumped_text): New decl.
6111         (class temp_dump_context): Add field "m_pp".
6112         * dumpfile.c (refresh_dumps_are_enabled): Convert to...
6113         (dump_context::refresh_dumps_are_enabled): ...and add a test for
6114         m_test_pp.
6115         (set_dump_file): Update for above change.
6116         (set_alt_dump_file): Likewise.
6117         (dump_loc): New overload, taking a pretty_printer *.
6118         (dump_context::dump_loc): Call end_any_optinfo.  Dump the location
6119         to any test pretty-printer.
6120         (make_item_for_dump_gimple_stmt): New function, adapted from
6121         optinfo::add_gimple_stmt.
6122         (dump_context::dump_gimple_stmt): Call it, and use the result,
6123         eliminating the direct usage of dump_file and alt_dump_file in
6124         favor of indirectly using them via emit_item.
6125         (make_item_for_dump_gimple_expr): New function, adapted from
6126         optinfo::add_gimple_expr.
6127         (dump_context::dump_gimple_expr): Call it, and use the result,
6128         eliminating the direct usage of dump_file and alt_dump_file in
6129         favor of indirectly using them via emit_item.
6130         (make_item_for_dump_generic_expr): New function, adapted from
6131         optinfo::add_tree.
6132         (dump_context::dump_generic_expr): Call it, and use the result,
6133         eliminating the direct usage of dump_file and alt_dump_file in
6134         favor of indirectly using them via emit_item.
6135         (make_item_for_dump_printf_va): New function, adapted from
6136         optinfo::add_printf_va.
6137         (make_item_for_dump_printf): New function.
6138         (dump_context::dump_printf_va): Call make_item_for_dump_printf_va,
6139         and use the result, eliminating the direct usage of dump_file and
6140         alt_dump_file in favor of indirectly using them via emit_item.
6141         (make_item_for_dump_dec): New function.
6142         (dump_context::dump_dec): Call it, and use the result,
6143         eliminating the direct usage of dump_file and alt_dump_file in
6144         favor of indirectly using them via emit_item.
6145         (make_item_for_dump_symtab_node): New function, adapted from
6146         optinfo::add_symtab_node.
6147         (dump_context::dump_symtab_node): Call it, and use the result,
6148         eliminating the direct usage of dump_file and alt_dump_file in
6149         favor of indirectly using them via emit_item.
6150         (dump_context::begin_scope): Reimplement, avoiding direct usage
6151         of dump_file and alt_dump_file in favor of indirectly using them
6152         via emit_item.
6153         (dump_context::emit_item): New member function.
6154         (temp_dump_context::temp_dump_context): Add param "test_pp_flags".
6155         Set up test pretty-printer on the underlying context.  Call
6156         refresh_dumps_are_enabled.
6157         (temp_dump_context::~temp_dump_context): Call
6158         refresh_dumps_are_enabled.
6159         (temp_dump_context::get_dumped_text): New member function.
6160         (selftest::verify_dumped_text): New function.
6161         (ASSERT_DUMPED_TEXT_EQ): New macro.
6162         (selftest::test_capture_of_dump_calls): Run all tests twice, with
6163         and then without optinfo enabled.  Add uses of
6164         ASSERT_DUMPED_TEXT_EQ to all tests.  Add test of nested scopes.
6165         * dumpfile.h: Update comment for the dump_* API.
6166         * optinfo-emit-json.cc
6167         (selftest::test_building_json_from_dump_calls): Update for new
6168         param for temp_dump_context ctor.
6169         * optinfo.cc (optinfo_item::optinfo_item): Remove "owned" param
6170         and "m_owned" field.
6171         (optinfo_item::~optinfo_item): Likewise.
6172         (optinfo::add_item): New member function.
6173         (optinfo::emit): Update comment.
6174         (optinfo::add_string): Delete.
6175         (optinfo::add_printf): Delete.
6176         (optinfo::add_printf_va): Delete.
6177         (optinfo::add_gimple_stmt): Delete.
6178         (optinfo::add_gimple_expr): Delete.
6179         (optinfo::add_tree): Delete.
6180         (optinfo::add_symtab_node): Delete.
6181         (optinfo::add_dec): Delete.
6182         * optinfo.h (class dump_context): New forward decl.
6183         (optinfo::add_item): New decl.
6184         (optinfo::add_string): Delete.
6185         (optinfo::add_printf): Delete.
6186         (optinfo::add_printf_va): Delete.
6187         (optinfo::add_gimple_stmt): Delete.
6188         (optinfo::add_gimple_expr): Delete.
6189         (optinfo::add_tree): Delete.
6190         (optinfo::add_symtab_node): Delete.
6191         (optinfo::add_dec): Delete.
6192         (optinfo::add_poly_int): Delete.
6193         (optinfo_item::optinfo_item): Remove "owned" param.
6194         (class optinfo_item): Remove field "m_owned".
6196 2018-07-31  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
6198         PR middle-end/86705
6199         * gcc/cfgexpand.c (set_parm_rtl): Use the alignment of Pmode when
6200         MAX_SUPPORTED_STACK_ALIGNMENT would otherwise be exceeded by the
6201         requested variable alignment.
6202         (expand_one_ssa_partition): Likewise.
6203         (expand_one_var): Likewise.
6205 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
6207         * config/pdp11/pdp11.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
6208         to speculation_safe_value_not_needed.
6210 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
6212         * targhooks.h (speculation_safe_value_not_needed): New prototype.
6213         * targhooks.c (speculation_safe_value_not_needed): New function.
6214         * target.def (have_speculation_safe_value): Update documentation.
6215         * doc/tm.texi: Regenerated.
6217 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
6219         * config/aarch64/iterators.md (ALLI_TI): New iterator.
6220         * config/aarch64/aarch64.md (despeculate_copy<ALLI_TI:mode>): New
6221         expand.
6222         (despeculate_copy<ALLI:mode>_insn): New insn.
6223         (despeculate_copyti_insn): New insn.
6224         (despeculate_simple<ALLI:mode>): New insn
6225         (despeculate_simpleti): New insn.
6226         * config/aarch64/aarch64.c (aarch64_speculation_safe_value): New
6227         function.
6228         (TARGET_SPECULATION_SAFE_VALUE): Redefine to
6229         aarch64_speculation_safe_value.
6230         (aarch64_print_operand): Handle const0_rtx in modifier 'H'.
6232 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
6234         * config/aarch64/aarch64-speculation.cc: New file.
6235         * config/aarch64/aarch64-passes.def (pass_track_speculation): Add
6236         before pass_reorder_blocks.
6237         * config/aarch64/aarch64-protos.h (make_pass_track_speculation): Add
6238         prototype.
6239         * config/aarch64/aarch64.c (aarch64_conditional_register_usage): Fix
6240         X14 and X15 when tracking speculation.
6241         * config/aarch64/aarch64.md (register name constants): Add
6242         SPECULATION_TRACKER_REGNUM and SPECULATION_SCRATCH_REGNUM.
6243         (unspec): Add UNSPEC_SPECULATION_TRACKER.
6244         (speculation_barrier): New insn attribute.
6245         (cmp<mode>): Allow SP in comparisons.
6246         (speculation_tracker): New insn.
6247         (speculation_barrier): Add speculation_barrier attribute.
6248         * config/aarch64/t-aarch64: Add make rule for aarch64-speculation.o.
6249         * config.gcc (aarch64*-*-*): Add aarch64-speculation.o to extra_objs.
6250         * doc/invoke.texi (AArch64 Options): Document -mtrack-speculation.
6252 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
6254         * config/aarch64/aarch64.md (cb<optab><mode>1): Disable when
6255         aarch64_track_speculation is true.
6256         (tb<optab><mode>1): Likewise.
6257         * config/aarch64/aarch64.c (aarch64_split_compare_regs): Do not
6258         generate CB[N]Z when tracking speculation.
6259         (aarch64_split_compare_and_swap): Likewise.
6260         (aarch64_split_atomic_op): Likewise.
6262 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
6264         * config/aarch64/aarch64.opt (mtrack-speculation): New target option.
6266 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
6268         * config/aarch64.md (unspecv): Add UNSPECV_SPECULAION_BARRIER.
6269         (speculation_barrier): New insn.
6271 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
6273         * config/arm/unspecs.md (unspecv): Add VUNSPEC_SPECULATION_BARRIER.
6274         * config/arm/arm.md (speculation_barrier): New expand.
6275         (speculation_barrier_insn): New pattern.
6277 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
6279         * builtin-types.def (BT_FN_PTR_PTR_VAR): New function type.
6280         (BT_FN_I1_I1_VAR, BT_FN_I2_I2_VAR, BT_FN_I4_I4_VAR): Likewise.
6281         (BT_FN_I8_I8_VAR, BT_FN_I16_I16_VAR): Likewise.
6282         * builtin-attrs.def (ATTR_NOVOPS_NOTHROW_LEAF_LIST): New attribute
6283         list.
6284         * builtins.def (BUILT_IN_SPECULATION_SAFE_VALUE_N): New builtin.
6285         (BUILT_IN_SPECULATION_SAFE_VALUE_PTR): New internal builtin.
6286         (BUILT_IN_SPECULATION_SAFE_VALUE_1): Likewise.
6287         (BUILT_IN_SPECULATION_SAFE_VALUE_2): Likewise.
6288         (BUILT_IN_SPECULATION_SAFE_VALUE_4): Likewise.
6289         (BUILT_IN_SPECULATION_SAFE_VALUE_8): Likewise.
6290         (BUILT_IN_SPECULATION_SAFE_VALUE_16): Likewise.
6291         * builtins.c (expand_speculation_safe_value): New function.
6292         (expand_builtin): Call it.
6293         * doc/cpp.texi: Document predefine __HAVE_SPECULATION_SAFE_VALUE.
6294         * doc/extend.texi: Document __builtin_speculation_safe_value.
6295         * doc/md.texi: Document "speculation_barrier" pattern.
6296         * doc/tm.texi.in: Pull in TARGET_SPECULATION_SAFE_VALUE and
6297         TARGET_HAVE_SPECULATION_SAFE_VALUE.
6298         * doc/tm.texi: Regenerated.
6299         * target.def (have_speculation_safe_value, speculation_safe_value): New
6300         hooks.
6301         * targhooks.c (default_have_speculation_safe_value): New function.
6302         (default_speculation_safe_value): New function.
6303         * targhooks.h (default_have_speculation_safe_value): Add prototype.
6304         (default_speculation_safe_value): Add prototype.
6306 2018-07-31  David Malcolm  <dmalcolm@redhat.com>
6308         * dump-context.h (dump_context::dump_loc): New decl.
6309         * dumpfile.c (dump_context::dump_loc): New member function.
6310         (dump_context::dump_gimple_stmt_loc): Reimplement using dump_loc
6311         and dump_gimple_stmt.
6312         (dump_context::dump_gimple_expr_loc): Likewise, using
6313         dump_gimple_expr.
6314         (dump_context::dump_generic_expr_loc): Likewise, using
6315         dump_generic_expr.
6316         (dump_context::dump_printf_loc_va): Likewise, using
6317         dump_printf_va.
6318         (dump_context::begin_scope): Explicitly using the global function
6319         "dump_loc", rather than the member function.
6321 2018-07-31  Martin Sebor  <msebor@redhat.com>
6323         PR tree-optimization/86741
6324         * tree-vrp.c (vrp_prop::check_mem_ref): Avoid incomplete types.
6326 2018-07-31  Andreas Krebbel  <krebbel@linux.ibm.com>
6328         * config/s390/s390.c (s390_expand_setmem): Make the unrolling to
6329         depend on whether prefetch instructions will be emitted or not.
6330         Use TARGET_SETMEM_PFD for checking whether prefetch instructions
6331         will be emitted or not.
6332         * config/s390/s390.h (TARGET_SETMEM_PREFETCH_DISTANCE)
6333         (TARGET_SETMEM_PFD): New macros.
6335 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
6337         * tree-vectorizer.h (stmt_vec_info): Turn back into a typedef.
6338         (NULL_STMT_VEC_INFO): Delete.
6339         (stmt_vec_info::operator*): Likewise.
6340         (stmt_vec_info::operator gimple *): Likewise.
6341         * tree-vect-loop.c (vectorizable_reduction): Use NULL instead
6342         of NULL_STMT_VEC_INFO.
6343         * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
6344         (vect_reassociating_reduction_p): Likewise.
6345         * tree-vect-stmts.c (vect_build_gather_load_calls): Likewise.
6346         (vectorizable_store): Likewise.
6347         * tree-vectorizer.c (vec_info::set_vinfo_for_stmt): Likewise.
6348         (vec_info::free_stmt_vec_infos): Likewise.
6350 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
6352         * tree-vectorizer.h (vect_stmt_in_region_p): Delete.
6353         * tree-vectorizer.c (vect_stmt_in_region_p): Likewise.
6355 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
6357         * tree-vectorizer.h (vec_info::new_vinfo_for_stmt)
6358         (vec_info::set_vinfo_for_stmt, vec_info::free_stmt_vec_infos)
6359         (vec_info::free_stmt_vec_info): New private member functions.
6360         (set_stmt_vec_info_vec, free_stmt_vec_infos, vinfo_for_stmt)
6361         (set_vinfo_for_stmt, new_stmt_vec_info, free_stmt_vec_info): Delete.
6362         * tree-parloops.c (gather_scalar_reductions): Remove calls to
6363         set_stmt_vec_info_vec and free_stmt_vec_infos.
6364         * tree-vect-loop.c (_loop_vec_info): Remove call to
6365         set_stmt_vec_info_vec.
6366         * tree-vect-stmts.c (new_stmt_vec_info, set_stmt_vec_info_vec)
6367         (free_stmt_vec_infos, free_stmt_vec_info): Delete in favor of...
6368         * tree-vectorizer.c (vec_info::new_stmt_vec_info)
6369         (vec_info::set_vinfo_for_stmt, vec_info::free_stmt_vec_infos)
6370         (vec_info::free_stmt_vec_info): ...these new functions.  Remove
6371         assignments in {vec_info::,}new_stmt_vec_info that are redundant
6372         with the clearing in the xcalloc.
6373         (stmt_vec_info_vec): Delete.
6374         (vec_info::vec_info): Don't call set_stmt_vec_info_vec.
6375         (vectorize_loops): Likewise.
6376         (vec_info::~vec_info): Remove argument from call to
6377         free_stmt_vec_infos.
6378         (vec_info::add_stmt): Remove vinfo argument from call to
6379         new_stmt_vec_info.
6381 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
6383         * tree-vectorizer.h (free_stmt_vec_info): Take a stmt_vec_info
6384         rather than a gimple stmt.
6385         * tree-vect-stmts.c (free_stmt_vec_info): Likewise.  Don't free
6386         information for pattern statements when passed the original
6387         statement; instead wait to be passed the pattern statement itself.
6388         Don't call set_vinfo_for_stmt here.
6389         (free_stmt_vec_infos): Update call to free_stmt_vec_info.
6390         * tree-vect-loop.c (_loop_vec_info::~loop_vec_info): Don't free
6391         stmt_vec_infos here.
6392         * tree-vect-slp.c (_bb_vec_info::~bb_vec_info): Likewise.
6393         * tree-vectorizer.c (vec_info::remove_stmt): Nullify the statement's
6394         stmt_vec_infos entry.
6396 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
6398         * tree-vectorizer.h (vec_info::replace_stmt): Declare.
6399         * tree-vectorizer.c (vec_info::replace_stmt): New function.
6400         * tree-vect-slp.c (vect_remove_slp_scalar_calls): Use it.
6401         * tree-vect-stmts.c (vectorizable_call): Likewise.
6402         (vectorizable_simd_clone_call): Likewise.
6404 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
6406         * tree-vectorizer.h (vec_info::remove_stmt): Declare.
6407         * tree-vectorizer.c (vec_info::remove_stmt): New function.
6408         * tree-vect-loop-manip.c (vect_set_loop_condition): Use it.
6409         * tree-vect-loop.c (vect_transform_loop): Likewise.
6410         * tree-vect-slp.c (vect_schedule_slp): Likewise.
6411         * tree-vect-stmts.c (vect_remove_stores): Likewise.
6413 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
6415         * tree-vectorizer.h (vec_info::lookup_dr): New member function.
6416         (vect_dr_stmt): Delete.
6417         * tree-vectorizer.c (vec_info::lookup_dr): New function.
6418         * tree-vect-loop-manip.c (vect_update_inits_of_drs): Use it instead
6419         of DR_VECT_AUX.
6420         * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
6421         (vect_analyze_data_ref_dependence, vect_record_base_alignments)
6422         (vect_verify_datarefs_alignment, vect_peeling_supportable)
6423         (vect_analyze_data_ref_accesses, vect_prune_runtime_alias_test_list)
6424         (vect_analyze_data_refs): Likewise.
6425         (vect_slp_analyze_data_ref_dependence): Likewise.  Take a vec_info
6426         argument.
6427         (vect_find_same_alignment_drs): Likewise.
6428         (vect_slp_analyze_node_dependences): Update calls accordingly.
6429         (vect_analyze_data_refs_alignment): Likewise.  Use vec_info::lookup_dr
6430         instead of DR_VECT_AUX.
6431         (vect_get_peeling_costs_all_drs): Take a loop_vec_info instead
6432         of a vector data references.  Use vec_info::lookup_dr instead of
6433         DR_VECT_AUX.
6434         (vect_peeling_hash_get_lowest_cost): Update calls accordingly.
6435         (vect_enhance_data_refs_alignment): Likewise.  Use vec_info::lookup_dr
6436         instead of DR_VECT_AUX.
6438 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
6440         * tree-vectorizer.h (_loop_vec_info::unaligned_dr): Change to
6441         dr_vec_info.
6442         * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Update
6443         accordingly.
6444         * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
6445         * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
6446         (vect_gen_prolog_loop_niters): Likewise.
6448 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
6450         * tree-vectorizer.h (set_dr_misalignment, dr_misalignment)
6451         (DR_TARGET_ALIGNMENT, aligned_access_p, known_alignment_for_access_p)
6452         (vect_known_alignment_in_bytes, vect_dr_behavior)
6453         (vect_get_scalar_dr_size): Take references as dr_vec_infos
6454         instead of data_references.  Update calls to other routines for
6455         which the same change has been made.
6456         * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Take
6457         dr_vec_infos instead of stmt_vec_infos.
6458         (vect_analyze_data_ref_dependence): Update call accordingly.
6459         (vect_slp_analyze_data_ref_dependence)
6460         (vect_record_base_alignments): Use DR_VECT_AUX.
6461         (vect_calculate_target_alignment, vect_compute_data_ref_alignment)
6462         (vect_update_misalignment_for_peel, verify_data_ref_alignment)
6463         (vector_alignment_reachable_p, vect_get_data_access_cost)
6464         (vect_peeling_supportable, vect_analyze_group_access_1)
6465         (vect_analyze_group_access, vect_analyze_data_ref_access)
6466         (vect_vfa_segment_size, vect_vfa_access_size, vect_vfa_align)
6467         (vect_compile_time_alias, vect_small_gap_p)
6468         (vectorizable_with_step_bound_p, vect_duplicate_ssa_name_ptr_info):
6469         (vect_supportable_dr_alignment): Take references as dr_vec_infos
6470         instead of data_references.  Update calls to other routines for
6471         which the same change has been made.
6472         (vect_verify_datarefs_alignment, vect_get_peeling_costs_all_drs)
6473         (vect_find_same_alignment_drs, vect_analyze_data_refs_alignment)
6474         (vect_slp_analyze_and_verify_node_alignment)
6475         (vect_analyze_data_ref_accesses, vect_prune_runtime_alias_test_list)
6476         (vect_create_addr_base_for_vector_ref, vect_create_data_ref_ptr)
6477         (vect_setup_realignment): Use dr_vec_infos.  Update calls after
6478         above changes.
6479         (_vect_peel_info::dr): Replace with...
6480         (_vect_peel_info::dr_info): ...this new field.
6481         (vect_peeling_hash_get_most_frequent)
6482         (vect_peeling_hash_choose_best_peeling): Update accordingly.
6483         (vect_peeling_hash_get_lowest_cost):
6484         (vect_enhance_data_refs_alignment): Likewise.  Update calls to other
6485         routines for which the same change has been made.
6486         (vect_peeling_hash_insert): Likewise.  Take a dr_vec_info instead of a
6487         data_reference.
6488         * tree-vect-loop-manip.c (get_misalign_in_elems)
6489         (vect_gen_prolog_loop_niters): Use dr_vec_infos.  Update calls after
6490         above changes.
6491         * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
6492         * tree-vect-stmts.c (vect_get_store_cost, vect_get_load_cost)
6493         (vect_truncate_gather_scatter_offset, compare_step_with_zero)
6494         (get_group_load_store_type, get_negative_load_store_type)
6495         (vect_get_data_ptr_increment, vectorizable_store)
6496         (vectorizable_load): Likewise.
6497         (ensure_base_align): Take a dr_vec_info instead of a data_reference.
6498         Update calls to other routines for which the same change has been made.
6500 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
6502         * tree-vectorizer.h (vec_info::move_dr): New member function.
6503         (dataref_aux): Rename to...
6504         (dr_vec_info): ...this and add "dr" and "stmt" fields.
6505         (_stmt_vec_info::dr_aux): Update accordingly.
6506         (_stmt_vec_info::data_ref_info): Delete.
6507         (STMT_VINFO_GROUPED_ACCESS, DR_GROUP_FIRST_ELEMENT)
6508         (DR_GROUP_NEXT_ELEMENT, DR_GROUP_SIZE, DR_GROUP_STORE_COUNT)
6509         (DR_GROUP_GAP, DR_GROUP_SAME_DR_STMT, REDUC_GROUP_FIRST_ELEMENT):
6510         (REDUC_GROUP_NEXT_ELEMENT, REDUC_GROUP_SIZE): Use dr_aux.dr instead
6511         of data_ref.
6512         (STMT_VINFO_DATA_REF): Likewise.  Turn into an lvalue.
6513         (STMT_VINFO_DR_INFO): New macro.
6514         (DR_VECT_AUX): Use STMT_VINFO_DR_INKFO and vect_dr_stmt.
6515         (set_dr_misalignment): Update after rename of dataref_aux.
6516         (vect_dr_stmt): Move earlier in file.  Return dr_aux.stmt.
6517         * tree-vect-stmts.c (new_stmt_vec_info): Remove redundant
6518         initialization of STMT_VINFO_DATA_REF.
6519         * tree-vectorizer.c (vec_info::move_dr): New function.
6520         * tree-vect-patterns.c (vect_recog_bool_pattern)
6521         (vect_recog_mask_conversion_pattern)
6522         (vect_recog_gather_scatter_pattern): Use it.
6523         * tree-vect-data-refs.c (vect_analyze_data_refs): Initialize
6524         the "dr" and "stmt" fields of dr_vec_info instead of
6525         STMT_VINFO_DATA_REF.
6527 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
6529         * tree-vectorizer.h (_stmt_vec_info::pattern_stmt_p): New field.
6530         (is_pattern_stmt_p): Use it.
6531         * tree-vect-patterns.c (vect_init_pattern_stmt): Set pattern_stmt_p
6532         on pattern statements.
6534 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
6536         * tree-vect-patterns.c (vect_mark_pattern_stmts): Take the
6537         original stmt as a stmt_vec_info rather than a gimple stmt.
6538         (vect_pattern_recog_1): Take the statement directly as a
6539         stmt_vec_info, rather than via a gimple_stmt_iterator.
6540         Update call to vect_mark_pattern_stmts.
6541         (vect_pattern_recog): Update calls accordingly.
6543 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
6545         * tree-vectorizer.h (vect_get_vec_defs_for_stmt_copy)
6546         (vect_get_vec_def_for_stmt_copy): Take a vec_info rather than
6547         a vect_def_type for the first argument.
6548         * tree-vect-stmts.c (vect_get_vec_defs_for_stmt_copy): Likewise.
6549         (vect_get_vec_def_for_stmt_copy): Likewise.  Return the original
6550         operand if it isn't defined by a vectorized statement.
6551         (vect_build_gather_load_calls): Remove the mask_dt argument and
6552         update calls to vect_get_vec_def_for_stmt_copy.
6553         (vectorizable_bswap): Likewise the dt argument.
6554         (vectorizable_call): Update calls to vectorizable_bswap and
6555         vect_get_vec_def_for_stmt_copy.
6556         (vectorizable_simd_clone_call, vectorizable_assignment)
6557         (vectorizable_shift, vectorizable_operation, vectorizable_condition)
6558         (vectorizable_comparison): Update calls to
6559         vect_get_vec_def_for_stmt_copy.
6560         (vectorizable_store): Likewise.  Remove now-unnecessary calls to
6561         vect_is_simple_use.
6562         (vect_get_loop_based_defs): Remove dt argument and update call
6563         to vect_get_vec_def_for_stmt_copy.
6564         (vectorizable_conversion): Update calls to vect_get_loop_based_defs
6565         and vect_get_vec_def_for_stmt_copy.
6566         (vectorizable_load): Update calls to vect_build_gather_load_calls
6567         and vect_get_vec_def_for_stmt_copy.
6568         * tree-vect-loop.c (vect_create_epilog_for_reduction)
6569         (vectorizable_reduction, vectorizable_live_operation): Update calls
6570         to vect_get_vec_def_for_stmt_copy.
6572 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
6574         * tree-vect-data-refs.c (vect_record_base_alignment): Replace vec_info
6575         and gimple stmt arguments with a stmt_vec_info.
6576         (vect_record_base_alignments): Update calls accordingly.
6577         * tree-vect-slp.c (vect_record_max_nunits): Replace vec_info
6578         and gimple stmt arguments with a stmt_vec_info.
6579         (vect_build_slp_tree_1): Remove vinfo argument and update call
6580         to vect_record_max_nunits.
6581         (vect_build_slp_tree_2): Update calls to vect_build_slp_tree_1
6582         and vect_record_max_nunits.
6584 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
6586         * tree-vectorizer.h (nested_in_vect_loop_p): Move further down
6587         file and take a stmt_vec_info instead of a gimple stmt.
6588         (supportable_widening_operation, vect_finish_replace_stmt)
6589         (vect_finish_stmt_generation, vect_get_store_rhs)
6590         (vect_get_vec_def_for_operand_1, vect_get_vec_def_for_operand)
6591         (vect_get_vec_defs, vect_init_vector, vect_transform_stmt)
6592         (vect_remove_stores, vect_analyze_stmt, vectorizable_condition)
6593         (vect_get_smallest_scalar_type, vect_check_gather_scatter)
6594         (vect_create_data_ref_ptr, bump_vector_ptr)
6595         (vect_permute_store_chain, vect_setup_realignment)
6596         (vect_transform_grouped_load, vect_record_grouped_load_vectors)
6597         (vect_create_addr_base_for_vector_ref, vectorizable_live_operation)
6598         (vectorizable_reduction, vectorizable_induction)
6599         (get_initial_def_for_reduction, is_simple_and_all_uses_invariant)
6600         (vect_get_place_in_interleaving_chain): Take stmt_vec_infos rather
6601         than gimple stmts as arguments.
6602         * tree-vect-data-refs.c (vect_get_smallest_scalar_type)
6603         (vect_preserves_scalar_order_p, vect_slp_analyze_node_dependences)
6604         (can_group_stmts_p, vect_check_gather_scatter)
6605         (vect_create_addr_base_for_vector_ref, vect_create_data_ref_ptr)
6606         (bump_vector_ptr, vect_permute_store_chain, vect_setup_realignment)
6607         (vect_permute_load_chain, vect_shift_permute_load_chain)
6608         (vect_transform_grouped_load)
6609         (vect_record_grouped_load_vectors): Likewise.
6610         * tree-vect-loop.c (vect_fixup_reduc_chain)
6611         (get_initial_def_for_reduction, vect_create_epilog_for_reduction)
6612         (vectorize_fold_left_reduction, is_nonwrapping_integer_induction)
6613         (vectorizable_reduction, vectorizable_induction)
6614         (vectorizable_live_operation, vect_loop_kill_debug_uses): Likewise.
6615         * tree-vect-patterns.c (type_conversion_p, adjust_bool_stmts)
6616         (vect_get_load_store_mask): Likewise.
6617         * tree-vect-slp.c (vect_get_place_in_interleaving_chain)
6618         (vect_analyze_slp_instance, vect_mask_constant_operand_p): Likewise.
6619         * tree-vect-stmts.c (vect_mark_relevant)
6620         (is_simple_and_all_uses_invariant)
6621         (exist_non_indexing_operands_for_use_p, process_use)
6622         (vect_init_vector_1, vect_init_vector, vect_get_vec_def_for_operand_1)
6623         (vect_get_vec_def_for_operand, vect_get_vec_defs)
6624         (vect_finish_stmt_generation_1, vect_finish_replace_stmt)
6625         (vect_finish_stmt_generation, vect_truncate_gather_scatter_offset)
6626         (compare_step_with_zero, vect_get_store_rhs, get_group_load_store_type)
6627         (get_negative_load_store_type, get_load_store_type)
6628         (vect_check_load_store_mask, vect_check_store_rhs)
6629         (vect_build_gather_load_calls, vect_get_strided_load_store_ops)
6630         (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
6631         (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
6632         (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
6633         (get_group_alias_ptr_type, vectorizable_store, hoist_defs_of_uses)
6634         (vectorizable_load, vectorizable_condition, vectorizable_comparison)
6635         (vect_analyze_stmt, vect_transform_stmt, vect_remove_stores)
6636         (supportable_widening_operation): Likewise.
6638 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
6640         * tree-vect-data-refs.c (vect_describe_gather_scatter_call): Take
6641         a stmt_vec_info instead of a gcall.
6642         (vect_check_gather_scatter): Update call accordingly.
6643         * tree-vect-loop-manip.c (iv_phi_p): Take a stmt_vec_info instead
6644         of a gphi.
6645         (vect_can_advance_ivs_p, vect_update_ivs_after_vectorizer)
6646         (slpeel_update_phi_nodes_for_loops):): Update calls accordingly.
6647         * tree-vect-loop.c (vect_transform_loop_stmt): Take a stmt_vec_info
6648         instead of a gimple stmt.
6649         (vect_transform_loop): Update calls accordingly.
6650         * tree-vect-slp.c (vect_split_slp_store_group): Take and return
6651         stmt_vec_infos instead of gimple stmts.
6652         (vect_analyze_slp_instance): Update use accordingly.
6653         * tree-vect-stmts.c (read_vector_array, write_vector_array)
6654         (vect_clobber_variable, vect_stmt_relevant_p, permute_vec_elements)
6655         (vect_use_strided_gather_scatters_p, vect_build_all_ones_mask)
6656         (vect_build_zero_merge_argument, vect_get_gather_scatter_ops)
6657         (vect_gen_widened_results_half, vect_get_loop_based_defs)
6658         (vect_create_vectorized_promotion_stmts, can_vectorize_live_stmts):
6659         Take a stmt_vec_info instead of a gimple stmt and pass stmt_vec_infos
6660         down to subroutines.
6662 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
6664         * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Change the type
6665         of the worklist from a vector of gimple stmts to a vector of
6666         stmt_vec_infos.
6667         * tree-vect-stmts.c (vect_mark_relevant, process_use)
6668         (vect_mark_stmts_to_be_vectorized): Likewise
6670 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
6672         * tree-vect-loop.c (vect_analyze_loop_operations): Look up the
6673         statement before passing it to vect_analyze_stmt.
6674         (vect_create_epilog_for_reduction): Use a stmt_vec_info to walk
6675         the chain of phi vector definitions.  Track the exit phi via its
6676         stmt_vec_info.
6677         (vectorizable_reduction): Set cond_stmt_vinfo directly from the
6678         STMT_VINFO_REDUC_DEF.
6679         * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Use
6680         stmt_vec_infos to handle the statement chains.
6681         (vect_get_slp_defs): Record the first statement in the node
6682         using a stmt_vec_info.
6683         * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Look up
6684         statements here and pass their stmt_vec_info down to subroutines.
6685         (vect_init_vector_1): Hoist call to vinfo_for_stmt and pass it
6686         down to vect_finish_stmt_generation.
6687         (vect_init_vector, vect_get_vec_defs, vect_finish_replace_stmt)
6688         (vect_finish_stmt_generation): Call vinfo_for_stmt and pass
6689         stmt_vec_infos to subroutines.
6690         (vect_remove_stores): Use stmt_vec_infos to handle the statement
6691         chains.
6693 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
6695         * tree-vect-data-refs.c (vect_slp_analyze_node_dependences):
6696         (vect_check_gather_scatter, vect_create_data_ref_ptr, bump_vector_ptr)
6697         (vect_permute_store_chain, vect_setup_realignment)
6698         (vect_permute_load_chain, vect_shift_permute_load_chain)
6699         (vect_transform_grouped_load): Use stmt_vec_info rather than gimple
6700         stmts internally, and when passing values to other vectorizer routines.
6701         * tree-vect-loop-manip.c (vect_can_advance_ivs_p): Likewise.
6702         * tree-vect-loop.c (vect_analyze_scalar_cycles_1)
6703         (vect_analyze_loop_operations, get_initial_def_for_reduction)
6704         (vect_create_epilog_for_reduction, vectorize_fold_left_reduction)
6705         (vectorizable_reduction, vectorizable_induction)
6706         (vectorizable_live_operation, vect_transform_loop_stmt)
6707         (vect_transform_loop): Likewise.
6708         * tree-vect-patterns.c (vect_reassociating_reduction_p)
6709         (vect_recog_widen_op_pattern, vect_recog_mixed_size_cond_pattern)
6710         (vect_recog_bool_pattern, vect_recog_gather_scatter_pattern): Likewise.
6711         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
6712         (vect_slp_analyze_node_operations_1): Likewise.
6713         * tree-vect-stmts.c (vect_mark_relevant, process_use)
6714         (exist_non_indexing_operands_for_use_p, vect_init_vector_1)
6715         (vect_mark_stmts_to_be_vectorized, vect_get_vec_def_for_operand)
6716         (vect_finish_stmt_generation_1, get_group_load_store_type)
6717         (get_load_store_type, vect_build_gather_load_calls)
6718         (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
6719         (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
6720         (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
6721         (vectorizable_store, vectorizable_load, vectorizable_condition)
6722         (vectorizable_comparison, vect_analyze_stmt, vect_transform_stmt)
6723         (supportable_widening_operation): Likewise.
6724         (vect_get_vector_types_for_stmt): Likewise.
6725         * tree-vectorizer.h (vect_dr_behavior): Likewise.
6727 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
6729         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
6730         (vect_slp_analyze_node_dependences, vect_analyze_data_ref_accesses)
6731         (vect_permute_store_chain, vect_permute_load_chain)
6732         (vect_shift_permute_load_chain, vect_transform_grouped_load): Avoid
6733         repeated stmt_vec_info lookups.
6734         * tree-vect-loop-manip.c (vect_can_advance_ivs_p): Likewise.
6735         (vect_update_ivs_after_vectorizer): Likewise.
6736         * tree-vect-loop.c (vect_is_simple_reduction): Likewise.
6737         (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
6738         * tree-vect-patterns.c (adjust_bool_stmts): Likewise.
6739         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
6740         (vect_bb_slp_scalar_cost): Likewise.
6741         * tree-vect-stmts.c (get_group_alias_ptr_type): Likewise.
6743 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
6745         * tree-vect-data-refs.c (vect_check_gather_scatter): Pass the
6746         gcall rather than the generic gimple stmt to gimple_call_internal_fn.
6747         (vect_get_smallest_scalar_type, can_group_stmts_p): Use dyn_cast
6748         to get gassigns and gcalls, rather than operating on generc gimple
6749         stmts.
6750         * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p)
6751         (vect_mark_stmts_to_be_vectorized, vectorizable_store)
6752         (vectorizable_load, vect_analyze_stmt): Likewise.
6753         * tree-vect-loop.c (vectorizable_reduction): Likewise gphi.
6755 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
6757         * tree-vectorizer.h (get_earlier_stmt, get_later_stmt): Take and
6758         return stmt_vec_infos rather than gimple stmts.  Do not accept
6759         null arguments.
6760         (vect_find_last_scalar_stmt_in_slp): Return a stmt_vec_info instead
6761         of a gimple stmt.
6762         * tree-vect-slp.c (vect_find_last_scalar_stmt_in_slp): Likewise.
6763         Update use of get_later_stmt.
6764         (vect_get_constant_vectors): Update call accordingly.
6765         (vect_schedule_slp_instance): Likewise
6766         * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Likewise.
6767         (vect_slp_analyze_instance_dependence): Likewise.
6768         (vect_preserves_scalar_order_p): Update use of get_earlier_stmt.
6770 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
6772         * tree-vectorizer.h (stmt_info_for_cost::stmt): Replace with...
6773         (stmt_info_for_cost::stmt_info): ...this new field.
6774         (add_stmt_costs): Update accordingly.
6775         * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost)
6776         (vect_get_known_peeling_cost): Likewise.
6777         (vect_estimate_min_profitable_iters): Likewise.
6778         * tree-vect-stmts.c (record_stmt_cost): Likewise.
6780 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
6782         * tree-vectorizer.h (_loop_vec_info::may_misalign_stmts): Change
6783         from an auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
6784         * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Update
6785         accordingly.
6786         * tree-vect-loop-manip.c (vect_create_cond_for_align_checks): Likewise.
6788 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
6790         * tree-vectorizer.h (_stmt_vec_info::same_dr_stmt): Change from
6791         a gimple stmt to a stmt_vec_info.
6792         * tree-vect-stmts.c (vectorizable_load): Update accordingly.
6794 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
6796         * tree-vectorizer.h (vec_info::grouped_stores): Change from
6797         an auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
6798         (_loop_vec_info::reduction_chains): Likewise.
6799         * tree-vect-loop.c (vect_fixup_scalar_cycles_with_patterns): Update
6800         accordingly.
6801         * tree-vect-slp.c (vect_analyze_slp): Likewise.
6803 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
6805         * tree-vectorizer.h (_stmt_vec_info::first_element): Change from
6806         a gimple stmt to a stmt_vec_info.
6807         (_stmt_vec_info::next_element): Likewise.
6808         * tree-vect-data-refs.c (vect_update_misalignment_for_peel)
6809         (vect_slp_analyze_and_verify_node_alignment)
6810         (vect_analyze_group_access_1, vect_analyze_group_access)
6811         (vect_small_gap_p, vect_prune_runtime_alias_test_list)
6812         (vect_create_data_ref_ptr, vect_record_grouped_load_vectors)
6813         (vect_supportable_dr_alignment): Update accordingly.
6814         * tree-vect-loop.c (vect_fixup_reduc_chain): Likewise.
6815         (vect_fixup_scalar_cycles_with_patterns, vect_is_slp_reduction)
6816         (vect_is_simple_reduction, vectorizable_reduction): Likewise.
6817         * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
6818         * tree-vect-slp.c (vect_build_slp_tree_1)
6819         (vect_attempt_slp_rearrange_stmts, vect_supported_load_permutation_p)
6820         (vect_split_slp_store_group, vect_analyze_slp_instance)
6821         (vect_analyze_slp, vect_transform_slp_perm_load): Likewise.
6822         * tree-vect-stmts.c (vect_model_store_cost, vect_model_load_cost)
6823         (get_group_load_store_type, get_load_store_type)
6824         (get_group_alias_ptr_type, vectorizable_store, vectorizable_load)
6825         (vect_transform_stmt, vect_remove_stores): Likewise.
6827 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
6829         * tree-vectorizer.h (vect_dr_stmt): Return a stmt_vec_info rather
6830         than a gimple stmt.
6831         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
6832         (vect_slp_analyze_data_ref_dependence, vect_record_base_alignments)
6833         (vect_calculate_target_alignmentm, vect_compute_data_ref_alignment)
6834         (vect_update_misalignment_for_peel, vect_verify_datarefs_alignment)
6835         (vector_alignment_reachable_p, vect_get_data_access_cost)
6836         (vect_get_peeling_costs_all_drs, vect_peeling_hash_get_lowest_cost)
6837         (vect_peeling_supportable, vect_enhance_data_refs_alignment)
6838         (vect_find_same_alignment_drs, vect_analyze_data_refs_alignment)
6839         (vect_analyze_group_access_1, vect_analyze_group_access)
6840         (vect_analyze_data_ref_access, vect_analyze_data_ref_accesses)
6841         (vect_vfa_access_size, vect_small_gap_p, vect_analyze_data_refs)
6842         (vect_supportable_dr_alignment): Remove vinfo_for_stmt from the
6843         result of vect_dr_stmt and use the stmt_vec_info instead of
6844         the associated gimple stmt.
6845         * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
6846         (vect_gen_prolog_loop_niters): Likewise.
6847         * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
6849 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
6851         * tree-vectorizer.h (_slp_tree::stmts): Change from a vec<gimple *>
6852         to a vec<stmt_vec_info>.
6853         * tree-vect-slp.c (vect_free_slp_tree): Update accordingly.
6854         (vect_create_new_slp_node): Take a vec<gimple *> instead of a
6855         vec<stmt_vec_info>.
6856         (_slp_oprnd_info::def_stmts): Change from a vec<gimple *>
6857         to a vec<stmt_vec_info>.
6858         (bst_traits::value_type, bst_traits::value_type): Likewise.
6859         (bst_traits::hash): Update accordingly.
6860         (vect_get_and_check_slp_defs): Change the stmts parameter from
6861         a vec<gimple *> to a vec<stmt_vec_info>.
6862         (vect_two_operations_perm_ok_p, vect_build_slp_tree_1): Likewise.
6863         (vect_build_slp_tree): Likewise.
6864         (vect_build_slp_tree_2): Likewise.  Update uses of
6865         SLP_TREE_SCALAR_STMTS.
6866         (vect_print_slp_tree): Update uses of SLP_TREE_SCALAR_STMTS.
6867         (vect_mark_slp_stmts, vect_mark_slp_stmts_relevant)
6868         (vect_slp_rearrange_stmts, vect_attempt_slp_rearrange_stmts)
6869         (vect_supported_load_permutation_p, vect_find_last_scalar_stmt_in_slp)
6870         (vect_detect_hybrid_slp_stmts, vect_slp_analyze_node_operations_1)
6871         (vect_slp_analyze_node_operations, vect_slp_analyze_operations)
6872         (vect_bb_slp_scalar_cost, vect_slp_analyze_bb_1)
6873         (vect_get_constant_vectors, vect_get_slp_defs)
6874         (vect_transform_slp_perm_load, vect_schedule_slp_instance)
6875         (vect_remove_slp_scalar_calls, vect_schedule_slp): Likewise.
6876         (vect_analyze_slp_instance): Build up a vec of stmt_vec_infos
6877         instead of gimple stmts.
6878         * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Change
6879         the stores parameter for a vec<gimple *> to a vec<stmt_vec_info>.
6880         (vect_slp_analyze_instance_dependence): Update uses of
6881         SLP_TREE_SCALAR_STMTS.
6882         (vect_slp_analyze_and_verify_node_alignment): Likewise.
6883         (vect_slp_analyze_and_verify_instance_alignment): Likewise.
6884         * tree-vect-loop.c (neutral_op_for_slp_reduction): Likewise.
6885         (get_initial_defs_for_reduction): Likewise.
6886         (vect_create_epilog_for_reduction): Likewise.
6887         (vectorize_fold_left_reduction): Likewise.
6888         * tree-vect-stmts.c (vect_prologue_cost_for_slp_op): Likewise.
6889         (vect_model_simple_cost, vectorizable_shift, vectorizable_load)
6890         (can_vectorize_live_stmts): Likewise.
6892 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
6894         * tree-vectorizer.h (_loop_vec_info::reductions): Change from an
6895         auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
6896         (vect_force_simple_reduction): Take and return stmt_vec_infos rather
6897         than gimple stmts.
6898         * tree-parloops.c (valid_reduction_p): Take a stmt_vec_info instead
6899         of a gimple stmt.
6900         (gather_scalar_reductions): Update after above interface changes.
6901         * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Likewise.
6902         (vect_is_simple_reduction): Take and return stmt_vec_infos rather
6903         than gimple stmts.
6904         (vect_force_simple_reduction): Likewise.
6905         * tree-vect-patterns.c (vect_pattern_recog_1): Update use of
6906         LOOP_VINFO_REDUCTIONS.
6907         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
6909 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
6911         * tree-vectorizer.h (_stmt_vec_info::reduc_def): Change from
6912         a gimple stmt to a stmt_vec_info.
6913         * tree-vect-loop.c (vect_active_double_reduction_p)
6914         (vect_force_simple_reduction, vectorizable_reduction): Update
6915         accordingly.
6917 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
6919         * tree-vectorizer.h (_slp_tree::vec_stmts): Change from a
6920         vec<gimple *> to a vec<stmt_vec_info>.
6921         * tree-vect-loop.c (vect_create_epilog_for_reduction): Change
6922         the reduction_phis argument from a vec<gimple *> to a
6923         vec<stmt_vec_info>.
6924         (vectorizable_reduction): Likewise the phis local variable that
6925         is passed to vect_create_epilog_for_reduction.  Update for new type
6926         of SLP_TREE_VEC_STMTS.
6927         (vectorizable_induction): Update for new type of SLP_TREE_VEC_STMTS.
6928         (vectorizable_live_operation): Likewise.
6929         * tree-vect-slp.c (vect_get_slp_vect_defs): Likewise.
6930         (vect_transform_slp_perm_load, vect_schedule_slp_instance): Likewise.
6932 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
6934         * tree-vectorizer.h (_stmt_vec_info::vectorized_stmt): Change from
6935         a gimple stmt to a stmt_vec_info.
6936         (vectorizable_condition, vectorizable_live_operation)
6937         (vectorizable_reduction, vectorizable_induction): Pass back the
6938         vectorized statement as a stmt_vec_info.
6939         * tree-vect-data-refs.c (vect_record_grouped_load_vectors): Update
6940         use of STMT_VINFO_VEC_STMT.
6941         * tree-vect-loop.c (vect_create_epilog_for_reduction): Likewise,
6942         accumulating the inner phis that feed the STMT_VINFO_VEC_STMT
6943         as stmt_vec_infos rather than gimple stmts.
6944         (vectorize_fold_left_reduction): Change vec_stmt from a gimple stmt
6945         to a stmt_vec_info.
6946         (vectorizable_live_operation): Likewise.
6947         (vectorizable_reduction, vectorizable_induction): Likewise,
6948         updating use of STMT_VINFO_VEC_STMT.
6949         * tree-vect-stmts.c (vect_get_vec_def_for_operand_1): Update use
6950         of STMT_VINFO_VEC_STMT.
6951         (vect_build_gather_load_calls, vectorizable_bswap, vectorizable_call)
6952         (vectorizable_simd_clone_call, vectorizable_conversion)
6953         (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
6954         (vectorizable_store, vectorizable_load, vectorizable_condition)
6955         (vectorizable_comparison, can_vectorize_live_stmts): Change vec_stmt
6956         from a gimple stmt to a stmt_vec_info.
6957         (vect_transform_stmt): Update use of STMT_VINFO_VEC_STMT.  Pass a
6958         pointer to a stmt_vec_info to the vectorizable_* routines.
6960 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
6962         * tree-vectorizer.h (_stmt_vec_info::related_stmt): Change from
6963         a gimple stmt to a stmt_vec_info.
6964         (is_pattern_stmt_p): Update accordingly.
6965         * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Likewise.
6966         (vect_record_grouped_load_vectors): Likewise.
6967         * tree-vect-loop.c (vect_determine_vf_for_stmt): Likewise.
6968         (vect_fixup_reduc_chain, vect_update_vf_for_slp): Likewise.
6969         (vect_model_reduction_cost): Likewise.
6970         (vect_create_epilog_for_reduction): Likewise.
6971         (vectorizable_reduction, vectorizable_induction): Likewise.
6972         * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
6973         Return the stmt_vec_info for the pattern statement.
6974         (vect_set_pattern_stmt): Update use of STMT_VINFO_RELATED_STMT.
6975         (vect_split_statement, vect_mark_pattern_stmts): Likewise.
6976         * tree-vect-slp.c (vect_detect_hybrid_slp_stmts): Likewise.
6977         (vect_detect_hybrid_slp, vect_get_slp_defs): Likewise.
6978         * tree-vect-stmts.c (vect_mark_relevant): Likewise.
6979         (vect_get_vec_def_for_operand_1, vectorizable_call): Likewise.
6980         (vectorizable_simd_clone_call, vect_analyze_stmt, new_stmt_vec_info)
6981         (free_stmt_vec_info, vect_is_simple_use): Likewise.
6983 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
6985         * tree-vectorizer.h (vect_finish_replace_stmt): Return a stmt_vec_info
6986         (vect_finish_stmt_generation): Likewise.
6987         * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
6988         (vect_finish_replace_stmt, vect_finish_stmt_generation): Likewise.
6989         (vect_build_gather_load_calls): Use the return value of the above
6990         functions instead of a separate call to vinfo_for_stmt.  Use narrow
6991         scopes for the input gimple stmt and wider scopes for the associated
6992         stmt_vec_info.  Use vec_info::lookup_def when setting these
6993         stmt_vec_infos from an SSA_NAME definition.
6994         (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
6995         (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
6996         (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
6997         (vectorizable_store, vectorizable_load, vectorizable_condition)
6998         (vectorizable_comparison): Likewise.
6999         * tree-vect-loop.c (vectorize_fold_left_reduction): Likewise.
7000         (vectorizable_reduction): Likewise.
7002 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
7004         * tree-vectorizer.h (vect_is_simple_use): Add an optional
7005         stmt_vec_info * parameter before the optional gimple **.
7006         * tree-vect-stmts.c (vect_is_simple_use): Likewise.
7007         (process_use, vect_get_vec_def_for_operand_1): Update callers.
7008         (vect_get_vec_def_for_operand, vectorizable_shift): Likewise.
7009         * tree-vect-loop.c (vectorizable_reduction): Likewise.
7010         (vectorizable_live_operation): Likewise.
7011         * tree-vect-patterns.c (type_conversion_p): Likewise.
7012         (vect_look_through_possible_promotion): Likewise.
7013         (vect_recog_rotate_pattern): Likewise.
7014         * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
7016 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
7018         * tree-vectorizer.h (stmt_vec_info): Temporarily change from
7019         a typedef to a wrapper class.
7020         (NULL_STMT_VEC_INFO): New macro.
7021         (vec_info::stmt_infos): Change to vec<stmt_vec_info>.
7022         (stmt_vec_info::operator*): New function.
7023         (stmt_vec_info::operator gimple *): Likewise.
7024         (set_vinfo_for_stmt): Use NULL_STMT_VEC_INFO.
7025         (add_stmt_costs): Likewise.
7026         * tree-vect-loop-manip.c (iv_phi_p): Likewise.
7027         * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost)
7028         (vect_get_known_peeling_cost): Likewise.
7029         (vect_estimate_min_profitable_iters): Likewise.
7030         * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
7031         * tree-vect-slp.c (vect_remove_slp_scalar_calls): Likewise.
7032         * tree-vect-stmts.c (vect_build_gather_load_calls): Likewise.
7033         (vectorizable_store, free_stmt_vec_infos): Likewise.
7034         (new_stmt_vec_info): Change return type of xcalloc to
7035         _stmt_vec_info *.
7037 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
7039         * tree-vectorizer.h (vec_info::lookup_single_use): Declare.
7040         * tree-vectorizer.c (vec_info::lookup_single_use): New function.
7041         * tree-vect-loop.c (vectorizable_reduction): Use it instead of
7042         a single_imm_use-based sequence.
7043         * tree-vect-stmts.c (supportable_widening_operation): Likewise.
7045 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
7047         * tree-vectorizer.h (vec_info::lookup_def): Declare.
7048         * tree-vectorizer.c (vec_info::lookup_def): New function.
7049         * tree-vect-patterns.c (vect_get_internal_def): Use it.
7050         (vect_widened_op_tree): Likewise.
7051         * tree-vect-stmts.c (vect_is_simple_use): Likewise.
7052         * tree-vect-loop.c (vect_analyze_loop_operations): Likewise.
7053         (vectorizable_reduction): Likewise.
7054         (vect_valid_reduction_input_p): Take a stmt_vec_info instead
7055         of a gimple *.
7056         (vect_is_slp_reduction): Update calls accordingly.  Use
7057         vec_info::lookup_def.
7058         (vect_is_simple_reduction): Likewise
7059         * tree-vect-slp.c (vect_detect_hybrid_slp_1): Use vec_info::lookup_def.
7061 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
7063         * tree-vectorizer.h (vec_info::lookup_stmt): Declare.
7064         * tree-vectorizer.c (vec_info::lookup_stmt): New function.
7065         * tree-vect-loop.c (vect_determine_vf_for_stmt): Use it instead
7066         of vinfo_for_stmt.
7067         (vect_determine_vectorization_factor, vect_analyze_scalar_cycles_1)
7068         (vect_compute_single_scalar_iteration_cost, vect_analyze_loop_form)
7069         (vect_update_vf_for_slp, vect_analyze_loop_operations)
7070         (vect_is_slp_reduction, vectorizable_induction)
7071         (vect_transform_loop_stmt, vect_transform_loop): Likewise.
7072         * tree-vect-patterns.c (vect_init_pattern_stmt):
7073         (vect_determine_min_output_precision_1, vect_determine_precisions)
7074         (vect_pattern_recog): Likewise.
7075         * tree-vect-stmts.c (vect_analyze_stmt, vect_transform_stmt): Likewise.
7076         * config/powerpcspe/powerpcspe.c (rs6000_density_test): Likewise.
7077         * config/rs6000/rs6000.c (rs6000_density_test): Likewise.
7078         * tree-vect-slp.c (vect_detect_hybrid_slp_stmts): Likewise.
7079         (vect_detect_hybrid_slp_1, vect_detect_hybrid_slp_2)
7080         (vect_detect_hybrid_slp): Likewise.  Change the walk_stmt_info
7081         info field from a loop to a loop_vec_info.
7083 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
7085         * tree-vectorizer.h (stmt_vec_info): Move typedef earlier in file.
7086         (vec_info::add_stmt): Declare.
7087         * tree-vectorizer.c (vec_info::add_stmt): New function.
7088         * tree-vect-data-refs.c (vect_create_data_ref_ptr): Use it.
7089         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Likewise.
7090         (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
7091         (vectorizable_induction): Likewise.
7092         * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Likewise.
7093         * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
7094         (vectorizable_simd_clone_call, vectorizable_store): Likewise.
7095         (vectorizable_load): Likewise.
7096         * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
7097         (vect_recog_bool_pattern, vect_recog_mask_conversion_pattern)
7098         (vect_recog_gather_scatter_pattern): Likewise.
7099         (append_pattern_def_seq): Likewise.  Remove a check that is
7100         performed by add_stmt itself.
7102 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
7104         * tree-vect-loop.c (vectorizable_reduction): Fix an instance in
7105         which make_ssa_name was called with new_stmt before new_stmt
7106         had been created.
7108 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
7110         * tree-vect-loop.c (vect_valid_reduction_input_p): New function,
7111         split out from...
7112         (vect_is_slp_reduction): ...here...
7113         (vect_is_simple_reduction): ...and here.  Remove repetition of tests
7114         that are already known to be false.
7116 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
7118         * tree-vectorizer.h (vect_free_slp_instance): Add a final_p parameter.
7119         * tree-vect-slp.c (vect_free_slp_tree): Likewise.  Don't update
7120         STMT_VINFO_NUM_SLP_USES when it's true.
7121         (vect_free_slp_instance): Add a final_p parameter and pass it to
7122         vect_free_slp_tree.
7123         (vect_build_slp_tree_2): Update call to vect_free_slp_instance.
7124         (vect_analyze_slp_instance): Likewise.
7125         (vect_slp_analyze_operations): Likewise.
7126         (vect_slp_analyze_bb_1): Likewise.
7127         * tree-vectorizer.c (vec_info): Likewise.
7128         * tree-vect-loop.c (vect_transform_loop): Likewise.
7130 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
7132         * tree-vect-loop.c (vectorizable_reduction): Assert that the
7133         function is not called for second and subsequent members of
7134         a reduction group.
7136 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
7138         * tree-vect-loop.c (get_initial_def_for_reduction): Move special
7139         cases for nested loops from here to ...
7140         (vect_create_epilog_for_reduction): ...here.  Only call
7141         vect_is_simple_use for inner-loop reductions.
7143 2018-07-31  Martin Liska  <mliska@suse.cz>
7145         PR gcov-profile/85338
7146         PR gcov-profile/85350
7147         PR gcov-profile/85372
7148         * profile.c (struct location_triplet): New.
7149         (struct location_triplet_hash): Likewise.
7150         (output_location): Do not output a BB that
7151         is already recorded for a line.
7152         (branch_prob): Use streamed_locations.
7154 2018-07-31  Martin Liska  <mliska@suse.cz>
7156         PR gcov-profile/85370
7157         * coverage.c (coverage_begin_function): Do not mark target
7158         clones as artificial functions.
7160 2018-07-31  Martin Liska  <mliska@suse.cz>
7162         PR gcov-profile/83813
7163         PR gcov-profile/84758
7164         PR gcov-profile/85217
7165         PR gcov-profile/85332
7166         * profile.c (branch_prob): Do not record GOTO expressions
7167         for GIMPLE statements which locations are already streamed.
7169 2018-07-31  Olivier Hainque  <hainque@adacore.com>
7171         * gcc.c (handle_spec_function): Accept a soft_matched_part
7172         argument, as do_spec_1.  Pass it down to ...
7173         (eval_spec_function): Accept a soft_matched_part argument,
7174         and pass it down to ...
7175         (do_spec_2): Accept a soft_matched_part argument, and pass
7176         it down to do_spec_1.
7177         (do_spec_1): Pass soft_matched_part to handle_spec_function.
7178         (handle_braces): Update call to handle_spec_function.
7179         (driver::set_up_specs): Update calls to do_spec_2.
7180         (compare_debug_dump_opt_spec_function): Likewise.
7181         (compare_debug_self_opt_spec_function): Likewise.
7183 2018-07-31  Olivier Hainque  <hainque@adacore.com>
7185         * common.opt (nolibc): New option.
7186         * doc/invoke.texi (Link Options): Document it.
7187         * gcc.c (LINK_GCC_C_SEQUENCE_SPEC): Honor nolibc.
7188         * config/alpha/linux.h: Likewise.
7189         * config/arc/elf.h: Likewise.
7190         * config/arm/uclinux-elf.h: Likewise.
7191         * config/arm/unknown-elf.h: Likewise.
7192         * config/avr/avrlibc.h: Likewise.
7193         * config/bfin/bfin.h: Likewise.
7194         * config/bfin/linux.h: Likewise.
7195         * config/bfin/uclinux.h: Likewise.
7196         * config/darwin.h: Likewise.
7197         * config/darwin10.h: Likewise.
7198         * config/darwin12.h: Likewise.
7199         * config/gnu-user.h: Likewise.
7200         * config/lm32/uclinux-elf.h: Likewise.
7201         * config/pa/pa-hpux11.h: Likewise.
7202         * config/pa/pa64-hpux.h: Likewise.
7203         * config/sparc/sparc.h: Likewise.
7205 2018-07-31  Olivier Hainque  <hainque@adacore.com>
7207         * gcc.c (getenv_spec_function): Prepend '/' to value for allowed
7208         undefined variables.
7210 2018-07-30  Segher Boessenkool  <segher@kernel.crashing.org>
7212         PR target/86640
7213         * config/arm/arm.c (arm_block_set_aligned_vect): Use gen_int_mode
7214         instead of GEN_INT.
7216 2018-07-30  Bernd Edlinger  <bernd.edlinger@hotmail.de>
7218         * tree-ssa-forwprop.c (simplify_builtin_call): Don't create a not NUL
7219         terminated string literal.
7221 2018-07-30  Segher Boessenkool  <segher@kernel.crashing.org>
7223         PR rtl-optimization/85160
7224         * combine.c (is_just_move): New function.
7225         (try_combine): Allow combining two instructions into two if neither of
7226         the original instructions was a move.
7228 2018-07-30  Alexander Monakov  <amonakov@ispras.ru>
7230         PR target/86673
7231         * doc/extend.texi (Global Register Variables): Discourage use of type
7232         qualifiers.
7233         (Local Register Variables): Likewise.
7235 2018-07-30  Richard Sandiford  <richard.sandiford@arm.com>
7237         PR tree-optimization/86506
7238         * hwint.h (ceil_log2): Resync with hwint.c implementation.
7240 2018-07-30  Ilya Leoshkevich  <iii@linux.ibm.com>
7242         PR target/86547
7243         * lra-constraints.c (spill_hard_reg_in_range): When selecting the
7244         hard_regno, make sure no insn between `from` and `to` clobbers it.
7246 2018-07-30  Cesar Philippidis  <cesar@codesourcery.com>
7247             Tom de Vries  <tdevries@suse.de>
7249         * config/nvptx/nvptx.c (PTX_GANG_DEFAULT): Rename to ...
7250         (PTX_DEFAULT_RUNTIME_DIM): ... this.
7251         (nvptx_goacc_validate_dims): Set default worker and gang dims to
7252         PTX_DEFAULT_RUNTIME_DIM.
7253         (nvptx_dim_limit): Ignore GOMP_DIM_WORKER.
7255 2018-07-29  John David Anglin  <danglin@gcc.gnu.org>
7257         * config/pa/pa.c (pa_output_addr_vec): Align address table.
7258         * config/pa/pa.h (JUMP_TABLES_IN_TEXT_SECTION): Revise comment.
7259         * config/pa/pa32-linux.h (JUMP_TABLES_IN_TEXT_SECTION): Define.
7261 2018-07-27  Michael Meissner  <meissner@linux.ibm.com>
7263         * config/rs6000/constraints.md (wG constraint): Delete, no longer
7264         used.
7265         * config/rs6000/predicates.md (p9_fusion_reg_operand): Rename
7266         predicate to reflect toc fusion has been deleted.
7267         (toc_fusion_mem_raw): Delete, no longer used.
7268         (toc_fusion_mem_wrapped): Likewise.
7269         * config/rs6000/rs6000-cpus.def (POWERPC_MASKS): Delete toc
7270         fusion mask bit.
7271         * config/rs6000/rs6000-protos.h (fusion_wrap_memory_address):
7272         Delete, no longer used.
7273         * config/rs6000/rs6000.c (struct rs6000_reg_addr): Delete fields
7274         meant to be used for toc fusion.
7275         (rs6000_debug_print_mode): Delete toc fusion debugging.
7276         (rs6000_debug_reg_global): Likewise.
7277         (rs6000_init_hard_regno_mode_ok): Delete setting up fields for toc
7278         fusion and secondary reload support that were never used.
7279         (rs6000_option_override_internal): Delete TOC fusion, that was only
7280         partially defined, and it did not work unless you also used the
7281         -mcmodel= switch.
7282         (rs6000_legitimate_address_p): Delete TOC fusion support.
7283         (rs6000_opt_masks): Likewise.
7284         (fusion_wrap_memory_address): Delete function, no longer used.
7285         (fusion_split_address); Delete TOC fusion support.
7286         * config/rs6000/rs6000.h (TARGET_TOC_FUSION_INT): Delete, no
7287         longer used with toc fusion being deleted.
7288         (TARGET_TOC_FUSION_FP): Likewise.
7289         * config/rs6000/rs6000.md (UNSPEC_FUSION_ADDIS): Delete TOC fusion
7290         UNSPEC.
7291         (toc fusion spliter): Delete TOC fusion support.
7292         (toc_fusionload_<mode>): Likewise.
7293         (toc_fusionload_di): Likewise.
7294         (fusion_gpr_load_<mode>): Delete generator function, this insn no
7295         longer needs to be named.  Rename predicate to delete TOC fusion.
7296         (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_load): Likewise.
7297         (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_store): Likewise.
7298         (fusion_vsx_<P:mode>_<GPR_FUSION:mode>_load): Likewise.
7299         (fusion_vsx_<P:mode>_<GPR_FUSION:mode>_store): Likewise.
7300         (p9 fusion peephole2s): Rename predicate to delete TOC fusion.
7302 2018-07-27  Kelvin Nilsen  <kelvin@gcc.gnu.org>
7304         * doc/extend.texi (Basic PowerPC Built-in Functions Available on
7305         ISA 2.05):  Replace __uint128_t with __uint128 and __int128_t with
7306         __int128 in built-in function prototypes.
7307         (PowerPC AltiVec Built-in Functions on ISA 2.07): Likewise.
7308         (PowerPC AltiVec Built-in Functions on ISA 3.0): Likewise.
7310 2018-07-27  Martin Sebor  <msebor@redhat.com>
7312         PR tree-optimization/86696
7313         * tree-ssa-strlen.c (get_min_string_length): Handle all integer
7314         types, including enums.
7315         (handle_char_store): Be prepared for the above function to fail.
7317 2018-07-26  Qing Zhao  <qing.zhao@oracle.com>
7319         * builtins.c (inline_expand_builtin_string_cmp): Disable inlining
7320         when optimization level is lower than 2 or optimize for size.
7322 2018-07-26  Martin Sebor  <msebor@redhat.com>
7324         PR tree-optimization/86043
7325         PR tree-optimization/86042
7326         * tree-ssa-strlen.c (handle_builtin_memcpy): Handle strict overlaps.
7327         (get_string_cst_length): Rename...
7328         (get_min_string_length): ...to this.  Add argument.
7329         (handle_char_store): Extend to handle multi-character stores by
7330         MEM_REF.
7331         * tree.c (initializer_zerop): Use new argument.  Handle MEM_REF.
7332         * tree.h (initializer_zerop): Add argument.
7334 2018-07-26  Jakub Jelinek  <jakub@redhat.com>
7336         PR middle-end/86660
7337         * omp-low.c (scan_sharing_clauses): Don't ignore map clauses for
7338         declare target to variables if they have always,{to,from,tofrom} map
7339         kinds.
7341 2018-07-26  Martin Liska  <mliska@suse.cz>
7343         PR lto/86548
7344         * lto-wrapper.c: Add linker_output as prefix
7345         for ltrans_output_file.
7347 2018-07-26  Segher Boessenkool  <segher@kernel.crashing.org>
7349         PR rtl-optimization/85805
7350         * combine.c (reg_nonzero_bits_for_combine): Only use the last set
7351         value for hard registers if that was written in the same mode.
7353 2018-07-26  Martin Liska  <mliska@suse.cz>
7355         PR gcov-profile/86536
7356         * gcov.c (format_gcov): Use printf format %.*f directly
7357         and do not handle special values.
7359 2018-07-25  Claudiu Zissulescu  <claziss@synopsys.com>
7361         * common/config/arc/arc-common.c (arc_option_optimization_table):
7362         Update default optimizations for size.
7364 2018-07-25  Claudiu Zissulescu  <claziss@synopsys.com>
7366         * config/arc/arc.md (movsf_insn): Add short instruction selection.
7367         * config/arc/constraints.md (CfZ): New constraint.
7368         * config/arc/fpu.md (addssf3_fpu): Use CfZ constraint.
7369         (subsf3_fpu): Likewise.
7370         (cmpsf_fpu): Likewise.
7371         (cmpsf_fpu_uneq): Likewise.
7373 2018-07-25  Claudiu Zissulescu  <claziss@synopsys.com>
7375         * config/arc/arc.c (compact_memory_operand_p): Check for uncached
7376         accesses as well.
7377         (arc_is_uncached_mem_p): uncached applies to both the variable and
7378         the pointer.
7380 2018-07-25  Claudiu Zissulescu  <claziss@synopsys.com>
7382         * config/arc/arc.h (ADDITIONAL_REGISTER_NAMES): Add additional
7383         register names.
7385 2018-07-25  David Malcolm  <dmalcolm@redhat.com>
7387         * optinfo-emit-json.cc (class optrecord_json_writer): Convert
7388         field "m_scopes" from vec to auto_vec.
7390 2018-07-25  Martin Liska  <mliska@suse.cz>
7392         * config/powerpcspe/powerpcspe-protos.h (rs6000_loop_align): Fix
7393         return type.
7395 2018-07-25  Richard Biener  <rguenther@suse.de>
7397         PR debug/86654
7398         * dwarf2out.c (dwarf2out_decl): Do not handle nested functions
7399         special wrt context_die late.
7400         (gen_subprogram_die): Re-use DIEs in local scope.
7402 2018-07-25  Richard Sandiford  <richard.sandiford@arm.com>
7404         PR tree-optimization/86644
7405         * hwint.c (ceil_log2): Fix comment.  Return 0 for 0.
7407 2018-07-25  Martin Liska  <mliska@suse.cz>
7409         PR middle-end/86645
7410         * dumpfile.c: And excluded values with TDF_ALL_VALUES.
7411         * dumpfile.h (enum dump_flag): Defince TDF_ALL_VALUES.
7413 2018-07-25  Martin Liska  <mliska@suse.cz>
7415         PR sanitizer/79635
7416         * params.def: Explain ASan abbreviation and provide
7417         a documentation link.
7419 2018-07-24  Martin Sebor  <msebor@redhat.com>
7421         PR tree-optimization/86622
7422         PR tree-optimization/86532
7423         * builtins.h (string_length): Declare.
7424         * builtins.c (c_strlen): Correct handling of non-constant offsets.
7425         (check_access): Be prepared for non-constant length ranges.
7426         (string_length): Make extern.
7427         * expr.c (string_constant): Only handle the minor non-constant
7428         array index.  Use string_constant to compute the length of
7429         a generic string constant.
7431 2018-07-24  Richard Sandiford  <richard.sandiford@arm.com>
7433         PR tree-optimization/86618
7434         * tree-vect-stmts.c (vectorizable_call): Don't take the address
7435         of LOOP_VINFO_MASKS (loop_vinfo) when loop_vinfo is null.
7437 2018-07-24  David Malcolm  <dmalcolm@redhat.com>
7439         PR tree-optimization/86636
7440         * json.cc (json::object::set): Fix comment.  Add assertions.
7441         (json::array::append): Move here from json.h.  Add comment and an
7442         assertion.
7443         (json::string::string): Likewise.
7444         * json.h (json::array::append): Move to json.cc.
7445         (json::string::string): Likewise.
7446         * optinfo-emit-json.cc
7447         (optrecord_json_writer::impl_location_to_json): Assert that we
7448         aren't attempting to write out UNKNOWN_LOCATION, or an ad-hoc
7449         wrapper around it.  Expand the location once, rather than three
7450         times.
7451         (optrecord_json_writer::inlining_chain_to_json): Fix the check for
7452         UNKNOWN_LOCATION, to use LOCATION_LOCUS to look through ad-hoc
7453         wrappers.
7454         (optrecord_json_writer::optinfo_to_json): Likewise, in four
7455         places.  Fix some overlong lines.
7457 2018-07-24  Matthew Malcomson  <matthew.malcomson@arm.com>
7459         * config/aarch64/aarch64-simd.md
7460         (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w<mode>): Split into...
7461         (aarch64_<ANY_EXTEND:su>subw<mode>): ... This...
7462         (aarch64_<ANY_EXTEND:su>addw<mode>): ... And this.
7463         (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w<mode>_internal): Split into...
7464         (aarch64_<ANY_EXTEND:su>subw<mode>_internal): ... This...
7465         (aarch64_<ANY_EXTEND:su>addw<mode>_internal): ... And this.
7466         (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w2<mode>_internal): Split into...
7467         (aarch64_<ANY_EXTEND:su>subw2<mode>_internal): ... This...
7468         (aarch64_<ANY_EXTEND:su>addw2<mode>_internal): ... And this.
7470 2018-07-24  Jakub Jelinek  <jakub@redhat.com>
7472         PR middle-end/86627
7473         * expmed.c (expand_divmod): Punt if d == HOST_WIDE_INT_MIN
7474         and size > HOST_BITS_PER_WIDE_INT.  For size > HOST_BITS_PER_WIDE_INT
7475         and abs_d == d, do the power of two handling if profitable.
7477 2018-07-24  Richard Biener  <rguenther@suse.de>
7479         * match.pd: Add BIT_FIELD_REF canonicalizations.
7481 2018-07-23  Bernd Edlinger  <bernd.edlinger@hotmail.de>
7483         PR c/86617
7484         * genmatch.c (dt_operand::gen_match_op): Avoid folding volatile values.
7486 2018-07-23  Bernd Edlinger  <bernd.edlinger@hotmail.de>
7488         * gimple-fold.c (gimple_fold_builtin_printf): Don't create a not NUL
7489         terminated STRING_CST object.
7491 2018-07-23  Bernd Edlinger  <bernd.edlinger@hotmail.de>
7493         hsa-dump.c (dump_hsa_symbol): Avoid out of scope access to buf.
7495 2018-07-23  Segher Boessenkool  <segher@kernel.crashing.org>
7497         * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Adjust.
7498         * config/rs6000/rs6000-protos.h (rs6000_split_v4si_init): Delete.
7499         * config/rs6000/rs6000.c (rs6000_expand_vector_init): Always force
7500         the elements into a register.
7501         (rs6000_split_v4si_init_di_reg): Delete.
7502         (rs6000_split_v4si_init): Delete.
7503         * config/rs6000/vsx.md (unspec): Delete UNSPEC_VSX_VEC_INIT.
7504         (vsx_init_v4si): Rewrite as a define_expand.
7506 2018-07-23  Segher Boessenkool  <segher@kernel.crashing.org>
7508         * config/rs6000/rs6000.md (splitters for rldimi and rlwimi with the
7509         zero_extend argument from memory): New.
7511 2018-07-22  Martin Sebor  <msebor@redhat.com>
7513         PR bootstrap/86621
7514         * gimple-ssa-warn-alloca.c (alloca_call_type_by_arg): Avoid
7515         diagnosing calls with unknown arguments unless -Walloca-larger-than
7516         is restricted to less than PTRDIFF_MAX bytes.
7518 2018-07-22  Gerald Pfeifer  <gerald@pfeifer.com>
7520         * doc/gcov.texi (Invoking Gcov): Editorial changes.
7522 2018-07-20  David Malcolm  <dmalcolm@redhat.com>
7524         * pretty-print.c (text_info::set_location): Remove redundant
7525         "line_table" parameter from call to rich_location::set_range.
7527 2018-07-20  Martin Sebor  <msebor@redhat.com>
7529         PR middle-end/82063
7530         * builtins.c (expand_builtin_alloca): Adjust.
7531         * calls.c (alloc_max_size): Simplify.
7532         * cgraphunit.c (cgraph_node::expand): Adjust.
7533         * common.opt (larger_than_size, warn_frame_larger_than): Remove
7534         variables.
7535         (frame_larger_than_size): Same.
7536         (-Wframe-larger-than, -Wlarger-than, -Wstack-usage): Change options
7537         to take a HOST_WIDE_INT argument and accept a byte-size suffix.
7538         Initialize.
7539         * doc/invoke.texi (GCC Command Options): Document option arguments.
7540         Explain byte-size arguments and suffixes.
7541         (-Wvla-larger-than, -Wno-alloc-size-larger-than): Update.
7542         (-Wno-alloca-larger-than, -Wno-vla-larger-than): Same.
7543         (-Wframe-larger-than, -Wlarger-than, -Wstack-usage): Same.
7544         * doc/options.texi (UInteger): Expand.
7545         (Host_Wide_Int, ByteSize): Document new properties.
7546         * final.c (final_start_function_1): Include sizes in an error message.
7547         * function.c (frame_offset_overflow): Same.
7548         * gimple-ssa-warn-alloca.c (pass_walloca::gate): Adjust.
7549         (alloca_call_type_by_arg): Change function argument to HOST_WIDE_INT.
7550         Diagnose unbounded alloca calls only for limits of less than
7551         PTRDIFF_MAX.
7552         (alloca_call_type): Adjust.  Diagnose possibly out-of-bounds alloca
7553         calls and VLA size only for limits of less than PTRDIFF_MAX.  Same
7554         for alloca(0).
7555         (pass_walloca::execute): Adjust.  Diagnose alloca calls in loops
7556         only for limits of less than PTRDIFF_MAX.
7557         * langhooks-def.h (lhd_handle_option): Change function argument
7558         to HOST_WIDE_INT.
7559         * langhooks.c (lhd_handle_option): Same.
7560         * langhooks.h (handle_option): Same.
7561         * opt-functions.awk (switch_bit_fields): Handle Host_Wide_Int and
7562         ByteSize flags.
7563         (var_type, var_type_struct): Same.
7564         (var_set): Handle ByteSize flag.
7565         * optc-gen.awk: Add comments to output to ease debugging.  Make
7566         use of HOST_WIDE_INT where appropriate.
7567         * opts-gen-save.awk:  Use %lx to format unsigned long.
7568         * opth-gen.awk: Change function argument to HOST_WIDE_INT.
7569         * opts-common.c (integral_argument): Return HOST_WIDE_INT and add
7570         arguments.  Parse bytes-size suffixes.
7571         (enum_arg_to_value): Change function argument to HOST_WIDE_INT.
7572         (enum_value_to_arg): Same.
7573         (decode_cmdline_option): Handle cl_host_wide_int.  Adjust.
7574         (handle_option): Adjust.
7575         (generate_option): Change function argument to HOST_WIDE_INT.
7576         (cmdline_handle_error): Adjust.
7577         (read_cmdline_option): Change function argument to HOST_WIDE_INT.
7578         (set_option): Change function argument to HOST_WIDE_INT.
7579         (option_enabled): Handle cl_host_wide_int.
7580         (get_option_state): Handle CLVC_SIZE.
7581         (control_warning_option): Same.
7582         * opts.c (common_handle_option): Change function argument to
7583         HOST_WIDE_INT.  Remove handling of OPT_Walloca_larger_than_ and
7584         OPT_Wvla_larger_than_.
7585         * opts.h (enum cl_var_type): Add an enumerator.
7586         * stor-layout.c (layout_decl): Print a more meaningful warning.
7587         * toplev.c (output_stack_usage): Adjust.
7589 2018-07-20  Qing Zhao  <qing.zhao@oracle.com>
7591         * builtins.c (expand_builtin_memcmp): Delete the last parameter for
7592         call to inline_expand_builtin_string_cmp.
7593         (expand_builtin_strcmp): Likewise.
7594         (expand_builtin_strncmp): Likewise.
7595         (inline_string_cmp): Delete the last parameter, change char_type_node
7596         to unsigned_char_type_node for strcmp/strncmp, add conversions to the
7597         two operands.
7598         (inline_expand_builtin_string_cmp): Delete the last parameter, give up
7599         the inlining expansion on target where the type of the call has same or
7600         narrower precision than unsigned char.
7602 2018-07-20  David Malcolm  <dmalcolm@redhat.com>
7604         * Makefile.in (OBJS): Add json.o and optinfo-emit-json.o.
7605         (CFLAGS-optinfo-emit-json.o): Define TARGET_NAME.
7606         * common.opt (fsave-optimization-record): New option.
7607         * coretypes.h (struct kv_pair): Move here from dumpfile.c.
7608         * doc/invoke.texi (-fsave-optimization-record): New option.
7609         * dumpfile.c: Include "optinfo-emit-json.h".
7610         (struct kv_pair): Move to coretypes.h.
7611         (optgroup_options): Make non-static.
7612         (dump_context::end_scope): Call
7613         optimization_records_maybe_pop_dump_scope.
7614         * dumpfile.h (optgroup_options): New decl.
7615         * json.cc: New file.
7616         * json.h: New file.
7617         * optinfo-emit-json.cc: New file.
7618         * optinfo-emit-json.h: New file.
7619         * optinfo.cc: Include "optinfo-emit-json.h".
7620         (optinfo::emit): Call optimization_records_maybe_record_optinfo.
7621         (optinfo_enabled_p): Check optimization_records_enabled_p.
7622         (optinfo_wants_inlining_info_p): Likewise.
7623         * optinfo.h: Update comment.
7624         * profile-count.c (profile_quality_as_string): New function.
7625         * profile-count.h (profile_quality_as_string): New decl.
7626         (profile_count::quality): New accessor.
7627         * selftest-run-tests.c (selftest::run_tests): Call json_cc_tests
7628         and optinfo_emit_json_cc_tests.
7629         * selftest.h (selftest::json_cc_tests): New decl.
7630         (selftest::optinfo_emit_json_cc_tests): New decl.
7631         * toplev.c: Include "optinfo-emit-json.h".
7632         (compile_file): Call optimization_records_finish.
7633         (do_compile): Call optimization_records_start.
7634         * tree-ssa-live.c: Include optinfo.h.
7635         (remove_unused_scope_block_p): Retain inlining information if
7636         optinfo_wants_inlining_info_p returns true.
7638 2018-07-20  Richard Biener  <rguenther@suse.de>
7640         PR debug/86585
7641         * dwarf2out.c (dwarf2out_die_ref_for_decl): Test in_lto_p
7642         to cover -flto-partition=none.
7644 2018-07-20  Martin Liska  <mliska@suse.cz>
7646         * tree.h (DECL_LOCATION_RANGE): Remove unused macro.
7647         (get_decl_source_range): Remove unused function.
7649 2018-07-20  Richard Biener  <rguenther@suse.de>
7651         * tree-ssa-sccvn.h (struct vn_nary_op_s): Add next member.
7652         (struct vn_phi_s): Likewise.
7653         (struct vn_reference_s): Likewise.
7654         * tree-ssa-sccvn.c (vn_nary_op_hasher::equal): Add shortcut
7655         for searching the slot of an entry known to be in the hash itself.
7656         (vn_phi_hasher::equal): Likewise.
7657         (vn_reference_hasher::equal): Likewise.
7658         (last_inserted_ref, last_inserted_phi, last_inserted_nary): New
7659         globals.
7660         (optimistic_info, current_info): Remove, keeping only valid_info.
7661         (vn_reference_lookup_1): Remove fallback lookup.
7662         (vn_reference_lookup_2): Likewise.
7663         (vn_nary_op_lookup_1): Likewise.
7664         (vn_phi_lookup): Likewise.
7665         (vn_nary_build_or_lookup_1): Make sure to not chain the built
7666         hash element.
7667         (vn_reference_insert): Adjust, chain the inserted hash element
7668         at last_inserted_ref.
7669         (vn_reference_insert_pieces): Likewise.
7670         (visit_reference_op_call): Likewise.
7671         (vn_nary_op_insert_into): Chain the inserted hash element at
7672         last_inserted_nary.
7673         (vn_nary_op_insert_pieces): Adjust.
7674         (vn_nary_op_insert): Likewise.
7675         (vn_nary_op_insert_stmt): Likewise.
7676         (vn_phi_insert): Adjust, chain the inserted hash element at
7677         last_inserted_phi.
7678         (process_scc): Remove clearing and copying the optimistic
7679         table.  Instead remove elements inserted during an optimistic
7680         iteration from the single table we maintain.
7681         (init_scc_vn): Adjust.
7682         (free_scc_vn): Likewise.
7683         (sccvn_dom_walker::record_cond): Likewise.
7684         (sccvn_dom_walker::after_dom_children): Likewise.
7686 2018-07-19  Martin Sebor  <msebor@redhat.com>
7688         PR tree-optimization/84047
7689         PR tree-optimization/83776
7690         * tree-vrp.c (vrp_prop::check_mem_ref): New function.
7691         (check_array_bounds): Call it.
7693 2018-07-19  Martin Sebor  <msebor@redhat.com>
7695         * align.h (align_flags): Use member initialization.
7697 2018-07-19  David Malcolm  <dmalcolm@redhat.com>
7699         * Makefile.in (OBJS): Add optinfo.o.
7700         * coretypes.h (class symtab_node): New forward decl.
7701         (struct cgraph_node): New forward decl.
7702         (class varpool_node): New forward decl.
7703         * dump-context.h: New file.
7704         * dumpfile.c: Include "optinfo.h", "dump-context.h", "cgraph.h",
7705         "tree-pass.h".
7706         (refresh_dumps_are_enabled): Use optinfo_enabled_p.
7707         (set_dump_file): Call dumpfile_ensure_any_optinfo_are_flushed.
7708         (set_alt_dump_file): Likewise.
7709         (dump_context::~dump_context): New dtor.
7710         (dump_gimple_stmt): Move implementation to...
7711         (dump_context::dump_gimple_stmt): ...this new member function.
7712         Add the stmt to any pending optinfo, creating one if need be.
7713         (dump_gimple_stmt_loc): Move implementation to...
7714         (dump_context::dump_gimple_stmt_loc): ...this new member function.
7715         Start a new optinfo and add the stmt to it.
7716         (dump_gimple_expr): Move implementation to...
7717         (dump_context::dump_gimple_expr): ...this new member function.
7718         Add the stmt to any pending optinfo, creating one if need be.
7719         (dump_gimple_expr_loc): Move implementation to...
7720         (dump_context::dump_gimple_expr_loc): ...this new member function.
7721         Start a new optinfo and add the stmt to it.
7722         (dump_generic_expr): Move implementation to...
7723         (dump_context::dump_generic_expr): ...this new member function.
7724         Add the tree to any pending optinfo, creating one if need be.
7725         (dump_generic_expr_loc): Move implementation to...
7726         (dump_context::dump_generic_expr_loc): ...this new member
7727         function.  Add the tree to any pending optinfo, creating one if
7728         need be.
7729         (dump_printf): Move implementation to...
7730         (dump_context::dump_printf_va): ...this new member function.  Add
7731         the text to any pending optinfo, creating one if need be.
7732         (dump_printf_loc): Move implementation to...
7733         (dump_context::dump_printf_loc_va): ...this new member function.
7734         Start a new optinfo and add the stmt to it.
7735         (dump_dec): Move implementation to...
7736         (dump_context::dump_dec): ...this new member function.  Add the
7737         value to any pending optinfo, creating one if need be.
7738         (dump_context::dump_symtab_node): New member function.
7739         (dump_context::get_scope_depth): New member function.
7740         (dump_context::begin_scope): New member function.
7741         (dump_context::end_scope): New member function.
7742         (dump_context::ensure_pending_optinfo): New member function.
7743         (dump_context::begin_next_optinfo): New member function.
7744         (dump_context::end_any_optinfo): New member function.
7745         (dump_context::s_current): New global.
7746         (dump_context::s_default): New global.
7747         (dump_scope_depth): Delete global.
7748         (dumpfile_ensure_any_optinfo_are_flushed): New function.
7749         (dump_symtab_node): New function.
7750         (get_dump_scope_depth): Reimplement in terms of dump_context.
7751         (dump_begin_scope): Likewise.
7752         (dump_end_scope): Likewise.
7753         (selftest::temp_dump_context::temp_dump_context): New ctor.
7754         (selftest::temp_dump_context::~temp_dump_context): New dtor.
7755         (selftest::verify_item): New function.
7756         (ASSERT_IS_TEXT): New macro.
7757         (ASSERT_IS_TREE): New macro.
7758         (ASSERT_IS_GIMPLE): New macro.
7759         (selftest::test_capture_of_dump_calls): New test.
7760         (selftest::dumpfile_c_tests): Call it.
7761         * dumpfile.h (dump_printf, dump_printf_loc, dump_basic_block)
7762         (dump_generic_expr_loc, dump_generic_expr, dump_gimple_stmt_loc)
7763         (dump_gimple_stmt, dump_dec): Gather these related decls and add a
7764         descriptive comment.
7765         (dump_function, print_combine_total_stats, enable_rtl_dump_file)
7766         (dump_node, dump_bb): Move these unrelated decls.
7767         (class dump_manager): Add leading comment.
7768         * optinfo.cc: New file.
7769         * optinfo.h: New file.
7771 2018-07-19  Michael Collison  <michael.collison@arm.com>
7772             Richard Henderson <rth@redhat.com>
7774         * config/aarch64/aarch64.md (subv<GPI>4, usubv<GPI>4): New patterns.
7775         (subti): Handle op1 zero.
7776         (subvti4, usub4ti4): New.
7777         (*sub<GPI>3_compare1_imm): New.
7778         (sub<GPI>3_carryinCV): New.
7779         (*sub<GPI>3_carryinCV_z1_z2, *sub<GPI>3_carryinCV_z1): New.
7780         (*sub<GPI>3_carryinCV_z2, *sub<GPI>3_carryinCV): New.
7782 2018-07-19  Michael Collison  <michael.collison@arm.com>
7783             Richard Henderson <rth@redhat.com>
7785         * config/aarch64/aarch64.md: (addv<GPI>4, uaddv<GPI>4): New.
7786         (addti3): Create simpler code if low part is already known to be 0.
7787         (addvti4, uaddvti4): New.
7788         (*add<GPI>3_compareC_cconly_imm): New.
7789         (*add<GPI>3_compareC_cconly): New.
7790         (*add<GPI>3_compareC_imm): New.
7791         (*add<GPI>3_compareC): Rename from add<GPI>3_compare1; do not
7792         handle constants within this pattern..
7793         (*add<GPI>3_compareV_cconly_imm): New.
7794         (*add<GPI>3_compareV_cconly): New.
7795         (*add<GPI>3_compareV_imm): New.
7796         (add<GPI>3_compareV): New.
7797         (add<GPI>3_carryinC, add<GPI>3_carryinV): New.
7798         (*add<GPI>3_carryinC_zero, *add<GPI>3_carryinV_zero): New.
7799         (*add<GPI>3_carryinC, *add<GPI>3_carryinV): New.
7800         ((*add<GPI>3_compareC_cconly_imm): Replace 'ne' operator
7801         with 'comparison' operator.
7802         (*add<GPI>3_compareV_cconly_imm): Ditto.
7803         (*add<GPI>3_compareV_cconly): Ditto.
7804         (*add<GPI>3_compareV_imm): Ditto.
7805         (add<GPI>3_compareV): Ditto.
7806         (add<mode>3_carryinC): Ditto.
7807         (*add<mode>3_carryinC_zero): Ditto.
7808         (*add<mode>3_carryinC): Ditto.
7809         (add<mode>3_carryinV): Ditto.
7810         (*add<mode>3_carryinV_zero): Ditto.
7811         (*add<mode>3_carryinV): Ditto.
7813 2018-07-19  Michael Collison  <michael.collison@arm.com>
7814             Richard Henderson <rth@redhat.com>
7816         * config/aarch64/aarch64-modes.def (CC_V): New.
7817         * config/aarch64/aarch64-protos.h
7818         (aarch64_addti_scratch_regs): Declare
7819         (aarch64_subvti_scratch_regs): Declare.
7820         (aarch64_expand_subvti): Declare.
7821         (aarch64_gen_unlikely_cbranch): Declare
7822         * config/aarch64/aarch64.c (aarch64_select_cc_mode): Test
7823         for signed overflow using CC_Vmode.
7824         (aarch64_get_condition_code_1): Handle CC_Vmode.
7825         (aarch64_gen_unlikely_cbranch): New function.
7826         (aarch64_addti_scratch_regs): New function.
7827         (aarch64_subvti_scratch_regs): New function.
7828         (aarch64_expand_subvti): New function.
7830 2018-07-19  Andre Vieira  <andre.simoesdiasvieira@arm.com>
7832         * config/aarch64/aarch64-option-extensions.def: New entry for profile
7833         extension.
7834         * config/aarch64/aarch64.h (AARCH64_FL_PROFILE): New.
7835         * doc/invoke.texi (aarch64-feature-modifiers): New entry for profile
7836         extension.
7838 2018-07-19  Andre Vieira  <andre.simoesdiasvieira@arm.com>
7840         PR target/83009
7841         * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
7842         address check not strict.
7844 2018-07-19  Andre Vieira  <andre.simoesdiasvieira@arm.com>
7846         * config/aarch64/aarch64-simd.md (aarch64_simd_mov<VQ:mode>): Replace
7847         Umq with Umn.
7848         (store_pair_lanes<mode>): Likewise.
7849         * config/aarch64/aarch64-protos.h (aarch64_addr_query_type): Add new
7850         enum value 'ADDR_QUERY_LDP_STP_N'.
7851         * config/aarch64/aarch64.c (aarch64_addr_query_type): Likewise.
7852         (aarch64_print_address_internal): Add declaration.
7853         (aarch64_print_ldpstp_address): Remove.
7854         (aarch64_classify_address): Adapt mode for 'ADDR_QUERY_LDP_STP_N'.
7855         (aarch64_print_operand): Change printing of 'y'.
7856         * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Use
7857         new enum value 'ADDR_QUERY_LDP_STP_N', don't hardcode mode and use
7858         'true' rather than '1'.
7859         * config/aarch64/constraints.md (Uml): Likewise.
7860         (Uml): Rename to Umn.
7861         (Umq): Remove.
7863 2018-07-19  Richard Biener  <rguenther@suse.de>
7865         * tree-ssa-sccvn.h (struct vn_phi_s): Make phiargs member
7866         a trailing array.
7867         * tree-ssa-sccvn.c: Remove alloc-pool.h use.
7868         (vn_phi_hasher): Derive from nofree_ptr_hash and remove remove method.
7869         (vn_reference_hasher): Likewise.
7870         (struct vn_tables_s): Remove obstack and alloc-pool members.
7871         (vn_tables_obstack, vn_tables_insert_obstack): New global obstacks.
7872         (vn_nary_build_or_lookup_1): Manually build in vn_tables_insert_obstack.
7873         (vn_reference_insert): Allocate from obstack instead of from alloc-pool.
7874         (vn_reference_insert_pieces): Likewise.
7875         (alloc_vn_nary_op_noinit): Adjust.
7876         (vn_nary_op_insert_stmt): Allocate phiargs in-place.
7877         (vn_phi_eq): Adjust.
7878         (shared_lookup_phiargs): Remove.
7879         (vn_phi_lookup): Allocate temporary vn_phi_s on the stack.
7880         (vn_phi_insert): Allocate from obstack instead of from alloc-pool.
7881         (visit_reference_op_call): Likewise.
7882         (copy_nary, copy_phi, copy_reference): Remove.
7883         (process_scc): Rewind the obstack when iterating.  Do not
7884         copy the elements to valid_info but just move them from one
7885         hashtable to the other.
7886         (allocate_vn_table): Adjust.
7887         (free_vn_table): Likewise.
7888         (init_scc_vn): Likewise.
7889         (free_scc_vn): Likewise.
7891 2018-07-19  H.J. Lu  <hongjiu.lu@intel.com>
7893         PR target/86560
7894         * config/i386/i386.c (rest_of_insert_endbranch): Lookup
7895         indirect_return as function type attribute.
7896         (ix86_attribute_table): Change indirect_return to function
7897         type attribute.
7898         * doc/extend.texi: Update indirect_return attribute.
7900 2018-07-19  Aldy Hernandez  <aldyh@redhat.com>
7902         * wide-int.h (widest2_int): New.
7903         * gimple-fold.c (arith_overflowed_p): Use it.
7904         * tree.h (widest2_int_cst): New.
7905         * tree-vrp.c (wide_int_binop_overflow): Rename from
7906         vrp_int_const_binop.
7907         Rewrite to work on trees.
7908         (extract_range_from_multiplicative_op_1): Abstract code to...
7909         (wide_int_range_min_max): ...here.
7910         (wide_int_range_cross_product): ...and here.
7911         (extract_range_from_binary_expr_1): Abstract overflow code to...
7912         (wide_int_range_mult_wrapping): ...here.
7913         * tree-vrp.h (wide_int_range_cross_product): New.
7914         (wide_int_range_mult_wrapping): New.
7916 2018-07-19  Andrew Senkevich  <andrew.senkevich@intel.com>
7917             Julia Koval  <julia.koval@intel.com>
7919         * config/i386/x86-tune-costs.h (skylake_memcpy,
7920         skylake_memset): Replace rep_prefix with unrolling for size 512.
7922 2018-07-18  Kugan Vivekanandarajah  <kuganv@linaro.org>
7924         PR middle-end/86544
7925         * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): Handle
7926         comparision with EQ_EXPR in last stmt.
7928 2018-07-18  Kelvin Nilsen  <kelvin@gcc.gnu.org>
7930         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Rename
7931         this subsection to "PowerPC AltiVec/VSX Built-in Functions".
7932         (PowerPC AltiVec/VSX Built-in Functions): New name for subsection
7933         previously known as "PowerPC AltiVec Built-in Functions".  Move
7934         some material to new subsubsections "PowerPC AltiVec Built-in
7935         Functions on ISA 2.06" and "PowerPC AltiVec Built-in Functions on
7936         ISA 2.07".
7937         (PowerPC Altivec Built-in Functions on ISA 2.05): New subsubsection.
7938         (PowerPC Altivec Built-in Functions on ISA 2.06): Likewise.
7939         (PowerPC Altivec Built-in Functions on ISA 2.07): Likewise.
7940         (PowerPC Altivec Built-in Functions on ISA 3.0): Likewise.
7942 2018-07-18  Richard Biener  <rguenther@suse.de>
7944         PR tree-optimization/86557
7945         * tree-vect-patterns.c (vect_recog_divmod_pattern): Also handle
7946         EXACT_DIV_EXPR.
7948 2018-07-18  Ilya Leoshkevich  <iii@linux.ibm.com>
7950         * config/s390/s390.c (s390_function_profiler): Generate CFI.
7952 2018-07-17  Jeff Law  <law@redhat.com>
7954         * config/arm/arm.c (get_label_padding): Update for recent
7955         changes to label_to_alignment.
7957         PR tree-optimization/86010
7958         * tree-ssa-dse.c (compute_trims): Fix typo/thinko.
7960         * config/mips/mips.c (vr4130_align_insns): Update for recent
7961         changes to label_to_alignment.
7963         * config/frv/frv.c (frv_label_align): Update for recent changes
7964         to label_to_alignment.
7966         * config/nios2/nios2.c (nios2_label_align): Update for recent
7967         changes which dropped ALIGN_LABELS_LOG.
7969 2018-07-17  Andreas Schwab  <schwab@linux-m68k.org>
7971         * config/m68k/m68k.md (umulsi3_highpart+1, const_umulsi3_highpart)
7972         (smulsi3_highpart+1, const_smulsi3_highpart): Add CC_STATUS_INIT.
7974 2018-07-17  Claudiu Zissulescu  <claziss@synopsys.com>
7976         * config/arc/arc.c (arc_label_align): Use align_labels instead of
7977         deprecated align_labels_log.
7979 2018-07-17  Richard Biener  <rguenther@suse.de>
7981         PR lto/86456
7982         * dwarf2out.c (init_sections_and_labels): Always generate
7983         a debug_line_str_section for early LTO debug.
7984         (dwarf2out_finish): Reset debug_line_str_hash output early.
7985         Bump counter for extra dwarf5 .debug_loc labels to not conflict
7986         with fat LTO part.
7987         (dwarf2out_early_finish): Output debug_line_str.
7989 2018-07-17  Robin Dapp  <rdapp@linux.ibm.com>
7991         * config/s390/s390.c (preferred_la_operand_p): Do not use LA with
7992         index register on z196 or later.
7994 2018-07-17  Robin Dapp  <rdapp@linux.ibm.com>
7996         * config/s390/s390.c (s390_default_align): Set default function
7997         alignment to 16.
7998         (s390_override_options_after_change): Call s390_default align.
7999         (s390_option_override_internal): Call s390_default_align.
8000         (TARGET_OVERRIDE_OPTIONS_AFTER_CHANGE): Define.
8002 2018-07-17  Jakub Jelinek  <jakub@redhat.com>
8004         PR middle-end/86542
8005         * omp-low.c (create_task_copyfn): Copy over also fields corresponding
8006         to _looptemp_ clauses, other than the first two.
8008 2018-07-17  Martin Liska  <mliska@suse.cz>
8010         * opts.c: Do not enable OPT_falign_* for -Os.
8012 2018-07-17  Martin Liska  <mliska@suse.cz>
8014         * align.h (MAX_CODE_ALIGN): New.
8015         (MAX_CODE_ALIGN_VALUE): New.
8016         * common/config/i386/i386-common.c (ix86_handle_option):
8017         (MAX_CODE_ALIGN): Moved to align.h.
8018         * final.c (MAX_CODE_ALIGN): Likewise.
8019         * opts.c (parse_and_check_align_values):
8020         (MAX_CODE_ALIGN): Likewise.
8021         (MAX_CODE_ALIGN_VALUE): Likewise.
8023 2018-07-17  Martin Liska  <mliska@suse.cz>
8025         * config/i386/att.h (ASM_OUTPUT_ALIGN): Fix spacing
8026         in order to fulfil coding style.
8027         * config/i386/cygming.h (ASM_OUTPUT_ALIGN): Likewise.
8028         * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
8029         * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
8030         * config/iq2000/iq2000.h (ASM_OUTPUT_ALIGN): Likewise.
8031         * config/pa/pa.h (ASM_OUTPUT_ALIGN): Likewise.
8032         * config/sparc/sol2.h (ASM_OUTPUT_ALIGN_WITH_NOP): Likewise.
8033         * config/sparc/sparc.h (ASM_OUTPUT_ALIGN): Likewise.
8034         * config/visium/visium.h (ASM_OUTPUT_ALIGN): Likewise.
8035         (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
8037 2018-07-17  Martin Liska  <mliska@suse.cz>
8039         * align.h: New file.
8040         * config/alpha/alpha.c (alpha_align_insns_1): Use align_functions
8041         directly.
8042         * config/i386/i386.c (ix86_avoid_jump_mispredicts): Use new return type
8043         align_flags of label_to_alignment.
8044         * config/m32r/m32r.h (LOOP_ALIGN): Wrap returned values into
8045         align_flags class.
8046         * config/m68k/m68k.c: Do not use removed align_labels_value and
8047         align_loops_value.
8048         * config/nds32/nds32.h (JUMP_ALIGN): Wrap result into align_flags class.
8049         (LOOP_ALIGN): Likewise.
8050         (LABEL_ALIGN): Likewise.
8051         * config/powerpcspe/powerpcspe.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
8052         Remove not used macro.
8053         (rs6000_loop_align): Change return type to align_flags.
8054         (rs6000_loop_align_max_skip): Remove.
8055         * config/rs6000/rs6000-protos.h (rs6000_loop_align):
8056         Change return type to align_flags.
8057         * config/rs6000/rs6000.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
8058         Remove not used macro.
8059         (rs6000_loop_align):  Change return type to align_flags.
8060         (rs6000_loop_align_max_skip): Remove.
8061         * config/rx/rx.h (JUMP_ALIGN): Wrap integer values
8062         * config/rx/rx-protos.h (rx_align_for_label): Make it
8063         static function.
8064         * config/rx/rx.c (rx_align_for_label): Change return type
8065         to align_flags.
8066         (rx_max_skip_for_label): Remove TARGET_ASM_*_ALIGN_MAX_SKIP
8067         macro definitions.
8068         into align_flags class.
8069         (LABEL_ALIGN): Likewise.
8070         (LOOP_ALIGN): Likewise.
8071         * config/s390/s390.c (s390_label_align): Use align_flags
8072         class member.
8073         (s390_asm_output_function_label): Likewise.
8074         * config/sh/sh.c (sh_override_options_after_change):
8075         Use align_flags class directly without macros.
8076         (find_barrier): Likewise.
8077         (barrier_align): Likewise.
8078         (sh_loop_align): Likewise.
8079         * config/spu/spu.c (spu_option_override):
8080         Use align_flags_tuple::get_value instead of removed macros.
8081         (spu_sched_init): Likewise.
8082         * config/spu/spu.h (GTY): Likewise.
8083         * config/visium/visium.c (visium_option_override):
8084         Set "8" as default secondary alignment.
8085         * config/visium/visium.h (SUBALIGN_LOG): Define to 3
8086         in order to guarantee secondary alignment of 8.
8087         * coretypes.h: Include align.h header file.
8088         * doc/tm.texi: Remove TARGET_ASM_JUMP_ALIGN_MAX_SKIP,
8089         TARGET_ASM_LOOP_ALIGN_MAX_SKIP, TARGET_ASM_LABEL_ALIGN_MAX_SKIP
8090         and TARGET_ASM_LABEL_ALIGN_AFTER_BARRIER_MAX_SKIP macros.
8091         * doc/tm.texi.in: Likewise.
8092         * final.c (struct label_alignment): Remove not used structure.
8093         (LABEL_ALIGN): Change type to align_flags.
8094         (LOOP_ALIGN): Likewise.
8095         (JUMP_ALIGN): Likewise.
8096         (default_loop_align_max_skip): Remove.
8097         (default_label_align_max_skip): Likewise.
8098         (default_jump_align_max_skip): Likewise.
8099         (default_label_align_after_barrier_max_skip):
8100         (LABEL_TO_ALIGNMENT): Change to access label_align vector.
8101         (LABEL_TO_MAX_SKIP): Remove.
8102         (label_to_alignment): Return align_flags type instead of integer.
8103         (label_to_max_skip): Remove.
8104         (align_fuzz): Use align_flags type.
8105         (compute_alignments): Use align_flags type and use align_flags::max
8106         to combine multiple alignments.
8107         (grow_label_align): Grow vec instead of C array.
8108         (update_alignments): Assign just LABEL_TO_ALIGNMENT.
8109         (shorten_branches):  Use align_flags type and use align_flags::max
8110         to combine multiple alignments.
8111         (final_scan_insn_1): Remove usage of secondary alignment that comes
8112         from label alignment, but instead use proper secondary alignment
8113         which is computed in grow_label_align.
8114         * flags.h (struct align_flags_tuple): Move to align.h.
8115         (struct align_flags): Likewise.
8116         (state_align_loops): Rename to align_loops.
8117         (state_align_jumps): Rename to align_jumps.
8118         (state_align_labels): Rename to align_labels.
8119         (state_align_functions): Rename to align_functions.
8120         (align_loops_log): Remove.
8121         (align_jumps_log): Remove.
8122         (align_labels_log): Remove.
8123         (align_functions_log): Remove.
8124         (align_loops_max_skip): Remove.
8125         (align_jumps_max_skip): Remove.
8126         (align_labels_max_skip): Remove.
8127         (align_functions_max_skip): Remove.
8128         (align_loops_value): Remove.
8129         (align_jumps_value): Remove.
8130         (align_labels_value): Remove.
8131         (align_functions_value): Remove.
8132         * output.h (label_to_alignment): Change return type to align_flags.
8133         (label_to_max_skip): Remove.
8134         * target.def: Remove loop_align_max_skip, label_align_max_skip,
8135         jump_align_max_skip macros.
8136         * targhooks.h (default_loop_align_max_skip): Remove.
8137         (default_label_align_max_skip): Likewise.
8138         (default_jump_align_max_skip): Likewise.
8139         (default_label_align_after_barrier_max_skip): Remove.
8140         * toplev.c (read_log_maxskip): Use ::normalize function.
8141         (parse_N_M): Remove not used argument and also call ::normalize.
8142         (parse_alignment_opts): Do not pass unused arguments.
8143         * varasm.c (assemble_start_function): Use directly align_functions
8144         instead of removed macros.
8145         * system.h: Do not poison removed macros.
8147 2018-07-17  Jakub Jelinek  <jakub@redhat.com>
8149         PR middle-end/86539
8150         * gimplify.c (gimplify_omp_for): Ensure taskloop firstprivatized init
8151         and cond temporaries don't have reference type if iterator has
8152         pointer type.  For init use &for_pre_body instead of pre_p if
8153         for_pre_body is non-empty.
8155 2018-07-16  Segher Boessenkool  <segher@kernel.crashing.org>
8157         * config/rs6000/rs6000.md (trunc<mode>sf2): Expand truncates of
8158         double-double modes to SFmode directly directly.
8159         (trunc<mode>sf2_fprs): Delete.
8161 2018-07-16  Segher Boessenkool  <segher@kernel.crashing.org>
8163         * config/rs6000/rs6000.c (init_float128_ibm): Use the correct names
8164         for conversions between IFmode and the decimal floating point modes.
8165         (init_float128_ieee): Use the correct names for conversions between
8166         KFmode and the decimal floating point modes.
8168 2018-07-16  Segher Boessenkool  <segher@kernel.crashing.org>
8170         * config/rs6000/rs6000.c (init_float128_ibm): Use more correct names
8171         for the conversions between TDmode and IFmode.
8172         (init_float128_ieee): Use more correct names for the conversions
8173         between TDmode and KFmode.
8175 2018-07-16  Jakub Jelinek  <jakub@redhat.com>
8177         PR tree-optimization/86526
8178         * builtins.c (expand_builtin_memcmp): Formatting fixes.
8179         (inline_expand_builtin_string_cmp): Likewise.
8180         (inline_string_cmp): Likewise.  Use c_readstr instead of
8181         builtin_memcpy_read_str.  Add unit_mode temporary.
8183 2018-07-16  Bernd Edlinger  <bernd.edlinger@hotmail.de>
8185         PR middle-end/86528
8186         * builtins.c (check_access): Bail out if range[0] is no INTEGER_CST.
8187         * expr.c (string_constant): Fix the element size of ARRAY_TYPE.
8189 2018-07-16  Kelvin Nilsen  <kelvin@gcc.gnu.org>
8191         * doc/extend.texi (PowerPC AltiVec Built-in Functions):
8192         Alphabetize prototypes of built-in functions, separating out
8193         built-in functions that are listed in this section but should be
8194         described elsewhere.
8196 2018-07-16  Uros Bizjak  <ubizjak@gmail.com>
8198         PR target/86511
8199         * expmed.c (emit_store_flag): Do not emit setcc followed by a
8200         conditional move when trapping comparison was split to a
8201         non-trapping one (and vice versa).
8203 2018-07-16  Ilya Leoshkevich  <iii@linux.ibm.com>
8205         * config/s390/s390.c (s390_function_profiler): Generate nops
8206         instead of profiler call sequences.
8207         * config/s390/s390.opt: Add the new option.
8209 2018-07-16  Ilya Leoshkevich  <iii@linux.ibm.com>
8211         * config/s390/s390.c (s390_function_profiler): Generate
8212         __mcount_loc section.
8213         * config/s390/s390.opt: Add the new option.
8215 2018-07-16  Ilya Leoshkevich  <iii@linux.ibm.com>
8217         * common.opt: Add the new warning.
8218         * config/s390/s390.c (s390_function_profiler): Emit "brasl
8219         %r0,__fentry__" when -mfentry is specified.
8220         (s390_option_override_internal): Disallow -mfentry for 31-bit
8221         CPUs.
8222         * config/s390/s390.opt: Add the new option.
8224 2018-07-16  Richard Biener  <rguenther@suse.de>
8226         PR lto/86523
8227         * dwarf2out.c (dwarf2out_register_external_die): Assign DIE parents
8228         for function-local FUNCTION_DECL and RESULT_DECL immediately.
8230 2018-07-16  Martin Liska  <mliska@suse.cz>
8232         PR ipa/86529
8233         * ipa-pure-const.c (malloc_candidate_p): Revert ::get
8234         to ::get_create.
8236 2017-07-16  Claudiu Zissulescu  <claziss@synopsys.com>
8238         * config/arc/arcHS.md: Update ARCHS scheduling rules.
8240 2017-07-16  Claudiu Zissulescu  <claziss@synopsys.com>
8242         * config/arc/arc-arch.h (arc_tune_attr): Add new tune parameters
8243         for ARCHS4x.
8244         * config/arc/arc-cpus.def (hs4x): New cpu.
8245         (hs4xd): Likewise.
8246         * config/arc/arc-tables.opt: Regenerate.
8247         * config/arc/arc.c (arc_sched_issue_rate): New function.
8248         (TARGET_SCHED_ISSUE_RATE): Define.
8249         (TARGET_SCHED_EXPOSED_PIPELINE): Likewise.
8250         * config/arc/arc.md (attr type): Add fpu_fuse, fpu_sdiv, fpu_ddiv,
8251         fpu_cvt.
8252         (attr tune): Add ARCHS4x tune values.
8253         (attr tune_dspmpy): Define.
8254         (*tst): Correct instruction type.
8255         * config/arc/arcHS.md: Don't use this automaton for ARCHS4x cpus.
8256         * config/arc/arcHS4x.md: New file.
8257         * config/arc/fpu.md: Update instruction type attributes.
8258         * config/arc/t-multilib: Regenerate.
8260 2018-07-16  Tom de Vries  <tdevries@suse.de>
8262         PR debug/86455
8263         * var-tracking.c (vt_initialize): Fix pre_dec handling.
8265 2018-07-16  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
8267         * config/aarch64/atomics.md (aarch64_store_execlusive<mode>): Add
8268         early clobber.
8270 2018-07-16  Eric Botcazou  <ebotcazou@adacore.com>
8272         PR tree-optimization/86514
8273         * tree-ssa-reassoc.c (init_range_entry) <CASE_CONVERT>: Return for a
8274         conversion to a boolean type from a type with greater precision.
8276 2018-07-16  Tom de Vries  <tdevries@suse.de>
8278         * var-tracking.c (vt_initialize): Print adjusted insn slim if
8279         dump_flags request TDF_SLIM.
8281 2018-07-16  Aldy Hernandez  <aldyh@redhat.com>
8283         * fold-const.c (int_const_binop_1): Abstract...
8284         (wide_int_binop): ...wide int code here.
8285         (poly_int_binop): ...poly int code here.
8286         Abstract the rest of int_const_binop_1 into int_const_binop.
8287         * fold-const.h (wide_int_binop): New.
8288         * tree-vrp.c (vrp_int_const_binop): Call wide_int_binop.
8289         Remove useless PLUS/MINUS_EXPR case.
8290         (zero_nonzero_bits_from_vr): Move wide int code...
8291         (zero_nonzero_bits_from_bounds): ...here.
8292         (extract_range_from_binary_expr_1): Move mask optimization code...
8293         (range_easy_mask_min_max): ...here.
8294         * tree-vrp.h (zero_nonzero_bits_from_bounds): New.
8295         (range_easy_mask_min_max): New.
8297 2018-07-15  Jeff Law  <law@redhat.com>
8299         PR target/85993
8300         * config/sh/sh.c (output_mi_thunk): Remove dead conditional
8301         block.
8303 2018-07-14  Jim Wilson  <jimw@sifive.com>
8305         * config/riscv/linux.h (TARGET_ASM_FILE_END): New.
8307 2018-07-14  Paul Koning  <ni1d@arrl.net>
8309         * config/pdp11/pdp11.c (pdp11_rtx_costs): Bugfixes.
8311 2018-07-13  Jan Hubicka  <hubicka@ucw.cz>
8313         * lto-streamer-out.c (copy_function_or_variable): Dump info about
8314         copying section.
8316 2018-07-13  Bill Schmidt  <wschmidt@linux.ibm.com>
8317             Steve Munroe  <munroesj52@gmail.com>
8319         * config/rs6000/emmintrin.h (_mm_and_si128): New function.
8320         (_mm_andnot_si128): Likewise.
8321         (_mm_or_si128): Likewise.
8322         (_mm_xor_si128): Likewise.
8324 2018-07-13  Qing Zhao  <qing.zhao@oracle.com>
8326         PR middle-end/78809
8327         * builtins.c (expand_builtin_memcmp): Inline the calls first
8328         when result_eq is false.
8329         (expand_builtin_strcmp): Inline the calls first.
8330         (expand_builtin_strncmp): Likewise.
8331         (inline_string_cmp): New routine. Expand a string compare
8332         call by using a sequence of char comparison.
8333         (inline_expand_builtin_string_cmp): New routine. Inline expansion
8334         a call to str(n)cmp/memcmp.
8335         * doc/invoke.texi (--param builtin-string-cmp-inline-length):
8336         New option.
8337         * params.def (BUILTIN_STRING_CMP_INLINE_LENGTH): New.
8339 2018-07-13  Richard Earnshaw  <rearnsha@arm.com>
8341         * config/arm/driver-arm.c: Include arm-native.h.
8342         (host_detect_local_cpu): Use auto-generated data tables.
8343         (vendors, arm_cpu_table): Delete.  Move part information to ...
8344         * config/arm/arm-cpus.in: ... here.
8345         * config/arm/parsecpu.awk (gen_native): New function.
8346         (vendor, part): New CPU fields.
8347         (END): Add support for building the native CPU detection tables.
8348         * config/arm/t-arm (arm-native.h): Add build rule.
8349         (driver-arm.o): Add dependency on arm-native.h.
8351 2018-07-13  Richard Biener  <rguenther@suse.de>
8353         PR middle-end/85974
8354         * match.pd (addr1 - addr2): Allow either of the operand to
8355         have a conversion.
8357 2018-07-13  Tom de Vries  <tdevries@suse.de>
8359         * tree-inline.c (remap_ssa_name): Save and reuse debug exprs generated
8360         in remap_ssa_name.
8362 2018-07-13  Jackson Woodruff  <jackson.woodruff@arm.com>
8364         * config/aarch64/aarch64.c (aarch64_operands_adjust_ok_for_ldpstp): Use
8365         arrays instead of numbered variables.
8367 2018-07-13  Eric Botcazou  <ebotcazou@adacore.com>
8369         * config/sparc/sparc-protos.h (sparc_compute_frame_size): Delete.
8370         * config/sparc/sparc.c (sparc_compute_frame_size): Make static.
8372 2018-07-13  Richard Biener  <rguenther@suse.de>
8374         PR debug/86452
8375         * dwarf2out.c (gen_type_die_with_usage): Use scope_die_for
8376         instead of get_context_die.
8378 2018-07-13  Kugan Vivekanandarajah  <kuganv@linaro.org>
8379             Richard Biener  <rguenther@suse.de>
8381         PR middle-end/86489
8382         * tree-ssa-loop-niter.c (number_of_iterations_popcount): Check
8383         that the loop latch destination where phi is defined.
8385 2018-07-12  Kito Cheng  <kito.cheng@gmail.com>
8387         * config/riscv/riscv.c (enum riscv_privilege_levels): Add UNKNOWN_MODE.
8388         (riscv_expand_epilogue): Add assertion to check interrupt mode.
8389         (riscv_set_current_function): Extract getting interrupt type to new
8390         function.
8391         (riscv_get_interrupt_type): New function.
8392         (riscv_merge_decl_attributes): New function, checking interrupt type is
8393         same.
8394         (TARGET_MERGE_DECL_ATTRIBUTES): Define.
8396 2018-07-12  Paul Koning  <ni1d@arrl.net>
8398         * config/pdp11/pdp11.c (pdp11_output_def): Fix typo in .set
8399         directive.
8401 2018-07-12  Paul Koning  <ni1d@arrl.net>
8403         * doc/rtl.texi (REG_NONNEG): Remove decrement and branch until
8404         zero reference, add doloop_end instead.
8405         * doc/md.texi (decrement_and_branch_until_zero): Remove.
8406         (Looping patterns): Remove decrement_and_branch_until_zero.  Add
8407         detail for doloop_end.
8409 2018-07-12  Martin Sebor  <msebor@redhat.com>
8411         PR c/86453
8412         * attribs.c (decl_attributes): Reject conflicting attributes before
8413         calling attribute handlers.
8415 2018-07-12  Jan Hubicka  <hubicka@ucw.cz>
8417         * dumpfile.c (gcc::dump_manager::get_dump_file_name): Add PART
8418          parameter.
8419         (gcc::dump_manager::get_dump_file_name): likewise.
8420         (dump_begin): Likewise.
8421         * dumpfile.h (dump_begin): Update prototype.
8422         (gcc::dump_manager::get_dump_file_name,
8423         gcc::dump_manager::get_dump_file_name): Update prototype.
8425 2018-07-12  Richard Sandiford  <richard.sandiford@linaro.org>
8427         * internal-fn.h (vectorizable_internal_fn_p): New function.
8428         * tree-vect-slp.c (compatible_calls_p): Likewise.
8429         (vect_build_slp_tree_1): Remove nops argument.  Handle calls
8430         to internal functions.
8431         (vect_build_slp_tree_2): Update call to vect_build_slp_tree_1.
8433 2018-07-12  Richard Sandiford  <richard.sandiford@linaro.org>
8435         * fold-const.h (inverse_conditions_p): Declare.
8436         * fold-const.c (inverse_conditions_p): New function.
8437         * match.pd: Use inverse_conditions_p.  Add folds of view_converts
8438         that test the inverse condition of a conditional internal function.
8439         * internal-fn.h (vectorized_internal_fn_supported_p): Declare.
8440         * internal-fn.c (internal_fn_mask_index): Handle conditional
8441         internal functions.
8442         (vectorized_internal_fn_supported_p): New function.
8443         * tree-if-conv.c: Include internal-fn.h and fold-const.h.
8444         (any_pred_load_store): Replace with...
8445         (need_to_predicate): ...this new variable.
8446         (redundant_ssa_names): New variable.
8447         (ifcvt_can_use_mask_load_store): Move initial checks to...
8448         (ifcvt_can_predicate): ...this new function.  Handle tree codes
8449         for which a conditional internal function exists.
8450         (if_convertible_gimple_assign_stmt_p): Use ifcvt_can_predicate
8451         instead of ifcvt_can_use_mask_load_store.  Update after variable
8452         name change.
8453         (predicate_load_or_store): New function, split out from
8454         predicate_mem_writes.
8455         (check_redundant_cond_expr): New function.
8456         (value_available_p): Likewise.
8457         (predicate_rhs_code): Likewise.
8458         (predicate_mem_writes): Rename to...
8459         (predicate_statements): ...this.  Use predicate_load_or_store
8460         and predicate_rhs_code.
8461         (combine_blocks, tree_if_conversion): Update after above name changes.
8462         (ifcvt_local_dce): Handle redundant_ssa_names.
8463         * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): Handle
8464         general conditional functions.
8465         * tree-vect-stmts.c (vectorizable_call): Likewise.
8467 2018-07-12  Richard Sandiford  <richard.sandiford@linaro.org>
8468             Alan Hayward  <alan.hayward@arm.com>
8469             David Sherwood  <david.sherwood@arm.com>
8471         * internal-fn.h (can_interpret_as_conditional_op_p): Declare.
8472         * internal-fn.c (can_interpret_as_conditional_op_p): New function.
8473         * tree-ssa-math-opts.c (convert_mult_to_fma_1): Handle conditional
8474         plus and minus and convert them into IFN_COND_FMA-based sequences.
8475         (convert_mult_to_fma): Handle conditional plus and minus.
8477 2018-07-12  Richard Sandiford  <richard.sandiford@linaro.org>
8479         * doc/md.texi (cond_fma, cond_fms, cond_fnma, cond_fnms): Document.
8480         * optabs.def (cond_fma_optab, cond_fms_optab, cond_fnma_optab)
8481         (cond_fnms_optab): New optabs.
8482         * internal-fn.def (COND_FMA, COND_FMS, COND_FNMA, COND_FNMS): New
8483         internal functions.
8484         (FMA): Use DEF_INTERNAL_FLT_FN rather than DEF_INTERNAL_FLT_FLOATN_FN.
8485         * internal-fn.h (get_conditional_internal_fn): Declare.
8486         (get_unconditional_internal_fn): Likewise.
8487         * internal-fn.c (cond_ternary_direct): New macro.
8488         (expand_cond_ternary_optab_fn): Likewise.
8489         (direct_cond_ternary_optab_supported_p): Likewise.
8490         (FOR_EACH_COND_FN_PAIR): Likewise.
8491         (get_conditional_internal_fn): New function.
8492         (get_unconditional_internal_fn): Likewise.
8493         * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 5.
8494         (gimple_match_op::gimple_match_op): Add a new overload for 5
8495         operands.
8496         (gimple_match_op::set_op): Likewise.
8497         (gimple_resimplify5): Declare.
8498         * genmatch.c (decision_tree::gen): Generate simplifications for
8499         5 operands.
8500         * gimple-match-head.c (gimple_simplify): Define an overload for
8501         5 operands.  Handle calls with 5 arguments in the top-level overload.
8502         (convert_conditional_op): Handle conversions from unconditional
8503         internal functions to conditional ones.
8504         (gimple_resimplify5): New function.
8505         (build_call_internal): Pass a fifth operand.
8506         (maybe_push_res_to_seq): Likewise.
8507         (try_conditional_simplification): Try converting conditional
8508         internal functions to unconditional internal functions.
8509         Handle 3-operand unconditional forms.
8510         * match.pd (UNCOND_TERNARY, COND_TERNARY): Operator lists.
8511         Define ternary equivalents of the current rules for binary conditional
8512         internal functions.
8513         * config/aarch64/aarch64.c (aarch64_preferred_else_value): Handle
8514         ternary operations.
8515         * config/aarch64/iterators.md (UNSPEC_COND_FMLA, UNSPEC_COND_FMLS)
8516         (UNSPEC_COND_FNMLA, UNSPEC_COND_FNMLS): New unspecs.
8517         (optab): Handle them.
8518         (SVE_COND_FP_TERNARY): New int iterator.
8519         (sve_fmla_op, sve_fmad_op): New int attributes.
8520         * config/aarch64/aarch64-sve.md (cond_<optab><mode>)
8521         (*cond_<optab><mode>_2, *cond_<optab><mode_4)
8522         (*cond_<optab><mode>_any): New SVE_COND_FP_TERNARY patterns.
8524 2018-07-12  Richard Sandiford  <richard.sandiford@linaro.org>
8526         * target.def (preferred_else_value): New target hook.
8527         * doc/tm.texi.in (TARGET_PREFERRED_ELSE_VALUE): New hook.
8528         * doc/tm.texi: Regenerate.
8529         * targhooks.h (default_preferred_else_value): Declare.
8530         * targhooks.c (default_preferred_else_value): New function.
8531         * internal-fn.h (conditional_internal_fn_code): Declare.
8532         * internal-fn.c (FOR_EACH_CODE_MAPPING): New macro.
8533         (get_conditional_internal_fn): Use it.
8534         (conditional_internal_fn_code): New function.
8535         * gimple-match.h (gimple_match_cond): New struct.
8536         (gimple_match_op): Add a cond member function.
8537         (gimple_match_op::gimple_match_op): Update all forms to take a
8538         gimple_match_cond.
8539         * genmatch.c (expr::gen_transform): Use the same condition as res_op
8540         for the suboperation, but don't specify a particular else_value.
8541         * tree-ssa-sccvn.c (vn_nary_simplify, vn_reference_lookup_3)
8542         (visit_nary_op, visit_reference_op_load): Pass
8543         gimple_match_cond::UNCOND to the gimple_match_op constructor.
8544         * gimple-match-head.c: Include tree-eh.h
8545         (convert_conditional_op): New function.
8546         (maybe_resimplify_conditional_op): Likewise.
8547         (gimple_resimplify1): Call maybe_resimplify_conditional_op.
8548         (gimple_resimplify2): Likewise.
8549         (gimple_resimplify3): Likewise.
8550         (gimple_resimplify4): Likewise.
8551         (maybe_push_res_to_seq): Return null for conditional operations.
8552         (try_conditional_simplification): New function.
8553         (gimple_simplify): Call it.  Pass conditions to the gimple_match_op
8554         constructor.
8555         * match.pd: Fold VEC_COND_EXPRs of an IFN_COND_* call to a new
8556         IFN_COND_* call.
8557         * config/aarch64/aarch64.c (aarch64_preferred_else_value): New
8558         function.
8559         (TARGET_PREFERRED_ELSE_VALUE): Redefine.
8561 2018-07-12  Jan Hubicka  <hubicka@ucw.cz>
8563         * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
8564         DECL_FCONTEXT
8565         (hash_tree): Do not hash DECL_FCONTEXT
8566         * tree-streamer-in.c (lto_input_ts_field_decl_tree_pointers):
8567         Do not stream DECL_FCONTEXT.
8568         * tree-streamer-out.c (write_ts_field_decl_tree_pointers): Likewise.
8569         * tree.c (free_lang_data_in_decl): Free DECL_FCONTEXT.
8571 2018-07-12  Richard Biener  <rguenther@suse.de>
8573         PR debug/86462
8574         * dwarf2out.c (gen_block_die): Only output blocks when they have
8575         at least one !DECL_IGNORED_P variable.
8577 2018-07-12  Richard Biener  <rguenther@suse.de>
8579         PR target/84829
8580         * config/gnu-user.h (GNU_USER_TARGET_NO_PTHREADS_LIB_SPEC):
8581         Remove -mieee-fp handling.
8583 2018-07-12  Richard Biener  <rguenther@suse.de>
8585         * tree-ssa-sccvn.c (vn_lookup_simplify_result): Remove bogus
8586         left-over from last patch.
8588 2018-07-12  Jakub Jelinek  <jakub@redhat.com>
8590         PR tree-optimization/86492
8591         * gimple-ssa-store-merging.c
8592         (imm_store_chain_info::coalesce_immediate_stores): Call
8593         check_no_overlap even for the merge_overlapping case.  Formatting fix.
8595 2018-07-12  Richard Biener  <rguenther@suse.de>
8597         PR middle-end/86479
8598         * fold-const.c (fold_binary_op_with_conditional_arg): Do not
8599         move possibly trapping operations into the conditional.
8601 2018-07-12  Richard Biener  <rguenther@suse.de>
8603         * tree-ssa-sccvn.c (mprts_hook_cnt): Remove.
8604         (vn_lookup_simplify_result): Remove recursion limit applied
8605         here.
8606         (vn_nary_build_or_lookup_1): Adjust.
8607         (try_to_simplify): Likewise.
8608         * gimple-match-head.c (gimple_resimplify1): Instead apply one
8609         here.
8610         (gimple_resimplify2): Likewise.
8611         (gimple_resimplify3): Likewise.
8612         (gimple_resimplify4): Likewise.
8614 2018-07-11  Jakub Jelinek  <jakub@redhat.com>
8616         * config/i386/avx512bitalgintrin.h (_mm512_mask_bitshuffle_epi64_mask):
8617         Use __mmask64 type instead of __mmask8 for __M argument.
8618         * config/i386/avx512fintrin.h (_mm512_mask_xor_epi64,
8619         _mm512_maskz_xor_epi64): Use __mmask8 type instead of __mmask16 for
8620         __U argument.
8621         (_mm512_mask_cmpneq_epi64_mask): Use __mmask8 type instead of
8622         __mmask16 for __M argument.
8623         (_mm512_maskz_insertf32x4, _mm512_maskz_inserti32x4,
8624         _mm512_mask_insertf32x4, _mm512_mask_inserti32x4): Cast last argument
8625         to __mmask16 instead of __mmask8.
8626         * config/i386/avx512vlintrin.h (_mm_mask_add_ps, _mm_maskz_add_ps,
8627         _mm256_mask_add_ps, _mm256_maskz_add_ps, _mm_mask_sub_ps,
8628         _mm_maskz_sub_ps, _mm256_mask_sub_ps, _mm256_maskz_sub_ps,
8629         _mm256_maskz_cvtepi32_ps, _mm_maskz_cvtepi32_ps): Use __mmask8 type
8630         instead of __mmask16 for __U argument.
8631         * config/i386/avx512vlbwintrin.h (_mm_mask_cmp_epi8_mask): Use
8632         __mmask16 instead of __mmask8 for __U argument.
8633         (_mm256_mask_cmp_epi8_mask): Use __mmask32 instead of __mmask16 for
8634         __U argument.
8635         (_mm256_cmp_epi8_mask): Use __mmask32 return type instead of
8636         __mmask16.
8637         (_mm_mask_cmp_epu8_mask): Use __mmask16 instead of __mmask8 for __U
8638         argument.
8639         (_mm256_mask_cmp_epu8_mask): Use __mmask32 instead of __mmask16 for
8640         __U argument.
8641         (_mm256_cmp_epu8_mask): Use __mmask32 return type instead of
8642         __mmask16.
8643         (_mm_mask_cmp_epi16_mask): Cast last argument to __mmask8 instead
8644         of __mmask16.
8645         (_mm256_mask_cvtepi8_epi16): Use __mmask16 instead of __mmask32 for
8646         __U argument.
8647         (_mm_mask_cvtepi8_epi16): Use __mmask8 instead of __mmask32 for
8648         __U argument.
8649         (_mm256_mask_cvtepu8_epi16): Use __mmask16 instead of __mmask32 for
8650         __U argument.
8651         (_mm_mask_cvtepu8_epi16): Use __mmask8 instead of __mmask32 for
8652         __U argument.
8653         (_mm256_mask_cmpneq_epu8_mask, _mm256_mask_cmplt_epu8_mask,
8654         _mm256_mask_cmpge_epu8_mask, _mm256_mask_cmple_epu8_mask): Change
8655         return type as well as __M argument type and all casts from __mmask8
8656         to __mmask32.
8657         (_mm256_mask_cmpneq_epu16_mask, _mm256_mask_cmplt_epu16_mask,
8658         _mm256_mask_cmpge_epu16_mask, _mm256_mask_cmple_epu16_mask): Change
8659         return type as well as __M argument type and all casts from __mmask8
8660         to __mmask16.
8661         (_mm256_mask_cmpneq_epi8_mask, _mm256_mask_cmplt_epi8_mask,
8662         _mm256_mask_cmpge_epi8_mask, _mm256_mask_cmple_epi8_mask): Change
8663         return type as well as __M argument type and all casts from __mmask8
8664         to __mmask32.
8665         (_mm256_mask_cmpneq_epi16_mask, _mm256_mask_cmplt_epi16_mask,
8666         _mm256_mask_cmpge_epi16_mask, _mm256_mask_cmple_epi16_mask): Change
8667         return type as well as __M argument type and all casts from __mmask8
8668         to __mmask16.
8669         * config/i386/avx512vbmi2vlintrin.h (_mm_mask_shrdi_epi32,
8670         _mm_mask_shldi_epi32): Cast last argument to __mmask8 instead of
8671         __mmask16.
8673 2018-07-11  Grazvydas Ignotas  <notasas@gmail.com>
8675         * config/i386/avx512bwintrin.h: (_mm512_mask_cmp_epi8_mask,
8676         _mm512_mask_cmp_epu8_mask): Use __mmask64 type instead of __mmask32
8677         for __U argument.
8679 2018-07-11  Paul Koning  <ni1d@arrl.net>
8681         * doc/md.texi (define_subst): Document how multiple occurrences of
8682         the same argument in the replacement pattern are handled.
8684 2018-07-11  Paul Koning  <ni1d@arrl.net>
8686         * doc/extend.texi (Common Variable Attributes): Move "mode" into
8687         alphabetical order.
8688         (Common Type Attributes): Add "mode" attribute.
8690 2018-07-11  Jan Hubicka  <hubicka@ucw.cz>
8692         * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not
8693         stream DECL_ORIGINAL_TYPE.
8694         (DFS::DFS_write_tree_body): Drop hack handling local external decls.
8695         (hash_tree): Do not walk DECL_ORIGINAL_TYPE.
8696         * tree-streamer-in.c (lto_input_ts_decl_non_common_tree_pointers):
8697         Do not walk original type.
8698         * tree-streamer-out.c (streamer_write_chain): Drop hack handling
8699         external decls.
8700         (write_ts_decl_non_common_tree_pointers): Do not stream
8701         DECL_ORIGINAL_TYPE
8702         * tree.c (free_lang_data_in_decl): Clear DECL_ORIGINAL_TYPE.
8703         (find_decls_types_r): Do not walk DEC_ORIGINAL_TYPE.
8705 2018-07-11  Aldy Hernandez  <aldyh@redhat.com>
8707         * tree-ssa-threadupdate.c (thread_through_all_blocks): Do not jump
8708         thread twice from the same starting edge.
8710 2018-07-11  Aldy Hernandez  <aldyh@redhat.com>
8712         * vr-values.c (gimple_stmt_nonzero_p): Abstract common code to...
8713         * gimple.c (gimple_call_nonnull_result_p): ...here...
8714         (gimple_call_nonnull_arg): ...and here.
8715         * gimple.h (gimple_call_nonnull_result_p): New.
8716         (gimple_call_nonnull_arg): New.
8718 2018-07-11  Richard Earnshaw  <rearnsha@arm.com>
8720         * config/arm/arm-cpus.in: Move information from fpu field of each
8721         cpu definition to the isa field.
8722         * config/arm/parsecpu.awk (fpu): Delete match rule.
8723         (gen_comm_data): Don't add bits from the CPU's FPU entry.
8725 2018-07-11  Richard Biener  <rguenther@suse.de>
8727         PR debug/86457
8728         * dwarf2out.c (init_sections_and_labels): Use
8729         output_asm_line_debug_info consistently.
8730         (dwarf2out_early_finish): Likewise.
8731         (dwarf2out_finish): Remove DW_AT_stmt_list from early generated
8732         type units.
8734 2018-07-11  Richard Biener  <rguenther@suse.de>
8736         * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1):
8737         Rework father_bb setting in a way to avoid propagating constants
8738         multiple times on a loop body.
8740 2018-07-10  Mark Wielaard  <mark@klomp.org>
8742         PR debug/86459
8743         * dwarf2out.c (output_macinfo_op): Fix dwarf_FORM typo in gcc_assert.
8745 2018-07-10  Richard Biener  <rguenther@suse.de>
8747         * hash-map.h (hash_map::iterator::operator*): Return
8748         references to key and value.
8750 2018-07-10  Jakub Jelinek  <jakub@redhat.com>
8752         PR c++/86443
8753         * gimplify.c (find_combined_omp_for): Add DATA argument, in addition
8754         to finding the inner OMP_FOR/OMP_SIMD stmt find non-trivial wrappers,
8755         BLOCKs with BLOCK_VARs, OMP_PARALLEL in between, OMP_FOR in between.
8756         (gimplify_omp_for): For composite loops, move outer
8757         OMP_{DISTRIBUTE,TASKLOOP,FOR,PARALLEL} right around innermost
8758         OMP_FOR/OMP_SIMD if there are any non-trivial wrappers.  For class
8759         iterators add any needed clauses.  Allow OMP_FOR_ORIG_DECLS to contain
8760         TREE_LIST for both the original class iterator and the "last" helper
8761         var.  Gimplify OMP_FOR_PRE_BODY before the outermost composite
8762         loop, remember has_decl_expr from outer composite loops for the
8763         innermost OMP_SIMD in TREE_PRIVATE bit on OMP_FOR_INIT.
8765 2018-07-09  Martin Sebor  <msebor@redhat.com>
8767         PR middle-end/77357
8768         PR middle-end/86428
8769         * builtins.c (c_strlen): Avoid out-of-bounds warnings when
8770         accessing implicitly initialized array elements.
8771         * expr.c (string_constant): Handle string initializers of
8772         character arrays within aggregates.
8773         * gimple-fold.c (fold_array_ctor_reference): Add argument.
8774         Store element offset.  As a special case, handle zero size.
8775         (fold_nonarray_ctor_reference): Same.
8776         (fold_ctor_reference): Add argument.  Store subobject offset.
8777         * gimple-fold.h (fold_ctor_reference): Add argument.
8779 2018-07-09  Paul Koning  <ni1d@arrl.net>
8781         * config/pdp11/pdp11.c (pdp11_addr_cost): New function.
8782         (pdp11_insn_cost): New function.
8783         (pdp11_md_asm_adjust): New function.
8784         (TARGET_INVALID_WITHIN_DOLOOP): Define.
8785         (pdp11_rtx_costs): Update to match machine better.
8786         (output_addr_const_pdp11): Correct format mismatch warnings.
8787         * config/pdp11/pdp11.h (SLOW_BYTE_ACCESS): Correct definition.
8788         * config/pdp11/pdp11.md: General change to add base_cost and/or
8789         length attributes for use by new pdp11_insn_cost function.
8790         (MIN_BRANCH): Correct definition.
8791         (MIN_SOB): Ditto.
8792         (doloop_end): Use standard pattern name for looping pattern.
8793         (doloop_end_nocc): New.
8794         (movsf): Add another constraint alternative.
8795         (zero_extendqihi2): Add constraint alternatives for not in place
8796         extend.
8797         (zero_extendhisi2): Remove.
8798         (shift patterns): Add CC handling variants.
8799         (bswaphi2): New.
8800         (bswapsi2): New.
8801         (rothi3): New.
8802         (define_peephole2): New peephole to recognize mov that sets CC for
8803         subsequent test.
8805 2018-07-09  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
8807         * config/sparc/sparc.c (sparc_fold_builtin) <SPARC_BUILTIN_PDIST,
8808         SPARC_BUILTIN_PDISTN>: Adapt for signature change of wi::neg,
8809         wi::add.
8811 2018-07-09  Jakub Jelinek  <jakub@redhat.com>
8813         PR c/86420
8814         * real.c (real_nextafter): Return true if result is denormal.
8816 2018-07-09  Martin Liska  <mliska@suse.cz>
8818         * common.opt: Add back wrongly removed attribute.
8820 2018-07-09  Richard Biener  <rguenther@suse.de>
8822         PR debug/86413
8823         * dwarf2out.c (gen_block_die): For an early generated DIE
8824         always output high/low PC attributes.
8826 2018-07-09  Tom de Vries  <tdevries@suse.de>
8828         * cfgexpand.c (expand_debug_source_expr): Handle VAR_DECL.
8829         * tree-inline.c (remap_ssa_name): Handle default def ssa_name mapping
8830         onto VAR_DECL with abstract origin.
8832 2018-07-07  Jim Wilson  <jimw@sifive.com>
8834         * config/riscv/riscv.c (TARGET_CUSTOM_FUNCTION_DESCRIPTORS): New.
8836 2018-07-07  Tom de Vries  <tdevries@suse.de>
8838         * tree-dfa.c (dump_enumerated_decls): Handle cfun->cfg == NULL.
8840 2018-07-07  Aldy Hernandez  <aldyh@redhat.com>
8842         * tree-vrp.c (vrp_int_const_binop): Change overflow type to
8843         overflow_type.
8844         (combine_bound): Use wide-int overflow calculation instead of
8845         rolling our own.
8846         * calls.c (maybe_warn_alloc_args_overflow): Change overflow type to
8847         overflow_type.
8848         * fold-const.c (int_const_binop_2): Same.
8849         (extract_muldiv_1): Same.
8850         (fold_div_compare): Same.
8851         (fold_abs_const): Same.
8852         * match.pd: Same.
8853         * poly-int.h (add): Same.
8854         (sub): Same.
8855         (neg): Same.
8856         (mul): Same.
8857         * predict.c (predict_iv_comparison): Same.
8858         * profile-count.c (slow_safe_scale_64bit): Same.
8859         * simplify-rtx.c (simplify_const_binary_operation): Same.
8860         * tree-chrec.c (tree_fold_binomial): Same.
8861         * tree-data-ref.c (split_constant_offset_1): Same.
8862         * tree-if-conv.c (idx_within_array_bound): Same.
8863         * tree-scalar-evolution.c (iv_can_overflow_p): Same.
8864         * tree-ssa-phiopt.c (minmax_replacement): Same.
8865         * tree-vect-loop.c (is_nonwrapping_integer_induction): Same.
8866         * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Same.
8867         * vr-values.c (vr_values::adjust_range_with_scev): Same.
8868         * wide-int.cc (wi::add_large): Same.
8869         (wi::mul_internal): Same.
8870         (wi::sub_large): Same.
8871         (wi::divmod_internal): Same.
8872         * wide-int.h: Change overflow type to overflow_type for neg, add,
8873         mul, smul, umul, div_trunc, div_floor, div_ceil, div_round,
8874         mod_trunc, mod_ceil, mod_round, add_large, sub_large,
8875         mul_internal, divmod_internal.
8876         (overflow_type): New enum.
8877         (accumulate_overflow): New.
8879 2018-07-06  Kugan Vivekanandarajah  <kugan.vivekanandarajah@linaro.org>
8881         * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): New.
8882         (tree_ssa_phiopt_worker): Call cond_removal_in_popcount_pattern.
8884 2018-07-06  Kugan Vivekanandarajah  <kuganv@linaro.org>
8886         * tree-ssa-loop-niter.c (number_of_iterations_popcount): If popcount
8887         argument is checked for zero before entering loop, avoid checking again.
8889 2018-07-06  Kugan Vivekanandarajah  <kuganv@linaro.org>
8891         * gimplify.h (generic_expr_could_trap_p): Set as global function.
8892         * gimplify.h (generic_expr_could_trap_p): Likwise.
8893         * tree-scalar-evolution.c (expression_expensive_p): Handle COND_EXPR.
8895 2018-07-06  Jakub Jelinek  <jakub@redhat.com>
8897         PR tree-optimization/86401
8898         * fold-const.c (fold_binary_loc) <case BIT_AND_EXPR>: Move the
8899         ((A & N) + B) & M -> (A + B) & M etc. optimization into ...
8900         (fold_bit_and_mask): ... here.  New helper function for match.pd.
8901         * fold-const.h (fold_bit_and_mask): Declare.
8902         * match.pd (((A & N) + B) & M -> (A + B) & M): New optimization.
8904 2018-07-06  Peter Bergner  <bergner@linux.ibm.com>
8906         PR target/86324
8907         * target.def (translate_mode_attribute): New hook.
8908         * targhooks.h (default_translate_mode_attribute): Declare.
8909         * targhooks.c (default_translate_mode_attribute): New function.
8910         * doc/tm.texi.in (TARGET_TRANSLATE_MODE_ATTRIBUTE): New hook.
8911         * doc/tm.texi: Regenerate.
8912         * config/rs6000/rs6000.c (TARGET_TRANSLATE_MODE_ATTRIBUTE): Define.
8913         (rs6000_translate_mode_attribute): New function.
8915 2018-07-06  Paul Koning  <ni1d@arrl.net>
8917         * doc/md.texi (define_split): Document DONE and FAIL.
8918         (define_peephole2): Ditto.
8920 2018-07-05  Jeff Law  <law@redhat.com>
8922         PR tree-optimization/86010
8923         * tree-ssa-dse.c (compute_trims): More aggressively trim at
8924         both the head and tail of mem* and str* calls.
8926 2018-07-05  Jim Wilson  <jimw@sifive.com>
8928         * config.gcc (riscv*-*-*): When setting xlen, handle riscv-*.
8930 2018-07-05  Indu Bhagat  <indu.bhagat@oracle.com>
8932         * config/aarch64/aarch64-simd.md: correct flags text for
8933         MIN_EXPR replacement.
8935 2018-07-05  James Clarke  <jrtc27@jrtc27.com>
8937         * configure: Regenerated.
8939 2018-07-05  Carl Love  <cel@us.ibm.com>
8941         * config/rs6000/rs6000-c.c: Map ALTIVEC_BUILTIN_VEC_UNPACKH for
8942         float argument to VSX_BUILTIN_DOUBLEH_V4SF.
8943         Map ALTIVEC_BUILTIN_VEC_UNPACKL for float argument to
8944         VSX_BUILTIN_DOUBLEL_V4SF.
8946 2018-07-05  Martin Sebor  <msebor@redhat.com>
8948         PR c++/86400
8949         * tree-ssa-strlen.c (maybe_set_strlen_range): Use type size rather
8950         than its domain to compute its the upper bound of a char array.
8952 2018-07-05  Nathan Sidwell  <nathan@acm.org>
8954         Replace NO_IMPLICIT_EXTERN_C with SYSTEM_IMPLICIT_EXTERN_C.
8955         * doc/cpp.texi: Update comment.
8956         * doc/tm.texi: Rebuilt.
8957         * doc/tm.texi.in (NO_IMPLICIT_EXTERN_C): Replace with ...
8958         (SYSTEM_IMPLICIT_EXTERN_C): ... this, opposite sense.
8959         * doc/extend.texi (Backwards Compatibility): Clarify it is system
8960         headers affected by extern "C".
8961         * system.h: Poison NO_IMPLICIT_EXTERN_C.
8962         * config/alpha/alpha.h, config/arm/uclinux-elf.h,
8963         config/bfin/elf.h, config/cris/cris.h, config/darwin.h,
8964         config/dragonfly.h, config/freebsd.h, config/gnu-user.h,
8965         config/i386/cygming.h, config/i386/djgpp.h, config/i386/nto.h,
8966         config/ia64/hpux.h, config/lm32/lm32.h, config/lm32/uclinux-elf.h,
8967         config/lynx.h, config/mips/elf.h, config/mmix/mmix.h,
8968         config/netbsd.h, config/pa/pa-hpux.h, config/powerpcspe/sysv4.h,
8969         config/riscv/elf.h, config/rs6000/sysv4.h, config/rtems.h,
8970         config/s390/tpf.h, config/sh/newlib.h, config/sol2.h,
8971         config/sparc/openbsd64.h, config/sparc/sp-elf.h,
8972         config/sparc/sp64-elf.h, config/spu/spu.h,
8973         config/stormy16/stormy16.h, config/v850/v850.h,
8974         config/visium/visium.h, config/vx-common.h, config/xtensa/elf.h: Don't
8975         define NO_IMPLICIT_EXTERN_C.
8976         * config/rs6000/aix.h: Set SYSTEM_IMPLICIT_EXTERN_C.
8978 2018-07-05  Tamar Christina  <tamar.christina@arm.com>
8980         PR target/84711
8981         * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
8982         instead of GET_MODE_SIZE when comparing Units.
8984 2018-07-05  Tamar Christina  <tamar.christina@arm.com>
8986         PR target/84711
8987         * rtlanal.c (set_noop_p): Constrain on mode change,
8988         include hard-reg-set.h
8990 2018-07-05  Tamar Christina  <tamar.christina@arm.com>
8992         * config/aarch64/aarch64.c (aarch64_expand_movmem): Fix mode size.
8994 2018-07-05  Jakub Jelinek  <jakub@redhat.com>
8996         Revert
8997         2018-07-04  Maxim Ostapenko  <m.ostapenko@samsung.com>
8999         PR sanitizer/84250
9000         * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
9001         libasan.
9002         * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
9004 2018-07-04  Maxim Ostapenko  <m.ostapenko@samsung.com>
9006         PR sanitizer/84250
9007         * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
9008         libasan.
9009         * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
9011 2018-07-04  Eric Botcazou  <ebotcazou@adacore.com>
9013         PR middle-end/86380
9014         * expmed.c (choose_multiplier): Fix incorrect comparison with mask.
9016 2018-07-04  Aldy Hernandez  <aldyh@redhat.com>
9018         * tree-vrp.c (extract_range_from_binary_expr_1): Initialize
9019         neg_*_op* variables.
9021 2018-07-04  Martin Liska  <mliska@suse.cz>
9023         * tree-switch-conversion.c: Define
9024         max_ratio_for_speed and max_ratio_for_size constants.
9026 2018-07-04  Denys Vlasenko  <dvlasenk@redhat.com>
9027             Martin Liska  <mliska@suse.cz>
9029         PR middle-end/66240
9030         PR target/45996
9031         PR c/84100
9032         * common.opt: Rename align options with 'str_' prefix.
9033         * common/config/i386/i386-common.c (set_malign_value): New
9034         function.
9035         (ix86_handle_option): Use it to set -falign-* options/
9036         * config/aarch64/aarch64-protos.h (struct tune_params): Change
9037         type from int to string.
9038         * config/aarch64/aarch64.c: Update default values from int
9039         to string.
9040         * config/alpha/alpha.c (alpha_override_options_after_change):
9041         Likewise.
9042         * config/arm/arm.c (arm_override_options_after_change_1): Likewise.
9043         * config/i386/dragonfly.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
9044         max skip conditionally.
9045         * config/i386/freebsd.h (SUBALIGN_LOG): New.
9046         (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
9047         max skip conditionally.
9048         * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
9049         max skip conditionally.
9050         * config/i386/gnu-user.h (SUBALIGN_LOG): New.
9051         (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
9052         max skip conditionally.
9053         * config/i386/i386.c (struct ptt): Change type from int to
9054         string.
9055         (ix86_default_align): Set default values.
9056         * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Print
9057         max skip conditionally.
9058         * config/i386/iamcu.h (SUBALIGN_LOG): New.
9059         (ASM_OUTPUT_MAX_SKIP_ALIGN):
9060         * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN):
9061         * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
9062         max skip conditionally.
9063         * config/i386/openbsdelf.h (SUBALIGN_LOG): New.
9064         (ASM_OUTPUT_MAX_SKIP_ALIGN) Print max skip conditionally.:
9065         * config/i386/x86-64.h (SUBALIGN_LOG): New.
9066         (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
9067         max skip conditionally.
9068         (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
9069         * config/ia64/ia64.c (ia64_option_override): Set default values
9070         for alignment options.
9071         * config/m68k/m68k.c: Handle new str_align_* options.
9072         * config/mips/mips.c (mips_set_compression_mode): Change
9073         type of constants.
9074         (mips_option_override): Set default values for options.
9075         * config/powerpcspe/powerpcspe.c (rs6000_option_override_internal):
9076         Likewise.
9077         * config/rs6000/rs6000.c (rs6000_option_override_internal):
9078         Likewise.
9079         * config/rx/rx.c (rx_option_override): Likewise.
9080         * config/rx/rx.h (JUMP_ALIGN): Use align_jumps_log.
9081         (LABEL_ALIGN): Use align_labels_log.
9082         (LOOP_ALIGN): Use align_loops_align.
9083         * config/s390/s390.c (s390_asm_output_function_label): Use new
9084         macros.
9085         * config/sh/sh.c (sh_override_options_after_change):
9086         Change type of constants.
9087         * config/spu/spu.c (spu_sched_init): Likewise.
9088         * config/sparc/sparc.c (sparc_option_override): Set default
9089         values for options.
9090         * config/visium/visium.c (visium_option_override): Likewise.
9091         * config/visium/visium.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Do not
9092         emit p2align format with last argument if it's not needed.
9093         * doc/invoke.texi: Document extended format of -falign-*.
9094         * final.c: Use align_labels alignment.
9095         * flags.h (struct target_flag_state): Change type to use
9096         align_flags.
9097         (struct align_flags_tuple): New.
9098         (struct align_flags): Likewise.
9099         (align_loops_log): Redefine macro to use new types.
9100         (align_loops_max_skip): Redefine macro to use new types.
9101         (align_jumps_log): Redefine macro to use new types.
9102         (align_jumps_max_skip): Redefine macro to use new types.
9103         (align_labels_log): Redefine macro to use new types.
9104         (align_labels_max_skip): Redefine macro to use new types.
9105         (align_functions_log): Redefine macro to use new types.
9106         (align_loops): Redefine macro to use new types.
9107         (align_jumps): Redefine macro to use new types.
9108         (align_labels): Redefine macro to use new types.
9109         (align_functions): Redefine macro to use new types.
9110         (align_functions_max_skip): Redefine macro to use new types.
9111         (align_loops_value): New macro.
9112         (align_jumps_value): New macro.
9113         (align_labels_value): New macro.
9114         (align_functions_value): New macro.
9115         * function.c (invoke_set_current_function_hook): Propagate
9116         alignment values from flags to global variables default in
9117         topleev.h.
9118         * ipa-icf.c (sem_function::equals_wpa): Use
9119         cl_optimization_option_eq instead of memcmp.
9120         * lto-streamer.h (cl_optimization_stream_out): Support streaming
9121         of string types.
9122         (cl_optimization_stream_in): Likewise.
9123         * optc-save-gen.awk: Support strings in cl_optimization.
9124         * opth-gen.awk: Likewise.
9125         * opts.c (finish_options): Remove error checking of invalid
9126         value ranges.
9127         (MAX_CODE_ALIGN): Remove.
9128         (MAX_CODE_ALIGN_VALUE): Likewise.
9129         (parse_and_check_align_values): New function.
9130         (check_alignment_argument): Likewise.
9131         (common_handle_option): Use check_alignment_argument.
9132         * opts.h (parse_and_check_align_values): Declare.
9133         * toplev.c (init_alignments): Remove.
9134         (read_log_maxskip): New.
9135         (parse_N_M): Likewise.
9136         (parse_alignment_opts): Likewise.
9137         (backend_init_target): Remove usage of init_alignments.
9138         * toplev.h (parse_alignment_opts): Declare.
9139         * tree-streamer-in.c (streamer_read_tree_bitfields): Add new
9140         argument.
9141         * tree-streamer-out.c (streamer_write_tree_bitfields): Likewise.
9142         * tree.c (cl_option_hasher::equal): New.
9143         * varasm.c: Use new global macros.
9145 2018-07-04  Denys Vlasenko  <dvlasenk@redhat.com>
9147         * config/i386/dragonfly.h: (ASM_OUTPUT_MAX_SKIP_ALIGN):
9148         Use a simpler align directive also if MAXSKIP = ALIGN-1.
9149         * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
9150         * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
9151         * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
9152         * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
9153         * config/i386/freebsd.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Remove "If N
9154         is large, do at least 8 byte alignment" code. Add SUBALIGN_LOG
9155         define. Use a simpler align directive also if MAXSKIP = ALIGN-1.
9156         * config/i386/gnu-user.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
9157         * config/i386/iamcu.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
9158         * config/i386/openbsdelf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
9159         * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
9161 2018-07-04  Martin Liska  <mliska@suse.cz>
9162             Jonathan Wakely  <jwakely@redhat.com>
9164         * coverage.c: Use correct type.
9165         * doc/invoke.texi: Language correction.
9167 2018-07-03  H.J. Lu  <hongjiu.lu@intel.com>
9169         PR target/85620
9170         * config/i386/i386.c (rest_of_insert_endbranch): Also generate
9171         ENDBRANCH for non-tail call which may return via indirect branch.
9172         * doc/extend.texi: Document indirect_return attribute.
9174 2018-07-03  Martin Sebor  <msebor@redhat.com>
9176         PR tree-optimization/86274
9177         * gimple-ssa-sprintf.c (fmtresult::type_max_digits): Verify
9178         precondition.
9179         (format_floating): Correct handling of infinities and NaNs.
9181 2018-07-03  Martin Sebor  <msebor@redhat.com>
9183         * print-tree.c (print_real_cst): New function.
9184         (print_node_brief): Call it.
9185         (print_node): Ditto.
9187 2018-07-03  Jeff Law  <law@redhat.com>
9189         * config/h8300/h8300.md (logical<mode>3_sn, logical<mode>3): Merge
9190         into a single pattern.
9192         * config/h8300/h8300.md (ors code_iterator): New.
9193         (bsetqi_msx, bnotqi_msx patterns and splitters): Consolidate into
9194         a single pattern and single splitter.
9195         (bsethi_msx, bnothi_msx patterns): Consolidate into a single pattern.
9196         (iorqi3_1, xorqi3_1): Likewise.
9197         (iorqi3, xorqi3 expanders): Similarly.
9199         * config/h8300/h8300.md (movmd_internal_normal): Consolidated with
9200         (movmd_internal) into a single pattern using the P mode iterator.
9201         (movmd splitters): Similarly.
9202         (stpcpy_internal_normal, stpcpy_internal): Similarly for thes patterns.
9203         (movsd splitters): Similarly.
9205         * config/h8300/h8300.c (h8300_insn_length_from_table): Consolidate
9206         ADDB, ADDW and ADDL into a single ADD attribute which selects the
9207         right table based on the size of the operand.
9208         * config/h8300/h8300.md (length_table): Corresponding changes. All
9209         references to "addb", "addw" and "addl" changed to "add".
9210         (btst patterns): Merge two variants into a single pattern.
9211         (tstqi, tsthi): Likewise.
9212         (addhi3_incdec, addsi3_incdec): Likewise.
9213         (subhi3_h8300hs, subsi3_h8300hs): Likewise.
9214         (mulhi3, mulsi3): Likewise.
9215         (udivhi3, udivsi3): Likewise.
9216         (divhi3, divsi3): Likewise.
9217         (andorqi3, andorhi3, andorsi3): Likewise.
9219 2018-07-03  Uros Bizjak  <ubizjak@gmail.com>
9221         PR target/85694
9222         * config/i386/sse.md (uavg<mode>3_ceil): New expander.
9223         (<sse2_avx2>_uavg<mode>3<mask_name>): Simplify expander.
9225 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
9227         PR tree-optimization/85694
9228         * config/aarch64/iterators.md (HADD, RHADD): New int iterators.
9229         (u): Handle UNSPEC_SHADD, UNSPEC_UHADD, UNSPEC_SRHADD and
9230         UNSPEC_URHADD.
9231         * config/aarch64/aarch64-simd.md (<u>avg<mode>3_floor)
9232         (<u>avg<mode>3_ceil): New patterns.
9234 2018-07-03  David Malcolm  <dmalcolm@redhat.com>
9236         * gcc.dg/vect/slp-perm-1.c: Remove "note: " prefix from
9237         scan-tree-dump directive.
9238         * gcc.dg/vect/slp-perm-2.c: Likewise.
9239         * gcc.dg/vect/slp-perm-3.c: Likewise.
9240         * gcc.dg/vect/slp-perm-5.c: Likewise.
9241         * gcc.dg/vect/slp-perm-6.c: Likewise.
9242         * gcc.dg/vect/slp-perm-7.c: Likewise.
9243         * gcc.dg/vect/slp-perm-8.c: Likewise.
9245 2018-07-03  Marek Polacek  <polacek@redhat.com>
9247         PR middle-end/86202
9248         * gimple-fold.c (size_must_be_zero_p): Check the type of the size.
9250 2018-07-03  Richard Biener  <rguenther@suse.de>
9252         PR ipa/86389
9253         * tree-ssa-structalias.c (find_func_clobbers): Properly
9254         handle indirect calls.
9256 2018-07-03  Jeff Law  <law@redhat.com>
9258         * config/h8300/h8300.md (HSI, QHSI, QHSIF): New mode iterators.
9259         (shifts): New code iterator.
9260         (movqi, movhi, movsi, movsf expanders): Consolidate into a single
9261         expander.  Fix HImode handling on H8/SX.
9262         (addqi3, addhi3, addsi3 expanders): Consolidate into a single expander.
9263         (subqi3, subhi3, subsi3 expanders): Likewise.
9264         (andqi3, andhi3, andsi3 expanders): Likewise.
9265         (iorqi3, iorhi3, iorsi3 expanders): Likewise.
9266         (xorqi3, xorhi3, xorsi3 expanders): Likewise.
9267         (negqi2, neghi2, negsi2, negsf2 expanders): Likewise.
9268         (one_cmplqi2, one_cmplhi2, one_cmplsi2): Likewise.
9269         (zero_extendqihi2, zero_extendqisi2): Likewise.
9270         (extendqihi2, extendqisi2): Likewise.
9271         (rotlqi3, rotlhi3, rotlsi3): Likewise.
9272         (neghi2_h8300, negsi2_h8300): Likewise for these patterns.
9273         (rotlqi3_1, rotlhi3_1): Likewise.
9274         (logicalhi3_sn, logicalsi3_sn): Likewise.
9275         (logicalhi3, logicalsi3): Likewise.
9277 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
9279         * tree-vect-patterns.c (vect_recog_rotate_pattern)
9280         (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
9281         (vect_recog_mixed_size_cond_pattern, adjust_bool_pattern_cast)
9282         (adjust_bool_pattern, vect_recog_bool_pattern): Pass the vector
9283         type to append_pattern_def_seq instead of creating a stmt_vec_info
9284         directly.
9285         (build_mask_conversion): Likewise.  Remove vinfo argument.
9286         (vect_add_conversion_to_patterm): Likewise, renaming to...
9287         (vect_add_conversion_to_pattern): ...this.
9288         (vect_recog_mask_conversion_pattern): Update call to
9289         build_mask_conversion.  Pass the vector type to
9290         append_pattern_def_seq here too.
9291         (vect_recog_gather_scatter_pattern): Update call to
9292         vect_add_conversion_to_pattern.
9294 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
9296         * tree-vect-patterns.c (new_pattern_def_seq): Delete.
9297         (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
9298         (vect_recog_widen_op_pattern, vect_recog_over_widening_pattern)
9299         (vect_recog_rotate_pattern, vect_synth_mult_by_constant): Don't set
9300         STMT_VINFO_PATTERN_DEF_SEQ to null here.
9301         (vect_recog_pow_pattern, vect_recog_vector_vector_shift_pattern)
9302         (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern): Use
9303         append_pattern_def_seq instead of new_pattern_def_seq.
9304         (vect_recog_divmod_pattern): Do both of the above.
9305         (vect_pattern_recog_1): Assert that STMT_VINO_PATTERN_DEF_SEQ
9306         is null.
9308 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
9310         * tree-vect-patterns.c (vect_recog_dot_prod_pattern):
9311         (vect_recog_sad_pattern, vect_recog_widen_op_pattern)
9312         (vect_recog_widen_mult_pattern, vect_recog_pow_pattern):
9313         (vect_recog_widen_sum_pattern, vect_recog_over_widening_pattern)
9314         (vect_recog_average_pattern, vect_recog_cast_forwprop_pattern)
9315         (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
9316         (vect_recog_vector_vector_shift_pattern, vect_synth_mult_by_constant)
9317         (vect_recog_mult_pattern, vect_recog_divmod_pattern)
9318         (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
9319         (vect_recog_mask_conversion_pattern): Replace vec<gimple *>
9320         parameter with a single stmt_vec_info.
9321         (vect_recog_func_ptr): Likewise.
9322         (vect_recog_gather_scatter_pattern): Likewise, folding in...
9323         (vect_try_gather_scatter_pattern): ...this.
9324         (vect_pattern_recog_1): Remove stmts_to_replace and just pass
9325         the stmt_vec_info of the statement to be matched.  Don't clear
9326         STMT_VINFO_RELATED_STMT.
9327         (vect_pattern_recog): Update call accordingly.
9329 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
9331         PR tree-optimization/85694
9332         * doc/md.texi (avgM3_floor, uavgM3_floor, avgM3_ceil)
9333         (uavgM3_ceil): Document new optabs.
9334         * doc/sourcebuild.texi (vect_avg_qi): Document new target selector.
9335         * internal-fn.def (IFN_AVG_FLOOR, IFN_AVG_CEIL): New internal
9336         functions.
9337         * optabs.def (savg_floor_optab, uavg_floor_optab, savg_ceil_optab)
9338         (savg_ceil_optab): New optabs.
9339         * tree-vect-patterns.c (vect_recog_average_pattern): New function.
9340         (vect_vect_recog_func_ptrs): Add it.
9341         * tree-vect-stmts.c (vectorizable_call): Get the type of the zero
9342         constant directly from the associated lhs.
9344 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
9346         * tree-vect-patterns.c (vect_split_statement): New function.
9347         (vect_convert_input): Use it to try to split an existing cast.
9349 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
9351         * poly-int.h (print_hex): New function.
9352         * dumpfile.h (dump_dec, dump_hex): Declare.
9353         * dumpfile.c (dump_dec, dump_hex): New poly_wide_int functions.
9354         * tree-vectorizer.h (_stmt_vec_info): Add min_output_precision,
9355         min_input_precision, operation_precision and operation_sign.
9356         * tree-vect-patterns.c (vect_get_range_info): New function.
9357         (vect_same_loop_or_bb_p, vect_single_imm_use)
9358         (vect_operation_fits_smaller_type): Delete.
9359         (vect_look_through_possible_promotion): Add an optional
9360         single_use_p parameter.
9361         (vect_recog_over_widening_pattern): Rewrite to use new
9362         stmt_vec_info infomration.  Handle one operation at a time.
9363         (vect_recog_cast_forwprop_pattern, vect_narrowable_type_p)
9364         (vect_truncatable_operation_p, vect_set_operation_type)
9365         (vect_set_min_input_precision): New functions.
9366         (vect_determine_min_output_precision_1): Likewise.
9367         (vect_determine_min_output_precision): Likewise.
9368         (vect_determine_precisions_from_range): Likewise.
9369         (vect_determine_precisions_from_users): Likewise.
9370         (vect_determine_stmt_precisions, vect_determine_precisions): Likewise.
9371         (vect_vect_recog_func_ptrs): Put over_widening first.
9372         Add cast_forwprop.
9373         (vect_pattern_recog): Call vect_determine_precisions.
9375 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
9377         * tree-vect-patterns.c (vect_mark_pattern_stmts): Remove pattern
9378         statements that have been replaced by further pattern statements.
9379         (vect_pattern_recog_1): Clear STMT_VINFO_PATTERN_DEF_SEQ on failure.
9381 2018-07-03  Richard Biener  <rguenther@suse.de>
9383         * tree-vect-stmts.c (vect_is_simple_use): Consolidate dumping,
9384         always set *dt.  Dump vectype in vectype overload.
9385         * dumpfile.h (dump_gimple_expr): New function.
9386         (dump_gimple_expr_loc): Likewise.
9387         * dumpfile.c (dump_gimple_expr): New function.
9388         (dump_gimple_expr_loc): Likewise.
9390 2018-07-02  Jeff Law  <law@redhat.com>
9392         * config/h8300/h8300.md (movqi_h8300, movqi_h8300hs): Consolidate
9393         the H8/300, H8/300H and H8/S variants into a single pattern.
9394         (movhi_h8300, movqi_h8300hs): Similarly.
9395         (pushqi_h8300hs, pushhi_h8300hs): Consolidate into a single pattern.
9396         (QHI mode iterator): New.
9398         * config/h8300/h8300.md: Remove trailing whitespace.
9400 2018-07-02  Jim Wilson  <jimw@sifive.com>
9402         * config/riscv/riscv.c (riscv_expand_epilogue): Use emit_jump_insn
9403         instead of emit_insn for interrupt returns.
9404         * config/riscv/riscv.md (riscv_met): Add (return) to rtl.
9405         (riscv_sret, riscv_uret): Likewise.
9407 2018-07-02  David Malcolm  <dmalcolm@redhat.com>
9409         * pretty-print.c (selftest::test_pp_format): Move save and restore
9410         of quotes to class auto_fix_quotes, and add an instance.
9411         * selftest.c: Include "intl.h".
9412         (selftest::auto_fix_quotes::auto_fix_quotes): New ctor.
9413         (selftest::auto_fix_quotes::~auto_fix_quotes): New dtor.
9414         * selftest.h (selftest::auto_fix_quotes): New class.
9416 2018-07-02  Richard Henderson  <richard.henderson@linaro.org>
9418         * config/aarch64/aarch64-protos.h, config/aarch64/aarch64.c
9419         (aarch64_sve_prepare_conditional_op): Remove.
9420         * config/aarch64/aarch64-sve.md (cond_<SVE_INT_BINARY><SVE_I>):
9421         Allow aarch64_simd_reg_or_zero as select operand; remove
9422         the aarch64_sve_prepare_conditional_op call.
9423         (cond_<SVE_INT_BINARY_SD><SVE_SDI>): Likewise.
9424         (cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
9425         (*cond_<SVE_INT_BINARY><SVE_I>_z): New pattern.
9426         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_z): New pattern.
9427         (*cond_<SVE_COND_FP_BINARY><SVE_F>_z): New pattern.
9428         (*cond_<SVE_INT_BINARY><SVE_I>_any): New pattern.
9429         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_any): New pattern.
9430         (*cond_<SVE_COND_FP_BINARY><SVE_F>_any): New pattern
9431         and a splitters to match all of the *_any patterns.
9432         * config/aarch64/predicates.md (aarch64_sve_any_binary_operator): New.
9434         * config/aarch64/iterators.md (SVE_INT_BINARY_REV): Remove.
9435         (SVE_COND_FP_BINARY_REV): Remove.
9436         (sve_int_op_rev, sve_fp_op_rev): New.
9437         * config/aarch64/aarch64-sve.md (*cond_<SVE_INT_BINARY><SVE_I>_0): New.
9438         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_0): New.
9439         (*cond_<SVE_COND_FP_BINARY><SVE_F>_0): New.
9440         (*cond_<SVE_INT_BINARY><SVE_I>_2): Rename, add movprfx alternative.
9441         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_2): Similarly.
9442         (*cond_<SVE_COND_FP_BINARY><SVE_F>_2): Similarly.
9443         (*cond_<SVE_INT_BINARY><SVE_I>_3): Similarly; use sve_int_op_rev.
9444         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_3): Similarly.
9445         (*cond_<SVE_COND_FP_BINARY><SVE_F>_3): Similarly; use sve_fp_op_rev.
9447         * config/aarch64/aarch64-sve.md (cond_<SVE_COND_FP_BINARY><SVE_F>):
9448         Remove match_dup 1 from the inner unspec.
9449         (*cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
9451         * config/aarch64/aarch64.md (movprfx): New attr.
9452         (length): Default movprfx to 8.
9453         * config/aarch64/aarch64-sve.md (*mul<SVE_I>3): Add movprfx alt.
9454         (*madd<SVE_I>, *msub<SVE_I): Likewise.
9455         (*<su>mul<SVE_I>3_highpart): Likewise.
9456         (*<SVE_INT_BINARY_SD><SVE_SDI>3): Likewise.
9457         (*v<ASHIFT><SVE_I>3): Likewise.
9458         (*<su><MAXMIN><SVE_I>3): Likewise.
9459         (*<su><MAXMIN><SVE_F>3): Likewise.
9460         (*fma<SVE_F>4, *fnma<SVE_F>4): Likewise.
9461         (*fms<SVE_F>4, *fnms<SVE_F>4): Likewise.
9462         (*div<SVE_F>4): Likewise.
9464 2018-07-02  Richard Sandiford  <richard.sandiford@arm.com>
9466         * tree-vect-patterns.c (vect_recog_widen_shift_pattern): Fix typo
9467         in dump string.
9469 2018-07-02  Richard Biener  <rguenther@suse.de>
9471         PR tree-optimization/86363
9472         * tree-ssa-sccvn.c (vn_reference_lookup_3): Check the
9473         memset argument refers to a non-variable address.
9475 2018-07-02  Aldy Hernandez  <aldyh@redhat.com>
9477         * tree-vrp.c (extract_range_from_binary_expr_1): Abstract a lot of the
9478         {PLUS,MINUS}_EXPR code to...
9479         (adjust_symbolic_bound): ...here,
9480         (combine_bound): ...here,
9481         (set_value_range_with_overflow): ...and here.
9483 2018-07-02  Aldy Hernandez  <aldyh@redhat.com>
9485         * tree-vrp.c (extract_range_from_unary_expr): Abstract ABS_EXPR
9486         code...
9487         (extract_range_from_abs_expr): ...here.
9489 2018-07-02  Eric Botcazou  <ebotcazou@adacore.com>
9491         * config/i386/i386.c (ix86_finalize_stack_frame_flags): Do not overrule
9492         -fno-omit-frame-pointer when not optimizing.
9494 2018-07-02  Martin Liska  <mliska@suse.cz>
9496         PR ipa/86279
9497         * ipa-pure-const.c (malloc_candidate_p): Revert usage of ::get.
9498         (propagate_nothrow): Likewise.
9500 2018-07-02  Martin Liska  <mliska@suse.cz>
9502         PR ipa/86323
9503         * ipa-inline.c (early_inliner): Revert wrongly added ::get call.
9505 2018-07-02  David Malcolm  <dmalcolm@redhat.com>
9507         * dumpfile.c (dump_generic_expr_loc): Undo removal of this
9508         function in r262149, changing "loc" param from source_location to
9509         const dump_location_t &.
9510         * dumpfile.h (dump_generic_expr_loc): Undo removal of this
9511         declaration, as above.
9513 2018-07-01  Paul Koning  <ni1d@arrl.net>
9515         * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
9516         -munit-asm, -mgnu-asm, -mdec-asm.
9517         * config/pdp11/pdp11-protos.h (pdp11_gen_int_label): New.
9518         (pdp11_output_labelref): New.
9519         (pdp11_output_def): New.
9520         (pdp11_output_addr_vec_elt): New.
9521         * config/pdp11/pdp11.c: Use tab between opcode and operands.  Use
9522         %# and %@ format codes.
9523         (pdp11_option_override): New.
9524         (TARGET_ASM_FILE_START_FILE_DIRECTIVE): Define.
9525         (pdp11_output_ident): New.
9526         (pdp11_asm_named_section): New.
9527         (pdp11_asm_init_sections): New.
9528         (pdp11_file_start): New.
9529         (pdp11_file_end): New.
9530         (output_ascii): Use .ascii/.asciz for -mdec-asm.
9531         (pdp11_asm_print_operand): Update %# and %$ for -mdec-asm.  Add
9532         %o, like %c but octal.
9533         (pdp11_option_override): New.
9534         * config/pdp11/pdp11.h (TEXT_SECTION_ASM_OP): Update for
9535         -mdec-asm.
9536         (DATA_SECTION_ASM_OP): Ditto.
9537         (READONLY_DATA_SECTION_ASM_OP): New.
9538         (IS_ASM_LOGICAL_LINE_SEPARATOR): New.
9539         (ASM_GENERATE_INTERNAL_LABEL): Use new function.
9540         (ASM_OUTPUT_LABELREF): Ditto.
9541         (ASM_OUTPUT_DEF): Ditto.
9542         (ASM_OUTPUT_EXTERNAL): New.
9543         (ASM_OUTPUT_SOURCE_FILENAME): New.
9544         (ASM_OUTPUT_ADDR_VEC_ELT): Use new function.
9545         (ASM_OUTPUT_SKIP): Update for -mdec-asm.
9546         * config/pdp11/pdp11.md: Use tab between opcode and operands.  Use
9547         %# and %@ format codes.
9548         * config/pdp11/pdp11.opt (mgnu-asm): New.
9549         (mdec-asm): Conflicts with -mgnu-asm and -munix-asm.
9550         (munix-asm): Conflicts with -mdec-asm and -mgnu-asm.
9551         * doc/invoke.txt (PDP-11 Options): Add -mgnu-asm.
9553 2018-07-01  Aldy Hernandez  <aldyh@redhat.com>
9555         * tree-ssa-threadupdate.c (mark_threaded_blocks): Avoid
9556         dereferencing path[] beyond its length.
9557         (debug_path): New.
9558         (debug_all_paths): New.
9559         (rewire_first_differing_edge): New.
9560         (adjust_paths_after_duplication): New.
9561         (duplicate_thread_path): Call adjust_paths_after_duplication.
9562         Add new argument.
9563         (thread_through_all_blocks): Add new argument to
9564         duplicate_thread_path.
9566 2018-06-30  Jim Wilson  <jimw@sifive.com>
9568         * config/riscv/predicates.md (p2m1_shift_operand): New.
9569         (high_mask_shift_operand): New.
9570         * config/riscv/riscv.md (lshrsi3_zero_extend_3+1): New combiner
9571         pattern using p2m1_shift_operand.
9572         (lshsi3_zero_extend_3+2): New combiner pattern using
9573         high_mask_shift_operand.
9575 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
9577         * tree-vect-patterns.c (vect_get_external_def_edge): New function,
9578         split out from...
9579         (vect_recog_rotate_pattern): ...here.
9580         (vect_convert_input): Try to insert casts of invariants in the
9581         preheader.
9582         * tree-vect-loop-manip.c (vect_loop_versioning): Don't require the
9583         preheader to be empty.
9585 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
9587         * tree-vect-patterns.c (append_pattern_def_seq): Take an optional
9588         vector type.  If given, install it in the new statement's
9589         STMT_VINFO_VECTYPE.
9590         (vect_element_precision): New function.
9591         (vect_unpromoted_value): New struct.
9592         (vect_unpromoted_value::vect_unpromoted_value): New function.
9593         (vect_unpromoted_value::set_op): Likewise.
9594         (vect_look_through_possible_promotion): Likewise.
9595         (vect_joust_widened_integer, vect_joust_widened_type): Likewise.
9596         (vect_widened_op_tree, vect_convert_input): Likewise.
9597         (vect_convert_inputs, vect_convert_output): Likewise.
9598         (vect_recog_dot_prod_pattern): Use vect_look_through_possible_promotion
9599         to handle the optional cast of the multiplication result and
9600         vect_widened_op_tree to detect the widened multiplication itself.
9601         Do not require the input and output of promotion casts to have
9602         the same sign, but base the signedness of the operation on the
9603         input rather than the result.  If the pattern includes two
9604         promotions, check that those promotions have the same sign.
9605         Do not restrict the MULT_EXPR handling to a double-width result;
9606         handle quadruple-width results and wider.  Use vect_convert_inputs
9607         to convert the inputs to the common type.
9608         (vect_recog_sad_pattern):  Use vect_look_through_possible_promotion
9609         to handle the optional cast of the ABS result.  Also allow a sign
9610         change or a sign extension between the ABS and MINUS.
9611         Use vect_widened_op_tree to detect the widened subtraction and use
9612         vect_convert_inputs to convert the inputs to the common type.
9613         (vect_handle_widen_op_by_const): Delete.
9614         (vect_recog_widen_op_pattern): New function.
9615         (vect_recog_widen_mult_pattern): Use it.
9616         (vect_recog_widen_shift_pattern): Likewise.
9617         (vect_recog_widen_sum_pattern): Use
9618         vect_look_through_possible_promotion to handle the promoted
9619         PLUS_EXPR operand.
9621 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
9623         * gimple-iterator.c (gsi_for_stmt): Add a new overload that takes
9624         the containing gimple_seq *.
9625         * gimple-iterator.h (gsi_for_stmt): Declare it.
9626         * tree-vect-patterns.c (vect_recog_dot_prod_pattern)
9627         (vect_recog_sad_pattern, vect_recog_widen_sum_pattern)
9628         (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
9629         (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
9630         (vect_recog_mask_conversion_pattern): Remove STMT_VINFO_IN_PATTERN_P
9631         checks.
9632         (vect_init_pattern_stmt, vect_set_pattern_stmt): New functions,
9633         split out from...
9634         (vect_mark_pattern_stmts): ...here.  Handle cases in which the
9635         statement being replaced is part of an existing pattern
9636         definition sequence, inserting the new pattern statements before
9637         the original one.
9638         (vect_pattern_recog_1): Don't return a bool.  If the statement
9639         is already part of a pattern, instead apply pattern matching
9640         to the pattern definition statements.  Don't clear the
9641         STMT_VINFO_RELATED_STMT if is_pattern_stmt_p.
9642         (vect_pattern_recog): Don't break after the first match;
9643         continue processing the pattern definition statements instead.
9644         Don't bail out for STMT_VINFO_IN_PATTERN_P here.
9646 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
9648         * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
9649         (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
9650         (vect_recog_widen_sum_pattern): Use it.
9652 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
9654         * tree-vect-loop.c (vectorizable_reduction): Assert that the
9655         phi is not a pattern statement and has not been replaced by
9656         a pattern statement.
9657         * tree-vect-patterns.c (type_conversion_p): Don't check
9658         STMT_VINFO_IN_PATTERN_P.
9659         (vect_recog_vector_vector_shift_pattern): Likewise.
9660         (vect_recog_dot_prod_pattern): Expect vect_is_simple_use to return
9661         the pattern statement rather than the original statement; check
9662         directly for a WIDEN_MULT_EXPR here.
9663         * tree-vect-slp.c (vect_get_and_check_slp_defs): Expect
9664         vect_is_simple_use to return the pattern statement rather
9665         than the original statement; use is_pattern_stmt_p to check
9666         for such a pattern statement.
9667         * tree-vect-stmts.c (process_use): Expect vect_is_simple_use
9668         to return the pattern statement rather than the original statement;
9669         don't do the same transformation here.
9670         (vect_is_simple_use): If the defining statement has been replaced
9671         by a pattern statement, return the pattern statement instead.
9672         Remove the corresponding (local) transformation from the vectype
9673         overload.
9675 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
9677         * tree-vectorizer.h (vect_is_simple_use): Move the gimple ** to the
9678         end and default to null.
9679         * tree-vect-loop.c (vect_create_epilog_for_reduction)
9680         (vectorizable_reduction): Update calls accordingly, dropping the
9681         gimple ** argument if the passed-back statement isn't needed.
9682         * tree-vect-patterns.c (vect_get_internal_def, type_conversion_p)
9683         (vect_recog_rotate_pattern): Likewise.
9684         (vect_recog_mask_conversion_pattern): Likewise.
9685         * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
9686         (vect_mask_constant_operand_p): Likewise.
9687         * tree-vect-stmts.c (is_simple_and_all_uses_invariant, process_use):
9688         (vect_model_simple_cost, vect_get_vec_def_for_operand): Likewise.
9689         (get_group_load_store_type, get_load_store_type): Likewise.
9690         (vect_check_load_store_mask, vect_check_store_rhs): Likewise.
9691         (vectorizable_call, vectorizable_simd_clone_call): Likewise.
9692         (vectorizable_conversion, vectorizable_assignment): Likewise.
9693         (vectorizable_shift, vectorizable_operation): Likewise.
9694         (vectorizable_store, vect_is_simple_cond): Likewise.
9695         (vectorizable_condition, vectorizable_comparison): Likewise.
9696         (get_same_sized_vectype, vect_get_mask_type_for_stmt): Likewise.
9697         (vect_is_simple_use): Rename the def_stmt argument to def_stmt_out
9698         and move it to the end.  Cope with null def_stmt_outs.
9700 2018-06-30  Bernd Edlinger  <bernd.edlinger@hotmail.de>
9702         * Makefile.in (FLAGS_TO_PASS): Add RANLIB_FOR_TARGET.
9704 2018-06-29  Jeff Law  <law@redhat.com>
9706         * config/v850/v850.c (v850_legitimate_address_p): Handle large
9707         displacements for TARGET_V850E2V3 and newer.
9708         (TARGET_LRA_P): Remove.  Defaults to LRA now.
9709         * config/v850/v850.md (sign23byte_load): Remove.
9710         (unsign23byte_load, sign23hword_load, unsign23hword_load): Likewise.
9711         (23word_load, 23byte_store, 23hword_store, 23word_store): Likewise.
9713 2018-06-29  Martin Liska  <mliska@suse.cz>
9715         PR lto/85759
9716         * coverage.c (coverage_init): Mangle full path name.
9717         * doc/invoke.texi: Document the change.
9718         * gcov-io.c (mangle_path): New.
9719         * gcov-io.h (mangle_path): Likewise.
9720         * gcov.c (mangle_name): Use mangle_path for path mangling.
9722 2018-06-29  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
9724         * config/arm/arm.c (output_move_double): Don't allow STRD instructions
9725         if starting source register is not even.
9727 2018-06-29  Martin Liska  <mliska@suse.cz>
9729         PR tree-optimization/86263
9730         * tree-switch-conversion.c (switch_decision_tree::try_switch_expansion):
9731         Make edge redirection.
9733 2018-06-29  David Malcolm  <dmalcolm@redhat.com>
9735         * dumpfile.c (dump_loc): Add indentation based on scope depth.
9736         (dump_scope_depth): New variable.
9737         (get_dump_scope_depth): New function.
9738         (dump_begin_scope): New function.
9739         (dump_end_scope): New function.
9740         * dumpfile.h (get_dump_scope_depth): New declaration.
9741         (dump_begin_scope): New declaration.
9742         (dump_end_scope): New declaration.
9743         (class auto_dump_scope): New class.
9744         (AUTO_DUMP_SCOPE): New macro.
9745         * tree-vectorizer.h (DUMP_VECT_SCOPE): Reimplement in terms of
9746         AUTO_DUMP_SCOPE.
9748 2018-06-29  Richard Biener  <rguenther@suse.de>
9750         * tree-vect-data-refs.c (vect_analyze_data_ref_dependences): Assert
9751         compute_all_dependences succeeds.
9752         * tree-vect-loop.c (vect_get_datarefs_in_loop): Fail early if we
9753         exceed --param loop-max-datarefs-for-datadeps.
9755 2018-06-29  Jakub Jelinek  <jakub@redhat.com>
9757         * config/rs6000/t-rs6000: Append rs6000-modes.h to TM_H.
9759 2018-06-28  Uros Bizjak  <ubizjak@gmail.com>
9761         PR target/86348
9762         * config/i386/sse.md (*vec_extractv4si_0_zext_sse4): Use
9763         alternative 0 in preferred_for_speed attribute.
9765 2018-06-28  Paul Koning  <ni1d@arrl.net>
9767         * config/pdp11/pdp11-protos.h (pdp11_shift_length): New function.
9768         * config/pdp11/pdp11.c (pdp11_shift_length): New function.
9769         * config/pdp11/pdp11.h (ADJUST_INSN_LENGTH): Remove.
9770         * config/pdp11/pdp11.md: Correct "length" attribute calculation
9771         for shift insn patterns.
9773 2018-06-28  David Malcolm  <dmalcolm@redhat.com>
9775         * cgraph.c (cgraph_node::get_body): Replace assignments to
9776         "dump_file" with calls to set_dump_file.
9777         * dumpfile.c (alt_dump_file): Make static, and group with...
9778         (alt_flags): ...this definition.
9779         (dumps_are_enabled): New variable.
9780         (refresh_dumps_are_enabled): New function.
9781         (set_dump_file): New function.
9782         (set_alt_dump_file): New function.
9783         (gcc::dump_manager::dump_start): Replace assignments to
9784         "dump_file" and "alt_dump_file" with calls to set_dump_file and
9785         set_alt_dump_file.
9786         (gcc::dump_manager::dump_finish): Likewise.
9787         * dumpfile.h (alt_dump_file): Delete decl.
9788         (dumps_are_enabled): New variable decl.
9789         (set_dump_file): New function decl.
9790         (dump_enabled_p): Rewrite in terms of new "dumps_are_enabled"
9791         global.
9792         * tree-nested.c (lower_nested_functions): Replace assignments to
9793         "dump_file" with calls to set_dump_file.
9795 2018-06-28  Eric Botcazou  <ebotcazou@adacore.com>
9797         * tree-cfg.c (verify_gimple_in_cfg): Call verify_location on the
9798         goto_locus of each outgoing edge of each basic block.
9800 2018-06-28  Richard Biener  <rguenther@suse.de>
9802         * dwarf2out.c (decl_scope_table): Remove.
9803         (push_decl_scope): Likewise.
9804         (pop_decl_scope): Likewise.
9805         (gen_type_die_for_member): Do not call push/pop_decl_scope.
9806         (gen_struct_or_union_type_die): Likewise.
9807         (gen_tagged_type_die): Likewise.
9808         (dwarf2out_init): Do not initialize decl_scope_table.
9809         (dwarf2out_c_finalize): Do not free it.
9811 2018-06-28  Richard Biener  <rguenther@suse.de>
9813         * dwarf2out.c (gen_subprogram_die): Use is_unit_die when
9814         deciding whether to not re-use a DIE.
9816 2018-06-28  Richard Biener  <rguenther@suse.de>
9818         * dwarf2out.c (gen_subprogram_die): Always re-use DIEs with an
9819         DW_AT_abstract_origin attribute.
9821 2018-06-28  Martin Liska  <mliska@suse.cz>
9823         * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
9824         Use newly introduced constants.
9825         * tree-switch-conversion.h (struct jump_table_cluster):
9826         Define max_ratio_for_size and max_ratio_for_speed.
9828 2018-06-28  Martin Liska  <mliska@suse.cz>
9830         * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
9831         Add new checking assert to catch invalid state.
9832         (jump_table_cluster::can_be_handled): Handle single case
9833         clusters.
9834         (jump_table_cluster::is_beneficial): Bail out for such case.
9835         (bit_test_cluster::find_bit_tests):
9836         Add new checking assert to catch invalid state.
9837         (bit_test_cluster::can_be_handled): Handle single case
9838         clusters.
9839         (bit_test_cluster::is_beneficial): Bail out for such case.
9840         (switch_decision_tree::analyze_switch_statement):
9841         Fix comment.
9843 2018-06-28  Martin Liska  <mliska@suse.cz>
9845         * common.opt: Introduce -completion option.
9846         * gcc.c (driver_handle_option): Handle it.
9847         (driver::main): Print completions if completion
9848         is set.
9849         * opt-suggestions.c (option_proposer::get_completions):
9850         New function.
9851         (option_proposer::suggest_completion): Likewise.
9852         (option_proposer::find_param_completions): Likewise.
9853         (verify_autocompletions): Likewise.
9854         (test_completion_valid_options): Likewise.
9855         (test_completion_valid_params): Likewise.
9856         (in_completion_p): Likewise.
9857         (empty_completion_p): Likewise.
9858         (test_completion_partial_match): Likewise.
9859         (test_completion_garbage): Likewise.
9860         (opt_proposer_c_tests): Likewise.
9861         * opt-suggestions.h: Declare new functions.
9862         * opts.c (common_handle_option): Handle OPT__completion_.
9863         * selftest-run-tests.c (selftest::run_tests): Add
9864         opt_proposer_c_tests.
9865         * selftest.c (assert_str_startswith): New.
9866         * selftest.h (assert_str_startswith): Likewise.
9867         (opt_proposer_c_tests): New.
9868         (ASSERT_STR_STARTSWITH): Likewise.
9870 2018-06-28  Martin Liska  <mliska@suse.cz>
9872         * Makefile.in: Add opt-suggestions.o.
9873         * gcc-main.c: Include opt-suggestions.h.
9874         * gcc.c (driver::driver): Likewise.
9875         (driver::~driver): Remove m_option_suggestions.
9876         (driver::build_option_suggestions): Moved to option_proposer.
9877         (driver::suggest_option): Likewise.
9878         (driver::handle_unrecognized_options): Use option_proposer.
9879         * gcc.h (class driver): Add new memver m_option_proposer.
9880         * opt-suggestions.c: New file.
9881         * opt-suggestions.h: New file.
9883 2018-06-28  Martin Liska  <mliska@suse.cz>
9885         * vec.h (class auto_string_vec): New (moved from auto_argvec).
9886         (auto_string_vec::~auto_string_vec): Likewise.
9888 2018-06-28  Eric Botcazou  <ebotcazou@adacore.com>
9890         * tree-inline.h (struct copy_body_data): Move remapping_type_depth and
9891         prevent_decl_creation_for_types fields up and add reset_location field.
9892         * tree-inline.c (remap_gimple_stmt): Force input_location on the new
9893         statement if id->reset_location is true.
9894         (copy_edges_for_bb): Do not set goto_locus on the new edges if
9895         id->reset_location is true.
9896         (copy_phis_for_bb): Force input_location on the arguments if
9897         id->reset_location is true.
9898         (expand_call_inline): Set id->reset_location if DECL_IGNORED_P
9899         is set on the function to be inlined.
9901 2018-06-27  Stephan Bergmann  <sbergman@redhat.com>
9903         * doc/invoke.texi (Debugging Options): Add -gsplit-dwarf.
9905 2018-06-27  Dimitar Dimitrov  <dimitar@dinux.eu>
9907         * lra-eliminations.c (update_reg_eliminate): Mark all spanning hard
9908         registers for Pmode.
9909         * lra-lives.c (check_pseudos_live_through_calls): Mark all spanning
9910         hard registers for the clobbered pseudo.
9912 2018-06-27  Paul Koning  <ni1d@arrl.net>
9914         * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
9915         mutually exclusive options.
9916         * config/pdp11/constraints.md (h): New constraint.
9917         (O): Update definition to match shift code generation.
9918         (D): New constraint.
9919         * config/pdp11/pdp11-modes.def (CCNZ): Define mode.
9920         (CCFP): Remove.
9921         * config/pdp11/pdp11-protos.h (int_no_side_effect_operand): New
9922         function.
9923         (output_jump): Change arguments.
9924         (pdp11_fixed_cc_regs): New function.
9925         (pdp11_cc_mode): Ditto.
9926         (pdp11_expand_shift): Ditto.
9927         (pdp11_assemble_shift): Ditto.
9928         (pdp11_small_shift): Ditto.
9929         (pdp11_branch_cost): Remove.
9930         * config/pdp11/pdp11.c (pdp11_assemble_integer): Remove comments
9931         from output.
9932         (pdp11_register_move_cost): Update for CC registers.
9933         (pdp11_rtx_costs): Add case for LSHIFTRT.
9934         (pdp11_output_jump): Add CCNZ mode conditional branches.
9935         (notice_update_cc_on_set): Remove.
9936         (pdp11_cc_mode): New function.
9937         (simple_memory_operand): Correct pre/post decrement case.
9938         (no_side_effect_operand): New function.
9939         (pdp11_regno_reg_class): Add CC_REGS class.
9940         (pdp11_fixed_cc_regs): New function.
9941         (pdp11_small_shift): New function.
9942         (pdp11_expand_shift): New function to expand shift insns.
9943         (pdp11_assemble_shift): New function to output shifts.
9944         (pdp11_branch_cost): Remove.
9945         (pdp11_modes_tieable_p): Make QI/HI modes tieable.
9946         * config/pdp11/pdp11.h (SIZE_TYPE): Ensure 16-bit type.
9947         (WCHAR_TYPE): Ditto.
9948         (PTRDIFF_TYPE): Ditto.
9949         (ADJUST_INSN_LENGTH): New macro.
9950         (FIXED_REGISTERS): Add CC registers.
9951         (CALL_USED_REGISTERS): Ditto.
9952         (reg_class): Ditto.
9953         (REG_CLASS_NAMES): Ditto.
9954         (REG_CLASS_CONTENTS): Ditto.
9955         (SELECT_CC_MODE): Use new function.
9956         (TARGET_FLAGS_REGNUM): New macro.
9957         (TARGET_FIXED_CONDITION_CODE_REGS): Ditto.
9958         (cc0_reg_rtx): Remove.
9959         (CC_STATUS_MDEP): Remove.
9960         (CC_STATUS_MDEFP_INIT): Remove.
9961         (CC_IN_FPU): Remove.
9962         (NOTICE_UPDATE_CC): Remove.
9963         (REGISTER_NAMES): Add CC registers.
9964         (BRANCH_COST): Change to constant 1.
9965         * config/pdp11/pdp11.md: Rewrite for CCmode condition code
9966         handling.
9967         * config/pdp11/pdp11.opt (mbcopy): Remove.
9968         (mbcopy-builtin): Remove.
9969         (mbranch-cheap): Remove.
9970         (mbranch-expensive): Remove.
9971         * config/pdp11/predicates.md (expand_shift_operand): Update to
9972         match shift code generation.
9973         (ccnz_operator): New predicate.
9974         * doc/invoke.texi (PDP-11 Options): Remove deleted options
9975         -mbcopy, -mbcopy-builtin, -mbranch-cheap, -mbranch-expensive.
9976         Remove non-existent option -mabshi, -mno-abshi.  Document mutually
9977         exclusive options.
9978         * doc/md.texi (PDP-11): Document new D and h constraints.  Update
9979         description of O constraint.
9981 2018-06-27  Jeff Law  <law@redhat.com>
9982             Austin Law  <austinklaw@gmail.com>
9984         * config/v850/v850.md (addsi3_set_flags): New pattern.
9985         (subsi3_set_flags, negsi2_set_flags, andsi3_set_flags): Likewise.
9986         (iorsi3_set_flags, xorsi3_set_flags, one_cmplsi2_set_flags): Likewise.
9987         (zero_extendhisi2_v850_set_flags): Likewise.
9988         (zero_extendqisi2_v850_set_flags): Likewise.
9989         (ashlsi3_set_flags, ashlsi3_v850e2_set_flags): Likewise.
9990         (lshrsi3_set_flags, lshrsi3_v850e2_set_flags): Likewise.
9991         (ashrsi3_set_flags, ashrsi3_v850e2_set_flags): Likewise.
9993         * config/v850/v850-protos.h (notice_update_cc): Remove.
9994         * config/v850/v850.c (v850_compare_op0, v850_compare_op1): Remove.
9995         (v850_print_operand): Handle 'D' and "d".
9996         (v850_select_cc_mode): Remove ATTRIBUTE_UNUSED for last argument.
9997         Add handling of arithmetic/logical operations compared against zero.
9998         (v850_gen_float_compare): Remove ATTRIBUTE_UNUSED for last argument.
9999         Do not look at v850_compare_op, instead get mode from last argument.
10000         (v850_gen_compare): Remove
10001         (increment_stack): Use addsi3_clobber_flags to avoid splitting failure
10002         after reload for prologue insns.
10003         (expand_prologue): Account for CLOBBER of CC_REGNUM in various
10004         patterns.
10005         (construct_save_jarl): Likewise.
10006         (TARGET_FLAGS_REGNUM): Define.
10007         * config/v850/v850.h (v850_compare_op0, v850_compare_op1): Remove.
10008         (NOTICE_UPDATE_CC): Remove.
10009         * config/v850/v850.md (v850_tst1): Use (reg:CCZ CC_REGNUM) rather
10010         than cc0.  Conditionalize on reload_completed.
10011         (cmpsi_insn, setfcc_insn): Likewise.
10012         (tst1 splitter): Turn into define_and_split which sets the flags
10013         after reload.
10014         (cstoresi4, cbranchsf4, cbranchdf4, cbranchsi4_insn): Likewise.
10015         (cbranchsi4, branch_normal, branch_invert): Do not expose cc0 here.
10016         (cstoresf4, cstoredf4): Clobber the flags.
10017         (cmpsi, cmpsf, cmpdf): Remove expanders.
10018         (setf_insn): Remove pattern.
10019         (addsi3): Turn into define_and_split which clobbers the flags after
10020         reload and a suitable pattern (addsi3_clobber_flags) for use after
10021         reload.
10022         (subsi3, negsi2, andsi3, iorsi3, xorsi3, one_cmplsi2) Likewise.
10023         (ashlsi3, ashlsi3_v850e2, lshrsi3, lsh4si3_v850e2): Likewise.
10024         (ashrsi3, ashrsi3_v850e2): Likewise.
10025         (bins): Clobber the flags.
10026         (movsicc_normal_cc, movsicc_normal, movsicc_tst1): Likewise.
10027         (movsicc_tst1_revesed, sasf, swap and rotate patterns): Likewise.
10028         (fix_loop_counter, call_internal_short, call_internal_long): Likewise.
10029         (call_value_internal_short, call_value_internal_long): Likewise.
10030         (callt_save_interrupt, callt_return_interrupt): Likewise.
10031         (save_interrupt, return_interrupt): Likewise.
10032         (callt_save_all_interrupt, save_all_interrupt): Likewise.
10033         (_save_all_interrupt, callt_restore_all_interrupt): Likewise.
10034         (restore_all_interrupt, _restore_all_interrupt): Likewise.
10035         (All FP comparisons): Only allow after reload has completed.
10036         (trfsr): Likewise.
10037         (divh, divhu): Tweak output template.
10038         (branch_z_normal, branch_z_invert): Remove
10039         (branch_nz_normal, branch_nz_invert): Likewise.
10040         (extendhisi_insn, extendqisi_insn): Do not clobber flags.
10042         * config/v850/v850-modes.def (CCZ, CCNZ): Add new modes.
10043         * config/v850/v850.c (notice_update_cc): Remove.
10044         * config/v850/v850.h  (CC_OVERFLOW_UNUSABLE): Remove
10045         (CC_NO_CARRY): Likewise.
10046         (NOTICE_UPDATE_CC): Define to nothing.
10047         * config/v850/v850.md: Remove block comment on cc0 handling
10048         Remove "cc" attribute from all patterns.  Remove cc_status handling
10049         from all patterns.  Minor formatting fixes.
10051 2018-06-27  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
10053         * config/aarch64/aarch64-cores.def (cortex-a76): New entry.
10054         (cortex-a76.cortex-a55): Likewise.
10055         * config/aarch64/aarch64-tune.md: Regenerate.
10056         * doc/invoke.texi (AArch64 Options): Document cortex-a76 and
10057         cortex-a76.cortex-a55.
10059 2018-06-27  Jeff Law  <law@redhat.com>
10061         * config/v850/t-v850 (MULTILIB_OPTIONS): Remove 8byte-align.
10062         (MULTILIB_DIRNAMES): Similarly.
10064 2018-06-27  Eric Botcazou  <ebotcazou@adacore.com>
10066         * gimple.h (gimple_return_retbnd): Delete.
10067         (gimple_return_set_retbnd): Likewise.
10068         * cgraphunit.c (cgraph_node::expand_thunk): Remove call to
10069         gimple_return_set_retbnd.
10070         * gimple-pretty-print.c (dump_gimple_return): Remove call to
10071         gimple_return_retbnd and adjust.
10072         * tree-inline.h (struct copy_body_data): Remove retbnd field.
10073         * tree-inline.c (remap_gimple_stmt): Remove handling of retbnd.
10074         Explicitly return NULL in a couple more cases.  Move assertion
10075         on debug statements and remove unreachable code.
10076         (reset_debug_binding): Do not test id->retbnd.
10077         (expand_call_inline): Do not set it.
10079 2018-06-27  Rasmus Villemoes  <rv@rasmusvillemoes.dk>
10081         * configure.ac: Add --disable-gcov option.
10082         * configure: Regenerate.
10083         * Makefile.in: Honour @enable_gcov@.
10084         * doc/install.texi: Document --disable-gcov.
10086 2018-06-27  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
10088         * config/arm/arm-cpus.in (cortex-a76): New entry.
10089         (cortex-a76.cortex-a55): Likewise.
10090         * config/arm/arm-tables.opt: Regenerate.
10091         * config/arm/arm-tune.md: Likewise.
10092         * config/arm/driver-arm.c (arm_cpu_table): Add Cortex-A76 entry.
10093         * doc/invoke.texi (ARM Options): Document cortex-a76 and
10094         cortex-a76.cortex-a55.
10096 2018-06-27  Tamar Christina  <tamar.christina@arm.com>
10098         PR target/85769
10099         * config/aarch64/aarch64.md (*movhf_aarch64): Add dup v0.4h pattern.
10101 2018-06-27  Siddhesh Poyarekar  <siddhesh@sourceware.org>
10103         * config/aarch64/aarch64.h (CALL_USE_REGISTERS): Fix obsolete
10104         comment.
10105         (EPILOGUE_USES): Likewise.
10107 2018-06-26  Eric Botcazou  <ebotcazou@adacore.com>
10109         * tree-inline.c (remap_location): New function extracted from...
10110         (copy_edges_for_bb): Add ID parameter.  Remap goto_locus.
10111         (copy_phis_for_bb): ...here.  Call remap_location.
10112         (copy_cfg_body): Adjust call to copy_edges_for_bb.
10114 2018-06-26  Aaron Sawdey  <acsawdey@linux.ibm.com>
10116         * config/rs6000/rs6000-string.c (expand_block_clear): Don't use
10117         unaligned vsx for 16B memset.
10119 2018-06-26  Segher Boessenkool  <segher@kernel.crashing.org>
10121         PR target/86285
10122         * config/rs6000/rs6000.c (rs6000_init_builtins): Do not set
10123         ieee128_float_type_node to long_double_type_node unless
10124         TARGET_LONG_DOUBLE_128 is set.
10126 2018-06-26  David Malcolm  <dmalcolm@redhat.com>
10128         * cfgloop.c (get_loop_location): Convert return type from
10129         location_t to dump_user_location_t, replacing INSN_LOCATION lookups
10130         by implicit construction from rtx_insn *, and using
10131         dump_user_location_t::from_function_decl for the fallback case.
10132         * cfgloop.h (get_loop_location): Convert return type from
10133         location_t to dump_user_location_t.
10134         * cgraphunit.c (walk_polymorphic_call_targets): Update call to
10135         dump_printf_loc to pass in a dump_location_t rather than a
10136         location_t, via the gimple stmt.
10137         * coverage.c (get_coverage_counts): Update calls to
10138         dump_printf_loc to pass in dump_location_t rather than a
10139         location_t.
10140         * doc/optinfo.texi (Dump types): Convert example of
10141         dump_printf_loc from taking "locus" to taking "insn".  Update
10142         description of the "_loc" calls to cover dump_location_t.
10143         * dumpfile.c: Include "backend.h", "gimple.h", "rtl.h", and
10144         "selftest.h".
10145         (dump_user_location_t::dump_user_location_t): New constructors,
10146         from gimple *stmt and rtx_insn *.
10147         (dump_user_location_t::from_function_decl): New function.
10148         (dump_loc): Make static.
10149         (dump_gimple_stmt_loc): Convert param "loc" from location_t to
10150         const dump_location_t &.
10151         (dump_generic_expr_loc): Delete.
10152         (dump_printf_loc): Convert param "loc" from location_t to
10153         const dump_location_t &.
10154         (selftest::test_impl_location): New function.
10155         (selftest::dumpfile_c_tests): New function.
10156         * dumpfile.h: Include "profile-count.h".
10157         (class dump_user_location_t): New class.
10158         (struct dump_impl_location_t): New struct.
10159         (class dump_location_t): New class.
10160         (dump_printf_loc): Convert 2nd param from source_location to
10161         const dump_location_t &.
10162         (dump_generic_expr_loc): Delete.
10163         (dump_gimple_stmt_loc): Convert 2nd param from source_location to
10164         const dump_location_t &.
10165         * gimple-fold.c (fold_gimple_assign): Update call to
10166         dump_printf_loc to pass in a dump_location_t rather than a
10167         location_t, via the gimple stmt.
10168         (gimple_fold_call): Likewise.
10169         * gimple-loop-interchange.cc
10170         (loop_cand::analyze_iloop_reduction_var): Update for change to
10171         check_reduction_path.
10172         (tree_loop_interchange::interchange): Update for change to
10173         find_loop_location.
10174         * graphite-isl-ast-to-gimple.c (scop_to_isl_ast): Update for
10175         change in return-type of find_loop_location.
10176         (graphite_regenerate_ast_isl): Likewise.
10177         * graphite-optimize-isl.c (optimize_isl): Likewise.
10178         * graphite.c (graphite_transform_loops): Likewise.
10179         * ipa-devirt.c (ipa_devirt): Update call to dump_printf_loc to
10180         pass in a dump_location_t rather than a location_t, via the
10181         gimple stmt.
10182         * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
10183         * ipa.c (walk_polymorphic_call_targets): Likewise.
10184         * loop-unroll.c (report_unroll): Convert "locus" param from
10185         location_t to dump_location_t.
10186         (decide_unrolling): Update for change to get_loop_location's
10187         return type.
10188         * omp-grid.c (struct grid_prop): Convert field "target_loc" from
10189         location_t to dump_user_location_t.
10190         (grid_find_single_omp_among_assignments_1): Updates calls to
10191         dump_printf_loc to pass in a dump_location_t rather than a
10192         location_t, via the gimple stmt.
10193         (grid_parallel_clauses_gridifiable): Convert "tloc" from
10194         location_t to dump_location_t.  Updates calls to dump_printf_loc
10195         to pass in a dump_location_t rather than a location_t, via the
10196         gimple stmt.
10197         (grid_inner_loop_gridifiable_p): Likewise.
10198         (grid_dist_follows_simple_pattern): Likewise.
10199         (grid_gfor_follows_tiling_pattern): Likewise.
10200         (grid_target_follows_gridifiable_pattern): Likewise.
10201         (grid_attempt_target_gridification): Convert initialization
10202         of local "grid" from memset to zero-initialization; FIXME: does
10203         this require C++11?  Update call to dump_printf_loc to pass in a
10204         optinfo_location rather than a location_t, via the gimple stmt.
10205         * profile.c (read_profile_edge_counts): Updates call to
10206         dump_printf_loc to pass in a dump_location_t rather than a
10207         location_t
10208         (compute_branch_probabilities): Likewise.
10209         * selftest-run-tests.c (selftest::run_tests): Call
10210         dumpfile_c_tests.
10211         * selftest.h (dumpfile_c_tests): New decl.
10212         * tree-loop-distribution.c (pass_loop_distribution::execute):
10213         Update for change in return type of find_loop_location.
10214         * tree-parloops.c (parallelize_loops): Likewise.
10215         * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Convert
10216         "locus" from location_t to dump_user_location_t.
10217         (canonicalize_loop_induction_variables): Likewise.
10218         * tree-ssa-loop-ivopts.c (tree_ssa_iv_optimize_loop): Update
10219         for change in return type of find_loop_location.
10220         * tree-ssa-loop-niter.c (number_of_iterations_exit): Update call
10221         to dump_printf_loc to pass in a dump_location_t rather than a
10222         location_t, via the stmt.
10223         * tree-ssa-sccvn.c (eliminate_dom_walker::before_dom_children):
10224         Likewise.
10225         * tree-vect-loop-manip.c (find_loop_location): Convert return
10226         type from source_location to dump_user_location_t.
10227         (vect_do_peeling): Update for above change.
10228         (vect_loop_versioning): Update for change in type of
10229         vect_location.
10230         * tree-vect-loop.c (check_reduction_path): Convert "loc" param
10231         from location_t to dump_user_location_t.
10232         (vect_estimate_min_profitable_iters): Update for change in type
10233         of vect_location.
10234         * tree-vect-slp.c (vect_print_slp_tree): Convert param "loc" from
10235         location_t to dump_location_t.
10236         (vect_slp_bb): Update for change in type of vect_location.
10237         * tree-vectorizer.c (vect_location): Convert from source_location
10238         to dump_user_location_t.
10239         (try_vectorize_loop_1): Update for change in vect_location's type.
10240         (vectorize_loops): Likewise.
10241         (increase_alignment): Likewise.
10242         * tree-vectorizer.h (vect_location): Convert from source_location
10243         to dump_user_location_t.
10244         (find_loop_location): Convert return type from source_location to
10245         dump_user_location_t.
10246         (check_reduction_path): Convert 1st param from location_t to
10247         dump_user_location_t.
10248         * value-prof.c (check_counter): Update call to dump_printf_loc to
10249         pass in a dump_user_location_t rather than a location_t; update
10250         call to error_at for change in type of "locus".
10251         (check_ic_target): Update call to dump_printf_loc to
10252         pass in a dump_user_location_t rather than a location_t, via the
10253         call_stmt.
10255 2018-06-26  Robin Dapp  <rdapp@linux.vnet.ibm.com>
10257         * config/s390/s390.h (enum processor_flags): Do not use
10258         default tune parameter when -march was specified.
10260 2018-06-26  Jakub Jelinek  <jakub@redhat.com>
10262         PR target/86314
10263         * config/i386/i386.md (setcc + movzbl to xor + setcc peephole2s):
10264         Check reg_overlap_mentioned_p in addition to reg_set_p with the same
10265         operands.
10267 2018-06-26  Richard Biener  <rguenther@suse.de>
10269         PR tree-optimization/86287
10270         PR bootstrap/86316
10271         * tree-vect-loop.c (vect_transform_loop_stmt): Fix read-after-free.
10272         (vect_analyze_loop): Initialize n_stmts.
10274 2018-06-26  Richard Biener  <rguenther@suse.de>
10276         PR middle-end/86271
10277         * fold-const.c (fold_convertible_p): Pointer extension
10278         isn't valid.
10280 2018-06-26  Alexandre Oliva <aoliva@redhat.com>
10282         PR debug/86064
10283         * dwarf2out.c (loc_list_has_views): Adjust comments.
10284         (dw_loc_list): Split single cross-partition range with
10285         nonzero locview.
10287 2018-06-25  Jeff Law  <law@redhat.com>
10289         * common/config/v850/v850-common.c (TARGET_DEFAULT_TARGET_FLAGS): Turn
10290         on -mbig-switch by default.
10292         * config/v850/predicates.md (const_float_1_operand): Fix match_code
10293         test.
10294         (const_float_0_operand): Remove unused predicate.
10295         * config/v850/v850.md (define_constants): Remove UNSPEC_LOOP.
10296         (define_c_enum unspec): Add LOOP, RCP and RSQRT constants.
10297         (recipsf2): New expander.  Original pattern now called
10298         (recipsf2_insn).
10299         (recipdf2, recipdf2_insn): Similarly.
10300         (rsqrtsf2, rsqrtsf2_insn): Similarly
10301         (rsqrtdf2, rsqrtdf2_insn): Similarly
10303 2018-06-26  Gerald Pfeifer  <gerald@pfeifer.com>
10305         * ginclude/stddef.h: Remove an obsolete comment on FreeBSD 5.
10306         Simplify logic for FreeBSD (twice).
10308 2018-06-25  Martin Sebor  <msebor@redhat.com>
10310         PR tree-optimization/86204
10311         * tree-ssa-strlen.c (handle_builtin_strlen): Avoid storing
10312         a strnlen result if it's less than the length of the string.
10314 2018-06-25  Martin Sebor  <msebor@redhat.com>
10316         PR tree-optimization/85700
10317         * gimple-fold.c (gimple_fold_builtin_strncat): Adjust comment.
10318         * tree-ssa-strlen.c (is_strlen_related_p): Handle integer subtraction.
10319         (maybe_diag_stxncpy_trunc): Distinguish strncat from strncpy.
10321 2018-06-25  Martin Sebor  <msebor@redhat.com>
10323         * doc/extend.texi (Zero-length arrays): Update and clarify.
10325 2018-06-25  Michael Meissner  <meissner@linux.ibm.com>
10327         * config.gcc (powerpc64le*): Revert January 16th, 2018 patch that
10328         added IEEE/IBM long double multilib support on PowerPC little
10329         endian Linux systems.
10330         * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Likewise.
10331         (MULTILIB_DEFAULTS): Likewise.
10332         * config/rs6000/rs6000.c (rs6000_option_override_internal):
10333         Likewise.
10334         * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Likewise.
10335         * config/rs6000/t-ldouble-linux64le-ibm: Delete, no longer used.
10336         * config/rs6000/t-ldouble-linux64le-ieee: Delete, no longer used.
10338 2018-06-25  Alexander Monakov  <amonakov@ispras.ru>
10340         PR middle-end/86311
10341         * sort.cc (REORDER_23): Avoid memcpy with same destination and source.
10342         (REORDER_45): Likewise.
10344 2018-06-25  Jeff Law  <law@redhat.com>
10346         * config/v850/v850.md (divmodhi4): Make sure to sign extend the
10347         dividend to 32 bits.  Adjust length.
10348         (udivmodhi4): Cleanup output template.  Fix length.
10350 2018-06-25  Carl Love  <cel@us.ibm.com>
10352         * config/rs6000/vsx.md: Change word selector to prefered location.
10354 2018-06-25  Richard Biener  <rguenther@suse.de>
10356         PR tree-optimization/86304
10357         * tree-vectorizer.c (vectorize_loops): Walk over new possibly
10358         epilogue-if-converted loops as well.
10360 2018-06-25  Jan Hubicka  <hubicka@ucw.cz>
10362         * lto-section-out.c (lto_begin_section): Do not print section
10363         name for noaddr and unnumbered dumps.
10365 2018-06-25  Richard Biener  <rguenther@suse.de>
10367         * tree-vectorizer.h (struct vec_info_shared): New structure
10368         with parts split out from struct vec_info and loop_nest from
10369         struct _loop_vec_info.
10370         (struct vec_info): Adjust accordingly.
10371         (struct _loop_vec_info): Likewise.
10372         (LOOP_VINFO_LOOP_NEST): Adjust.
10373         (LOOP_VINFO_DATAREFS): Likewise.
10374         (LOOP_VINFO_DDRS): Likewise.
10375         (struct _bb_vec_info): Likewise.
10376         (BB_VINFO_DATAREFS): Likewise.
10377         (BB_VINFO_DDRS): Likewise.
10378         (struct _stmt_vec_info): Add dr_aux member.
10379         (DR_VECT_AUX): Adjust to refer to member of DR_STMTs vinfo.
10380         (DR_MISALIGNMENT_UNINITIALIZED): New.
10381         (set_dr_misalignment): Adjust.
10382         (dr_misalignment): Assert misalign isn't DR_MISALIGNMENT_UNINITIALIZED.
10383         (vect_analyze_loop): Adjust prototype.
10384         (vect_analyze_loop_form): Likewise.
10385         * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
10386         Compute dependences lazily.
10387         (vect_record_base_alignments): Use shared datarefs/ddrs.
10388         (vect_verify_datarefs_alignment): Likewise.
10389         (vect_analyze_data_refs_alignment): Likewise.
10390         (vect_analyze_data_ref_accesses): Likewise.
10391         (vect_analyze_data_refs): Likewise.
10392         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Add
10393         constructor parameter for shared part.
10394         (vect_analyze_loop_form): Pass in shared part and adjust.
10395         (vect_analyze_loop_2): Pass in storage for the number of
10396         stmts.  Move loop nest finding to the caller.  Compute
10397         datarefs lazily.
10398         (vect_analyze_loop): Pass in shared part.
10399         (vect_transform_loop): Verify shared datarefs are unchanged.
10400         * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Add
10401         constructor parameter for shared part.
10402         (vect_slp_analyze_bb_1): Pass in shared part and adjust.
10403         (vect_slp_bb): Verify shared datarefs are unchanged before
10404         transform.
10405         * tree-vect-stmts.c (ensure_base_align): Adjust for DR_AUX
10406         change.
10407         (new_stmt_vec_info): Initialize DR_AUX misalignment to
10408         DR_MISALIGNMENT_UNINITIALIZED.
10409         * tree-vectorizer.c (vec_info::vec_info): Add constructor
10410         parameter for shared part.
10411         (vec_info::~vec_info): Adjust.
10412         (vec_info_shared::vec_info_shared): New.
10413         (vec_info_shared::~vec_info_shared): Likewise.
10414         (vec_info_shared::save_datarefs): Likewise.
10415         (vec_info_shared::check_datarefs): Likewise.
10416         (try_vectorize_loop_1): Construct shared part live for analyses
10417         of a single loop for multiple vector sizes.
10418         * tree-parloops.c (gather_scalar_reductions): Adjust.
10420 2018-06-25  Richard Biener  <rguenther@suse.de>
10422         * tree-vect-data-refs.c (vect_find_stmt_data_reference): Modify
10423         DR for SIMD lane accesses here and mark DR with (void *)-1 aux.
10424         (vect_analyze_data_refs): Remove similar code from here and
10425         simplify accordingly.
10427 2018-06-25  Richard Biener  <rguenther@suse.de>
10429         * tree-vect-data-refs.c (vect_check_gather_scatter): Fail
10430         for reverse storage order accesses rather than asserting
10431         they cannot happen here.
10433 2018-06-25  Tom de Vries  <tdevries@suse.de>
10435         PR debug/86257
10436         * config/i386/i386.md (define_insn "*tls_global_dynamic_64_<mode>"):
10437         Use data16 instead of .byte for insn prefix.
10439 2018-06-25  Andreas Krebbel  <krebbel@linux.ibm.com>
10441         PR C++/86082
10442         * parser.c (make_char_string_pack): Pass this literal chars
10443         through cpp_interpret_string.
10444         (cp_parser_userdef_numeric_literal): Check the result of
10445         make_char_string_pack.
10447 2018-06-24  Maya Rashish  <coypu@sdf.org>
10449         * ginclude/stddef.h: Simplify conditions around avoiding
10450         re-definition of __size_t.
10452 2018-06-22  Jan Hubicka  <hubicka@ucw.cz>
10454         * lto-streamer-out.c (tree_is_indexable): Make LABEL_DECL nonindexable
10455         unless it is forced or nonlocal; assert that we stream no IMPORTED_DECL.
10457 2018-06-22  Maya Rashish  <coypu@sdf.org>
10459         * doc/invoke.texi (mno-fancy-math-387): Update for changes
10460         made to OpenBSD and NetBSD through the years.
10462 2018-06-22  Kelvin Nilsen  <kelvin@gcc.gnu.org>
10464         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
10465         behavior of vec_pack (vector double, vector double) to match
10466         behavior of vec_float2 (vector double, vector double).
10468 2018-06-22  Olivier Hainque  <hainque@adacore.com>
10470         * gimplify.c (gimplify_function_tree): Prevent creation
10471         of a trampoline for the address of the current function
10472         passed to entry/exit instrumentation hooks.
10474 2018-06-22  Aaron Sawdey  <acsawdey@linux.ibm.com>
10476         PR target/86222
10477         * config/rs6000/rs6000-string.c (expand_strn_compare): Handle -m32
10478         correctly.
10480 2018-06-22  Martin Liska  <mliska@suse.cz>
10482         PR tree-optimization/86263
10483         * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
10484         Bail out if is_enabled is false.
10485         * tree-switch-conversion.h (jump_table_cluster::is_enabled):
10486         New declaration.
10487         (jump_table_cluster::is_enabled): New function.
10489 2018-06-22  Jan Hubicka  <hubicka@ucw.cz>
10491         * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
10492         BINFO_BASE_ACCESSES and BINFO_VPTR_FIELD.
10493         * tree-streamer-in.c (streamer_read_tree_bitfields): Likewise.
10494         (lto_input_ts_binfo_tree_pointers): Likewise.
10495         * tree-streamer-out.c (streamer_write_tree_bitfields,
10496         write_ts_binfo_tree_pointers): Likewise.
10497         * tree.c (free_lang_data_in_binfo): Clear BINFO_VPTR_FIELD.
10499 2018-06-22  Jan Hubicka  <hubicka@ucw.cz>
10501         * tree.c (free_lang_data_in_type): Free all TYPE_VFIELDs.
10503 2018-06-22  Martin Liska  <mliska@suse.cz>
10505         * symbol-summary.h (get): Make it pure and inline move
10506         functionality from ::get function.
10507         (get): Remove and inline into ::get and ::get_create.
10508         (get_create): Move code from ::get function.
10510 2018-06-22  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
10512         PR target/85994
10513         * config/i386/sol2.h (CPP_SPEC): Don't pass -P for
10514         -x assembler-with-cpp.
10516 2018-06-22  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
10518         * config/sol2.h (TARGET_OS_CPP_BUILTINS): Define
10519         _FILE_OFFSET_BITS=64 for C++.
10521 2018-06-21  Michael Meissner  <meissner@linux.ibm.com>
10523         * config/rs6000/rs6000.md (extendtfif2): Add missing 128-bit
10524         conversion insn that shows up when pr85657-3.c is compiled using
10525         IEEE 128-bit long double.
10526         (neg<mode>2_internal): Use the correct mode to check whether the
10527         mode is IBM extended.
10528         * config/rs6000/rs6000.c (init_float128_ieee): Prevent complex
10529         multiply and divide external functions from being created more
10530         than once.
10532 2018-06-21  Eric Botcazou  <ebotcazou@adacore.com>
10534         * cfgrtl.c (fixup_reorder_chain): Do not emit NOPs in DECL_IGNORED_P
10535         functions.
10536         (rtl_merge_blocks): Likewise.  Do not emit a NOP if the location of
10537         the edge can be forwarded.
10538         (cfg_layout_merge_blocks): Likewise.
10540 2018-06-21  Eric Botcazou  <ebotcazou@adacore.com>
10542         * except.c (finish_eh_generation): Commit edge insertions only after
10543         the EH edges have been redirected from post-landing to landing pads.
10545 2018-06-21  Eric Botcazou  <ebotcazou@adacore.com>
10547         * tree-nested.c (get_frame_type): Use create_tmp_var_raw instead of
10548         create_tmp_var_for to create the FRAME decl.
10549         (finalize_nesting_tree_1): Do not unchain the FRAME decl.
10551 2018-06-21  Eric Botcazou  <ebotcazou@adacore.com>
10553         * tree-inline.c (copy_edges_for_bb): Minor tweak.
10554         (maybe_move_debug_stmts_to_successors): Also reset the locus of the
10555         debug statement when resetting its value.
10556         (expand_call_inline): Copy the locus of the call onto the assignment
10557         of the return value, if any.  Use local variable in more cases.
10559 2018-06-21  Martin Liska  <mliska@suse.cz>
10561         * ipa-pure-const.c (propagate_nothrow): Use
10562         funct_state_summaries->get.
10563         (dump_malloc_lattice): Likewise.
10564         (propagate_malloc): Likewise.
10566 2018-06-21  Richard Biener  <rguenther@suse.de>
10568         * lto-streamer-out.c (DFS::DFS_write_tree_body): Update outdated
10569         comment.  Follow BLOCK_ABSTRACT_ORIGIN unconditionally.
10570         * tree-streamer-in.c (lto_input_ts_block_tree_pointers): Update
10571         comment.
10572         * tree-streamer-out.c (write_ts_block_tree_pointers): Stream
10573         BLOCK_ABSTRACT_ORIGIN unconditionally.
10575 2018-06-21  David Malcolm  <dmalcolm@redhat.com>
10577         * ipa-cp.c (ipcp_driver): Set edge_clone_summaries to NULL after
10578         deleting it.
10579         * ipa-reference.c (ipa_reference_c_finalize): Delete
10580         ipa_ref_opt_sum_summaries and set it to NULL.
10582 2018-06-21  Tom de Vries  <tdevries@suse.de>
10584         PR tree-optimization/85859
10585         * tree-ssa-tail-merge.c (stmt_local_def): Copy gimple_is_call
10586         test with comment from bb_no_side_effects_p.
10588 2018-06-21  Richard Biener  <rguenther@suse.de>
10590         PR tree-optimization/86232
10591         * tree-ssa-loop-niter.c (number_of_iterations_popcount): Adjust
10592         max for constant niter.
10594 2018-06-21  Andre Vieira  <andre.simoesdiasvieira@arm.com>
10596         * config/aarch64/aarch64-simd.md
10597         (*aarch64_crypto_aes<aes_op>v16qi_xor_combine): New.
10599 2018-06-21  Andre Vieira  <andre.simoesdiasvieira@arm.com>
10601         * config/aarch64/aarch64-simd.md (aarch64_crypto_aes<aes_op>v16qi):
10602         Make opernads of the unspec commutative.
10604 2018-06-21  Richard Biener  <rguenther@suse.de>
10606         * tree-data-ref.c (dr_step_indicator): Handle NULL DR_STEP.
10607         * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr):
10608         Avoid calling vect_mark_for_runtime_alias_test with gathers or scatters.
10609         (vect_analyze_data_ref_dependence): Re-order checks to deal with
10610         NULL DR_STEP.
10611         (vect_record_base_alignments): Do not record base alignment
10612         for gathers or scatters.
10613         (vect_compute_data_ref_alignment): Drop return value that is always
10614         true.  Bail out early for gathers or scatters.
10615         (vect_enhance_data_refs_alignment): Bail out early for gathers
10616         or scatters.
10617         (vect_find_same_alignment_drs): Likewise.
10618         (vect_analyze_data_refs_alignment): Remove dead code.
10619         (vect_slp_analyze_and_verify_node_alignment): Likewise.
10620         (vect_analyze_data_refs): For possible gathers or scatters do
10621         not create an alternate DR, just check their possible validity
10622         and mark them.  Adjust DECL_NONALIASED handling to not rely
10623         on DR_BASE_ADDRESS.
10624         * tree-vect-loop-manip.c (vect_update_inits_of_drs): Do not
10625         update inits of gathers or scatters.
10626         * tree-vect-patterns.c (vect_recog_mask_conversion_pattern):
10627         Also copy gather/scatter flag to pattern vinfo.
10629 2018-06-20  Kelvin Nilsen  <kelvin@gcc.gnu.org>
10631         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
10632         behavior of vec_packsu (vector unsigned long long, vector unsigned
10633         long long) to match behavior of vec_packs with same signature.
10635 2018-06-20  Chung-Lin Tang <cltang@codesourcery.com>
10636             Thomas Schwinge <thomas@codesourcery.com>
10637             Cesar Philippidis  <cesar@codesourcery.com>
10639         * gimplify.c (gimplify_scan_omp_clauses): Add support for
10640         OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
10641         (gimplify_adjust_omp_clauses): Likewise.
10642         (gimplify_oacc_declare_1): Add support for GOMP_MAP_RELEASE, remove
10643         support for GOMP_MAP_FORCE_{ALLOC,TO,FROM,TOFROM}.
10644         (gimplify_omp_target_update): Update handling of acc update and
10645         enter/exit data.
10646         * omp-low.c (install_var_field): Remove unused parameter
10647         base_pointers_restrict.
10648         (scan_sharing_clauses): Remove base_pointers_restrict parameter.
10649         Update call to install_var_field. Handle OMP_CLAUSE_{IF_PRESENT,
10650         FINALIZE}
10651         (omp_target_base_pointers_restrict_p): Delete.
10652         (scan_omp_target): Update call to scan_sharing_clauses.
10653         * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_{IF_PRESENT,
10654         FINALIZE}.
10655         * tree-nested.c (convert_nonlocal_omp_clauses): Handle
10656         OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
10657         (convert_local_omp_clauses): Likewise.
10658         * tree-pretty-print.c (dump_omp_clause): Likewise.
10659         * tree.c (omp_clause_num_ops): Add entries for  OMP_CLAUSE_{IF_PRESENT,
10660         FINALIZE}.
10661         (omp_clause_code_name): Likewise.
10663 2018-06-20  Jakub Jelinek  <jakub@redhat.com>
10665         PR debug/86194
10666         * var-tracking.c (use_narrower_mode_test): Check if shift amount can
10667         be narrowed.
10669         PR tree-optimization/86231
10670         * tree-vrp.c (union_ranges): For (  [  )  ] or (   )[   ] range and
10671         anti-range don't overwrite *vr0min before using it to compute *vr0max.
10673 2018-06-20  Tom de Vries  <tdevries@suse.de>
10675         PR tree-optimization/86097
10676         * tree-ssa-loop-manip.c (canonicalize_loop_ivs): Also convert *nit to
10677         iv type if signedness of iv type is not the same as that of *nit.
10679 2018-06-20  Jakub Jelinek  <jakub@redhat.com>
10681         * cfgrtl.c (rtl_verify_edges): Formatting fix.  If bb->preds has any
10682         EDGE_EH edges, verify they are all EDGE_EH.
10684 2018-06-20  Maya Rashish  <coypu@sdf.org>
10686         * ginclude/stddef.h: Limit #include <machine/ansi.h> to NetBSD.
10688 2018-06-20  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
10690         * config/aarch64/aarch64-tuning-flags.def (no_ldp_stp_qregs): New.
10691         * config/aarch64/aarch64.c (xgene1_tunings): Add
10692         AARCH64_EXTRA_TUNE_NO_LDP_STP_QREGS to tune_flags.
10693         (aarch64_mode_valid_for_sched_fusion_p):
10694         Allow 16-byte modes.
10695         (aarch64_classify_address): Allow 16-byte modes for load_store_pair_p.
10696         * config/aarch64/aarch64-ldpstp.md: Add peepholes for LDP STP of
10697         128-bit modes.
10698         * config/aarch64/aarch64-simd.md (load_pair<VQ:mode><VQ2:mode>):
10699         New pattern.
10700         (vec_store_pair<VQ:mode><VQ2:mode>): Likewise.
10701         * config/aarch64/iterators.md (VQ2): New mode iterator.
10703 2018-06-20  Martin Liska  <mliska@suse.cz>
10705         * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
10706         Change default ratio from 10 to 8.
10708 2018-06-20  Martin Liska  <mliska@suse.cz>
10710         * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
10711         New.
10712         (bit_test_cluster::find_bit_tests): Likewise.
10713         (switch_decision_tree::analyze_switch_statement): Find clusters.
10714         * tree-switch-conversion.h (struct jump_table_cluster): Document
10715         hierarchy.
10717 2018-06-20  Martin Liska  <mliska@suse.cz>
10719         * tree-switch-conversion.c (switch_conversion::collect):
10720         Record m_uniq property.
10721         (switch_conversion::expand): Bail out for special conditions.
10722         (group_cluster::~group_cluster): New.
10723         (group_cluster::group_cluster): Likewise.
10724         (group_cluster::dump): Likewise.
10725         (jump_table_cluster::emit): New.
10726         (switch_decision_tree::fix_phi_operands_for_edges): New.
10727         (struct case_node): Remove struct.
10728         (jump_table_cluster::can_be_handled): New.
10729         (case_values_threshold): Moved to header.
10730         (reset_out_edges_aux): Likewise.
10731         (jump_table_cluster::is_beneficial): New.
10732         (bit_test_cluster::can_be_handled): Likewise.
10733         (add_case_node): Remove.
10734         (bit_test_cluster::is_beneficial): New.
10735         (case_bit_test::cmp): New.
10736         (bit_test_cluster::emit): New.
10737         (expand_switch_as_decision_tree_p): Remove.
10738         (bit_test_cluster::hoist_edge_and_branch_if_true): New.
10739         (fix_phi_operands_for_edge): Likewise.
10740         (switch_decision_tree::analyze_switch_statement): New.
10741         (compute_cases_per_edge): Move ...
10742         (switch_decision_tree::compute_cases_per_edge): ... here.
10743         (try_switch_expansion): Likewise.
10744         (switch_decision_tree::try_switch_expansion): Likewise.
10745         (record_phi_operand_mapping): Likewise.
10746         (switch_decision_tree::record_phi_operand_mapping): Likewise.
10747         (emit_case_decision_tree): Likewise.
10748         (switch_decision_tree::emit): Likewise.
10749         (balance_case_nodes): Likewise.
10750         (switch_decision_tree::balance_case_nodes): Likewise.
10751         (dump_case_nodes): Likewise.
10752         (switch_decision_tree::dump_case_nodes): Likewise.
10753         (emit_jump): Likewise.
10754         (switch_decision_tree::emit_jump): Likewise.
10755         (emit_cmp_and_jump_insns): Likewise.
10756         (switch_decision_tree::emit_cmp_and_jump_insns): Likewise.
10757         (emit_case_nodes): Likewise.
10758         (switch_decision_tree::emit_case_nodes): Likewise.
10759         (conditional_probability): Remove.
10760         * tree-switch-conversion.h (enum cluster_type): New.
10761         (PRINT_CASE): New.
10762         (struct cluster): Likewise.
10763         (cluster::cluster): Likewise.
10764         (struct simple_cluster): Likewise.
10765         (simple_cluster::simple_cluster): Likewise.
10766         (struct group_cluster): Likewise.
10767         (struct jump_table_cluster): Likewise.
10768         (struct bit_test_cluster): Likewise.
10769         (struct min_cluster_item): Likewise.
10770         (struct case_tree_node): Likewise.
10771         (case_tree_node::case_tree_node): Likewise.
10772         (jump_table_cluster::case_values_threshold): Likewise.
10773         (struct case_bit_test): Likewise.
10774         (struct switch_decision_tree): Likewise.
10775         (struct switch_conversion): Likewise.
10776         (switch_decision_tree::reset_out_edges_aux): Likewise.
10778 2018-06-20  Martin Liska  <mliska@suse.cz>
10780         * tree-switch-conversion.c (MAX_CASE_BIT_TESTS): Remove.
10781         (hoist_edge_and_branch_if_true): Likewise.
10782         (expand_switch_using_bit_tests_p): Likewise.
10783         (struct case_bit_test): Likewise.
10784         (case_bit_test_cmp): Likewise.
10785         (emit_case_bit_tests): Likewise.
10786         (switch_conversion::switch_conversion): New class.
10787         (struct switch_conv_info): Remove old struct.
10788         (collect_switch_conv_info): More to ...
10789         (switch_conversion::collect): ... this.
10790         (check_range): Likewise.
10791         (switch_conversion::check_range): Likewise.
10792         (check_all_empty_except_final): Likewise.
10793         (switch_conversion::check_all_empty_except_final): Likewise.
10794         (check_final_bb): Likewise.
10795         (switch_conversion::check_final_bb): Likewise.
10796         (create_temp_arrays): Likewise.
10797         (switch_conversion::create_temp_arrays): Likewise.
10798         (free_temp_arrays): Likewise.
10799         (gather_default_values): Likewise.
10800         (switch_conversion::gather_default_values): Likewise.
10801         (build_constructors): Likewise.
10802         (switch_conversion::build_constructors): Likewise.
10803         (constructor_contains_same_values_p): Likewise.
10804         (switch_conversion::contains_same_values_p): Likewise.
10805         (array_value_type): Likewise.
10806         (switch_conversion::array_value_type): Likewise.
10807         (build_one_array): Likewise.
10808         (switch_conversion::build_one_array): Likewise.
10809         (build_arrays): Likewise.
10810         (switch_conversion::build_arrays): Likewise.
10811         (gen_def_assigns): Likewise.
10812         (switch_conversion::gen_def_assigns): Likewise.
10813         (prune_bbs): Likewise.
10814         (switch_conversion::prune_bbs): Likewise.
10815         (fix_phi_nodes): Likewise.
10816         (switch_conversion::fix_phi_nodes): Likewise.
10817         (gen_inbound_check): Likewise.
10818         (switch_conversion::gen_inbound_check): Likewise.
10819         (process_switch): Use the newly created class.
10820         (switch_conversion::expand): New.
10821         (switch_conversion::~switch_conversion): New.
10822         * tree-switch-conversion.h: New file.
10824 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
10826         * tree-vectorizer.h (NUM_PATTERNS, vect_recog_func_ptr): Move to
10827         tree-vect-patterns.c.
10828         * tree-vect-patterns.c (vect_supportable_direct_optab_p): New function.
10829         (vect_recog_dot_prod_pattern): Use it.  Remove the type_in argument.
10830         (vect_recog_sad_pattern): Likewise.
10831         (vect_recog_widen_sum_pattern): Likewise.
10832         (vect_recog_pow_pattern): Likewise.  Check for a null vectype.
10833         (vect_recog_widen_shift_pattern): Remove the type_in argument.
10834         (vect_recog_rotate_pattern): Likewise.
10835         (vect_recog_mult_pattern): Likewise.
10836         (vect_recog_vector_vector_shift_pattern): Likewise.
10837         (vect_recog_divmod_pattern): Likewise.
10838         (vect_recog_mixed_size_cond_pattern): Likewise.
10839         (vect_recog_bool_pattern): Likewise.
10840         (vect_recog_mask_conversion_pattern): Likewise.
10841         (vect_try_gather_scatter_pattern): Likewise.
10842         (vect_recog_widen_mult_pattern): Likewise.  Check for a null vectype.
10843         (vect_recog_over_widening_pattern): Likewise.
10844         (vect_recog_gather_scatter_pattern): Likewise.
10845         (vect_recog_func_ptr): Move from tree-vectorizer.h
10846         (vect_vect_recog_func_ptrs): Move further down the file.
10847         (vect_recog_func): Likewise.  Remove the third argument.
10848         (NUM_PATTERNS): Define based on vect_vect_recog_func_ptrs.
10849         (vect_pattern_recog_1): Expect the pattern function to do any
10850         necessary target tests.  Also expect it to provide a vector type.
10851         Remove the type_in handling.
10853 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
10855         * tree-vect-patterns.c (vect_pattern_detected): New function.
10856         (vect_recog_dot_prod_patternm, vect_recog_sad_pattern)
10857         (vect_recog_widen_mult_pattern, vect_recog_widen_sum_pattern)
10858         (vect_recog_over_widening_pattern, vect_recog_widen_shift_pattern
10859         (vect_recog_rotate_pattern, vect_recog_vector_vector_shift_pattern)
10860         (vect_recog_mult_pattern, vect_recog_divmod_pattern)
10861         (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
10862         (vect_recog_mask_conversion_pattern)
10863         (vect_try_gather_scatter_pattern): Likewise.
10865 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
10867         * tree-vect-patterns.c (vect_get_internal_def): New function.
10868         (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
10869         (vect_recog_vector_vector_shift_pattern, check_bool_pattern)
10870         (search_type_for_mask_1): Use it.
10872 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
10874         * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
10875         redundant WIDEN_SUM_EXPR handling.
10876         (vect_recog_sad_pattern): Likewise.
10878 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
10880         * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
10881         redundant check that the types of a PLUS_EXPR or MULT_EXPR agree.
10882         (vect_recog_sad_pattern): Likewise PLUS_EXPR, ABS_EXPR and MINUS_EXPR.
10883         (vect_recog_widen_mult_pattern): Likewise MULT_EXPR.
10884         (vect_recog_widen_sum_pattern): Likewise PLUS_EXPR.
10886 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
10888         * tree-vect-stmts.c (vectorizable_call): Make sure that we
10889         use the stmt_vec_info of the original bb statement for the
10890         new zero assignment, even if the call is part of a pattern.
10892 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
10894         * tree-vectorizer.h (_stmt_vec_info): Note above pattern_def_seq
10895         that the sequence is attached to the original statement rather
10896         than the pattern statement.
10897         * tree-vect-loop.c (vect_determine_vf_for_stmt): Take the
10898         PATTERN_DEF_SEQ from the original statement rather than
10899         the main pattern statement.
10900         * tree-vect-stmts.c (free_stmt_vec_info): Likewise.
10901         * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Likewise.
10902         (vect_mark_pattern_stmts): Don't copy the PATTERN_DEF_SEQ.
10904 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
10906         * tree-vect-stmts.c (vect_analyze_stmt): Move the handling of pattern
10907         definition statements before the early exit for statements that aren't
10908         live or relevant.
10909         * tree-vect-loop.c (vect_transform_loop_stmt): New function,
10910         split out from...
10911         (vect_transform_loop): ...here.  Process pattern definition
10912         statements without first checking whether the main pattern
10913         statement is live or relevant.
10915 2018-06-19  Eric Botcazou  <ebotcazou@adacore.com>
10917         * tree-cfgcleanup.c (tree_forwarder_block_p): Do not return false at
10918         -O0 if the locus represent UNKNOWN_LOCATION but have different values.
10920 2018-06-19  Aaron Sawdey  <acsawdey@linux.ibm.com>
10922         * config/rs6000/rs6000-string.c (select_block_compare_mode): Check
10923         TARGET_EFFICIENT_OVERLAPPING_UNALIGNED here instead of in caller.
10924         (do_and3, do_and3_mask, do_compb3, do_rotl3): New functions.
10925         (expand_block_compare): Change select_block_compare_mode call.
10926         (expand_strncmp_align_check): Use new functions, fix comment.
10927         (emit_final_str_compare_gpr): New function.
10928         (expand_strn_compare): Refactor and clean up code.
10929         * config/rs6000/vsx.md (vsx_mov<mode>_64bit): Remove *.
10931 2018-06-19  Tony Reix  <tony.reix@atos.com>
10932             Damien Bergamini  <damien.bergamini@atos.com>
10933             David Edelsohn  <dje.gcc@gmail.com>
10935         * collect2.c (static_obj): New variable.
10936         (static_libs): New variable.
10937         (is_in_list): Uncomment declaration.
10938         (main): Track AIX libraries linked statically.
10939         (is_in_list): Uncomment definition.
10940         (scan_prog_file): Don't add AIX shared libraries initializer
10941         to constructor list if linking statically.
10943 2018-06-19  Max Filippov  <jcmvbkbc@gmail.com>
10945         * config/xtensa/xtensa.md (UNSPEC_FRAME_BLOCKAGE): New unspec
10946         constant.
10947         (allocate_stack, frame_blockage, *frame_blockage): New patterns.
10949 2018-06-19  Jan Hubicka  <hubicka@ucw.cz>
10951         * tree.c (find_decls_types_r): Remove all non-VAR_DECLs from
10952         blocks.
10954 2018-06-19  Martin Liska  <mliska@suse.cz>
10956         * config/i386/i386.c (ix86_can_inline_p): Do not use
10957         ipa_fn_summaries::get_create.
10958         * ipa-cp.c (ipcp_cloning_candidate_p): Replace get_create with
10959         get.
10960         (devirtualization_time_bonus): Likewise.
10961         (ipcp_propagate_stage): Likewise.
10962         * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
10963         (edge_set_predicate): Likewise.
10964         (evaluate_conditions_for_known_args): Likewise.
10965         (evaluate_properties_for_edge): Likewise.
10966         (ipa_call_summary::reset): Tranform to ...
10967         (ipa_call_summary::~ipa_call_summary): ... this.
10968         (ipa_fn_summary::reset): Transform to ...
10969         (ipa_fn_summary::~ipa_fn_summary): ... this.
10970         (ipa_fn_summary_t::remove): Rename to ...
10971         (ipa_fn_summary_t::remove_callees): ... this.
10972         (ipa_fn_summary_t::duplicate): Use placement new
10973         instead of memory copy.
10974         (ipa_call_summary_t::duplicate): Likewise.
10975         (ipa_call_summary_t::remove): Remove.
10976         (dump_ipa_call_summary): Change get_create to get.
10977         (ipa_dump_fn_summary): Dump only when summary exists.
10978         (analyze_function_body): Use symbol_summary::get instead
10979         of get_create.
10980         (compute_fn_summary): Likewise.
10981         (estimate_edge_devirt_benefit): Likewise.
10982         (estimate_edge_size_and_time): Likewise.
10983         (inline_update_callee_summaries): Likewise.
10984         (remap_edge_change_prob): Likewise.
10985         (remap_edge_summaries): Likewise.
10986         (ipa_merge_fn_summary_after_inlining): Likewise.
10987         (write_ipa_call_summary): Likewise.
10988         (ipa_fn_summary_write): Likewise.
10989         (ipa_free_fn_summary): Likewise.
10990         * ipa-fnsummary.h (struct GTY): Add new ctor and copy ctor.
10991         (struct ipa_call_summary): Likewise.
10992         * ipa-icf.c (sem_function::merge): Use symbol_summary::get instead
10993         of get_create.
10994         * ipa-inline-analysis.c (do_estimate_edge_time): Likewise.
10995         (estimate_size_after_inlining): Likewise.
10996         (estimate_growth): Likewise.
10997         (growth_likely_positive): Likewise.
10998         * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
10999         (inline_call): Likewise.
11000         * ipa-inline.c (caller_growth_limits): Likewise.
11001         (can_inline_edge_p): Likewise.
11002         (can_inline_edge_by_limits_p): Likewise.
11003         (compute_uninlined_call_time): Likewise.
11004         (compute_inlined_call_time): Likewise.
11005         (want_inline_small_function_p): Likewise.
11006         (edge_badness): Likewise.
11007         (update_caller_keys): Likewise.
11008         (update_callee_keys): Likewise.
11009         (inline_small_functions): Likewise.
11010         (inline_to_all_callers_1): Likewise.
11011         (dump_overall_stats): Likewise.
11012         (early_inline_small_functions): Likewise.
11013         (early_inliner): Likewise.
11014         * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
11015         * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
11016         * ipa-pure-const.c (malloc_candidate_p): Likewise.
11017         * ipa-split.c (execute_split_functions): Likewise.
11018         * symbol-summary.h: Likewise.
11019         * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
11021 2018-06-19  Richard Biener  <rguenther@suse.de>
11023         * tree-vectorizer.c (try_vectorize_loop_1): Split out of ...
11024         (vectorize_loops): ... here.  Fix dbgcnt handling.
11025         (try_vectorize_loop): Wrap try_vectorize_loop_1.
11027 2018-06-19  Segher Boessenkool  <segher@kernel.crashing.org>
11029         PR target/86197
11030         * config/rs6000/rs6000.md (rs6000_discover_homogeneous_aggregate): An
11031         ieee128 argument takes up only one (vector) register, not two (floating
11032         point) registers.
11034 2018-06-19  Eric Botcazou  <ebotcazou@adacore.com>
11036         * gimplify.c (gimplify_init_constructor): Really never clear for an
11037         incomplete constructor if CONSTRUCTOR_NO_CLEARING is set.
11039 2018-06-19  Richard Biener  <rguenther@suse.de>
11041         PR tree-optimization/86179
11042         * tree-vect-patterns.c (vect_pattern_recog_1): Clean up
11043         after failed recognition.
11045 2018-06-18  Martin Sebor  <msebor@redhat.com>
11047         PR middle-end/85602
11048         * calls.c (maybe_warn_nonstring_arg): Handle strncat.
11049         * tree-ssa-strlen.c (is_strlen_related_p): Make extern.
11050         Handle integer subtraction.
11051         (maybe_diag_stxncpy_trunc): Handle nonstring source arguments.
11052         * tree-ssa-strlen.h (is_strlen_related_p): Declare.
11054 2018-06-18  David Malcolm  <dmalcolm@redhat.com>
11056         * config/frv/frv-protos.h (frv_ifcvt_modify_insn): Strengthen 3rd
11057         param from rtx to rtx_insn *.
11058         * config/frv/frv.c (frv_ifcvt_add_insn): Likewise for "insn"
11059         param.
11060         (frv_ifcvt_modify_insn): Likwise.
11061         (frv_ifcvt_modify_final): Likwise for local "existing_insn",
11062         adding an as_a <rtx_insn *> cast.  Likewise for local "insn".
11063         * config/mips/mips.c (r10k_insert_cache_barriers): Add an
11064         as_a <rtx_insn *> cast to local "unprotected_region" once
11065         it's been established that it's not NULL or pc_rtx.
11066         * config/nds32/nds32-relax-opt.c (nds32_group_insns): Strengthen
11067         param "sethi" from rtx to rtx_insn *.
11068         (nds32_group_float_insns): Likewise for param "insn".
11069         * config/vax/vax-protos.h (vax_output_int_add): Likewise for 1st
11070         param.
11071         (vax_output_int_subtract): Likewise.
11072         * config/vax/vax.c (vax_output_int_add): Likewise for param
11073         "insn".
11074         (vax_output_int_subtract): Likewise.
11075         * emit-rtl.c (set_insn_deleted): Likewise, removing cast.
11076         (emit_pattern_after): Likewise for param "after".
11077         (emit_insn_after): Likewise.
11078         (emit_jump_insn_after): Likewise.
11079         (emit_call_insn_after): Likewise.
11080         (emit_debug_insn_after): Likewise.
11081         (emit_pattern_before): Likewise for param "before".
11082         (emit_insn_before): Likewise.
11083         (emit_jump_insn_before): Likewise.
11084         * final.c (get_insn_template): Likewise for param "insn", removing
11085         a cast.
11086         * output.h (get_insn_template): Likewise for 2nd param.
11087         * rtl.h (emit_insn_before): Likewise.
11088         (emit_jump_insn_before): Likewise.
11089         (emit_debug_insn_before_noloc): Likewise.
11090         (emit_insn_after): Likewise.
11091         (emit_jump_insn_after): Likewise.
11092         (emit_call_insn_after): Likewise.
11093         (emit_debug_insn_after): Likewise.
11094         (set_insn_deleted): Likewise for param.
11096 2018-06-18  Michael Meissner  <meissner@linux.ibm.com>
11098         PR target/85358
11099         * config/rs6000/rs6000-modes.def (toplevel): Rework the 128-bit
11100         floating point modes, so that IFmode is numerically greater than
11101         TFmode, which is greater than KFmode using FRACTIONAL_FLOAT_MODE
11102         to declare the ordering.  This prevents IFmode from being
11103         converted to TFmode when long double is IEEE 128-bit on an ISA 3.0
11104         machine.  Include rs6000-modes.h to share the fractional values
11105         between genmodes* and the rest of the compiler.
11106         (IFmode): Likewise.
11107         (KFmode): Likewise.
11108         (TFmode): Likewise.
11109         * config/rs6000/rs6000-modes.h: New file.
11110         * config/rs6000/rs6000.c (rs6000_debug_reg_global): Change the
11111         meaning of rs6000_long_double_size so that 126..128 selects an
11112         appropriate 128-bit floating point type.
11113         (rs6000_option_override_internal): Likewise.
11114         * config/rs6000/rs6000.h (toplevel): Include rs6000-modes.h.
11115         (TARGET_LONG_DOUBLE_128): Change the meaning of
11116         rs6000_long_double_size so that 126..128 selects an appropriate
11117         128-bit floating point type.
11118         (LONG_DOUBLE_TYPE_SIZE): Update comment.
11119         * config/rs6000/rs6000.md (trunciftf2): Correct the modes of the
11120         source and destination to match the standard usage.
11121         (truncifkf2): Likewise.
11122         (copysign<mode>3, IEEE iterator): Rework copysign of float128 on
11123         ISA 2.07 to use an explicit clobber, instead of passing in a
11124         temporary.
11125         (copysign<mode>3_soft): Likewise.
11127 2018-06-18  David Malcolm  <dmalcolm@redhat.com>
11129         * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
11130         Replace dump_printf_loc call with DUMP_VECT_SCOPE.
11131         (vect_slp_analyze_instance_dependence): Likewise.
11132         (vect_enhance_data_refs_alignment): Likewise.
11133         (vect_analyze_data_refs_alignment): Likewise.
11134         (vect_slp_analyze_and_verify_instance_alignment
11135         (vect_analyze_data_ref_accesses): Likewise.
11136         (vect_prune_runtime_alias_test_list): Likewise.
11137         (vect_analyze_data_refs): Likewise.
11138         * tree-vect-loop-manip.c (vect_update_inits_of_drs): Likewise.
11139         * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
11140         (vect_analyze_scalar_cycles_1): Likewise.
11141         (vect_get_loop_niters): Likewise.
11142         (vect_analyze_loop_form_1): Likewise.
11143         (vect_update_vf_for_slp): Likewise.
11144         (vect_analyze_loop_operations): Likewise.
11145         (vect_analyze_loop): Likewise.
11146         (vectorizable_induction): Likewise.
11147         (vect_transform_loop): Likewise.
11148         * tree-vect-patterns.c (vect_pattern_recog): Likewise.
11149         * tree-vect-slp.c (vect_analyze_slp): Likewise.
11150         (vect_make_slp_decision): Likewise.
11151         (vect_detect_hybrid_slp): Likewise.
11152         (vect_slp_analyze_operations): Likewise.
11153         (vect_slp_bb): Likewise.
11154         * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Likewise.
11155         (vectorizable_bswap): Likewise.
11156         (vectorizable_call): Likewise.
11157         (vectorizable_simd_clone_call): Likewise.
11158         (vectorizable_conversion): Likewise.
11159         (vectorizable_assignment): Likewise.
11160         (vectorizable_shift): Likewise.
11161         (vectorizable_operation): Likewise.
11162         * tree-vectorizer.h (DUMP_VECT_SCOPE): New macro.
11164 2018-06-18  Martin Sebor  <msebor@redhat.com>
11166         PR tree-optimization/81384
11167         * builtin-types.def (BT_FN_SIZE_CONST_STRING_SIZE): New.
11168         * builtins.c (expand_builtin_strnlen): New function.
11169         (expand_builtin): Call it.
11170         (fold_builtin_n): Avoid setting TREE_NO_WARNING.
11171         * builtins.def (BUILT_IN_STRNLEN): New.
11172         * calls.c (maybe_warn_nonstring_arg): Handle BUILT_IN_STRNLEN.
11173         Warn for bounds in excess of maximum object size.
11174         * tree-ssa-strlen.c (maybe_set_strlen_range): Return tree representing
11175         single-value ranges.  Handle strnlen.
11176         (handle_builtin_strlen): Handle strnlen.
11177         (strlen_check_and_optimize_stmt): Same.
11178         * doc/extend.texi (Other Builtins): Document strnlen.
11180 2018-06-18  Maya Rashish  <coypu@sdf.org>
11182         * config/alpha/openbsd.h (TARGET_DEFAULT): Define.
11183         (LINK_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): Likewise.
11184         (INTMAX_TYPE, UINTMAX_TYPE, WINT_TYPE): Likewise.
11186         * config/alpha/elf.h (STARTFILE_SPEC, ENDFILE_SPEC): Move from
11187         here to ...
11188         * config/alpha/linux.h (STARTFILE_SPEC, ENDFILE_SPEC): Here.
11190 2018-06-18  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
11192         * tree.c (escaped_string::escape): Replace cast to char * by
11193         const_cast<char *> (unescaped).
11195 2018-06-18  Nick Clifton  <nickc@redhat.com>
11197         PR 84195
11198         * tree.c (escaped_string): New class.  Converts an unescaped
11199         string into its escaped equivalent.
11200         (warn_deprecated_use): Use the new class to convert the
11201         deprecation message, if present.
11202         (test_escaped_strings): New self test.
11203         (test_c_tests): Add test_escaped_strings.
11204         * doc/extend.texi (deprecated): Add a note that the
11205         deprecation message is affected by the -fmessage-length
11206         option, and that control characters will be escaped.
11207         (#pragma GCC error): Document this pragma.
11208         (#pragma GCC warning): Likewise.
11209         * doc/invoke.texi (-fmessage-length): Document this option's
11210         effect on the #warning and #error preprocessor directives and
11211         the deprecated attribute.
11213 2018-06-18  Eric Botcazou  <ebotcazou@adacore.com>
11215         * tree.c (decl_value_expr_lookup): Revert latest change.
11216         (decl_value_expr_insert): Likewise.
11218 2018-06-17  Eric Botcazou  <ebotcazou@adacore.com>
11220         * gimplify.c (nonlocal_vlas): Delete.
11221         (nonlocal_vla_vars): Likewise.
11222         (gimplify_var_or_parm_decl): Do not add debug VAR_DECLs for non-local
11223         referenced VLAs.
11224         (gimplify_body): Do not create and destroy nonlocal_vlas.
11225         * tree-nested.c: Include diagnostic.h.
11226         (use_pointer_in_frame): Tweak.
11227         (lookup_field_for_decl): Add assertion and declare the transformation.
11228         (convert_nonlocal_reference_op) <PARM_DECL>: Rework and issue an
11229         internal error when the reference is in a wrong context.  Do not
11230         create a debug decl by default.
11231         (note_nonlocal_block_vlas): Delete.
11232         (convert_nonlocal_reference_stmt) <GIMPLE_BIND>: Do not call it.
11233         (convert_local_reference_op) <PARM_DECL>: Skip the frame decl.  Do not
11234         create a debug decl by default.
11235         (convert_gimple_call) <GIMPLE_CALL>: Issue an internal error when the
11236         call is in a wrong context.
11237         (fixup_vla_decls): New function.
11238         (finalize_nesting_tree_1): Adjust comment.  Call fixup_vla_decls if no
11239         debug variables were created.
11240         * tree.c (decl_value_expr_lookup): Add checking assertion.
11241         (decl_value_expr_insert): Likewise.
11243 2018-06-16  Kugan Vivekanandarajah  <kuganv@linaro.org>
11245         PR middle-end/82479
11246         * ipa-fnsummary.c (will_be_nonconstant_expr_predicate): Handle CALL_EXPR.
11247         * tree-scalar-evolution.c (interpret_expr): Likewise.
11248         (expression_expensive_p): Likewise.
11249         * tree-ssa-loop-ivopts.c (contains_abnormal_ssa_name_p): Likewise.
11250         * tree-ssa-loop-niter.c (number_of_iterations_popcount): New.
11251         (number_of_iterations_exit_assumptions): Use number_of_iterations_popcount.
11252         (ssa_defined_by_minus_one_stmt_p): New.
11254 2018-06-16  Kugan Vivekanandarajah  <kuganv@linaro.org>
11256         PR middle-end/64946
11257         * cfgexpand.c (expand_debug_expr): Hande ABSU_EXPR.
11258         * config/i386/i386.c (ix86_add_stmt_cost): Likewise.
11259         * dojump.c (do_jump): Likewise.
11260         * expr.c (expand_expr_real_2): Check operand type's sign.
11261         * fold-const.c (const_unop): Handle ABSU_EXPR.
11262         (fold_abs_const): Likewise.
11263         * gimple-pretty-print.c (dump_unary_rhs): Likewise.
11264         * gimple-ssa-backprop.c (backprop::process_assign_use): Likesie.
11265         (strip_sign_op_1): Likesise.
11266         * match.pd: Add new pattern to generate ABSU_EXPR.
11267         * optabs-tree.c (optab_for_tree_code): Handle ABSU_EXPR.
11268         * tree-cfg.c (verify_gimple_assign_unary): Likewise.
11269         * tree-eh.c (operation_could_trap_helper_p): Likewise.
11270         * tree-inline.c (estimate_operator_cost): Likewise.
11271         * tree-pretty-print.c (dump_generic_node): Likewise.
11272         * tree-vect-patterns.c (vect_recog_sad_pattern): Likewise.
11273         * tree.def (ABSU_EXPR): New.
11275 2018-06-16  Jakub Jelinek  <jakub@redhat.com>
11277         PR middle-end/86095
11278         * common.opt (Wunsafe-loop-optimizations): Add Ignore, remove Var,
11279         documented as preserved for backward compatibility only.
11280         * doc/invoke.texi: Remove -Wunsafe-loop-optimizations documentation.
11282         PR rtl-optimization/86108
11283         * bb-reorder.c (create_forwarder_block): Renamed to ...
11284         (create_eh_forwarder_block): ... this.  Split OLD_BB after labels and
11285         jump from new landing pad to the second part.
11286         (sjlj_fix_up_crossing_landing_pad, dw2_fix_up_crossing_landing_pad):
11287         Adjust callers.
11289 2018-06-15  Jakub Jelinek  <jakub@redhat.com>
11291         PR middle-end/85878
11292         * expr.c (expand_assignment): Remove now redundant COMPLEX_MODE_P
11293         check from first store_expr, use to_mode instead of GET_MODE (to_rtx).
11294         Only call store_expr for halves if the mode is the same.
11296         PR middle-end/86123
11297         * match.pd ((X / Y) == 0 -> X < Y): Don't transform complex divisions.
11298         Fix up comment formatting.
11300 2018-06-15  Bernd Edlinger  <bernd.edlinger@hotmail.de>
11302         * typed-splay-tree.h (typed_splay_tree::remove): New function.
11303         (typed_splay_tree::closure,
11304         typed_splay_tree::inner_foreach_fn, typed_splay_tree::m_inner): Deleted.
11305         (typed_splay_tree::typed_splay_tree,
11306         typed_splay_tree::operator =): Declared private.
11307         (typed_splay_tree::splay_tree_key, typed_splay_tree::splay_tree_value,
11308         typed_splay_tree::splay_tree_node_s, typed_splay_tree::KDEL,
11309         typed_splay_tree::VDEL, typed_splay_tree::splay_tree_delete_helper,
11310         typed_splay_tree::rotate_left, typed_splay_tree::rotate_right,
11311         typed_splay_tree::splay_tree_splay,
11312         typed_splay_tree::splay_tree_foreach_helper,
11313         typed_splay_tree::splay_tree_insert,
11314         typed_splay_tree::splay_tree_remove,
11315         typed_splay_tree::splay_tree_lookup,
11316         typed_splay_tree::splay_tree_predecessor,
11317         typed_splay_tree::splay_tree_successor,
11318         typed_splay_tree::splay_tree_min,
11319         typed_splay_tree::splay_tree_max): Took over from splay-tree.c/.h.
11320         (typed_splay_tree::root, typed_splay_tree::comp,
11321         typed_splay_tree::delete_key,
11322         typed_splay_tree::delete_value): New data members.
11323         * typed-splay-tree.c (selftest::test_str_to_int): Add a test for
11324         typed_splay_tree::remove.
11326 2018-06-15  Matthew Fortune  <matthew.fortune@mips.com>
11328         * config/mips/mips.h (ASM_SPEC): Pass through -mcrc, -mno-crc,
11329         -mginv and -mno-ginv to the assembler.
11330         * config/mips/mips.opt (-mcrc): New option.
11331         (-mginv): Likewise.
11332         * doc/invoke.text (-mcrc): Document.
11333         (-mginv): Likewise.
11335 2018-06-15  Nick Clifton  <nickc@redhat.com>
11337         PR 84195
11338         * tree.c (escaped_string): New class.  Converts an unescaped
11339         string into its escaped equivalent.
11340         (warn_deprecated_use): Use the new class to convert the
11341         deprecation message, if present.
11342         (test_escaped_strings): New self test.
11343         (test_c_tests): Add test_escaped_strings.
11344         * doc/extend.texi (deprecated): Add a note that the
11345         deprecation message is affected by the -fmessage-length
11346         option, and that control characters will be escaped.
11347         (#pragma GCC error): Document this pragma.
11348         (#pragma GCC warning): Likewise.
11349         * doc/invoke.texi (-fmessage-length): Document this option's
11350         effect on the #warning and #error preprocessor directives and
11351         the deprecated attribute.
11353 2018-06-15  Richard Biener  <rguenther@suse.de>
11355         * tree-vect-slp.c (vect_slp_bb): Dump MSG_OPTIMIZED_LOCATIONS
11356         here, also noting vector size used.
11357         * tree-vectorizer.c (vectorize_loops): Adjust.  Note vector
11358         size used in MSG_OPTIMIZED_LOCATIONS dump.
11359         (pass_slp_vectorize::execute): Adjust.
11361 2018-06-15  Claudiu Zissulescu  <claziss@synopsys.com>
11363         PR target/85968
11364         * config/arc/arc.c (arc_return_address_register): Fix
11365         if-condition.
11367 2018-06-15  Richard Biener  <rguenther@suse.de>
11369         PR middle-end/86159
11370         * tree-cfg.c (gimplify_build3): Do not strip sign conversions,
11371         leave useless conversion stripping to force_gimple_operand_gsi.
11372         (gimplify_build2): Likewise.
11373         (gimplify_build1): Likewise.
11375 2018-06-15  Richard Biener  <rguenther@suse.de>
11377         PR middle-end/86076
11378         * tree-cfg.c (move_stmt_op): unshare invariant addresses
11379         before adjusting their block.
11381 2018-06-15  Sebastian Huber  <sebastian.huber@embedded-brains.de>
11383         * config.gcc (riscv*-*-elf* | riscv*-*-rtems*): Use custom
11384         multilibs for *-*-rtems*.
11385         * config/riscv/t-rtems: New file.
11387 2018-06-14  Jakub Jelinek  <jakub@redhat.com>
11389         PR middle-end/86122
11390         * match.pd ((A +- CST1) +- CST2): Punt if last resort
11391         unsigned_type_for returns NULL.
11393         PR target/85945
11394         * lower-subreg.c (find_decomposable_subregs): Don't decompose float
11395         subregs of multi-word pseudos unless the float mode has word size.
11397 2018-06-14  Richard Biener  <rguenther@suse.de>
11399         PR middle-end/86139
11400         * tree-vect-generic.c (build_word_mode_vector_type): Remove
11401         duplicate and harmful type_hash_canon.
11402         * tree.c (type_hash_canon): Assert we didn't find ourselves.
11404 2018-06-14  Richard Biener  <rguenther@suse.de>
11406         PR ipa/86124
11407         * tree-ssa-struct-alias.c (create_variable_info_for): Handle
11408         NULL cgraph_node.
11410 2018-06-14  Sebastian Huber  <sebastian.huber@embedded-brains.de>
11412         * config/rtems.h (STDINT_LONG32): Define.
11414 2018-06-13  Matthew Fortune  <matthew.fortune@mips.com>
11415             Prachi Godbole  <prachi.godbole@imgtec.com>
11417         * config/mips/mips-cpus.def: Define P6600.
11418         * config/mips/mips-tables.opt: Regenerate.
11419         * config/mips/mips.c (mips_ucbranch_type): New enum.
11420         (mips_rtx_cost_data): Add support for P6600.
11421         (mips_issue_rate): Likewise.
11422         (mips_multipass_dfa_lookahead): Likewise.
11423         (mips_avoid_hazard): Likewise.
11424         (mips_reorg_process_insns): Likewise.
11425         (mips_classify_branch_p6600): New function.
11426         * config/mips/mips.h (TUNE_P6600): New define.
11427         (MIPS_ISA_LEVEL_SPEC): Infer mips64r6 from p6600.
11428         (ENABLE_LD_ST_PAIRS): Enable load/store bonding for p6600.
11429         * config/mips/mips.md: Include p6600.md.
11430         (processor): Add p6600.
11431         * config/mips/p6600.md: New file.
11432         * doc/invoke.texi: Add p6600 to supported architectures.
11434 2018-06-13  Martin Sebor  <msebor@redhat.com>
11436         PR tree-optimization/86114
11437         * gimple-fold.c (gimple_fold_builtin_strlen): Only handle LHS
11438         of integer types.
11439         * tree-ssa-strlen.c (maybe_set_strlen_range): Same.
11441 2018-06-13  Richard Biener  <rguenther@suse.de>
11443         * tree-vect-patterns.c (vect_recog_vector_vector_shift_pattern):
11444         Properly set vector type of the intermediate stmt.
11445         * tree-vect-stmts.c (vectorizable_operation): The destination
11446         var always has vectype_out type.
11448 2018-06-13  Jeff Law  <law@redhat.com>
11450         * config/rl78/rl78.c (move_elim_pass): Use TDF_NONE rather than
11451         integer 0 for argument to print_rtl_with_bb.
11452         (rl78_reorg): Likewise.
11454 2018-06-13  David Malcolm  <dmalcolm@redhat.com>
11456         * config/arc/arc.c (hwloop_optimize): Strengthen local "end_label"
11457         from rtx to rtx_insn *.
11458         * config/bfin/bfin.c (hwloop_optimize): Likewise for local
11459         "label".
11460         (add_sched_insns_for_speculation): Likewise for local "target",
11461         converting usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
11462         * config/c6x/c6x.c (reorg_split_calls): Strengthen param "call_labels"
11463         from rtx_insn ** to rtx_code_label **.
11464         (reorg_emit_nops): Likewise.
11465         (c6x_reorg): Likewise for local "call_labels".
11466         * config/sh/sh-protos.h (get_dest_uid): Strengthen 1st param from
11467         rtx to rtx_insn *.
11468         * config/sh/sh.c (dump_table): Strengthen local "lab" from rtx to
11469         rtx_code_label *, adding safe_as_a <rtx_code_label *> casts to
11470         the loops over LABEL_REFS.
11471         (fixup_addr_diff_vecs): Add as_a <rtx_insn *> to usage of
11472         braf_label.
11473         (barrier_align): Convert usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
11474         (get_dest_uid): Strengthen param "label" from rtx to rtx_insn *.
11475         (split_branches): Strengthen local "olabel" from rtx to
11476         rtx_insn *, adding a safe_as_a cast.
11477         * emit-rtl.c (next_real_insn): Strengthen param from "rtx"
11478         to "rtx_insn *".
11479         (add_insn_after): Likewise for first two params.
11480         (add_insn_before): Likewise.
11481         (remove_insn): Likewise for param.
11482         (emit_pattern_before_noloc): Likewise for second and third params.
11483         (emit_jump_insn_before_noloc): Convert NULL_RTX to NULL.
11484         (emit_call_insn_before_noloc): Likewise.
11485         (emit_debug_insn_before_noloc): Strengthen "before" param from "rtx"
11486         to "rtx_insn *".
11487         (emit_barrier_before): Likewise.
11488         (emit_label_before): Strengthen "label" param from "rtx" to
11489         "rtx_code_label *".  Strengthen "before" param from "rtx" to
11490         "rtx_insn *".
11491         (emit_insn_after_1): Strengthen "after" param from "rtx" to
11492         "rtx_insn *".
11493         (emit_pattern_after_noloc): Likewise.
11494         (emit_insn_after_noloc): Likewise.
11495         (emit_jump_insn_after_noloc): Likewise.
11496         (emit_call_insn_after_noloc): Likewise.
11497         (emit_debug_insn_after_noloc): Likewise.
11498         (emit_barrier_after): Likewise.
11499         (emit_label_after): Likewise for both params.
11500         (emit_pattern_after_setloc): Likewise for "after" param.  Convert
11501         "loc" param from "int" to "location_t".
11502         (emit_insn_after_setloc): Likewise.
11503         (emit_jump_insn_after_setloc): Likewise.
11504         (emit_call_insn_after_setloc): Likewise.
11505         (emit_debug_insn_after_setloc): Likewise.
11506         (emit_pattern_before_setloc): Likewise for "before" param.  Convert
11507         "loc" param from "int" to "location_t".
11508         (emit_pattern_before): Convert NULL_RTX to NULL.
11509         (emit_insn_before_setloc): Convert "loc" param from "int" to
11510         "location_t".
11511         (emit_jump_insn_before_setloc): Likewise.
11512         (emit_call_insn_before_setloc): Likewise.
11513         (emit_debug_insn_before_setloc): Strengthen "before" param from rtx to
11514         rtx_insn *.  Convert "loc" param from "int" to "location_t".
11515         * rtl.h (emit_insn_before_setloc, emit_jump_insn_before_setloc,
11516         emit_call_insn_before_setloc, emit_debug_insn_before_setloc):
11517         Convert 3rd param from "int" to "location_t".
11518         (emit_barrier_before, emit_barrier_after, next_real_insn):
11519         Strengthen param from rtx to rtx_insn *.
11520         (emit_label_before): Strengthen 1st param from "rtx" to
11521         "rtx_code_label *".  Strengthen 2nd param from "rtx" to
11522         "rtx_insn *".
11523         (emit_insn_after_noloc, emit_jump_insn_after_noloc,
11524         emit_call_insn_after_noloc, emit_debug_insn_after_noloc):
11525         Strengthen 2nd param from "rtx" to "rtx_insn *".
11526         (emit_insn_after_setloc, emit_jump_insn_after_setloc)
11527         emit_call_insn_after_setloc, emit_debug_insn_after_setloc):
11528         Likewise. Convert 3rd param from "int" to "location_t".
11529         (emit_label_after): Strengthen 1st param from "rtx" to
11530         "rtx_code_label *".
11531         (next_real_insn, remove_insn): Strengthen param from "rtx" to
11532         "rtx_insn *".
11533         (add_insn_before, add_insn_after): Strengthen 1st and 2nd params
11534         from "rtx" to "rtx_insn *".
11536 2018-06-13  Jan Hubicka  <hubicka@gcc.gnu.org>
11538         * cgraph.c (cgraph_node::get_untransformed_body): Dump function
11539         bodies streamed in with -Q.
11540         * dumpfile.c (dump_files): Add lto-stream-out dump file.
11541         * dumpfile.h (tree_dump_index): Add lto_stream_out.
11542         * gimple-streamer-out.c: Include gimple-pretty-print.h
11543         (output_bb): Dump stmts streamed.
11544         * lto-section-out.c: Include print-tree.h
11545         (lto_begin_section): Dump sections created.
11546         (lto_output_decl_index): Dump decl encoded.
11547         * lto-streamer-out.c: Include print-tree.h
11548         (create_output_block): Dump output block created.
11549         (DFS::DFS_write_tree_body): Dump DFS SCCs streamed.
11550         (output_function): Dump function output.
11551         (output_constructor): Dump constructor streamed.
11552         (write_global_stream): Output indexes encoded.
11553         (produce_asm_for_decls): Dump streams encoded.
11554         * lto-streamer.c (streamer_dump_file): New global var.
11555         * lto-streamer.h (streamer_dump_file): Declare.
11556         * passes.c (ipa_write_summaries): Initialize streamer dump.
11557         * varpool.c (varpool_node::get_constructor): Dump constructors streamed
11558         in.
11560 2018-06-13  Eric Botcazou  <ebotcazou@adacore.com>
11562         PR target/86048
11563         * config/i386/winnt.c (i386_pe_seh_cold_init): Do not emit negative
11564         offsets for register save directives.  Emit a second batch of save
11565         directives, if need be, when the function accesses prior frames.
11567 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
11569         * config/arc/fpu.md (fmasf4): Force operand to register.
11570         (fnmasf4): Likewise.
11572 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
11574         * config/arc/arc-protos.h (arc_pad_return): Remove.
11575         * config/arc/arc.c (machine_function): Remove force_short_suffix
11576         and size_reason.
11577         (arc_print_operand): Adjust printing of '&'.
11578         (arc_verify_short): Remove conditional printing of short suffix.
11579         (arc_final_prescan_insn): Remove reference to size_reason.
11580         (pad_return): New function.
11581         (arc_reorg): Call pad_return.
11582         (arc_pad_return): Remove.
11583         (arc_init_machine_status): Remove reference to force_short_suffix.
11584         * config/arc/arc.md (vunspec): Add VUNSPEC_ARC_BLOCKAGE.
11585         (attr length): When attribute iscompact is true force to 2
11586         regardless; in the case of maybe check if we want to force the
11587         instruction to have 4 bytes length.
11588         (nopv): Change it to generate 4 byte long nop as well.
11589         (blockage): New pattern.
11590         (simple_return): Remove call to arc_pad_return.
11591         (p_return_i): Likewise.
11593 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
11595         * config/arc/elf.h (LINK_GCC_C_SEQUENCE_SPEC): Define.
11597 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
11599         * config/arc/builtins.def (SYNC): SYNC instruction is valid on all
11600         ARC cores.
11602 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
11604         * config/arc/arc.c (atomic_exchangesi): EX instruction is default
11605         for ARC700 and ARCv2.
11607 2018-06-13  Chenghua Xu <paul.hua.gm@gmail.com>
11609         PR target/86076
11610         * config/mips/loongson.md (vec_setv4hi): Gen_lowpart for
11611         operands[2] instead of operands[1].
11614 2018-06-12  Richard Sandiford  <richard.sandiford@linaro.org>
11616         * lra-constraints.c (simplify_operand_subreg): In the paradoxical
11617         case, check whether the outer register overlaps an unallocatable
11618         register, not just whether it fits the required class.
11620 2018-06-12  Richard Sandiford  <richard.sandiford@linaro.org>
11622         * poly-int.h (can_div_trunc_p): Add new overload in which all values
11623         are poly_ints.
11624         * alias.c (get_addr): Extend CONST_INT handling to poly_int_rtx_p.
11625         (memrefs_conflict_p): Likewise.
11626         (init_alias_analysis): Likewise.
11627         * cfgexpand.c (expand_debug_expr): Likewise.
11628         * combine.c (combine_simplify_rtx, force_int_to_mode): Likewise.
11629         * cse.c (fold_rtx): Likewise.
11630         * explow.c (adjust_stack, anti_adjust_stack): Likewise.
11631         * expr.c (emit_block_move_hints): Likewise.
11632         (clear_storage_hints, push_block, emit_push_insn): Likewise.
11633         (store_expr_with_bounds, reduce_to_bit_field_precision): Likewise.
11634         (emit_group_load_1): Use rtx_to_poly_int64 for group offsets.
11635         (emit_group_store): Likewise.
11636         (find_args_size_adjust): Use strip_offset.  Use rtx_to_poly_int64
11637         to read the PRE/POST_MODIFY increment.
11638         * calls.c (store_one_arg): Use strip_offset.
11639         * rtlanal.c (rtx_addr_can_trap_p_1): Extend CONST_INT handling to
11640         poly_int_rtx_p.
11641         (set_noop_p): Use rtx_to_poly_int64 for the elements selected
11642         by a VEC_SELECT.
11643         * simplify-rtx.c (avoid_constant_pool_reference): Use strip_offset.
11644         (simplify_binary_operation_1): Extend CONST_INT handling to
11645         poly_int_rtx_p.
11646         * var-tracking.c (compute_cfa_pointer): Take a poly_int64 rather
11647         than a HOST_WIDE_INT.
11648         (hard_frame_pointer_adjustment): Change from HOST_WIDE_INT to
11649         poly_int64.
11650         (adjust_mems, add_stores): Update accodingly.
11651         (vt_canonicalize_addr): Track polynomial offsets.
11652         (emit_note_insn_var_location): Likewise.
11653         (vt_add_function_parameter): Likewise.
11654         (vt_initialize): Likewise.
11656 2018-06-12  Jeff Law  <law@redhat.com>
11658         * config.gcc (alpha*-*-freebsd*): Remove.
11659         * config/alpha/freebsd.h: Remove.
11661 2018-06-12  David Malcolm  <dmalcolm@redhat.com>
11663         PR other/69968
11664         * spellcheck-tree.c (levenshtein_distance): Rename to...
11665         (get_edit_distance): ...this, and update for underlying renaming.
11666         * spellcheck-tree.h (levenshtein_distance): Rename to...
11667         (get_edit_distance): ...this.
11668         * spellcheck.c (levenshtein_distance): Rename to...
11669         (get_edit_distance): ...this.  Convert from Levenshtein distance
11670         to Damerau-Levenshtein distance by supporting transpositions of
11671         adjacent characters.  Rename "v1" to "v_next" and "v0" to
11672         "v_one_ago".
11673         (selftest::levenshtein_distance_unit_test_oneway): Rename to...
11674         (selftest::test_edit_distance_unit_test_oneway): ...this, and
11675         update for underlying renaming.
11676         (selftest::levenshtein_distance_unit_test): Rename to...
11677         (selftest::test_get_edit_distance_unit): ...this, and update for
11678         underlying renaming.
11679         (selftest::test_find_closest_string): Add example from PR 69968
11680         where transposition helps
11681         (selftest::test_metric_conditions): Update for renaming.
11682         (selftest::test_metric_conditions): Likewise.
11683         (selftest::spellcheck_c_tests): Likewise.
11684         * spellcheck.h (levenshtein_distance): Rename both overloads to...
11685         (get_edit_distance): ...this.
11686         (best_match::consider): Update for renaming.
11688 2018-06-12  Martin Sebor  <msebor@redhat.com>
11690         PR tree-optimization/85259
11691         * builtins.c (compute_objsize): Handle constant offsets.
11692         * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Return
11693         true iff a warning has been issued.
11694         * gimple.h (gimple_nonartificial_location): New function.
11695         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Call
11696         gimple_nonartificial_location and handle -Wno-system-headers.
11697         (handle_builtin_stxncpy): Same.
11699 2018-06-12  Martin Sebor  <msebor@redhat.com>
11701         PR c/85931
11702         * fold-const.c (operand_equal_p): Handle SAVE_EXPR.
11704 2018-06-12  Will Schmidt  <will_schmidt@vnet.ibm.com>
11706         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
11707         BUILTIN_VEC_XST entries for pointer to double and long long.
11709 2018-06-12  H.J. Lu  <hongjiu.lu@intel.com>
11711         PR target/85990
11712         * config/i386/gnu-user.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
11713         Update comments.
11714         * config/i386/gnu-user64.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
11715         Likewise.
11717 2018-06-12  Martin Liska  <mliska@suse.cz>
11719         * doc/options.texi: Document IntegerRange.
11721 2018-06-12  Martin Liska  <mliska@suse.cz>
11723         * config/i386/i386.opt: Make MPX-related options as Deprecated.
11724         * opt-functions.awk: Handle Deprecated flag.
11725         * opts-common.c (decode_cmdline_option): Handle cl_deprecated
11726         and report error.
11727         (read_cmdline_option): Report warning for a deprecated option.
11728         * opts.h (struct cl_option): Add new field cl_deprecated.
11729         (CL_ERR_DEPRECATED): New.
11731 2018-06-12  Martin Liska  <mliska@suse.cz>
11733         * doc/options.texi: Document Deprecated option flag.
11735 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
11737         * config/arc/arc-arch.h (arc_extras): New enum.
11738         (arc_cpu_t):Add field extra.
11739         (arc_cpu_types): Consider the extras.
11740         * config/arc/arc-cpus.def: Add extras info.
11741         * config/arc/arc-opts.h (processor_type): Consider extra field.
11742         * config/arc/arc.c (arc_override_options): Handle extra field.
11744 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
11746         * config/arc/arc-arch.h: Update ARC_OPTX macro.
11747         * config/arc/arc-options.def (ARC_OPTX): Introduce a new doc
11748         field.
11749         * config/arc/arc.c (arc_init): Update pic warning.
11750         (irq_range): Update irq range parsing warnings.
11751         (arc_override_options): Update various warning messages.
11752         (arc_handle_aux_attribute): Likewise.
11754 2018-06-12  Robert Suchanek  <robert.suchanek@mips.com>
11756         * config/mips/i6400.md (i6400_fpu_fadd): Remove frint.
11758 2018-06-12  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
11760         * doc/sourcebuild.texi: Document usage of line number 0 in verify
11761         compiler messages directives.
11763 2018-06-12  Matthew Fortune  <mfortune@gmail.com>
11765         * config/mips/mips-cpus.def: New MIPS_CPU for i6500.
11766         * config/mips/mips-tables.opt: Regenerate.
11767         * config/mips/mips.h (MIPS_ISA_LEVEL_SPEC): Mark i6500 as
11768         mips64r6.
11769         * doc/invoke.texi: Document -march=i6500.
11771 2018-06-12  Prachi Godbole  <prachi.godbole@imgtec.com>
11773         * config/mips/i6400.md (i6400_gpmuldiv): Remove cpu_unit.
11774         (i6400_gpmul): Add cpu_unit.
11775         (i6400_gpdiv): Likewise.
11776         (i6400_msa_add_d): Update reservations.
11777         (i6400_msa_int_add) Likewise.
11778         (i6400_msa_short_logic3) Likewise.
11779         (i6400_msa_short_logic2) Likewise.
11780         (i6400_msa_short_logic) Likewise.
11781         (i6400_msa_move) Likewise.
11782         (i6400_msa_cmp) Likewise.
11783         (i6400_msa_short_float2) Likewise.
11784         (i6400_msa_div_d) Likewise.
11785         (i6400_msa_long_logic1) Likewise.
11786         (i6400_msa_long_logic2) Likewise.
11787         (i6400_msa_mult) Likewise.
11788         (i6400_msa_long_float2) Likewise.
11789         (i6400_msa_long_float4) Likewise.
11790         (i6400_msa_long_float5) Likewise.
11791         (i6400_msa_long_float8) Likewise.
11792         (i6400_fpu_fadd): Include frint type.
11793         (i6400_fpu_store): New define_insn_reservation.
11794         (i6400_fpu_load): Likewise.
11795         (i6400_fpu_move): Likewise.
11796         (i6400_fpu_fcmp): Likewise.
11797         (i6400_fpu_fmadd): Likewise.
11798         (i6400_int_mult): Include imul3nc type and update reservation.
11799         (i6400_int_div): Include idiv3 type and update reservation.
11800         (i6400_int_load): Update to check type not move_type.
11801         (i6400_int_store): Likewise.
11802         (i6400_int_prefetch): Set zero latency.
11804 2018-06-12  Eric Botcazou  <ebotcazou@adacore.com>
11806         * gcc.c: Document new %@{...} sequence.
11807         (LINK_COMMAND_SPEC): Use it for the -L switches.
11808         (cpp_unique_options): Use it for the -I switches.
11809         (at_file_argbuf): New global variable.
11810         (in_at_file): Likewise.
11811         (alloc_args): Create at_file_argbuf.
11812         (clear_args): Truncate at_file_argbuf.
11813         (store_arg): If in_at_file, push the argument onto at_file_argbuf.
11814         (open_at_file): New function.
11815         (close_at_file): Likewise.
11816         (create_at_file): Delete.
11817         (do_spec_1) <'i'>: Use open_at_file/close_at_file.
11818         <'o'>: Likewise.
11819         <'@'>: New case.
11820         (validate_switches_from_spec): Deal with %@{...} sequence.
11821         (validate_switches): Likewise.
11822         (driver::finalize): Call clear_args.
11824 2018-06-11  Rasmus Villemoes  <rasmus.villemoes@prevas.dk>
11826         * config/vx-common.h (USE_TM_CLONE_REGISTRY): #define to 0.
11828 2018-06-11  Martin Sebor  <msebor@redhat.com>
11830         * doc/invoke.texi (-Wall): List -Wc++17-compat.
11831         (Wno-class-memaccess): Add @opindex.
11832         (Wno-templates, Wno-multiple-inheritance): Same.
11833         (Wno-virtual-inheritance, Wno-namespaces): Same.
11834         (Wno-pedantic, Wno-chkp, Wcoverage-mismatch): Same.
11835         (Wno-format-overflow, Wno-format-truncation): Same.
11836         (Wno-unused-local-typedefs, Walloc-size-larger-than): Same.
11837         (Wno-alloc-size-larger-than, Wframe-larger-than): Same
11838         (Wno-stack-usage, Wno-c++-compat, Wno-c++11-compat): Same.
11839         (Wno-c++14-compat, Wno-c++17-compat, Wno-openmp-simd): Same.
11840         (Wno-unsuffixed-float-constants, Wno-addr-space-convert): Same.
11841         (Wno-misspelled-isr): Same.
11843 2018-06-11  Martin Sebor  <msebor@redhat.com>
11845         * PR tree-optimization/86083
11846         * tree-ssa-strlen.c (handle_char_store): Use tree_expr_nonzero_p.
11848 2018-06-11  Zhouyi Zhou <zhouzhouyi@gmail.com>
11850         * tree-eh.c (lower_eh_constructs_2): Add a comma to comment.
11852 2018-06-11  Segher Boessenkool  <segher@kernel.crashing.org>
11854         PR target/85755
11855         * config/rs6000/rs6000.md (*movdi_internal32): Put constraint modifiers
11856         on the correct operand.
11857         (*movdi_internal64): Ditto.
11859 2018-06-11  Martin Liska  <mliska@suse.cz>
11861         PR tree-optimization/86089
11862         * tree-ssa-strlen.c (get_string_length): Move back removed hunk.
11864 2018-06-11  Julia Koval  <julia.koval@intel.com>
11866         * config/i386/vaesintrin.h (_mm_aesdec_epi128, _mm_aesdeclast_epi128)
11867         _mm_aesenc_epi128, _mm_aesenclast_epi128: Remove.
11868         * config/i386/vpclmulqdqintrin.h (_mm_clmulepi64_epi128): Remove.
11870 2018-06-11  Olivier Hainque  <hainque@adacore.com>
11872         * dwarf2out.c (gen_compile_unit_die): Fallout to DW_LANG_Ada83
11873         for Ada with strict dwarf2.
11875 2018-06-08  Peter Bergner  <bergner@vnet.ibm.com>
11877         PR target/85755
11878         * config/rs6000/rs6000.c (mem_operand_gpr): Enable PRE_INC and PRE_DEC
11879         addresses.
11881 2018-06-08  Jan Hubicka  <hubicka@ucw.cz>
11883         * dumpfile.c (FIRST_ME_AUTO_NUMBERED_DUMP): Bump to 4.
11885 2018-06-08  David Edelsohn  <dje.gcc@gmail.com>
11887         * config/rs6000/rs6000.c (rs6000_passes_ieee128): Protect with #if
11888         TARGET_ELF.
11890 2018-06-08  Martin Liska  <mliska@suse.cz>
11892         * tree-cfg.h (debug_function): Fix argument type to match
11893         implementation.
11895 2018-06-08  Martin Liska  <mliska@suse.cz>
11897         * config/powerpcspe/powerpcspe.c (rs6000_xcoff_visibility):
11898         Remove usage of MPX-related (and removed) fields.
11899         * config/rs6000/rs6000.c (rs6000_xcoff_visibility): Likewise.
11901 2018-06-08  David Malcolm  <dmalcolm@redhat.com>
11903         * cfg.c (debug): Use TDF_NONE rather than 0.
11904         * cfghooks.c (debug): Likewise.
11905         * dumpfile.c (DUMP_FILE_INFO): Likewise; also for OPTGROUP.
11906         (struct dump_option_value_info): Convert to...
11907         (struct kv_pair): ...this template type.
11908         (dump_options): Convert to kv_pair<dump_flags_t>; use TDF_NONE
11909         rather than 0.
11910         (optinfo_verbosity_options): Likewise.
11911         (optgroup_options): Convert to kv_pair<optgroup_flags_t>; use
11912         OPTGROUP_NONE.
11913         (gcc::dump_manager::dump_register): Use optgroup_flags_t rather
11914         than int for "optgroup_flags" param.
11915         (dump_generic_expr_loc): Use dump_flags_t rather than int for
11916         "dump_kind" param.
11917         (dump_dec): Likewise.
11918         (dump_finish): Use TDF_NONE rather than 0.
11919         (gcc::dump_manager::opt_info_enable_passes): Use optgroup_flags_t
11920         rather than int for "optgroup_flags" param.  Use TDF_NONE rather
11921         than 0.  Update for change to option_ptr.
11922         (opt_info_switch_p_1): Convert "optgroup_flags" param from int *
11923         to optgroup_flags_t *.  Use TDF_NONE and OPTGROUP_NONE rather than
11924         0.  Update for changes to optinfo_verbosity_options and
11925         optgroup_options.
11926         (opt_info_switch_p): Convert optgroup_flags from int to
11927         optgroup_flags_t.
11928         (dump_basic_block): Use dump_flags_t rather than int
11929         for "dump_kind" param.
11930         * dumpfile.h (TDF_ADDRESS, TDF_SLIM, TDF_RAW, TDF_DETAILS,
11931         TDF_STATS, TDF_BLOCKS, TDF_VOPS, TDF_LINENO, TDF_UID)
11932         TDF_STMTADDR, TDF_GRAPH, TDF_MEMSYMS, TDF_RHS_ONLY, TDF_ASMNAME,
11933         TDF_EH, TDF_NOUID, TDF_ALIAS, TDF_ENUMERATE_LOCALS, TDF_CSELIB,
11934         TDF_SCEV, TDF_GIMPLE, TDF_FOLDING, MSG_OPTIMIZED_LOCATIONS,
11935         MSG_MISSED_OPTIMIZATION, MSG_NOTE, MSG_ALL, TDF_COMPARE_DEBUG,
11936         TDF_NONE): Convert from macros to...
11937         (enum dump_flag): ...this new enum.
11938         (dump_flags_t): Update to use enum.
11939         (operator|, operator&, operator~, operator|=, operator&=):
11940         Implement for dump_flags_t.
11941         (OPTGROUP_NONE, OPTGROUP_IPA, OPTGROUP_LOOP, OPTGROUP_INLINE,
11942         OPTGROUP_OMP, OPTGROUP_VEC, OPTGROUP_OTHER, OPTGROUP_ALL):
11943         Convert from macros to...
11944         (enum optgroup_flag): ...this new enum.
11945         (optgroup_flags_t): New typedef.
11946         (operator|, operator|=): Implement for optgroup_flags_t.
11947         (struct dump_file_info): Convert field "alt_flags" to
11948         dump_flags_t.  Convert field "optgroup_flags" to
11949         optgroup_flags_t.
11950         (dump_basic_block): Use dump_flags_t rather than int for param.
11951         (dump_generic_expr_loc): Likewise.
11952         (dump_dec): Likewise.
11953         (dump_register): Convert param "optgroup_flags" to
11954         optgroup_flags_t.
11955         (opt_info_enable_passes): Likewise.
11956         * early-remat.c (early_remat::dump_edge_list): Use TDF_NONE rather
11957         than 0.
11958         * gimple-pretty-print.c (debug): Likewise.
11959         * gimple-ssa-store-merging.c (bswap_replace): Likewise.
11960         (merged_store_group::apply_stores): Likewise.
11961         * gimple-ssa-strength-reduction.c (insert_initializers): Likewise.
11962         * gimple.c (verify_gimple_pp): Likewise.
11963         * graphite-poly.c (print_pbb_body): Likewise.
11964         * passes.c (pass_manager::register_one_dump_file): Convert
11965         local "optgroup_flags" to optgroup_flags_t.
11966         * print-tree.c (print_node): Use TDF_NONE rather than 0.
11967         (debug): Likewise.
11968         (debug_body): Likewise.
11969         * tree-pass.h (struct pass_data): Convert field "optgroup_flags"
11970         to optgroup_flags_t.
11971         * tree-pretty-print.c (print_struct_decl): Use TDF_NONE rather
11972         than 0.
11973         * tree-ssa-math-opts.c (convert_mult_to_fma_1): Likewise.
11974         (convert_mult_to_fma): Likewise.
11975         * tree-ssa-reassoc.c (undistribute_ops_list): Likewise.
11976         * tree-ssa-sccvn.c (vn_eliminate): Likewise.
11977         * tree-vect-data-refs.c (dump_lower_bound): Convert param
11978         "dump_kind" to dump_flags_t.
11980 2018-06-08  Segher Boessenkool  <segher@kernel.crashing.org>
11982         * config/rs6000/rs6000.c (min, max): Delete.
11984 2018-06-08  Segher Boessenkool  <segher@kernel.crashing.org>
11986         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete mention of
11987         -mabi=spe and -mabi=no-spe.
11989 2018-06-08  Martin Liska  <mliska@suse.cz>
11991         * ipa-pure-const.c (propagate_pure_const): Use ::get at places
11992         where we expect an existing summary.
11994 2018-06-08  Martin Liska  <mliska@suse.cz>
11996         * ipa-inline-analysis.c (simple_edge_hints): Use ::get method.
11997         * ipa-inline.h (estimate_edge_growth): Likewise.
11999 2018-06-08  Martin Liska  <mliska@suse.cz>
12001         * cgraph.c (function_version_hasher::hash): Use
12002         cgraph_node::get_uid ().
12003         (function_version_hasher::equal):
12004         * cgraph.h (cgraph_node::get_uid): New method.
12005         * ipa-inline.c (update_caller_keys): Use
12006         cgraph_node::get_uid ().
12007         (update_callee_keys): Likewise.
12008         * ipa-utils.c (searchc): Likewise.
12009         (ipa_reduced_postorder): Likewise.
12010         * lto-cgraph.c (input_node): Likewise.
12011         * passes.c (is_pass_explicitly_enabled_or_disabled): Likewise.
12012         * symbol-summary.h (symtab_insertion): Likewise.
12013         (symtab_removal): Likewise.
12014         (symtab_duplication): Likewise.
12015         * tree-pretty-print.c (dump_function_header): Likewise.
12016         * tree-sra.c (convert_callers_for_node): Likewise.
12018 2018-06-08  Martin Liska  <mliska@suse.cz>
12020         * cgraph.c (symbol_table::create_edge): Always assign a new
12021         unique number.
12022         (symbol_table::free_edge): Do not recycle numbers.
12023         * cgraph.h (cgraph_edge::get): New method.
12024         * symbol-summary.h (symtab_removal): Use it.
12025         (symtab_duplication): Likewise.
12026         (call_summary::hashable_uid): Remove.
12028 2018-06-08  Martin Liska  <mliska@suse.cz>
12030         * ipa-inline-analysis.c (inline_edge_removal_hook): Remove.
12031         (initialize_growth_caches): Remove.
12032         (free_growth_caches): Likewise.
12033         (do_estimate_edge_time): Use edge_growth_cache.
12034         (do_estimate_edge_size): Likewise.
12035         (do_estimate_edge_hints): Likewise.
12036         * ipa-inline.c (reset_edge_caches): Likewise.
12037         (recursive_inlining): Likewise.
12038         (inline_small_functions): Likewise.
12039         * ipa-inline.h (initialize_growth_caches): Remove.
12040         (estimate_edge_size): Likewise.
12041         (estimate_edge_time): Likewise.
12042         (estimate_edge_hints): Likewise.
12043         (reset_edge_growth_cache): Likewise.
12044         * symbol-summary.h (call_summary::remove): New method.
12046 2018-06-08  Martin Liska  <mliska@suse.cz>
12048         * ipa-cp.c (class edge_clone_summary): New summary.
12049         (grow_edge_clone_vectors): Remove.
12050         (ipcp_edge_duplication_hook): Remove.
12051         (class edge_clone_summary_t): New call_summary class.
12052         (ipcp_edge_removal_hook): Remove.
12053         (edge_clone_summary_t::duplicate): New function.
12054         (get_next_cgraph_edge_clone): Use edge_clone_summaries.
12055         (create_specialized_node): Likewise.
12056         (ipcp_driver): Initialize edge_clone_summaries and do not
12057         register hooks.
12059 2018-06-08  Martin Liska  <mliska@suse.cz>
12061         * symbol-summary.h (get): New function.
12062         (call_summary::m_initialize_when_cloning): New class member.
12064 2018-06-08  Martin Liska  <mliska@suse.cz>
12066         * cgraph.c (cgraph_node::remove): Do not recycle uid.
12067         * cgraph.h (symbol_table::release_symbol): Do not pass uid.
12068         (symbol_table::allocate_cgraph_symbol): Do not set uid.
12069         * passes.c (uid_hash_t): Record removed_nodes by their uids.
12070         (remove_cgraph_node_from_order): Use the removed_nodes set.
12071         (do_per_function_toporder): Likwise.
12072         * symbol-summary.h (symtab_insertion): Use cgraph_node::uid
12073         instead of summary_uid.
12074         (symtab_removal): Likewise.
12075         (symtab_duplication): Likewise.
12077 2018-06-08  Martin Liska  <mliska@suse.cz>
12079         * ipa-cp.c (ipcp_store_bits_results): Use
12080         ipcp_transformation_sum.
12081         (ipcp_store_vr_results): Likewise.
12082         * ipa-prop.c (ipcp_grow_transformations_if_necessary): Renamed
12083         to ...
12084         (ipcp_transformation_initialize): ... this.
12085         (ipa_set_node_agg_value_chain):
12086         (ipa_node_params_t::duplicate): Use ipcp_transformation_sum.
12087         (write_ipcp_transformation_info): Likewise.
12088         (read_ipcp_transformation_info): Likewise.
12089         (ipcp_update_bits): Likewise.
12090         (ipcp_update_vr): Likewise.
12091         (ipcp_transform_function): Likewise.
12092         * ipa-prop.h: Rename ipcp_transformation_summary to
12093         ipcp_transformation.
12094         (class ipcp_transformation_t): New function summary.
12095         (ipcp_get_transformation_summary): Use ipcp_transformation_sum.
12096         (ipa_get_agg_replacements_for_node): Likewise.
12098 2018-06-08  Martin Liska  <mliska@suse.cz>
12100         * ipa-pure-const.c (struct funct_state_d): Do it class instead
12101         of struct.
12102         (class funct_state_summary_t): New function_summary class.
12103         (has_function_state): Remove.
12104         (get_function_state): Likewise.
12105         (set_function_state): Likewise.
12106         (add_new_function): Likewise.
12107         (funct_state_summary_t::insert): New function.
12108         (duplicate_node_data): Remove.
12109         (remove_node_data): Remove.
12110         (funct_state_summary_t::duplicate): New function.
12111         (register_hooks): Create new funct_state_summaries.
12112         (pure_const_generate_summary): Use it.
12113         (pure_const_write_summary): Likewise.
12114         (pure_const_read_summary): Likewise.
12115         (propagate_pure_const): Likewise.
12116         (propagate_nothrow): Likewise.
12117         (dump_malloc_lattice): Likewise.
12118         (propagate_malloc): Likewise.
12119         (execute): Do not register hooks, just remove summary
12120         instead.
12121         (pass_ipa_pure_const::pass_ipa_pure_const): Simplify
12122         constructor.
12124 2018-06-08  Martin Liska  <mliska@suse.cz>
12126         * ipa-reference.c (remove_node_data): Remove.
12127         (duplicate_node_data): Likewise.
12128         (class ipa_ref_var_info_summary_t): New class.
12129         (class ipa_ref_opt_summary_t): Likewise.
12130         (get_reference_vars_info): Use ipa_ref_var_info_summaries.
12131         (get_reference_optimization_summary): Use
12132         ipa_ref_opt_sum_summaries.
12133         (set_reference_vars_info): Remove.
12134         (set_reference_optimization_summary): Likewise.
12135         (ipa_init): Create summaries.
12136         (init_function_info): Use function summary.
12137         (ipa_ref_opt_summary_t::duplicate): New function.
12138         (ipa_ref_opt_summary_t::remove): New function.
12139         (get_read_write_all_from_node): Fix GNU coding style.
12140         (propagate): Use function summary.
12141         (write_node_summary_p): Fix GNU coding style.
12142         (stream_out_bitmap): Likewise.
12143         (ipa_reference_read_optimization_summary): Use function summary.
12144         (ipa_reference_c_finalize): Do not release hooks.
12146 2018-06-08  Martin Liska  <mliska@suse.cz>
12148         * ipa-fnsummary.c (dump_ipa_call_summary): Use ::get method.
12149         (analyze_function_body): Extract multiple calls of get_create.
12150         * ipa-inline-analysis.c (simple_edge_hints): Likewise.
12151         * ipa-inline.c (recursive_inlining): Use ::get method.
12152         * ipa-inline.h (estimate_edge_growth): Likewise.
12154 2018-06-08  Martin Liska  <mliska@suse.cz>
12156         * hsa-common.h (enum hsa_function_kind): Rename HSA_NONE to
12157         HSA_INVALID.
12158         (hsa_function_summary::hsa_function_summary): Use the new enum
12159         value.
12160         (hsa_gpu_implementation_p): Use hsa_summaries::get.
12161         * hsa-gen.c (hsa_get_host_function): Likewise.
12162         (get_brig_function_name): Likewise.
12163         * ipa-hsa.c (process_hsa_functions): Likewise.
12164         (ipa_hsa_write_summary): Likewise.
12165         * symbol-summary.h (symtab_duplication): Use ::get function/
12166         (get): New function.
12168 2018-06-08  Martin Liska  <mliska@suse.cz>
12170         * config/i386/i386.c (ix86_can_inline_p): Use get_create instead
12171         of get.
12172         * hsa-common.c (hsa_summary_t::link_functions): Likewise.
12173         (hsa_register_kernel): Likewise.
12174         * hsa-common.h (hsa_gpu_implementation_p): Likewise.
12175         * hsa-gen.c (hsa_get_host_function): Likewise.
12176         (get_brig_function_name): Likewise.
12177         (generate_hsa): Likewise.
12178         (pass_gen_hsail::execute): Likewise.
12179         * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
12180         (devirtualization_time_bonus): Likewise.
12181         (ipcp_propagate_stage): Likewise.
12182         * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
12183         (edge_set_predicate): Likewise.
12184         (evaluate_conditions_for_known_args): Likewise.
12185         (evaluate_properties_for_edge): Likewise.
12186         (ipa_fn_summary::reset): Likewise.
12187         (ipa_fn_summary_t::duplicate): Likewise.
12188         (dump_ipa_call_summary): Likewise.
12189         (ipa_dump_fn_summary): Likewise.
12190         (analyze_function_body): Likewise.
12191         (compute_fn_summary): Likewise.
12192         (estimate_edge_devirt_benefit): Likewise.
12193         (estimate_edge_size_and_time): Likewise.
12194         (estimate_calls_size_and_time): Likewise.
12195         (estimate_node_size_and_time): Likewise.
12196         (inline_update_callee_summaries): Likewise.
12197         (remap_edge_change_prob): Likewise.
12198         (remap_edge_summaries): Likewise.
12199         (ipa_merge_fn_summary_after_inlining): Likewise.
12200         (ipa_update_overall_fn_summary): Likewise.
12201         (read_ipa_call_summary): Likewise.
12202         (inline_read_section): Likewise.
12203         (write_ipa_call_summary): Likewise.
12204         (ipa_fn_summary_write): Likewise.
12205         (ipa_free_fn_summary): Likewise.
12206         * ipa-hsa.c (process_hsa_functions): Likewise.
12207         (ipa_hsa_write_summary): Likewise.
12208         (ipa_hsa_read_section): Likewise.
12209         * ipa-icf.c (sem_function::merge): Likewise.
12210         * ipa-inline-analysis.c (simple_edge_hints): Likewise.
12211         (do_estimate_edge_time): Likewise.
12212         (estimate_size_after_inlining): Likewise.
12213         (estimate_growth): Likewise.
12214         (growth_likely_positive): Likewise.
12215         * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
12216         (inline_call): Likewise.
12217         * ipa-inline.c (caller_growth_limits): Likewise.
12218         (can_inline_edge_p): Likewise.
12219         (can_inline_edge_by_limits_p): Likewise.
12220         (compute_uninlined_call_time): Likewise.
12221         (compute_inlined_call_time): Likewise.
12222         (want_inline_small_function_p): Likewise.
12223         (edge_badness): Likewise.
12224         (update_caller_keys): Likewise.
12225         (update_callee_keys): Likewise.
12226         (recursive_inlining): Likewise.
12227         (inline_small_functions): Likewise.
12228         (inline_to_all_callers_1): Likewise.
12229         (dump_overall_stats): Likewise.
12230         (early_inline_small_functions): Likewise.
12231         (early_inliner): Likewise.
12232         * ipa-inline.h (estimate_edge_growth): Likewise.
12233         * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
12234         * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
12235         * ipa-prop.h (IPA_NODE_REF): Likewise.
12236         (IPA_EDGE_REF): Likewise.
12237         * ipa-pure-const.c (malloc_candidate_p): Likewise.
12238         (propagate_malloc): Likewise.
12239         * ipa-split.c (execute_split_functions): Likewise.
12240         * symbol-summary.h: Rename get to get_create.
12241         (get): Likewise.
12242         (get_create): Likewise.
12243         * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
12245 2018-06-08  Martin Liska  <mliska@suse.cz>
12247         * symbol-summary.h (release): Move definition out of class
12248         declaration.
12249         (symtab_removal): Likewise.
12250         (symtab_duplication): Likewise.
12252 2018-06-08  Martin Liska  <mliska@suse.cz>
12254         * symbol-summary.h (function_summary): Move constructor
12255         implementation out of class declaration.
12256         (release): Likewise.
12257         (symtab_insertion): Likewise.
12258         (symtab_removal): Likewise.
12259         (symtab_duplication): Likewise.
12260         (get): Likewise.
12262 2018-06-08  Martin Liska  <mliska@suse.cz>
12264         * Makefile.in: Remove support for MPX (macros, related functions,
12265         fields in cgraph_node, ...).
12266         * builtin-types.def (BT_BND): Likewise.
12267         (BT_FN_BND_CONST_PTR): Likewise.
12268         (BT_FN_CONST_PTR_BND): Likewise.
12269         (BT_FN_VOID_PTR_BND): Likewise.
12270         (BT_FN_BND_CONST_PTR_SIZE): Likewise.
12271         (BT_FN_VOID_CONST_PTR_BND_CONST_PTR): Likewise.
12272         * builtins.c (expand_builtin_memcpy_with_bounds): Likewise.
12273         (expand_builtin_mempcpy_with_bounds): Likewise.
12274         (expand_builtin_memset_with_bounds): Likewise.
12275         (expand_builtin_memset_args): Likewise.
12276         (std_expand_builtin_va_start): Likewise.
12277         (expand_builtin): Likewise.
12278         (expand_builtin_with_bounds): Likewise.
12279         * builtins.def (DEF_BUILTIN_CHKP): Likewise.
12280         (DEF_LIB_BUILTIN_CHKP): Likewise.
12281         (DEF_EXT_LIB_BUILTIN_CHKP): Likewise.
12282         (DEF_CHKP_BUILTIN): Likewise.
12283         (BUILT_IN_MEMCPY): Likewise.
12284         (BUILT_IN_MEMMOVE): Likewise.
12285         (BUILT_IN_MEMPCPY): Likewise.
12286         (BUILT_IN_MEMSET): Likewise.
12287         (BUILT_IN_STPCPY): Likewise.
12288         (BUILT_IN_STRCAT): Likewise.
12289         (BUILT_IN_STRCHR): Likewise.
12290         (BUILT_IN_STRCPY): Likewise.
12291         (BUILT_IN_STRLEN): Likewise.
12292         (BUILT_IN_MEMCPY_CHK): Likewise.
12293         (BUILT_IN_MEMMOVE_CHK): Likewise.
12294         (BUILT_IN_MEMPCPY_CHK): Likewise.
12295         (BUILT_IN_MEMSET_CHK): Likewise.
12296         (BUILT_IN_STPCPY_CHK): Likewise.
12297         (BUILT_IN_STRCAT_CHK): Likewise.
12298         (BUILT_IN_STRCPY_CHK): Likewise.
12299         * calls.c (store_bounds): Likewise.
12300         (emit_call_1): Likewise.
12301         (special_function_p): Likewise.
12302         (maybe_warn_nonstring_arg): Likewise.
12303         (initialize_argument_information): Likewise.
12304         (finalize_must_preallocate): Likewise.
12305         (compute_argument_addresses): Likewise.
12306         (expand_call): Likewise.
12307         * cfgexpand.c (expand_call_stmt): Likewise.
12308         (expand_return): Likewise.
12309         (expand_gimple_stmt_1): Likewise.
12310         (pass_expand::execute): Likewise.
12311         * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
12312         (cgraph_node::remove): Likewise.
12313         (cgraph_node::dump): Likewise.
12314         (cgraph_node::verify_node): Likewise.
12315         * cgraph.h (chkp_function_instrumented_p): Likewise.
12316         (symtab_node::get_alias_target): Likewise.
12317         (cgraph_node::can_remove_if_no_direct_calls_and_refs_p): Likewise.
12318         (cgraph_local_p): Likewise.
12319         * cgraphbuild.c (cgraph_edge::rebuild_edges): Likewise.
12320         (cgraph_edge::rebuild_references): Likewise.
12321         * cgraphunit.c (varpool_node::finalize_decl): Likewise.
12322         (walk_polymorphic_call_targets): Likewise.
12323         (cgraph_node::expand_thunk): Likewise.
12324         (symbol_table::output_weakrefs): Likewise.
12325         * common/config/i386/i386-common.c (OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET): Likewise.
12326         (ix86_handle_option): Likewise.
12327         * config/i386/constraints.md: Likewise.
12328         * config/i386/i386-builtin-types.def (BND): Likewise.
12329         (VOID): Likewise.
12330         (PVOID): Likewise.
12331         (ULONG): Likewise.
12332         * config/i386/i386-builtin.def (BDESC_END): Likewise.
12333         (BDESC_FIRST): Likewise.
12334         (BDESC): Likewise.
12335         * config/i386/i386-c.c (ix86_target_macros_internal): Likewise.
12336         * config/i386/i386-protos.h (ix86_bnd_prefixed_insn_p): Likewise.
12337         * config/i386/i386.c (enum reg_class): Likewise.
12338         (ix86_target_string): Likewise.
12339         (ix86_option_override_internal): Likewise.
12340         (ix86_conditional_register_usage): Likewise.
12341         (ix86_valid_target_attribute_inner_p): Likewise.
12342         (ix86_set_indirect_branch_type): Likewise.
12343         (ix86_set_current_function): Likewise.
12344         (ix86_function_arg_regno_p): Likewise.
12345         (init_cumulative_args): Likewise.
12346         (ix86_function_arg_advance): Likewise.
12347         (ix86_function_arg): Likewise.
12348         (ix86_pass_by_reference): Likewise.
12349         (ix86_function_value_regno_p): Likewise.
12350         (ix86_function_value_1): Likewise.
12351         (ix86_function_value_bounds): Likewise.
12352         (ix86_return_in_memory): Likewise.
12353         (ix86_setup_incoming_vararg_bounds): Likewise.
12354         (ix86_va_start): Likewise.
12355         (indirect_thunk_need_prefix): Likewise.
12356         (print_reg): Likewise.
12357         (ix86_print_operand): Likewise.
12358         (ix86_expand_call): Likewise.
12359         (ix86_output_function_return): Likewise.
12360         (reg_encoded_number): Likewise.
12361         (BDESC_VERIFYS): Likewise.
12362         (ix86_init_mpx_builtins): Likewise.
12363         (ix86_init_builtins): Likewise.
12364         (ix86_emit_cmove): Likewise.
12365         (ix86_emit_move_max): Likewise.
12366         (ix86_expand_builtin): Likewise.
12367         (ix86_builtin_mpx_function): Likewise.
12368         (ix86_get_arg_address_for_bt): Likewise.
12369         (ix86_load_bounds): Likewise.
12370         (ix86_store_bounds): Likewise.
12371         (ix86_load_returned_bounds): Likewise.
12372         (ix86_store_returned_bounds): Likewise.
12373         (ix86_class_likely_spilled_p): Likewise.
12374         (ix86_hard_regno_mode_ok): Likewise.
12375         (x86_order_regs_for_local_alloc): Likewise.
12376         (ix86_mitigate_rop): Likewise.
12377         (ix86_bnd_prefixed_insn_p): Likewise.
12378         (ix86_mpx_bound_mode): Likewise.
12379         (ix86_make_bounds_constant): Likewise.
12380         (ix86_initialize_bounds): Likewise.
12381         (TARGET_LOAD_BOUNDS_FOR_ARG): Likewise.
12382         (TARGET_STORE_BOUNDS_FOR_ARG): Likewise.
12383         (TARGET_LOAD_RETURNED_BOUNDS): Likewise.
12384         (TARGET_STORE_RETURNED_BOUNDS): Likewise.
12385         (TARGET_CHKP_BOUND_MODE): Likewise.
12386         (TARGET_BUILTIN_CHKP_FUNCTION): Likewise.
12387         (TARGET_CHKP_FUNCTION_VALUE_BOUNDS): Likewise.
12388         (TARGET_CHKP_MAKE_BOUNDS_CONSTANT): Likewise.
12389         (TARGET_CHKP_INITIALIZE_BOUNDS): Likewise.
12390         * config/i386/i386.h (TARGET_MPX): Likewise.
12391         (TARGET_MPX_P): Likewise.
12392         (VALID_BND_REG_MODE): Likewise.
12393         (FIRST_BND_REG): Likewise.
12394         (LAST_BND_REG): Likewise.
12395         (enum reg_class): Likewise.
12396         (BND_REG_P): Likewise.
12397         (BND_REGNO_P): Likewise.
12398         (BNDmode): Likewise.
12399         (ADJUST_INSN_LENGTH): Likewise.
12400         * config/i386/i386.md: Likewise.
12401         * config/i386/i386.opt: Likewise.
12402         * config/i386/linux-common.h (LIBMPX_LIBS): Likewise.
12403         (defined): Likewise.
12404         (LINK_MPX): Likewise.
12405         (MPX_SPEC): Likewise.
12406         (LIBMPX_SPEC): Likewise.
12407         (LIBMPXWRAPPERS_SPEC): Likewise.
12408         (CHKP_SPEC): Likewise.
12409         * config/i386/predicates.md: Likewise.
12410         * dbxout.c (dbxout_type): Likewise.
12411         * doc/extend.texi: Likewise.
12412         * doc/invoke.texi: Likewise.
12413         * doc/md.texi: Likewise.
12414         * doc/tm.texi: Likewise.
12415         * doc/tm.texi.in: Likewise.
12416         * dwarf2out.c (is_base_type): Likewise.
12417         (gen_formal_types_die): Likewise.
12418         (gen_subprogram_die): Likewise.
12419         (gen_type_die_with_usage): Likewise.
12420         (gen_decl_die): Likewise.
12421         (dwarf2out_late_global_decl): Likewise.
12422         * expr.c (expand_assignment): Likewise.
12423         (emit_storent_insn): Likewise.
12424         (store_expr_with_bounds): Likewise.
12425         (store_expr): Likewise.
12426         (expand_expr_real_1): Likewise.
12427         * expr.h (store_expr_with_bounds): Likewise.
12428         * function.c (use_register_for_decl): Likewise.
12429         (struct bounds_parm_data): Likewise.
12430         (assign_parms_augmented_arg_list): Likewise.
12431         (assign_parm_find_entry_rtl): Likewise.
12432         (assign_parm_is_stack_parm): Likewise.
12433         (assign_parm_load_bounds): Likewise.
12434         (assign_bounds): Likewise.
12435         (assign_parms): Likewise.
12436         (expand_function_start): Likewise.
12437         * gcc.c (CHKP_SPEC): Likewise.
12438         * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
12439         * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Likewise.
12440         (wrestrict_dom_walker::check_call): Likewise.
12441         * gimple.c (gimple_build_call_from_tree): Likewise.
12442         * gimple.h (enum gf_mask): Likewise.
12443         (gimple_call_with_bounds_p): Likewise.
12444         (gimple_call_set_with_bounds): Likewise.
12445         * gimplify.c (gimplify_init_constructor): Likewise.
12446         * ipa-cp.c (initialize_node_lattices): Likewise.
12447         (propagate_constants_across_call): Likewise.
12448         (find_more_scalar_values_for_callers_subset): Likewise.
12449         * ipa-hsa.c (process_hsa_functions): Likewise.
12450         * ipa-icf-gimple.c (func_checker::compare_gimple_call): Likewise.
12451         * ipa-icf.c (sem_function::merge): Likewise.
12452         * ipa-inline.c (early_inliner): Likewise.
12453         * ipa-pure-const.c (warn_function_noreturn): Likewise.
12454         (warn_function_cold): Likewise.
12455         (propagate_pure_const): Likewise.
12456         * ipa-ref.h (enum GTY): Likewise.
12457         * ipa-split.c (find_retbnd): Likewise.
12458         (consider_split): Likewise.
12459         (split_function): Likewise.
12460         * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
12461         * ipa.c (walk_polymorphic_call_targets): Likewise.
12462         (symbol_table::remove_unreachable_nodes): Likewise.
12463         (process_references): Likewise.
12464         (cgraph_build_static_cdtor_1): Likewise.
12465         * lto-cgraph.c (lto_output_node): Likewise.
12466         (output_refs): Likewise.
12467         (compute_ltrans_boundary): Likewise.
12468         (input_overwrite_node): Likewise.
12469         (input_node): Likewise.
12470         (input_cgraph_1): Likewise.
12471         * params.def (PARAM_CHKP_MAX_CTOR_SIZE): Likewise.
12472         * passes.c (pass_manager::execute_early_local_passes): Likewise.
12473         (class pass_chkp_instrumentation_passes): Likewise.
12474         (make_pass_chkp_instrumentation_passes): Likewise.
12475         * passes.def: Likewise.
12476         * rtl.h (struct GTY): Likewise.
12477         (CALL_EXPR_WITH_BOUNDS_P): Likewise.
12478         * stor-layout.c (layout_type): Likewise.
12479         * symtab.c: Likewise.
12480         * target.def: Likewise.
12481         * targhooks.c (default_chkp_bound_type): Likewise.
12482         (default_chkp_bound_mode): Likewise.
12483         (default_builtin_chkp_function): Likewise.
12484         (default_chkp_function_value_bounds): Likewise.
12485         (default_chkp_make_bounds_constant): Likewise.
12486         (default_chkp_initialize_bounds): Likewise.
12487         * targhooks.h (default_chkp_bound_type): Likewise.
12488         (default_chkp_bound_mode): Likewise.
12489         (default_builtin_chkp_function): Likewise.
12490         (default_chkp_function_value_bounds): Likewise.
12491         (default_chkp_make_bounds_constant): Likewise.
12492         (default_chkp_initialize_bounds): Likewise.
12493         * toplev.c (compile_file): Likewise.
12494         (process_options): Likewise.
12495         * tree-core.h (DEF_BUILTIN): Likewise.
12496         (DEF_BUILTIN_CHKP): Likewise.
12497         * tree-inline.c (declare_return_variable): Likewise.
12498         (remap_gimple_stmt): Likewise.
12499         (copy_bb): Likewise.
12500         (initialize_inlined_parameters): Likewise.
12501         (expand_call_inline): Likewise.
12502         * tree-pass.h (make_pass_ipa_chkp_versioning): Likewise.
12503         (make_pass_ipa_chkp_early_produce_thunks): Likewise.
12504         (make_pass_ipa_chkp_produce_thunks): Likewise.
12505         (make_pass_chkp): Likewise.
12506         (make_pass_chkp_opt): Likewise.
12507         (make_pass_chkp_instrumentation_passes): Likewise.
12508         * tree-pretty-print.c (dump_generic_node): Likewise.
12509         * tree-ssa-ccp.c (insert_clobber_before_stack_restore): Likewise.
12510         * tree-ssa-dce.c (propagate_necessity): Likewise.
12511         (eliminate_unnecessary_stmts): Likewise.
12512         * tree-ssa-pre.c (create_expression_by_pieces): Likewise.
12513         * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
12514         * tree-ssa-sccvn.h: Likewise.
12515         * tree-ssa-strlen.c (get_string_length): Likewise.
12516         (valid_builtin_call): Likewise.
12517         (adjust_last_stmt): Likewise.
12518         (handle_builtin_strchr): Likewise.
12519         (handle_builtin_strcpy): Likewise.
12520         (handle_builtin_stxncpy): Likewise.
12521         (handle_builtin_memcpy): Likewise.
12522         (handle_builtin_strcat): Likewise.
12523         (strlen_check_and_optimize_stmt): Likewise.
12524         * tree-stdarg.c (expand_ifn_va_arg_1): Likewise.
12525         * tree-streamer-in.c: Likewise.
12526         * tree-streamer.c (record_common_node): Likewise.
12527         * tree.c (tree_code_size): Likewise.
12528         (wide_int_to_tree_1): Likewise.
12529         (type_contains_placeholder_1): Likewise.
12530         (build_common_tree_nodes): Likewise.
12531         * tree.def (POINTER_BOUNDS_TYPE): Likewise.
12532         * tree.h (POINTER_BOUNDS_TYPE_P): Likewise.
12533         (POINTER_BOUNDS_P): Likewise.
12534         (BOUNDED_TYPE_P): Likewise.
12535         (BOUNDED_P): Likewise.
12536         (CALL_WITH_BOUNDS_P): Likewise.
12537         (pointer_bounds_type_node): Likewise.
12538         * value-prof.c (gimple_ic): Likewise.
12539         * var-tracking.c (vt_add_function_parameters): Likewise.
12540         * varasm.c (make_decl_rtl): Likewise.
12541         (assemble_start_function): Likewise.
12542         (output_constant): Likewise.
12543         (maybe_assemble_visibility): Likewise.
12544         * varpool.c (ctor_for_folding): Likewise.
12545         * chkp-builtins.def: Remove.
12546         * ipa-chkp.c: Remove.
12547         * ipa-chkp.h: Remove.
12548         * rtl-chkp.c: Remove.
12549         * rtl-chkp.h: Remove.
12550         * tree-chkp-opt.c: Remove.
12551         * tree-chkp.c: Remove.
12552         * tree-chkp.h: Remove.
12554 2018-06-07  Carl Love  <cel@us.ibm.com>
12556         * config/rs6000/vsx.md (vextract_fp_from_shorth,
12557         vextract_fp_from_shortl): Add BE support.
12559 2018-06-07  Paul Koning  <ni1d@arrl.net>
12561         * compare-elim.c (try_merge_compare): Don't merge compare if
12562         address contains a side effect.
12563         (try_eliminate_compare): Likewise.
12565 2018-06-07  Olga Makhotina  <olga.makhotina@intel.com>
12567         * config.gcc: Support "tremont".
12568         * config/i386/driver-i386.c (host_detect_local_cpu): Detect "tremont".
12569         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
12570         PROCESSOR_TREMONT.
12571         * config/i386/i386.c (m_TREMONT): Define.
12572         (processor_target_table): Add "tremont".
12573         (PTA_TREMONT): Define.
12574         (ix86_lea_outperforms): Add TARGET_TREMONT.
12575         (get_builtin_code_for_version): Handle PROCESSOR_TREMONT.
12576         (fold_builtin_cpu): Add M_INTEL_TREMONT, replace M_INTEL_GOLDMONT
12577         and M_INTEL_GOLDMONT_PLUS.
12578         (fold_builtin_cpu): Add "tremont".
12579         (ix86_add_stmt_cost): Add TARGET_TREMONT.
12580         (ix86_option_override_internal): Add "tremont".
12581         * config/i386/i386.h (processor_costs): Define TARGET_TREMONT.
12582         (processor_type): Add PROCESSOR_TREMONT.
12583         * config/i386/x86-tune.def: Add m_TREMONT.
12584         * doc/invoke.texi: Add tremont as x86 -march=/-mtune= CPU type.
12586 2018-06-07  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
12588         * config/msp430/msp430.c (msp430_mcu_name): Set the "i" in the
12589         symbol defined for msp430i* devices to be lower case.
12591 2018-06-07  Richard Biener  <rguenther@suse.de>
12593         * graphite-sese-to-poly.c (extract_affine): Avoid unneded
12594         wrapping.  Properly wrap the result of a BIT_NOT_EXPR.
12595         Properly wrap signed arithmetic if overflow wraps.
12597 2018-06-07  Jakub Jelinek  <jakub@redhat.com>
12599         PR tree-optimization/69615
12600         * tree-ssa-reassoc.c (optimize_range_tests_var_bound): If rhs2 is lhs
12601         of a cast from a same precision integral SSA_NAME in a bb dominated
12602         by first_bb, retry with rhs2 set to the rhs1 of the cast.  Don't emit
12603         cast to utype if rhs2 has already a compatible type.
12605 2018-06-07  Richard Biener  <rguenther@suse.de>
12607         PR tree-optimization/85935
12608         * graphite-scop-detection.c (find_params_in_bb): Analyze
12609         condition operands with respect to the correct loop.  Assert
12610         the analysis doesn't fail.
12612 2018-06-04  Carl Love  <cel@us.ibm.com>
12614         * config/rs6000/vsx.md (first_match_index_<mode>): Calculate index
12615         using natural element order.  Use gen_lshrsi3 instead of gen_ashrsi3
12616         as it is slightly cheaper.
12617         (first_match_or_eos_index_<mode>):
12618         Calculate index using natural element order.
12619         (first_match_index_<mode>):
12620         Calculate index using natural element order.
12621         (first_match_or_eos_index_<mode>):
12622         Calculate index using natural order.
12623         (define_insn vclzlsbb): Change to define_insn vclzlsbb_<mode>.
12624         for BE and LE modes.
12625         * config/rs6000/rs6000-c.c: Rename P9V_BUILTIN_VCLZLSBB,
12626         P9V_BUILTIN_VCLZLSBB_V16QI.
12627         * config/rs6000/rs6000-builtin.def: Make VCLZLSBB mode
12628         specific.
12630 2018-06-06  Kelvin Nilsen  <kelvin@gcc.gnu.org>
12632         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Adjust
12633         indentation and line wrap for many prototypes.  Add missing
12634         @smallexample directives around block of prototypes for vec_xl and
12635         vec_xst.
12637 2018-06-05  Michael Meissner  <meissner@linux.ibm.com>
12639         * config/rs6000/rs6000.c (rs6000_passes_ieee128): New boolean to
12640         track if we pass or return IEEE 128-bit floating point.
12641         (ieee128_mangling_gcc_8_1): New boolean to say whether to generate
12642         C++ mangling that is compatible with GCC 8.1.
12643         (TARGET_ASM_GLOBALIZE_DECL_NAME): Override target hook.
12644         (init_cumulative_args): Note if we pass or return IEEE 128-bit
12645         floating point types.
12646         (rs6000_function_arg_advance_1): Likewise.
12647         (rs6000_mangle_type): Optionally generate mangled names that match
12648         what GCC 8.1 generated for IEEE 128-bit floating point types.
12649         (rs6000_globalize_decl_name): If we have an external function that
12650         passes or returns IEEE 128-bit types, generate a weak reference
12651         from the mangled name used in GCC 8.1 to the current mangled
12652         name.
12653         (rs6000_init_builtins): Make __ibm128 use the long double type if
12654         long double is IBM extended double.  Make __float128 use the long
12655         double type if long double is IEEE 128-bit.
12657         PR target/85657
12658         * config/rs6000/rs6000-builtin.def (BU_IBM128_2): New helper
12659         macro for __ibm128 built-in functions.
12660         (PACK_IF): Add __ibm128 pack/unpack functions.
12661         (UNPACK_IF): Likewise.
12662         * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Do not
12663         enable long double built-in functions if long double is IEEE
12664         128-bit floating point.
12665         (rs6000_invalid_builtin): Update long double built-in function
12666         error message.
12667         (rs6000_expand_builtin): For PACK_IF and UNPACK_IF built-in
12668         functions, adjust the built-in function to use the long double
12669         built-in function if __ibm128 and long double are the same type.
12670         * doc/extend.texi (PowerPC builtins): Update documention for
12671         __builtin_{,un}pack_longdouble.  Add documentation for
12672         __builtin_{,un}pack_ibm128.
12674 2018-06-06  Jim Wilson  <jimw@sifive.com>
12676         * config/riscv/riscv.c (enum riscv_privilege_levels): New.
12677         (struct machine_function): New field interrupt_mode.
12678         (riscv_handle_type_attribute): New function.  Add forward declaration.
12679         (riscv_attribute_table) <interrupt>: Use riscv_handle_type_attribute.
12680         (riscv_expand_epilogue): Check interrupt_mode field.
12681         (riscv_set_current_function): Check interrupt attribute args and
12682         set interrupt_mode field.
12683         * config/riscv/riscv.md (UNSPECV_SRET, UNSPECV_URET): New.
12684         (riscv_sret, riscv_uret): New.
12685         * doc/extend.texi (RISC-V Function Attributes) <interrupt>: Document
12686         new arguments to interrupt attribute.
12688 2018-06-06  Peter Bergner  <bergner@vnet.ibm.com>
12690         PR target/63177
12691         * /config/rs6000/rs6000.h (ASM_CPU_SPEC): Add support for -mpower9.
12692         Don't handle -mcpu=power8 if -mpower9-vector is also used.
12694 2018-06-06  Kelvin Nilsen  <kelvin@gcc.gnu.org>
12696         * config/rs6000/rs6000-builtin.def (VSX_BUILTIN_VEC_LD,
12697         VSX_BUILTIN_VEC_ST): Add comment to explain non-traditional uses.
12698         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
12699         several redundant entries.
12701 2018-06-06  David Malcolm  <dmalcolm@redhat.com>
12703         * config/i386/i386-protos.h (ix86_expand_call): Strengthen return
12704         type from "rtx" to "rtx_insn *".
12705         * config/i386/i386.c (ix86_expand_split_stack_prologue): Likewise
12706         for local "call_insn", removing cast.
12707         (ix86_expand_call): Likewise, introducing a "call_insn" local.
12709 2018-06-06  Eric Botcazou  <ebotcazou@adacore.com>
12711         PR tree-optimization/86066
12712         * gimple-ssa-store-merging.c (process_store): Do not bypass BIT_NOT_EXPR
12713         for BIT_INSERT_EXPR stores.
12715 2018-06-06  Richard Biener  <rguenther@suse.de>
12717         PR tree-optimization/86062
12718         * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
12719         component refs ontop
12720         of to be offsetted base.
12722 2018-06-06  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
12724         * config/msp430/msp430.c (msp430_attr): Allow interrupt handlers
12725         to be static and remove check on interrupt attribute name.
12727 2018-06-05  Kelvin Nilsen  <kelvin@gcc.gnu.org>
12729         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
12730         volatile qualifier from vec_lvsl and vec_lvsr argument prototypes.
12732 2018-06-05  Steve Ellcey  <sellcey@cavium.com>
12734         PR target/79924
12735         * config/aarch64/aarch64-protos.h (aarch64_err_no_fpadvsimd): Remove
12736         second argument.
12737         * config/aarch64/aarch64-protos..c (aarch64_err_no_fpadvsimd):
12738         Remove second argument, change how error is called.
12739         (aarch64_layout_arg): Remove second argument from
12740         aarch64_err_no_fpadvsimd call.
12741         (aarch64_init_cumulative_args): Ditto.
12742         (aarch64_gimplify_va_arg_expr): Ditto.
12743         * config/aarch64/aarch64.md (mov<mode>): Ditto.
12745 2018-06-05  Uros Bizjak  <ubizjak@gmail.com>
12747         * config/i386/i386.md (simple_return_indirect_internal): New expander.
12748         (*simple_return_indirect_internal<mode>): Rename from
12749         simple_return_indirect_internal.  Use W mode iterator.
12750         (rstorssp): New expander.
12751         (*rstorssp<mode>): Rename from rstorssp.  Use P mode iterator.
12752         (clrssbsy): New expander.
12753         (*clrssbsy<mode>): Rename from clrssbsy.  Use P mode iterator.
12755 2018-06-05  Andre Vieira  <andre.simoesdiasvieira@arm.com>
12757         * config/arm/arm_cmse.h (cmse_nsfptr_create): Change typeof to
12758         __typeof__.
12759         (cmse_check_pointed_object): Likewise.
12761 2018-06-05  Martin Liska  <mliska@suse.cz>
12763         PR gcov-profile/47618
12764         * doc/invoke.texi: Document how -fprofile-dir format
12765         is extended.
12767 2018-06-05  Richard Biener  <rguenther@suse.de>
12769         * tree-cfgcleanup.c (cleanup_control_flow_pre): For edge
12770         removal pretend DOM info isn't available so we do not update
12771         it and only remove edges, not dominated blocks.  Actually free
12772         DOM info in case we removed something.  Remove unreachable blocks.
12773         (mfb_keep_latches): Work with either DOM info or marked backedges.
12774         (cleanup_tree_cfg_noloop): Do not remove unreachable blocks
12775         first.  Mark backedges if DOM info isn't available.
12776         (Re-)compute DOM info after cleanup_control_flow_pre.
12778 2018-06-05  Richard Biener  <rguenther@suse.de>
12780         * tree-cfg.c (struct locus_discrim_map): Store line, not location.
12781         (locus_discrim_hasher::hash): Adjust.
12782         (locus_discrim_hasher::equal): Likewise.
12783         (next_discriminator_for_locus): Work on line directly.
12784         (same_line_p): Pass in expanded locus1 as well.
12785         (assign_discriminators): Avoid redundant location expansions.
12787 2018-06-05  Richard Biener  <rguenther@suse.de>
12789         PR tree-optimization/86046
12790         * tree-ssa.c (maybe_optimize_var): Clear DECL_GIMPLE_REG_P
12791         if required after clearing TREE_ADDRESSABLE.
12793 2018-06-05  Richard Biener  <rguenther@suse.de>
12795         PR tree-optimization/86047
12796         * tree-ssa-loop.c (for_each_index): Glob handling of all
12797         decls and constants and really handle all of them.
12799 2018-06-05  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
12801         PR target/81497
12802         * config/arm/arm-builtins.c (arm_type_qualifiers): Add
12803         qualifier_void_pointer and qualifier_const_void_pointer.
12804         (arm_ldc_qualifiers, arm_stc_qualifiers): Use the above.
12805         (arm_init_builtins): Handle the above.
12806         * config/arm/arm_acle.h (__arm_cdp, __arm_ldc, __arm_ldcl, __arm_stc,
12807         __arm_stcl, __arm_mcr, __arm_cdp2, __arm_ldc2, __arm_ldcl2, __arm_stc2,
12808         __arm_stcl2,__arm_mcr2, __arm_mcrr, __arm_mcrr2): Remove return for
12809         void intrinsics.
12811 2018-06-05  Martin Liska  <mliska@suse.cz>
12813         * auto-profile.c (read_autofdo_file): Do not use
12814         gcov_ctr_summary struct.
12815         (afdo_callsite_hot_enough_for_early_inline): Likewise.
12816         * coverage.c (struct counts_entry): Likewise.
12817         (read_counts_file): Read just single summary entry.
12818         (get_coverage_counts): Use gcov_summary struct.
12819         * coverage.h (get_coverage_counts): Likewise.
12820         * gcov-dump.c (dump_working_sets): Likewise.
12821         (tag_summary): Dump just single summary.
12822         * gcov-io.c (gcov_write_summary): Write just histogram
12823         summary.
12824         (gcov_read_summary): Read just single summary.
12825         (compute_working_sets): Use gcov_summary struct.
12826         * gcov-io.h (GCOV_TAG_SUMMARY_LENGTH): Remove usage
12827         of GCOV_COUNTERS_SUMMABLE.
12828         (GCOV_COUNTERS_SUMMABLE): Remove.
12829         (GCOV_FIRST_VALUE_COUNTER): Replace with
12830         GCOV_COUNTER_V_INTERVAL.
12831         (struct gcov_ctr_summary): Remove.
12832         (struct gcov_summary): Directly use fields of former
12833         gcov_ctr_summary.
12834         (compute_working_sets): Use gcov_summary struct.
12835         * gcov.c (read_count_file): Do not use ctrs fields.
12836         * lto-cgraph.c (merge_profile_summaries): Use gcov_summary
12837         struct.
12838         * lto-streamer.h (struct GTY): Make profile_info gcov_summary
12839         struct.
12840         * profile.c: Likewise.
12841         * profile.h: Likewise.
12843 2018-06-05  Martin Liska  <mliska@suse.cz>
12845         PR gcov-profile/84846
12846         * gcov.c (output_lines): Print working directory only
12847         in intermediate format.
12849 2018-06-05  Andreas Krebbel  <krebbel@linux.ibm.com>
12851         * config/s390/s390-builtin-types.def: Add void function type.
12852         * config/s390/s390-builtins.def: Use the function type for the
12853         tbeginc builtin.
12855 2018-06-04  Jim Wilson  <jimw@sifive.com>
12857         * config/riscv/riscv-protos.h (riscv_expand_epilogue): Change bool arg
12858         to int.
12859         * config/riscv/riscv.c (riscv_for_each_saved_reg): New args epilogue
12860         and maybe_eh_return.  Change regno to unsigned int.  Use new args to
12861         handle EH_RETURN_DATA_REGNO registers properly.
12862         (riscv_expand_prologue): Pass new args to riscv_for_each_saved_reg.
12863         (riscv_expand_epilogue): Update comment.  Change argument name and
12864         type.  Update code to use new name and type.  Pass new args to
12865         riscv_for_each_saved_reg.  Only use EH_RETURN_STACKADJ_RTX when
12866         EXCEPTION_RETURN.
12867         * config/riscv/riscv.md (NORMAL_RETURN): New.
12868         (SIBCALL_RETURN, EXCEPTION_RETURN): New.
12869         (epilogue, sibcall_epilogue): Update riscv_expand_epilogue arg.
12870         (eh_return): Call gen_eh_return_internal and emit barrier.
12871         (eh_return_internal): Call riscv_expand_epilogue.
12873 2018-06-04  Eric Botcazou  <ebotcazou@adacore.com>
12875         * gimple-ssa-store-merging.c (struct merged_store_group): Move up
12876         bit_insertion field and declare can_be_merged_into method.
12877         (merged_store_group::can_be_merged_into): New method.
12878         (imm_store_chain_info::coalesce_immediate): Call it to decide whether
12879         consecutive non-overlapping stores can be merged.  Turn MEM_REF stores
12880         into BIT_INSERT_EXPR stores if the group contains a non-MEM_REF store.
12882 2018-06-04  Richard Biener  <rguenther@suse.de>
12884         PR tree-optimization/85955
12885         * builtins.c (fold_builtin_sincos): Convert pointers to
12886         destination to appropriate type before dereferencing.
12888 2018-06-04  Segher Boessenkool  <segher@kernel.crashing.org>
12890         * config/rs6000/rs6000.md (abs<mode>2 for FLOAT128): Handle IFmode.
12892 2018-06-04  Richard Sandiford  <richard.sandiford@linaro.org>
12894         * expr.c (expand_expr_real_1): Force the operand into memory if
12895         its TYPE_MODE is BLKmode and if there is no integer mode for
12896         the number of bits being extracted.
12898 2018-06-04  Jakub Jelinek  <jakub@redhat.com>
12900         PR target/85832
12901         PR target/86036
12902         * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
12903         Use vptestnm rather than vptestm in (=Yc,v,C) variant.
12905 2018-06-04  Richard Biener  <rguenther@suse.de>
12907         * tree-cfgcleanup.c (cleanup_tree_cfg_1): Fold into...
12908         (cleanup_tree_cfg_noloop): ... single caller.  Do
12909         start_recording_case_labels later.
12911 2018-06-04  Sebastian Peryt  <sebastian.peryt@intel.com>
12913         * config/i386/cldemoteintrin.h: Change define from _X86INTRIN_H_INCLUDED
12914         to _IMMINTRIN_H_INCLUDED.
12915         * config/i386/pconfigintrin.h: Ditto.
12916         * config/i386/waitpkgintrin.h: Ditto.
12917         * config/i386/immintrin.h: Add includes for sgxintrin.h,
12918         pconfigintrin.h, waitpkgintrin.h and cldemoteintrin.h.
12919         * config/i386/x86intrin.h: Remove includes for mintrin.h, xmmintrin.h,
12920         emmintrin.h, pmmintrin.h, tmmintrin.h, smmintrin.h, wmmintrin.h,
12921         bmiintrin.h, bmi2intrin.h, lzcntintrin.h, sgxintrin.h, pconfigintrin.h,
12922         waitpkgintrin.h and cldemoteintrin.h.
12924 2018-06-04  Richard Biener  <rguenther@suse.de>
12926         PR tree-optimization/86038
12927         * tracer.c (find_best_successor): Check probability for
12928         being initialized, bail out if not.
12930 2018-06-04  Richard Earnshaw  <rearnsha@arm.com>
12932         PR target/86003
12933         * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
12934         of bits to ignore when comparing architectures.
12936 2018-06-04  Jakub Jelinek  <jakub@redhat.com>
12938         PR tree-optimization/69615
12939         * fold-const.c (merge_ranges): If range1 is - [x, x] and x is the
12940         maximum or minimum of the type, try to merge it also as if
12941         range1 is + [-, x - 1] or + [x + 1, -].
12943         PR c++/86025
12944         * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
12946 2018-06-03  Eric Botcazou  <ebotcazou@adacore.com>
12948         PR tree-optimization/86034
12949         * gimple-ssa-store-merging.c (output_merged_store): Convert the RHS to
12950         the unsigned bitfield type in a bit insertion sequence if it does not
12951         have a larger precision than the bitfield size.
12952         (process_store): Also bypass widening conversions for BIT_INSERT_EXPR.
12954 2018-06-03  Kito Cheng  <kito.cheng@gmail.com>
12956         * config/nds32/nds32-peephole2.md: Add new patterns for code size.
12958 2018-06-03  Chung-Ju Wu  <jasonwucj@gmail.com>
12960         * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3J.
12961         * config/nds32/nds32.c (nds32_option_override): Consider ARCH_V3J.
12962         * config/nds32/nds32.h (TARGET_ISA_V3): Add ARCH_V3J checking.
12963         * config/nds32/nds32.opt (march): Add enum value Value(ARCH_V3J).
12965 2018-06-03  Chung-Ju Wu  <jasonwucj@gmail.com>
12967         * common/config/nds32/nds32-common.c (nds32_option_optimization_table):
12968         Disable -fdelete-null-pointer-checks for ELF toolchain.
12970 2018-06-02  Chung-Ju Wu  <jasonwucj@gmail.com>
12971             Kito Cheng  <kito.cheng@gmail.com>
12973         * config.gcc (nds32*): Use nds32-linux.opt and nds32-elf.opt.
12974         (nds32le-*-*, nds32be-*-*): Integrate checking process.
12975         (nds32*-*-*): Add glibc and uclibc conditions.
12976         * common/config/nds32/nds32-common.c (nds32_except_unwind_info): New.
12977         (TARGET_EXCEPT_UNWIND_INFO): Define.
12978         * config/nds32/elf.h: New file.
12979         * config/nds32/linux.h: New file.
12980         * config/nds32/nds32-elf.opt: New file.
12981         * config/nds32/nds32-linux.opt: New file.
12982         * config/nds32/nds32-fp-as-gp.c
12983         (pass_nds32_fp_as_gp::gate): Consider TARGET_LINUX_ABI.
12984         * config/nds32/nds32.c (nds32_conditional_register_usage): Consider
12985         TARGET_LINUX_ABI.
12986         (nds32_asm_file_end): Ditto.
12987         (nds32_print_operand): Ditto.
12988         (nds32_insert_attributes): Ditto.
12989         (nds32_init_libfuncs): New function.
12990         (TARGET_HAVE_TLS): Define.
12991         (TARGET_INIT_LIBFUNCS): Define.
12992         * config/nds32/nds32.h (TARGET_DEFAULT_RELAX): Apply different relax
12993         spec content.
12994         (TARGET_ELF): Apply different mcmodel setting.
12995         (LINK_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): The content has
12996         been migrated into elf.h and linux.h files.
12997         * config/nds32/nds32.md (add_pc): Consider TARGET_LINUX_ABI.
12998         * config/nds32/nds32.opt (mvh): Consider TARGET_LINUX_ABI.
12999         (mcmodel): The content has been migrated into nds32-elf.opt and
13000         nds32-linux.opt files.
13001         * config/nds32/t-elf: New file.
13002         * config/nds32/t-linux: New file.
13004 2018-06-02  Chung-Ju Wu  <jasonwucj@gmail.com>
13005             Shiva Chen  <shiva0217@gmail.com>
13007         * config/nds32/constants.md (unspec_volatile_element): Add
13008         UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END.
13009         * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp
13010         optimization.
13011         * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare.
13012         (make_pass_nds32_fp_as_gp): Declare.
13013         * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one
13014         optmization pass.
13015         (nds32_asm_function_end_prologue): Remove unused asm output.
13016         (nds32_asm_function_begin_epilogue): Remove unused asm output.
13017         (nds32_asm_file_start): Output necessary fp_as_gp information.
13018         (nds32_option_override): Adjust register usage.
13019         (nds32_expand_prologue): Consider fp_as_gp situation.
13020         (nds32_expand_prologue_v3push): Consider fp_as_gp situation.
13021         * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p.
13022         (epilogue): Ditto.
13023         (return): Ditto.
13024         (simple_return): Ditto.
13025         (omit_fp_begin): Output special directive for fp_as_gp.
13026         (omit_fp_end): Output special directive for fp_as_gp.
13027         * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp,
13028         mforbid-fp-as-gp): New options.
13030 2018-06-01  Mark Wielaard  <mark@klomp.org>
13032         * dwarf2out.c (dwarf2out_finish): Remove generation of
13033         DW_AT_loclists_base.
13035 2018-06-01  Eric Botcazou  <ebotcazou@adacore.com>
13037         * gimple-ssa-store-merging.c: Include gimple-fold.h.
13038         (struct store_immediate_info): Document BIT_INSERT_EXPR stores.
13039         (struct merged_store_group): Add bit_insertion field.
13040         (dump_char_array): Use standard hexadecimal format.
13041         (merged_store_group::merged_store_group): Set bit_insertion to false.
13042         (merged_store_group::apply_stores): Use optimal buffer size.  Deal
13043         with BIT_INSERT_EXPR stores.  Move up code updating the mask and
13044         also print the mask in the dump file.
13045         (pass_store_merging::gate): Minor tweak.
13046         (imm_store_chain_info::coalesce_immediate): Fix wrong association
13047         of stores with groups in dump.  Allow coalescing of BIT_INSERT_EXPR
13048         stores with INTEGER_CST stores.
13049         (count_multiple_uses) <BIT_INSERT_EXPR>: New case.
13050         (imm_store_chain_info::output_merged_store): Add try_bitpos variable
13051         and use it throughout.  Generate bit insertion sequences if need be.
13052         (pass_store_merging::process_store): Remove redundant condition.
13053         Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR.
13055 2018-06-01  Segher Boessenkool  <segher@kernel.crashing.org>
13057         * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
13058         the 128-bit floating point types.  Fix function comment.
13060 2018-06-01  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
13062         * config/aarch64/aarch64-simd.md
13063         (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler
13064         mnemonics.
13065         (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
13066         mnemonics.
13068 2018-06-01  Richard Sandiford  <richard.sandiford@linaro.org>
13070         PR tree-optimization/85989
13071         * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
13072         variable.
13073         (backprop::intersect_uses): Check it when deciding whether this
13074         is a backedge reference.
13075         (backprop::process_block): Add each phi to m_visited_phis
13076         after visiting it, then clear it at the end.
13078 2018-06-01  Richard Biener  <rguenther@suse.de>
13080         * tree-vectorizer.h (vect_dr_stmt): New function.
13081         (vect_get_load_cost): Adjust.
13082         (vect_get_store_cost): Likewise.
13083         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
13084         Use vect_dr_stmt instead of DR_SMTT.
13085         (vect_record_base_alignments): Likewise.
13086         (vect_calculate_target_alignment): Likewise.
13087         (vect_compute_data_ref_alignment): Likewise and make static.
13088         (vect_update_misalignment_for_peel): Likewise.
13089         (vect_verify_datarefs_alignment): Likewise.
13090         (vector_alignment_reachable_p): Likewise.
13091         (vect_get_data_access_cost): Likewise.  Pass down
13092         vinfo to vect_get_load_cost/vect_get_store_cost instead of DR.
13093         (vect_get_peeling_costs_all_drs): Likewise.
13094         (vect_peeling_hash_get_lowest_cost): Likewise.
13095         (vect_enhance_data_refs_alignment): Likewise.
13096         (vect_find_same_alignment_drs): Likewise.
13097         (vect_analyze_data_refs_alignment): Likewise.
13098         (vect_analyze_group_access_1): Likewise.
13099         (vect_analyze_group_access): Likewise.
13100         (vect_analyze_data_ref_access): Likewise.
13101         (vect_analyze_data_ref_accesses): Likewise.
13102         (vect_vfa_segment_size): Likewise.
13103         (vect_small_gap_p): Likewise.
13104         (vectorizable_with_step_bound_p): Likewise.
13105         (vect_prune_runtime_alias_test_list): Likewise.
13106         (vect_analyze_data_refs): Likewise.
13107         (vect_supportable_dr_alignment): Likewise.
13108         * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
13109         (vect_gen_prolog_loop_niters): Likewise.
13110         * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
13111         * tree-vect-patterns.c (vect_recog_bool_pattern): Do not
13112         modify DR_STMT.
13113         (vect_recog_mask_conversion_pattern): Likewise.
13114         (vect_try_gather_scatter_pattern): Likewise.
13115         * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info
13116         to vect_get_store_cost.
13117         (vect_get_store_cost): Get stmt_info instead of DR.
13118         (vect_model_load_cost): Pass stmt_info to vect_get_load_cost.
13119         (vect_get_load_cost): Get stmt_info instead of DR.
13121 2018-06-01  Richard Biener  <rguenther@suse.de>
13123         PR middle-end/86017
13124         * gimple-fold.c (var_decl_component_p): Also allow offsetted
13125         vars wrapped in MEM_REFs.
13127 2018-06-01  Richard Sandiford  <richard.sandiford@linaro.org>
13129         * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
13130         Fix subreg tests so that we only return a choice between
13131         GENERAL_REGS and FP_REGS if the original classes included both.
13133 2018-06-01  Richard Biener  <rguenther@suse.de>
13135         PR ipa/85960
13136         * tree-ssa-structalias.c (get_function_part_constraint):
13137         Handle NULL fi->decl.
13138         (find_func_aliases_for_call): Properly handle indirect
13139         fi from direct call.
13140         (find_func_clobbers): Likewise.
13141         (ipa_pta_execute): Likewise.
13142         (create_variable_info_for): For functions that are ifunc_resolver
13143         resolve to a varinfo that contains the result of the resolver call.
13144         (associate_varinfo_to_alias): Do not treat ifunc resolvers as
13145         aliases.
13147 2018-05-31  Michael Collison  <michael.collison@arm.com>
13149         * config/aarch64/aarch64.md:
13150         (*fix_to_zero_extenddfdi2): New pattern.
13151         * gcc.target/aarch64/fix_extend1.c: New testcase.
13153 2018-05-31  Qing Zhao <qing.zhao@oracle.com>
13155         PR middle-end/78809
13156         PR middle-end/83026
13157         * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
13158         and BUILT_IN_STRNCMP_EQ.
13159         * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
13160         BUILT_IN_STRNCMP_EQ.
13161         * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
13162         handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
13163         (gimple_fold_builtin): Likewise.
13164         * tree-ssa-strlen.c (compute_string_length): New function.
13165         (determine_min_obsize): New function.
13166         (handle_builtin_string_cmp): New function to handle calls to
13167         string compare functions.
13168         (strlen_optimize_stmt): Add handling to builtin string compare
13169         calls.
13170         * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
13171         Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
13172         * tree.c (build_common_builtin_nodes): Add new defines of
13173         BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
13175 2018-05-31  Jakub Jelinek  <jakub@redhat.com>
13177         PR target/85984
13178         * bb-reorder.c (pass_partition_blocks::gate): Return false for
13179         functions with naked attribute.
13181 2018-05-31  Uros Bizjak  <ubizjak@gmail.com>
13183         * config/i386/sse.md (avx_vec_concat<mode>):
13184         Substitute concat_tg_mode mode attribute with xtg_mode.
13185         (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
13186         (concat_tg_mode): Remove mode attribute.
13188 2018-05-31  Martin Sebor  <msebor@redhat.com>
13190         PR c/82063
13191         * calls.c (alloc_max_size): Correct a logic error/typo.
13192         Treat excessive arguments as infinite.  Warn for invalid arguments.
13193         * doc/invoke.texi (-Walloc-size-larger-than): Update.
13195 2018-05-31  H.J. Lu  <hongjiu.lu@intel.com>
13197         PR target/85829
13198         * config/i386/x86-tune.def: Re-enable partial_reg_dependency
13199         and movx for Haswell.
13201 2018-05-31  Chung-Lin Tang  <cltang@codesourcery.com>
13202             Cesar Philippidis  <cesar@codesourcery.com>
13204         PR middle-end/85879
13205         * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
13206         when emitting error on private/firstprivate reductions.
13207         * omp-low.c (lower_omp_target): Avoid reference-type processing
13208         on pointers for firstprivate clause.
13210 2018-05-31  Sameera Deshpande <sameera.deshpande@linaro.org>
13212         * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
13213         (st1x2): Likewise.
13214         (st1x3): Likewise.
13215         * config/aarch64/aarch64-simd.md
13216         (aarch64_ld1x3<VALLDIF:mode>): New pattern.
13217         (aarch64_ld1_x3_<mode>): Likewise
13218         (aarch64_st1x2<VALLDIF:mode>): Likewise
13219         (aarch64_st1_x2_<mode>): Likewise
13220         (aarch64_st1x3<VALLDIF:mode>): Likewise
13221         (aarch64_st1_x3_<mode>): Likewise
13222         * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
13223         (vld1_s8_x3): Likewise.
13224         (vld1_u16_x3): Likewise.
13225         (vld1_s16_x3): Likewise.
13226         (vld1_u32_x3): Likewise.
13227         (vld1_s32_x3): Likewise.
13228         (vld1_u64_x3): Likewise.
13229         (vld1_s64_x3): Likewise.
13230         (vld1_f16_x3): Likewise.
13231         (vld1_f32_x3): Likewise.
13232         (vld1_f64_x3): Likewise.
13233         (vld1_p8_x3): Likewise.
13234         (vld1_p16_x3): Likewise.
13235         (vld1_p64_x3): Likewise.
13236         (vld1q_u8_x3): Likewise.
13237         (vld1q_s8_x3): Likewise.
13238         (vld1q_u16_x3): Likewise.
13239         (vld1q_s16_x3): Likewise.
13240         (vld1q_u32_x3): Likewise.
13241         (vld1q_s32_x3): Likewise.
13242         (vld1q_u64_x3): Likewise.
13243         (vld1q_s64_x3): Likewise.
13244         (vld1q_f16_x3): Likewise.
13245         (vld1q_f32_x3): Likewise.
13246         (vld1q_f64_x3): Likewise.
13247         (vld1q_p8_x3): Likewise.
13248         (vld1q_p16_x3): Likewise.
13249         (vld1q_p64_x3): Likewise.
13250         (vst1_s64_x2): Likewise.
13251         (vst1_u64_x2): Likewise.
13252         (vst1_f64_x2): Likewise.
13253         (vst1_s8_x2): Likewise.
13254         (vst1_p8_x2): Likewise.
13255         (vst1_s16_x2): Likewise.
13256         (vst1_p16_x2): Likewise.
13257         (vst1_s32_x2): Likewise.
13258         (vst1_u8_x2): Likewise.
13259         (vst1_u16_x2): Likewise.
13260         (vst1_u32_x2): Likewise.
13261         (vst1_f16_x2): Likewise.
13262         (vst1_f32_x2): Likewise.
13263         (vst1_p64_x2): Likewise.
13264         (vst1q_s8_x2): Likewise.
13265         (vst1q_p8_x2): Likewise.
13266         (vst1q_s16_x2): Likewise.
13267         (vst1q_p16_x2): Likewise.
13268         (vst1q_s32_x2): Likewise.
13269         (vst1q_s64_x2): Likewise.
13270         (vst1q_u8_x2): Likewise.
13271         (vst1q_u16_x2): Likewise.
13272         (vst1q_u32_x2): Likewise.
13273         (vst1q_u64_x2): Likewise.
13274         (vst1q_f16_x2): Likewise.
13275         (vst1q_f32_x2): Likewise.
13276         (vst1q_f64_x2): Likewise.
13277         (vst1q_p64_x2): Likewise.
13278         (vst1_s64_x3): Likewise.
13279         (vst1_u64_x3): Likewise.
13280         (vst1_f64_x3): Likewise.
13281         (vst1_s8_x3): Likewise.
13282         (vst1_p8_x3): Likewise.
13283         (vst1_s16_x3): Likewise.
13284         (vst1_p16_x3): Likewise.
13285         (vst1_s32_x3): Likewise.
13286         (vst1_u8_x3): Likewise.
13287         (vst1_u16_x3): Likewise.
13288         (vst1_u32_x3): Likewise.
13289         (vst1_f16_x3): Likewise.
13290         (vst1_f32_x3): Likewise.
13291         (vst1_p64_x3): Likewise.
13292         (vst1q_s8_x3): Likewise.
13293         (vst1q_p8_x3): Likewise.
13294         (vst1q_s16_x3): Likewise.
13295         (vst1q_p16_x3): Likewise.
13296         (vst1q_s32_x3): Likewise.
13297         (vst1q_s64_x3): Likewise.
13298         (vst1q_u8_x3): Likewise.
13299         (vst1q_u16_x3): Likewise.
13300         (vst1q_u32_x3): Likewise.
13301         (vst1q_u64_x3): Likewise.
13302         (vst1q_f16_x3): Likewise.
13303         (vst1q_f32_x3): Likewise.
13304         (vst1q_f64_x3): Likewise.
13305         (vst1q_p64_x3): Likewise.
13307 2018-05-30  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
13309         * config/msp430/msp430.c (msp430_output_labelref): Prepend
13310         user_label_prefix to name.
13312         * tree-core.h: Update comment about the format of NAME string
13313         passed to handler in attribute_spec.
13315         * config/msp430/msp430.md: Remove erroneous subreg expression from
13316         zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
13317         zero_extend{q,h}isi2.
13319 2018-05-30  Borislav Petkov  <bp@suse.de>
13321         * doc/extend.texi: Document some architecture specific
13322         constraints and sort entries.
13324 2018-05-30  Martin Sebor  <msebor@redhat.com>
13326         PR middle-end/85369
13327         * builtins.c (expand_builtin_stpcpy_1): New function.
13328         (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
13329         only if the former succeeds.
13331 2018-05-31  Sameera Deshpande <sameera.deshpande@linaro.org>
13333         * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
13334         in saphira.
13336 2018-05-30  Jan Hubicka  <hubicka@ucw.cz>
13338         * doc/invoke.texi (-flinker-output): Document
13340 2018-05-30  Jan Hubicka  <hubicka@ucw.cz>
13342         * passes.c (ipa_write_summaries): Only modify statements if body
13343         is in memory.
13344         * cgraphunit.c (ipa_passes): Also produce intermeidate code when
13345         incrementally linking.
13346         (ipa_passes): Likewise.
13347         * lto-cgraph.c (lto_output_node): When incrementally linking do not
13348         pass down resolution info.
13349         * common.opt (flag_incremental_link): Update info.
13350         * gcc.c (plugin specs): Turn flinker-output=* to
13351         -plugin-opt=-linker-output-known
13352         * toplev.c (compile_file): Also cut compilation when doing incremental
13353         link.
13354         * flag-types. (enum lto_partition_model): Add
13355         LTO_LINKER_OUTPUT_NOLTOREL.
13356         (invoke.texi): Add -flinker-output docs.
13357         * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
13358         link same way as WPA; do not stream in dead initializers.
13360         * dwarf2out.c (dwarf2out_die_ref_for_decl,
13361         darf2out_register_external_decl): Support incremental link.
13363 2018-05-30  Jan Hubicka  <hubicka@ucw.cz>
13365         * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
13367 2018-05-30  Jan Hubicka  <hubicka@ucw.cz>
13369         * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
13370         it down to simple_object_copy_lto_debug_sections.
13371         (run_gcc): Determine incremental LTO link time and configure
13372         lto1 into non-wpa mode, disable renaming of debug sections.
13374 2018-05-30  Kelvin Nilsen  <kelvin@gcc.gnu.org>
13376         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
13377         descriptions of various incorrectly documented functions.
13379 2018-05-30  Andre Vieira  <andre.simoesdiasvieira@arm.com>
13381         Revert:
13382         * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
13383         address check not strict.
13385 2018-05-30  Richard Biener  <rguenther@suse.de>
13387         PR tree-optimization/85964
13388         * tracer.c (better_p): Drop initialized count check, we only
13389         call the function with initialized counts now.
13390         (find_best_successor): Do find a best edge if one
13391         has uninitialized count.
13392         (find_best_predecessor): Likewise.  Do BB frequency check only
13393         if count is initialized.
13395 2017-05-30  Jackson Woodruff  <jackson.woodruff@arm.com>
13397         * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
13398         (aarch64_ldrstr_offset_compare): New.
13399         (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
13400         load/store orderings.
13401         (aarch64_gen_adjusted_ldpstp): Likewise.
13403 2018-05-30  Wilco Dijkstra  <wdijkstr@arm.com>
13405         * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
13406         Check for subset of GENERAL_REGS and FP_REGS.
13407         * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
13408         r=w alternative.
13410 2018-05-30  Richard Sandiford  <richard.sandiford@linaro.org>
13412         * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
13413         and wi::to_poly_offset.  Add the current offset and then check
13414         whether the sum fits, rather than using an unchecked addition of
13415         a checked term.  Check for a shwi rather than a uhwi.
13416         * expr.c (get_bit_range): Use tree_to_poly_uint64.
13417         (store_constructor): Use poly_int_tree_p.
13418         (expand_expr_real_1): Likewise.
13419         * function.c (assign_temp): Likewise.
13420         * fold-const.c (const_binop): Use poly_int_tree_p and
13421         wi::to_poly_offset.
13422         (fold_indirect_ref_1): Likewise.  Use multiple_p to attempt an exact
13423         division.
13424         * ipa-icf-gimple.c (func_checker::compare_operand): Use
13425         to_poly_offset for MEM offsets.
13426         * ipa-icf.c (sem_variable::equals): Likewise.
13427         * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
13428         * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
13429         wi::to_poly_offset for BIT_FIELD_REF offsets.
13430         (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
13431         wi::to_poly_offset.
13432         * var-tracking.c (emit_note_insn_var_location): Use
13433         tree_to_poly_uint64.
13435 2018-05-29  Jim Wilson  <jimw@sifive.com>
13437         * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
13439 2018-05-29  Uros Bizjak  <ubizjak@gmail.com>
13441         PR target/85950
13442         * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
13443         Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
13444         sequence.
13445         (sse4_1_round<mode>2): Use nonimmediate_operand
13446         for operand 1 predicate.
13448 2018-05-29  Martin Sebor  <msebor@redhat.com>
13449             Richard Biener  <rguenther@suse.de>
13451         PR testsuite/85888
13452         * calls.c (get_size_range): Call determine_value_range instead
13453         of get_value_range..
13454         * tree-vrp.h (determine_value_range): Declared new function.
13455         * tree-vrp.c (determine_value_range_1, determine_value_range): New.
13457 2018-05-29  Richard Biener  <rguenther@suse.de>
13459         * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
13460         sure to use non-pattern stmts for get_earlier_stmt arguments.
13461         * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
13462         called on pattern stmts.
13463         (get_later_stmt): Likewise.
13465 2018-05-29  Martin Liska  <mliska@suse.cz>
13467         PR gcov-profile/85759
13468         * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
13469         env variables.
13471 2018-05-29  Jakub Jelinek  <jakub@redhat.com>
13473         * tree-cfg.c (verify_gimple_assign_unary): Add checking for
13474         VEC_UNPACK_*_EXPR.
13475         (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
13476         VEC_PACK_*_EXPR.
13478         PR target/85918
13479         * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
13480         VEC_PACK_FLOAT_EXPR): New tree codes.
13481         * tree-pretty-print.c (op_code_prio): Handle
13482         VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
13483         (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
13484         VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
13485         * tree-inline.c (estimate_operator_cost): Likewise.
13486         * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
13487         * fold-const.c (const_binop): Likewise.
13488         (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
13489         VEC_UNPACK_FIX_TRUNC_LO_EXPR.
13490         * tree-cfg.c (verify_gimple_assign_unary): Likewise.
13491         (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
13492         * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
13493         VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
13494         * expr.c (expand_expr_real_2): Likewise.
13495         * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
13496         vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
13497         vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
13498         optabs.
13499         * optabs.c (expand_widen_pattern_expr): For
13500         VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
13501         sign from result type rather than operand's type.
13502         (expand_binop_directly): For vec_packu_float_optab and
13503         vec_packs_float_optab allow result type to be different from operand's
13504         type.
13505         * optabs-tree.c (optab_for_tree_code): Handle
13506         VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
13507         VEC_PACK_FLOAT_EXPR.  Formatting fixes.
13508         * tree-vect-generic.c (expand_vector_operations_1):  Handle
13509         VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
13510         VEC_PACK_FLOAT_EXPR.
13511         * tree-vect-stmts.c (supportable_widening_operation): Handle
13512         FIX_TRUNC_EXPR.
13513         (supportable_narrowing_operation): Handle FLOAT_EXPR.
13514         * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
13515         * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
13516         (float<floatunssuffix>v2div2sf2): ... this.  Formatting fix.
13517         (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
13518         mode attributes.
13519         (vec_pack<floatprefix>_float_<mode>): New expander.
13520         (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
13521         attributes.
13522         (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
13523         vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
13524         * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
13525         vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
13526         vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
13527         Document.
13528         * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
13529         VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
13530         (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
13531         VEC_PACK_FLOAT_EXPR): Document.
13533 2018-05-29  Richard Biener  <rguenther@suse.de>
13535         * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
13536         member.
13537         (stmt_vec_info_vec): Make pointer.
13538         (init_stmt_vec_info_vec): Remove.
13539         (free_stmt_vec_info_vec): Likewise.
13540         (set_stmt_vec_info_vec): New function.
13541         (free_stmt_vec_infos): Likewise.
13542         (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
13543         (set_vinfo_for_stmt): Likewise.
13544         (get_earlier_stmt): Likewise.
13545         (get_later_stmt): Likewise.
13546         * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
13547         (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
13548         (vec_info::~vec_info): Free stmt_vec_infos.
13549         (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
13550         Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
13551         (pass_slp_vectorize::execute): Likewise.
13552         * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
13553         (free_stmt_vec_info_vec): Likewise.
13554         (set_stmt_vec_info_vec): New function.
13555         (free_stmt_vec_infos): Likewise.
13556         * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
13557         the global stmt_vec_info_vec.
13558         * tree-parloops.c (gather_scalar_reductions): Use
13559         set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
13560         vector.
13562 2018-05-29  Richard Biener  <rguenther@suse.de>
13564         * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
13566 2018-05-29  Martin Liska  <mliska@suse.cz>
13567             David Malcolm  <dmalcolm@redhat.com>
13569         * vec.c (test_reverse): New.
13570         (vec_c_tests): Add new test.
13571         * vec.h (vl_ptr>::reverse): New function.
13573 2018-05-29  Gerald Pfeifer  <gerald@pfeifer.com>
13575         * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
13577         * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
13578         and later.
13580 2018-05-28  Bernd Edlinger  <bernd.edlinger@hotmail.de>
13582         * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
13584 2018-05-28  Richard Biener  <rguenther@suse.de>
13586         PR tree-optimization/85933
13587         * tree-vect-data-refs.c (vect_record_base_alignments): Only
13588         look at stmts marked as vectorizable.
13590 2018-05-28  Richard Biener  <rguenther@suse.de>
13592         PR tree-optimization/85934
13593         * tree-vect-generic.c (expand_vector_operations_1): Hoist
13594         vector boolean check before scalar optimization.
13596 2018-05-28  Jakub Jelinek  <jakub@redhat.com>
13598         * doc/invoke.texi (ARM Options): Use @item instead of @itemx
13599         for armv5te.
13601 2018-05-28  Mark Wielaard  <mark@klomp.org>
13603         * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
13604         if it is an expression containing a minus sign.
13606 2018-05-27  John David Anglin  <danglin@gcc.gnu.org>
13608         * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
13610 2018-05-27  Paul Koning  <ni1d@arrl.net>
13612         * config/pdp11/pdp11.md (truncsihi2): Remove.
13614 2018-05-27  Monk Chiang  <sh.chiang04@gmail.com>
13615             Chung-Ju Wu  <jasonwucj@gmail.com>
13617         * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
13618         implementation.
13619         (unaligned_store_dw): Ditto.
13620         * config/nds32/nds32-memory-manipulation.c
13621         (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
13622         (nds32_gen_dup_4_byte_to_word_value): Rename to ...
13623         (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
13624         (emit_setmem_word_loop): Rename to ...
13625         (emit_setmem_doubleword_loop): ... this.
13626         (nds32_gen_dup_4_byte_to_word_value): New function.
13627         (nds32_gen_dup_8_byte_to_double_word_value): New function.
13628         (nds32_expand_setmem_loop): Refine implementation.
13629         (nds32_expand_setmem_loop_v3m): Ditto.
13630         * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
13631         pattern.
13633 2018-05-27  Chung-Ju Wu  <jasonwucj@gmail.com>
13635         * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
13637 2018-05-27  Chung-Ju Wu  <jasonwucj@gmail.com>
13639         * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
13640         (nds32_init_machine_status): Initialize machine->attr_naked_p and
13641         machine->attr_no_prologue_p.
13642         (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
13643         (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
13644         (nds32_expand_epilogue): Consider attr_naked_p.
13645         (nds32_expand_epilogue_v3pop): Likewise.
13646         (nds32_can_use_return_insn): Likewise.
13647         * config/nds32/nds32.h (machine_function): Add attr_naked_p and
13648         attr_no_prologue_p fields.
13649         * config/nds32/nds32.opt (mret-in-naked-func): New option.
13651 2018-05-27  Jakub Jelinek  <jakub@redhat.com>
13653         PR target/85918
13654         * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
13655         attributes.
13656         * config/i386/sse.md
13657         (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
13658         Rename to ...
13659         (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
13660         ... this.
13661         (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
13662         Rename to ...
13663         (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
13664         ... this.
13665         (*<floatsuffix>floatv2div2sf2): Rename to ...
13666         (*float<floatunssuffix>v2div2sf2): ... this.
13667         (<floatsuffix>floatv2div2sf2_mask): Rename to ...
13668         (float<floatunssuffix>v2div2sf2_mask): ... this.
13669         (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
13670         (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
13671         (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
13672         to ...
13673         (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
13674         ... this.
13675         (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
13676         Rename to ...
13677         (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
13678         ... this.
13679         (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
13680         Rename to ...
13681         (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
13682         ... this.
13683         (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
13684         (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
13685         (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
13686         gen_ufix_truncv8dfv8si2.
13687         * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
13688         __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
13689         __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
13690         __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
13691         __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
13692         __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
13693         __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
13694         Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
13696 2018-05-24  H.J. Lu  <hongjiu.lu@intel.com>
13698         PR target/85900
13699         PR target/85345
13700         * varasm.c (assemble_alias): Lookup ifunc attribute on error.
13702 2018-05-25  Jim Wilson  <jimw@sifive.com>
13704         * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
13705         * config/riscv/riscv.c (struct machine_function): Add
13706         interrupt_handler_p and attribute_checked_p fields.
13707         (riscv_attribute_table): Add interrupt.
13708         (riscv_interrupt_type_p): New.
13709         (riscv_save_reg_p): Save extra regs for interrupt handler.
13710         (riscv_use_save_libcall): Return false  for interrupt handler.
13711         (riscv_first_stack_step): Add forward declaration.
13712         (riscv_compute_frame_info): New local interrupt_save_t1.  Set it
13713         for interrupt handler with large frame.  Use it for saved reg list.
13714         (riscv_expand_prologue): Move flag_stack_usage_info support to
13715         eliminate duplication.
13716         (riscv_expand_epilogue): Generate mret for interrupt handler.
13717         (riscv_epilogue_uses): New.
13718         (riscv_can_use_return_insn): Return false for interrupt handler.
13719         (riscv_function_ok_for_sibcall): Likewise.
13720         (riscv_set_current_function): Add interrupt handler support.
13721         * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
13722         * config/riscv/riscv.md (UNSPECV_MRET): New.
13723         (GP_REGNUM): New.
13724         (riscv_frflags, riscv_fsflags): Use tab after opcode.
13725         (riscv_mret): New.
13726         * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
13728 2018-05-25  Bill Schmidt  <wschmidt@linux.ibm.com>
13730         PR tree-optimization/85712
13731         * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
13732         this candidate has already been replaced in-situ by a copy.
13734 2018-05-25  Jason Merrill  <jason@redhat.com>
13736         PR c++/80485 - inline function non-zero address.
13737         * symtab.c (nonzero_address): Check DECL_COMDAT.
13739 2018-05-25  Uros Bizjak  <ubizjak@gmail.com>
13741         PR target/83628
13742         * config/alpha/alpha.md (ashlsi3): New insn pattern.
13743         (*ashlsi_se): Rename from *ashldi_se.  Define as sign
13744         extension of SImode operation.  Use const123_operand predicate.
13745         (*saddsi_1): Remove.
13746         (*saddl_se_1): Ditto.
13747         (*ssubsi_1): Ditto.
13748         (*ssubl_se_1): Ditto.
13749         * config/alpha/predicates.md (const123_operand): New predicate.
13750         * config/alpha/constraints.md (P): Use IN_RANGE.
13752 2018-05-25  Richard Biener  <rguenther@suse.de>
13754         * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
13755         defaulted to true.
13756         (ref_maybe_used_by_stmt_p): Likewise.
13757         (stmt_may_clobber_ref_p): Likewise.
13758         (stmt_may_clobber_ref_p_1): Likewise.
13759         * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
13760         and pass it along.
13761         (ref_maybe_used_by_stmt_p): Likewise.
13762         (stmt_may_clobber_ref_p): Likewise.
13763         (stmt_may_clobber_ref_p_1): Likewise.
13764         * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
13765         the alias oracle to disambiguate DRs with stmts DR analysis
13766         couldn't handle.
13767         (vect_analyze_data_refs): Do not give up on not analyzable
13768         DRs for BB vectorization.  Remove code truncating the dataref
13769         vector.
13771 2018-05-25  Jakub Jelinek  <jakub@redhat.com>
13773         PR target/85832
13774         * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
13775         Add (=Yk,v,C) variant using vptestm insn.  Use TARGET_AVX512BW
13776         in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
13778 2018-05-25  Richard Biener  <rguenther@suse.de>
13780         * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
13781         function, combining stmt data ref gathering and fatal analysis
13782         parts.
13783         (vect_analyze_data_refs): Remove now redudnant code and simplify.
13784         * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
13785         vect_analyze_loop_2 and use vect_find_stmt_data_reference.
13786         * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
13787         * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
13789 2018-05-25  Bin Cheng  <bin.cheng@arm.com>
13791         PR tree-optimization/85720
13792         * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
13793         SCC if all partitions are builtins.
13794         (version_loop_by_alias_check): New parameter.  Generate cancelable
13795         runtime alias check if all partitions are builtins.
13796         (distribute_loop): Update call to above function.
13798 2018-05-25  Bin Cheng  <bin.cheng@arm.com>
13800         * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
13801         (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
13802         (parm_default_def_partition_arg): Ditto.
13803         (set_parm_default_def_partition): Ditto.
13804         (get_parm_default_def_partitions): Ditto and make it static.
13805         (get_undefined_value_partitions): Ditto and make it static.
13806         (remove_ssa_form): Refactor call to init_var_map here.
13807         * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
13808         computation for loop region.
13809         (coalesce_partitions, compute_optimized_partition_bases): Ditto.
13810         (register_default_def): Delete.
13811         (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
13812         (parm_default_def_partition_arg): Ditto.
13813         (set_parm_default_def_partition): Ditto.
13814         (get_parm_default_def_partitions): Ditto and make it static.
13815         (get_undefined_value_partitions): Ditto and make it static.
13816         (coalesce_with_default, coalesce_with_default): Update comment.
13817         (create_coalesce_list_for_region): New func factored out from
13818         create_outofssa_var_map.
13819         (populate_coalesce_list_for_outofssa): New func factored out from
13820         create_outofssa_var_map and coalesce_ssa_name.
13821         (create_outofssa_var_map): Delete.
13822         (coalesce_ssa_name): Refactor to support live range computation.
13823         * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
13824         (get_parm_default_def_partitions): Delete.
13825         (get_undefined_value_partitions): Ditto.
13826         * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
13827         computation for loop region.
13828         (new_tree_live_info, loe_visit_block): Ditto.
13829         (live_worklist, set_var_live_on_entry): Ditto.
13830         (calculate_live_on_exit, verify_live_on_entry): Ditto.
13831         * tree-ssa-live.h (struct _var_map): New fields.
13832         (init_var_map): Change decl.
13833         (region_contains_p): New.
13835 2018-05-25  Bin Cheng  <bin.cheng@arm.com>
13837         * tree-ssa-live.h (live_merge_and_clear): Delete.
13839 2018-05-25  Richard Biener  <rguenther@suse.de>
13841         PR c++/85912
13842         * tree-dump.c (dequeue_and_dump): Remove access to removed
13843         operand 2 of a SWITCH_EXPR.
13845 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
13847         * doc/sourcebuild.texi (vect_double_cond_arith): Include
13848         multiplication and division.
13849         * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
13850         (cond_udiv@var{m}, cond_umod@var{m}): Document.
13851         * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
13852         (cond_udiv_optab, cond_umod_optab): New optabs.
13853         * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
13854         (IFN_COND_RDIV): New internal functions.
13855         * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
13856         TRUNC_MOD_EXPR and RDIV_EXPR.
13857         * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
13858         * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
13859         New unspecs.
13860         (SVE_INT_BINARY): Include mult.
13861         (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
13862         (optab, sve_int_op): Handle mult.
13863         (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
13864         UNSPEC_COND_DIV.
13865         * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
13866         for SVE_INT_BINARY_SD.
13868 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
13870         * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
13871         (optab, sve_int_op): Handle div and udiv.
13872         * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
13873         for SVE_INT_BINARY_SD.
13874         (*<optab><mode>3): New insn for the same.
13876 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
13878         * tree-vect-patterns.c: Include predict.h.
13879         (vect_recog_divmod_pattern): Restrict check for division support
13880         to when optimizing for size.
13882 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
13884         * doc/sourcebuild.texi (vect_double_cond_arith: Document.
13885         * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
13886         (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
13887         (gimple_match_op::set_op): Likewise.
13888         (gimple_resimplify4): Declare.
13889         * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
13890         (expr::gen_transform): Likewise.
13891         (decision_tree::gen): Generate a simplification routine for 4 operands.
13892         * gimple-match-head.c (gimple_simplify): Add an overload for
13893         4 operands.  In the top-level function, handle up to 4 call
13894         arguments and call gimple_resimplify4.
13895         (gimple_resimplify4): New function.
13896         (build_call_internal): Pass a fourth operand.
13897         (maybe_push_to_seq): Likewise.
13898         * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
13899         Fold VEC_COND_EXPRs of an operation and a default value into
13900         an IFN_COND_* function if possible.
13901         * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
13902         New unspecs.
13903         (SVE_COND_FP_BINARY): Include them.
13904         (optab, sve_fp_op): Handle them.
13905         (SVE_INT_BINARY_REV): New code iterator.
13906         (SVE_COND_FP_BINARY_REV): New int iterator.
13907         (commutative): New int attribute.
13908         * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
13909         Declare.
13910         * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
13911         function.
13912         * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
13913         (*cond_<optab><mode>): New patterns for reversed operands.
13915 2018-05-25  Richard Biener  <rguenther@suse.de>
13917         * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
13918         (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
13919         (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
13920         (STMT_VINFO_GROUPED_ACCESS): Adjust.
13921         * tree-vect-data-refs.c (everywhere): Adjust users.
13922         * tree-vect-loop.c (everywhere): Likewise.
13923         * tree-vect-slp.c (everywhere): Likewise.
13924         * tree-vect-stmts.c (everywhere): Likewise.
13925         * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
13927 2018-05-25  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
13929         * configure.ac (gcc_cv_as_section_has_e): Move to common section.
13930         Rename to...
13931         (gcc_cv_as_section_exclude): ... this.
13932         Try Solaris as #exclude syntax.
13933         * configure: Regenerate.
13934         * config.in: Regenerate.
13935         * config/i386/i386.c (i386_solaris_elf_named_section): Handle
13936         SECTION_EXCLUDE.
13937         * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
13938         [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
13940         * varasm.c (default_elf_asm_named_section): Don't check if
13941         HAVE_GAS_SECTION_EXCLUDE is defined.
13943 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
13945         * doc/md.texi: Update the documentation of the cond_* optabs
13946         to mention the new final operand.  Fix GET_MODE_NUNITS call.
13947         Describe the scalar case too.
13948         * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
13949         * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
13950         instead of 2.
13951         (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
13952         (get_conditional_internal_fn): Update comment.
13953         * tree-vect-loop.c (vectorizable_reduction): Pass the original
13954         accumulator value as a final argument to conditional functions.
13955         * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
13956         a define_expand and add an "else" operand.  Assert for now that
13957         the else operand is equal to operand 2.  Use SVE_INT_BINARY and
13958         SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
13959         (*cond_<optab><mode>): New patterns.
13960         * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
13961         (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
13962         (UNSPEC_COND_EOR): Delete.
13963         (optab): Remove associated mappings.
13964         (SVE_INT_BINARY): New code iterator.
13965         (sve_int_op): Remove int attribute and add "minus" to the code
13966         attribute.
13967         (SVE_COND_INT_OP): Delete.
13968         (SVE_COND_FP_OP): Rename to...
13969         (SVE_COND_FP_BINARY): ...this.
13971 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
13973         * optabs.c (can_reuse_operands_p): New function.
13974         (maybe_legitimize_operands): Try to reuse the results for
13975         earlier operands.
13977 2018-05-24  Uros Bizjak  <ubizjak@gmail.com>
13979         * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
13980         Add {q} suffix to insn mnemonic.
13982 2018-05-23  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
13984         * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
13985         (msp430_warn_func_return): New.
13987 2018-05-24  Roger Sayle  <roger@nextmovesoftware.com>
13989         * fold-const.c (tree_nonzero_bits): New function.
13990         * fold-const.h (tree_nonzero_bits): Likewise.
13991         * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
13992         friends.  POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
13994 2018-05-24  H.J. Lu  <hongjiu.lu@intel.com>
13996         PR target/85900
13997         PR target/85345
13998         * varasm.c (assemble_alias): Check ifunc_resolver only on
13999         FUNCTION_DECL.
14001 2018-05-24  Uros Bizjak  <ubizjak@gmail.com>
14003         PR target/85903
14004         * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
14005         when memory input operand is handled.
14007 2018-05-24  Luis Machado  <luis.machado@linaro.org>
14009         * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
14010         global.
14011         (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
14013 2018-05-24  Richard Sandiford  <richard.sandiford@linaro.org>
14015         * match.pd: Delay FMA folds until after vectorization.
14017 2018-05-24  Andre Vieira  <andre.simoesdiasvieira@arm.com>
14019         PR target/83009
14020         * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
14021         address check not strict.
14023 2018-05-24  Richard Sandiford  <richard.sandiford@linaro.org>
14025         * gimple-match.h (gimple_match_op): New class.
14026         (mprts_hook): Replace parameters with a gimple_match_op *.
14027         (maybe_build_generic_op): Likewise.
14028         (gimple_simplified_result_is_gimple_val): Replace parameters with
14029         a const gimple_match_op *.
14030         (gimple_simplify): Replace code_helper * and tree * parameters with
14031         a gimple_match_op * parameter.
14032         (gimple_resimplify1): Replace code_helper *, tree and tree *
14033         parameters with a gimple_match_op * parameter.
14034         (gimple_resimplify2): Likewise.
14035         (gimple_resimplify3): Likewise.
14036         (maybe_push_res_to_seq): Replace code_helper, tree and tree *
14037         parameters with a gimple_match_op * parameter.
14038         * gimple-match-head.c (gimple_simplify): Change prototypes of
14039         auto-generated functions to take a gimple_match_op * instead of
14040         separate code_helper * and tree * parameters.  Make the same
14041         change in the top-level overload and update calls to the
14042         gimple_resimplify routines.  Update calls to the auto-generated
14043         functions and to maybe_push_res_to_seq in the publicly-facing
14044         operation-specific gimple_simplify overloads.
14045         (gimple_match_op::MAX_NUM_OPS): Define.
14046         (gimple_resimplify1): Replace rcode and ops with a single res_op
14047         parameter.  Update call to gimple_simplify.
14048         (gimple_resimplify2): Likewise.
14049         (gimple_resimplify3): Likewise.
14050         (mprts_hook): Replace parameters with a gimple_match_op *.
14051         (maybe_build_generic_op): Likewise.
14052         (build_call_internal): Replace type, nargs and ops with
14053         a gimple_match_op *.
14054         (maybe_push_res_to_seq): Replace res_code, type and ops parameters
14055         with a single gimple_match_op *.  Update calls to mprts_hook,
14056         build_call_internal and gimple_simplified_result_is_gimple_val.
14057         Factor out code that is common to the tree_code and combined_fn cases.
14058         * genmatch.c (expr::gen_transform): Replace tem_code and
14059         tem_ops with a gimple_match_op called tem_op.  Update calls
14060         to the gimple_resimplify functions and maybe_push_res_to_seq.
14061         (dt_simplify::gen_1): Manipulate res_op instead of res_code and
14062         res_ops.  Update call to the gimple_resimplify functions.
14063         (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
14064         (decision_tree::gen): Make the functions take a gimple_match_op *
14065         called res_op instead of separate res_code and res_ops parameters.
14066         Update call accordingly.
14067         * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
14068         and ops with a single res_op parameter.  Update calls to
14069         maybe_build_generic_op and maybe_push_res_to_seq.
14070         (fold_stmt_1): Update calls to gimple_simplify and
14071         replace_stmt_with_simplification.
14072         (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
14073         and gimple_simplified_result_is_gimple_val.
14074         * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
14075         gimple_simplify.
14076         * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
14077         with a gimple_match_op *.
14078         (vn_nary_build_or_lookup): Likewise.  Update call to
14079         vn_nary_build_or_lookup_1.
14080         (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
14081         gimple_match_op *.  Update calls to the gimple_resimplify routines
14082         and to gimple_simplified_result_is_gimple_val.
14083         (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
14084         Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
14085         (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
14086         (visit_nary_op): Likewise.
14087         (visit_reference_op_load): Likewise.
14089 2018-05-23  Luis Machado  <luis.machado@linaro.org>
14091         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
14092         modifier for printing the step amount.
14094 2018-05-23  Jozef Lawrynowicz  <jozef.l@somniumtech.com>
14096         PR target/78849
14097         * tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
14098         types.
14100 2018-05-23  Segher Boessenkool  <segher@kernel.crashing.org>
14102         * doc/sourcebuild.texi (Endianness): New subsubsection.
14104 2018-05-23  Luis Machado  <luis.machado@linaro.org>
14106         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
14107         <prefetch_dynamic_strides>: New const bool field.
14108         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
14109         prefetch_dynamic_strides.
14110         (exynosm1_prefetch_tune): Likewise.
14111         (thunderxt88_prefetch_tune): Likewise.
14112         (thunderx_prefetch_tune): Likewise.
14113         (thunderx2t99_prefetch_tune): Likewise.
14114         (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
14115         false.
14116         (aarch64_override_options_internal): Update to set
14117         PARAM_PREFETCH_DYNAMIC_STRIDES.
14118         * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
14119         * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
14120         * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
14121         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
14122         prefetch-dynamic-strides setting.
14124 2018-05-23  Luis Machado  <luis.machado@linaro.org>
14126         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
14127         <minimum_stride>: New const int field.
14128         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
14129         minimum_stride field defaulting to -1.
14130         (exynosm1_prefetch_tune): Likewise.
14131         (thunderxt88_prefetch_tune): Likewise.
14132         (thunderx_prefetch_tune): Likewise.
14133         (thunderx2t99_prefetch_tune): Likewise.
14134         (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
14135         <default_opt_level>: Set to 3.
14136         (aarch64_override_options_internal): Update to set
14137         PARAM_PREFETCH_MINIMUM_STRIDE.
14138         * doc/invoke.texi (prefetch-minimum-stride): Document new option.
14139         * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
14140         * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
14141         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
14142         stride is constant and is below the minimum stride threshold.
14144 2018-05-23  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
14146         * config/arm/arm-cpus.in (mode26): Delete.
14147         (armv4): Delete mode26 reference.
14148         * config/arm/arm.c (arm_configure_build_target): Delete use of
14149         isa_bit_mode26.
14151 2018-05-23  Uros Bizjak  <ubizjak@gmail.com>
14153         * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
14154         New insn pattern.
14155         (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
14156         Rewrite expander pattern.  Emit gen_floatunssi<mode>2_i387_with_xmm
14157         for non-SSE modes.
14158         (floatunsdisf2): Rewrite expander pattern.  Hanlde TARGET_AVX512F.
14159         (floatunsdidf2): Ditto.
14161 2018-05-23  Uros Bizjak  <ubizjak@gmail.com>
14163         * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
14164         (fixuns_trunc<mode>si2_avx512f): Ditto.
14165         (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
14166         (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
14167         Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
14169 2018-05-23  Alexander Monakov  <amonakov@ispras.ru>
14171         PR rtl-optimization/79985
14172         * df-scan.c (df_insn_refs_collect): Remove special case for
14173         global registers and asm statements.
14175 2018-05-23  Alexander Monakov  <amonakov@ispras.ru>
14177         * extend.texi (Global Register Variables): Rewrite the bullet list.
14178         Note that the register is available for allocation. Note that access
14179         via inline asm must use constraints. Add note about async-signal
14180         handlers. Remove paragraph about automagic register selection.
14182 2018-05-23  Richard Biener  <rguenther@suse.de>
14184         * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
14185         of fixed offset from memset VN.
14187 2018-05-23  Bill Schmidt  <wschmidt@linux.ibm.com>
14189         * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
14190         first_interp field.
14191         (alloc_cand_and_find_basis): Initialize first_interp field.
14192         (slsr_process_mul): Modify first_interp field.
14193         (slsr_process_add): Likewise.
14194         (slsr_process_cast): Modify first_interp field for each new
14195         interpretation.
14196         (slsr_process_copy): Likewise.
14197         (dump_candidate): Dump first_interp field.
14198         (replace_mult_candidate): Process all interpretations, not just
14199         subsequent ones.
14200         (replace_rhs_if_not_dup): Likewise.
14201         (replace_one_candidate): Likewise.
14203 2018-05-23  Wilco Dijkstra  <wdijkstr@arm.com>
14205         * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
14206         Add new boolean.
14207         (aarch64_needs_frame_chain): New function.
14208         (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
14210 2018-05-23  Sudakshina Das  <sudi.das@arm.com>
14212         PR target/84882
14213         * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
14214         Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
14215         * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
14216         * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
14217         as true for strict-align.
14218         (aarch64_can_inline_p): Perform checks even when callee has no
14219         attributes to check for strict alignment.
14220         * doc/extend.texi (AArch64 Function Attributes): Document
14221         no-strict-align.
14222         * doc/invoke.texi: (AArch64 Options): Likewise.
14224 2018-05-23  Richard Sandiford  <richard.sandiford@linaro.org>
14226         PR tree-optimization/85853
14227         * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
14228         the handling of the root of the node to...
14229         (vect_slp_analyze_node_operations_1): ...this new function,
14230         and run the whole thing with the child nodes' def types
14231         set according to their SLP node's def type.
14233 2018-05-23  Richard Biener  <rguenther@suse.de>
14235         PR middle-end/85874
14236         * tree-data-ref.c (create_runtime_alias_checks): Defer
14237         and ignore overflow warnings.
14239 2018-05-23  Yury Gribov  <tetra2005@gmail.com>
14241         PR tree-optimization/85822
14242         * tree-vrp.c (is_masked_range_test): Fix handling of negative
14243         constants.
14245 2018-05-23  Richard Biener  <rguenther@suse.de>
14247         * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
14248         memset constants via native_interpret_expr.
14250 2018-05-22  H.J. Lu  <hongjiu.lu@intel.com>
14252         PR target/85345
14253         * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
14254         attribute.
14255         (cgraph_node::create_alias): Likewise.
14256         (cgraph_node::get_availability): Check ifunc_resolver instead
14257         of looking up ifunc attribute.
14258         * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
14259         * varasm.c (do_assemble_alias): Likewise.
14260         (assemble_alias): Likewise.
14261         (default_binds_local_p_3): Likewise.
14262         * cgraph.h (cgraph_node): Add ifunc_resolver.
14263         (cgraph_node::only_called_directly_or_aliased_p): Return false
14264         for IFUNC resolver.
14265         * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
14266         attribute.
14267         * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
14268         is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
14269         (symtab_node::binds_to_current_def_p): Check ifunc_resolver
14270         instead of looking up ifunc attribute.
14272 2018-05-22  Luis Machado  <luis.machado@linaro.org>
14274         * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
14276 2018-05-22  Martin Sebor  <msebor@redhat.com>
14278         PR middle-end/85359
14279         * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
14280         only when expasion succeeds.
14281         (expand_builtin_strcmp): Same.
14282         (expand_builtin_strncmp): Same.
14284 2018-05-22  Martin Sebor  <msebor@redhat.com>
14286         * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
14288 2018-05-22  Jackson Woodruff  <jackson.woodruff@arm.com>
14289             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
14291         * config/aarch64/aarch64-ldpstp.md: Replace uses of
14292         aarch64_mem_pair_operand with memory_operand and delete operand swapping
14293         code.
14294         * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
14295         Add check for legitimate_address.
14296         (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
14297         (aarch64_swap_ldrstr_operands): New.
14298         * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
14299         Define prototype.
14301 2018-05-22  Jackson Woodruff  <jackson.woodruff@arm.com>
14302             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
14304         * config/aarch64/aarch64.md: New patterns to generate stp
14305         and ldp.
14306         (store_pair_sw, store_pair_dw): New patterns to generate stp for
14307         single words and double words.
14308         (load_pair_sw, load_pair_dw): Likewise.
14309         (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
14310         Delete.
14311         (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
14312         Delete.
14313         * config/aarch64/aarch64-ldpstp.md: Modify peephole
14314         for different mode ldpstp and add peephole for merged zero stores.
14315         Likewise for loads.
14316         * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
14317         Add size check.
14318         (aarch64_gen_store_pair): Rename calls to match new patterns.
14319         (aarch64_gen_load_pair): Rename calls to match new patterns.
14320         * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
14321         (load_pair<DREG:mode><DREG2:mode>): ... This.
14322         (store_pair<mode>): Rename to...
14323         (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
14324         * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
14325         New mode iterators.
14326         (V_INT_EQUIV): Handle SImode.
14327         * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
14328         New predicate.
14330 2018-05-22  Martin Sebor  <msebor@redhat.com>
14332         PR c/85623
14333         * calls.c (maybe_warn_nonstring_arg): Use string length to set
14334         or ajust the presumed bound on an operation to avoid unnecessary
14335         warnings.
14337 2018-05-22  Martin Sebor  <msebor@redhat.com>
14339         PR tree-optimization/85826
14340         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
14341         assuming that a DECL necesarily has a constant size.
14343 2018-05-22  Richard Sandiford  <richard.sandiford@linaro.org>
14345         PR middle-end/85862
14346         * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
14348 2018-05-22  Richard Biener  <rguenther@suse.de>
14350         PR tree-optimization/85834
14351         * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
14352         non-constant and non-zero memset arguments.
14354 2018-05-22  Martin Liska  <mliska@suse.cz>
14356         PR ipa/85607
14357         * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
14359 2018-05-22  Richard Biener  <rguenther@suse.de>
14361         PR tree-optimization/85863
14362         * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
14363         comparisons when vectype is specified.
14364         (vectorizable_condition): Do not specify vectype for
14365         vect_is_simple_cond when SLP vectorizing.
14367 2018-05-21  Michael Meissner  <meissner@linux.ibm.com>
14369         PR target/85657
14370         * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
14371         define __ibm128 as long double.
14372         * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
14373         as a distinct type when IEEE 128-bit support is enabled.
14374         (init_float128_ieee): Fix up conversions between IFmode and IEEE
14375         128-bit types to use the correct functions.
14376         (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
14377         convert between 128-bit floating point types that have different
14378         modes but the same representation, instead of using gen_lowpart to
14379         makean alias.
14380         * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
14381         KFmode.
14382         (IFKF_reg): New attributes to give the register constraints for
14383         IFmode and KFmode.
14384         (extend<mode>tf2_internal): New insns to mark an explicit
14385         conversion between 128-bit floating point types that have a
14386         different mode but share the same representation.
14388 2018-05-21  Richard Sandiford  <richard.sandiford@linaro.org>
14390         PR tree-optimization/85814
14391         * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
14392         a null return from get_strinfo when unsharing the next
14393         strinfo in the chain.
14395 2018-05-21  Vladimir Mezentsev  <vladimir.mezentsev@oracle.com>
14397         PR gcc/84923
14398         * varasm.c (weak_finish): Clean up weak_decls.
14400 2018-05-21  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
14402         * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
14403         UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
14404         UNSPEC_UADALP values.
14405         * config/aarch64/iterators.md (ABAL): New int iterator.
14406         (ABDL2): Likewise.
14407         (ADALP): Likewise.
14408         (sur): Add mappings for the above.
14409         * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
14410         New define_insn.
14411         (aarch64_<sur>abal<mode>_4): Likewise.
14412         (aarch64_<sur>adalp<mode>_3): Likewise.
14413         (<sur>sadv16qi): New define_expand.
14415 2018-05-21  Alexander Nesterovskiy  <alexander.nesterovskiy@intel.com>
14417         * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
14418         (*movdf_internal): Ditto.
14419         (*rcpsf2_sse): Ditto.
14420         (*rsqrtsf2_sse): Ditto.
14421         (*sqrt<mode>2_sse): Ditto.
14423 2018-05-21  Tamar Christina  <tamar.christina@arm.com>
14425         * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
14426         eor3q<mode>4.
14427         (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
14428         * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
14429         veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
14430         vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
14431         vbcaxq_s64): New.
14432         * config/aarch64/arm_neon.h: Likewise.
14433         * config/aarch64/iterators.md (VQ_I): New.
14435 2018-05-21  Alexey Brodkin <abrodkin@synopsys.com>
14437         * config.gcc: Add arc/t-multilib-linux to tmake_file for
14438         arc*-*-linux*.
14439         * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
14440         MULTILIB_DIRNAMES
14442 2018-05-20  Chung-Ju Wu  <jasonwucj@gmail.com>
14444         * config/nds32/constraints.md (S): New constraint.
14445         * config/nds32/nds32.md (call_internal): Use constraint S.
14446         (call_value_internal): Likewise.
14447         (sibcall_internal): Likewise.
14448         (sibcall_value_internal): Likewise.
14450 2018-05-20  Kito Cheng  <kito.cheng@gmail.com>
14451             Chung-Ju Wu  <jasonwucj@gmail.com>
14453         * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
14454         into consideration.
14456 2018-05-20  Kito Cheng  <kito.cheng@gmail.com>
14457             Chung-Ju Wu  <jasonwucj@gmail.com>
14459         * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
14460         (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
14461         (nds32_rtx_costs_impl): Simplify.
14462         (nds32_address_cost_impl): Simplify.
14463         (nds32_init_rtx_costs): New function.
14464         (nds32_rtx_costs_speed_prefer): Likewise.
14465         (nds32_rtx_costs_size_prefer): Likewise.
14466         (nds32_address_cost_speed_prefer): Likewise.
14467         (nds32_address_cost_speed_fwprop): Likewise.
14468         (nds32_address_cost_size_prefer): Likewise.
14469         * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
14470         * config/nds32/nds32.c (nds32_option_override): Use
14471         nds32_init_rtx_costs function.
14473 2018-05-20  Chung-Ju Wu  <jasonwucj@gmail.com>
14475         * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
14476         * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
14477         (TARGET_PIPELINE_N8): Likewise.
14478         (TARGET_PIPELINE_N10): Likewise.
14479         (TARGET_PIPELINE_N13): Likewise.
14480         (TARGET_PIPELINE_GRAYWOLF): Likewise.
14482 2018-05-19  Monk Chiang  <sh.chiang04@gmail.com>
14484         * config/nds32/nds32-fpu.md: Update copyright year.
14486 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
14488         * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
14490 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
14492         * config/nds32/nds32.c
14493         (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
14494         * config/nds32/nds32.opt (minline-asm-r15): New option.
14496 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
14498         * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
14499         MASK_HW_ABS.
14500         * config/nds32/nds32.md (abssi2): New pattern.
14502 2018-05-19  Uros Bizjak  <ubizjak@gmail.com>
14504         * config/i386/i386.md (rex64namesuffix): New mode attribute.
14505         * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
14506         Merge insn pattern from sse_cvtsi2ss<round_name> and
14507         sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
14508         (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
14509         from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
14510         using SWI48 mode iterator.
14511         (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
14512         sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
14513         (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
14514         pattern from sse_cvttss2si<round_saeonly_name>
14515         and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
14516         (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
14517         from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
14518         using SWI48 mode iterator.
14519         (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
14520         insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
14521         avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
14522         (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
14523         from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
14524         using SWI48 mode iterator.
14525         (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
14526         insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
14527         avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
14528         (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
14529         sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
14530         SWI48 mode iterator.
14531         (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
14532         sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
14533         (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
14534         pattern from sse_cvttsd2si<round_saeonly_name>
14535         and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
14537 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
14539         * config/nds32/nds32-md-auxiliary.c
14540         (nds32_valid_smw_lwm_base_p): Refine.
14541         (nds32_output_smw_single_word): Refine.
14542         (nds32_output_smw_double_word): New.
14543         * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
14545 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
14547         * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
14548         (nds32_output_stack_pop): Refine.
14549         (nds32_expand_unaligned_load): Refine.
14550         (nds32_expand_unaligned_store): Refine.
14552 2018-05-19  Kuan-Lin Chen <kuanlinchentw@gmail.com>
14553             Chung-Ju Wu  <jasonwucj@gmail.com>
14555         * config/nds32/constants.md: Add TP_REGNUM constant.
14556         (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
14557         UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
14558         UNSPEC_ADD32.
14559         * config/nds32/nds32-doubleword.md: Consider flag_pic.
14560         * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
14561         * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
14562         * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
14563         and PIC code generation.
14564         * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
14565         code generation.
14566         * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
14567         optimization.
14568         * config/nds32/nds32.md: Support TLS and PIC.
14569         * config/nds32/nds32.c: Support TLS and PIC.
14570         * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
14571         * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
14572         predicate.
14574 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
14576         * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
14577         mode with E_ prefix.
14579 2018-05-19  Kuan-Lin Chen <kuanlinchentw@gmail.com>
14580             Chung-Ju Wu  <jasonwucj@gmail.com>
14582         * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
14583         * config/nds32/nds32-md-auxiliary.c
14584         (symbolic_reference_mentioned_p): New.
14585         (nds32_legitimize_ict_address): New.
14586         (nds32_expand_ict_move): New.
14587         (nds32_indirect_call_referenced_p): New.
14588         (nds32_symbol_binds_local_p): Delete.
14589         (nds32_long_call_p): Modify.
14590         * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
14591         * config/nds32/nds32-protos.h
14592         (symbolic_reference_mentioned_p): Declare.
14593         (nds32_legitimize_ict_address): Declare.
14594         (nds32_expand_ict_move): Declare.
14595         (nds32_indirect_call_referenced_p): Declare.
14596         * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
14597         (nds32_relax_group): Use nds32_ict_const_p as condition.
14598         * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
14599         (nds32_asm_file_start): Output ict_model directive in asm code.
14600         (nds32_legitimate_address_p): Consider indirect call.
14601         (nds32_print_operand): Consider indirect call.
14602         (nds32_print_operand_address): Consider indirect call.
14603         (nds32_insert_attributes): Handle "indirect_call" attribute.
14604         (TARGET_LEGITIMATE_ADDRESS_P): Define.
14605         (TARGET_LEGITIMATE_CONSTANT_P): Define.
14606         (TARGET_CANNOT_FORCE_CONST_MEM): Define.
14607         (TARGET_DELEGITIMIZE_ADDRESS): Define.
14608         (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
14609         * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
14610         (TARGET_ICT_MODEL_SMALL): Define.
14611         (TARGET_ICT_MODEL_LARGE): Define.
14612         * config/nds32/nds32.md (movsi): Consider ict model.
14613         (call, call_value): Consider ict model.
14614         (sibcall, sibcall_value): Consider ict model.
14615         * config/nds32/nds32.opt (mict-model): New option.
14616         * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
14617         model.
14619 2018-05-18  Kito Cheng  <kito.cheng@gmail.com>
14620             Monk Chiang  <sh.chiang04@gmail.com>
14621             Jim Wilson <jimw@sifive.com>
14623         * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
14624         Add support to parse rv32e*.  Clear MASK_RVE for rv32i and rv64i.
14625         * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
14626         * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
14627         __riscv_32e when TARGET_RVE.  Handle ABI_ILP32E as soft-float ABI.
14628         * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
14629         * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
14630         compute save_libcall_adjustment properly.
14631         (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
14632         (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
14633         * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
14634         (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
14635         (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
14636         (ABI_SPEC): Handle mabi=ilp32e.
14637         * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
14638         (RVE): Add RVE mask.
14639         * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
14640         <-march>: Add rv32e as an example.
14642 2018-05-18  Marc Glisse  <marc.glisse@inria.fr>
14644         PR c++/82899
14645         * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
14646         (intra_create_variable_infos): Handle C++ constructors.
14648 2018-05-18  Martin Liska  <mliska@suse.cz>
14650         * passes.def: Remove a redundant pass.
14652 2018-05-18  Eric Botcazou  <ebotcazou@adacore.com>
14654         PR bootstrap/85838
14655         * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
14657 2018-05-18  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
14659         * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
14660         (ARMv4): Update.
14661         (ARMv2, ARMv3, ARMv3m): Delete fgroups.
14662         (ARMv6m): Update.
14663         (armv2, armv2a, armv3, armv3m): Delete architectures.
14664         (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
14665         arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
14666         arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
14667         Delete cpus.
14668         * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
14669         (*mulsidi3adddi): Likewise.
14670         (mulsidi3): Likewise.
14671         (*mulsidi3_nov6): Likewise.
14672         (umulsidi3): Likewise.
14673         (umulsidi3_nov6): Likewise.
14674         (umaddsidi4): Likewise.
14675         (*umulsidi3adddi): Likewise.
14676         (smulsi3_highpart): Likewise.
14677         (*smulsi3_highpart_nov6): Likewise.
14678         (umulsi3_highpart): Likewise.
14679         (*umulsi3_highpart_nov6): Likewise.
14680         * config/arm/arm.h (arm_arch3m): Delete.
14681         * config/arm/arm.c (arm_arch3m): Delete.
14682         (arm_option_override_internal): Update armv3-related comment.
14683         (arm_configure_build_target): Delete use of isa_bit_mode32.
14684         (arm_option_reconfigure_globals): Delete set of arm_ach3m.
14685         (arm_rtx_costs_internal): Delete check of arm_arch3m.
14686         * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
14687         (mulsa3): Likewise.
14688         (mulusa3): Likewise.
14689         * config/arm/arm-protos.h (arm_arch3m): Delete.
14690         * config/arm/arm-tables.opt: Regenerate.
14691         * config/arm/arm-tune.md: Likewise.
14692         * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
14693         deleted architectures.
14695 2018-05-18  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
14697         * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
14698         (armv5t, armv5te): New features.
14699         (ARMv5, ARMv5e): Delete fgroups.
14700         (ARMv5t, ARMv5te): Adjust for above changes.
14701         (ARMv6m): Likewise.
14702         (armv5, armv5e): Delete arches.
14703         * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
14704         arm_arch5.
14705         (*call_reg_arm): Likewise.
14706         (*call_value_reg_armv5): Likewise.
14707         (*call_value_reg_arm): Likewise.
14708         (*call_symbol): Likewise.
14709         (*call_value_symbol): Likewise.
14710         (*sibcall_insn): Likewise.
14711         (*sibcall_value_insn): Likewise.
14712         (clzsi2): Likewise.
14713         (prefetch): Likewise.
14714         (define_split and define_peephole2 dependent on arm_arch5):
14715         Likewise.
14716         * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
14717         arm_arch5e.
14718         (TARGET_ARM_QBIT): Likewise.
14719         (TARGET_DSP_MULTIPLY): Likewise.
14720         (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
14721         (arm_arch5, arm_arch5e): Delete.
14722         (arm_arch5t, arm_arch5te): Declare.
14723         * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
14724         (arm_arch5t): Declare.
14725         (arm_option_reconfigure_globals): Update for the above.
14726         (arm_options_perform_arch_sanity_checks): Update comment, replace
14727         use of arm_arch5 with arm_arch5t.
14728         (use_return_insn): Likewise.
14729         (arm_emit_call_insn): Likewise.
14730         (output_return_instruction): Likewise.
14731         (arm_final_prescan_insn): Likewise.
14732         (arm_coproc_builtin_available): Likewise.
14733         * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
14734         arm_arch5e with arm_arch5t and arm_arch5te.
14735         * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
14736         (arm_arch5t, arm_arch5te): Declare.
14737         * config/arm/arm-tables.opt: Regenerate.
14738         * config/arm/t-arm-elf: Remove references to armv5, armv5e.
14739         * config/arm/t-multilib: Likewise.
14740         * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
14741         instead of arm_arch5.
14742         (*call_reg_thumb1): Likewise.
14743         (*call_value_reg_thumb1_v5): Likewise.
14744         (*call_value_reg_thumb1): Likewise.
14745         * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
14746         unreachable path.
14747         * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
14749 2018-05-18  Martin Liska  <mliska@suse.cz>
14751         PR gcov-profile/84846
14752         * doc/gcov.texi: Document -t option of gcov tool.
14754 2018-05-18  Martin Liska  <mliska@suse.cz>
14756         PR gcov-profile/84846
14757         * gcov.c (print_usage): Add new -t option.
14758         (process_args): Handle the option.
14759         (generate_results): Use stdout as output when requested by
14760         the option.
14762 2018-05-18  Martin Liska  <mliska@suse.cz>
14764         PR gcov-profile/84846
14765         * coverage.c (coverage_init): Write PWD to .gcno file.
14766         * doc/gcov.texi: Document how working directory is printed.
14767         * gcov-dump.c (dump_gcov_file): Print PWD.
14768         * gcov.c (output_intermediate_file): Likewise.
14769         (read_graph_file): Read PWD string.
14770         (output_lines): Print PWD.
14772 2018-05-18  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
14774         PR middle-end/85817
14775         * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
14776         for retval and return false if all args to phi are zero.
14778 2018-05-18  Richard Biener  <rguenther@suse.de>
14780         * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
14781         method.
14782         (evrp_dom_walker::before_dom_children): Call it.
14784 2018-05-18  Richard Biener  <rguenther@suse.de>
14786         * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
14787         results when processing array refs with variable index.
14789 2018-05-18  Toon Moene  <toon@moene.org>
14791         * doc/invoke.texi: Move -floop-unroll-and-jam documentation
14792         directly after that of -floop-interchange. Indicate that both
14793         options are enabled by default when specifying -O3.
14795 2018-05-18  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
14797         * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
14798         iterator.  Delete separate integer-mode vec_set<mode> expander.
14799         (aarch64_simd_vec_setv2di): Delete.
14800         (vec_setv2di): Delete.
14801         (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
14802         Use VALL_F16 mode iterator.  Add LD1 alternative and use vwcore for
14803         the "w, r" alternative.
14805 2018-05-18  Martin Liska  <mliska@suse.cz>
14807         * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
14808         * tree-pass.h (make_pass_lower_switch_O0): New function.
14809         * tree-switch-conversion.c (node_has_low_bound): Remove.
14810         (node_has_high_bound): Likewise.
14811         (node_is_bounded): Likewise.
14812         (class pass_lower_switch): Make it a template type and create
14813         two instances.
14814         (pass_lower_switch::execute): Add template argument.
14815         (make_pass_lower_switch): New function.
14816         (make_pass_lower_switch_O0): New function.
14817         (do_jump_if_equal): Remove.
14818         (emit_case_nodes): Simplify to just handle all 3 cases and leave
14819         all the hard work to tree optimization passes.
14821 2018-05-18  Martin Liska  <mliska@suse.cz>
14823         * dbgcnt.c (limit_low): Renamed from limit.
14824         (limit_high): New variable.
14825         (dbg_cnt_is_enabled): Check for upper limit.
14826         (dbg_cnt): Adjust dumping.
14827         (dbg_cnt_set_limit_by_index): Add new argument for high
14828         value.
14829         (dbg_cnt_set_limit_by_name): Likewise.
14830         (dbg_cnt_process_single_pair): Parse new format.
14831         (dbg_cnt_process_opt): Use strtok.
14832         (dbg_cnt_list_all_counters): Remove 'value' and add
14833         'limit_high'.
14834         * doc/invoke.texi: Document changes.
14836 2018-05-18  Richard Sandiford  <richard.sandiford@linaro.org>
14838         * doc/sourcebuild.texi (scalar_all_fma): Document.
14839         * tree.def (FMA_EXPR): Delete.
14840         * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
14841         * internal-fn.c (ternary_direct): New macro.
14842         (expand_ternary_optab_fn): Likewise.
14843         (direct_ternary_optab_supported_p): Likewise.
14844         * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
14845         * builtins.c (fold_builtin_fma): Delete.
14846         (fold_builtin_3): Don't call it.
14847         * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
14848         * expr.c (expand_expr_real_2): Likewise.
14849         * fold-const.c (operand_equal_p): Likewise.
14850         (fold_ternary_loc): Likewise.
14851         * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
14852         * gimple.c (DEFTREECODE): Likewise.
14853         * gimplify.c (gimplify_expr): Likewise.
14854         * optabs-tree.c (optab_for_tree_code): Likewise.
14855         * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
14856         * tree-eh.c (operation_could_trap_p): Likewise.
14857         (stmt_could_throw_1_p): Likewise.
14858         * tree-inline.c (estimate_operator_cost): Likewise.
14859         * tree-pretty-print.c (dump_generic_node): Likewise.
14860         (op_code_prio): Likewise.
14861         * tree-ssa-loop-im.c (stmt_cost): Likewise.
14862         * tree-ssa-operands.c (get_expr_operands): Likewise.
14863         * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
14864         * fold-const-call.h (fold_fma): Delete.
14865         * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
14866         CFN_FNMA and CFN_FNMS.
14867         (fold_fma): Delete.
14868         * genmatch.c (combined_fn): New enum.
14869         (commutative_ternary_tree_code): Remove FMA_EXPR handling.
14870         (commutative_op): New function.
14871         (commutate): Use it.  Handle more than 2 operands.
14872         (dt_operand::gen_gimple_expr): Use commutative_op.
14873         (parser::parse_expr): Allow :c to be used with non-binary
14874         operators if the commutative operand is known.
14875         * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
14876         CFN_FMS, CFN_FNMA and CFN_FNMS.
14877         (backprop::process_assign_use): Remove FMA_EXPR handling.
14878         * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
14879         (gen_hsa_fma): New function.
14880         (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
14881         IFN_FNMA and IFN_FNMS.
14882         * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
14883         * gimple-fold.h (follow_all_ssa_edges): Declare.
14884         * gimple-fold.c (follow_all_ssa_edges): New function.
14885         * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
14886         gimple_build interface and use follow_all_ssa_edges to fold the result.
14887         (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
14888         instead of checking for optabs directly.
14889         * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
14890         rather than FMA_EXPRs.
14891         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
14892         call to IFN_FMA instead of an FMA_EXPR.
14894 2018-05-17  Jim Wilson  <jimw@sifive.com>
14896         * expr.c (do_tablejump): When converting index to Pmode, if we have a
14897         sign extended promoted subreg, and the range does not have the sign bit
14898         set, then do a sign extend.
14900         * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
14901         test, check for sign extended subreg and/or constant operands, and
14902         do a sign extend in that case.
14904 2018-05-17  Steve Ellcey  <sellcey@cavium.com>
14906         * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
14907         (thunderx2t99_multiple): Delete psuedo-units from used cpus.
14908         Add untyped.
14909         (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
14910         Change logics_shift_reg to logics_shift_imm.
14911         (thunderx2t99_fp_loadpair_basic): Delete.
14912         (thunderx2t99_fp_storepair_basic): Delete.
14913         (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
14914         (thunderx2t99_asimd_polynomial): Delete.
14915         (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
14916         and neon_fp_mul_d_scalar_q.
14917         (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
14918         (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
14919         (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
14920         (thunderx2t99_asimd_lut): Add missing tbl types.
14921         (thunderx2t99_asimd_ext): Delete.
14922         (thunderx2t99_asimd_load1_1_mult): Delete.
14923         (thunderx2t99_asimd_load1_2_mult): Delete.
14924         (thunderx2t99_asimd_load1_ldp): New.
14925         (thunderx2t99_asimd_load1): New.
14926         (thunderx2t99_asimd_load2): Add missing *load2* types.
14927         (thunderx2t99_asimd_load3): New.
14928         (thunderx2t99_asimd_load4): New.
14929         (thunderx2t99_asimd_store1_1_mult): Delete.
14930         (thunderx2t99_asimd_store1_2_mult): Delete.
14931         (thunderx2t99_asimd_store2_mult): Delete.
14932         (thunderx2t99_asimd_store2_onelane): Delete.
14933         (thunderx2t99_asimd_store_stp): New.
14934         (thunderx2t99_asimd_store1): New.
14935         (thunderx2t99_asimd_store2): New.
14936         (thunderx2t99_asimd_store3): New.
14937         (thunderx2t99_asimd_store4): New.
14939 2018-05-17  Jerome Lambourg  <lambourg@adacore.com>
14941         * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
14942         #include <stdint.h>.  Replace intptr_t with __INTPTR_TYPE__.
14944 2018-05-17  Pat Haugen  <pthaugen@us.ibm.com>
14945             Segher Boessenkool  <segher@kernel.crashing.org>
14947         PR target/85698
14948         * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
14949         operand.
14951 2018-05-17  Richard Biener  <rguenther@suse.de>
14953         * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
14954         for pruning loop and prune defs feeding only already visited PHIs.
14956 2018-05-17  Richard Biener  <rguenther@suse.de>
14958         * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
14960 2018-05-17  Bin Cheng  <bin.cheng@arm.com>
14961             Richard Biener  <rguenther@suse.de>
14963         PR tree-optimization/85793
14964         * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
14965         for VMAT_ELEMENTWISE.
14967 2018-05-17  Richard Sandiford  <richard.sandiford@linaro.org>
14969         * internal-fn.h (lookup_internal_fn): Declare
14970         * internal-fn.c (lookup_internal_fn): New function.
14971         * gimple.c (gimple_build_call_from_tree): Handle calls to
14972         internal functions.
14973         * gimple-pretty-print.c (dump_gimple_call): Print "." before
14974         internal function names.
14975         * tree-pretty-print.c (dump_generic_node): Likewise.
14976         * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
14978 2018-05-17  Richard Sandiford  <richard.sandiford@linaro.org>
14980         * gimple-fold.h (gimple_build): Make the function forms take
14981         combined_fn rather than built_in_function.
14982         (gimple_simplify): Likewise.
14983         * gimple-match-head.c (gimple_simplify): Likewise.
14984         * gimple-fold.c (gimple_build): Likewise.
14985         * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
14986         rather than gimple_build_call_internal.
14987         (get_initial_defs_for_reduction): Likewise.
14988         (vect_create_epilog_for_reduction): Likewise.
14989         (vectorizable_live_operation): Likewise.
14991 2018-05-17  Martin Liska  <mliska@suse.cz>
14993         * gimple-ssa-sprintf.c (format_directive): Do not use
14994         space in between 'G_' and '('.
14996 2018-05-17  Jakub Jelinek  <jakub@redhat.com>
14998         PR target/85323
14999         * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
15000         even if the mask is not all ones.
15002         PR target/85323
15003         * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
15004         vector.
15005         (ix86_gimple_fold_builtin): Likewise.
15007         PR target/85323
15008         * config/i386/i386.c: Include tree-vector-builder.h.
15009         (ix86_vector_shift_count): New function.
15010         (ix86_fold_builtin): Fold shift builtins by scalar count.
15011         (ix86_gimple_fold_builtin): Likewise.
15013         * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
15014         _mm512_setzero): New intrinsics.
15016 2018-05-17  James Greenhalgh  <james.greenhalgh@arm.com>
15017             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
15019         * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
15020         code generation for cases where splatting a value is not useful.
15021         * simplify-rtx.c (simplify_ternary_operation): Simplify
15022         vec_merge across a vec_duplicate and a paradoxical subreg forming
15023         a vector mode to a vec_concat.
15025 2018-05-17  Olga Makhotina  <olga.makhotina@intel.com>
15027         * config.gcc: Support "goldmont-plus".
15028         * config/i386/driver-i386.c (host_detect_local_cpu): Detect
15029         "goldmont-plus".
15030         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
15031         PROCESSOR_GOLDMONT_PLUS.
15032         * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
15033         (processor_target_table): Add "goldmont-plus".
15034         (PTA_GOLDMONT_PLUS): Define.
15035         (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
15036         (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
15037         (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
15038         (fold_builtin_cpu): Add "goldmont-plus".
15039         (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
15040         (ix86_option_override_internal): Add "goldmont-plus".
15041         * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
15042         (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
15043         * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
15044         * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
15046 2018-05-17  Richard Biener  <rguenther@suse.de>
15048         PR tree-optimization/85757
15049         * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
15050         remove defs that only feed that PHI from further processing.
15052 2018-05-16  Jim Wilson  <jimw@sifive.com>
15054         * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
15055         asterisk to name.
15056         (<optab>di3_mask, <optab>di3_mask_1): Likewise.
15058 2018-05-16  Mark Wielaard  <mark@klomp.org>
15060         * dwarf2out.c (count_index_strings): New function.
15061         (output_indirect_strings): Call count_index_strings and generate
15062         header for dwarf_version >= 5.
15064 2018-05-16  Mark Wielaard  <mark@klomp.org>
15066         * dwarf2out.c (dwarf_FORM): New function.
15067         (set_indirect_string): Use dwarf_FORM.
15068         (reset_indirect_string): Likewise.
15069         (size_of_die): Likewise.
15070         (value_format): Likewise.
15071         (output_die): Likewise.
15072         (add_skeleton_AT_string): Likewise.
15073         (output_macinfo_op): Likewise.
15074         (index_string): Likewise.
15075         (output_index_string_offset): Likewise.
15076         (output_index_string): Likewise.
15077         (count_index_strings): Likewise.
15079 2018-05-16  Carl Love  <cel@us.ibm.com>
15081         * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
15082         dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
15084 2018-05-16  Martin Jambor  <mjambor@suse.cz>
15086         * ipa-prop.c (ipa_free_all_edge_args): Remove.
15087         * ipa-prop.h (ipa_free_all_edge_args): Likewise.
15089 2018-05-16  Wilco Dijkstra  <wdijkstr@arm.com>
15091         * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
15092         (fnma<mode>4): Likewise.
15093         (fms<mode>4): Likewise.
15094         (fnms<mode>4): Likewise.
15095         (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
15096         (aarch64_fnma<mode>4): Likewise.
15097         (aarch64_fms<mode>4): Likewise.
15098         (aarch64_fnms<mode>4): Likewise.
15099         (aarch64_fnmadd<mode>4): Likewise.
15101 2018-05-16  Jason Merrill  <jason@redhat.com>
15103         * tree.c (warn_deprecated_use): Return bool.  Simplify logic.
15105 2018-05-16  Richard Biener  <rguenther@suse.de>
15107         * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
15108         (dump_stmt_cost): Declare.
15109         (add_stmt_cost): Dump cost we add.
15110         (add_stmt_costs): New function.
15111         (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
15112         No longer exported.
15113         (vect_analyze_stmt): Adjust prototype.
15114         (vectorizable_condition): Likewise.
15115         (vectorizable_live_operation): Likewise.
15116         (vectorizable_reduction): Likewise.
15117         (vectorizable_induction): Likewise.
15118         * tree-vect-loop.c (vect_analyze_loop_operations): Create local
15119         cost vector to pass to vectorizable_ and record afterwards.
15120         (vect_model_reduction_cost): Take cost vector argument and adjust.
15121         (vect_model_induction_cost): Likewise.
15122         (vectorizable_reduction): Likewise.
15123         (vectorizable_induction): Likewise.
15124         (vectorizable_live_operation): Likewise.
15125         * tree-vect-slp.c (vect_create_new_slp_node): Initialize
15126         SLP_TREE_NUMBER_OF_VEC_STMTS.
15127         (vect_analyze_slp_cost_1): Remove.
15128         (vect_analyze_slp_cost): Likewise.
15129         (vect_slp_analyze_node_operations): Take visited args and
15130         a target cost vector.  Avoid processing already visited stmt sets.
15131         (vect_slp_analyze_operations): Use a local cost vector to gather
15132         costs and register those of non-discarded instances.
15133         (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
15134         (vect_schedule_slp_instance): Remove copying of
15135         SLP_TREE_NUMBER_OF_VEC_STMTS.  Instead assert that it is not
15136         zero.
15137         * tree-vect-stmts.c (record_stmt_cost): Remove path directly
15138         adding cost.  Record cost entry location.
15139         (vect_prologue_cost_for_slp_op): Function to compute cost of
15140         a constant or invariant generated for SLP vect in the prologue,
15141         split out from vect_analyze_slp_cost_1.
15142         (vect_model_simple_cost): Make static.  Adjust for SLP costing.
15143         (vect_model_promotion_demotion_cost): Likewise.
15144         (vect_model_store_cost): Likewise, make static.
15145         (vect_model_load_cost): Likewise.
15146         (vectorizable_bswap): Add cost vector arg and adjust.
15147         (vectorizable_call): Likewise.
15148         (vectorizable_simd_clone_call): Likewise.
15149         (vectorizable_conversion): Likewise.
15150         (vectorizable_assignment): Likewise.
15151         (vectorizable_shift): Likewise.
15152         (vectorizable_operation): Likewise.
15153         (vectorizable_store): Likewise.
15154         (vectorizable_load): Likewise.
15155         (vectorizable_condition): Likewise.
15156         (vectorizable_comparison): Likewise.
15157         (can_vectorize_live_stmts): Likewise.
15158         (vect_analyze_stmt): Likewise.
15159         (vect_transform_stmt): Adjust calls to vectorizable_*.
15160         * tree-vectorizer.c: Include gimple-pretty-print.h.
15161         (dump_stmt_cost): New function.
15163 2018-05-16  Richard Biener  <rguenther@suse.de>
15165         * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
15166         * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
15167         * tree-ssa-dse.c: Include tree-ssa-loop.h.
15168         (check_name): New callback.
15169         (dse_classify_store): Track cycles via a visited bitmap of PHI
15170         defs and simplify handling of in-loop and across loop dead stores
15171         and properly fail for loop-variant refs.  Handle byte-tracking with
15172         multiple defs.  Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
15173         limiting the walk.
15175 2018-05-16  Richard Sandiford  <richard.sandiford@linaro.org>
15177         * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
15178         (vect_get_mask_type_for_stmt): Likewise.
15179         * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
15180         split out from...
15181         (vect_build_slp_tree_1): ...here.  Use vect_get_vector_types_for_stmt
15182         to determine the statement's vector type and the vector type that
15183         should be used for calculating nunits.  Deal with cases in which
15184         the type has to be deferred.
15185         (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
15186         and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
15187         * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
15188         (vect_determine_vf_for_stmt): New functions, split out from...
15189         (vect_determine_vectorization_factor): ...here.
15190         * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
15191         (vect_get_mask_type_for_stmt): New functions, split out from
15192         vect_determine_vectorization_factor.
15194 2018-05-16  Richard Biener  <rguenther@suse.de>
15196         * tree-cfg.c (verify_gimple_assign_ternary): Properly
15197         verify the [VEC_]COND_EXPR embedded comparison.
15199 2018-05-15  Martin Sebor  <msebor@redhat.com>
15201         PR tree-optimization/85753
15202         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
15203         RECORD_TYPE in addition to ARRAY_TYPE.
15205 2018-05-15  Martin Sebor  <msebor@redhat.com>
15207         PR middle-end/85643
15208         * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
15210 2018-05-15  Richard Biener  <rguenther@suse.de>
15212         * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
15213         add by_clobber_p one.  Change algorithm to collect all defs
15214         representing uses we need to walk and try reducing them to
15215         a single one before failing.
15216         (dse_dom_walker::dse_optimize_stmt): Adjust.
15218 2018-05-13  Mark Wielaard  <mark@klomp.org>
15220         * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
15221         (size_of_loc_descr): Likewise.
15222         (output_loc_operands): Likewise.
15223         (output_loc_operands_raw): Likewise.
15224         (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
15225         (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
15226         (hash_loc_operands): Likewise.
15227         (compare_loc_operands): Likewise.
15229 2018-05-14  Mark Wielaard  <mark@klomp.org>
15231         * dwarf2out.c (count_index_addrs): New function.
15232         (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
15234 2018-05-15  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
15236         PR tree-optimization/83648
15237         * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
15238         return value as malloc candidate.
15240 2018-05-15  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
15242         PR ipa/85734
15243         * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite
15244         param as true in call to suggest_attribute.
15246 2018-05-14  Segher Boessenkool  <segher@kernel.crashing.org>
15248         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
15249         -mreadonly-in-sdata.
15251 2018-05-14  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
15253         * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
15254         New pattern.
15255         (aarch64_crypto_aesd_fused): Likewise.
15257 2018-05-14  Wilco Dijkstra  <wdijkstr@arm.com>
15259         * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
15260         (movsi_aarch64): Likewise.
15261         (load_pairsi): Likewise.
15262         (load_pairdi): Likewise.
15263         (store_pairsi): Likewise.
15264         (store_pairdi): Likewise.
15265         (load_pairsf): Likewise.
15266         (load_pairdf): Likewise.
15267         (store_pairsf): Likewise.
15268         (store_pairdf): Likewise.
15269         (zero_extend): Likewise.
15270         (trunc): Swap alternatives.
15271         (fcvt_target): Add '?' to prefer w over r.
15273 2018-05-14  Jakub Jelinek  <jakub@redhat.com>
15275         PR target/85756
15276         * config/i386/i386.md: Disallow non-commutative arithmetics in
15277         last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
15278         optimization.  Use COMMUTATIVE_ARITH_P test rather than != MINUS
15279         in the peephole2 before it.
15281 2018-05-14  Sebastian Peryt  <sebastian.peryt@intel.com>
15283         * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
15284         OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
15285         (ix86_handle_option): Handle -mcldemote.
15286         * config.gcc: New header.
15287         * config/i386/cldemoteintrin.h: New file.
15288         * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
15289         * config/i386/driver-i386.c (host_detect_local_cpu): Detect
15290         -mcldemote.
15291         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
15292         OPTION_MASK_ISA_CLDEMOTE.
15293         * config/i386/i386.c (ix86_target_string): Add -mcldemote.
15294         (ix86_valid_target_attribute_inner_p): Ditto.
15295         (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
15296         (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
15297         (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
15298         * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
15299         * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
15300         (cldemote): New.
15301         * config/i386/i386.opt: Add -mcldemote.
15302         * config/i386/x86intrin.h: New header.
15303         * doc/invoke.texi: Add -mcldemote.
15305 2018-05-14  Richard Biener  <rguenther@suse.de>
15307         * doc/match-and-simplify.texi: Adjust :s documentation.
15309 2018-05-14  Alexander Monakov  <amonakov@ispras.ru>
15311         * sort.cc (REORDER_23): Pass the type for the temporaries instead of
15312         intended memcpy size.
15313         (REORDER_45): Likewise.
15315 2018-05-13  Alexander Monakov  <amonakov@ispras.ru>
15317         * sort.cc: New file.
15318         * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
15319         * vec.c (qsort_chk): Use gcc_qsort.
15320         * Makefile.in (OBJS-libcommon): Add sort.o.
15321         (build/sort.o): New target.  Use it...
15322         (BUILD_RTL): ... here, and...
15323         (build/gencfn-macros): ... here, and...
15324         (build/genmatch): ... here.
15326 2018-05-13  Kito Cheng  <kito.cheng@gmail.com>
15327             Chung-Ju Wu  <jasonwucj@gmail.com>
15329         * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
15330         * config/nds32/nds32-graywolf.md: New file.
15331         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
15332         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
15333         pipeline.
15334         * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
15335         * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
15336         * config/nds32/nds32.md (pipeline_model): Add graywolf.
15337         * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
15338         * config/nds32/pipelines.md: Include n15 settings.
15340 2018-05-13  Kito Cheng  <kito.cheng@gmail.com>
15341             Chung-Ju Wu  <jasonwucj@gmail.com>
15343         * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
15344         * config/nds32/nds32-n13.md: New file.
15345         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
15346         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
15347         pipeline.
15348         * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
15349         * config/nds32/nds32.md (pipeline_model): Add n13.
15350         * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
15351         * config/nds32/pipelines.md: Include n13 settings.
15353 2018-05-13  Kito Cheng  <kito.cheng@gmail.com>
15354             Chung-Ju Wu  <jasonwucj@gmail.com>
15356         * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
15357         * config/nds32/nds32-n10.md: New file.
15358         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
15359         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
15360         pipeline.
15361         * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
15362         * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
15363         * config/nds32/nds32.md (pipeline_model): Add n10.
15364         * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
15365         * config/nds32/pipelines.md: Include n10 settings.
15367 2018-05-13  Monk Chiang  <sh.chiang04@gmail.com>
15368             Kito Cheng  <kito.cheng@gmail.com>
15369             Chung-Ju Wu  <jasonwucj@gmail.com>
15371         * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
15372         * config/nds32/constants.md (unspec_element, unspec_volatile_element):
15373         Add enum values for DSP extension instructions.
15374         * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
15375         New constraints.
15376         * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
15377         sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
15378         New code iterators.
15379         (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
15380         * config/nds32/nds32-dspext.md: New file for DSP implementation.
15381         * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
15382         * config/nds32/nds32-intrinsic.md: Likewise.
15383         * config/nds32/nds32_intrinsic.h: Likewise.
15384         * config/nds32/nds32-md-auxiliary.c: Likewise.
15385         * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
15386         * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
15387         (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
15388         (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
15389         * config/nds32/nds32-protos.h: New declarations for DSP extension.
15390         * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
15391         TYPE_DMAC in switch statement.
15392         * config/nds32/nds32.c: New checking and implementation for DSP
15393         extension instructions.
15394         * config/nds32/nds32.h: Likewise.
15395         * config/nds32/nds32.md: Likewise.
15396         * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
15397         * config/nds32/predicates.md: Implement new predicates for DSP
15398         extension.
15400 2018-05-11  Michael Meissner  <meissner@linux.vnet.ibm.com>
15402         * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
15403         Reformat alternatives and attributes so it is easier to identify
15404         which constraints/attributes go with which instruction.
15405         (mov<mode>_hardfloat32, FMOVE64): Likewise.
15406         (mov<mode>_softfloat32, FMOVE64): Likewise.
15407         (mov<mode>_hardfloat64, FMOVE64): Likewise.
15408         (mov<mode>_softfloat64, FMOVE64): Likewise.
15410 2018-05-11  Kelvin Nilsen  <kelvin@gcc.gnu.org>
15412         * doc/extend.texi (PowerPC Built-in Functions): Rename this
15413         subsection.
15414         (Basic PowerPC Built-in Functions): The new name of the
15415         subsection previously known as "PowerPC Built-in Functions".
15416         (Basic PowerPC Built-in Functions Available on all Configurations):
15417         New subsubsection.
15418         (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
15419         (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
15420         (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
15421         (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
15423 2018-05-11  Martin Jambor  <mjambor@suse.cz>
15425         PR ipa/85655
15426         * ipa-cp.c (intersect_with_plats): Check that the lattice contains
15427         single const.
15429 2018-05-11  Richard Earnshaw  <rearnsha@arm.com>
15431         PR target/85733
15432         * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
15434 2018-05-11  Sebastian Peryt  <sebastian.peryt@intel.com>
15436         * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
15437         OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
15438         (ix86_handle_option): Handle -mwaitpkg.
15439         * config.gcc: New header.
15440         * config/i386/cpuid.h (bit_WAITPKG): New bit.
15441         * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
15442         * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
15443         function type.
15444         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
15445         OPTION_MASK_ISA_WAITPKG.
15446         * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
15447         (ix86_option_override_internal): Add PTA_WAITPKG.
15448         (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
15449         (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
15450         IX86_BUILTIN_TPAUSE.
15451         (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
15452         __builtin_ia32_umwait and __builtin_ia32_tpause.
15453         (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
15454         IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
15455         * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
15456         * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
15457         UNSPECV_TPAUSE): New.
15458         (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
15459         * config/i386/i386.opt: Add -mwaitpkg.
15460         * config/i386/waitpkgintrin.h: New file.
15461         * config/i386/x86intrin.h: New header.
15462         * doc/invoke.texi: Add -mwaitpkg.
15464 2018-05-11  Richard Earnshaw  <rearnsha@arm.com>
15466         PR target/85606
15467         * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
15468         equivalent.
15469         (cortex-m0): Use armv6s-m isa.
15470         (cortex-m0plus): Likewise.
15471         (cortex-m1): Likewise.
15472         (cortex-m0.small-multiply): Likewise.
15473         (cortex-m0plus.small-multiply): Likewise.
15474         (cortex-m1.small-multiply): Likewise.
15476 2018-05-11  Allan Sandfeld Jensen  <allan.jensen@qt.io>
15477             Jakub Jelinek  <jakub@redhat.com>
15479         PR tree-optimization/85692
15480         * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
15481         source permute as well.
15483 2018-05-11  Martin Liska  <mliska@suse.cz>
15485         PR sanitizer/85556
15486         * doc/extend.texi: Document LLVM style format for no_sanitize
15487         attribute.
15489 2018-05-10  Michael Meissner  <meissner@linux.vnet.ibm.com>
15491         * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
15492         mode_supports_vsx_dform_quad to mode_supports_dq_form.
15493         (mode_supports_vsx_dform_quad): Likewise.
15494         (mode_supports_vmx_dform): Move these functions to be next to the
15495         other mode_supports functions.
15496         (mode_supports_dq_form): Likewise.
15497         (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
15498         mode_supports_dq_form.
15499         (reg_offset_addressing_ok_p): Likewise.
15500         (offsettable_ok_by_alignment): Likewise.
15501         (rs6000_legitimate_offset_address_p): Likewise.
15502         (legitimate_lo_sum_address_p): Likewise.
15503         (rs6000_legitimize_address): Likewise.
15504         (rs6000_legitimize_reload_address): Likewise.
15505         (rs6000_secondary_reload_inner): Likewise.
15506         (rs6000_preferred_reload_class): Likewise.
15507         (rs6000_output_move_128bit): Likewise.
15509 2018-05-10  Uros Bizjak  <ubizjak@gmail.com>
15511         * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
15512         Generate SImode target register for null target.
15513         <case IX86_BUILTIN_XGETBV>: Ditto.
15514         <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
15515         * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
15517 2018-05-10  Carl Love  <cel@us.ibm.com>
15519         * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
15520         dcbtt and dcbtstt if operands[2] is 0.
15522 2018-05-10  Uros Bizjak  <ubizjak@gmail.com>
15524         PR target/85693
15525         * config/i386/sse.md (usadv64qi): New expander.
15527 2018-05-10  Segher Boessenkool  <segher@kernel.crashing.org>
15529         * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
15530         altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
15531         -maltivec=be support.
15532         (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
15533         vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
15534         vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
15535         vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
15536         vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
15537         vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
15538         altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
15539         altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
15540         altivec_vsumsws): Adjust.
15541         (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
15542         *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
15543         altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
15544         support.
15545         (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
15546         altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
15547         altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
15548         (altivec_lve<VI_char>x): Delete expand.
15549         (*altivec_lve<VI_char>x_internal): Rename to...
15550         (altivec_lve<VI_char>x): ... this.
15551         (altivec_lvxl_<mode>): Delete expand.
15552         (*altivec_lvxl_<mode>_internal): Rename to ...
15553         (altivec_lvxl_<mode>): ... this.
15554         (altivec_stvxl_<mode>): Delete expand.
15555         (*altivec_stvxl_<mode>_internal): Rename to ...
15556         (altivec_stvxl_<mode>): ... this.
15557         (altivec_stve<VI_char>x): Delete expand.
15558         (*altivec_stve<VI_char>x_internal): Rename to ...
15559         (altivec_stve<VI_char>x): ... this.
15560         (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
15561         doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
15562         reduc_plus_scal_<mode>): Adjust.
15563         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
15564         comment.
15565         (rs6000_cpu_cpp_builtins): Adjust.
15566         (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
15567         * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
15568         altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
15569         * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
15570         -maltivec=be support.
15571         (rs6000_split_vec_extract_var): Adjust.
15572         (rs6000_split_v4si_init): Adjust.
15573         (swap_selector_for_mode): Delete.
15574         (altivec_expand_lvx_be, altivec_expand_stvx_be,
15575         altivec_expand_stvex_be): Delete.
15576         (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
15577         -maltivec=be support.
15578         (rs6000_gimple_fold_builtin): Ditto.
15579         (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
15580         Adjust.
15581         * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
15582         (TARGET_DIRECT_MOVE_64BIT): Adjust.
15583         * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
15584         * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
15585         * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
15586         unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
15587         vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
15588         *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
15589         *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
15590         *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
15591         anonymous split): Adjust.
15592         (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
15593         (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
15595 2018-05-10  Eric Botcazou  <ebotcazou@adacore.com>
15597         * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
15598         when --with-gxx-include-dir is also specified.
15599         * configure: Regenerate.
15601 2018-05-09  Jim Wilson  <jimw@sifive.com>
15603         PR target/84797
15604         * config.gcc (riscv*-*-*): Handle --with-multilib-list.
15605         * config/riscv/t-withmultilib: New.
15606         * config/riscv/withmultilib.h: New.
15607         * doc/install.texi: Document RISC-V --with-multilib-list support.
15609 2018-05-09  Richard Biener  <rguenther@suse.de>
15611         * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
15612         vector.
15613         (vect_bb_vectorization_profitable_p): Adjust.  Compute
15614         actual scalar cost using the cost vector and the add_stmt_cost
15615         machinery.
15617 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
15619         PR rtl-optimization/85645
15620         * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
15621         in the REG_CFA_REGISTER note for LR, don't leave it empty.
15623 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
15625         PR rtl-optimization/85645
15626         * shrink-wrap.c (spread_components): Return a boolean saying if
15627         anything was changed.
15628         (try_shrink_wrapping_separate): Iterate spread_components until
15629         nothing changes anymore.
15631 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
15633         PR rtl-optimization/85645
15634         * regrename.c (build_def_use): Also kill the chains that include the
15635         destination of a REG_CFA_REGISTER note.
15637 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
15639         PR rtl-optimization/85645
15640         *  regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
15641         insn that has a REG_CFA_REGISTER note.
15643 2018-05-09  Richard Sandiford  <richard.sandiford@linaro.org>
15645         * cfgexpand.c (expand_clobber): New function.
15646         (expand_gimple_stmt_1): Use it.
15647         * tree-vect-stmts.c (vect_clobber_variable): New function,
15648         split out from...
15649         (vectorizable_simd_clone_call): ...here.
15650         (vectorizable_store): Emit a clobber either side of an
15651         IFN_STORE_LANES sequence.
15652         (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
15654 2018-05-09  Tom de Vries  <tom@codesourcery.com>
15656         PR target/85626
15657         * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
15658         (define_insn "trap_if_false"): Add exit after trap.
15660 2018-05-09  Eric Botcazou  <ebotcazou@adacore.com>
15662         PR rtl-optimization/85638
15663         * bb-reorder.c: Include common/common-target.h.
15664         (create_forwarder_block): New function extracted from...
15665         (fix_up_crossing_landing_pad): ...here.  Rename into...
15666         (dw2_fix_up_crossing_landing_pad): ...this.
15667         (sjlj_fix_up_crossing_landing_pad): New function.
15668         (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
15669         call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
15670         from both partitions and exit the loop after one iteration.
15672 2018-05-08  Kelvin Nilsen  <kelvin@gcc.gnu.org>
15674         Revert:
15675         * doc/extend.texi (PowerPC Built-in Functions): Rename this
15676         subsection.
15677         (Basic PowerPC Built-in Functions): The new name of the
15678         subsection previously known as "PowerPC Built-in Functions".
15679         (Basic PowerPC Built-in Functions Available on all Configurations):
15680         New subsubsection.
15681         (Basic PowerPC Built-in Functions Available on ISA 2.05): New
15682         subsubsection.
15683         (Basic PowerPC Built-in Functions Available on ISA 2.06): New
15684         subsubsection.
15685         (Basic PowerPC Built-in Functions Available on ISA 2.07): New
15686         subsubsection.
15687         (Basic PowerPC Built-in Functions Available on ISA 3.0): New
15688         subsubsection.
15690 2018-05-08  Jim Wilson  <jimw@sifive.com>
15692         * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
15693         (LD_EMUL_SUFFIX): New.
15694         (LINK_SPEC): Use it.
15696 2018-05-08  Kelvin Nilsen  <kelvin@gcc.gnu.org>
15698         * doc/extend.texi (PowerPC Built-in Functions): Rename this
15699         subsection.
15700         (Basic PowerPC Built-in Functions): The new name of the
15701         subsection previously known as "PowerPC Built-in Functions".
15702         (Basic PowerPC Built-in Functions Available on all Configurations):
15703         New subsubsection.
15704         (Basic PowerPC Built-in Functions Available on ISA 2.05): New
15705         subsubsection.
15706         (Basic PowerPC Built-in Functions Available on ISA 2.06): New
15707         subsubsection.
15708         (Basic PowerPC Built-in Functions Available on ISA 2.07): New
15709         subsubsection.
15710         (Basic PowerPC Built-in Functions Available on ISA 3.0): New
15711         subsubsection.
15713 2018-05-08  Jakub Jelinek  <jakub@redhat.com>
15715         PR target/85683
15716         * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
15717         after cmpelim optimization.
15719 2018-05-08  Olga Makhotina  <olga.makhotina@intel.com>
15721         * config.gcc: Support "goldmont".
15722         * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
15723         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
15724         PROCESSOR_GOLDMONT.
15725         * config/i386/i386.c (m_GOLDMONT): Define.
15726         (processor_target_table): Add "goldmont".
15727         (PTA_GOLDMONT): Define.
15728         (ix86_lea_outperforms): Add TARGET_GOLDMONT.
15729         (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
15730         (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
15731         (fold_builtin_cpu): Add "goldmont".
15732         (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
15733         (ix86_option_override_internal): Add "goldmont".
15734         * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
15735         (processor_type): Add PROCESSOR_GOLDMONT.
15736         * config/i386/i386.md: Add CPU "glm".
15737         * config/i386/glm.md: New file.
15738         * config/i386/x86-tune.def: Add m_GOLDMONT.
15739         * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
15741 2018-05-08  Jakub Jelinek  <jakub@redhat.com>
15743         PR target/85572
15744         * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
15745         E_V4DImode.
15746         * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
15747         VI1248_AVX512VL_AVX512BW.  Handle V2DImode and V4DImode if not
15748         TARGET_AVX512VL using ix86_expand_sse2_abs.  Formatting fixes.
15750         PR target/85317
15751         * config/i386/i386.c (ix86_fold_builtin): Handle
15752         IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
15754         PR target/85480
15755         * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
15756         (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
15758 2018-05-08  Richard Earnshaw  <rearnsha@arm.com>
15760         PR target/85658
15761         * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
15762         (check_arch): Likewise.
15763         (check_fpu): Return the result rather than printing it.
15764         (end arch): Fix operator precedence.
15765         (end cpu): Likewise.
15766         (END): Print the result from check_fpu.
15768 2018-05-08  Richard Sandiford  <richard.sandiford@linaro.org>
15769             Alan Hayward  <alan.hayward@arm.com>
15770             David Sherwood  <david.sherwood@arm.com>
15772         * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
15773         (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
15774         (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
15775         (*fcmuo<mode>_and): New patterns.
15777 2018-05-08  Richard Sandiford  <richard.sandiford@linaro.org>
15779         * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
15780         (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
15781         (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
15782         (cmp_op, sve_imm_con): New code attributes.
15783         (SVE_COND_INT_CMP, imm_con): Delete.
15784         (cmp_op): Remove above unspecs from int attribute.
15785         * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
15786         to...
15787         (*cmp<cmp_op><mode>): ...this.  Use UNSPEC_MERGE_PTRUE instead of
15788         comparison-specific unspecs.
15789         (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
15790         (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
15791         (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
15792         (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
15793         (*vec_fcm<cmp_op><mode>): Rename to...
15794         (*fcm<cmp_op><mode>): ...this and adjust likewise.
15795         (*vec_fcmuo<mode>): Rename to...
15796         (*fcmuo<mode>): ...this and adjust likewise.
15797         (*pred_fcm<cmp_op><mode>): New pattern.
15798         * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
15799         (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
15800         functions.
15801         (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
15802         and UNORDERED.
15803         (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
15804         (aarch64_emit_sve_predicated_cond): New function.
15805         (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
15806         (aarch64_emit_unspec_cond_or): Replace with...
15807         (aarch64_emit_sve_or_conds): ...this new function.  Use
15808         aarch64_emit_sve_ptrue_op for the individual comparisons and
15809         aarch64_emit_binop to OR them together.
15810         (aarch64_emit_inverted_unspec_cond): Replace with...
15811         (aarch64_emit_sve_inverted_cond): ...this new function.  Use
15812         aarch64_emit_sve_ptrue_op for the comparison and
15813         aarch64_emit_unop to invert the result.
15814         (aarch64_expand_sve_vec_cmp_float): Update after the above
15815         changes.  Use aarch64_emit_sve_ptrue_op for native comparisons.
15817 2018-05-07  Nathan Sidwell  <nathan@acm.org>
15819         * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
15820         * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
15821         (Backwards Compatibility): Likewise.
15823 2018-05-07  Luis Machado  <luis.machado@linaro.org>
15825         PR bootstrap/85681
15826         Revert:
15827         2018-05-07  Luis Machado  <luis.machado@linaro.org>
15829         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
15830         <prefetch_dynamic_strides>: New const bool field.
15831         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
15832         prefetch_dynamic_strides.
15833         (exynosm1_prefetch_tune): Likewise.
15834         (thunderxt88_prefetch_tune): Likewise.
15835         (thunderx_prefetch_tune): Likewise.
15836         (thunderx2t99_prefetch_tune): Likewise.
15837         (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
15838         to false.
15839         (aarch64_override_options_internal): Update to set
15840         PARAM_PREFETCH_DYNAMIC_STRIDES.
15841         * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
15842         * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
15843         * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
15844         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
15845         prefetch-dynamic-strides setting.
15847         2018-05-07  Luis Machado  <luis.machado@linaro.org>
15849         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
15850         <minimum_stride>: New const int field.
15851         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
15852         minimum_stride field.
15853         (exynosm1_prefetch_tune): Likewise.
15854         (thunderxt88_prefetch_tune): Likewise.
15855         (thunderx_prefetch_tune): Likewise.
15856         (thunderx2t99_prefetch_tune): Likewise.
15857         (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
15858         (aarch64_override_options_internal): Update to set
15859         PARAM_PREFETCH_MINIMUM_STRIDE.
15860         * doc/invoke.texi (prefetch-minimum-stride): Document new option.
15861         * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
15862         * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
15863         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
15864         stride is constant and is below the minimum stride threshold.
15866 2018-05-07  Luis Machado  <luis.machado@linaro.org>
15868         * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
15869         to 512.
15871 2018-05-07  Luis Machado  <luis.machado@linaro.org>
15873         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
15874         <prefetch_dynamic_strides>: New const bool field.
15875         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
15876         prefetch_dynamic_strides.
15877         (exynosm1_prefetch_tune): Likewise.
15878         (thunderxt88_prefetch_tune): Likewise.
15879         (thunderx_prefetch_tune): Likewise.
15880         (thunderx2t99_prefetch_tune): Likewise.
15881         (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
15882         to false.
15883         (aarch64_override_options_internal): Update to set
15884         PARAM_PREFETCH_DYNAMIC_STRIDES.
15885         * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
15886         * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
15887         * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
15888         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
15889         prefetch-dynamic-strides setting.
15891 2018-05-07  Luis Machado  <luis.machado@linaro.org>
15893         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
15894         <minimum_stride>: New const int field.
15895         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
15896         minimum_stride field.
15897         (exynosm1_prefetch_tune): Likewise.
15898         (thunderxt88_prefetch_tune): Likewise.
15899         (thunderx_prefetch_tune): Likewise.
15900         (thunderx2t99_prefetch_tune): Likewise.
15901         (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
15902         (aarch64_override_options_internal): Update to set
15903         PARAM_PREFETCH_MINIMUM_STRIDE.
15904         * doc/invoke.texi (prefetch-minimum-stride): Document new option.
15905         * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
15906         * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
15907         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
15908         stride is constant and is below the minimum stride threshold.
15910 2018-05-06  Jakub Jelinek  <jakub@redhat.com>
15912         PR c++/85659
15913         * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
15914         the type is addressable.  Don't force op into register if it has
15915         BLKmode.
15917 2018-05-05  Roland McGrath  <mcgrathr@google.com>
15919         PR other/77609
15920         * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
15921         any section for which we don't know a specific type it should have,
15922         regardless of name.  Previously this was done only for the exact
15923         names ".init_array", ".fini_array", and ".preinit_array".
15924         (default_elf_asm_named_section): Add comment about
15925         relationship with default_section_type_flags and SECTION_NOTYPE.
15926         (get_section): Don't consider it a type conflict if one side has
15927         SECTION_NOTYPE and the other doesn't, as long as neither has the
15928         SECTION_BSS et al used in the default_section_type_flags logic.
15930 2018-05-05  Tom de Vries  <tom@codesourcery.com>
15932         PR target/85653
15933         * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
15934         (workaround_barsyncs): New function.
15935         (nvptx_reorg): Use workaround_barsyncs.
15936         * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
15937         (define_expand "nvptx_membar_cta"): New define_expand.
15938         (define_insn "*nvptx_membar_cta"): New insn.
15940 2018-05-04  Pekka Jääskeläinen  <pekka.jaaskelainen@parmance.com>
15942         * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
15943         To improve optimization opportunities.
15944         * builtin-types.def: The new needed builtin types for the above.
15946 2018-05-04  Richard Biener  <rguenther@suse.de>
15948         * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
15949         * gimple-ssa-store-merging.c
15950         (imm_store_chain_info::output_merged_store): Remove redundant create,
15951         release split_store vector contents on failure.
15952         * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
15953         scalar stmt vector on cache hit.
15955 2018-05-04  Segher Boessenkool  <segher@kernel.crashing.org>
15957         * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
15958         Xilinx FP support.
15959         * config.gcc (powerpc-xilinx-eabi*): Remove.
15960         * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
15961         support.
15962         (fusion_addis_mem_combo_load): Ditto.
15963         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
15964         FP support.
15965         (rs6000_cpu_cpp_builtins): Ditto.
15966         * config/rs6000/rs6000-linux.c
15967         (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
15968         * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
15969         * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
15970         support.
15971         (rs6000_setup_reg_addr_masks): Ditto.
15972         (rs6000_init_hard_regno_mode_ok): Ditto.
15973         (rs6000_option_override_internal): Ditto.
15974         (legitimate_lo_sum_address_p): Ditto.
15975         (rs6000_legitimize_address): Ditto.
15976         (rs6000_legitimize_reload_address): Ditto.
15977         (rs6000_legitimate_address_p): Ditto.
15978         (abi_v4_pass_in_fpr): Ditto.
15979         (setup_incoming_varargs): Ditto.
15980         (rs6000_gimplify_va_arg): Ditto.
15981         (rs6000_split_multireg_move): Ditto.
15982         (rs6000_savres_strategy): Ditto.
15983         (rs6000_emit_prologue_components): Ditto.
15984         (rs6000_emit_epilogue_components): Ditto.
15985         (rs6000_emit_prologue): Ditto.
15986         (rs6000_emit_epilogue): Ditto.
15987         (rs6000_elf_file_end): Ditto.
15988         (rs6000_function_value): Ditto.
15989         (rs6000_libcall_value): Ditto.
15990         * config/rs6000/rs6000.h: Ditto.
15991         (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
15992         (TARGET_MINMAX): ... this.  New.
15993         (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
15994         * config/rs6000/rs6000.md: Remove Xilinx FP support.
15995         (*movsi_internal1_single): Delete.
15996         * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
15997         mfpu=, mxilinx-fpu): Delete.
15998         * config/rs6000/singlefp.h: Delete.
15999         * config/rs6000/sysv4.h: Remove Xilinx FP support.
16000         * config/rs6000/t-rs6000: Ditto.
16001         * config/rs6000/t-xilinx: Delete.
16002         * config/rs6000/titan.md: Adjust for fp_type removal.
16003         * config/rs6000/vsx.md: Remove Xilinx FP support.
16004         (VStype_simple): Delete.
16005         (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
16006         * config/rs6000/xfpu.h: Delete.
16007         * config/rs6000/xfpu.md: Delete.
16008         * config/rs6000/xilinx.h: Delete.
16009         * config/rs6000/xilinx.opt: Delete.
16010         * doc/invoke.texi (RS/6000 and PowerPC Options): Remove
16011         -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
16013 2018-05-04  Tom de Vries  <tom@codesourcery.com>
16015         PR libgomp/85639
16016         * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
16017         if ignore == 0.
16019 2018-05-04  Richard Biener  <rguenther@suse.de>
16021         PR middle-end/85627
16022         * tree-complex.c (update_complex_assignment): We are always in SSA form.
16023         (expand_complex_div_wide): Likewise.
16024         (expand_complex_operations_1): Likewise.
16025         (expand_complex_libcall): Preserve EH info of the original stmt.
16026         (tree_lower_complex): Handle removed blocks.
16027         * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
16028         on complex multiplication and division libcall builtins.
16030 2018-05-04  Richard Biener  <rguenther@suse.de>
16032         PR middle-end/85574
16033         * fold-const.c (negate_expr_p): Restrict negation of operand
16034         zero of a division to when we know that can happen without
16035         overflow.
16036         (fold_negate_expr_1): Likewise.
16038 2018-05-04  Jakub Jelinek  <jakub@redhat.com>
16040         PR libstdc++/85466
16041         * real.h (real_nextafter): Declare.
16042         * real.c (real_nextafter): New function.
16043         * fold-const-call.c (fold_const_nextafter): New function.
16044         (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
16045         CASE_CFN_NEXTTOWARD.
16046         (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
16047         even when arg1_mode is different from arg0_mode.
16049 2018-05-03  Nathan Sidwell  <nathan@acm.org>
16051         * doc/extend.texi (Deprecated Features): Remove
16052         -ffriend-injection.
16053         (Backwards Compatibility): Likewise.
16054         * doc/invoke.texi (C++ Language Options): Likewise.
16055         (C++ Dialect Options): Likewise.
16057 2018-05-03  Jakub Jelinek  <jakub@redhat.com>
16059         PR target/85530
16060         * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
16061         _mm512_mask_mullox_epi64): New intrinsics.
16063 2018-05-03  Tom de Vries  <tom@codesourcery.com>
16065         PR testsuite/85106
16066         * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
16067         dump files): Add offload-tree.
16069 2018-05-03  Richard Biener  <rguenther@suse.de>
16071         PR tree-optimization/85615
16072         * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
16073         to loops not nested in BBs loop father to avoid creating multi-entry
16074         loops.
16076 2018-05-03  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
16078         PR tree-optimization/70291
16079         * tree-complex.c (expand_complex_libcall): Add type, inplace_p
16080         arguments.  Change return type to tree.  Emit libcall as a new
16081         statement rather than replacing existing one when inplace_p is true.
16082         (expand_complex_multiplication_components): New function.
16083         (expand_complex_multiplication): Expand floating-point complex
16084         multiplication using the above.
16085         (expand_complex_division): Rename inner_type parameter to type.
16086         Update expand_complex_libcall call-site.
16087         (expand_complex_operations_1): Update expand_complex_multiplication
16088         and expand_complex_division call-sites.
16090 2018-05-02  Jakub Jelinek  <jakub@redhat.com>
16092         PR target/85582
16093         * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
16094         *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
16095         *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
16096         the highest significant bit of the shift count mask is clear.  In
16097         check whether and[sq]i3 is needed verify that all significant bits
16098         of the shift count other than the highest are set.
16100 2018-05-02  Tom de Vries  <tom@codesourcery.com>
16102         PR libgomp/82428
16103         * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
16104         * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
16105         (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
16106         * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
16107         (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
16108         * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
16109         __builtin_goacc_parlevel_size.
16111 2018-05-02  Richard Biener  <rguenther@suse.de>
16113         PR tree-optimization/85597
16114         * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
16115         do not use split vect_get_vec_defs call but call vect_get_slp_defs
16116         directly.
16118 2018-05-02  Tom de Vries  <tom@codesourcery.com>
16120         PR testsuite/85106
16121         * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
16122         dump files): Add ltrans-tree.
16124 2018-05-02  Tom de Vries  <tom@codesourcery.com>
16126         PR testsuite/85106
16127         * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
16128         dump files): Add wpa-ipa.
16130 2018-05-02  Segher Boessenkool  <segher@kernel.crashing.org>
16132         * config.gcc (powerpc*-*-*): Remove paired.h.  Unsupport the
16133         powerpc*-*-linux*paired* target.
16134         * config/rs6000/750cl.h: Delete.
16135         * config/rs6000/paired.h: Delete.
16136         * config/rs6000/paired.md: Delete.
16137         * config/rs6000/predicates.md (easy_vector_constant): Remove paired
16138         float support.
16139         * config/rs6000/rs6000-builtin.def: Remove paired float support.
16140         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
16141         comment.  Remove paired float support.
16142         * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
16143         * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
16144         VECTOR_PAIRED.
16145         * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
16146         paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
16147         declarations.
16148         * config/rs6000/rs6000.c: Remove paired float support.
16149         (paired_expand_vector_init, paired_expand_vector_move,
16150         paired_emit_vector_compare, paired_emit_vector_cond_expr,
16151         (paired_expand_lv_builtin, paired_expand_stv_builtin,
16152         paired_expand_builtin, paired_expand_predicate_builtin,
16153         paired_init_builtins): Delete.
16154         * config/rs6000/rs6000.h: Remove paired float support.
16155         * config/rs6000/rs6000.md: Remove paired float support.
16156         (move_from_CR_ov_bit): Delete.
16157         * config/rs6000/rs6000.opt (mpaired): Delete.
16158         * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
16159         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
16161 2018-05-02  Richard Biener  <rguenther@suse.de>
16163         PR middle-end/85567
16164         * gimplify.c (gimplify_save_expr): When in SSA form allow
16165         SAVE_EXPRs to compute to SSA vars.
16167 2018-05-02  Jakub Jelinek  <jakub@redhat.com>
16169         PR target/85582
16170         * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
16171         *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
16172         *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
16173         clobber operands[2], instead use a new pseudo.  Formatting fixes.
16175 2018-05-02  Richard Sandiford  <richard.sandiford@linaro.org>
16177         PR tree-optimization/85586
16178         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
16179         exit early for statements in the same group if the accesses are
16180         not strided.
16182 2018-05-02  Tom de Vries  <tom@codesourcery.com>
16184         PR lto/85451
16185         * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
16186         error message.
16188 2018-05-01  Marc Glisse  <marc.glisse@inria.fr>
16190         PR tree-optimization/85143
16191         * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
16193 2018-05-01  Tom de Vries  <tom@codesourcery.com>
16195         PR lto/85451
16196         * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
16197         not found" error message.
16199 2018-05-01  Tom de Vries  <tom@codesourcery.com>
16201         PR other/83786
16202         * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
16203         * vec.c (test_ordered_remove_if): New function.
16204         (vec_c_tests): Call test_ordered_remove_if.
16205         * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
16206         * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
16207         * tree-vect-patterns.c (vect_pattern_recog_1): Use
16208         VEC_ORDERED_REMOVE_IF.
16210 2018-05-01  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
16212         PR tree-optimization/82665
16213         * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
16214         pointer subtraction where arguments come from a memchr call.
16216 2018-05-01  Jakub Jelinek  <jakub@redhat.com>
16218         * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
16219         --push-state --as-needed and --pop-state instead of --as-needed and
16220         --no-as-needed if ld supports it.
16221         * configure: Regenerated.
16223         PR web/85578
16224         * doc/install.texi2html: Replace _002d with - and _002a with * in
16225         generated html files using sed.
16227 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
16229         PR c++/85523
16230         * gcc-rich-location.c (blank_line_before_p): New function.
16231         (use_new_line): New function.
16232         (gcc_rich_location::add_fixit_insert_formatted): New function.
16233         * gcc-rich-location.h
16234         (gcc_rich_location::add_fixit_insert_formatted): New function.
16236 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
16238         * selftest.c (assert_streq): Rename "expected" and "actual" to
16239         "val1" and "val2".  Extend NULL-handling to cover both inputs
16240         symmetrically, while still requiring both to be non-NULL for a pass.
16241         * selftest.h (assert_streq): Rename "expected" and "actual" to
16242         "val1" and "val2".
16243         (ASSERT_EQ): Likewise.
16244         (ASSERT_EQ_AT): Likewise.
16245         (ASSERT_KNOWN_EQ): Likewise.
16246         (ASSERT_KNOWN_EQ_AT): Likewise.
16247         (ASSERT_NE): Likewise.
16248         (ASSERT_MAYBE_NE): Likewise.
16249         (ASSERT_MAYBE_NE_AT): Likewise.
16250         (ASSERT_STREQ): Likewise.  Clarify that both must be non-NULL for
16251         the assertion to pass.
16252         (ASSERT_STREQ_AT): Likewise.
16254 2018-04-30  Jonathan Wakely  <jwakely@redhat.com>
16256         * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
16257         interaction with -pie.
16259 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
16261         * selftest.h: Fix alphabetization of per-source-file selftest
16262         declarations.
16264 2018-04-30  Jason Merrill  <jason@redhat.com>
16266         PR c++/61982 - dead stores to destroyed objects.
16267         * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
16268         of clobber.
16270 2018-04-30  Jason Merrill  <jason@redhat.com>
16272         * tree.c (build_clobber): New.
16273         * tree.h: Declare it.
16274         * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
16276 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
16278         * diagnostic-show-locus.c (layout::layout): Update for
16279         location_get_source_line returning a char_span.
16280         (struct char_span): Move to input.h.
16281         (struct correction): Update for fields in char_span becoming
16282         private.
16283         (struct source_line): Update for location_get_source_line
16284         returning a char_span.
16285         (layout::print_line): Likewise.
16286         * edit-context.c (edited_file::print_content): Likewise.
16287         (edited_file::print_diff_hunk): Likewise.
16288         (edited_file::print_run_of_changed_lines): Likewise.
16289         (edited_file::get_num_lines): Likewise.
16290         (edited_line::edited_line): Likewise.
16291         * final.c (asm_show_source): Likewise.
16292         * input.c (location_get_source_line): Convert return type
16293         from const char * to char_span, losing the final "line_len"
16294         param.
16295         (dump_location_info): Update for the above.
16296         (get_substring_ranges_for_loc): Likewise.  Use a char_span
16297         when handling the literal within the line.
16298         (test_reading_source_line): Update for location_get_source_line
16299         returning a char_span.
16300         * input.h (class char_span): Move here from
16301         diagnostic-show-locus.c, converting from a struct to a class.
16302         Make data members private.
16303         (char_span::operator bool): New.
16304         (char_span::length): New.
16305         (char_span::get_buffer): New.
16306         (char_span::operator[]): New.
16307         (char_span::subspan): Make const.
16308         (char_span::xstrdup): New.
16309         (location_get_source_line): Convert return type from const char *
16310         to char_span, losing the final "line_size" param.
16312 2018-04-30  Jan Hubicka  <jh@suse.cz>
16314         * lto-wrapper.c (ltrans_priorities): New static var.
16315         (cmp_priority): New.
16316         (run_gcc): Read priorities and if doing parallel build order
16317         the Makefile by them.
16319 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
16321         * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
16323 2018-04-30  Richard Biener  <rguenther@suse.de>
16325         * tree-cfg.c (verify_address): Remove base argument, add
16326         flag whether to check TREE_ADDRESSABLE and do that.
16327         (verify_expr): Remove.
16328         (verify_types_in_gimple_reference): Add pieces from verify_expr.
16329         (verify_gimple_assign_single): Likewise.
16330         (verify_gimple_switch): Likewise.
16331         (verify_expr_location_1): Dereference tp once.  Add (disabled)
16332         piece from verify_expr.
16333         (verify_gimple_in_cfg): Do not call verify_expr on all ops.
16335 2018-04-30  Claudiu Zissulescu  <claziss@synopsys.com>
16337         * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
16339 2018-04-30  Claudiu Zissulescu  <claziss@synopsys.com>
16341         * config/arc/arc-protos.h (prepare_extend_operands): Remove.
16342         (small_data_pattern): Likewise.
16343         (arc_rewrite_small_data): Likewise.
16344         * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
16345         (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
16346         (get_symbol_alignment): New function.
16347         (legitimate_small_data_address_p): Likewise.
16348         (legitimate_scaled_address): Update, call
16349         legitimate_small_data_address_p.
16350         (output_sdata): New static variable.
16351         (arc_print_operand): Update how we handle small data operands.
16352         (arc_print_operand_address): Likewise.
16353         (arc_legitimate_address_p): Update, use
16354         legitimate_small_data_address_p.
16355         (arc_rewrite_small_data_p): Remove.
16356         (arc_rewrite_small_data_1): Likewise.
16357         (arc_rewrite_small_data): Likewise.
16358         (small_data_pattern): Likewise.
16359         (compact_sda_memory_operand): Update to use
16360         legitimate_small_data_address_p and get_symbol_alignment.
16361         (prepare_move_operands): Don't rewite sdata pattern.
16362         (prepare_extend_operands): Remove.
16363         * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
16364         pattern.
16365         (zero_extendqisi2): Likewise.
16366         (zero_extendhisi2): Likewise.
16367         (extendqihi2): Likewise.
16368         (extendqisi2): Likewise.
16369         (extendhisi2): Likewise.
16370         (addsi3): Likewise.
16371         (subsi3): Likewise.
16372         (andsi3): Likewise.
16373         * config/arc/constraints.md (Usd): Change it to memory constraint.
16375 2018-04-30  Claudiu Zissulescu  <claziss@synopsys.com>
16377         * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
16378         as source of std instructions.
16379         * config/arc/arc.md (movsi_insn): Update pattern predicate to
16380         allow 6-bit constants as source for store instructions.
16381         (movdi_insn): Update instruction pattern to allow 6-bit constants
16382         as source for store instructions.
16384 2018-04-30  Jonathan Wakely  <jwakely@redhat.com>
16386         * doc/invoke.texi (-fdebug-types-section): Fix grammar.
16388 2018-04-30  Nathan Sidwell  <nathan@acm.org>
16389             Sandra Loosemore <sandra@codesourcery.com>
16391         * dumpfile.c (dump_open): Allow '-' for stdout.
16392         * doc/invoke.texi (Developer Options): Document dump filename
16393         determination early.  Document stdin/stdout selection.
16395 2018-04-30  Andrew Sadek  <andrew.sadek.se@gmail.com>
16397         Microblaze Target: PIC data text relative
16399         * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
16400         * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
16401         Add declaration.
16402         * config/microblaze/microblaze.h (microblaze_constant_address_p):
16403         CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
16404         * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
16405         New addressing mode for data-text relative position indepenedent code.
16406         (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
16407         'ADDRESS_SYMBOLIC_TXT_REL'.
16408         (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
16409         (microblaze_legitimate_pic_operand): Exclude function calls from
16410         pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
16411         (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
16412         addresses cases.
16413         (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
16414         (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
16415         (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
16416         for 'address + offset'.
16417         (microblaze_expand_prologue): Add new function prologue call for
16418         'r20' assignation.
16419         (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
16420         'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
16421         table in case of TARGET_PIC_DATA_TEXT_REL.
16422         (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
16423         * config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
16424         Add new macros 'UNSPEC_TEXT',
16425         'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
16426         + exclude function calls from 'UNSPEC_PLT' in case of data text
16427         relative mode.
16428         * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
16429         new target hook for generating address diff vector tables in case of
16430         flag_pic.
16431         * doc/tm.texi : Regenerate.
16432         * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
16433         'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
16434         of addr diff vector generation.
16435         * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
16436         target hook definition.
16437         * targhooks.h, targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
16438         Add default function for generate_pic_addr_diff_vec -> flag_pic.
16439         * doc/invoke.texi (Add new pic option): Add new microblaze pic
16440         option for data text relative.
16442 2018-04-30  Richard Biener  <rguenther@suse.de>
16444         * tree-chrec.h (evolution_function_is_constant_p): Remove
16445         redundant check.
16446         * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
16448 2018-04-30  Richard Biener  <rguenther@suse.de>
16450         PR bootstrap/85571
16451         * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
16453 2018-04-30  Richard Biener  <rguenther@suse.de>
16455         PR tree-optimization/28364
16456         PR tree-optimization/85275
16457         * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
16458         copying first exit test.
16460 2018-04-28  Mark Wielaard  <mark@klomp.org>
16462         * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
16463         dwarf_version >= 5.
16464         (dwarf_AT): Handle DW_AT_addr_base.
16465         (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
16467 2018-04-28  Uros Bizjak  <ubizjak@gmail.com>
16469         PR target/84431
16470         * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
16471         (*ashl<dwi>3_doubleword_mask_1): Ditto.
16472         (*<shift_insn><dwi>3_doubleword_mask): Ditto.
16473         (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
16475 2018-04-28  Richard Biener  <rguenther@suse.de>
16477         * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
16478         (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
16479         to reflect use.  Only add interesting stmts.
16481 2018-04-27  Martin Jambor  <mjambor@suse.cz>
16483         PR ipa/85549
16484         * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
16485         the jump function allows for passing through aggregate values.
16487 2018-04-27  David Malcolm  <dmalcolm@redhat.com>
16489         * input.h (in_system_header_at): Convert from macro to inline
16490         function.
16491         (from_macro_expansion_at): Likewise.
16492         (from_macro_definition_at): Likewise.
16494 2018-04-27  Jeff Law  <law@redhat.com>
16496         * config.gcc: Mark tile* targets as deprecated/obsolete.
16498 2018-04-27  Richard Biener  <rguenther@suse.de>
16500         * config/aarch64/aarch64.c: Simplify ap.__stack advance and
16501         fix for ILP32.
16503 2018-04-27  Richard Biener  <rguenther@suse.de>
16505         * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
16507 2018-04-27  Uros Bizjak  <ubizjak@gmail.com>
16509         * config/i386/i386.md (*movti_internal): Substitute Ye constraint
16510         with Yd constraint. Set "preferred_for_speed" attribute from
16511         TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
16512         with Yd constraint.
16513         (*movdi_internal): Ditto.
16514         (movti_interunit splitters): Remove
16515         TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
16516         (movdi_interunit splitters): Ditto.
16517         * config/i386/constraints.md (Ye): Remove.
16518         (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
16520 2018-04-27  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
16522         PR target/85512
16523         * config/aarch64/constraints.md (Usg): Limit to 31.
16524         (Usj): Limit to 63.
16526 2018-04-27  Jakub Jelinek  <jakub@redhat.com>
16528         PR tree-optimization/85529
16529         * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
16530         argument.  Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
16531         rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
16532         zero extension or masking of the MSB bit.
16533         (optimize_range_tests): Add FIRST_BB argument, pass it through
16534         to optimize_range_tests_var_bound.
16535         (maybe_optimize_range_tests, reassociate_bb): Adjust
16536         optimize_range_tests callers.
16538 2018-04-26  Richard Biener  <rguenther@suse.de>
16539             Jakub Jelinek  <jakub@redhat.com>
16541         * cgraph.h (symbol_table): Just declare debug method here.
16542         * symtab.c (symbol_table::debug): Define.
16544 2018-04-26  Eric Botcazou  <ebotcazou@adacore.com>
16546         * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
16548 2018-04-26  Uros Bizjak  <ubizjak@gmail.com>
16550         * config/i386/i386.md ("isa" attribute): Add x64_sse2.
16551         ("enabled" attribute): Handle x64_sse2 "isa" attribute.
16552         (*movdi_internal): Substitute Yi and Yj constraint with x
16553         and Ym and Yn constraint with y constraint.  Update "isa"
16554         attribute and set "preferred_for_speed" attribute from
16555         TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
16556         (*movsi_internal): Ditto.
16557         (*movdf_internal): Ditto.
16558         (*movsf_internal): Ditto.
16559         (*zero_extendsidi2): Ditto.
16560         * config/i386/sse.md (vec_set<mode>_0): Ditto.
16561         (sse2_loadld): Ditto.
16562         (*vec_extract<ssevecmodelower>_0): Ditto.
16563         (*vec_extractv4si_0_zext_sse4): Ditto.
16564         (vec_concatv2di): Ditto.
16565         (*vec_dup<mode>): Ditto.
16566         * config/i386/mmx.md (*mov<mode>_internal): Ditto.
16567         * config/i386/constraints.md (Yi): Remove.
16568         (Yj): Remove.
16569         (Ym): Remove.
16570         (Yn): Remove.
16572 2018-04-26  Nathan Sidwell  <nathan@acm.org>
16574         * dumpfile.c (dump_open): New.
16575         (dump_open_alternate_stream, dump_start, dump_begin): Call it.
16576         (dump_finish): Detect stdio/stderr by value not name.
16578 2018-04-26  Jonathan Wakely  <jwakely@redhat.com>
16580         * doc/invoke.texi (-Wreturn-type): Document default status for C++.
16582 2018-04-26  Tom de Vries  <tom@codesourcery.com>
16584         PR target/84952
16585         * config/nvptx/nvptx.c (verify_neutering_jumps)
16586         (verify_neutering_labels): New function
16587         (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
16589 2018-04-26  Tom de Vries  <tom@codesourcery.com>
16591         PR target/84025
16592         * config/nvptx/nvptx.c (needs_neutering_p): New function.
16593         (nvptx_single): Use needs_neutering_p to skip over insns that do not
16594         need neutering.
16596 2018-04-26  Richard Biener <rguenther@suse.de>
16597             Tom de Vries  <tom@codesourcery.com>
16599         PR lto/85422
16600         * lto-streamer-out.c (output_function): Fixup loops if required to match
16601         discovery done in the reader.
16603 2018-04-26  Richard Biener  <rguenther@suse.de>
16605         PR tree-optimization/85116
16606         * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
16607         have a loop exit from the single latch predecessor.  Remove
16608         case of header with just condition.
16609         (ch_base::copy_headers): Exclude infinite loops from any
16610         processing.
16611         (pass_ch::execute): Record exits.
16613 2018-04-26  Richard Biener  <rguenther@suse.de>
16615         * tree-vect-data-refs.c (vect_get_data_access_cost): Get
16616         prologue cost vector and pass it to vect_get_load_cost.
16617         (vect_get_peeling_costs_all_drs): Likewise.
16618         (vect_peeling_hash_get_lowest_cost): Likewise.
16619         (vect_enhance_data_refs_alignment): Likewise.
16621 2018-04-26  Richard Biener  <rguenther@suse.de>
16623         PR middle-end/85450
16624         * tree-cfg.c (verify_gimple_assign_unary): Restore proper
16625         checking of integer<->pointer conversions.
16626         * omp-expand.c (expand_omp_for_static_nochunk): Avoid
16627         sign-/zero-extending pointer types.
16628         (expand_omp_for_static_chunk): Likewise.
16630 2018-03-22  Hans-Peter Nilsson  <hp@axis.com>
16631             Jean Lee  <xiaoyur347@gmail.com>
16633         * config/mips/mips.c (mips_asan_shadow_offset): New function.
16634         (TARGET_ASAN_SHADOW_OFFSET): Define.
16635         * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
16636         true for -fsanitize=address.
16638 2018-04-25  Mark Wielaard  <mark@klomp.org>
16640         * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
16641         shorter ones.
16643 2018-04-25  Jakub Jelinek  <jakub@redhat.com>
16645         * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
16646         than "alu", remove explicit "memory" and "imm_disp" attributes.
16647         (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
16649         PR middle-end/85414
16650         * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
16651         case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
16652         gen_lowpart_no_emit.
16654 2018-04-25  Sebastian Peryt  <sebastian.peryt@intel.com>
16656         PR target/85473
16657         * config/i386/i386.c (ix86_expand_builtin): Change memory
16658         operand to XI, extend p0 to Pmode.
16659         * config/i386/i386.md: Change unspec volatile and operand
16660         1 mode to XI, change operand 0 mode to P.
16662 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
16664         * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
16665         GET_MODE_MASK before any checking.
16666         (nds32_can_use_bset_p): Likewise.
16667         (nds32_can_use_btgl_p): Likewise.
16669 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
16671         * config/nds32/nds32-doubleword.md: New define_split pattern for
16672         illegal register number.
16674 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
16676         * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
16678 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
16680         * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
16682 2018-04-25  Richard Biener  <rguenther@suse.de>
16684         * lto-streamer.h (LTO_major_version): Bump to 8.
16686 2018-04-25  Jakub Jelinek  <jakub@redhat.com>
16688         * BASE-VER: Set to 9.0.0.
16690 2018-04-24  Segher Boessenkool  <segher@kernel.crashing.org>
16692         * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
16693         in __abskf2 and __powikf2.
16695 2018-04-24  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
16697         PR target/85512
16698         * config/aarch64/constraints.md (Usg, Usj): New constraints.
16699         * config/aarch64/iterators.md (cmode_simd): New mode attribute.
16700         * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
16701         Use the above on operand 2.  Reindent.
16702         (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
16704 2018-04-24  H.J. Lu  <hongjiu.lu@intel.com>
16706         PR target/85485
16707         * common/config/i386/i386-common.c (ix86_handle_option): Don't
16708         handle OPT_mcet.
16709         * config/i386/i386.opt (mcet): Removed.
16710         * doc/install.texi: Remove -mcet documentation.
16711         * doc/invoke.texi: Likewise.
16713 2018-04-24  H.J. Lu  <hongjiu.lu@intel.com>
16715         PR target/85485
16716         * doc/install.texi: Remove -mcet from bootstrap-cet.
16718 2018-04-24  Jakub Jelinek  <jakub@redhat.com>
16720         PR target/85511
16721         * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
16722         __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
16723         if TARGET_64BIT.
16725         PR target/85503
16726         * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
16727         const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
16728         containing a CONST_VECTOR.
16730 2018-04-24  Cesar Philippidis  <cesar@codesourcery.com>
16732         * doc/install.texi: Update newlib dependency for nvptx.
16734 2018-04-24  Jakub Jelinek  <jakub@redhat.com>
16736         PR target/85508
16737         * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
16738         instead of INTVAL when shifting x left.
16740 2018-04-24  Andreas Krebbel  <krebbel@linux.ibm.com>
16742         PR tree-optimization/85478
16743         * tree-vect-loop.c (vect_analyze_loop_2): Do not call
16744         vect_grouped_store_supported for single element vectors.
16746 2018-04-24  Richard Biener  <rguenther@suse.de>
16748         PR target/85491
16749         * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
16750         load cost increase to the case of non-constant step.
16752 2018-04-24  Jakub Jelinek  <jakub@redhat.com>
16754         PR target/84828
16755         * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
16756         destination if any_malformed_asm.
16758 2018-04-23  Eric Botcazou  <ebotcazou@adacore.com>
16760         PR middle-end/85496
16761         * expr.c (store_field): In the bitfield case, if the value comes from
16762         a function call and is returned in registers by means of a PARALLEL,
16763         do not change the mode of the temporary unless BLKmode and VOIDmode.
16765 2018-04-23  Andrey Belevantsev  <abel@ispras.ru>
16767         PR rtl-optimization/85423
16768         * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
16769         dependencies to debug insns when the previous insn is non-debug.
16771 2018-04-23  Claudiu Zissulescu  <claziss@synopsys.com>
16773         * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
16774         enums into a single definition.
16775         (fls): Fix predicates and printing.
16776         (seti): Likewise.
16778 2018-04-23  Claudiu Zissulescu  <claziss@synopsys.com>
16780         * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
16781         * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
16782         and short u6 immediate.
16783         (check_if_valid_sleep_operand): Remove.
16784         * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
16786 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
16788         * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
16789         flag_always_save_lp condition.
16790         * config/nds32/nds32.opt (malways-save-lp): New option.
16792 2018-04-22  Shiva Chen  <shiva0217@gmail.com>
16794         * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
16795         * config/nds32/nds32.c (nds32_use_load_post_increment): New.
16796         * config/nds32/nds32.h
16797         (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
16798         (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
16800 2018-04-22  Shiva Chen  <shiva0217@gmail.com>
16802         * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
16803         * config/nds32/nds32.c (nds32_ls_333_p): Remove.
16805 2018-04-22  Shiva Chen  <shiva0217@gmail.com>
16806             Chung-Ju Wu  <jasonwucj@gmail.com>
16808         * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
16809         Declare.
16810         * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
16811         * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
16813 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
16815         * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
16817 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
16819         * config/nds32/nds32-protos.h (nds32_data_alignment,
16820         nds32_local_alignment): Declare.
16821         * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
16822         nds32_local_alignment): New functions.
16823         (TARGET_CONSTANT_ALIGNMENT): Define.
16824         * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
16826 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
16828         * config/nds32/nds32.c
16829         (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
16830         (TARGET_MODES_TIEABLE_P): Likewise.
16832 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
16834         * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
16835         level Ofast and Og.
16837 2018-04-22  Monk Chiang  <sh.chiang04@gmail.com>
16838             Chung-Ju Wu  <jasonwucj@gmail.com>
16840         * config/nds32/constants.md (unspec_volatile_element): Add enum values
16841         for unaligned access.
16842         * config/nds32/nds32-intrinsic.c: Implementation of expanding
16843         unaligned access.
16844         * config/nds32/nds32-intrinsic.md: Likewise.
16845         * config/nds32/nds32_intrinsic.h: Likewise.
16846         * config/nds32/nds32.h (nds32_builtins): Likewise.
16847         * config/nds32/nds32.opt (munaligned-access): New option.
16848         * config/nds32/nds32.c (nds32_asm_file_start): Display
16849         flag_unaligned_access status.
16851 2018-04-20  Kito Cheng  <kito.cheng@gmail.com>
16853         * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
16854         -mno-relax is present.
16855         * config/riscv/linux.h (LINK_SPEC): Ditto.
16857 2018-04-20  Martin Sebor  <msebor@redhat.com>
16859         PR c/85365
16860         * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
16861         for null pointers.
16862         (gimple_fold_builtin_stxcpy_chk): Same.
16863         * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
16865 2018-04-20  Michael Meissner  <meissner@linux.ibm.com>
16867         PR target/85456
16868         * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
16869         __powikf2 when long double is IEEE 128-bit.
16871 2018-04-20  Kito Cheng  <kito.cheng@gmail.com>
16873         * config/riscv/riscv.c (riscv_first_stack_step): Round up min
16874         step to make sure stack always aligned.
16876 2018-04-20  Carl Love  <cel@us.ibm.com>
16878         PR target/83402
16879         * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
16880         size check for arg0.
16882 2018-04-20  Nathan Sidwell  <nathan@codesourcery.com>
16883             Tom de Vries  <tom@codesourcery.com>
16885         PR target/85445
16886         * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
16887         Emit insns for calls too.
16888         (nvptx_find_par): Always look for worker-level predecessor insn.
16889         (nvptx_propagate): Add is_call parm, return bool.  Copy frame for
16890         calls.
16891         (nvptx_vpropagate, nvptx_wpropagate): Adjust.
16892         (nvptx_process_pars): Propagate frames for calls.
16894 2018-04-20  H.J. Lu  <hongjiu.lu@intel.com>
16896         PR target/85469
16897         * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
16898         Removed.
16899         (OPTION_MASK_ISA_IBT_UNSET): Likewise.
16900         (ix86_handle_option): Don't handle OPT_mibt.
16901         * config/i386/cet.h: Check __CET__ instead of __IBT__ and
16902         __SHSTK__.
16903         * config/i386/driver-i386.c (host_detect_local_cpu): Remove
16904         has_ibt and ibt.
16905         * config/i386/i386-c.c (ix86_target_macros_internal): Don't
16906         check OPTION_MASK_ISA_IBT nor flag_cf_protection.
16907         (ix86_target_macros): Define __CET__ with flag_cf_protection
16908         for -fcf-protection.
16909         * config/i386/i386.c (isa2_opts): Remove -mibt.
16910         * config/i386/i386.h (TARGET_IBT): Removed.
16911         (TARGET_IBT_P): Likewise.
16912         (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
16913         * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
16914         * config/i386/i386.opt (mcet): Update help message.
16915         (mshstk): Likewise.
16916         (mibt): Removed.
16917         * doc/invoke.texi: Remove -mibt.  Document __CET__.  Document
16918         -mcet as an alias for -mshstk.
16920 2018-04-20  Richard Biener <rguenther@suse.de>
16922         PR middle-end/85475
16923         * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
16924         complexity by forcing a single use of the multiply operand.
16926 2018-04-20  Martin Jambor  <mjambor@suse.cz>
16928         ipa/85449
16929         * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
16930         recursion dependency to only apply to non-clones.
16932 2018-04-20  Martin Jambor  <mjambor@suse.cz>
16934         ipa/85447
16935         * ipa-cp.c (create_specialized_node): Check that clones of
16936         self-recursive edges exist during IPA-CP.
16938 2018-04-19  Toon Moene  <toon@moene.org>
16940         * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
16941         by -O3.
16943 2018-04-19  Jakub Jelinek  <jakub@redhat.com>
16945         PR tree-optimization/85467
16946         * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
16947         VECTOR_TYPE_P macro.  If type is vector type, VIEW_CONVERT_EXPR the
16948         VECTOR_CST element to type.
16950 2018-04-19  H.J. Lu  <hongjiu.lu@intel.com>
16952         PR target/85397
16953         * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
16954         * config/i386/i386.md (builtin_setjmp_setup): Removed.
16955         (builtin_longjmp): Likewise.
16956         (save_stack_nonlocal): New pattern.
16957         (restore_stack_nonlocal): Likewise.
16959 2018-04-19  H.J. Lu  <hongjiu.lu@intel.com>
16961         PR target/85404
16962         * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
16963         Replace ASM_OUTPUT_LABEL with fprintf.
16965 2018-04-19  H.J. Lu  <hongjiu.lu@intel.com>
16967         PR target/85417
16968         * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
16969         Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
16970         * config/i386/i386-c.c (ix86_target_macros_internal): Also
16971         define __IBT__ and __SHSTK__ for -fcf-protection.
16972         * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
16973         TARGET_IBT.
16974         (ix86_trampoline_init): Likewise.
16975         (x86_output_mi_thunk): Likewise.
16976         (ix86_notrack_prefixed_insn_p): Likewise.
16977         (ix86_option_override_internal): Don't disallow -fcf-protection.
16978         * config/i386/i386.md (rdssp<mode>): Also enable for
16979         -fcf-protection.
16980         (incssp<mode>): Likewise.
16981         (nop_endbr): Likewise.
16982         * config/i386/i386.opt (mcet): Change help message to built-in
16983         functions only.
16984         (mibt): Likewise.
16985         (mshstk): Likewise.
16986         * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
16987         on -fcf-protection.  Change -mcet, -mibt and -mshstk to only
16988         enable CET built-in functions.
16990 2018-04-19  Sebastian Peryt  <sebastian.peryt@intel.com>
16992         * common/config/i386/i386-common.c
16993         (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
16994         OPTION_MASK_ISA_MOVDIRI_UNSET,
16995         OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
16996         (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
16997         * config.gcc (movdirintrin.h): New header.
16998         * config/i386/cpuid.h (bit_MOVDIRI,
16999         bit_MOVDIR64B): New bits.
17000         * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
17001         and -mmvodir64b.
17002         * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
17003         (VOID, PVOID, PCVOID)): New function types.
17004         * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
17005         __builtin_ia32_directstoreu_u64,
17006         __builtin_ia32_movdir64b): New builtins.
17007         * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
17008         * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
17009         and -mmovdiri.
17010         (ix86_valid_target_attribute_inner_p): Ditto.
17011         (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
17012         and VOID_FTYPE_PUNSIGNED_UNSIGNED.
17013         (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
17014         * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
17015         TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
17016         * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
17017         (movdiri<mode>, movdir64b_<mode>): New.
17018         * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
17019         * config/i386/immintrin.h: Include movdirintrin.h.
17020         * config/i386/movdirintrin.h: New file.
17021         * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
17023 2018-04-19  Richard Biener  <rguenther@suse.de>
17025         PR middle-end/85455
17026         * cfg.c (clear_bb_flags): When loop state says we have
17027         marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
17029 2018-04-19  Richard Biener  <rguenther@suse.de>
17031         PR tree-optimization/84737
17032         * tree-vect-data-refs.c (vect_copy_ref_info): New function
17033         copying restrict info.
17034         (vect_setup_realignment): Use it.
17035         * tree-vectorizer.h (vect_copy_ref_info): Declare.
17036         * tree-vect-stmts.c (vectorizable_store): Copy ref info from
17037         the first DR to all generated stores.
17038         (vectorizable_load): Likewise for loads.
17040 2018-04-19  Jakub Jelinek  <jakub@redhat.com>
17042         PR tree-optimization/85446
17043         * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
17044         the integral and pointer types to have the same precision.
17046         * doc/install.texi: Document --disable-cet being the default and
17047         --enable-cet=auto.
17049 2018-04-18  Martin Liska  <mliska@suse.cz>
17051         * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
17052         style.
17054 2018-04-18  Martin Liska  <mliska@suse.cz>
17056         Revert
17057         2018-03-02  Eric Botcazou  <ebotcazou@adacore.com>
17059         PR ipa/83983
17060         * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
17061         arguments if they are comparable.
17063 2018-04-18  Martin Liska  <mliska@suse.cz>
17065         Revert
17066         2018-03-13  Eric Botcazou  <ebotcazou@adacore.com>
17068         PR lto/84805
17069         * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
17070         incomplete types.
17072 2018-04-18  H.J. Lu  <hongjiu.lu@intel.com>
17074         PR target/85388
17075         * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
17076         ENDBR after calling __morestack.
17078 2018-04-18  David Malcolm  <dmalcolm@redhat.com>
17080         PR jit/85384
17081         * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
17082         by using gcc_base_ver to generate a gcc_driver_version, and use
17083         it when generating GCC_DRIVER_NAME.
17084         * configure: Regenerate.
17086 2018-04-18  Jakub Jelinek  <jakub@redhat.com>
17088         PR target/81084
17089         * config.gcc: Obsolete powerpc*-*-*spe*.
17091 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
17093         PR debug/84637
17094         * dbxout.c (dbxout_int): Perform negation in unsigned int type.
17095         (stabstr_D): Change type of unum from unsigned int to
17096         unsigned HOST_WIDE_INT.  Perform negation in unsigned HOST_WIDE_INT
17097         type.
17099 2018-04-17  Jim Wilson  <jimw@sifive.com>
17101         PR 84856
17102         * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
17103         RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
17104         Set arg_pointer_offset after using pretend_args_size.
17106 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
17108         PR rtl-optimization/85431
17109         * dse.c (record_store): Ignore zero width stores.
17111         PR sanitizer/85230
17112         * asan.c (handle_builtin_stack_restore): Adjust comment.  Emit
17113         __asan_allocas_unpoison call and last_alloca_addr = new_sp before
17114         __builtin_stack_restore rather than after it.
17115         * builtins.c (expand_asan_emit_allocas_unpoison): Pass
17116         arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
17117         argument instead of virtual_dynamic_stack_rtx.
17119 2018-04-17  Kelvin Nilsen  <kelvin@gcc.gnu.org>
17121         * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
17122         New prototype.
17123         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
17124         Add note to error message to explain internal mapping of overloaded
17125         built-in function name to non-overloaded built-in function name.
17126         * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
17127         function.
17129 2018-04-17  Michael Meissner  <meissner@linux.vnet.ibm.com>
17131         PR target/85424
17132         * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
17133         where the inputs overlap with the output.
17135 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
17137         * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
17138         (=v, v) alternative and explicit "memory" attribute.
17139         (vec_extract_lo_<mode><mask_name>): Likewise.  Also add
17140         "type", "prefix", "prefix_extra", "length_immediate" and "mode"
17141         attributes.
17142         (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
17143         "sselog1" type instead of "sselog".
17144         (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
17145         "sselog".  Remove explicit "memory" attribute.
17146         (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
17147         "type", "prefix", "prefix_extra", "length_immediate" and "mode"
17148         attributes.
17149         (vec_extract_hi_v32hi): Merge all alternatives into one, use
17150         "sselog1" type instead of "sselog".  Remove explicit "memory"
17151         attribute.
17152         (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
17153         use "sselog1" type instead of "sselog".  Remove explicit "memory"
17154         attribute.
17155         (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
17156         "type", "prefix", "prefix_extra", "length_immediate" and "mode"
17157         attributes.
17158         (vec_extract_hi_v64qi): Merge all alternatives into one, use
17159         "sselog1" type instead of "sselog".  Remove explicit "memory"
17160         attribute.
17161         (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
17162         use "sselog1" type instead of "sselog".  Remove explicit "memory"
17163         attribute.
17165         PR target/85430
17166         * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
17168         PR middle-end/85414
17169         * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
17170         on a SUBREG.
17172 2018-04-17  Martin Jambor  <mjambor@suse.cz>
17174         PR ipa/85421
17175         * ipa-cp.c (create_specialized_node): Call
17176         expand_all_artificial_thunks if necessary.
17178 2018-04-17  Martin Liska  <mliska@suse.cz>
17180         PR lto/85405
17181         * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
17182         in message, remote space in between '_G' and '('.
17184 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
17186         PR target/85281
17187         * config/i386/sse.md (reduces<mode><mask_scalar_name>,
17188         avx512f_vmcmp<mode>3<round_saeonly_name>,
17189         avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
17190         avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
17191         avx512f_rndscale<mode><round_saeonly_name>,
17192         avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
17193         avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
17194         Use %<iptr>2 instead of %2 for -masm=intel.
17195         (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
17196         avx512f_vcvttss2usi<round_saeonly_name>,
17197         avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
17198         -masm=intel.
17199         (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
17200         avx512f_vcvttsd2usi<round_saeonly_name>,
17201         avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
17202         Use %q1 instead of %1 for -masm=intel.
17203         (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
17204         avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
17205         of %3 for -masm=intel.
17206         (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
17207         -masm=intel.
17208         (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
17209         -masm=intel.
17210         (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
17211         -masm=intel.
17212         (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
17213         %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
17214         %g1.
17215         (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
17216         -masm=intel.
17217         (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
17218         %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
17219         %g1 and one with %0 and %1.
17220         (avx512er_vmrcp28<mode><round_saeonly_name>,
17221         avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
17222         %1 for -masm=intel.
17223         (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
17224         avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
17225         avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
17226         of %0 and %{%4%} for -masm=intel.
17227         (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
17228         avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
17229         avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
17230         order of %0 and %{%5%}%{z%} for -masm=intel.
17232 2018-04-17  Jan Hubicka  <jh@suse.cz>
17234         PR lto/85405
17235         * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
17237 2018-04-17  Martin Liska  <mliska@suse.cz>
17239         PR ipa/85329
17240         * multiple_target.c (create_dispatcher_calls): Set apostrophes
17241         for target_clone error message.  Make default implementation
17242         clone to be a local declaration.
17243         (separate_attrs): Add new argument and check for an empty
17244         string.
17245         (expand_target_clones): Handle it.
17246         (ipa_target_clone): Make redirection just for target_clones
17247         functions.
17249 2018-04-16  Cesar Philippidis  <cesar@codesourcery.com>
17250             Tom de Vries  <tom@codesourcery.com>
17252         PR middle-end/84955
17253         * omp-expand.c (expand_oacc_for): Add dummy false branch for
17254         tiled basic blocks without omp continue statements.
17256 2018-04-16  Aaron Sawdey  <acsawdey@linux.ibm.com>
17258         PR target/83660
17259         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
17260         vec_extract expression as having side effects to make sure it gets
17261         a cleanup point.
17263 2018-04-16  H.J. Lu  <hongjiu.lu@intel.com>
17265         PR target/85403
17266         * config/i386/i386.c (get_builtin_code_for_version): Check
17267         error_mark_node.
17269 2018-04-16  Olga Makhotina  <olga.makhotina@intel.com>
17271         PR target/84331
17272         * config.gcc: Support "skylake".
17273         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
17274         PROCESSOR_SKYLAKE.
17275         * config/i386/i386.c (m_SKYLAKE): Define.
17276         (processor_target_table): Add "skylake".
17277         (ix86_option_override_internal): Add "skylake".
17278         (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
17279         PROCESSOR_CANNONLAKE.
17280         (get_builtin_code_for_version): Fix priority for
17281         PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
17282         PROCESSOR_SKYLAKE-AVX512.
17283         * config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
17284         (processor_type): Add PROCESSOR_SKYLAKE.
17286 2018-04-16  Paolo Carlini  <paolo.carlini@oracle.com>
17287             Jason Merrill  <jason@redhat.com>
17289         PR c++/85112
17290         * convert.c (convert_to_integer_1): Use direct recursion for
17291         enumeral types and types with a precision less than the number
17292         of bits in their mode.
17294 2018-04-16  Julia Koval  <julia.koval@intel.com>
17296         PR target/84413
17297         * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
17298         X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
17300 2018-04-14  Segher Boessenkool  <segher@kernel.crashing.org>
17302         PR target/85293
17303         * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
17304         * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
17305         and -mno-direct-move.
17307 2018-04-13  Paul A. Clarke  <pc@us.ibm.com>
17309         PR target/83402
17310         * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
17311         Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
17312         Ensure negative shifts result in {0}.
17314 2018-04-13  Vladimir Makarov  <vmakarov@redhat.com>
17316         PR rtl-optimization/79916
17317         * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
17318         regs (if any) to define how to gnerate SD moves when LRA is in
17319         progress.
17321 2018-04-13  Jakub Jelinek  <jakub@redhat.com>
17323         PR rtl-optimization/85393
17324         * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
17325         * except.c (expand_dw2_landing_pad_for_region): Make static.
17326         * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
17327         a label and unconditional jump to old_bb, rather than
17328         expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
17329         basic block.
17331         PR rtl-optimization/85376
17332         * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
17333         zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
17334         instead of a specific value.
17336 2018-04-13  Jan Hubicka  <hubicka@ucw.cz>
17337             Bin Cheng  <bin.cheng@arm.com>
17339         PR tree-optimization/82965
17340         PR tree-optimization/83991
17341         * cfgloopanal.c (expected_loop_iterations_unbounded): Add
17342         by_profile_only parameter.
17343         * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
17344         information if the loop was predicted to iterate too many times.
17345         * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
17347 2018-04-13  Jan Hubicka  <hubicka@ucw.cz>
17349         PR lto/71991
17350         * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
17351         always inline.
17353 2018-04-13  Martin Liska  <mliska@suse.cz>
17354             Jakub Jelinek  <jakub@redhat.com>
17356         PR middle-end/81657
17357         * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
17358         * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
17359         * builtins.c (expand_builtin_memory_copy_args): Use
17360         BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
17361         handle dest_addr == pc_rtx.
17363 2018-04-12  Segher Boessenkool  <segher@kernel.crashing.org>
17365         PR target/85291
17366         * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
17367         asked to not generate direct moves.
17368         (fix_trunc<mode>si2_stfiwx): Similar.
17369         (fix_trunc<mode>si2_internal): Similar.
17371 2018-04-12  Jakub Jelinek  <jakub@redhat.com>
17373         PR debug/83157
17374         * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
17375         * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
17376         lookup if dest in some wider mode is known to be const0_rtx and
17377         if so, record permanent equivalence for it to be ZERO_EXTEND of
17378         the narrower mode destination.
17380 2018-04-12  Cesar Philippidis  <cesar@codesourcery.com>
17382         * lto-streamer-out.c (output_function): Revert 259346.
17383         * omp-expand.c (expand_oacc_for): Likewise.
17385 2018-04-12  Alexander Monakov  <amonakov@ispras.ru>
17387         PR rtl-optimization/85354
17388         * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
17389         * sel-sched.c (sel_global_init): ... here.
17391 2018-04-12  Eric Botcazou  <ebotcazou@adacore.com>
17393         PR target/85238
17394         * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
17395         * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
17396         mode for PE-COFF targets.
17397         * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
17398         (i386_pe_asm_lto_end): Likewise.
17399         * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
17400         (TARGET_ASM_LTO_END): Likewise.
17401         * config/i386/winnt.c (saved_debug_info_level): New static variable.
17402         (i386_pe_asm_lto_start): New function.
17403         (i386_pe_asm_lto_end): Likewise.
17405 2018-04-12  Cesar Philippidis  <cesar@codesourcery.com>
17406             Richard Biener  <rguenther@suse.de>
17408         PR middle-end/84955
17409         * lto-streamer-out.c (output_function): Fix CFG loop state before
17410         streaming out.
17411         * omp-expand.c (expand_oacc_for): Handle calls to internal
17412         functions like regular functions.
17414 2018-04-12  Richard Biener  <rguenther@suse.de>
17416         PR lto/85371
17417         * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
17418         for the early LTO debug to properly generate references to it
17419         during DIE emission.  Do not re-use that for the skeleton for
17420         split-dwarf.
17421         (dwarf2out_early_finish): Likewise.
17423 2018-04-12  Jakub Jelinek  <jakub@redhat.com>
17425         PR target/85328
17426         * config/i386/sse.md
17427         (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
17428         <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
17429         vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
17430         vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
17431         and output is a reg, avoid creating invalid lowpart subreg, but
17432         instead split into a 512-bit move.  Don't split if not AVX512VL,
17433         input is xmm16+ reg and output is a mem.
17434         (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
17435         vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
17436         xmm16+ reg and output is a mem.
17438 2018-04-12  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
17440         * config/s390/s390.c (s390_output_indirect_thunk_function): Check
17441         also for flag_dwarf2_cfi_asm.
17443 2018-04-12  Jakub Jelinek  <jakub@redhat.com>
17445         PR rtl-optimization/85342
17446         * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
17447         a bool scalar var inside of the loop instead.  Don't try to update
17448         recog_data.operand after failed apply_change_group.
17450 2018-04-12  Tom de Vries  <tom@codesourcery.com>
17452         PR target/85296
17453         * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
17454         (nvptx_assemble_decl_begin): Add undefined param.  Declare undefined
17455         array with flexible array member as array without given dimension.
17456         (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
17457         argument for undefined param to true.
17459 2018-04-11  Aaron Sawdey  <acsawdey@linux.ibm.com>
17461         PR target/85321
17462         * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
17463         -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
17464         from PowerPC section.
17465         * config/rs6000/sysv4.opt (mcall-): Improve help text.
17466         * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
17467         help text that is too long.
17468         * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
17469         help text that is too long.
17470         * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
17471         help text that is too long.
17473 2018-04-11  Uros Bizjak  <ubizjak@gmail.com>
17475         * config/alpha/alpha.md (stack_probe_internal): Rename
17476         from "probe_stack".  Update all callers.
17478 2018-04-11  Alexander Monakov  <amonakov@ispras.ru>
17480         PR rtl-optimization/84566
17481         * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
17482         sched_macro_fuse_insns.
17484 2018-04-11  Alexander Monakov  <amonakov@ispras.ru>
17486         PR target/84301
17487         * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
17488         (compute_block_dependences): ... from here.
17490 2018-04-11  Jakub Jelinek  <jakub@redhat.com>
17492         PR tree-optimization/85331
17493         * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
17494         from int to HOST_WIDE_INT.
17496 2018-04-11  Martin Jambor  <mjambor@suse.cz>
17498         PR ipa/84149
17499         * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
17500         (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
17501         not the same as the source val.
17502         (cgraph_edge_brings_value_p): New parameter.
17503         (gather_edges_for_value): Pass destination value to
17504         cgraph_edge_brings_value_p.
17505         (perhaps_add_new_callers): Likewise.
17506         (get_info_about_necessary_edges): Likewise and exclude values brought
17507         only by self-recursive edges.
17508         (create_specialized_node): Redirect only clones of self-calling edges.
17509         (+self_recursive_pass_through_p): New function.
17510         (find_more_scalar_values_for_callers_subset): Use it.
17511         (find_aggregate_values_for_callers_subset): Likewise.
17512         (known_aggs_to_agg_replacement_list): Removed.
17513         (decide_whether_version_node): Re-calculate known constants for all
17514         remaining context clones.
17516 2018-04-11  Richard Biener  <rguenther@suse.de>
17518         PR lto/85339
17519         * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
17520         from early DWARF output.
17521         (dwarf2out_early_finish): Output line info unconditionally into
17522         early DWARF and add reference to it.
17524 2018-04-11  Jakub Jelinek  <jakub@redhat.com>
17526         PR target/85281
17527         * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
17528         (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
17529         other than V2DFmode using iptr mode attribute.
17530         (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
17532 2018-04-11  Alexander Monakov  <amonakov@ispras.ru>
17534         PR rtl-optimization/84659
17535         * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
17537 2018-04-11  Jakub Jelinek  <jakub@redhat.com>
17539         PR debug/85302
17540         * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
17541         SIZEP is NULL.
17542         (output_loc_list): Pass address of a dummy size variable even in the
17543         locview handling loop.
17544         (index_location_lists): Add comment on why skip_loc_list_entry can't
17545         call size_of_locs.
17547 2018-04-11  Thomas Preud'homme  <thomas.preudhomme@arm.com>
17549         PR target/85261
17550         * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
17551         into register.
17553 2018-04-10  Aaron Sawdey  <acsawdey@linux.ibm.com>
17555         PR target/85321
17556         * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
17557         -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
17558         and -mstring-compare-inline-limit.
17560 2018-04-10  Segher Boessenkool  <segher@kernel.crashing.org>
17562         PR target/85287
17563         * config/rs6000/rs6000.md (allocate_stack): Put the residual size
17564         for stack clash protection in a register whenever we need it to be in
17565         a register.
17567 2018-04-10  Segher Boessenkool  <segher@kernel.crashing.org>
17569         * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
17570         Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
17572 2018-04-10  Segher Boessenkool  <segher@kernel.crashing.org>
17574         PR target/85321
17575         * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
17576         the help text.
17577         (mlong-double-): Ditto.
17578         * config/rs6000/sysv4.opt (msdata=): Ditto.
17579         (mtls-size=): Ditto.
17581 2018-04-10  Kelvin Nilsen  <kelvin@gcc.gnu.org>
17583         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
17584         erroneous entries for
17585         "vector int vec_ldl (int, long int *)", and
17586         "vector unsigned int vec_ldl (int, unsigned long int *)".
17587         Add comments and entries for
17588         "vector bool char vec_ldl (int, bool char *)",
17589         "vector bool short vec_ldl (int, bool short *)",
17590         "vector bool int vec_ldl (int, bool int *)",
17591         "vector bool long long vec_ldl (int, bool long long *)",
17592         "vector pixel vec_ldl (int, pixel *)",
17593         "vector long long vec_ldl (int, long long *)",
17594         "vector unsigned long long vec_ldl (int, unsigned long long *)".
17595         * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
17596         type tree bool_long_long_type_node and correct definition of
17597         bool_V2DI_type_node to make reference to this new type tree.
17598         (rs6000_mangle_type): Replace erroneous reference to
17599         bool_long_type_node with bool_long_long_type_node.
17600         * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
17601         comments to emphasize sign distinctions for char and int types and
17602         replace RS6000_BTI_bool_long constant with
17603         RS6000_BTI_bool_long_long constant.  Also add comment to restrict
17604         use of RS6000_BTI_pixel.
17605         (bool_long_type_node): Remove this macro definition.
17606         (bool_long_long_type_node): New macro definition
17608 2018-04-10  Jakub Jelinek  <jakub@redhat.com>
17610         PR rtl-optimization/85300
17611         * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
17612         into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
17613         simplify_unary_operation fails.
17615 2018-04-10  Martin Liska  <mliska@suse.cz>
17617         * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
17618         cgraph_edge and ipa_ref.
17620 2018-04-10  Jakub Jelinek  <jakub@redhat.com>
17622         PR target/85177
17623         PR target/85255
17624         * config/i386/sse.md
17625         (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
17626         computation of the VEC_MERGE selector from mask.
17627         (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
17628         Fix decoding of the VEC_MERGE selector into mask.
17630 2018-04-10  Richard Sandiford  <richard.sandiford@linaro.org>
17632         PR tree-optimization/85286
17633         * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
17635 2018-04-10  Richard Sandiford  <richard.sandiford@linaro.org>
17637         * final.c (final_1): Set insn_last_address as well as
17638         insn_current_address.
17640 2018-04-10  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
17642         PR target/85173
17643         * explow.c (emit_stack_probe): Call validize_mem on memory location
17644         before passing it to gen_probe_stack.  Create address operand and
17645         legitimize it for the probe_stack_address case.
17647 2018-04-09  Jan Hubicka  <jh@suse.cz>
17649         PR lto/85078
17650         * ipa-devirt.c (rebuild_type_inheritance-hash): New.
17651         * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
17652         * tree.c (free_lang_data_in_type): Fix handling of binfos;
17653         walk basetypes.
17654         (free_lang_data): Rebuild type inheritance graph.
17656 2018-04-09  Martin Sebor  <msebor@redhat.com>
17658         * invoke.texi (-finline-small-functions): Mention other optimization
17659         options.
17660         (-findirect-inlining, -fpartial-inlining): Same.
17661         (-finline-functions-called-once): Same.
17662         (-freorder-blocks-and-partition): Same.
17664 2018-04-09  Jan Hubicka  <jh@suse.cz>
17666         PR rtl/84058
17667         * cfgcleanup.c (try_forward_edges): Do not give up on crossing
17668         jumps; choose last target that matches the criteria (i.e.
17669         no partition changes for non-crossing jumps).
17670         * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
17671         support for redirecting crossing jumps to non-crossing.
17673 2018-04-09  Alexey Brodkin  <abrodkin@synopsys.com>
17675         * config/arc/arc.c (arc_expand_prologue): Set stack usage info
17676         also for naked functions.
17678 2018-04-09  Claudiu Zissulescu  <claziss@synopsys.com>
17680         * config/arc/arc.md (add_shift): New pattern.
17681         (add_shift2): Likewise.
17682         (sub_shift): Likewise.
17683         (sub_shift_cmp0_noout): Likewise.
17684         (compare_si_ashiftsi): Likewise.
17685         (xbfu_cmp0_noout): New combine pattern.
17686         (xbfu_cmp0"): Likewise.
17687         (movsi_set_cc_insn): Place the predicable variant first.
17688         (commutative_binary_cmp0_noout): Remove clobber.
17689         (commutative_binary_cmp0): New pattern.
17690         (noncommutative_binary_cmp0): Likewise.
17691         (noncommutative_binary_cmp0_noout): Likewise.
17692         (noncommutative_binary_comparison_result_used): Removed.
17693         (rsub_cmp0): New pattern.
17694         (rsub_cmp0_noout): Likewise.
17695         (extzvsi): Changed, keep only meaningful variants.
17696         (SQH, SEZ): New iterators.
17697         (SQH_postfix): New mode attribute.
17698         (SEZ_prefix): New code attribute.
17699         (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
17700         (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
17701         * config/arc/predicates.md (cc_set_register): Use CC_REG instead
17702         of numerical value.
17703         (noncommutative_operator): Check the availability of barrel
17704         shifter option.
17706 2018-04-09  Richard Biener  <rguenther@suse.de>
17708         PR tree-optimization/85284
17709         * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
17710         Only use the niter constraining form of simple_iv when the exit
17711         is always executed.
17713 2018-04-09  Tom de Vries  <tom@codesourcery.com>
17715         PR target/84041
17716         * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
17717         (define_expand "*memory_barrier"): New define_expand.
17718         (define_insn "memory_barrier"): New insn.
17720 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
17722         PR rtl-optimization/80463
17723         PR rtl-optimization/83972
17724         PR rtl-optimization/83480
17726         * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
17727         correct producer for the insn.
17728         (tidy_control_flow): Fixup seqnos in case of debug insns.
17730 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
17732         PR rtl-optimization/83913
17734         * sel-sched-ir.c (merge_expr_data): Choose the middle between two
17735         different sched-times when merging exprs.
17737 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
17739         PR rtl-optimization/83962
17741         * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
17742         tidy_fallthru_edge and tidy_control_flow.
17744 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
17746         PR rtl-optimization/83530
17748         * sel-sched.c (force_next_insn): New global variable.
17749         (remove_insn_for_debug): When force_next_insn is true, also leave only
17750         next insn in the ready list.
17751         (sel_sched_region): When the region wasn't scheduled, make another pass
17752         over it with force_next_insn set to 1.
17754 2018-04-08  Monk Chiang  <sh.chiang04@gmail.com>
17756         * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
17757         into tm_file.
17758         * config/nds32/constants.md (unspec_volatile_element): Add enum values
17759         for interrupt control.
17760         * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
17761         functions for interrupt control.
17762         * config/nds32/nds32-intrinsic.md: Likewise.
17763         * config/nds32/nds32_intrinsic.h: Likewise.
17764         * config/nds32/nds32.h (nds32_builtins): Likewise.
17766 2018-04-08  Chung-Ju Wu  <jasonwucj@gmail.com>
17768         * config/nds32/nds32.c (nds32_init_machine_status,
17769         nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
17770         strict_aligned_p field.
17771         (nds32_expand_to_rtl_hook): New function.
17772         (TARGET_EXPAND_TO_RTL_HOOK): Define.
17773         * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
17775 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
17776             Chung-Ju Wu  <jasonwucj@gmail.com>
17778         * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
17779         * config/nds32/nds32-n7.md: New file.
17780         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
17781         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
17782         pipeline.
17783         * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
17784         * config/nds32/nds32.md (pipeline_model): Add n7.
17785         * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
17786         * config/nds32/pipelines.md: Include n7 settings.
17788 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
17789             Chung-Ju Wu  <jasonwucj@gmail.com>
17791         * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
17792         * config/nds32/nds32-e8.md: New file.
17793         * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
17794         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
17795         pipeline.
17796         * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
17797         * config/nds32/nds32.md (pipeline_model): Add e8.
17798         * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
17799         * config/nds32/pipelines.md: Include e8 settings.
17801 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
17802             Chung-Ju Wu  <jasonwucj@gmail.com>
17804         * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
17805         * config/nds32/nds32-n8.md: New file.
17806         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
17807         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
17808         pipeline.
17809         * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
17810         * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
17811         * config/nds32/nds32.md (pipeline_model): Add n8.
17812         * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
17813         * config/nds32/pipelines.md: Include n8 settings.
17815 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
17816             Chung-Ju Wu  <jasonwucj@gmail.com>
17818         * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
17819         * config/nds32/nds32-n9-2r1w.md: New file.
17820         * config/nds32/nds32-n9-3r2w.md: New file.
17821         * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
17822         nds32_register_ports): New or modify for cpu n9.
17823         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
17824         pipeline.
17825         * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
17826         * config/nds32/nds32-utils.c: New file.
17827         * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
17828         TARGET_MUL_SLOW): Define.
17829         * config/nds32/nds32.md (pipeline_model): New attribute.
17830         * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
17831         New options that support cpu n9.
17832         * config/nds32/pipelines.md: Include n9 settings.
17833         * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
17835 2018-04-08  Chung-Ju Wu  <jasonwucj@gmail.com>
17837         * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
17838         information if necessary.
17839         (output_cond_branch_compare_zero): Likewise.
17840         * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
17841         (nds32_target_alignment): Refine for alignment.
17842         * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
17843         (FUNCTION_BOUNDARY): Modify.
17844         * config/nds32/nds32.md (call_internal, call_value_internal): Consider
17845         align case.
17846         * config/nds32/nds32.opt (malways-align, malign-functions): New.
17848 2018-04-08  Monk Chiang  <sh.chiang04@gmail.com>
17850         * config/nds32/constants.md (unspec_volatile_element): Add values for
17851         TLB operation and data prefetch.
17852         * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
17853         functions for TLB operation and data prefetch.
17854         * config/nds32/nds32-intrinsic.md: Likewise.
17855         * config/nds32/nds32_intrinsic.h: Likewise.
17856         * config/nds32/nds32.c (nds32_dpref_names): Likewise.
17857         (nds32_print_operand): Likewise.
17858         * config/nds32/nds32.h (nds32_builtins): Likewise.
17860 2018-04-07  Thomas Koenig  <tkoenig@gcc.gnu.org>
17861         Andrew Pinski <pinsika@gcc.gnu.org>
17863         PR middle-end/82976
17864         * match.pd: Use constant_boolean_node of correct type instead of
17865         boolean_true_node or boolean_false_node for simplifying
17866         pointer comparisons to zero.
17868 2018-04-07  Jakub Jelinek  <jakub@redhat.com>
17870         PR tree-optimization/80021
17871         * tree.c (verify_type_variant): Make error call in verify_variant_match
17872         translatable and remove final full stop.
17874 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
17876         * config/nds32/constants.md (unspec_volatile_element): Add
17877         UNSPEC_VOLATILE_EH_RETURN.
17878         * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
17879         nds32_output_stack_pop): Support dwarf exception handling process.
17880         * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
17881         * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
17882         exception handling process.
17883         (nds32_compute_stack_frame): Likewise.
17884         (nds32_return_addr_rtx): Likewise.
17885         (nds32_initial_elimination_offset): Likewise.
17886         (nds32_expand_prologue): Likewise.
17887         (nds32_expand_epilogue): Likewise.
17888         (nds32_dynamic_chain_address): New function.
17889         * config/nds32/nds32.h (machine_function): Add fields for dwarf
17890         exception handling.
17891         (DYNAMIC_CHAIN_ADDRESS): Define.
17892         (EH_RETURN_DATA_REGNO): Define.
17893         (EH_RETURN_STACKADJ_RTX): Define.
17894         * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
17895         patterns for dwarf exception handling.
17897 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
17899         * config/nds32/nds32.h: Clean up obsolete macros.
17901 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
17903         * config/nds32/constants.md (unspec_element, unspec_volatile_element):
17904         Add enum values for particular instructions.
17905         * config/nds32/nds32-intrinsic.c: Implementation of expanding
17906         particular intrinsic functions.
17907         * config/nds32/nds32-intrinsic.md: Likewise.
17908         * config/nds32/nds32_intrinsic.h: Likewise.
17909         * config/nds32/nds32.h (nds32_builtins): Likewise.
17910         * config/nds32/nds32.md (type): Add pbsad and pbsada.
17911         (btst, ave): New patterns for particular instructions.
17913 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
17915         * config/nds32/constants.md (unspec_element, unspec_volatile_element):
17916         Add enum values for atomic load/store and memory sync.
17917         * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
17918         and memory sync.
17919         * config/nds32/nds32-intrinsic.md: Likewise.
17920         * config/nds32/nds32_intrinsic.h: Likewise.
17921         * config/nds32/nds32.h (nds32_builtins): Likewise.
17923 2018-04-07  Jakub Jelinek  <jakub@redhat.com>
17925         PR tree-optimization/85257
17926         * fold-const.c (native_encode_vector): If not all elts could fit
17927         and off is -1, return 0 rather than offset.
17928         * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
17929         (offseti - offset2) / BITS_PER_UNIT as 4th argument to
17930         native_encode_expr.  Verify len * BITS_PER_UNIT >= maxsizei.  Don't
17931         adjust buffer in native_interpret_expr call.
17933 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
17935         * config/nds32/constants.md (unspec_volatile_element): Add cache
17936         control enum values.
17937         * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
17938         * config/nds32/nds32-intrinsic.md: Add cache control patterns.
17939         * config/nds32/nds32.c (nds32_cctl_names): New.
17940         (nds32_print_operand): Handle cache control register names.
17941         * config/nds32/nds32.h (nds32_builtins): New enum values.
17942         * config/nds32/nds32_intrinsic.h: Add cache control enum types and
17943         macros.
17944         * config/nds32/nds32.md (type): Add mmu.
17945         * config/nds32/pipelines.md (simple_insn): Add mmu.
17947 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
17949         * config/nds32/nds32.md (type): Remove call.
17950         * config/nds32/pipelines.md (simple_insn): Likewise.
17952 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
17954         * config/nds32/constants.md (unspec_volatile_element): Add
17955         UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
17956         UNSPEC_VOLATILE_FMFCFG.
17957         * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
17958         description for fmfcfg and fmfcsr.
17959         (bdesc_1arg): Add fmtcsr.
17960         (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
17961         (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
17962         * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
17963         unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
17964         unspec_fmfcfg): New patterns.
17965         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
17966         NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
17967         NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
17968         * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
17969         __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
17970         __nds32__fmfcfg): Define.
17972 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
17974         * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
17975         intrinsic register names.
17976         * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
17977         intrinsic register enum values and macros.
17979 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
17981         * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
17982         for load/store addressing form.
17983         (nds32_print_operand_address): Likewise.
17985 2018-04-06  Eric Botcazou  <ebotcazou@adacore.com>
17987         PR target/85196
17988         * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
17989         based on LABEL_REF.  Remove useless assertion.
17990         (pic_address_needs_scratch): Fix formatting.
17991         (sparc_legitimize_pic_address): Minor tweaks.
17992         (sparc_delegitimize_address): Adjust assertion accordingly.
17993         * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
17994         into symbolic_operand.
17995         (movsi_high_pic_label_ref): Likewise.
17996         (movsi_lo_sum_pic_label_ref): Likewise.
17997         (movdi_pic_label_ref): Likewise.
17998         (movdi_high_pic_label_ref): Likewise.
17999         (movdi_lo_sum_pic_label_ref): Likewise.
18001 2018-04-06  Amaan Cheval  <amaan.cheval@gmail.com>
18003         * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
18004         custom LIB_SPEC setup.
18006 2018-04-06  Ruslan Bukin  <br@bsdpad.com>
18007             Kito Cheng  <kito.cheng@gmail.com>
18009         * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
18010         * config/riscv/freebsd.h: New.
18012 2018-04-06  Chung-Ju Wu  <jasonwucj@gmail.com>
18014         * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
18015         * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
18016         file.
18018 2018-04-06  Chung-Ju Wu  <jasonwucj@gmail.com>
18019             Kito Cheng  <kito.cheng@gmail.com>
18021         * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
18022         nds32_output_call, nds32_symbol_binds_local_p): New functions.
18023         * config/nds32/nds32-protos.h (nds32_output_call,
18024         nds32_output_return): Declare.
18025         * config/nds32/nds32.md: Refine all the call and return patterns.
18027 2018-04-06  Jakub Jelinek  <jakub@redhat.com>
18029         PR debug/85252
18030         * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
18031         build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
18033         PR rtl-optimization/84872
18034         * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
18035         nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
18036         EDGE_CROSSING edge.
18038 2018-04-06  Tamar Christina  <tamar.christina@arm.com>
18040         * expr.c (copy_blkmode_to_reg): Revert 254862.
18041         * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
18043 2018-04-06  Richard Biener  <rguenther@suse.de>
18045         PR middle-end/85244
18046         * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
18047         after seeing a component reference with an adjacent field.  Treat
18048         refs to arrays at struct end of external decls similar to
18049         refs to unconstrained commons.
18051 2018-04-06  Jakub Jelinek  <jakub@redhat.com>
18053         PR sanitizer/85213
18054         * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
18055         look through SAVE_EXPRs with non-side-effects argument.  Adjust
18056         recursive calls.
18057         (fold_comparison): Adjust twoval_comparison_p caller, don't handle
18058         save_p here.
18060 2018-04-06  Richard Biener  <rguenther@suse.de>
18062         PR middle-end/85180
18063         * alias.c (find_base_term): New wrapper around find_base_term
18064         unwinding CSELIB_VAL_PTR changes.
18065         (find_base_term): Do not restore CSELIB_VAL_PTR during the
18066         recursion.
18068 2018-04-06  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
18070         * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
18071         instructions.
18072         * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
18073         constant definitions.
18074         ("nop"): lr 0,0 -> nopr r0
18075         ("nop_lr0", "nop_lr1"): New insn definitions.
18077 2018-04-06  Chung-Ju Wu  <jasonwucj@gmail.com>
18079         * config/nds32/nds32.md (*stack_push, *stack_pop): Use
18080         NDS32_V3PUSH_AVAILABLE_P macro.
18082 2018-04-06  Monk Chiang  <sh.chiang04@gmail.com>
18083             Chung-Ju Wu  <jasonwucj@gmail.com>
18085         * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
18086         (nds32*-*-*): Add float and fpu_config into supported_defaults.
18087         * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
18088         Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
18089         * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
18090         UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
18091         * config/nds32/constraints.md: New constraints and checking for hard
18092         float configuration.
18093         * config/nds32/iterators.md: New mode iterator and attribute for hard
18094         float configuration.
18095         * config/nds32/nds32-doubleword.md: Use hard float alternatives and
18096         patterns.
18097         * config/nds32/nds32-fpu.md: New file.
18098         * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
18099         deal with hard float code generation.
18100         * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
18101         ARCH_V3S.
18102         (abi_type, float_reg_number): New enum type.
18103         * config/nds32/nds32-predicates.c: New predicates for hard float.
18104         * config/nds32/nds32-protos.h: Declare functions for hard float.
18105         * config/nds32/nds32.c: Implementation for hard float configuration.
18106         * config/nds32/nds32.h: Definitions for hard float configuration.
18107         * config/nds32/nds32.md: Include hard float machine description and
18108         modify patterns for hard float configuration.
18109         * config/nds32/nds32.opt: New options for hard float configuration.
18110         * config/nds32/predicates.md: New predicates for hard float
18111         configuration.
18113 2018-04-06  Kuan-Lin Chen  <kuanlinchentw@gmail.com>
18115         * common/config/nds32/nds32-common.c
18116         (nds32_option_optimization_table): Enable -mreleax-hint by default.
18118 2018-04-05  Jakub Jelinek  <jakub@redhat.com>
18120         PR middle-end/85195
18121         * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
18122         CONSTRUCTOR_ELT (ctor, ...)->value.
18124 2018-04-05  Uros Bizjak  <ubizjak@gmail.com>
18126         PR target/85193
18127         * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
18129 2018-04-05  Tom de Vries  <tom@codesourcery.com>
18131         PR target/85204
18132         * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
18133         cond jump.
18135 2018-04-05  Shiva Chen  <shiva0217@gmail.com>
18136             Kito Cheng  <kito.cheng@gmail.com>
18138         * config/nds32/constraints.md (U33): Fine-tune checking condition.
18139         * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
18140         * config/nds32/nds32.h (nds32_16bit_address_type): Add
18141         ADDRESS_POST_MODIFY_LO_REG_IMM3U.
18143 2018-04-05  Shiva Chen  <shiva0217@gmail.com>
18144             Kito Cheng  <kito.cheng@gmail.com>
18146         * config/nds32/constraints.md (Ufe): New memory constraint.
18147         * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
18148         nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
18149         * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
18150         operands.
18151         * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
18152         * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
18154 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
18156         * config/nds32/nds32.md: Use optimize_size in the condition for
18157         alu-shift instructions.
18159 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
18161         * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
18163 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
18165         * config/nds32/nds32.md (negsi2): Refine pattern.
18167 2018-04-05  Kito Cheng  <kito.cheng@gmail.com>
18168             Chung-Ju Wu  <jasonwucj@gmail.com>
18170         * config/nds32/iterators.md (shift_rotate): New code iterator.
18171         (shift): New code attribute.
18172         * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
18173         * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
18174         * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
18175         * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
18176         bit-wise operations.
18177         (andsi3, *andsi3): Ditto.
18178         (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
18179         (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
18180         (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
18181         * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
18182         nds32_ior_operand, nds32_xor_operand): New predicates.
18184 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
18186         * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
18187         (addsi3, subsi3): ... this.
18189 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
18191         * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
18193 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
18195         * config/nds32/nds32.md: Adjust indention.
18197 2018-04-05  Kito Cheng  <kito.cheng@gmail.com>
18199         * config/nds32/nds32.md (feature): New attribute.
18201 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
18203         * config/nds32/nds32.md (subtype): New attribute.
18205 2018-04-04  Thomas Preud'homme  <thomas.preudhomme@arm.com>
18207         PR target/85203
18208         * config/arm/arm-builtins.c (arm_expand_builtin): Change
18209         expansion to perform a bitwise AND of the argument followed by a
18210         boolean negation of the result.
18212 2018-04-04  Peter Bergner  <bergner@vnet.ibm.com>
18214         PR rtl-optimization/84878
18215         * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
18216         the basic block.  Assert the use reference is not artificial and that
18217         it has an associated insn.
18219 2018-04-04  Michael Matz  <matz@suse.de>
18221         * builtins.c (compute_objsize): Pass correct operand
18222         to array_at_struct_end_p.
18224 2018-04-04  Richard Biener  <rguenther@suse.de>
18226         PR lto/85176
18227         * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
18228         from contexts for DINFO_LEVEL_TERSE and below.
18230 2018-04-04  Kito Cheng  <kito.cheng@gmail.com>
18232         * config/nds32/nds32-doubleword.md (move_<mode>): Require
18233         resiter_operand condition.
18234         * config/nds32/nds32.md (*move<mode>): Ditto.
18236 2018-04-04  Kito Cheng  <kito.cheng@gmail.com>
18237             Monk Chiang  <sh.chiang04@gmail.com>
18239         * config/nds32/nds32.md (movmisalign<mode>): New pattern.
18241 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
18243         * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
18245 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
18246             Kito Cheng  <kito.cheng@gmail.com>
18248         * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
18249         nds32_cond_code_str, output_cond_branch,
18250         output_cond_branch_compare_zero, nds32_expand_cbranch,
18251         nds32_expand_cstore, nds32_expand_movcc,
18252         nds32_output_cbranchsi4_equality_zero,
18253         nds32_output_cbranchsi4_equality_reg,
18254         nds32_output_cbranchsi4_equality_reg_or_const_int,
18255         nds32_output_cbranchsi4_greater_less_zero: New functions.
18256         * config/nds32/nds32-protos.h (nds32_expand_cbranch,
18257         nds32_expand_cstore, nds32_expand_movcc,
18258         nds32_output_cbranchsi4_equality_zero,
18259         nds32_output_cbranchsi4_equality_reg,
18260         nds32_output_cbranchsi4_equality_reg_or_const_int,
18261         nds32_output_cbranchsi4_greater_less_zero): Declare.
18262         * config/nds32/predicates.md (nds32_movecc_comparison_operator,
18263         nds32_rimm11s_operand): New predicates.
18264         * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
18265         * config/nds32/nds32.md: Rewrite all the branch and conditional move
18266         patterns.
18268 2018-04-04  Kito Cheng  <kito.cheng@gmail.com>
18270         * config/nds32/nds32-doubleword.md: Refine all the instruction type.
18271         * config/nds32/nds32.md: Ditto.
18272         * config/nds32/pipelines.md: Ditto.
18274 2018-04-04  Richard Biener  <rguenther@suse.de>
18276         PR tree-optimization/85168
18277         * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
18278         propagating abnormals.
18280 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
18282         * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
18284 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
18285             Kito Cheng  <kito.cheng@gmail.com>
18287         * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
18288         * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
18289         * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
18290         (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
18291         * config/nds32/nds32.md (sibcall_internal): New.
18292         (sibcall_register): Remove.
18293         (sibcall_immediate): Remove.
18294         (sibcall_value_internal): New.
18295         (sibcall_value_register): Remove.
18296         (sibcall_value_immediate): Remove.
18297         * config/nds32/predicates.md (nds32_general_register_operand): New.
18298         (nds32_call_address_operand): New.
18300 2018-04-03  Jakub Jelinek  <jakub@redhat.com>
18302         PR rtl-optimization/85167
18303         * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
18304         bb_defs if *split_p, instead preinitialize it to NULL.
18306         PR tree-optimization/85156
18307         * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
18308         evaluating the argument multiple times.
18310 2018-04-03  Bill Schmidt  <wschmidt@linux.ibm.com>
18312         * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
18313         than vector.
18314         (_mm_cvtpd_ps): Likewise.
18315         (_mm_cvttpd_epi32): Likewise.
18316         * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
18317         * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
18318         vector, pixel, and bool following altivec.h include.
18320 2018-04-03  Martin Sebor  <msebor@redhat.com>
18322         * doc/extend.texi (Common Function Attributes): Clarify.
18323         (const attribute): Likewise.
18324         (pure attribute): Likewise.
18326 2018-04-03  Jakub Jelinek  <jakub@redhat.com>
18328         PR target/85169
18329         * config/i386/i386.c (ix86_expand_vector_set): Use
18330         HOST_WIDE_INT_1U << elt instead of 1 << elt.  Formatting fix.
18332 2018-04-03  Uros Bizjak  <ubizjak@gmail.com>
18334         * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
18335         instructions when changing rounding bits to preserve precision bits
18336         in the x87 control word.
18338 2018-04-03  Martin Liska  <mliska@suse.cz>
18340         PR tree-optimization/82491
18341         * rtl.h (strip_offset_and_add): Replace += suboffset with
18342         poly_uint64 () + suboffset.
18344 2018-03-29  Martin Liska  <mliska@suse.cz>
18345             Martin Jambor  <mjambor@suse.cz>
18347         PR ipa/84947
18348         * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
18349         param_type is not an integral or pointer type.
18351 2018-04-03  Richard Biener  <rguenther@suse.de>
18353         * sese.h (recompute_all_dominators): Remove.
18355 2018-04-02  Martin Sebor  <msebor@redhat.com>
18357         * doc/invoke.texi (-Wrestrict): Fix typos.
18359 2018-04-02  Jim Wilson  <jimw@sifive.com>
18361         * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
18362         * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
18363         (<optab>di3, <optab>si3_extend): Likewise.
18364         (<optab>si3_mask, <optab>si3_mask_1): New.
18365         (<optab>di3_mask, <optab>di3_mask_1): New.
18366         (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
18367         (lshrsi3_zero_extend_1): Use VOIDmode shift count.
18368         * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
18370 2018-04-02  Gerald Pfeifer  <gerald@pfeifer.com>
18372         * doc/cpp.texi (Variadic Macros): Fix line continuation in an
18373         example.
18375 2018-04-02  Chung-Ju Wu  <jasonwucj@gmail.com>
18377         * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
18378         (nds32_canonicalize_comparison): New function.
18380 2018-04-01  Chung-Ju Wu  <jasonwucj@gmail.com>
18381             Kito Cheng  <kito.cheng@gmail.com>
18382             Kuan-Lin Chen  <kuanlinchentw@gmail.com>
18384         * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
18385         * config/nds32/constants.md (unspec_volatile_element): Add
18386         UNSPEC_VOLATILE_RELAX_GROUP.
18387         * config/nds32/nds32-relax-opt.c: New file.
18388         * config/nds32/nds32-predicates.c
18389         (nds32_symbol_load_store_p): New function.
18390         * config/nds32/nds32-protos.h
18391         (nds32_symbol_load_store_p): Declare function.
18392         (make_pass_nds32_relax_opt): Declare new rtl pass function.
18393         * config/nds32/nds32.c
18394         (nds32_register_pass): New function to register pass.
18395         (nds32_register_passes): New function to register passes.
18396         * config/nds32/nds32.md (relax_group): New pattern.
18397         * config/nds32/nds32.opt (mrelax-hint): New option.
18398         * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
18400 2018-04-01  Kito Cheng  <kito.cheng@gmail.com>
18402         * config/nds32/t-nds32: Modify files dependency.
18404 2018-04-01  Chung-Ju Wu  <jasonwucj@gmail.com>
18406         * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
18407         (PROFILE_HOOK): Define its implementation.
18409 2018-04-01  Chung-Ju Wu  <jasonwucj@gmail.com>
18411         * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
18412         type and 32-bit size.
18414 2018-04-01  Jakub Jelinek  <jakub@redhat.com>
18416         PR middle-end/85090
18417         * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
18418         (V_128_256): New mode iterator.
18419         (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
18420         (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
18421         (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
18422         of V.
18423         * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
18424         V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
18426 2018-03-31  Segher Boessenkool  <segher@kernel.crashing.org>
18428         PR target/83315
18429         * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
18430         NaN inputs correctly.
18432 2018-03-30  Peter Bergner  <bergner@vnet.ibm.com>
18434         PR target/80546
18435         * config/rs6000/vsx.md (??r): New mode attribute.
18436         (*vsx_mov<mode>_64bit): Use it.
18437         (*vsx_mov<mode>_32bit): Likewise.
18439 2018-03-30  Martin Sebor  <msebor@redhat.com>
18441         PR tree-optimization/84818
18442         * builtins.c (check_access): Use warning_n.
18444 2018-03-30  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
18446         PR target/83822
18447         * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
18448         condition.
18449         * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
18450         condition.
18452 2018-03-30  Julia Koval  <julia.koval@intel.com>
18454         PR target/84413
18455         * x86-tune.def (movx, partial_reg_dependency): Enable for
18456         m_SKYLAKE_AVX512.
18458 2018-03-29  Vladimir Makarov  <vmakarov@redhat.com>
18460         PR inline-asm/84985
18461         * lra-constraints.c (process_alt_operands): Move setting
18462         this_alternative_matches below.
18464 2018-03-29  Martin Liska  <mliska@suse.cz>
18466         PR lto/84995.
18467         * doc/invoke.texi: Document how LTO works with debug info.
18468         Describe auto-load support of binutils.  Mention 'x86-64'
18469         as valid option value of -march option.
18471 2018-03-29  Jakub Jelinek  <jakub@redhat.com>
18473         * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
18475         PR c/85094
18476         * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
18477         For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
18478         OEP_NO_HASH_CHECK for recursive call, to avoid exponential
18479         checking.
18481 2018-03-28  Peter Bergner  <bergner@vnet.ibm.com>
18483         PR target/84912
18484         * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
18485         (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
18486         * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
18487         for RS6000_BTM_POWERPC64.
18488         (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
18489         (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
18490         * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
18491         definition.
18492         (DIVDE): Use it.
18493         (DIVDEU): Likewise.
18495 2018-03-28  Carl Love  <cel@us.ibm.com>
18497         Revert
18498         2017-09-27  Carl Love  <cel@us.ibm.com>
18500         * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
18501         (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
18502         * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
18503         fctiw instruction.
18505 2018-03-28  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
18507         * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
18508         instead of __vector bool.
18509         (_mm_max_pu8): Likewise.
18510         (_mm_min_pi16): Likewise.
18512 2018-03-28  Peter Bergner  <bergner@vnet.ibm.com>
18514         PR target/84912
18515         * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
18516         (DIVWEUO): Likewise.
18517         (DIVDEO): Likewise.
18518         (DIVDEUO): Likewise.
18519         * config/rs6000/rs6000.c (builtin_function_type): Remove support for
18520         DIVWEUO and DIVDEUO.
18521         * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
18522         (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
18523         (div_extend): Likewise.
18524         * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
18525         builtin function.
18526         (__builtin_divweuo): Likewise.
18527         (__builtin_divdeo): Likewise.
18528         (__builtin_divdeuo): Likewise.
18530 2018-03-28  Jakub Jelinek  <jakub@redhat.com>
18532         PR target/85095
18533         * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
18534         *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
18536         PR tree-optimization/82004
18537         * gimple-match-head.c (optimize_pow_to_exp): New function.
18538         * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
18539         Don't fold to exp if optimize_pow_to_exp is false.
18541 2018-03-28  Martin Liska  <mliska@suse.cz>
18543         PR other/84819
18544         * calls.c (initialize_argument_information): Fix trailing space.
18545         * common.opt: Fix typo and provide better explanation for
18546         -fsanitize-coverage option.
18547         * config/i386/i386.opt: Fix typo.
18549 2018-03-28  Jakub Jelinek  <jakub@redhat.com>
18550             Martin Liska  <mliska@suse.cz>
18552         PR sanitizer/85081
18553         * gimplify.c (asan_poison_variable): Don't do the check for
18554         gimplify_omp_ctxp here.
18555         (gimplify_decl_expr): Do it here.
18556         (gimplify_target_expr): Likewise.
18558 2018-03-28  Martin Liska  <mliska@suse.cz>
18560         PR target/84988
18561         * config/i386/i386.c (ix86_function_arg_advance): Do not call
18562         chkp_type_bounds_count if MPX is not enabled.
18564 2018-03-27  Chung-Ju Wu  <jasonwucj@gmail.com>
18566         * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
18568 2018-03-27  Michael Meissner  <meissner@linux.vnet.ibm.com>
18570         PR target/84914
18571         * config/rs6000/rs6000.c (create_complex_muldiv): New helper
18572         function to create the function decl for complex long double
18573         multiply and divide for -mabi=ieeelongdouble.
18574         (init_float128_ieee): Call it.
18576 2018-03-27  H.J. Lu  <hongjiu.lu@intel.com>
18578         PR target/85044
18579         * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
18580         -fcf-protection=branch -mibt.
18581         * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
18583 2018-03-27  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
18585         PR target/81863
18586         * config/arm/arm.c (arm_valid_symbolic_address): Handle
18587         arm_word_relocations.
18589 2018-03-27  Cesar Philippidis  <cesar@codesourcery.com>
18591         PR target/85056
18592         * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
18593         extern array declarations.
18595 2018-03-27  Richard Biener  <rguenther@suse.de>
18597         PR middle-end/84067
18598         * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
18599         explicit single_use checks.
18601 2018-03-27  Richard Biener  <rguenther@suse.de>
18603         PR tree-optimization/85082
18604         * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
18605         Valueize the VUSE.
18607 2018-03-27  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
18609         * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
18610         * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
18611         Turn on fasynchronous-unwind-tables and funwind-tables.
18613 2018-03-26  Uros Bizjak  <ubizjak@gmail.com>
18615         PR target/85073
18616         * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
18617         (*bmi_blsr_<mode>_ccz): Ditto.
18619 2018-03-26  Tom de Vries  <tom@codesourcery.com>
18621         PR tree-optimization/85063
18622         * omp-general.c (offloading_function_p): New function.  Factor out
18623         of ...
18624         * omp-offload.c (pass_omp_target_link::gate): ... here.
18625         * omp-general.h (offloading_function_p): Declare.
18626         * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
18627         with attribute omp declare target for offloading functions.
18629 2018-03-24  Richard Sandiford  <richard.sandiford@linaro.org>
18631         PR tree-optimization/84005
18632         * tree-data-ref.h (get_base_for_alignment): Declare.
18633         * tree-data-ref.c (get_base_for_alignment_1): New function.
18634         (get_base_for_alignment): Likewise.
18635         * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
18636         get_base_for_alignment to find a suitable base object, instead
18637         of always using drb->base_address.
18639 2018-03-23  Jakub Jelinek  <jakub@redhat.com>
18641         PR inline-asm/85022
18642         * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
18643         known size by default.
18645 2018-03-23  Vladimir Makarov  <vmakarov@redhat.com>
18647         PR inline-asm/85030
18648         * lra-constraints.c (process_alt_operands): Don't match BLKmode
18649         and non BLKmode operands.
18651 2018-03-23  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
18653         PR target/85026
18654         * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
18655         Clean up attributes.
18657 2018-03-23  Richard Biener  <rguenther@suse.de>
18659         PR debug/85020
18660         * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
18661         we are going to emit early debug for LTO.
18663 2018-03-23  Jakub Jelinek  <jakub@redhat.com>
18665         PR inline-asm/85034
18666         * function.c (match_asm_constraints_1): Don't optimize if input
18667         doesn't satisfy general_operand predicate for output's mode.
18669         PR inline-asm/85022
18670         * alias.c (write_dependence_p): Don't require for x_canonicalized
18671         non-VOIDmode if x has VOIDmode.
18673         PR sanitizer/85029
18674         * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
18675         just don't try to optimize it rather than assert it never happens.
18677 2018-03-22  Kelvin Nilsen  <kelvin@gcc.gnu.org>
18679         * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
18680         macro expansions for definition of ST_INTERNAL_<mode> and
18681         LD_INTERNAL_<mode> builtins.
18682         * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
18683         Remove prototype.
18684         * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
18685         function.
18686         (altivec_expand_st_builtin): Likewise.
18687         (altivec_expand_builtin): Remove calls to deleted functions.
18688         (rs6000_address_for_altivec): Delete this function.
18689         * config/rs6000/vector.md: Remove expands for
18690         vector_altivec_load_<mode> and vector_altivec_store_<mode>.
18692 2018-03-22  Sudakshina Das  <sudi.das@arm.com>
18694         PR target/84826
18695         * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
18696         * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
18697         re-computing once computed.
18698         (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
18699         (arm_init_machine_status): Initialize
18700         machine->static_chain_stack_bytes.
18702 2018-03-22  Kelvin Nilsen  <kelvin@gcc.gnu.org>
18704         PR target/84760
18705         * doc/extend.texi: Add four new prototypes for vec_ld.
18706         * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
18707         definitions for more logical presentation.
18708         * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
18709         entries for V1TI variants of __builtin_altivec_ld builtin.
18710         * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
18711         handling of V1TI variant of LVX icode pattern.
18712         (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
18713         (rs6000_gimple_fold_builtin): Likewise.
18714         (altivec_init_builtins): Add code to define
18715         __builtin_altivec_lvx_v1ti function.
18717 2018-03-22  Jakub Jelinek  <jakub@redhat.com>
18719         PR inline-asm/84941
18720         * function.c (match_asm_constraints_1): Don't do the optimization
18721         if input isn't a REG, SUBREG, MEM or constant.
18723 2018-03-22  Tom de Vries  <tom@codesourcery.com>
18725         PR tree-optimization/84956
18726         * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
18727         bb_has_abnormal_pred.
18729 2018-03-22  Jakub Jelinek  <jakub@redhat.com>
18731         PR sanitizer/85018
18732         * dwarf2asm.c (dw2_output_indirect_constant_1): Set
18733         DECL_INITIAL (decl) to decl at the end.
18734         * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
18735         adjust the comment.
18737 2018-03-21  Joseph Myers  <joseph@codesourcery.com>
18739         * doc/extend.texi (__builtin_tgmath): Document when complex
18740         integer types are treated as _Complex _Float64.
18742 2018-03-21  Tom de Vries  <tom@codesourcery.com>
18744         * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
18746 2018-03-21  Jakub Jelinek  <jakub@redhat.com>
18748         PR tree-optimization/84960
18749         * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
18750         if it is ENTRY block, move them into single succ of ENTRY in that case.
18752 2018-03-21  Richard Sandiford  <richard.sandiford@linaro.org>
18754         PR tree-optimization/84811
18755         * poly-int.h (poly_span_traits): Remove the T3 parameter and
18756         promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
18757         (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
18758         (known_subrange_p): Update accordingly.  Cast each value involved
18759         in the size comparison, rather than casting the result of the
18760         subtraction.
18762 2018-03-21  Jakub Jelinek  <jakub@redhat.com>
18764         PR tree-optimization/84982
18765         * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
18766         by flipping the least significant bit rather than all bits from
18767         bitpos to bitpos + bitsize - 1.
18769 2018-03-21  Nathan Sidwell  <nathan@acm.org>
18771         * doc/extend.texi (Deprecated Features): Remove mention of
18772         long-deleted deprecations.
18774 2018-03-21  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
18776         PR jit/84288
18777         * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
18778         * configure: Regenerate.
18780 2018-03-21  Tom de Vries  <tom@codesourcery.com>
18782         PR tree-optimization/83126
18783         * tree-parloops.c (num_phis): New function.
18784         (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
18786 2018-03-21  Nathan Sidwell  <nathan@acm.org>
18788         * doc/extend.texi (Deprecated Features): Update deprecated flags,
18789         mention anon-struct/union members and trailing attributes.
18791 2018-03-21  Bin Cheng  <bin.cheng@arm.com>
18793         PR tree-optimization/84969
18794         * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
18795         builtin memset partitions if they set different rhs values.
18797 2018-03-21  Jakub Jelinek  <jakub@redhat.com>
18799         PR rtl-optimization/84989
18800         * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
18801         VEC_DUPLICATE with scalar result mode.
18803 2018-03-21  Martin Liska  <mliska@suse.cz>
18805         PR ipa/84963
18806         * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
18807         not intended return statement.
18809 2018-03-21  Martin Liska  <mliska@suse.cz>
18811         PR target/84988
18812         * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
18813         (chkp_find_bound_slots_1): Limit number of iterations.
18815 2018-03-20  David H. Gutteridge  <dhgutteridge@sympatico.ca>
18817         PR target/84838
18818         * Minor grammar fixes for x86 options.
18820 2018-03-20  Jakub Jelinek  <jakub@redhat.com>
18822         PR debug/84875
18823         * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
18824         holding REG_CFA_RESTORE notes, instead turn them into a USE.
18826 2018-03-20  Peter Bergner  <bergner@vnet.ibm.com>
18828         PR target/83789
18829         * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
18830         (altivec_lvx_<mode>_1op): Likewise.
18831         (altivec_stvx_<mode>_2op): Likewise.
18832         (altivec_stvx_<mode>_1op): Likewise.
18833         (altivec_lvx_<VM2:mode>): New define_expand.
18834         (altivec_stvx_<VM2:mode>): Likewise.
18835         (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
18836         (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
18837         (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
18838         (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
18839         * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
18840         (rs6000_gen_lvx): Likewise.
18841         * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
18842         (altivec_expand_stv_builtin): Likewise.
18843         (altivec_expand_builtin): Likewise.
18844         * config/rs6000/vector.md: Likewise.
18846 2018-03-20  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
18848         PR target/82518
18849         * config/arm/arm.c (arm_array_mode_supported_p): Return false for
18850         BYTES_BIG_ENDIAN.
18852 2018-03-20  Richard Biener  <rguenther@suse.de>
18854         PR target/84986
18855         * config/i386/i386.c (ix86_add_stmt_cost): Only cost
18856         sign-conversions as zero, fall back to standard scalar_stmt
18857         cost for the rest.
18859 2018-03-20  Martin Liska  <mliska@suse.cz>
18861         PR ipa/84825
18862         * predict.c (rebuild_frequencies): Handle case when we have
18863         PROFILE_ABSENT, but flag_guess_branch_prob is false.
18865 2018-03-20  Jakub Jelinek  <jakub@redhat.com>
18867         PR target/84990
18868         * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
18869         flag_section_anchors.
18870         * varasm.c (use_blocks_for_decl_p): Remove hack for
18871         dw2_force_const_mem.
18873         PR target/84845
18874         * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
18875         to ...
18876         (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this.  If pseudos can't
18877         be created, use lowpart_subreg of operands[0] rather than operands[0]
18878         itself.
18879         (*aarch64_reg_<mode>3_minus_mask): Rename to ...
18880         (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
18881         (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
18882         and n constraint instead of aarch64_shift_imm_di and Usd.
18883         (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
18884         (*aarch64_<optab>_reg_minus<mode>3): ... this.
18886 2018-03-20  Sudakshina Das  <sudi.das@arm.com>
18888         PR target/82989
18889         * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
18890         to favor GPR over NEON registers.
18891         (<shift>di3_neon): Likewise.
18893 2018-03-20  Tom de Vries  <tom@codesourcery.com>
18895         PR target/84952
18896         * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
18897         (nvptx_process_pars): Emit bar.sync asap and alap.
18899 2018-03-20  Tom de Vries  <tom@codesourcery.com>
18901         PR target/84954
18902         * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
18903         seen_label if seen_label is already set.
18905 2018-03-20  Jakub Jelinek  <jakub@redhat.com>
18907         PR target/84945
18908         * config/i386/i386.c (fold_builtin_cpu): For features above 31
18909         use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
18910         Use 1U instead of 1.  Formatting fixes.
18912         PR c/84953
18913         * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
18914         instead of TREE_TYPE (s1) for the return value.
18916 2018-03-19  Jakub Jelinek  <jakub@redhat.com>
18918         PR tree-optimization/84946
18919         * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
18920         bitsize + bitsize in poly_uint64 rather than poly_int64.
18922         PR sanitizer/78651
18923         * dwarf2asm.c: Include fold-const.c.
18924         (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
18925         of decl rather than decl itself.
18927         PR rtl-optimization/84643
18928         * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
18930 2018-03-19  Maxim Ostapenko  <m.ostapenko@samsung.com>
18932         PR sanitizer/78651
18933         * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
18934         calling assemble_variable.
18936 2018-03-19  Sudakshina Das  <sudi.das@arm.com>
18938         PR target/81647
18939         * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
18940         instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
18942 2018-03-19  Jim Wilson  <jimw@sifive.com>
18944         PR bootstrap/84856
18945         * config/riscv/riscv.c (riscv_function_arg_boundary): Use
18946         PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
18947         (riscv_first_stack_step): Likewise.
18948         (riscv_option_override): Use STACK_BOUNDARY instead of
18949         MIN_STACK_BOUNDARY.
18950         * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
18951         MIN_STACK_BOUNDARY.
18952         (BIGGEST_ALIGNMENT): Set to 128.
18953         (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
18954         (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
18955         STACK_BOUNDARY.
18957 2018-03-19  Richard Biener  <rguenther@suse.de>
18959         PR tree-optimization/84933
18960         * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
18961         values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
18963 2018-03-19  Richard Biener  <rguenther@suse.de>
18965         PR tree-optimization/84859
18966         * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
18967         (cond_if_else_store_replacement): Perform sinking operation on
18968         single-store BBs regardless of MAX_STORES_TO_SINK setting.
18969         Generalize what a BB with a single eligible store is.
18971 2018-03-19  Richard Biener  <rguenther@suse.de>
18973         PR tree-optimization/84929
18974         * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
18975         chrec_is_positive against non-chrec arg.
18977 2018-03-19  Tamar Christina  <tamar.christina@arm.com>
18979         PR target/84711
18980         * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
18982 2018-03-18  Martin Liska  <mliska@suse.cz>
18984         PR rtl-optimization/84635
18985         * regrename.c (build_def_use): Use matches_mode only when
18986         matches >= 0.
18988 2018-03-18  Richard Sandiford  <richard.sandiford@linaro.org>
18990         PR tree-optimization/84913
18991         * tree-vect-loop.c (vectorizable_reduction): Don't try to
18992         vectorize chains of COND_EXPRs.
18994 2018-03-18  Chung-Ju Wu  <jasonwucj@gmail.com>
18996         * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
18998 2018-03-18  Chung-Ju Wu  <jasonwucj@gmail.com>
19000         * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
19002 2018-03-18  Chung-Ju Wu  <jasonwucj@gmail.com>
19004         * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
19006 2018-03-17  Chung-Ju Wu  <jasonwucj@gmail.com>
19007             Kito Cheng  <kito.cheng@gmail.com>
19009         * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
19010         * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
19011         (nds32_adjust_reg_alloc_order): New function.
19012         * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
19014 2018-03-17  Kito Cheng  <kito.cheng@gmail.com>
19016         * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
19017         nds32_print_operand, nds32_print_operand_address): Use
19018         HOST_WIDE_INT_PRINT_DEC instead.
19020 2018-03-17  Chung-Ju Wu  <jasonwucj@gmail.com>
19022         * config/nds32/nds32.c (nds32_register_priority): Modify cost.
19024 2018-03-17  Jakub Jelinek  <jakub@redhat.com>
19026         PR target/84902
19027         * config/i386/i386.c (initial_ix86_tune_features,
19028         initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
19029         unsigned long long.
19030         (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
19031         to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
19032         rather than 1u << ix86_tune.  Formatting fix.
19033         (ix86_option_override_internal): Change ix86_arch_mask from
19034         unsigned int to unsigned HOST_WIDE_INT, initialize to
19035         HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
19036         (ix86_function_specific_restore): Likewise.
19038 2018-03-16  Jakub Jelinek  <jakub@redhat.com>
19040         PR target/84899
19041         * postreload.c (reload_combine_recognize_pattern): Perform
19042         INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
19043         truncate_int_for_mode the result for the destination's mode.
19045         PR c/84909
19046         * hsa-gen.c (mem_type_for_type): Fix comment typo.
19047         * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
19048         Likewise.
19049         * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
19050         Likewise.
19052 2018-03-16  Vladimir Makarov  <vmakarov@redhat.com>
19054         PR target/84876
19055         * lra-assigns.c (lra_split_hard_reg_for): Don't use
19056         regno_allocno_class_array and sorted_pseudos.
19057         * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
19058         insns where regno is used.
19060 2018-03-16  Martin Liska  <mliska@suse.cz>
19062         PR ipa/84833
19063         * multiple_target.c (create_dispatcher_calls): Redirect
19064         reference in the symbol table.
19066 2018-03-16  Martin Liska  <mliska@suse.cz>
19068         PR ipa/84722
19069         * multiple_target.c (create_dispatcher_calls): Redirect also
19070         an alias.
19072 2018-03-16  Jakub Jelinek  <jakub@redhat.com>
19074         PR c++/79937
19075         PR c++/82410
19076         * tree.h (TARGET_EXPR_NO_ELIDE): Define.
19077         * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
19078         TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
19080 2018-03-16  Julia Koval  <julia.koval@intel.com>
19082         * doc/invoke.texi (Skylake Server): Add CLWB.
19083         Cannonlake): Remove CLWB.
19085 2018-03-16  Jakub Jelinek  <jakub@redhat.com>
19087         PR tree-optimization/84841
19088         * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
19089         1 << 3.
19090         (FLOAT_ONE_CONST_TYPE): Define.
19091         (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
19092         (sort_by_operand_rank): Put entries with higher constant_type last
19093         rather than first to match comments.
19095 2018-03-15  Sandra Loosemore  <sandra@codesourcery.com>
19097         * config/nios2/nios2.md (movsi_internal): Fix thinko in
19098         split predicate.
19100 2018-03-15  Jakub Jelinek  <jakub@redhat.com>
19102         PR c++/79085
19103         * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
19104         check and use address of target always.
19106 2018-03-15  H.J. Lu  <hongjiu.lu@intel.com>
19108         PR target/84574
19109         * config/i386/i386.c (indirect_thunk_needed): Update comments.
19110         (indirect_thunk_bnd_needed): Likewise.
19111         (indirect_thunks_used): Likewise.
19112         (indirect_thunks_bnd_used): Likewise.
19113         (indirect_return_needed): New.
19114         (indirect_return_bnd_needed): Likewise.
19115         (output_indirect_thunk_function): Add a bool argument for
19116         function return.
19117         (output_indirect_thunk_function): Don't generate alias for
19118         function return thunk.
19119         (ix86_code_end): Call output_indirect_thunk_function to generate
19120         function return thunks.
19121         (ix86_output_function_return): Set indirect_return_bnd_needed
19122         and indirect_return_needed instead of indirect_thunk_bnd_needed
19123         and indirect_thunk_needed.
19125 2018-03-15  Olga Makhotina  <olga.makhotina@intel.com>
19127         * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
19128         (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
19129         (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
19131 2018-03-15  David Malcolm  <dmalcolm@redhat.com>
19132             Paul Hua <paul.hua.gm@gmail.com>
19134         PR c/84852
19135         * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
19137 2018-03-15  Segher Boessenkool  <segher@kernel.crashing.org>
19139         * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
19140         TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
19141         resp. SFmode cases.
19143 2018-03-15  Tamar Christina  <tamar.christina@arm.com>
19145         PR target/84711
19146         * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
19147         instead of GET_MODE_SIZE when comparing Units.
19149 2018-03-15  Vladimir Mezentsev  <vladimir.mezentsev@oracle.com>
19151         PR target/68256
19152         * varasm.c (hash_section): Return an unchangeble hash value
19153         * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
19154         Return !aarch64_can_use_per_function_literal_pools_p ().
19156 2018-03-15  Jakub Jelinek  <jakub@redhat.com>
19158         PR target/84860
19159         * optabs.c (emit_conditional_move): Pass address of cmode's copy
19160         rather than address of cmode as last argument to prepare_cmp_insn.
19162 2018-03-15  Julia Koval  <julia.koval@intel.com>
19164         * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
19165         F_AVX512VNNI, F_AVX512BITALG): New.
19167 2018-03-14  John David Anglin  <danglin@gcc.gnu.org>
19169         PR target/83451
19170         * config/pa/pa.c (pa_emit_move_sequence):  Always emit secondary reload
19171         insn for floating-point loads and stores.
19173 2018-03-14  Carl Love  <cel@us.ibm.com>
19175         * config/rs6000/rs6000-c.c: Add macro definitions for
19176         ALTIVEC_BUILTIN_VEC_PERMXOR.
19177         * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
19178         * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
19179         * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
19180         * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
19181         UNSPEC_VPERMXOR.
19182         * config/doc/extend.texi: Add prototypes for vec_permxor.
19184 2018-03-14  David Malcolm  <dmalcolm@redhat.com>
19186         PR c/84852
19187         * diagnostic-show-locus.c (class layout_point): Convert m_line
19188         from int to linenum_type.
19189         (line_span::comparator): Use linenum "compare" function when
19190         comparing line numbers.
19191         (test_line_span): New function.
19192         (layout_range::contains_point): Convert param "row" from int to
19193         linenum_type.
19194         (layout_range::intersects_line_p): Likewise.
19195         (layout::will_show_line_p): Likewise.
19196         (layout::print_source_line): Likewise.
19197         (layout::should_print_annotation_line_p): Likewise.
19198         (layout::print_annotation_line): Likewise.
19199         (layout::print_leading_fixits): Likewise.
19200         (layout::annotation_line_showed_range_p): Likewise.
19201         (struct line_corrections): Likewise for field m_row.
19202         (line_corrections::line_corrections): Likewise for param "row".
19203         (layout::print_trailing_fixits): Likewise.
19204         (layout::get_state_at_point): Likewise.
19205         (layout::get_x_bound_for_row): Likewise.
19206         (layout::print_line): Likewise.
19207         (diagnostic_show_locus): Likewise for locals "last_line" and "row".
19208         (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
19209         * input.c (selftest::test_linenum_comparisons): New function.
19210         (selftest::input_c_tests): Call it.
19211         * selftest.c (selftest::test_assertions): Test ASSERT_GT,
19212         ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
19213         * selftest.h (ASSERT_GT): New macro.
19214         (ASSERT_GT_AT): New macro.
19215         (ASSERT_LT): New macro.
19216         (ASSERT_LT_AT): New macro.
19218 2018-03-14  Segher Boessenkool  <segher@kernel.crashing.org>
19220         PR rtl-optimization/84780
19221         * combine.c (distribute_links): Don't make a link based on pc_rtx.
19223 2018-03-14  Martin Liska  <mliska@suse.cz>
19225         * tree.c (record_node_allocation_statistics): Use
19226         get_stats_node_kind.
19227         (get_stats_node_kind): New function extracted from
19228         record_node_allocation_statistics.
19229         (free_node): Use get_stats_node_kind.
19231 2018-03-14  Richard Biener  <rguenther@suse.de>
19233         * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
19234         that the value-set of ANTIC_IN doesn't grow.
19236         Revert
19237         * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
19238         member.
19239         (BB_VISITED_WITH_VISITED_SUCCS): New define.
19240         (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
19242 2018-03-14  Julia Koval  <julia.koval@intel.com>
19244         * config.gcc (icelake-client, icelake-server): New.
19245         (icelake): Remove.
19246         * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
19247         (initial_ix86_arch_features): Ditto.
19248         (PTA_SKYLAKE): Add SGX.
19249         (PTA_ICELAKE): Remove.
19250         (PTA_ICELAKE_CLIENT): New.
19251         (PTA_ICELAKE_SERVER): New.
19252         (ix86_option_override_internal): Split up icelake on icelake client and
19253         icelake server.
19254         (get_builtin_code_for_version): Ditto.
19255         (fold_builtin_cpu): Ditto.
19256         * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
19257         * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
19258         * config/i386/i386.h (processor_type): Ditto.
19259         * doc/invoke.texi: Ditto.
19261 2018-03-14  Jakub Jelinek  <jakub@redhat.com>
19263         PR sanitizer/83392
19264         * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
19265         INTEGER_CST offset, add it together with bitpos / 8 and
19266         sign extend based on POINTER_SIZE.
19268         PR target/84844
19269         Revert
19270         2017-04-20  Uros Bizjak  <ubizjak@gmail.com>
19272         PR target/78090
19273         * config/i386/constraints.md (Yc): New register constraint.
19274         * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
19275         Use Yc constraint for alternative 2 of operand 0.  Remove
19276         preferred_for_speed attribute.
19278 2018-03-14  Richard Biener  <rguenther@suse.de>
19280         PR tree-optimization/84830
19281         * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
19282         with the old one to avoid oscillations.
19284 2018-03-13  Vladimir Makarov  <vmakarov@redhat.com>
19286         PR target/83712
19287         * lra-assigns.c (find_all_spills_for): Ignore uninteresting
19288         pseudos.
19289         (assign_by_spills): Return a flag of reload assignment failure.
19290         Do not process the reload assignment failures.  Do not spill other
19291         reload pseudos if they has the same reg class.  Update n if
19292         necessary.
19293         (lra_assign): Add a return arg.  Set up from the result of
19294         assign_by_spills call.
19295         (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
19296         * lra-constraints.c (split_reg): Add a new arg.  Use it instead of
19297         usage_insns if it is not NULL.
19298         (spill_hard_reg_in_range): New function.
19299         (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
19300         * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
19301         function prototypes.
19302         (lra_assign): Change prototype.
19303         * lra.c (lra): Add code to deal with fails by splitting hard reg
19304         live ranges.
19306 2018-03-01  Palmer Dabbelt  <palmer@sifive.com>
19308         * config/riscv/riscv.opt (mrelax): New option.
19309         * config/riscv/riscv.c (riscv_file_start): Emit ".option
19310         "norelax" when riscv_mrelax is disabled.
19311         * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
19313 2018-03-13  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
19315         PR target/84743
19316         * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
19317         reassociation for int modes.
19319 2018-03-13  Richard Sandiford  <richard.sandiford@linaro.org>
19321         * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
19322         Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
19323         for big-endian.
19324         * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
19325         * config/aarch64/aarch64-sve.md
19326         (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
19327         (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
19328         (*extend<mode><Vwide>2): Rename to...
19329         (aarch64_sve_extend<mode><Vwide>2): ...this.
19330         (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
19331         renaming the old pattern to...
19332         (aarch64_sve_punpk<perm_hilo>_<mode>): ...this.  Only define
19333         unsigned packs.
19334         (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
19335         define_expand, renaming the old pattern to...
19336         (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
19337         (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
19338         (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
19339         account when deciding which SVE instruction the optab should use.
19340         (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
19342 2018-03-13  Richard Sandiford  <richard.sandiford@linaro.org>
19344         * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
19345         (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
19346         (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
19347         (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
19348         (tlsdesc_small_<mode>): Turn a define_expand and use
19349         tlsdesc_small_sve_<mode> for SVE.  Rename original define_insn to...
19350         (tlsdesc_small_advsimd_<mode>): ...this.
19351         (tlsdesc_small_sve_<mode>): New pattern.
19353 2018-03-13  Richard Sandiford  <richard.sandiford@linaro.org>
19355         * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
19356         (UNSPEC_UMUL_HIGHPART): New constants.
19357         (MUL_HIGHPART): New int iteraor.
19358         (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
19359         * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
19360         define_expand.
19361         (*<su>mul<mode>3_highpart): New define_insn.
19363 2018-03-13  Eric Botcazou  <ebotcazou@adacore.com>
19365         PR lto/84805
19366         * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
19367         incomplete types.
19369 2018-03-13  Martin Liska  <mliska@suse.cz>
19371         PR ipa/84658.
19372         * (sem_item_optimizer::sem_item_optimizer): Initialize new
19373         vector.
19374         (sem_item_optimizer::~sem_item_optimizer): Release it.
19375         (sem_item_optimizer::merge_classes): Register variable aliases.
19376         (sem_item_optimizer::fixup_pt_set): New function.
19377         (sem_item_optimizer::fixup_points_to_sets): Likewise.
19378         * ipa-icf.h: Declare new variables and functions.
19380 2018-03-13  Jakub Jelinek  <jakub@redhat.com>
19382         PR middle-end/84834
19383         * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
19384         integer_pow2p@2 and test integer_pow2p in condition.
19385         (A < 0 ? C : 0): Similarly for @1.
19387         PR middle-end/84831
19388         * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
19389         characters starting at p contain '\0' character, don't look beyond
19390         that.
19392         PR target/84827
19393         * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
19394         pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
19396         PR target/84828
19397         * reg-stack.c (change_stack): Change update_end var from int to
19398         rtx_insn *, if non-NULL don't update just BB_END (current_block), but
19399         also call set_block_for_insn on the newly added insns and rescan.
19401         PR target/84786
19402         * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
19403         on the last operand.
19405         PR c++/84704
19406         * tree.c (stabilize_reference_1): Return save_expr (e) for
19407         STATEMENT_LIST even if it doesn't have side-effects.
19409 2018-03-12  Jonathan Wakely  <jwakely@redhat.com>
19411         * doc/invoke.texi (-mclflushopt): Fix spelling of option.
19413 2018-03-12  Renlin Li  <renlin.li@arm.com>
19415         * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
19416         aarch64_output_scalar_simd_mov_immediate.
19418 2018-03-12  Martin Sebor  <msebor@redhat.com>
19420         PR tree-optimization/83456
19421         * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
19422         for perfectly overlapping calls to memcpy.
19423         (gimple_fold_builtin_memory_chk): Same.
19424         (gimple_fold_builtin_strcpy): Handle no-warning.
19425         (gimple_fold_builtin_stxcpy_chk): Same.
19426         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
19428 2018-03-12  Segher Boessenkool  <segher@kernel.crashing.org>
19430         * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
19431         parameter.  Use it for SFmode.
19432         (rs6000_function_arg_advance_1): Adjust.
19433         (rs6000_function_arg): Adjust.
19434         (rs6000_gimplify_va_arg): Pass false for that new parameter.
19436 2018-03-12  Segher Boessenkool  <segher@kernel.crashing.org>
19438         PR rtl-optimization/84169
19439         PR rtl-optimization/84780
19440         * combine.c (can_combine_p): Check for a 2-insn combination whether
19441         the destination register is used between the two insns, too.
19443 2018-03-12  Richard Biener  <rguenther@suse.de>
19445         PR tree-optimization/84803
19446         * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
19447         for refs DR analysis didn't process.
19449 2018-03-12  Richard Biener  <rguenther@suse.de>
19451         PR tree-optimization/84777
19452         * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
19453         force-vectorize loops ignore whether we are optimizing for size.
19455 2018-03-12  Chung-Ju Wu  <jasonwucj@gmail.com>
19457         * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
19458         (TARGET_MD_ASM_ADJUST): Define.
19460 2018-03-12  Monk Chiang  <sh.chiang04@gmail.com>
19461             Kito Cheng  <kito.cheng@gmail.com>
19462             Chung-Ju Wu  <jasonwucj@gmail.com>
19464         * config/nds32/nds32.c (nds32_compute_stack_frame,
19465         nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
19466         nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
19467         nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
19468         nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
19469         * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
19470         NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
19471         * config/nds32/nds32.md (prologue, epilogue): Use macro
19472         NDS32_V3PUSH_AVAILABLE_P to do checking.
19474 2018-03-11  Jakub Jelinek  <jakub@redhat.com>
19476         PR debug/58150
19477         * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
19478         DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
19479         but on TYPE_SIZE.  Don't do anything for ENUM_IS_OPAQUE if not creating
19480         a new die.  Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE.  Guard
19481         addition of most attributes on !orig_type_die or the attribute not
19482         being present already.  Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
19484 2018-03-11  Kito Cheng  <kito.cheng@gmail.com>
19485             Chung-Ju Wu  <jasonwucj@gmail.com>
19487         * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
19488         __NDS32_VH__ macro.
19489         * config/nds32/nds32.opt (mvh): New option.
19491 2018-03-11  Kito Cheng  <kito.cheng@gmail.com>
19492             Chung-Ju Wu  <jasonwucj@gmail.com>
19494         * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
19495         function.
19496         * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
19497         * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
19498         definition.
19500 2018-03-11  Kito Cheng  <kito.cheng@gmail.com>
19501             Chung-Ju Wu  <jasonwucj@gmail.com>
19503         * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
19504         function.
19505         * config/nds32/nds32-multiple.md (strlensi): New pattern.
19506         * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
19508 2018-03-11  Monk Chiang  <sh.chiang04@gmail.com>
19509             Kito Cheng  <kito.cheng@gmail.com>
19510             Chung-Ju Wu  <jasonwucj@gmail.com>
19512         * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
19513         UNSPEC_FFMISM and UNSPEC_FLMISM.
19514         * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
19515         for ffb, ffmism and flmism.
19516         * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
19517         (unspec_ffmism): Ditto.
19518         (unspec_flmism): Ditto.
19519         (nds32_expand_builtin_impl): Check if string extension is available.
19520         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
19521         NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
19523 2018-03-10  Vladimir Makarov  <vmakarov@redhat.com>
19525         Reverting patch:
19526         2018-03-09  Vladimir Makarov  <vmakarov@redhat.com>
19528         PR target/83712
19529         * lra-assigns.c (assign_by_spills): Return a flag of reload
19530         assignment failure.  Do not process the reload assignment
19531         failures.  Do not spill other reload pseudos if they has the same
19532         reg class.
19533         (lra_assign): Add a return arg.  Set up from the result of
19534         assign_by_spills call.
19535         (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
19536         * lra-constraints.c (split_reg): Add a new arg.  Use it instead of
19537         usage_insns if it is not NULL.
19538         (spill_hard_reg_in_range): New function.
19539         (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
19540         * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
19541         function prototypes.
19542         (lra_assign): Change prototype.
19543         * lra.c (lra): Add code to deal with fails by splitting hard reg
19544         live ranges.
19546 2018-03-10  H.J. Lu  <hongjiu.lu@intel.com>
19548         PR target/84807
19549         * config/i386/i386.opt: Replace Enforcment with Enforcement.
19551 2018-03-10  Alexandre Oliva  <aoliva@redhat.com>
19553         PR debug/84620
19554         * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
19555         (dw_val_node): Add val_symbolic_view.
19556         * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
19557         (symview_upper_bound): New.
19558         (new_line_info_table): Initialize symviews_since_reset.
19559         (dwarf2out_source_line): Count symviews_since_reset and set
19560         symview_upper_bound.
19561         (dw_val_equal_p): Handle symview.
19562         (add_AT_symview): New.
19563         (print_dw_val): Handle symview.
19564         (attr_checksum, attr_checksum_ordered): Likewise.
19565         (same_dw_val_p, size_of_die): Likewise.
19566         (value_format, output_die): Likewise.
19567         (add_high_low_attributes): Use add_AT_symview for entry_view.
19568         (dwarf2out_finish): Reset symview_upper_bound, clear
19569         zero_view_p.
19571 2018-03-09  Peter Bergner  <bergner@vnet.ibm.com>
19573         PR target/83969
19574         * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
19575         Add strict argument and use it.
19576         (rs6000_split_multireg_move): Update for new strict argument.
19577         (mem_operand_gpr): Disallow all non-offsettable addresses.
19578         * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
19580 2018-03-09  Jakub Jelinek  <jakub@redhat.com>
19582         PR target/84772
19583         * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
19584         temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
19585         * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
19587         PR c++/84767
19588         * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
19589         decl, use remap_type if we want to use the type.
19591 2018-03-09  Martin Sebor  <msebor@redhat.com>
19593         PR tree-optimization/84526
19594         * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
19595         Remove dead code.
19596         (builtin_access::generic_overlap): Be prepared to handle non-array
19597         base objects.
19599 2018-03-09  Alexandre Oliva  <aoliva@redhat.com>
19601         PR rtl-optimization/84682
19602         * lra-constraints.c (process_address_1): Check is_address flag
19603         for address constraints.
19604         (process_alt_operands): Likewise.
19605         * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
19606         preprocess_constraints.
19607         * recog.h (preprocess_constraints): Add oploc parameter.
19608         Adjust callers.
19609         * recog.c (preprocess_constraints): Test address_operand for
19610         CT_ADDRESS constraints.
19612 2018-03-09  Vladimir Makarov  <vmakarov@redhat.com>
19614         PR target/83712
19615         * lra-assigns.c (assign_by_spills): Return a flag of reload
19616         assignment failure.  Do not process the reload assignment
19617         failures.  Do not spill other reload pseudos if they has the same
19618         reg class.
19619         (lra_assign): Add a return arg.  Set up from the result of
19620         assign_by_spills call.
19621         (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
19622         * lra-constraints.c (split_reg): Add a new arg.  Use it instead of
19623         usage_insns if it is not NULL.
19624         (spill_hard_reg_in_range): New function.
19625         (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
19626         * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
19627         function prototypes.
19628         (lra_assign): Change prototype.
19629         * lra.c (lra): Add code to deal with fails by splitting hard reg
19630         live ranges.
19632 2018-03-09  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
19634         PR target/83193
19635         * common/config/arm/arm-common.c (arm_parse_arch_option_name):
19636         Accept complain bool parameter.  Only emit errors if it is true.
19637         (arm_parse_cpu_option_name): Likewise.
19638         (arm_target_thumb_only): Adjust callers of the above.
19639         * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
19640         prototype to take a default true bool parameter.
19641         (arm_parse_arch_option_name): Likewise.
19643 2018-03-09  David Malcolm  <dmalcolm@redhat.com>
19644             Francois-Xavier Coudert  <fxcoudert@gcc.gnu.org>
19646         PR jit/64089
19647         PR jit/84288
19648         * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
19649         * configure: Regenerate.
19650         * configure.ac ("linker --version-script option"): New.
19651         ("linker soname option"): New.
19653 2018-03-09  Richard Biener  <rguenther@suse.de>
19655         PR tree-optimization/84775
19656         * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
19657         immediate uses of predicate stmts and mark them modified.
19659         Revert
19660         PR tree-optimization/84178
19661         * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
19662         to caller.
19663         (version_loop_for_if_conversion): Delay update_ssa call.
19664         (tree_if_conversion): Delay update_ssa until after predicate
19665         insertion.
19667 2018-03-09  Eric Botcazou  <ebotcazou@adacore.com>
19669         PR target/84763
19670         * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
19671         when the function accesses prior frames.
19673 2018-03-08  Jakub Jelinek  <jakub@redhat.com>
19675         PR debug/84456
19676         * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
19677         gen_llsym, otherwise call maybe_gen_llsym.
19679         PR inline-asm/84742
19680         * recog.c (asm_operand_ok): Return 0 if multi-character constraint
19681         has ',' character inside of it.
19683 2018-03-08  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
19685         PR target/84748
19686         * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
19687         as clobbering CC_REGNUM.
19689 2018-03-08  Richard Biener  <rguenther@suse.de>
19691         PR middle-end/84552
19692         * tree-scalar-evolution.c: Include tree-into-ssa.h.
19693         (follow_copies_to_constant): Do not follow SSA names registered
19694         for update.
19696 2018-03-08  Richard Biener  <rguenther@suse.de>
19698         PR tree-optimization/84178
19699         * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
19700         to caller.
19701         (version_loop_for_if_conversion): Delay update_ssa call.
19702         (tree_if_conversion): Delay update_ssa until after predicate
19703         insertion.
19705 2018-03-08  David Malcolm  <dmalcolm@redhat.com>
19707         PR tree-optimization/84178
19708         * tree-if-conv.c (release_bb_predicate): Remove the
19709         the assertion that the stmts have NULL use_ops.
19710         Discard the statements, asserting that they haven't
19711         yet been added to a BB.
19713 2018-03-08  Richard Biener  <rguenther@suse.de>
19715         PR tree-optimization/84746
19716         * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
19717         (phi_translate): Pass in destination ANTIC_OUT set.
19718         (phi_translate_1): Likewise.  For a simplified result lookup
19719         a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
19720         (phi_translate_set): Adjust.
19721         (do_pre_regular_insertion): Likewise.
19722         (do_pre_partial_partial_insertion): Likewise.
19724 2018-03-08  Martin Liska  <mliska@suse.cz>
19726         PR gcov-profile/84735
19727         * doc/gcov.texi: Document usage of profile files.
19728         * gcov-io.h: Document changes in the format.
19730 2018-03-08  Alexandre Oliva  <aoliva@redhat.com>
19732         PR debug/84404
19733         PR debug/84408
19734         * dwarf2out.c (struct dw_line_info_table): Update comments for
19735         view == -1.
19736         (FORCE_RESET_NEXT_VIEW): New.
19737         (FORCE_RESETTING_VIEW_P): New.
19738         (RESETTING_VIEW_P): Check for -1 too.
19739         (ZERO_VIEW_P): Likewise.
19740         (new_line_info_table): Force-reset next view.
19741         (dwarf2out_begin_function): Likewise.
19742         (dwarf2out_source_line): Simplify zero_view_p initialization.
19743         Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
19744         view directly.  Omit view when omitting .loc at line 0.
19746 2018-03-08  Jakub Jelinek  <jakub@redhat.com>
19748         PR tree-optimization/84740
19749         * tree-switch-conversion.c (process_switch): Call build_constructors
19750         only if info.phi_count is non-zero.
19752         PR tree-optimization/84739
19753         * tree-tailcall.c (find_tail_calls): Check call arguments against
19754         DECL_ARGUMENTS (current_function_decl) rather than
19755         DECL_ARGUMENTS (func) when checking for tail recursion.
19757 2018-03-07  Jakub Jelinek  <jakub@redhat.com>
19759         * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
19760         Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
19761         Volker Reichelt's entry and add entries for people that perform
19762         GCC fuzzy testing and report numerous bugs.
19764 2018-03-07  Segher Boessenkool  <segher@kernel.crashing.org>
19766         PR target/82411
19767         * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
19768         readonly data in sdata, if that is disabled.
19769         * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
19770         * doc/invoke.texi (RS/6000 and PowerPC Options): Document
19771         -mreadonly-in-sdata option.
19773 2018-03-07  Martin Sebor  <msebor@redhat.com>
19775         PR tree-optimization/84468
19776         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
19777         basic block when looking for nul assignment.
19779 2018-03-07  Eric Botcazou  <ebotcazou@adacore.com>
19781         PR target/84277
19782         * except.h (output_function_exception_table): Adjust prototype.
19783         * except.c (output_function_exception_table): Remove FNNAME parameter
19784         and add SECTION parameter.  Ouput one part of the table at a time.
19785         * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
19786         the first part of the exception table and emit unwind directives.
19787         * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
19788         (i386_pe_seh_cold_init): Likewise.
19789         * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
19790         (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
19791         * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
19792         (ix86_output_call_insn): Emit a nop in one more case for SEH.
19793         * config/i386/winnt.c: Include except.h.
19794         (struct seh_frame_state): Add reg_offset, after_prologue and
19795         in_cold_section fields.
19796         (i386_pe_seh_end_prologue): Set seh->after_prologue.
19797         (i386_pe_seh_cold_init): New function.
19798         (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
19799         to seh->in_cold_section.
19800         (seh_emit_push): Record the offset of the push.
19801         (seh_emit_save): Record the offet of the save.
19802         (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
19803         Test seh->after_prologue to disregard the epilogue.
19804         (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
19805         (i386_pe_end_cold_function): New function.
19807 2018-03-07  Jakub Jelinek  <jakub@redhat.com>
19809         PR fortran/84565
19810         * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
19811         aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
19813         PR c++/84704
19814         * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
19815         on tmp_var.
19816         * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
19817         don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
19819         PR middle-end/84723
19820         * multiple_target.c: Include tree-inline.h and intl.h.
19821         (expand_target_clones): Diagnose and fail if node->definition and
19822         !tree_versionable_function_p (node->decl).
19824 2018-03-06  John David Anglin  <danglin@gcc.gnu.org>
19826         * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
19827         sprint_ul.
19828         (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
19829         (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
19830         * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
19832 2018-03-06  Jakub Jelinek  <jakub@redhat.com>
19834         PR target/84710
19835         * combine.c (try_combine): Use reg_or_subregno instead of handling
19836         just paradoxical SUBREGs and REGs.
19838 2018-03-06  Claudiu Zissulescu  <claziss@synopsys.com>
19840         * config/arc/arc.c (arc_finalize_pic): Remove function.
19841         (arc_must_save_register): We use single base PIC register, remove
19842         checks to save/restore the PIC register.
19843         (arc_expand_prologue): Likewise.
19844         * config/arc/arc-protos.h (arc_set_default_type_attributes):
19845         Remove.
19846         (arc_verify_short): Likewise.
19847         (arc_attr_type): Likewise.
19848         * config/arc/arc.c (arc_set_default_type_attributes): Remove.
19849         (walk_stores): Likewise.
19850         (arc_address_cost): Make it static.
19851         (arc_verify_short): Likewise.
19852         (branch_dest): Likewise.
19853         (arc_attr_type): Likewise.
19854         * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
19855         (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
19856         (arc_final_prescan_insn): Remove inserting the nops due to
19857         hardware hazards.  It is done in reorg step.
19858         (insn_length_variant_t): Remove.
19859         (insn_length_parameters_t): Likewise.
19860         (arc_insn_length_parameters): Likewise.
19861         (arc_get_insn_variants): Likewise.
19862         * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
19864 2018-03-06  Jakub Jelinek  <jakub@redhat.com>
19866         PR inline-asm/84683
19867         * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
19868         assertion failure.
19870         PR tree-optimization/84687
19871         * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
19872         on new_node->decl.
19873         * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
19875 2018-03-05  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
19877         * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
19878         Rename to ppc_speculation_barrier.
19879         * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
19880         __builtin_ppc_speculation_barrier.
19882 2018-03-05  Jakub Jelinek  <jakub@redhat.com>
19884         PR target/84700
19885         * combine.c (combine_simplify_rtx): Don't try to simplify if
19886         if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
19887         are equal to x.
19889 2018-03-05  Segher Boessenkool  <segher@kernel.crashing.org>
19891         * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
19892         to 32 bytes when compiling for POWER9.
19894 2018-03-05  Jakub Jelinek  <jakub@redhat.com>
19896         PR target/84564
19897         * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
19898         regparm >= 3 with no arg reg available also for calls with
19899         flag_force_indirect_call.  Pass decl to ix86_function_regparm.
19901         PR target/84524
19902         * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
19903         orig,vex.
19904         (*<plusminus_insn><mode>3): Likewise.  Remove <mask_operand3> uses.
19906 2018-03-05  Peter Bergner  <bergner@vnet.ibm.com>
19908         PR target/84264
19909         * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
19911 2018-03-05  Richard Biener  <rguenther@suse.de>
19913         PR tree-optimization/84486
19914         * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
19915         When inserting a __builtin_assume_aligned call set the LHS
19916         SSA name alignment info accordingly.
19918 2018-03-05  Wilco Dijkstra  <wdijkstr@arm.com>
19920         PR tree-optimization/84114
19921         * config/aarch64/aarch64.c (aarch64_reassociation_width)
19922         Avoid reassociation of FLOAT_MODE addition.
19924 2018-03-05  Olga Makhotina  <olga.makhotina@intel.com>
19926         * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
19927         OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
19928         OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
19929         (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
19930         * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
19931         * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
19932         * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
19933         and -mwbnoinvd.
19934         * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
19935         __builtin_ia32_wbinvd): New builtins.
19936         (SPECIAL_ARGS2): New.
19937         * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
19938         (SPECIAL_ARGS2): New.
19939         * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
19940         (ix86_valid_target_attribute_inner_p): Ditto.
19941         (ix86_init_mmx_sse_builtins): Add special_args2.
19942         * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
19943         TARGET_WBNOINVD_P): New.
19944         * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
19945         (define_insn "wbinvd", define_insn "wbnoinvd"): New.
19946         * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
19947         * config/i386/immintrin.h (_wbinvd): New intrinsic.
19948         * config/i386/pconfigintrin.h: New file.
19949         * config/i386/wbnoinvdintrin.h: Ditto.
19950         * config/i386/x86intrin.h: Add headers pconfigintrin.h and
19951         wbnoinvdintrin.h.
19952         * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
19954 2018-03-05  Richard Biener  <rguenther@suse.de>
19956         PR tree-optimization/84670
19957         * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
19958         member.
19959         (BB_VISITED_WITH_VISITED_SUCCS): New define.
19960         (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
19961         (compute_antic_aux): Only assert the number of values in ANTIC_IN
19962         doesn't grow if all successors (recursively) were visited at least
19963         once.
19965 2018-03-05  Richard Biener  <rguenther@suse.de>
19967         PR tree-optimization/84650
19968         * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
19969         if executed in the loop pipeline.
19971 2018-03-05  Sandra Loosemore  <sandra@codesourcery.com>
19973         * doc/configfiles.texi (Configuration Files): Move info about
19974         conditionalizing $target-protos.h to...
19975         * doc/sourcebuild.texi (Back End): Here.  Explain how $target.h
19976         differs from $target-protos.h.
19978 2018-03-05  Kito Cheng  <kito.cheng@gmail.com>
19979             Chung-Ju Wu  <jasonwucj@gmail.com>
19981         * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
19982         * config/nds32/nds32-multiple.md (setmemsi): Define.
19983         * config/nds32/nds32-memory-manipulation.c
19984         (nds32_gen_dup_4_byte_to_word_value): New.
19985         (emit_setmem_word_loop): New.
19986         (emit_setmem_byte_loop): New.
19987         (nds32_expand_setmem_loop): New.
19988         (nds32_expand_setmem_loop_v3m): New.
19989         (nds32_expand_setmem_unroll): New.
19990         (nds32_expand_setmem): New.
19992 2018-03-04  Kito Cheng  <kito.cheng@gmail.com>
19993             Chung-Ju Wu  <jasonwucj@gmail.com>
19995         * config/nds32/nds32-memory-manipulation.c
19996         (nds32_emit_load_store): New.
19997         (nds32_emit_post_inc_load_store): New.
19998         (nds32_emit_mem_move): New.
19999         (nds32_emit_mem_move_block): New.
20000         (nds32_expand_movmemsi_loop_unknown_size): New.
20001         (nds32_expand_movmemsi_loop_known_size): New.
20002         (nds32_expand_movmemsi_loop): New.
20003         (nds32_expand_movmemsi_unroll): New.
20004         (nds32_expand_movmemqi): Rename ...
20005         (nds32_expand_movmemsi): ... to this.
20006         * config/nds32/nds32-multiple.md (movmemqi): Rename ...
20007         (movmemsi): ... to this.
20008         * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
20009         (nds32_expand_movmemsi): ... to this.
20011 2018-03-04  Kito Cheng  <kito.cheng@gmail.com>
20012             Monk Chiang  <sh.chiang04@gmail.com>
20013             Chung-Ju Wu  <jasonwucj@gmail.com>
20015         * config/nds32/nds32-protos.h
20016         (nds32_expand_load_multiple): New arguments.
20017         (nds32_expand_store_multiple): Ditto.
20018         (nds32_valid_multiple_load_store): Rename ...
20019         (nds32_valid_multiple_load_store_p): ... to this.
20020         * config/nds32/nds32-memory-manipulation.c
20021         (nds32_expand_load_multiple): Refine implementation.
20022         (nds32_expand_store_multiple): Ditto.
20023         * config/nds32/nds32-multiple.md
20024         (load_multiple): Update nds32_expand_load_multiple interface.
20025         (store_multiple): Update nds32_expand_store_multiple interface.
20026         * config/nds32/nds32-predicates.c
20027         (nds32_valid_multiple_load_store): Rename ...
20028         (nds32_valid_multiple_load_store_p): ... to this and refine
20029         implementation.
20030         * config/nds32/predicates.md
20031         (nds32_load_multiple_and_update_address_operation): New predicate.
20032         (nds32_store_multiple_and_update_address_operation): New predicate.
20034 2018-03-04  Kito Cheng  <kito.cheng@gmail.com>
20035             Chung-Ju Wu  <jasonwucj@gmail.com>
20037         * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
20038         (combo): New attribute.
20039         * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
20041 2018-03-03  Chung-Ju Wu  <jasonwucj@gmail.com>
20043         * config/nds32/nds32.opt: Change -mcmodel= default value.
20045 2018-03-03  Kito Cheng  <kito.cheng@gmail.com>
20046             Monk Chiang  <sh.chiang04@gmail.com>
20047             Chung-Ju Wu  <jasonwucj@gmail.com>
20049         * config/nds32/constants.md (unspec_element): New enum.
20050         * config/nds32/constraints.md (Umw): New constraint.
20051         * config/nds32/nds32-intrinsic.c: Add more builtin functions.
20052         * config/nds32/nds32-intrinsic.md: Likewise.
20053         * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
20054         (nds32_valid_smw_lwm_base_p): New.
20055         (nds32_output_smw_single_word): New.
20056         (nds32_output_lmw_single_word): New.
20057         (nds32_expand_unaligned_load): New.
20058         (nds32_expand_unaligned_store): New.
20059         * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
20060         (nds32_output_smw_single_word): Declare.
20061         (nds32_output_lmw_single_word): Declare.
20062         (nds32_expand_unaligned_load): Declare.
20063         (nds32_expand_unaligned_store): Declare.
20064         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
20065         NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
20066         NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
20067         NDS32_BUILTIN_UASTORE_DW.
20068         * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
20069         predicate.
20071 2018-03-03  Monk Chiang  <sh.chiang04@gmail.com>
20072             Kito Cheng  <kito.cheng@gmail.com>
20073             Chung-Ju Wu  <jasonwucj@gmail.com>
20075         * config/nds32/nds32-intrinsic.c
20076         (nds32_expand_builtin_null_ftype_reg): Delete.
20077         (nds32_expand_builtin_reg_ftype_imm): Ditto.
20078         (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
20079         (nds32_read_argument): New.
20080         (nds32_legitimize_target): Ditto.
20081         (nds32_legitimize_argument): Ditto.
20082         (nds32_check_constant_argument): Ditto.
20083         (nds32_expand_unop_builtin): Ditto.
20084         (nds32_expand_unopimm_builtin): Ditto.
20085         (nds32_expand_binop_builtin): Ditto.
20086         (nds32_builtin_decl_impl): Ditto.
20087         (builtin_description): Ditto.
20088         (nds32_expand_builtin_impl): Rewrite with new infrastructure.
20089         (nds32_init_builtins_impl): Ditto.
20090         * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
20091         (nds32_builtin_decl): New.
20092         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
20093         * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
20095 2018-03-02  Jeff Law  <law@redhat.com>
20097         * reorg.c (stop_search_p): Handle DEBUG_INSN.
20098         (redundant_insn, fill_simple_delay_slots): Likewise.
20099         (fill_slots_from_thread): Likewise.
20100         * resource.c (mark_referenced_resources): Likewise.
20101         (mark_set_resources, find_dead_or_set_registers): Likewise.
20103 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
20105         * substring-locations.h (format_warning_va): Formatting fix for
20106         ATTRIBUTE_GCC_DIAG.
20107         (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
20108         argument.
20109         (format_warning_n_va, format_warning_at_substring_n): New prototypes.
20110         * substring-locations.c: Include intl.h.
20111         (format_warning_va): Turned into small wrapper around
20112         format_warning_n_va, renamed to ...
20113         (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
20114         rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
20115         use ngettext.
20116         (format_warning_at_substring_n): New function.
20117         * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
20118         (fmtwarn): Add ATTRIBUTE_GCC_DIAG.  Turn into a copy of
20119         format_warning_at_substring with just a shorter name instead of
20120         const function pointer.
20121         (fmtwarn_n): New function.
20122         (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
20123         appropriate, get rid of all the fmtstr temporaries, move conditionals
20124         with G_() wrapped string literals directly into fmtwarn arguments,
20125         cast dir.len to (int), formatting fixes.
20127 2018-03-02  Thomas Schwinge  <thomas@codesourcery.com>
20129         * doc/invoke.texi: Remove "Cilk Plus" references.
20131 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
20132             Richard Biener  <rguenther@suse.de>
20134         PR ipa/84628
20135         * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
20136         for error or warning attributes if CALL_FROM_THUNK_P is set.
20137         Formatting fixes.
20139 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
20141         PR target/56540
20142         * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
20143         __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
20145         PR target/56540
20146         * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
20147         __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
20149         * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
20150         instead of -1U in last predictors element's probability member.
20152 2018-03-02  Eric Botcazou  <ebotcazou@adacore.com>
20154         PR ipa/83983
20155         * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
20156         arguments if they are comparable.
20158 2018-03-02  Richard Sandiford  <richard.sandiford@linaro.org>
20160         PR tree-optimization/84634
20161         * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
20162         masks and masked_loop_p with a single loop_masks, making sure it's
20163         null for bb vectorization.
20165 2018-03-02  Richard Sandiford  <richard.sandiford@linaro.org>
20167         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
20168         (vect_analyze_data_ref_access): Use loop->safe_len rather than
20169         loop->force_vectorize to check whether there is no alias.
20171 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
20173         PR target/84614
20174         * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
20175         prototypes.
20176         * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
20177         comments.
20178         (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
20179         * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
20180         instead of a loop around prev_real_insn.
20181         * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
20182         prev_real_insn.
20184         PR inline-asm/84625
20185         * config/i386/i386.c (ix86_print_operand): Use conditional
20186         output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
20187         zero vector.
20189 2018-03-02  Richard Biener  <rguenther@suse.de>
20191         PR tree-optimization/84427
20192         * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
20193         (bitmap_set_subtract_values): Rewrite to handle multiple
20194         exprs per value.
20195         (clean): Likewise.
20196         (prune_clobbered_mems): Likewise.
20197         (phi_translate): Take edge instead of pred/phiblock.
20198         (phi_translate_1): Likewise.
20199         (phi_translate_set): Likewise.  Insert all translated
20200         exprs for a value into the set, keeping possibly multiple
20201         expressions per value.
20202         (compute_antic_aux): Adjust for phi_translate changes.
20203         When intersecting union the expressions and prune those
20204         not in the final value set, keeping possibly multiple
20205         expressions per value.  Do not use value-insertion
20206         for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
20207         all expressions.  Add verification that the value-sets
20208         only shrink during iteration.
20209         (compute_partial_antic_aux): Adjust for the phi_translate changes.
20210         (do_pre_regular_insertion): Likewise.
20211         (do_pre_partial_partial_insertion): Likewise.
20213 2018-03-02  Richard Biener  <rguenther@suse.de>
20215         PR target/82005
20216         * config/darwin.c (saved_debug_info_level): New static global.
20217         (darwin_asm_lto_start): Disable debug info generation for LTO out.
20218         (darwin_asm_lto_end): Restore debug info generation settings.
20220 2018-03-01  Martin Liska  <mliska@suse.cz>
20222         PR sanitizer/82484
20223         * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
20224         volatile arguments.
20226 2018-03-01  Richard Biener  <rguenther@suse.de>
20228         PR debug/84645
20229         * dwarf2out.c (gen_variable_die): Properly handle late VLA
20230         type annotation with LTO when debug was disabled at compile-time.
20232 2018-03-01  Matthew Fortune  <mfortune@gmail.com>
20234         * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
20235         XINT with INTVAL.
20236         (mips_final_postscan_insn): Likewise.
20238 2018-03-01  Richard Sandiford  <richard.sandiford@linaro.org>
20240         PR rtl-optimization/84528
20241         * alias.c (init_alias_target): Add commentary.
20242         (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
20243         a unique base value if the frame pointer is not eliminated
20244         to the stack pointer.
20246 2018-03-01  Tom de Vries  <tom@codesourcery.com>
20248         PR rtl-optimization/83327
20249         * lra-int.h (hard_regs_spilled_into): Declare.
20250         * lra.c (hard_regs_spilled_into): Define.
20251         (init_reg_info): Init hard_regs_spilled_into.
20252         * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
20253         * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
20254         (process_bb_lives): Handle hard_regs_spilled_into.
20255         (lra_create_live_ranges_1): Before doing liveness propagation, clear
20256         regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
20258 2018-02-28  David Edelsohn  <dje.gcc@gmail.com>
20260         * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
20261         (powerpc-ibm-aix[789]*): Default to AIX 7.2.
20262         * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
20263         * config/rs6000/aix72.h: New file.
20265 2018-02-28  Jakub Jelinek  <jakub@redhat.com>
20267         * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
20268         instead of warning_at with conditional singular and plural messages
20269         where possible.
20271         PR target/52991
20272         * stor-layout.c (update_alignment_for_field): For
20273         targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
20274         && !DECL_PACKED (field), do the alignment update, just use
20275         only desired_align instead of MAX (type_align, desired_align)
20276         as the alignment.
20277         (place_field): Don't do known_align < desired_align handling
20278         early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
20279         is non-NULL, instead do it after rli->prev_field handling and
20280         only if not within a bitfield word.  For DECL_PACKED (field)
20281         use type_align of BITS_PER_UNIT.
20283 2018-02-28  Eric Botcazou  <ebotcazou@adacore.com>
20285         * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
20286         superfluous parentheses and trailing spaces.
20288 2018-02-28  Richard Biener  <rguenther@suse.de>
20290         PR tree-optimization/84584
20291         * graphite-scop-detection.c (scop_detection::add_scop): Discard
20292         SCoPs with fake exit edge.
20294 2018-02-28  Martin Liska  <mliska@suse.cz>
20296         PR testsuite/84597
20297         * timevar.c (timer::print): Fix format to properly print 100%
20298         values.
20300 2018-02-28  Richard Biener  <rguenther@suse.de>
20302         PR middle-end/84607
20303         * genmatch.c (capture_info::walk_match): Do not mark
20304         captured expressions without operands as expr_p given
20305         they act more like predicates and should be subject to
20306         "lost tail" side-effect preserving.
20308 2018-02-28  Alexandre Oliva  <aoliva@redhat.com>
20310         PR rtl-optimization/81611
20311         * auto-inc-dec.c (attempt_change): Move dead note from
20312         mem_insn if it's the next use of regno
20313         (find_address): Take address use of reg holding
20314         non-incremented value.  Add parm to limit search to the named
20315         reg only.
20316         (merge_in_block): Attempt to use a mem insn that is the next
20317         use of the original regno.
20319 2018-02-27  Martin Sebor  <msebor@redhat.com>
20321         PR c++/83871
20322         * doc/invoke.texi (-Wmissing-attributes): New option.
20323         * print-tree.c (print_node): Handle DECL_UNINLINABLE.
20325 2018-02-27  Martin Sebor  <msebor@redhat.com>
20327         PR translation/84207
20328         * diagnostic-core.h (warning_n, error_n, inform_n): Change
20329         n argument to unsigned HOST_WIDE_INT.
20330         * diagnostic.c (warning_n, error_n, inform_n): Ditto.
20331         (diagnostic_n_impl): Ditto.  Handle arguments in excess of LONG_MAX.
20332         * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
20333         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
20335 2018-02-27  Richard Biener  <rguenther@suse.de>
20337         PR tree-optimization/84512
20338         * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
20339         Do not use the estimate returned from record_stmt_cost for
20340         the scalar iteration cost but sum properly using add_stmt_cost.
20342 2018-02-27  Richard Biener  <rguenther@suse.de>
20344         PR tree-optimization/84466
20345         * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
20346         Adjust last change to less strictly validate use operands.
20348 2018-02-27  Martin Liska  <mliska@suse.cz>
20350         PR gcov-profile/84548
20351         * gcov.c (process_file): Allow partial overlap and consider it
20352         also as group functions.
20353         (output_lines): Properly calculate range of lines for a group.
20355 2018-02-27  Martin Liska  <mliska@suse.cz>
20357         * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
20358         'ggc' suffixes.  Change first column width.
20359         (timer::print): Fix formatting of the column.
20361 2018-02-27  Alexandre Oliva  <aoliva@redhat.com>
20363         * tree-ssa-live.c (remove_unused_scope_block_p): Do not
20364         preserve inline entry blocks for the sake of debug inline
20365         entry point markers alone.
20366         (remove_unused_locals): Suggest in comments a better place to
20367         force the preservation of inline entry blocks that are
20368         otherwise unused, but do not preserve them.
20370 2018-02-26  H.J. Lu  <hongjiu.lu@intel.com>
20372         * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
20374 2018-02-26  H.J. Lu  <hongjiu.lu@intel.com>
20376         PR target/84039
20377         * config/i386/constraints.md (Bs): Replace
20378         ix86_indirect_branch_register with
20379         TARGET_INDIRECT_BRANCH_REGISTER.
20380         (Bw): Likewise.
20381         * config/i386/i386.md (indirect_jump): Likewise.
20382         (tablejump): Likewise.
20383         (*sibcall_memory): Likewise.
20384         (*sibcall_value_memory): Likewise.
20385         Peepholes of indirect call and jump via memory: Likewise.
20386         (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
20387         (*sibcall_value_GOT_32): Likewise.
20388         * config/i386/predicates.md (indirect_branch_operand): Likewise.
20389         (GOT_memory_operand): Likewise.
20390         (call_insn_operand): Likewise.
20391         (sibcall_insn_operand): Likewise.
20392         (GOT32_symbol_operand): Likewise.
20393         * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
20395 2018-02-26  Eric Botcazou  <ebotcazou@adacore.com>
20397         PR rtl-optimization/83496
20398         * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
20399         booleans to RTXes.  Call fix_reg_dead_note on every non-null element.
20400         (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
20401         redundant insn, if any.
20402         (relax_delay_slots): Likewise.
20403         (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
20405 2018-02-26  Richard Sandiford  <richard.sandiford@linaro.org>
20407         PR tree-optimization/83965
20408         * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
20409         that grouped statements are part of a reduction chain.  Return
20410         true if the statement is not marked as a reduction itself but
20411         is part of a group.
20412         (vect_recog_dot_prod_pattern): Don't check whether the statement
20413         is part of a group here.
20414         (vect_recog_sad_pattern): Likewise.
20415         (vect_recog_widen_sum_pattern): Likewise.
20417 2018-02-26  Eric Botcazou  <ebotcazou@adacore.com>
20419         PR debug/84545
20420         * final.c (rest_of_clean_state): Also look for calls inside sequences.
20422 2018-02-26  H.J. Lu  <hongjiu.lu@intel.com>
20424         PR target/84530
20425         * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
20426         the bool argument.
20427         (ix86_output_indirect_function_return): New prototype.
20428         (ix86_split_simple_return_pop_internal): Likewise.
20429         * config/i386/i386.c (indirect_return_via_cx): New.
20430         (indirect_return_via_cx_bnd): Likewise.
20431         (indirect_thunk_name): Handle return va CX_REG.
20432         (output_indirect_thunk_function): Create alias for
20433         __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
20434         (ix86_output_indirect_jmp): Remove the bool argument.
20435         (ix86_output_indirect_function_return): New function.
20436         (ix86_split_simple_return_pop_internal): Likewise.
20437         * config/i386/i386.md (*indirect_jump): Don't pass false
20438         to ix86_output_indirect_jmp.
20439         (*tablejump_1): Likewise.
20440         (simple_return_pop_internal): Change it to define_insn_and_split.
20441         Call ix86_split_simple_return_pop_internal to split it for
20442         -mfunction-return=.
20443         (simple_return_indirect_internal): Call
20444         ix86_output_indirect_function_return instead of
20445         ix86_output_indirect_jmp.
20447 2018-02-26  Jakub Jelinek  <jakub@redhat.com>
20449         PR bootstrap/84405
20450         * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
20451         memset and value initialization afterwards.
20453 2018-02-26  Christophe Lyon  <christophe.lyon@linaro.org>
20455         * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
20457 2018-02-26  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
20459         PR target/84521
20460         * common/config/aarch64/aarch64-common.c
20461         (aarch_option_optimization_table[]): Switch
20462         off fomit-frame-pointer
20464 2018-02-26  Kito Cheng  <kito.cheng@gmail.com>
20465             Chung-Ju Wu  <jasonwucj@gmail.com>
20467         * config/nds32/nds32-multiple.md (load_multiple): Disallow
20468         volatile memory.
20469         (store_multiple): Ditto.
20471 2018-02-26  Kito Cheng  <kito.cheng@gmail.com>
20473         * config.gcc: Add --with-cpu support for nds32 target.
20474         * config/nds32/nds32-opts.h (nds32_cpu_type): New.
20475         * config/nds32/nds32.opt: Add -mcpu= option.
20477 2018-02-25  Segher Boessenkool  <segher@kernel.crashing.org>
20479         * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
20480         isel=yes): Warn for these deprecated options.
20482 2018-02-23  David Edelsohn  <dje.gcc@gmail.com>
20484         * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
20485         ISA_2_5_MASKS_EMBEDDED.
20487 2018-02-23  Jakub Jelinek  <jakub@redhat.com>
20489         * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
20490         p->max as pointers rather than using iterative_hash_expr.
20492 2018-02-23  Carl Love  <cel@us.ibm.com>
20494         * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
20495         macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
20496         BU_P8V_OVERLOAD_2.
20497         * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
20498         P8V_BUILTIN_VEC_VSIGNED2.  Change VSX_BUILTIN_VEC_VUNSIGNED2 to
20499         P8V_BUILTIN_VEC_VUNSIGNED2.
20501 2018-02-22  Vladimir Makarov  <vmakarov@redhat.com>
20503         PR target/81572
20504         * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
20505         * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
20506         LRA_UNKNOWN_ALT.
20507         * lra-constraints.c (curr_insn_transform): Set up
20508         LRA_NON_CLOBBERED_ALT for moves processed on the fast path.  Use
20509         LRA_UNKNOWN_ALT.
20510         (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
20511         * lra-eliminations.c (spill_pseudos): Ditto.
20512         (process_insn_for_elimination): Ditto.
20513         * lra-lives.c (reg_early_clobber_p): Use the new macros.
20514         * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
20515         LRA_NON_CLOBBERED_ALT.
20517 2018-02-22  Martin Sebor  <msebor@redhat.com>
20519         PR tree-optimization/84480
20520         * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
20521         to maybe_diag_stxncpy_trunc.  Call it.
20522         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
20523         from gimple_fold_builtin_strcpy.  Print inlining stack.
20524         (handle_builtin_stxncpy): Print inlining stack.
20525         * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
20527 2018-02-22  H.J. Lu  <hongjiu.lu@intel.com>
20529         PR target/84176
20530         * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
20531         error when -mindirect-branch=thunk-extern, -fcf-protection=branch
20532         and -fcheck-pointer-bounds are used together.
20533         (indirect_thunk_prefix): New enum.
20534         (indirect_thunk_need_prefix): New function.
20535         (indirect_thunk_name): Replace need_bnd_p with need_prefix.  Use
20536         "_nt" instead of "_bnd" for NOTRACK prefix.
20537         (output_indirect_thunk): Replace need_bnd_p with need_prefix.
20538         (output_indirect_thunk_function): Likewise.
20539         (): Likewise.
20540         (ix86_code_end): Update output_indirect_thunk_function calls.
20541         (ix86_output_indirect_branch_via_reg): Replace
20542         ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
20543         (ix86_output_indirect_branch_via_push): Likewise.
20544         (ix86_output_function_return): Likewise.
20545         * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
20546         incompatible with -fcf-protection=branch and
20547         -fcheck-pointer-bounds.
20549 2018-02-22  Steve Ellcey  <sellcey@cavium.com>
20551         PR target/83335
20552         * config/aarch64/aarch64.c (aarch64_print_address_internal):
20553         Change gcc_assert call to output_operand_lossage.
20555 2018-02-22  Steve Ellcey  <sellcey@cavium.com>
20557         * doc/extend.texi (__builtin_extend_pointer): Document builtin.
20559 2018-02-22  DJ Delorie  <dj@redhat.com>
20560             Sebastian Perta  <sebastian.perta@renesas.com>
20561             Oleg Endo  <olegendo@gcc.gnu.org>
20563         * config/rx/rx.c (rx_rtx_costs): New function.
20564         (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
20566 2018-02-22  Thomas Preud'homme  <thomas.preudhomme@arm.com>
20568         * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
20570 2018-02-22  Martin Liska  <mliska@suse.cz>
20572         PR driver/83193
20573         * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
20574         Add "native" as a possible value.
20576 2018-02-22  Martin Liska  <mliska@suse.cz>
20578         PR driver/83193
20579         * config/i386/i386.c (ix86_option_override_internal):
20580         Add "native" as a possible value for -march and -mtune.
20582 2018-02-22  Jakub Jelinek  <jakub@redhat.com>
20584         PR target/84502
20585         * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
20586         to all type variants.
20588         PR tree-optimization/84503
20589         * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
20590         width as info->bitpos + info->bitsize - start.
20591         (merged_store_group::merge_overlapping): Simplify width computation.
20592         (check_no_overlap): New function.
20593         (imm_store_chain_info::try_coalesce_bswap): Compute expected
20594         start + width and last_order of the group, fail if check_no_overlap
20595         fails.
20596         (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
20597         to group if check_no_overlap fails.
20599 2018-02-21  Segher Boessenkool  <segher@kernel.crashing.org>
20601         * config/rs6000/altivec.md: Delete contraint arguments to
20602         define_expand, define_split, and define_peephole2, and in
20603         define_insn_and_split if always unused.
20604         * config/rs6000/darwin.md: Ditto.
20605         * config/rs6000/dfp.md: Ditto.
20606         * config/rs6000/rs6000.md: Ditto.
20607         * config/rs6000/sync.md: Ditto.
20608         * config/rs6000/vector.md: Ditto.
20609         * config/rs6000/vsx.md: Ditto.
20611 2018-02-21  Segher Boessenkool  <segher@kernel.crashing.org>
20613         * config/rs6000/altivec.md: Write output control strings as braced
20614         blocks instead of double-quoted strings.
20615         * config/rs6000/darwin.md: Ditto.
20616         * config/rs6000/rs6000.md: Ditto.
20617         * config/rs6000/vector.md: Ditto.
20618         * config/rs6000/vsx.md: Ditto.
20620 2018-02-21  Jason Merrill  <jason@redhat.com>
20622         PR c++/84314 - ICE with templates and fastcall attribute.
20623         * attribs.c (build_type_attribute_qual_variant): Remove assert.
20625 2018-02-21  Jan Hubicka  <hubicka@ucw.cz>
20627         * ipa-cp.c (determine_versionability): Fix comment typos.
20629 2018-02-21  Jan Hubicka  <hubicka@ucw.cz>
20631         PR c/84229
20632         * ipa-cp.c (determine_versionability): Do not version functions caling
20633         va_arg_pack.
20635 2018-02-21  Martin Liska  <mliska@suse.cz>
20637         PR driver/83193
20638         * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
20639         Add "native" as a possible value.
20640         * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT):  Define
20641         the macro when native cpu detection is available.
20643 2018-02-21  Martin Liska  <mliska@suse.cz>
20645         PR driver/83193
20646         * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
20647         Add "native" as a possible value.
20648         * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
20649         when native cpu detection is available.
20651 2018-02-21  Jakub Jelinek  <jakub@redhat.com>
20652             Martin Sebor  <msebor@redhat.com>
20654         PR tree-optimization/84478
20655         * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
20656         false.
20657         * gimple-fold.c (get_range_strlen): Make minlen const and assume it
20658         can't be NULL.  Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
20659         support which is conservatively correct, for 2 only stay conservative
20660         for maxlen.  Formatting and comment capitalization fixes.  Add STRICT
20661         argument to the 2 argument get_range_strlen, adjust 6 arg
20662         get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
20663         false.
20664         (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
20665         (gimple_fold_builtin_strlen): Pass true as last argument to
20666         get_range_strlen.
20668 2018-02-20  Martin Sebor  <msebor@redhat.com>
20670         PR middle-end/84095
20671         * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
20672         (builtin_memref::set_base_and_offset): Same.  Handle inner references.
20673         (builtin_memref::builtin_memref): Factor out parts into
20674         set_base_and_offset and call it.
20676 2018-02-20  Richard Sandiford  <richard.sandiford@linaro.org>
20678         PR middle-end/84406
20679         * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
20680         is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
20681         greater precision.  If to_mode is a MODE_PARTIAL_INT, stop the
20682         search at the associated MODE_INT.
20684 2018-02-20  Jeff Law  <law@redhat.com>
20686         PR middle-end/82123
20687         PR tree-optimization/81592
20688         PR middle-end/79257
20689         * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
20690         for range data rather than using global data.
20691         * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
20692         range data rather than using global data.
20693         * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
20694         pass it to children as needed.
20695         (struct directive::fmtresult): Similarly.
20696         (struct directive::set_width): Similarly.
20697         (struct directive::set_precision): Similarly.
20698         (format_integer, format_directive, parse_directive): Similarly.
20699         (format_none): Accept unnamed vr_values parameter.
20700         (format_percent, format_floating, format_character): Similarly.
20701         (format_string, format_plain): Similarly.
20702         * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
20703         the EVRP range analyzer for range data rather than using global data.
20704         * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
20705         gimple-ssa-evrp-analyze.h
20706         (class sprintf_dom_walker): Add after_dom_children member function.
20707         Add evrp_range_analyzer member.
20708         (sprintf_dom_walker::before_dom_children): Call into the EVRP
20709         range analyzer as needed.
20710         (sprintf_dom_walker::after_dom_children): New member function.
20711         * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
20712         if not optimizing.
20713         (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
20714         (evrp_range_analyzer::pop_to_marker): Likewise.
20716 2018-02-20  Richard Sandiford  <richard.sandiford@linaro.org>
20718         PR tree-optimization/84419
20719         * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
20720         with the required type if its current type is compatible but
20721         different.
20723 2018-02-20  Jakub Jelinek  <jakub@redhat.com>
20725         PR middle-end/82004
20726         * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
20727         after vectorization.
20729 2018-02-20  Martin Liska  <mliska@suse.cz>
20731         PR driver/83193
20732         * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
20733         possible values if we don't have a hint.
20735 2018-02-20  Martin Liska  <mliska@suse.cz>
20737         PR c/84310
20738         PR target/79747
20739         * final.c (shorten_branches): Build align_tab array with one
20740         more element.
20741         * opts.c (finish_options): Add alignment option limit check.
20742         (MAX_CODE_ALIGN): Likewise.
20743         (MAX_CODE_ALIGN_VALUE): Likewise.
20744         * doc/invoke.texi: Document maximum allowed option value for
20745         all -falign-* options.
20747 2018-02-19  Jakub Jelinek  <jakub@redhat.com>
20749         PR target/84146
20750         * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
20751         * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
20752         * var-tracking.c (emit_note_insn_var_location): Remove all references
20753         to NOTE_INSN_CALL_ARG_LOCATION.
20754         (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
20755         the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
20756         Use copy_rtx_if_shared.
20757         * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
20758         NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
20759         (dwarf2out_var_location): Remove handling of
20760         NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
20761         on call_insn.
20762         * final.c (final_scan_insn): Remove all references to
20763         NOTE_INSN_CALL_ARG_LOCATION.
20764         (rest_of_clean_state): Likewise.  Remove REG_CALL_ARG_LOCATION notes
20765         before dumping final insns.
20766         * except.c (emit_note_eh_region_end): Remove all references to
20767         NOTE_INSN_CALL_ARG_LOCATION.
20768         * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
20769         * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
20770         * config/arc/arc.c (hwloop_optimize): Likewise.
20771         * config/arm/arm.c (create_fix_barrier): Likewise.
20772         * config/s390/s390.c (s390_chunkify_start): Likewise.
20773         * config/sh/sh.c (find_barrier): Likewise.
20774         * config/i386/i386.c (rest_of_insert_endbranch,
20775         ix86_seh_fixup_eh_fallthru): Likewise.
20776         * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
20777         * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
20778         * config/frv/frv.c (frv_function_prologue): Likewise.
20779         * emit-rtl.c (try_split): Likewise.  Copy over REG_CALL_ARG_LOCATION
20780         reg note.
20781         (note_outside_basic_block_p): Remove all references to
20782         NOTE_INSN_CALL_ARG_LOCATION.
20783         * gengtype.c (adjust_field_rtx_def): Likewise.
20784         * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
20785         Likewise.
20786         * jump.c (cleanup_barriers, delete_related_insns): Likewise.
20787         * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
20789         PR c++/84444
20790         * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
20791         is ADDR_EXPR.
20793         PR tree-optimization/84452
20794         * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
20795         expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
20796         is NULL.
20798 2018-02-19  Martin Liska  <mliska@suse.cz>
20800         PR sanitizer/82183
20801         * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
20803 2018-02-19  Martin Liska  <mliska@suse.cz>
20804             Richard Sandiford  <richard.sandiford@linaro.org>
20806         PR tree-optimization/82491
20807         * gimple-fold.c (get_base_constructor): Make earlier bail out
20808         to prevent ubsan.
20810 2018-02-19  Carl Love  <cel@us.ibm.com>
20812         * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
20813         BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
20814         BU_P8V_OVERLOAD_1.
20815         * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
20816         P8V_BUILTIN_VEC_NEG.
20818 2018-02-19  Sebastian Perta  <sebastian.perta@renesas.com>
20820         * config/rl78/rl78.md (movdf): New define expand.
20822 2018-02-19  Martin Liska  <mliska@suse.cz>
20824         PR other/80589
20825         * doc/invoke.texi: Fix typo.
20826         * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
20828 2018-02-18  Segher Boessenkool  <segher@kernel.crashing.org>
20830         * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
20831         handle rs6000_single_float and rs6000_double_float specially for
20832         e500 family CPUs.
20834 2018-02-16  Jeff Law  <law@redhat.com>
20836         * config/rx/rx.c (add_pop_cfi_notes): New function.;
20837         (pop_regs): Use it.
20839 2018-02-16  Jakub Jelinek  <jakub@redhat.com>
20841         PR ipa/84425
20842         * ipa-inline.c (inline_small_functions): Fix a typo.
20844 2018-02-16  Nathan Sidwell  <nathan@acm.org>
20846         * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
20848 2018-02-16  Carl Love  <cel@us.ibm.com>
20850         * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
20851         Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
20852         from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
20853         * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
20854         expansion to P8V_BUILTIN_VEC_FLOAT2.
20856 2018-02-16  Vladimir Makarov  <vmakarov@redhat.com>
20858         PR rtl-optimization/70023
20859         * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
20860         src_regno into account.
20862 2018-02-16  Carl Love  <cel@us.ibm.com>
20864         * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
20865         * config/rs6000/rs6000-builtin.def: Remove macro expansion for
20866         VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
20867         * config/rs6000/rs6000.c: Remove case statements for
20868         P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
20869         P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
20870         and P9V_BUILTIN_VEC_VINSERT4B.
20871         * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
20872         P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
20873         * config/rs6000/vsx.md:
20874         * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
20875         vec_insert4b.
20877 2018-02-16  Carl Love  <cel@us.ibm.com>
20879         * config/rs6000/altivec.h: Add builtin names vec_extract4b
20880         vec_insert4b.
20881         * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
20882         definitions.
20883         * config/rs6000/rs6000-c.c: Add the definitions for
20884         P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
20885         * config/rs6000/rs6000.c (altivec_expand_builtin): Add
20886         P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
20887         * config/rs6000/vsx.md: Add define_insn extract4b.  Add define_expand
20888         definition for insert4b and define insn *insert3b_internal.
20889         * doc/extend.texi: Add documentation for vec_extract4b.
20891 2018-02-16  Nathan Sidwell  <nathan@acm.org>
20893         * doc/extend.texi (Backwards Compatibility): Mention friend
20894         injection.  Note for-scope is deprecated.
20895         * doc/invoke.texi (-ffriend-injection): Deprecate.
20897 2018-02-16  Segher Boessenkool  <segher@kernel.crashing.org>
20899         * combine.c (try_combine): When adjusting LOG_LINKS for the destination
20900         that moved to I2, also allow destinations that are a paradoxical
20901         subreg (instead of a normal reg).
20903 2018-02-16  Oleg Endo  <olegendo@gcc.gnu.org>
20905         PR target/83831
20906         * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
20907         to QImode.
20909 2018-02-16  Richard Biener  <rguenther@suse.de>
20911         PR tree-optimization/84037
20912         PR tree-optimization/84016
20913         PR target/82862
20914         * config/i386/i386.c (ix86_builtin_vectorization_cost):
20915         Adjust vec_construct for the fact we need additional higher latency
20916         128bit inserts for AVX256 and AVX512 vector builds.
20917         (ix86_add_stmt_cost): Scale vector construction cost for
20918         elementwise loads.
20920 2018-02-16  Richard Biener  <rguenther@suse.de>
20922         PR tree-optimization/84417
20923         * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
20924         the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
20925         (non_rewritable_lvalue_p): Likewise, use poly-ints.
20927 2018-02-16  Martin Liska  <mliska@suse.cz>
20929         PR sanitizer/84307
20930         * internal-fn.def (ASAN_CHECK): Set proper flags.
20931         (ASAN_MARK): Likewise.
20933 2018-02-16  Julia Koval  <julia.koval@intel.com>
20935         * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
20936         from PTA_CANNONLAKE.
20938 2018-02-16  Jakub Jelinek  <jakub@redhat.com>
20940         PR target/84272
20941         * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
20942         Use ++iter rather than iter++ for std::list iterators.
20943         (func_fma_steering::dfs): Likewise.  Don't delete nodes right away,
20944         defer deleting them until all nodes in the forest are processed.  Do
20945         free even leaf nodes.  Change to_process into auto_vec.
20947         PR bootstrap/84405
20948         * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
20949         * vec.h (vec_default_construct): Use memset instead of placement new
20950         if BROKEN_VALUE_INITIALIZATION is defined.
20951         * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
20952         memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
20953         is defined.
20955         PR rtl-optimization/83723
20956         * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
20957         * lra.c (lra_substitute_pseudo): Likewise.  If true, use
20958         gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG.  Pass DEBUG_P to
20959         recursive calls.
20960         (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
20961         callers.
20962         * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
20964 2018-02-16  Eric Botcazou  <ebotcazou@adacore.com>
20966         PR rtl-optimization/81443
20967         * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
20968         from inner REGs to paradoxical SUBREGs.
20970 2018-02-16  Richard Biener  <rguenther@suse.de>
20972         PR tree-optimization/84399
20973         * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
20974         For operands we can analyze at their definition make sure we can
20975         analyze them at each use as well.
20977 2018-02-16  Richard Biener  <rguenther@suse.de>
20979         PR tree-optimization/84190
20980         * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
20981         volatile accesses if the decl isn't volatile.
20983 2018-02-15  Jason Merrill  <jason@redhat.com>
20985         PR c++/84314 - ICE with templates and fastcall attribute.
20986         * attribs.c (build_type_attribute_qual_variant): Don't clobber
20987         TYPE_CANONICAL on an existing type.
20989 2018-02-15  Jakub Jelinek  <jakub@redhat.com>
20991         PR tree-optimization/84383
20992         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
20993         dstoff nor call operand_equal_p if dstbase is NULL.
20995         PR tree-optimization/84334
20996         * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
20997         also a CONSTANT_CLASS_P, punt.
20999 2018-02-14  Jim Wilson  <jimw@sifive.com>
21001         * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
21002         first SMALL_OPERAND check.  New local min_second_step.  Move assert
21003         to where locals are set.  Add TARGET_RVC support.
21004         * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
21006 2018-02-14  Indu Bhagat  <indu.bhagat@oracle.com>
21008         * doc/invoke.texi: Correct -Wformat-overflow code sample.
21010 2018-02-14  Martin Sebor  <msebor@redhat.com>
21012         PR tree-optimization/83698
21013         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
21014         arrays constrain the offset range to their bounds.
21015         (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
21016         (builtin_access::overlap): Avoid setting the size of overlap if it's
21017         already been set.
21018         (maybe_diag_overlap): Also consider arrays when deciding what values
21019         of offsets to include in diagnostics.
21021 2018-02-14  Martin Sebor  <msebor@redhat.com>
21023         PR c/84108
21024         * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
21025         that correspond to the kind of a declaration.
21027 2018-02-14  John David Anglin  <danglin@gcc.gnu.org>
21029         PR target/83984
21030         * config/pa/pa.md: Load address of PIC label using the linkage table
21031         if the label is nonlocal.
21033 2018-02-14  Kelvin Nilsen  <kelvin@gcc.gnu.org>
21035         * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
21036         warning message if user requests -maltivec=be.
21037         * doc/invoke.texi: Document deprecation of -maltivec=be.
21039 2018-02-14  Will Schmidt  <will_schmidt@vnet.ibm.com>
21041         PR target/84220
21042         * config/rs6000/rs6000-c.c: Update definitions for
21043         ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
21044         VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
21046 2018-02-14  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
21048         PR target/84239
21049         * config/i386/cetintrin.h: Remove _rdssp[d|q] and
21050         add _get_ssp intrinsics. Remove argument from
21051         __builtin_ia32_rdssp[d|q].
21052         * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
21053         * config/i386/i386-builtin.def: Remove argument from
21054         __builtin_ia32_rdssp[d|q].
21055         * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
21056         ix86_expand_special_args_builtin for _rdssp[d|q].
21057         * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
21058         Clear register before usage.
21059         * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
21060         Add documentation for new _get_ssp and _inc_ssp intrinsics.
21062 2018-02-14  Richard Sandiford  <richard.sandiford@linaro.org>
21064         PR tree-optimization/84357
21065         * tree-data-ref.c (object_address_invariant_in_loop_p): Check
21066         operand 1 of an ARRAY_REF too.
21068 2018-02-14  Oleg Endo  <olegendo@gcc.gnu.org>
21070         PR target/83831
21071         * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
21072         rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
21073         declarations.
21074         (set_of_reg): New struct.
21075         (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
21076         * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
21077         rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
21078         functions.
21079         * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
21080         Split into bitclr, bitset, bitinvert patterns if appropriate.
21081         (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
21082         use rx_fuse_in_memory_bitop.
21083         (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
21084         to named insn, correct maximum insn length.
21086 2018-02-14  Jozef Lawrynowicz  <jozefl.gcc@gmail.com>
21088         PR target/79242
21089         * machmode.def: Define a complex mode for PARTIAL_INT.
21090         * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
21091         MODE_PARTIAL_INT.
21092         * doc/rtl.texi: Document CSPImode.
21093         * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
21094         handling.
21095         (msp430_hard_regno_nregs_with_padding): Likewise.
21097 2018-02-13  Peter Bergner  <bergner@vnet.ibm.com>
21099         PR target/84279
21100         * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
21102 2018-02-13  Segher Boessenkool  <segher@kernel.crashing.org>
21104         PR rtl-optimization/84169
21105         * combine.c (try_combine): New variable split_i2i3.  Set it to true if
21106         we generated a parallel as new i3 and we split that to new i2 and i3
21107         instructions.  Handle split_i2i3 similar to swap_i2i3: scan the
21108         LOG_LINKs of i3 to see which of those need to link to i2 now.  Link
21109         those to i2, not i1.  Partially rewrite this scan code.
21111 2018-02-13  Jakub Jelinek  <jakub@redhat.com>
21113         PR c/82210
21114         * stor-layout.c (place_field): For variable length fields, adjust
21115         offset_align afterwards not just based on the field's alignment,
21116         but also on the size.
21118         PR middle-end/84309
21119         * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
21120         of exps and logs in the use_exp2 case.
21122 2018-02-13  Jeff Law  <law@redhat.com>
21124         * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
21125         entry for "vector".
21127         * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
21128         ARGS as unused.
21130 2018-02-13  Alexandre Oliva  <aoliva@redhat.com>
21132         PR debug/84342
21133         PR debug/84319
21134         * common.opt (gas-loc-support, gas-locview-support): New.
21135         (ginline-points, ginternal-reset-location-views): New.
21136         * doc/invoke.texi: Document them.  Use @itemx where intended.
21137         (gvariable-location-views): Adjust.
21138         * target.def (reset_location_view): New.
21139         * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
21140         (TARGET_RESET_LOCATION_VIEW): New.
21141         * doc/tm.texi: Rebuilt.
21142         * dwarf2out.c (dwarf2out_default_as_loc_support): New.
21143         (dwarf2out_default_as_locview_support): New.
21144         (output_asm_line_debug_info): Use option variables.
21145         (dwarf2out_maybe_output_loclist_view_pair): Likewise.
21146         (output_loc_list): Likewise.
21147         (add_high_low_attributes): Check option variables.
21148         Don't output entry view attribute in strict mode.
21149         (gen_inlined_subroutine_die): Check option variables.
21150         (dwarf2out_inline_entry): Likewise.
21151         (init_sections_and_labels): Likewise.
21152         (dwarf2out_early_finish): Likewise.
21153         (maybe_reset_location_view): New, from...
21154         (dwarf2out_var_location): ... here.  Call it.
21155         * debug.h (dwarf2out_default_as_loc_support): Declare.
21156         (dwarf2out_default_as_locview_support): Declare.
21157         * hooks.c (hook_int_rtx_insn_0): New.
21158         * hooks.h (hook_int_rtx_insn_0): Declare.
21159         * toplev.c (process_options): Take -gas-loc-support and
21160         -gas-locview-support from dwarf2out.  Enable
21161         -gvariable-location-views by default only with locview
21162         assembler support.  Enable -ginternal-reset-location-views by
21163         default only if the target defines the corresponding hook.
21164         Enable -ginline-points by default if location views are
21165         enabled; force it disabled if statement frontiers are
21166         disabled.
21167         * tree-inline.c (expand_call_inline): Check option variables.
21168         * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
21170 2018-02-13  Richard Sandiford  <richard.sandiford@linaro.org>
21172         PR tree-optimization/84321
21173         * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
21174         handling.  Also check whether the anti-range contains any values
21175         that satisfy the mask; switch to a VR_RANGE if not.
21177 2018-02-13  Paolo Bonzini  <bonzini@gnu.org>
21179         PR sanitizer/84340
21180         * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
21182 2018-02-13  Martin Jambor  <mjambor@suse.cz>
21184         PR c++/83990
21185         * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
21186         of call statements, also set location of a load to a temporary.
21188 2018-02-13  Sebastian Perta  <sebastian.perta@renesas.com>
21190         * config/rl78/rl78.c (add_vector_labels): New function.
21191         * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
21192         * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
21193         * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
21194         which checks that no arguments are passed.
21195         * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
21196         * doc/extend.texi: Documentation for the new attribute.
21198 2018-02-13  Andreas Schwab  <schwab@suse.de>
21200         * config/riscv/linux.h (CPP_SPEC): Define.
21202 2018-02-13  Jakub Jelinek  <jakub@redhat.com>
21204         PR target/84335
21205         * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
21206         OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
21207         OPTION_MASK_ISA_AES as first argument to def_builtin_const
21208         for AES builtins.  Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
21209         instead of OPTION_MASK_ISA_PCLMUL as first argument to
21210         def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
21211         * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
21212         temporarily for AES and PCLMUL builtins.
21214         PR tree-optimization/84339
21215         * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
21216         ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
21217         Formatting fixes.
21219         PR middle-end/84309
21220         * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
21221         exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
21222         * generic-match-head.c (canonicalize_math_after_vectorization_p): New
21223         inline function.
21224         * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
21225         inline function.
21226         * omp-simd-clone.h: New file.
21227         * omp-simd-clone.c: Include omp-simd-clone.h.
21228         (expand_simd_clones): No longer static.
21229         * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
21230         cgraph.h and omp-simd-clone.h.
21231         (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
21232         (vect_recog_widen_shift_pattern): Formatting fix.
21233         (vect_pattern_recog_1): Don't check optab for calls.
21235         PR target/84336
21236         * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
21237         operands[2] into a REG before using gen_lowpart on it.
21239 2018-02-12  Jeff Law  <law@redhat.com>
21241         PR target/83760
21242         * config/sh/sh.c (find_barrier): Consider a sibling call
21243         a barrier as well.
21245         * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
21246         successfully back substituting a reg.
21248 2018-02-12  Richard Biener  <rguenther@suse.de>
21250         PR tree-optimization/84037
21251         * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
21252         parameter, move visited init to caller.
21253         (vect_slp_analyze_operations): Separate cost from validity
21254         check, initialize visited once for all instances.
21255         (vect_schedule_slp): Analyze map to CSE vectorized nodes once
21256         for all instances.
21257         * tree-vect-stmts.c (vect_model_simple_cost): Make early
21258         out an assert.
21259         (vect_model_promotion_demotion_cost): Likewise.
21260         (vectorizable_bswap): Guard cost modeling with !slp_node
21261         instead of !PURE_SLP_STMT to avoid double-counting on hybrid
21262         SLP stmts.
21263         (vectorizable_call): Likewise.
21264         (vectorizable_conversion): Likewise.
21265         (vectorizable_assignment): Likewise.
21266         (vectorizable_shift): Likewise.
21267         (vectorizable_operation): Likewise.
21268         (vectorizable_store): Likewise.
21269         (vectorizable_load): Likewise.
21270         (vectorizable_condition): Likewise.
21271         (vectorizable_comparison): Likewise.
21273 2018-02-12  Paolo Bonzini  <bonzini@gnu.org>
21275         PR sanitizer/84307
21276         * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
21277         (ASAN_MARK): Fix fnspec to account for return value, change pointer
21278         argument from 'R' to 'W' so that the pointed-to datum is clobbered.
21280 2018-02-08  Jan Hubicka  <hubicka@ucw.cz>
21282         PR middle-end/83665
21283         * params.def (inline-min-speedup): Increase from 8 to 15.
21284         (max-inline-insns-auto): Decrease from 40 to 30.
21285         * ipa-split.c (consider_split): Add some buffer for function to
21286         be considered inlining candidate.
21287         * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
21288         default values.
21290 2018-02-12  Richard Biener  <rguenther@suse.de>
21292         PR tree-optimization/84037
21293         * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
21294         matched stmts if we cannot swap the non-matched ones.
21296 2018-02-12  Olga Makhotina  <olga.makhotina@intel.com>
21298         * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
21299         _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
21300         _mm_maskz_scalef_round_ss): New intrinsics.
21301         (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
21302         * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
21303         __builtin_ia32_scalefss_round): Remove.
21304         (__builtin_ia32_scalefsd_mask_round,
21305         __builtin_ia32_scalefss_mask_round): New intrinsics.
21306         * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
21307         (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
21308         ((match_operand:VF_128 2 "<round_nimm_predicate>"
21309         "<round_constraint>")): Changed to ...
21310         ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
21311         "<round_scalar_constraint>")): ... this.
21312         ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
21313         %0, %1, %2<round_op3>}"): Changed to ...
21314         ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
21315         %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
21316         %2<round_scalar_mask_op3>}"): ... this.
21317         * config/i386/subst.md (round_scalar_nimm_predicate): New.
21319 2018-02-12  Olga Makhotina  <olga.makhotina@intel.com>
21321         * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
21322         (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
21323         (_mm_maskz_sqrt_round_ss): New intrinsics.
21324         (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
21325         (__builtin_ia32_sqrtsd_mask_round)
21326         (__builtin_ia32_sqrtss_mask_round): New builtins.
21327         * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
21328         (__builtin_ia32_sqrtss_round): Remove.
21329         (__builtin_ia32_sqrtsd_mask_round)
21330         (__builtin_ia32_sqrtss_mask_round): New builtins.
21331         * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
21332         (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
21333         ((match_operand:VF_128 1 "vector_operand"
21334         "xBm,<round_constraint>")): Changed to ...
21335         ((match_operand:VF_128 1 "vector_operand"
21336         "xBm,<round_scalar_constraint>")): ... this.
21337         (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
21338         %0, %2, %<iptr>1<round_op3>}): Changed to ...
21339         (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
21340         %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
21341         %<iptr>1<round_scalar_mask_op3>}): ... this.
21342         ((set_attr "prefix" "<round_prefix>")): Changed to ...
21343         ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
21345 2018-02-11  Steven Munroe  <munroesj@gcc.gnu.org>
21347         PR target/84266
21348         * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
21349         Cast vec_cmpeq result to correct type.
21350         * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
21351         Cast vec_cmpgt result to correct type.
21353 2018-02-11  Alexandre Oliva  <aoliva@redhat.com>
21355         * final.c (final_scan_insn_1): Renamed from...
21356         (final_scan_insn): ... this.  New wrapper, to recover
21357         seen from the outermost call in recursive ones.
21358         * config/sparc/sparc.c (output_return): Drop seen from call.
21359         (output_sibcall): Likewise.
21360         * config/visium/visium.c (output_branch): Likewise.
21362 2018-02-10  John David Anglin  <danglin@gcc.gnu.org>
21364         * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
21365         function label.
21367 2018-02-10  Alan Modra  <amodra@gmail.com>
21369         PR target/84300
21370         * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
21371         Specify LR as an input.
21373 2018-02-10  Jakub Jelinek  <jakub@redhat.com>
21375         PR sanitizer/83987
21376         * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
21377         remove_member_access_dummy_vars): New functions.
21378         (lower_omp_for, lower_omp_taskreg, lower_omp_target,
21379         lower_omp_1, execute_lower_omp): Use them.
21381         PR rtl-optimization/84308
21382         * shrink-wrap.c (spread_components): Release todo vector.
21384 2018-02-09  Vladimir Makarov  <vmakarov@redhat.com>
21386         PR rtl-optimization/57193
21387         * ira-color.c (struct allocno_color_data): Add member
21388         conflict_allocno_hard_prefs.
21389         (update_conflict_allocno_hard_prefs): New.
21390         (bucket_allocno_compare_func): Add a preference based on
21391         conflict_allocno_hard_prefs.
21392         (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
21393         (color_allocnos): Remove a dead code.  Initiate
21394         conflict_allocno_hard_prefs.  Call update_costs_from_prefs.
21396 2018-02-09  Jakub Jelinek  <jakub@redhat.com>
21398         PR target/84226
21399         * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
21400         constraint from =wa to wa.  Avoid a subreg on the output operand,
21401         instead use a pseudo and subreg it in a move.
21402         (p9_xxbrd_<mode>): Changed to ...
21403         (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
21404         (p9_xxbrd_v2df): New expander.
21405         (p9_xxbrw_<mode>): Changed to ...
21406         (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
21407         (p9_xxbrw_v4sf): New expander.
21409 2018-02-09  Sebastian Perta  <sebastian.perta@renesas.com>
21411         * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
21413 2018-02-09  Peter Bergner  <bergner@vnet.ibm.com>
21415         PR target/83926
21416         * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
21417         multiply in 32-bit mode.
21418         (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
21419         (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
21420         mode.
21422 2018-02-09  Sebastian Perta  <sebastian.perta@renesas.com>
21424         * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
21425         to allow or block "symbol_ref" depending on the value of TARGET_JSR.
21426         * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
21427         * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
21429 2018-02-09  Pierre-Marie de Rodat  <derodat@adacore.com>
21431         PR lto/84213
21432         * dwarf2out.c (is_trivial_indirect_ref): New function.
21433         (dwarf2out_late_global_decl): Do not generate a location
21434         attribute for variables that have a non-trivial DECL_VALUE_EXPR
21435         and that are not defined in the current unit.
21437 2018-02-09  Eric Botcazou  <ebotcazou@adacore.com>
21439         * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
21440         instead of a libcall for UNORDERED.
21442 2018-02-09  Tamar Christina  <tamar.christina@arm.com>
21444         PR target/82641
21445         * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
21446         __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
21448 2018-02-09  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
21450         PR target/PR84295
21451         * config/s390/s390.c (s390_set_current_function): Invoke
21452         s390_indirect_branch_settings also if fndecl didn't change.
21454 2018-02-09  Alexandre Oliva  <aoliva@redhat.com>
21456         * config/rs6000/rs6000.md (blockage): Set length to zero.
21458 2018-02-09  Eric Botcazou  <ebotcazou@adacore.com>
21460         * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
21462 2018-02-09  Jakub Jelinek  <jakub@redhat.com>
21464         PR sanitizer/84285
21465         * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
21466         STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
21467         -static-lib*san.
21469         PR debug/84252
21470         * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
21471         PARALLEL incoming that failed vt_get_decl_and_offset check.
21473         PR middle-end/84237
21474         * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
21475         * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
21476         TREE_READONLY bit.
21477         (get_variable_section): For decls in named .bss* sections pass true as
21478         second argument to bss_initializer_p.
21480 2018-02-09  Marek Polacek  <polacek@redhat.com>
21481             Jakub Jelinek  <jakub@redhat.com>
21483         PR c++/83659
21484         * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
21485         Formatting fixes.  Verify first that tree_fits_poly_int64_p (op01).
21486         Sync some changes from cxx_fold_indirect_ref.
21488 2018-02-09  Alexandre Oliva  <aoliva@redhat.com>
21490         * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
21491         markers.
21492         * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
21493         (BLOCK_INLINE_ENTRY_LABEL): New.
21494         (dwarf2out_var_location): Disregard inline entry markers.
21495         (inline_entry_data): New struct.
21496         (inline_entry_data_hasher): New hashtable type.
21497         (inline_entry_data_hasher::hash): New.
21498         (inline_entry_data_hasher::equal): New.
21499         (inline_entry_data_table): New variable.
21500         (add_high_low_attributes): Add DW_AT_entry_pc and
21501         DW_AT_GNU_entry_view attributes if a pending entry is found
21502         in inline_entry_data_table.  Add old entry_pc attribute only
21503         if debug nonbinding markers are disabled.
21504         (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
21505         markers are enabled.
21506         (block_within_block_p, dwarf2out_inline_entry): New.
21507         (dwarf2out_finish): Check that no entries remained in
21508         inline_entry_data_table.
21509         * final.c (reemit_insn_block_notes): Handle inline entry notes.
21510         (final_scan_insn, notice_source_line): Likewise.
21511         (rest_of_clean_state): Skip inline entry markers.
21512         * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
21513         markers.
21514         * gimple.c (gimple_build_debug_inline_entry): New.
21515         * gimple.h (enum gimple_debug_subcode): Add
21516         GIMPLE_DEBUG_INLINE_ENTRY.
21517         (gimple_build_debug_inline_entry): Declare.
21518         (gimple_debug_inline_entry_p): New.
21519         (gimple_debug_nonbind_marker_p): Adjust.
21520         * insn-notes.def (INLINE_ENTRY): New.
21521         * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
21522         inline entry marker notes.
21523         (print_insn): Likewise.
21524         * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
21525         (INSN_DEBUG_MARKER_KIND): Likewise.
21526         (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
21527         * tree-inline.c (expand_call_inline): Build and insert
21528         debug_inline_entry stmt.
21529         * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
21530         inline entry blocks early, if nonbind markers are enabled.
21531         (dump_scope_block): Dump fragment info.
21532         * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
21533         * doc/gimple.texi (gimple_debug_inline_entry_p): New.
21534         (gimple_build_debug_inline_entry): New.
21535         * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
21536         Enable/disable inline entry points too.
21537         * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
21538         (DEBUG_INSN): Describe inline entry markers.
21540         * common.opt (gvariable-location-views): New.
21541         (gvariable-location-views=incompat5): New.
21542         * config.in: Rebuilt.
21543         * configure: Rebuilt.
21544         * configure.ac: Test assembler for view support.
21545         * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
21546         * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
21547         * dwarf2out.c (var_loc_view): New typedef.
21548         (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
21549         (dwarf2out_locviews_in_attribute): New.
21550         (dwarf2out_locviews_in_loclist): New.
21551         (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
21552         (enum dw_line_info_opcode): Add LI_adv_address.
21553         (struct dw_line_info_table): Add view.
21554         (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
21555         (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
21556         (zero_view_p): New variable.
21557         (ZERO_VIEW_P): New macro.
21558         (output_asm_line_debug_info): New.
21559         (struct var_loc_node): Add view.
21560         (add_AT_view_list, AT_loc_list): New.
21561         (add_var_loc_to_decl): Add view param.  Test it against last.
21562         (new_loc_list): Add view params.  Record them.
21563         (AT_loc_list_ptr): Handle loc and view lists.
21564         (view_list_to_loc_list_val_node): New.
21565         (print_dw_val): Handle dw_val_class_view_list.
21566         (size_of_die): Likewise.
21567         (value_format): Likewise.
21568         (loc_list_has_views): New.
21569         (gen_llsym): Set vl_symbol too.
21570         (maybe_gen_llsym, skip_loc_list_entry): New.
21571         (dwarf2out_maybe_output_loclist_view_pair): New.
21572         (output_loc_list): Output view list or entries too.
21573         (output_view_list_offset): New.
21574         (output_die): Handle dw_val_class_view_list.
21575         (output_dwarf_version): New.
21576         (output_compilation_unit_header): Use it.
21577         (output_skeleton_debug_sections): Likewise.
21578         (output_rnglists, output_line_info): Likewise.
21579         (output_pubnames, output_aranges): Update version comments.
21580         (output_one_line_info_table): Output view numbers in asm comments.
21581         (dw_loc_list): Determine current endview, pass it to new_loc_list.
21582         Call maybe_gen_llsym.
21583         (loc_list_from_tree_1): Adjust.
21584         (add_AT_location_description): Create view list attribute if
21585         needed, check it's absent otherwise.
21586         (convert_cfa_to_fb_loc_list): Adjust.
21587         (maybe_emit_file): Call output_asm_line_debug_info for test.
21588         (dwarf2out_var_location): Reset views as needed.  Precompute
21589         add_var_loc_to_decl args.  Call get_attr_min_length only if we have the
21590         attribute.  Set view.
21591         (new_line_info_table): Reset next view.
21592         (set_cur_line_info_table): Call output_asm_line_debug_info for test.
21593         (dwarf2out_source_line): Likewise.  Output view resets and labels to
21594         the assembler, or select appropriate line info opcodes.
21595         (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
21596         (optimize_string_length): Catch it.  Adjust.
21597         (resolve_addr): Copy vl_symbol along with ll_symbol.  Handle
21598         dw_val_class_view_list, and remove it if no longer needed.
21599         (hash_loc_list): Hash view numbers.
21600         (loc_list_hasher::equal): Compare them.
21601         (optimize_location_lists): Check whether a view list symbol is
21602         needed, and whether the locview attribute is present, and
21603         whether they match.  Remove the locview attribute if no longer
21604         needed.
21605         (index_location_lists): Call skip_loc_list_entry for test.
21606         (dwarf2out_finish): Call output_asm_line_debug_info for test.
21607         Use output_dwarf_version.
21608         * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
21609         (struct dw_val_node): Add val_view_list.
21610         * final.c (SEEN_NEXT_VIEW): New.
21611         (set_next_view_needed): New.
21612         (clear_next_view_needed): New.
21613         (maybe_output_next_view): New.
21614         (final_start_function): Rename to...
21615         (final_start_function_1): ... this.  Take pointer to FIRST,
21616         add SEEN parameter.  Emit param bindings in the initial view.
21617         (final_start_function): Reintroduce SEEN-less interface.
21618         (final): Rename to...
21619         (final_1): ... this.  Take SEEN parameter.  Output final pending
21620         next view at the end.
21621         (final): Reintroduce seen-less interface.
21622         (final_scan_insn): Output pending next view before switching
21623         sections or ending a block.  Mark the next view as needed when
21624         outputting variable locations.  Notify debug backend of section
21625         changes, and of location view changes.
21626         (rest_of_handle_final): Adjust.
21627         * toplev.c (process_options): Autodetect value for debug variable
21628         location views option.  Warn on incompat5 without -gdwarf-5.
21629         * doc/invoke.texi (gvariable-location-views): New.
21630         (gvariable-location-views=incompat5): New.
21631         (gno-variable-location-views): New.
21633 2018-02-08  David Malcolm  <dmalcolm@redhat.com>
21635         PR tree-optimization/84136
21636         * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
21637         that the result of find_edge is non-NULL.
21639 2018-02-08  Sergey Shalnov  <sergey.shalnov@intel.com>
21641         PR target/83008
21642         * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
21643         storing integer register in SImode.  Fix cost of 256 and 512
21644         byte aligned SSE register store.
21646 2018-02-08  Sergey Shalnov  <sergey.shalnov@intel.com>
21648         * config/i386/i386.c (ix86_multiplication_cost): Fix
21649         multiplication cost for TARGET_AVX512DQ.
21651 2018-02-08  Marek Polacek  <polacek@redhat.com>
21653         PR tree-optimization/84238
21654         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
21655         get_range_strlen.
21657 2018-02-08  Richard Sandiford  <richard.sandiford@linaro.org>
21659         PR tree-optimization/84265
21660         * tree-vect-stmts.c (vectorizable_store): Don't treat
21661         VMAT_CONTIGUOUS accesses as grouped.
21662         (vectorizable_load): Likewise.
21664 2018-02-08  Richard Sandiford  <richard.sandiford@linaro.org>
21666         PR tree-optimization/81635
21667         * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
21668         * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
21669         (test_round_for_mask): New functions.
21670         (wide_int_cc_tests): Call test_round_for_mask.
21671         * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
21672         * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
21673         * tree-data-ref.c (split_constant_offset_1): Use it to refine the
21674         range returned by get_range_info.
21676 2018-02-08  Jan Hubicka  <hubicka@ucw.cz>
21678         PR ipa/81360
21679         * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
21680         * symtab.c: Include builtins.h
21681         (symtab_node::output_to_lto_symbol_table_p): Move here
21682         from lto-streamer-out.c:output_symbol_p.
21683         * lto-streamer-out.c (write_symbol): Turn early exit to assert.
21684         (output_symbol_p): Move all logic to symtab.c
21685         (produce_symtab): Update.
21687 2018-02-08  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
21689         * config/s390/s390-opts.h (enum indirect_branch): Define.
21690         * config/s390/s390-protos.h (s390_return_addr_from_memory)
21691         (s390_indirect_branch_via_thunk)
21692         (s390_indirect_branch_via_inline_thunk): Add function prototypes.
21693         (enum s390_indirect_branch_type): Define.
21694         * config/s390/s390.c (struct s390_frame_layout, struct
21695         machine_function): Remove.
21696         (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
21697         (indirect_branch_table_label_no, indirect_branch_table_name):
21698         Define variables.
21699         (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
21700         (enum s390_indirect_branch_option): Define.
21701         (s390_return_addr_from_memory): New function.
21702         (s390_handle_string_attribute): New function.
21703         (s390_attribute_table): Add new attribute handler.
21704         (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
21705         (s390_indirect_branch_via_thunk): New function.
21706         (s390_indirect_branch_via_inline_thunk): New function.
21707         (s390_function_ok_for_sibcall): When jumping via thunk disallow
21708         sibling call optimization for non z10 compiles.
21709         (s390_emit_call): Force indirect branch target to be a single
21710         register.  Add r1 clobber for non-z10 compiles.
21711         (s390_emit_epilogue): Emit return jump via return_use expander.
21712         (s390_reorg): Handle JUMP_INSNs as execute targets.
21713         (s390_option_override_internal): Perform validity checks for the
21714         new command line options.
21715         (s390_indirect_branch_attrvalue): New function.
21716         (s390_indirect_branch_settings): New function.
21717         (s390_set_current_function): Invoke s390_indirect_branch_settings.
21718         (s390_output_indirect_thunk_function):  New function.
21719         (s390_code_end): Implement target hook.
21720         (s390_case_values_threshold): Implement target hook.
21721         (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
21722         macros.
21723         * config/s390/s390.h (struct s390_frame_layout)
21724         (struct machine_function): Move here from s390.c.
21725         (TARGET_INDIRECT_BRANCH_NOBP_RET)
21726         (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
21727         (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
21728         (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
21729         (TARGET_INDIRECT_BRANCH_NOBP_CALL)
21730         (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
21731         (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
21732         (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
21733         (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
21734         * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
21735         (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
21736         (mnemonic attribute): Add values which aren't recognized
21737         automatically.
21738         ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
21739         pattern for branch conversion.  Fix mnemonic attribute.
21740         ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
21741         indirect branch via thunk if requested.
21742         ("indirect_jump", "<code>"): Expand patterns for branch conversion.
21743         ("*indirect_jump"): Disable for branch conversion using out of
21744         line thunks.
21745         ("indirect_jump_via_thunk<mode>_z10")
21746         ("indirect_jump_via_thunk<mode>")
21747         ("indirect_jump_via_inlinethunk<mode>_z10")
21748         ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
21749         ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
21750         ("casesi_jump_via_inlinethunk<mode>_z10")
21751         ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
21752         ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
21753         ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
21754         ("*indirect2_jump"): Disable for branch conversion.
21755         ("casesi_jump"): Turn into expander and expand patterns for branch
21756         conversion.
21757         ("return_use"): New expander.
21758         ("*return"): Emit return via thunk and rename it to ...
21759         ("*return<mode>"): ... this one.
21760         * config/s390/s390.opt: Add new options and and enum for the
21761         option values.
21763 2018-02-08  Richard Sandiford  <richard.sandiford@linaro.org>
21765         * lra-constraints.c (match_reload): Unconditionally use
21766         gen_lowpart_SUBREG, rather than selecting between that
21767         and equivalent gen_rtx_SUBREG code.
21769 2018-02-08  Richard Biener  <rguenther@suse.de>
21771         PR tree-optimization/84233
21772         * tree-ssa-phiprop.c (propagate_with_phi): Use separate
21773         changed flag instead of boguously re-using phi_inserted.
21775 2018-02-08  Martin Jambor  <mjambor@suse.cz>
21777         * hsa-gen.c (get_symbol_for_decl): Set program allocation for
21778         static local variables.
21780 2018-02-08  Richard Biener  <rguenther@suse.de>
21782         PR tree-optimization/84278
21783         * tree-vect-stmts.c (vectorizable_store): When looking for
21784         smaller vector types to perform grouped strided loads/stores
21785         make sure the mode is supported by the target.
21786         (vectorizable_load): Likewise.
21788 2018-02-08  Wilco Dijkstra  <wdijkstr@arm.com>
21790         * config/aarch64/aarch64.c (aarch64_components_for_bb):
21791         Increase LDP/STP opportunities by adding adjacent callee-saves.
21793 2018-02-08  Wilco Dijkstra  <wdijkstr@arm.com>
21795         PR rtl-optimization/84068
21796         PR rtl-optimization/83459
21797         * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
21799 2018-02-08  Aldy Hernandez  <aldyh@redhat.com>
21801         PR tree-optimization/84224
21802         * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
21803         * calls.c (gimple_alloca_call_p): Only return TRUE when we have
21804         non-zero arguments.
21806 2018-02-07  Iain Sandoe  <iain@codesourcery.com>
21808         PR target/84113
21809         * config/rs6000/altivec.md (*restore_world): Remove LR use.
21810         * config/rs6000/predicates.md (restore_world_operation): Adjust op
21811         count, remove one USE.
21813 2018-02-07  Michael Meissner  <meissner@linux.vnet.ibm.com>
21815         * doc/install.texi (Configuration): Document the
21816         --with-long-double-format={ibm,ieee} PowerPC configuration
21817         options.
21819         PR target/84154
21820         * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
21821         Convert from define_expand to be define_insn_and_split.  Rework
21822         float/double/_Float128 conversions to QI/HI/SImode to work with
21823         both ISA 2.07 (power8) or ISA 3.0 (power9).  Fix regression where
21824         conversions to QI/HImode types did a store and then a load to
21825         truncate the value.  For conversions to VSX registers, don't split
21826         the insn, instead emit the code directly.  Use the code iterator
21827         any_fix to combine signed and unsigned conversions.
21828         (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
21829         (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
21830         (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
21831         (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
21832         (fix_<mode>di2_hw): Likewise.
21833         (fixuns_<mode>di2_hw): Likewise.
21834         (fix_<mode>si2_hw): Likewise.
21835         (fixuns_<mode>si2_hw): Likewise.
21836         (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
21837         (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
21838         (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
21839         fix<uns>_trunc<SFDF:mode>si2_p8.
21840         (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer used.
21841         (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
21842         (fix<uns>_<mode>_mem): Likewise.
21843         (fctiw<u>z_<mode>_mem): Likewise.
21844         (fix<uns>_<mode>_mem): Likewise.
21845         (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
21846         the register allocator from doing a direct move to the GPRs to do
21847         a store, and instead use the ISA 3.0 store byte/half-word from
21848         vector register instruction.  For IEEE 128-bit floating point,
21849         also optimize stores of 32-bit ints.
21850         (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
21852 2018-02-07  Alan Hayward  <alan.hayward@arm.com>
21854         * genextract.c (push_pathstr_operand): New function to support
21855         [a-zA-Z].
21856         (walk_rtx): Call push_pathstr_operand.
21857         (print_path): Support [a-zA-Z].
21859 2018-02-07  Richard Biener  <rguenther@suse.de>
21861         PR tree-optimization/84037
21862         * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
21863         (cse_and_gimplify_to_preheader): Declare.
21864         (vect_get_place_in_interleaving_chain): Likewise.
21865         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
21866         ivexpr_map.
21867         (_loop_vec_info::~_loop_vec_info): Delete it.
21868         (cse_and_gimplify_to_preheader): New function.
21869         * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
21870         * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
21871         (vectorizable_load): Likewise.  For grouped stores always base
21872         the IV on the first element.
21873         * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
21874         condition before gimplifying.
21876 2018-02-07  Jakub Jelinek  <jakub@redhat.com>
21878         * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
21879         *DIV_EXPR and *MOD_EXPR.
21881 2018-02-07  H.J. Lu  <hongjiu.lu@intel.com>
21883         PR target/84248
21884         * config/i386/i386.c (ix86_option_override_internal): Mask out
21885         the CF_SET bit when checking -fcf-protection.
21887 2018-02-07  Tom de Vries  <tom@codesourcery.com>
21889         PR libgomp/84217
21890         * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
21891         enough.
21893 2018-02-07  Richard Biener  <rguenther@suse.de>
21895         PR tree-optimization/84204
21896         * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
21897         this place.
21899         PR tree-optimization/84205
21900         * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
21901         special-case isl_ast_op_zdiv_r.
21903         PR tree-optimization/84223
21904         * graphite-scop-detection.c (gather_bbs::before_dom_children):
21905         Only add conditions from within the region.
21906         (gather_bbs::after_dom_children): Adjust.
21908 2018-02-07  Georg-Johann Lay  <avr@gjlay.de>
21910         PR target/84209
21911         * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
21912         * config/avr/avr.md: Only post-reload split REG-REG moves if
21913         either register is GENERAL_REG_P.
21915 2018-02-07  Jakub Jelinek  <jakub@redhat.com>
21917         PR tree-optimization/84235
21918         * tree-ssa-scopedtables.c
21919         (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
21920         if the subtraction is performed in floating point type where NaNs are
21921         honored.  For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
21922         build 1.  Formatting fix.
21924 2018-02-06  Jakub Jelinek  <jakub@redhat.com>
21926         PR target/84146
21927         * config/i386/i386.c (rest_of_insert_endbranch): Only skip
21928         NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
21929         and skip it regardless of bb boundaries.  Use CALL_P macro,
21930         don't test INSN_P (insn) together with CALL_P or JUMP_P check
21931         unnecessarily, formatting fix.
21933 2018-02-06  Michael Collison  <michael.collison@arm.com>
21935         * config/arm/thumb2.md:
21936         (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
21937         (*thumb_mov_notscc): Ditto.
21939 2018-02-06  Michael Meissner  <meissner@linux.vnet.ibm.com>
21941         PR target/84154
21942         * config/rs6000/rs6000.md (su code attribute): Use "u" for
21943         unsigned_fix, not "s".
21945 2018-02-06  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
21947         * configure.ac (gcc_fn_eh_frame_ro): New function.
21948         (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
21949         correct .eh_frame permissions.
21950         * configure: Regenerate.
21952 2018-02-06  Andrew Jenner  <andrew@codeourcery.com>
21954         * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
21955         irrelevant options.
21957 2018-02-06  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
21959         * config/rs6000/rs6000.c (rs6000_option_override_internal):
21960         Display warning message for -mno-speculate-indirect-jumps.
21962 2018-02-06  Andrew Jenner  <andrew@codesourcery.com>
21964         * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
21965         Undocumented.
21966         * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
21968 2018-02-06  Aldy Hernandez  <aldyh@redhat.com>
21970         PR tree-optimization/84225
21971         * tree-eh.c (find_trapping_overflow): Only call
21972         operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
21974 2018-02-06  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
21976         PR target/84145
21977         * config/i386/i386.c: Reimplement the check of possible options
21978         -mibt/-mshstk conbination. Change error messages.
21979         * doc/invoke.texi: Fix a typo: remove extra '='.
21981 2018-02-06  Marek Polacek  <polacek@redhat.com>
21983         PR tree-optimization/84228
21984         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
21986 2018-02-06  Tamar Christina  <tamar.christina@arm.com>
21988         PR target/82641
21989         * config/arm/arm.c (arm_print_asm_arch_directives): Record already
21990         emitted arch directives.
21991         * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
21992         __ARM_FEATURE_COPROC before changing architectures.
21994 2018-02-06  Richard Biener  <rguenther@suse.de>
21996         * config/i386/i386.c (print_reg): Fix typo.
21997         (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
21999 2018-02-06  Eric Botcazou  <ebotcazou@adacore.com>
22001         * configure: Regenerate.
22003 2018-02-05  Martin Sebor  <msebor@redhat.com>
22005         PR tree-optimization/83369
22006         * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
22007         inlining context.
22009 2018-02-05  Martin Liska  <mliska@suse.cz>
22011         * doc/invoke.texi: Cherry-pick upstream r323995.
22013 2018-02-05  Richard Sandiford  <richard.sandiford@linaro.org>
22015         * ira.c (ira_init_register_move_cost): Adjust comment.
22017 2018-02-05  Martin Liska  <mliska@suse.cz>
22019         PR gcov-profile/84137
22020         * doc/gcov.texi: Fix typo in documentation.
22022 2018-02-05  Martin Liska  <mliska@suse.cz>
22024         PR gcov-profile/83879
22025         * doc/gcov.texi: Document necessity of --dynamic-list-data when
22026         using dlopen functionality.
22028 2018-02-05  Olga Makhotina  <olga.makhotina@intel.com>
22030         * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
22031         _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
22032         _mm_maskz_range_ss, _mm_mask_range_round_ss,
22033         _mm_maskz_range_round_ss): New intrinsics.
22034         (__builtin_ia32_rangesd128_round)
22035         (__builtin_ia32_rangess128_round): Remove.
22036         (__builtin_ia32_rangesd128_mask_round,
22037         __builtin_ia32_rangess128_mask_round): New builtins.
22038         * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
22039         __builtin_ia32_rangess128_round): Remove.
22040         (__builtin_ia32_rangesd128_mask_round,
22041         __builtin_ia32_rangess128_mask_round): New builtins.
22042         * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
22043         (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
22044         ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
22045         "<round_saeonly_constraint>")): Changed to ...
22046         ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
22047         "<round_saeonly_scalar_constraint>")): ... this.
22048         ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
22049         %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
22050         ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
22051         %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
22052         %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
22054 2018-02-02  Andrew Jenner  <andrew@codesourcery.com>
22056         * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
22057         options.
22058         * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
22059         Remove all values except native, 8540 and 8548.
22061 2018-02-02  H.J. Lu  <hongjiu.lu@intel.com>
22063         * config/i386/i386.c (ix86_output_function_return): Pass
22064         INVALID_REGNUM, instead of -1, as invalid register number to
22065         indirect_thunk_name and output_indirect_thunk.
22067 2018-02-02  Julia Koval  <julia.koval@intel.com>
22069         * config.gcc: Add -march=icelake.
22070         * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
22071         * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
22072         * config/i386/i386.c (processor_costs): Add m_ICELAKE.
22073         (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
22074         PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
22075         (processor_target_table): Add icelake.
22076         (ix86_option_override_internal): Handle new PTAs.
22077         (get_builtin_code_for_version): Handle icelake.
22078         (M_INTEL_COREI7_ICELAKE): New.
22079         (fold_builtin_cpu): Handle icelake.
22080         * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
22081         * doc/invoke.texi: Add -march=icelake.
22083 2018-02-02  Julia Koval  <julia.koval@intel.com>
22085         * config/i386/i386.c (ix86_option_override_internal): Change flags type
22086         to wide_int_bitmask.
22087         * wide-int-bitmask.h: New.
22089 2018-02-02  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
22091         PR target/84066
22092         * config/i386/i386.md: Replace Pmode with word_mode in
22093         builtin_setjmp_setup and builtin_longjmp to support x32.
22095 2018-02-01  Peter Bergner  <bergner@vnet.ibm.com>
22097         PR target/56010
22098         PR target/83743
22099         * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
22100         #include "opts.h".
22101         (rs6000_supported_cpu_names): New static variable.
22102         (linux_cpu_translation_table): Likewise.
22103         (elf_platform) <cpu>: Define new static variable and use it.
22104         Translate kernel AT_PLATFORM name to canonical name if needed.
22105         Error if platform name is unknown.
22107 2018-02-01  Aldy Hernandez  <aldyh@redhat.com>
22109         PR target/84089
22110         * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
22112 2018-02-01  Jeff Law  <law@redhat.com>
22114         PR target/84128
22115         * config/i386/i386.c (release_scratch_register_on_entry): Add new
22116         OFFSET and RELEASE_VIA_POP arguments.  Use SP+OFFSET to restore
22117         the scratch if RELEASE_VIA_POP is false.
22118         (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
22119         If we have to save a temporary register, decrement SIZE appropriately.
22120         Pass new arguments to release_scratch_register_on_entry.
22121         (ix86_adjust_stack_and_probe): Likewise.
22122         (ix86_emit_probe_stack_range): Pass new arguments to
22123         release_scratch_register_on_entry.
22125 2018-02-01  Uros Bizjak  <ubizjak@gmail.com>
22127         PR rtl-optimization/84157
22128         * combine.c (change_zero_ext): Use REG_P predicate in
22129         front of HARD_REGISTER_P predicate.
22131 2018-02-01  Georg-Johann Lay  <avr@gjlay.de>
22133         * config/avr/avr.c (avr_option_override): Move disabling of
22134         -fdelete-null-pointer-checks to...
22135         * common/config/avr/avr-common.c (avr_option_optimization_table):
22136         ...here.
22138 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
22140         PR tree-optimization/81635
22141         * tree-data-ref.c (split_constant_offset_1): For types that
22142         wrap on overflow, try to use range info to prove that wrapping
22143         cannot occur.
22145 2018-02-01  Renlin Li  <renlin.li@arm.com>
22147         PR target/83370
22148         * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
22149         TAILCALL_ADDR_REGS.
22150         (aarch64_register_move_cost): Likewise.
22151         * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
22152         TAILCALL_ADDR_REGS.
22153         (REG_CLASS_NAMES): Likewise.
22154         (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
22155         TAILCALL_ADDR_REGS. Remove IP registers.
22156         * config/aarch64/aarch64.md (Ucs): Update register constraint.
22158 2018-02-01  Richard Biener  <rguenther@suse.de>
22160         * domwalk.h (dom_walker::dom_walker): Add additional constructor
22161         for specifying RPO order and allow NULL for that.
22162         * domwalk.c (dom_walker::dom_walker): Likewise.
22163         (dom_walker::walk): Handle NULL RPO order.
22164         * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
22165         in RPO order.
22166         (rewrite_update_dom_walker): Likewise.
22167         (mark_def_dom_walker): Likewise.
22169 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
22171         * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
22172         (aarch64_maybe_expand_sve_subreg_move): Declare.
22173         * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
22174         * config/aarch64/predicates.md (aarch64_any_register_operand): New
22175         predicate.
22176         * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
22177         that are semantically a reverse operation.
22178         (*aarch64_sve_mov<mode>_subreg_be): New pattern.
22179         * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
22180         (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
22181         functions.
22182         (aarch64_can_change_mode_class): For big-endian, forbid changes
22183         between two SVE modes if they have different element sizes.
22185 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
22187         * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
22188         the TImode handling for big-endian targets.
22190 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
22192         * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
22193         (*sve_ld1rq<Vesize>): ... this new pattern.  Handle all element sizes,
22194         not just bytes.
22195         * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
22196         Remove BSWAP handing for big-endian targets and use the form of
22197         LD1RQ appropariate for the mode.
22199 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
22201         * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
22202         all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
22203         duplicated element.
22205 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
22207         PR tearget/83845
22208         * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
22209         check for operands that need to go through aarch64_sve_reload_be.
22211 2018-02-01  Jakub Jelinek  <jakub@redhat.com>
22213         PR tree-optimization/81661
22214         PR tree-optimization/84117
22215         * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
22216         * tree-eh.c: Include gimplify.h.
22217         (find_trapping_overflow, replace_trapping_overflow,
22218         rewrite_to_non_trapping_overflow): New functions.
22219         * tree-vect-loop.c: Include tree-eh.h.
22220         (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
22221         * tree-data-ref.c: Include tree-eh.h.
22222         (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
22224 2018-01-31  Uros Bizjak  <ubizjak@gmail.com>
22226         PR rtl-optimization/84123
22227         * combine.c (change_zero_ext): Check if hard register satisfies
22228         can_change_dest_mode before calling gen_lowpart_SUBREG.
22230 2018-01-31  Vladimir Makarov  <vmakarov@redhat.com>
22232         PR target/82444
22233         * ira.c (ira_init_register_move_cost): Remove assert.
22235 2018-01-31  Eric Botcazou  <ebotcazou@adacore.com>
22237         PR rtl-optimization/84071
22238         * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
22239         * doc/tm.texi: Regenerate.
22241 2018-01-31  Richard Biener  <rguenther@suse.de>
22243         PR tree-optimization/84132
22244         * tree-data-ref.c (analyze_miv_subscript): Properly
22245         check whether evolution_function_is_affine_multivariate_p
22246         before calling gcd_of_steps_may_divide_p.
22248 2018-01-31  Julia Koval  <julia.koval@intel.com>
22250         PR target/83618
22251         * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
22252         * config/i386/i386.md (rdpid_rex64) New.
22253         (rdpid): Make 32bit only.
22255 2018-01-29  Aldy Hernandez  <aldyh@redhat.com>
22257         PR lto/84105
22258         * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
22259         an IDENTIFIER_NODE for FUNCTION_TYPE's.
22261 2018-01-31  Eric Botcazou  <ebotcazou@adacore.com>
22263         Revert
22264         2018-01-12  Eric Botcazou  <ebotcazou@adacore.com>
22266         * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
22268 2018-01-31  Eric Botcazou  <ebotcazou@adacore.com>
22270         PR rtl-optimization/84071
22271         * combine.c (record_dead_and_set_regs_1): Record the source unmodified
22272         for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
22274 2018-01-31  Claudiu Zissulescu  <claziss@synopsys.com>
22276         * config/arc/arc.c (arc_handle_aux_attribute): New function.
22277         (arc_attribute_table): Add 'aux' attribute.
22278         (arc_in_small_data_p): Consider aux like variables.
22279         (arc_is_aux_reg_p): New function.
22280         (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
22281         (arc_get_aux_arg): New function.
22282         (prepare_move_operands): Handle aux-register access.
22283         (arc_handle_aux_attribute): New function.
22284         * doc/extend.texi (ARC Variable attributes): Add subsection.
22286 2018-01-31  Claudiu Zissulescu  <claziss@synopsys.com>
22288         * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
22289         * config/arc/arc.c (arc_handle_uncached_attribute): New function.
22290         (arc_attribute_table): Add 'uncached' attribute.
22291         (arc_print_operand): Print '.di' flag for uncached memory
22292         accesses.
22293         (arc_in_small_data_p): Do not consider for small data the uncached
22294         types.
22295         (arc_is_uncached_mem_p): New function.
22296         * config/arc/predicates.md (compact_store_memory_operand): Check
22297         for uncached memory accesses.
22298         (nonvol_nonimm_operand): Likewise.
22299         * doc/extend.texi (ARC Type Attribute): New subsection.
22301 2018-01-31  Jakub Jelinek  <jakub@redhat.com>
22303         PR c/84100
22304         * common.opt (falign-functions=, falign-jumps=, falign-labels=,
22305         falign-loops=): Add Optimization flag.
22307 2018-01-30  Jeff Law  <law@redhat.com>
22309         PR target/84064
22310         * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
22311         INT_REGISTERS_SAVED.  Check it prior to calling
22312         get_scratch_register_on_entry.
22313         (ix86_adjust_stack_and_probe): Similarly.
22314         (ix86_emit_probe_stack_range): Similarly.
22315         (ix86_expand_prologue): Corresponding changes.
22317 2018-01-30  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
22319         PR target/40411
22320         * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
22321         -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
22323 2018-01-30  Vladimir Makarov  <vmakarov@redhat.com>
22325         PR target/84112
22326         * lra-constraints.c (curr_insn_transform): Process AND in the
22327         address.
22329 2018-01-30  Jakub Jelinek  <jakub@redhat.com>
22331         PR rtl-optimization/83986
22332         * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
22333         dependence against last_pending_memory_flush in addition to
22334         pending_jump_insns.
22336 2018-01-30  Alexandre Oliva  <aoliva@redhat.com>
22338         PR tree-optimization/81611
22339         * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
22340         copies.
22342 2018-01-30  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
22344         PR target/83758
22345         * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
22346         a reg rtx.
22348 2018-01-30  Richard Biener  <rguenther@suse.de>
22349             Jakub Jelinek  <jakub@redhat.com>
22351         PR tree-optimization/84111
22352         * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
22353         inner loops added during recursion, as they don't have up-to-date
22354         SSA form.
22356 2018-01-30  Jan Hubicka  <hubicka@ucw.cz>
22358         PR ipa/81360
22359         * ipa-inline.c (can_inline_edge_p): Break out late tests to...
22360         (can_inline_edge_by_limits_p): ... here.
22361         (can_early_inline_edge_p, check_callers,
22362         update_caller_keys, update_callee_keys, recursive_inlining,
22363         add_new_edges_to_heap, speculation_useful_p,
22364         inline_small_functions,
22365         inline_small_functions, flatten_function,
22366         inline_to_all_callers_1): Update.
22368 2018-01-30  Jan Hubicka  <hubicka@ucw.cz>
22370         * profile-count.c (profile_count::combine_with_ipa_count): Handle
22371         zeros correctly.
22373 2018-01-30  Richard Biener  <rguenther@suse.de>
22375         PR tree-optimization/83008
22376         * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
22377         invariant and constant vector uses in stmts when they need
22378         more than one stmt.
22380 2018-01-30  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
22382         PR bootstrap/84017
22383         * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
22384         * configure: Regenerate.
22386 2018-01-30  Richard Sandiford  <richard.sandiford@linaro.org>
22388         * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
22389         pattern.
22390         (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
22391         Use gen_rtx_REG rather than gen_lowpart.
22393 2018-01-30  Richard Sandiford  <richard.sandiford@linaro.org>
22395         * lra-constraints.c (match_reload): Use subreg_lowpart_offset
22396         rather than 0 when creating partial subregs.
22398 2018-01-30  Richard Sandiford  <richard.sandiford@linaro.org>
22400         * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
22401         of usage.
22403 2018-01-29  Michael Meissner  <meissner@linux.vnet.ibm.com>
22405         PR target/81550
22406         * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
22407         and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
22408         -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
22409         flags.  This restores the settings used before the 2017-07-24.
22410         Turning off pre increment/decrement/modify allows IVOPTS to
22411         optimize DF/SF loops where the index is an int.
22413 2018-01-29  Richard Biener  <rguenther@suse.de>
22414             Kelvin Nilsen  <kelvin@gcc.gnu.org>
22416         PR bootstrap/80867
22417         * tree-vect-stmts.c (vectorizable_call): Don't call
22418         targetm.vectorize_builtin_md_vectorized_function if callee is
22419         NULL.
22421 2018-01-22  Carl Love  <cel@us.ibm.com>
22423         * doc/extend.tex: Fix typo in second arg in
22424         __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
22426 2018-01-29  Richard Biener  <rguenther@suse.de>
22428         PR tree-optimization/84086
22429         * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
22430         (flush_ssaname_freelist): When SSA names were released reset
22431         the SCEV hash table.
22433 2018-01-29  Richard Biener  <rguenther@suse.de>
22435         PR tree-optimization/84057
22436         * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
22437         removed paths when removing edges.
22439 2018-01-27  H.J. Lu  <hongjiu.lu@intel.com>
22441         * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
22442         -mfunction-return=@var{choice}.
22444 2018-01-27  Bernd Edlinger  <bernd.edlinger@hotmail.de>
22446         PR diagnostic/84034
22447         * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
22448         Handle CR like TAB.
22449         (layout::print_source_line): Likewise.
22450         (test_get_line_width_without_trailing_whitespace): Add test cases.
22452 2018-01-27  Jakub Jelinek  <jakub@redhat.com>
22454         PR middle-end/84040
22455         * sched-deps.c (sched_macro_fuse_insns): Return immediately for
22456         debug insns.
22458 2018-01-26  Jim Wilson  <jimw@sifive.com>
22460         * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
22462         * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
22463         specified.
22465 2018-01-26  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
22467         * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
22468         and CMP + SUB-immediate -> SUBS.
22470 2018-01-26  Martin Sebor  <msebor@redhat.com>
22472         PR tree-optimization/83896
22473         * tree-ssa-strlen.c (get_string_len): Rename...
22474         (get_string_cst_length): ...to this.  Return HOST_WIDE_INT.
22475         Avoid assuming length is constant.
22476         (handle_char_store): Use HOST_WIDE_INT for string length.
22478 2018-01-26  Uros Bizjak  <ubizjak@gmail.com>
22480         PR target/81763
22481         * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
22482         to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
22484 2018-01-26  Richard Biener  <rguenther@suse.de>
22486         PR rtl-optimization/84003
22487         * dse.c (record_store): Only record redundant stores when
22488         the earlier store aliases at least all accesses the later one does.
22490 2018-01-26  Jakub Jelinek  <jakub@redhat.com>
22492         PR rtl-optimization/83985
22493         * dce.c (deletable_insn_p): Return false for separate shrink wrapping
22494         REG_CFA_RESTORE insns.
22495         (delete_unmarked_insns): Don't ignore separate shrink wrapping
22496         REG_CFA_RESTORE insns here.
22498         PR c/83989
22499         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
22500         use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
22502 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
22504         * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
22505         * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
22506         (arc_init): Likewise.
22507         (arc_override_options): Likewise.
22508         (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
22509         value.
22510         (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
22511         support.
22512         * config/arc/arc.h (TARGET_DBNZ): Define.
22513         * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
22514         properly set the tune attribute.
22515         (dbnz): Use TARGET_DBNZ guard.
22516         * config/arc/arc.opt (mtune): Add core3 option.
22518 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
22520         * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
22521         recognize new pic like addresses.
22522         (arc_delegitimize_address): Clean up.
22524 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
22526         * config/arc/arc-arches.def: Option mrf16 valid for all
22527         architectures.
22528         * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
22529         * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
22530         * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
22531         * config/arc/arc-tables.opt: Regenerate.
22532         * config/arc/arc.c (arc_conditional_register_usage): Handle
22533         reduced register file case.
22534         (arc_file_start): Set must have build attributes.
22535         * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
22536         mrf16 option value.
22537         * config/arc/arc.opt (mrf16): Add new option.
22538         * config/arc/elf.h (ATTRIBUTE_PCS): Define.
22539         * config/arc/genmultilib.awk: Handle new mrf16 option.
22540         * config/arc/linux.h (ATTRIBUTE_PCS): Define.
22541         * config/arc/t-multilib: Regenerate.
22542         * doc/invoke.texi (ARC Options): Document mrf16 option.
22544 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
22546         * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
22547         * config/arc/arc.c (arc_handle_secure_attribute): New function.
22548         (arc_attribute_table): Add 'secure_call' attribute.
22549         (arc_print_operand): Print secure call operand.
22550         (arc_function_ok_for_sibcall): Don't optimize tail calls when
22551         secure.
22552         (arc_is_secure_call_p): New function.  * config/arc/arc.md
22553         (call_i): Add support for sjli instruction.
22554         (call_value_i): Likewise.
22555         * config/arc/constraints.md (Csc): New constraint.
22557 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
22558             John Eric Martin  <John.Martin@emmicro-us.com>
22560         * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
22561         * config/arc/arc.c (_arc_jli_section): New struct.
22562         (arc_jli_section): New type.
22563         (rc_jli_sections): New static variable.
22564         (arc_handle_jli_attribute): New function.
22565         (arc_attribute_table): Add jli_always and jli_fixed attribute.
22566         (arc_file_end): New function.
22567         (TARGET_ASM_FILE_END): Define.
22568         (arc_print_operand): Reuse 'S' letter for JLI output instruction.
22569         (arc_add_jli_section): New function.
22570         (jli_call_scan): Likewise.
22571         (arc_reorg): Call jli_call_scan.
22572         (arc_output_addsi): Remove 'S' from printing asm operand.
22573         (arc_is_jli_call_p): New function.
22574         * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
22575         operand.
22576         (movhi_insn): Likewise.
22577         (movsi_insn): Likewise.
22578         (movsi_set_cc_insn): Likewise.
22579         (loadqi_update): Likewise.
22580         (load_zeroextendqisi_update): Likewise.
22581         (load_signextendqisi_update): Likewise.
22582         (loadhi_update): Likewise.
22583         (load_zeroextendhisi_update): Likewise.
22584         (load_signextendhisi_update): Likewise.
22585         (loadsi_update): Likewise.
22586         (loadsf_update): Likewise.
22587         (movsicc_insn): Likewise.
22588         (bset_insn): Likewise.
22589         (bxor_insn): Likewise.
22590         (bclr_insn): Likewise.
22591         (bmsk_insn): Likewise.
22592         (bicsi3_insn): Likewise.
22593         (cmpsi_cc_c_insn): Likewise.
22594         (movsi_ne): Likewise.
22595         (movsi_cond_exec): Likewise.
22596         (clrsbsi2): Likewise.
22597         (norm_f): Likewise.
22598         (normw): Likewise.
22599         (swap): Likewise.
22600         (divaw): Likewise.
22601         (flag): Likewise.
22602         (sr): Likewise.
22603         (kflag): Likewise.
22604         (ffs): Likewise.
22605         (ffs_f): Likewise.
22606         (fls): Likewise.
22607         (call_i): Remove 'S' asm letter, add jli instruction.
22608         (call_value_i): Likewise.
22609         * config/arc/arc.op (mjli-always): New option.
22610         * config/arc/constraints.md (Cji): New constraint.
22611         * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
22612         operand.
22613         (subsf3_fpx): Likewise.
22614         (mulsf3_fpx): Likewise.
22615         * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
22616         asm operand.
22617         * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
22618         function attrbutes.
22619         * doc/invoke.texi (ARC): Document mjli-always option.
22621 2018-01-26  Sebastian Perta  <sebastian.perta@renesas.com>
22623         * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
22624         avoid addition with 0 and use incw and decw where possible.
22626 2018-01-26  Richard Biener  <rguenther@suse.de>
22628         PR tree-optimization/81082
22629         * fold-const.c (fold_plusminus_mult_expr): Do not perform the
22630         association if it requires casting to unsigned.
22631         * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
22632         from fold_plusminus_mult_expr to catch important cases late when
22633         range info is available.
22635 2018-01-26  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
22637         * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
22638         * configure.ac (hidden_linkonce): New test.
22639         * configure: Regenerate.
22640         * config.in: Regenerate.
22642 2018-01-26  Julia Koval  <julia.koval@intel.com>
22644         * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
22645         _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
22646         _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
22647         _mm_mask_bitshuffle_epi64_mask): Fix type.
22648         * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
22649         USI_FTYPE_V4DI_V4DI_USI): Remove.
22650         * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
22651         __builtin_ia32_vpshufbitqmb256_mask,
22652         __builtin_ia32_vpshufbitqmb128_mask): Fix types.
22653         * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
22654         * config/i386/sse.md (VI1_AVX512VLBW): Change types.
22656 2018-01-26  Alan Modra  <amodra@gmail.com>
22658         PR target/84033
22659         * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
22660         UNSPEC_VBPERMQ.  Sort other unspecs.
22662 2018-01-25  David Edelsohn  <dje.gcc@gmail.com>
22664         * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
22666 2018-01-25  Jan Hubicka  <hubicka@ucw.cz>
22668         PR middle-end/83055
22669         * predict.c (drop_profile): Do not push/pop cfun; update also
22670         node->count.
22671         (handle_missing_profiles): Fix logic looking for zero profiles.
22673 2018-01-25  Jakub Jelinek  <jakub@redhat.com>
22675         PR middle-end/83977
22676         * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
22677         on functions with #pragma omp declare simd or functions with simd
22678         attribute.
22679         * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
22680         * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
22681         Remove trailing \n from warning_at calls.
22683 2018-01-25  Tom de Vries  <tom@codesourcery.com>
22685         PR target/84028
22686         * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
22687         for neutered workers.
22689 2018-01-24  Joseph Myers  <joseph@codesourcery.com>
22691         PR target/68467
22692         * config/m68k/m68k.c (m68k_promote_function_mode): New function.
22693         (TARGET_PROMOTE_FUNCTION_MODE): New macro.
22695 2018-01-24  Jeff Law  <law@redhat.com>
22697         PR target/83994
22698         * i386.c (get_probe_interval): Move to earlier point.
22699         (ix86_compute_frame_layout): If -fstack-clash-protection and
22700         the frame is larger than the probe interval, then use pushes
22701         to save registers rather than reg->mem moves.
22702         (ix86_expand_prologue): Remove conditional for int_registers_saved
22703         assertion.
22705 2018-01-24  Vladimir Makarov  <vmakarov@redhat.com>
22707         PR target/84014
22708         * ira-build.c (setup_min_max_allocno_live_range_point): Set up
22709         min/max for never referenced object.
22711 2018-01-24  Jakub Jelinek  <jakub@redhat.com>
22713         PR middle-end/83977
22714         * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
22715         here.
22716         * omp-low.c (create_omp_child_function): Remove "omp declare simd"
22717         attributes from DECL_ATTRIBUTES (decl) without affecting
22718         DECL_ATTRIBUTES (current_function_decl).
22719         * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
22720         functions with non-NULL DECL_ABSTRACT_ORIGIN.
22722 2018-01-24  Richard Sandiford  <richard.sandiford@linaro.org>
22724         PR tree-optimization/83979
22725         * fold-const.c (fold_comparison): Use constant_boolean_node
22726         instead of boolean_{true,false}_node.
22728 2018-01-24  Jan Hubicka  <hubicka@ucw.cz>
22730         * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
22731         with zero counts.
22733 2018-01-24  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
22735         * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
22736         Simplify the clause that sets the length attribute.
22737         (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
22738         (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
22739         clause that sets the length attribute.
22740         (*sibcall_value_nonlocal_sysv<mode>): Likewise.
22742 2018-01-24  Tom de Vries  <tom@codesourcery.com>
22744         PR target/83589
22745         * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
22746         (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
22747         Add strict parameter.
22748         (prevent_branch_around_nothing): Insert dummy insn between branch to
22749         label and label with no ptx insn inbetween.
22750         * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
22752 2018-01-24  Tom de Vries  <tom@codesourcery.com>
22754         PR target/81352
22755         * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
22756         for neutered threads in warp.
22757         * config/nvptx/nvptx.md (define_insn "exit"): New insn.
22759 2018-01-24  Richard Biener  <rguenther@suse.de>
22761         PR tree-optimization/83176
22762         * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
22763         operands.
22765 2018-01-24  Richard Biener  <rguenther@suse.de>
22767         PR tree-optimization/82819
22768         * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
22769         code generating pluses that are no-ops in the target precision.
22771 2018-01-24  Richard Biener  <rguenther@suse.de>
22773         PR middle-end/84000
22774         * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
22776 2018-01-23  Jan Hubicka  <hubicka@ucw.cz>
22778         * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
22779         to merge probabilities.
22780         * predict.c (probably_never_executed): Also mark as cold functions
22781         with global 0 profile and guessed local profile.
22782         * profile-count.c (profile_probability::combine_with_count): New
22783         member function.
22784         * profile-count.h (profile_probability::operator*,
22785         profile_probability::operator*=, profile_probability::operator/,
22786         profile_probability::operator/=): Reduce precision to adjusted
22787         and set value to guessed on contradictory divisions.
22788         (profile_probability::combine_with_freq): Remove.
22789         (profile_probability::combine_wiht_count): Declare.
22790         (profile_count::force_nonzero):: Set to adjusted.
22791         (profile_count::probability_in):: Set quality to adjusted.
22792         * tree-ssa-tail-merge.c (replace_block_by): Use
22793         combine_with_count.
22795 2018-01-23  Andrew Waterman  <andrew@sifive.com>
22796             Jim Wilson  <jimw@sifive.com>
22798         * config/riscv/riscv.c (riscv_stack_boundary): New.
22799         (riscv_option_override): Set riscv_stack_boundary.  Handle
22800         riscv_preferred_stack_boundary_arg.
22801         * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
22802         (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
22803         (STACK_BOUNDARY): Set to riscv_stack_boundary.
22804         (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
22805         * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
22806         * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
22808 2018-01-23  H.J. Lu  <hongjiu.lu@intel.com>
22810         PR target/83905
22811         * config/i386/i386.c (ix86_expand_prologue): Use cost reference
22812         of struct ix86_frame.
22813         (ix86_expand_epilogue): Likewise.  Add a local variable for
22814         the reg_save_offset field in struct ix86_frame.
22816 2018-01-23  Bin Cheng  <bin.cheng@arm.com>
22818         PR tree-optimization/82604
22819         * tree-loop-distribution.c (enum partition_kind): New enum item
22820         PKIND_PARTIAL_MEMSET.
22821         (partition_builtin_p): Support above new enum item.
22822         (generate_code_for_partition): Ditto.
22823         (compute_access_range): Differentiate cases that equality can be
22824         proven at all loops, the innermost loops or no loops.
22825         (classify_builtin_st, classify_builtin_ldst): Adjust call to above
22826         function.  Set PKIND_PARTIAL_MEMSET for partition appropriately.
22827         (finalize_partitions, distribute_loop): Don't fuse partition of
22828         PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
22829         (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
22830         parloop is enabled.
22832 2018-01-23  Martin Liska  <mliska@suse.cz>
22834         * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
22835         order to ignore the predictor.
22836         (PRED_POLYMORPHIC_CALL): Likewise.
22837         (PRED_RECURSIVE_CALL): Likewise.
22839 2018-01-23  Martin Liska  <mliska@suse.cz>
22841         * tree-profile.c (tree_profiling): Print function header to
22842         aware reader which function we are working on.
22843         * value-prof.c (gimple_find_values_to_profile): Do not print
22844         not interesting value histograms.
22846 2018-01-23  Martin Liska  <mliska@suse.cz>
22848         * profile-count.h (enum profile_quality): Add
22849         profile_uninitialized as the first value. Do not number values
22850         as they are zero based.
22851         (profile_count::verify): Update sanity check.
22852         (profile_probability::verify): Likewise.
22854 2018-01-23  Nathan Sidwell  <nathan@acm.org>
22856         * doc/invoke.texi (ffor-scope): Deprecate.
22858 2018-01-23  David Malcolm  <dmalcolm@redhat.com>
22860         PR tree-optimization/83510
22861         * domwalk.c (set_all_edges_as_executable): New function.
22862         (dom_walker::dom_walker): Convert bool param
22863         "skip_unreachable_blocks" to enum reachability.  Move setup of
22864         edge flags to set_all_edges_as_executable and only do it when
22865         reachability is REACHABLE_BLOCKS.
22866         * domwalk.h (enum dom_walker::reachability): New enum.
22867         (dom_walker::dom_walker): Convert bool param
22868         "skip_unreachable_blocks" to enum reachability.
22869         (set_all_edges_as_executable): New decl.
22870         * graphite-scop-detection.c  (gather_bbs::gather_bbs): Convert
22871         from false for "skip_unreachable_blocks" to ALL_BLOCKS for
22872         "reachability".
22873         * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
22874         but converting true to REACHABLE_BLOCKS.
22875         * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
22876         * tree-vrp.c
22877         (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
22878         Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
22879         (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
22880         REACHABLE_BLOCKS.
22881         (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
22882         if check_all_array_refs will be called.
22884 2018-01-23  David Malcolm  <dmalcolm@redhat.com>
22886         * tree.c (selftest::test_location_wrappers): Add more test
22887         coverage.
22889 2018-01-23  David Malcolm  <dmalcolm@redhat.com>
22891         * sbitmap.c (selftest::test_set_range): Fix memory leaks.
22892         (selftest::test_bit_in_range): Likewise.
22894 2018-01-23  Richard Sandiford  <richard.sandiford@linaro.org>
22896         PR testsuite/83888
22897         * doc/sourcebuild.texi (vect_float): Say that the selector
22898         only describes the situation when -funsafe-math-optimizations is on.
22899         (vect_float_strict): Document.
22901 2018-01-23  Richard Sandiford  <richard.sandiford@linaro.org>
22903         PR tree-optimization/83965
22904         * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
22905         (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
22906         instead of checking only for a reduction.
22907         (vect_recog_widen_sum_pattern): Likewise.
22909 2018-01-23  Jan Hubicka  <hubicka@ucw.cz>
22911         * predict.c (probably_never_executed): Only use precise profile info.
22912         (compute_function_frequency): Skip after inlining hack since we now
22913         have quality checking.
22915 2018-01-23  Jan Hubicka  <hubicka@ucw.cz>
22917         * profile-count.h (profile_probability::very_unlikely,
22918         profile_probability::unlikely, profile_probability::even): Set
22919         precision to guessed.
22921 2018-01-23  Richard Biener  <rguenther@suse.de>
22923         PR tree-optimization/83963
22924         * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
22925         Properly terminate dominator walk when crossing the exit edge not
22926         when visiting its source block.
22928 2018-01-23  Jakub Jelinek  <jakub@redhat.com>
22930         PR c++/83918
22931         * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
22932         VIEW_CONVERT_EXPR to wrap CONST_DECLs.
22934 2018-01-22  Jakub Jelinek  <jakub@redhat.com>
22936         PR tree-optimization/83957
22937         * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs.  Remove
22938         semicolon after for body surrounded by braces.
22940         PR tree-optimization/83081
22941         * profile-count.h (profile_probability::split): New method.
22942         * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
22943         Use profile_probability::split.
22944         (do_compare_rtx_and_jump): Fix adjustment of probabilities
22945         when splitting a single conditional jump into 2.
22947 2018-01-22  David Malcolm  <dmalcolm@redhat.com>
22949         PR tree-optimization/69452
22950         * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
22951         decl.
22953 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
22955         * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
22956         * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
22957         * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
22959 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
22961         * config/rl78/rl78-protos.h (rl78_split_movdi): New function
22962         declaration.
22963         * config/rl78/rl78.md (movdi): New define_expand.
22964         * config/rl78/rl78.c (rl78_split_movdi): New function.
22966 2018-01-22  Michael Meissner  <meissner@linux.vnet.ibm.com>
22968         PR target/83862
22969         * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
22970         no longer used.
22971         * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
22972         * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
22973         128-bit to produce an UNSPEC move to get the double word with the
22974         signbit and then a shift directly to do signbit.
22975         (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
22976         implementation with a new version that just does either a direct
22977         move or a regular move.  Move memory interface to separate insns.
22978         Move insns so they are next to the expander.
22979         (signbit<mode>2_dm_mem_be): New combiner insns to combine load
22980         with signbit move.  Split big and little endian case.
22981         (signbit<mode>2_dm_mem_le): Likewise.
22982         (signbit<mode>2_dm_<su>ext): Delete, no longer used.
22983         (signbit<mode>2_dm2): Likewise.
22985 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
22987         * config/rl78/rl78.md (anddi3): New define_expand.
22989 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
22991         * config/rl78/rl78.md (umindi3): New define_expand.
22993 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
22995         * config/rl78/rl78.md (smindi3): New define_expand.
22997 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
22999         * config/rl78/rl78.md (smaxdi3): New define_expand.
23001 2018-01-22  Carl Love  <cel@us.ibm.com>
23003         * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
23004         LVX_V1TI): Add macro expansion.
23005         * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
23006         definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
23007         VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
23008         * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
23009         Change check to determine if the instruction is a byte reversing
23010         entry.  Fix typo in comment.
23011         * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
23012         for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
23013         Add def_builtin calls for new builtins.
23014         * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
23015         Add define_insn expansion.
23017 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
23019         * config/rl78/rl78.md (umaxdi3): New define_expand.
23021 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
23023         * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
23024         for non-QImode registers.
23026 2018-01-22  Richard Biener  <rguenther@suse.de>
23028         PR tree-optimization/83963
23029         * graphite-scop-detection.c (scop_detection::get_sese): Delay
23030         including the loop exit block.
23031         (scop_detection::merge_sese): Likewise.
23032         (scop_detection::add_scop): Do it here instead.
23034 2018-01-22  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
23036         * doc/sourcebuild.texi (arm_softfloat): Document.
23038 2018-01-21  John David Anglin  <danglin@gcc.gnu.org>
23040         PR gcc/77734
23041         * config/pa/pa.c (pa_function_ok_for_sibcall): Use
23042         targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
23043         Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
23045 2018-01-21  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
23046             David Edelsohn  <dje.gcc@gmail.com>
23048         PR target/83946
23049         * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
23050         Change "crset eq" to "crset 2".
23051         (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
23052         (*call_indirect_aix<mode>_nospec): Likewise.
23053         (*call_value_indirect_aix<mode>_nospec): Likewise.
23054         (*call_indirect_elfv2<mode>_nospec): Likewise.
23055         (*call_value_indirect_elfv2<mode>_nospec): Likewise.
23056         (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
23057         change assembly output from . to $.
23058         (*sibcall_value_nonlocal_sysv<mode>): Likewise.
23059         (indirect_jump<mode>_nospec): Change assembly output from . to $.
23060         (*tablejump<mode>_internal1_nospec): Likewise.
23062 2018-01-21  Oleg Endo  <olegendo@gcc.gnu.org>
23064         PR target/80870
23065         * config/sh/sh_optimize_sett_clrt.cc:
23066         Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
23068 2018-01-20  Richard Sandiford  <richard.sandiford@linaro.org>
23070         PR tree-optimization/83940
23071         * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
23072         offset_dt to vect_constant_def rather than vect_unknown_def_type.
23073         (vect_check_load_store_mask): Add a mask_dt_out parameter and
23074         use it to pass back the definition type.
23075         (vect_check_store_rhs): Likewise rhs_dt_out.
23076         (vect_build_gather_load_calls): Add a mask_dt argument and use
23077         it instead of a call to vect_is_simple_use.
23078         (vectorizable_store): Update calls to vect_check_load_store_mask
23079         and vect_check_store_rhs.  Use the dt returned by the latter instead
23080         of scatter_src_dt.  Use the cached mask_dt and gs_info.offset_dt
23081         instead of calls to vect_is_simple_use.  Pass the scalar rather
23082         than the vector operand to vect_is_simple_use when handling
23083         second and subsequent copies of an rhs value.
23084         (vectorizable_load): Update calls to vect_check_load_store_mask
23085         and vect_build_gather_load_calls.  Use the cached mask_dt and
23086         gs_info.offset_dt instead of calls to vect_is_simple_use.
23088 2018-01-20  Jakub Jelinek  <jakub@redhat.com>
23090         PR middle-end/83945
23091         * tree-emutls.c: Include gimplify.h.
23092         (lower_emutls_2): New function.
23093         (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
23094         with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
23095         it before further processing.
23097         PR target/83930
23098         * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
23099         UINTVAL (trueop1) instead of INTVAL (op1).
23101 2018-01-19  Jakub Jelinek  <jakub@redhat.com>
23103         PR debug/81570
23104         PR debug/83728
23105         * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
23106         INCOMING_FRAME_SP_OFFSET if not defined.
23107         (scan_trace): Add ENTRY argument.  If true and
23108         DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
23109         emit a note to adjust the CFA offset.
23110         (create_cfi_notes): Adjust scan_trace callers.
23111         (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
23112         INCOMING_FRAME_SP_OFFSET in the CIE.
23113         * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
23114         * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
23115         Likewise.
23116         * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
23117         * doc/tm.texi: Regenerated.
23119 2018-01-19  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
23121         PR rtl-optimization/83147
23122         * lra-constraints.c (remove_inheritance_pseudos): Use
23123         lra_substitute_pseudo_within_insn.
23125 2018-01-19  Tom de Vries  <tom@codesourcery.com>
23126             Cesar Philippidis  <cesar@codesourcery.com>
23128         PR target/83920
23129         * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
23131 2018-01-19  Cesar Philippidis  <cesar@codesourcery.com>
23133         PR target/83790
23134         * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
23135         spaces for function labels.
23137 2018-01-19  Martin Liska  <mliska@suse.cz>
23139         * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
23140         (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
23141         (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
23142         (PRED_OPCODE_POSITIVE): Change from 64 to 59.
23143         (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
23144         (PRED_CONST_RETURN): Change from 69 to 65.
23145         (PRED_NULL_RETURN): Change from 91 to 71.
23146         (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
23147         (PRED_LOOP_GUARD): Change from 66 to 73.
23149 2018-01-19  Martin Liska  <mliska@suse.cz>
23151         * predict.c (predict_insn_def): Add new assert.
23152         (struct branch_predictor): Change type to signed integer.
23153         (test_prediction_value_range): Amend test to cover
23154         PROB_UNINITIALIZED.
23155         * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
23156         (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
23157         (PRED_LOOP_ITERATIONS_MAX): Likewise.
23158         (PRED_LOOP_IV_COMPARE): Likewise.
23159         * predict.h (PROB_UNINITIALIZED): Define new constant.
23161 2018-01-19  Martin Liska  <mliska@suse.cz>
23163         * predict.c (dump_prediction): Add new format for
23164         analyze_brprob.py script which is enabled with -details
23165         suboption.
23166         * profile-count.h (precise_p): New function.
23168 2018-01-19  Richard Sandiford  <richard.sandiford@linaro.org>
23170         PR tree-optimization/83922
23171         * tree-vect-loop.c (vect_verify_full_masking): Return false if
23172         there are no statements that need masking.
23173         (vect_active_double_reduction_p): New function.
23174         (vect_analyze_loop_operations): Use it when handling phis that
23175         are not in the loop header.
23177 2018-01-19  Richard Sandiford  <richard.sandiford@linaro.org>
23179         PR tree-optimization/83914
23180         * tree-vect-loop.c (vectorizable_induction): Don't convert
23181         init_expr or apply the peeling adjustment for inductions
23182         that are nested within the vectorized loop.
23184 2018-01-19  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
23186         * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
23187         instead of NEG.
23189 2018-01-18  Jakub Jelinek  <jakub@redhat.com>
23191         PR sanitizer/81715
23192         PR testsuite/83882
23193         * function.h (gimplify_parameters): Add gimple_seq * argument.
23194         * function.c: Include gimple.h and options.h.
23195         (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
23196         for the added local temporaries if needed.
23197         * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
23198         if there are any parameter cleanups, wrap whole body into a
23199         try/finally with the cleanups.
23201 2018-01-18  Wilco Dijkstra  <wdijkstr@arm.com>
23203         PR target/82964
23204         * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
23205         Use GET_MODE_CLASS for scalar floating point.
23207 2018-01-18  Jan Hubicka  <hubicka@ucw.cz>
23209         PR ipa/82256
23210         patch by PaX Team
23211         * cgraphclones.c (cgraph_node::create_version_clone_with_body):
23212         Fix call of call_cgraph_insertion_hooks.
23214 2018-01-18  Martin Sebor  <msebor@redhat.com>
23216         * doc/invoke.texi (-Wclass-memaccess): Tweak text.
23218 2018-01-18  Jan Hubicka  <hubicka@ucw.cz>
23220         PR ipa/83619
23221         * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
23222         frequencies.
23224 2018-01-18  Boris Kolpackov  <boris@codesynthesis.com>
23226         PR other/70268
23227         * common.opt: (-ffile-prefix-map): New option.
23228         * opts.c (common_handle_option): Defer it.
23229         * opts-global.c (handle_common_deferred_options): Handle it.
23230         * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
23231         * file-prefix-map.h: New file.
23232         (remap_debug_filename, add_debug_prefix_map): ...here.
23233         (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
23234         * final.c (debug_prefix_map, add_debug_prefix_map
23235         remap_debug_filename): Move to...
23236         * file-prefix-map.c: New file.
23237         (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
23238         generalize, get rid of alloca(), use strrchr() instead of strchr().
23239         (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
23240         Implement in terms of add_prefix_map().
23241         (remap_macro_filename, remap_debug_filename): Implement in term of
23242         remap_filename().
23243         * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
23244         * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
23245         * dbxout.c: Include file-prefix-map.h.
23246         * varasm.c: Likewise.
23247         * vmsdbgout.c: Likewise.
23248         * xcoffout.c: Likewise.
23249         * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
23250         * doc/cppopts.texi (-fmacro-prefix-map): Document.
23251         * doc/invoke.texi (-ffile-prefix-map): Document.
23252         (-fdebug-prefix-map): Update description.
23254 2018-01-18  Martin Liska  <mliska@suse.cz>
23256         * config/i386/i386.c (indirect_thunk_name): Document that also
23257         lfence is emitted.
23258         (output_indirect_thunk): Document why both instructions
23259         (pause and lfence) are generated.
23261 2018-01-18  Richard Biener  <rguenther@suse.de>
23263         PR tree-optimization/83887
23264         * graphite-scop-detection.c
23265         (scop_detection::get_nearest_dom_with_single_entry): Remove.
23266         (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
23267         (scop_detection::merge_sese): Re-implement with a flood-fill
23268         algorithm that properly finds a SESE region if it exists.
23270 2018-01-18  Jakub Jelinek  <jakub@redhat.com>
23272         PR c/61240
23273         * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
23274         pointer_diff optimizations use view_convert instead of convert.
23276 2018-01-17  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
23278         * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
23279         Generate different code for -mno-speculate-indirect-jumps.
23280         (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
23281         (*call_indirect_aix<mode>): Disable for
23282         -mno-speculate-indirect-jumps.
23283         (*call_indirect_aix<mode>_nospec): New define_insn.
23284         (*call_value_indirect_aix<mode>): Disable for
23285         -mno-speculate-indirect-jumps.
23286         (*call_value_indirect_aix<mode>_nospec): New define_insn.
23287         (*sibcall_nonlocal_sysv<mode>): Generate different code for
23288         -mno-speculate-indirect-jumps.
23289         (*sibcall_value_nonlocal_sysv<mode>): Likewise.
23291 2018-01-17  Michael Meissner  <meissner@linux.vnet.ibm.com>
23293         * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
23294         long double type, set the flags for noting the default long double
23295         type, even if we don't pass or return a long double type.
23297 2018-01-17  Jan Hubicka  <hubicka@ucw.cz>
23299         PR ipa/83051
23300         * ipa-inline.c (flatten_function): Do not overwrite final inlining
23301         failure.
23303 2018-01-17  Will Schmidt  <will_schmidt@vnet.ibm.com>
23305         * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
23306         support for merge[hl].
23307         (fold_mergehl_helper): New helper function.
23308         (tree-vector-builder.h): New #include for tree_vector_builder usage.
23309         * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
23310         (altivec_vmrglw_direct): Add xxmrglw insn.
23312 2018-01-17  Andrew Waterman  <andrew@sifive.com>
23314         * config/riscv/riscv.c (riscv_conditional_register_usage): If
23315         UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
23317 2018-01-17  David Malcolm  <dmalcolm@redhat.com>
23319         PR lto/83121
23320         * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
23321         call the lto_location_cache before reading the
23322         DECL_SOURCE_LOCATION of the types.
23324 2018-01-17  Wilco Dijkstra  <wdijkstr@arm.com>
23325             Richard Sandiford  <richard.sandiford@linaro.org>
23327         * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
23328         * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
23329         (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
23330         SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
23331         * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
23332         Add declaration.
23333         * config/aarch64/constraints.md (aarch64_movti_operand):
23334         Limit immediates.
23335         * config/aarch64/predicates.md (Uti): Add new constraint.
23337 2018-01-17  Carl Love  <cel@us.ibm.com>
23339         * config/rs6000/vsx.md (define_expand xl_len_r,
23340         define_expand stxvl, define_expand *stxvl): Add match_dup argument.
23341         (define_insn): Add, match_dup 1 argument to define_insn stxvll and
23342         lxvll.
23343         (define_expand, define_insn): Move the shift left from  the
23344         define_insn to the define_expand for lxvl and stxvl instructions.
23345         * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
23346         and XL_LEN_R definitions to PURE.
23348 2018-01-17  Uros Bizjak  <ubizjak@gmail.com>
23350         * config/i386/i386.c (indirect_thunk_name): Declare regno
23351         as unsigned int.  Compare regno with INVALID_REGNUM.
23352         (output_indirect_thunk): Ditto.
23353         (output_indirect_thunk_function): Ditto.
23354         (ix86_code_end): Declare regno as unsigned int.  Use INVALID_REGNUM
23355         in the call to output_indirect_thunk_function.
23357 2018-01-17  Richard Sandiford  <richard.sandiford@linaro.org>
23359         PR middle-end/83884
23360         * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
23361         rather than the size of inner_type to determine the stack slot size
23362         when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
23364 2018-01-16  Sebastian Peryt  <sebastian.peryt@intel.com>
23366         PR target/83546
23367         * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
23368         to PTA_SILVERMONT.
23370 2018-01-16  Michael Meissner  <meissner@linux.vnet.ibm.com>
23372         * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
23373         endian Linux systems to optionally enable multilibs for selecting
23374         the long double type if the user configured an explicit type.
23375         * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
23376         have no long double multilibs if not defined.
23377         * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
23378         warn if the user used -mabi={ieee,ibm}longdouble and we built
23379         multilibs for long double.
23380         * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
23381         appropriate multilib option.
23382         (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
23383         multilib options.
23384         * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
23385         for building long double multilibs.
23386         * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
23388 2018-01-16  John David Anglin  <danglin@gcc.gnu.org>
23390         * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
23391         copies.
23393         * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
23394         64 bits.
23395         * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
23396         128 bits.
23398         * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
23399         variables.
23401         * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
23402         return value.
23404 2018-01-16  Eric Botcazou  <ebotcazou@adacore.com>
23406         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
23407         ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
23409 2018-01-16  Kelvin Nilsen  <kelvin@gcc.gnu.org>
23411         * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
23412         different rtl trees depending on TARGET_64BIT.
23413         (rs6000_gen_lvx): Likewise.
23415 2018-01-16  Eric Botcazou  <ebotcazou@adacore.com>
23417         * config/visium/visium.md (nop): Tweak comment.
23418         (hazard_nop): Likewise.
23420 2018-01-16  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
23422         * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
23423         -mspeculate-indirect-jumps.
23424         * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
23425         for -mno-speculate-indirect-jumps.
23426         (*call_indirect_elfv2<mode>_nospec): New define_insn.
23427         (*call_value_indirect_elfv2<mode>): Disable for
23428         -mno-speculate-indirect-jumps.
23429         (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
23430         (indirect_jump): Emit different RTL for
23431         -mno-speculate-indirect-jumps.
23432         (*indirect_jump<mode>): Disable for
23433         -mno-speculate-indirect-jumps.
23434         (*indirect_jump<mode>_nospec): New define_insn.
23435         (tablejump): Emit different RTL for
23436         -mno-speculate-indirect-jumps.
23437         (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
23438         (tablejumpsi_nospec): New define_expand.
23439         (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
23440         (tablejumpdi_nospec): New define_expand.
23441         (*tablejump<mode>_internal1): Disable for
23442         -mno-speculate-indirect-jumps.
23443         (*tablejump<mode>_internal1_nospec): New define_insn.
23444         * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
23445         option.
23447 2018-01-16  Artyom Skrobov tyomitch@gmail.com
23449         * caller-save.c (insert_save): Drop unnecessary parameter.  All
23450         callers updated.
23452 2018-01-16  Jakub Jelinek  <jakub@redhat.com>
23453             Richard Biener  <rguenth@suse.de>
23455         PR libgomp/83590
23456         * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
23457         return early, inline manually is_gimple_sizepos.  Make sure if we
23458         call gimplify_expr we don't end up with a gimple constant.
23459         * tree.c (variably_modified_type_p): Don't return true for
23460         is_gimple_constant (_t).  Inline manually is_gimple_sizepos.
23461         * gimplify.h (is_gimple_sizepos): Remove.
23463 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
23465         PR tree-optimization/83857
23466         * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
23467         vectorizable_live_operation for pure SLP statements.
23468         (vectorizable_live_operation): Handle PHIs.
23470 2018-01-16  Richard Biener  <rguenther@suse.de>
23472         PR tree-optimization/83867
23473         * tree-vect-stmts.c (vect_transform_stmt): Precompute
23474         nested_in_vect_loop_p since the scalar stmt may get invalidated.
23476 2018-01-16  Jakub Jelinek  <jakub@redhat.com>
23478         PR c/83844
23479         * stor-layout.c (handle_warn_if_not_align): Use byte_position and
23480         multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
23481         If off is not INTEGER_CST, issue a may not be aligned warning
23482         rather than isn't aligned.  Use isn%'t rather than isn't.
23483         * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
23484         into MULT_EXPR.
23485         <case MULT_EXPR>: Improve the case when bottom and one of the
23486         MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
23487         operand, in that case check if the other operand is multiple of
23488         bottom divided by the INTEGER_CST operand.
23490 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
23492         PR target/83858
23493         * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
23494         * config/pa/pa-protos.h (pa_function_arg_size): Declare.
23495         * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
23496         pa_function_arg_size instead of FUNCTION_ARG_SIZE.
23497         * config/pa/pa.c (pa_function_arg_advance): Likewise.
23498         (pa_function_arg, pa_arg_partial_bytes): Likewise.
23499         (pa_function_arg_size): New function.
23501 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
23503         * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
23504         in a separate statement.
23506 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
23508         PR tree-optimization/83847
23509         * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
23510         group gathers and scatters.
23512 2018-01-16  Jakub Jelinek  <jakub@redhat.com>
23514         PR rtl-optimization/86620
23515         * params.def (max-sched-ready-insns): Bump minimum value to 1.
23517         PR rtl-optimization/83213
23518         * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
23519         to last if both are JUMP_INSNs.
23521         PR tree-optimization/83843
23522         * gimple-ssa-store-merging.c
23523         (imm_store_chain_info::output_merged_store): Handle bit_not_p on
23524         store_immediate_info for bswap/nop orig_stores.
23526 2018-01-15  Andrew Waterman  <andrew@sifive.com>
23528         * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
23529         !TARGET_MUL.
23530         <UDIV>: Increase cost if !TARGET_DIV.
23532 2018-01-15  Segher Boessenkool  <segher@kernel.crashing.org>
23534         * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
23535         (define_attr "cr_logical_3op"): New.
23536         (cceq_ior_compare): Adjust.
23537         (cceq_ior_compare_complement): Adjust.
23538         (*cceq_rev_compare): Adjust.
23539         * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
23540         (is_cracked_insn): Adjust.
23541         (insn_must_be_first_in_group): Adjust.
23542         * config/rs6000/40x.md: Adjust.
23543         * config/rs6000/440.md: Adjust.
23544         * config/rs6000/476.md: Adjust.
23545         * config/rs6000/601.md: Adjust.
23546         * config/rs6000/603.md: Adjust.
23547         * config/rs6000/6xx.md: Adjust.
23548         * config/rs6000/7450.md: Adjust.
23549         * config/rs6000/7xx.md: Adjust.
23550         * config/rs6000/8540.md: Adjust.
23551         * config/rs6000/cell.md: Adjust.
23552         * config/rs6000/e300c2c3.md: Adjust.
23553         * config/rs6000/e500mc.md: Adjust.
23554         * config/rs6000/e500mc64.md: Adjust.
23555         * config/rs6000/e5500.md: Adjust.
23556         * config/rs6000/e6500.md: Adjust.
23557         * config/rs6000/mpc.md: Adjust.
23558         * config/rs6000/power4.md: Adjust.
23559         * config/rs6000/power5.md: Adjust.
23560         * config/rs6000/power6.md: Adjust.
23561         * config/rs6000/power7.md: Adjust.
23562         * config/rs6000/power8.md: Adjust.
23563         * config/rs6000/power9.md: Adjust.
23564         * config/rs6000/rs64.md: Adjust.
23565         * config/rs6000/titan.md: Adjust.
23567 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
23569         * config/i386/predicates.md (indirect_branch_operand): Rewrite
23570         ix86_indirect_branch_register logic.
23572 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
23574         * config/i386/constraints.md (Bs): Update
23575         ix86_indirect_branch_register check.  Don't check
23576         ix86_indirect_branch_register with GOT_memory_operand.
23577         (Bw): Likewise.
23578         * config/i386/predicates.md (GOT_memory_operand): Don't check
23579         ix86_indirect_branch_register here.
23580         (GOT32_symbol_operand): Likewise.
23582 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
23584         * config/i386/predicates.md (constant_call_address_operand):
23585         Rewrite ix86_indirect_branch_register logic.
23586         (sibcall_insn_operand): Likewise.
23588 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
23590         * config/i386/constraints.md (Bs): Replace
23591         ix86_indirect_branch_thunk_register with
23592         ix86_indirect_branch_register.
23593         (Bw): Likewise.
23594         * config/i386/i386.md (indirect_jump): Likewise.
23595         (tablejump): Likewise.
23596         (*sibcall_memory): Likewise.
23597         (*sibcall_value_memory): Likewise.
23598         Peepholes of indirect call and jump via memory: Likewise.
23599         * config/i386/i386.opt: Likewise.
23600         * config/i386/predicates.md (indirect_branch_operand): Likewise.
23601         (GOT_memory_operand): Likewise.
23602         (call_insn_operand): Likewise.
23603         (sibcall_insn_operand): Likewise.
23604         (GOT32_symbol_operand): Likewise.
23606 2018-01-15  Jakub Jelinek  <jakub@redhat.com>
23608         PR middle-end/83837
23609         * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
23610         type rather than type addr's type points to.
23611         (expand_omp_atomic_mutex): Likewise.
23612         (expand_omp_atomic): Likewise.
23614 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
23616         PR target/83839
23617         * config/i386/i386.c (output_indirect_thunk_function): Use
23618         ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
23619         for  __x86_return_thunk.
23621 2018-01-15  Richard Biener  <rguenther@suse.de>
23623         PR middle-end/83850
23624         * expmed.c (extract_bit_field_1): Fix typo.
23626 2018-01-15  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
23628         PR target/83687
23629         * config/arm/iterators.md (VF): New mode iterator.
23630         * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
23631         Remove integer-related logic from pattern.
23632         (neon_vabd<mode>_3): Likewise.
23634 2018-01-15  Jakub Jelinek  <jakub@redhat.com>
23636         PR middle-end/82694
23637         * common.opt (fstrict-overflow): No longer an alias.
23638         (fwrapv-pointer): New option.
23639         * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
23640         also for pointer types based on flag_wrapv_pointer.
23641         * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
23642         opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
23643         opts->x_flag_wrapv got set.
23644         * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
23645         changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
23646         POINTER_TYPE_OVERFLOW_UNDEFINED.
23647         * match.pd: Likewise in address comparison pattern.
23648         * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
23650 2018-01-15  Richard Biener  <rguenther@suse.de>
23652         PR lto/83804
23653         * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
23654         from TYPE_FIELDS.  Free TYPE_BINFO if not used by devirtualization.
23655         Reset type names to their identifier if their TYPE_DECL doesn't
23656         have linkage (and thus is used for ODR and devirt).
23657         (save_debug_info_for_decl): Remove.
23658         (save_debug_info_for_type): Likewise.
23659         (add_tree_to_fld_list): Adjust.
23660         * tree-pretty-print.c (dump_generic_node): Make dumping of
23661         type names more robust.
23663 2018-01-15  Richard Biener  <rguenther@suse.de>
23665         * BASE-VER: Bump to 8.0.1.
23667 2018-01-14  Martin Sebor  <msebor@redhat.com>
23669         PR other/83508
23670         * builtins.c (check_access): Avoid warning when the no-warning bit
23671         is set.
23673 2018-01-14  Cory Fields  <cory-nospam-@coryfields.com>
23675         * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
23676         * ira-color (allocno_hard_regs_compare): Likewise.
23678 2018-01-14  Nathan Rossi  <nathan@nathanrossi.com>
23680         PR target/83013
23681         * config/microblaze/microblaze.c (microblaze_asm_output_ident):
23682         Use .pushsection/.popsection.
23684 2018-01-14  Martin Sebor  <msebor@redhat.com>
23686         PR c++/81327
23687         * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
23689 2018-01-14  Jakub Jelinek  <jakub@redhat.com>
23691         * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
23692         entry from extra_headers.
23693         (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
23694         extra_headers, make the list bitwise identical to the i?86-*-* one.
23696 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
23698         * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
23699         -mcmodel=large with -mindirect-branch=thunk,
23700         -mindirect-branch=thunk-extern, -mfunction-return=thunk and
23701         -mfunction-return=thunk-extern.
23702         * doc/invoke.texi: Document -mcmodel=large is incompatible with
23703         -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
23704         -mfunction-return=thunk and -mfunction-return=thunk-extern.
23706 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
23708         * config/i386/i386.c (print_reg): Print the name of the full
23709         integer register without '%'.
23710         (ix86_print_operand): Handle 'V'.
23711         * doc/extend.texi: Document 'V' modifier.
23713 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
23715         * config/i386/constraints.md (Bs): Disallow memory operand for
23716         -mindirect-branch-register.
23717         (Bw): Likewise.
23718         * config/i386/predicates.md (indirect_branch_operand): Likewise.
23719         (GOT_memory_operand): Likewise.
23720         (call_insn_operand): Likewise.
23721         (sibcall_insn_operand): Likewise.
23722         (GOT32_symbol_operand): Likewise.
23723         * config/i386/i386.md (indirect_jump): Call convert_memory_address
23724         for -mindirect-branch-register.
23725         (tablejump): Likewise.
23726         (*sibcall_memory): Likewise.
23727         (*sibcall_value_memory): Likewise.
23728         Disallow peepholes of indirect call and jump via memory for
23729         -mindirect-branch-register.
23730         (*call_pop): Replace m with Bw.
23731         (*call_value_pop): Likewise.
23732         (*sibcall_pop_memory): Replace m with Bs.
23733         * config/i386/i386.opt (mindirect-branch-register): New option.
23734         * doc/invoke.texi: Document -mindirect-branch-register option.
23736 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
23738         * config/i386/i386-protos.h (ix86_output_function_return): New.
23739         * config/i386/i386.c (ix86_set_indirect_branch_type): Also
23740         set function_return_type.
23741         (indirect_thunk_name): Add ret_p to indicate thunk for function
23742         return.
23743         (output_indirect_thunk_function): Pass false to
23744         indirect_thunk_name.
23745         (ix86_output_indirect_branch_via_reg): Likewise.
23746         (ix86_output_indirect_branch_via_push): Likewise.
23747         (output_indirect_thunk_function): Create alias for function
23748         return thunk if regno < 0.
23749         (ix86_output_function_return): New function.
23750         (ix86_handle_fndecl_attribute): Handle function_return.
23751         (ix86_attribute_table): Add function_return.
23752         * config/i386/i386.h (machine_function): Add
23753         function_return_type.
23754         * config/i386/i386.md (simple_return_internal): Use
23755         ix86_output_function_return.
23756         (simple_return_internal_long): Likewise.
23757         * config/i386/i386.opt (mfunction-return=): New option.
23758         (indirect_branch): Mention -mfunction-return=.
23759         * doc/extend.texi: Document function_return function attribute.
23760         * doc/invoke.texi: Document -mfunction-return= option.
23762 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
23764         * config/i386/i386-opts.h (indirect_branch): New.
23765         * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
23766         * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
23767         with local indirect jump when converting indirect call and jump.
23768         (ix86_set_indirect_branch_type): New.
23769         (ix86_set_current_function): Call ix86_set_indirect_branch_type.
23770         (indirectlabelno): New.
23771         (indirect_thunk_needed): Likewise.
23772         (indirect_thunk_bnd_needed): Likewise.
23773         (indirect_thunks_used): Likewise.
23774         (indirect_thunks_bnd_used): Likewise.
23775         (INDIRECT_LABEL): Likewise.
23776         (indirect_thunk_name): Likewise.
23777         (output_indirect_thunk): Likewise.
23778         (output_indirect_thunk_function): Likewise.
23779         (ix86_output_indirect_branch_via_reg): Likewise.
23780         (ix86_output_indirect_branch_via_push): Likewise.
23781         (ix86_output_indirect_branch): Likewise.
23782         (ix86_output_indirect_jmp): Likewise.
23783         (ix86_code_end): Call output_indirect_thunk_function if needed.
23784         (ix86_output_call_insn): Call ix86_output_indirect_branch if
23785         needed.
23786         (ix86_handle_fndecl_attribute): Handle indirect_branch.
23787         (ix86_attribute_table): Add indirect_branch.
23788         * config/i386/i386.h (machine_function): Add indirect_branch_type
23789         and has_local_indirect_jump.
23790         * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
23791         to true.
23792         (tablejump): Likewise.
23793         (*indirect_jump): Use ix86_output_indirect_jmp.
23794         (*tablejump_1): Likewise.
23795         (simple_return_indirect_internal): Likewise.
23796         * config/i386/i386.opt (mindirect-branch=): New option.
23797         (indirect_branch): New.
23798         (keep): Likewise.
23799         (thunk): Likewise.
23800         (thunk-inline): Likewise.
23801         (thunk-extern): Likewise.
23802         * doc/extend.texi: Document indirect_branch function attribute.
23803         * doc/invoke.texi: Document -mindirect-branch= option.
23805 2018-01-14  Jan Hubicka  <hubicka@ucw.cz>
23807         PR ipa/83051
23808         * ipa-inline.c (edge_badness): Tolerate roundoff errors.
23810 2018-01-14  Richard Sandiford  <richard.sandiford@linaro.org>
23812         * ipa-inline.c (want_inline_small_function_p): Return false if
23813         inlining has already failed with CIF_FINAL_ERROR.
23814         (update_caller_keys): Call want_inline_small_function_p before
23815         can_inline_edge_p.
23816         (update_callee_keys): Likewise.
23818 2018-01-10  Kelvin Nilsen  <kelvin@gcc.gnu.org>
23820         * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
23821         New function.
23822         (rs6000_quadword_masked_address_p): Likewise.
23823         (quad_aligned_load_p): Likewise.
23824         (quad_aligned_store_p): Likewise.
23825         (const_load_sequence_p): Add comment to describe the outer-most loop.
23826         (mimic_memory_attributes_and_flags): New function.
23827         (rs6000_gen_stvx): Likewise.
23828         (replace_swapped_aligned_store): Likewise.
23829         (rs6000_gen_lvx): Likewise.
23830         (replace_swapped_aligned_load): Likewise.
23831         (replace_swapped_load_constant): Capitalize argument name in
23832         comment describing this function.
23833         (rs6000_analyze_swaps): Add a third pass to search for vector loads
23834         and stores that access quad-word aligned addresses and replace
23835         with stvx or lvx instructions when appropriate.
23836         * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
23837         New function prototype.
23838         (rs6000_quadword_masked_address_p): Likewise.
23839         (rs6000_gen_lvx): Likewise.
23840         (rs6000_gen_stvx): Likewise.
23841         * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
23842         VSX_D (V2DF, V2DI), modify this split to select lvx instruction
23843         when memory address is aligned.
23844         (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
23845         this split to select lvx instruction when memory address is aligned.
23846         (*vsx_le_perm_load_v8hi): Modify this split to select lvx
23847         instruction when memory address is aligned.
23848         (*vsx_le_perm_load_v16qi): Likewise.
23849         (four unnamed splitters): Modify to select the stvx instruction
23850         when memory is aligned.
23852 2018-01-13  Jan Hubicka  <hubicka@ucw.cz>
23854         * predict.c (determine_unlikely_bbs): Handle correctly BBs
23855         which appears in the queue multiple times.
23857 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
23858             Alan Hayward  <alan.hayward@arm.com>
23859             David Sherwood  <david.sherwood@arm.com>
23861         * tree-vectorizer.h (vec_lower_bound): New structure.
23862         (_loop_vec_info): Add check_nonzero and lower_bounds.
23863         (LOOP_VINFO_CHECK_NONZERO): New macro.
23864         (LOOP_VINFO_LOWER_BOUNDS): Likewise.
23865         (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
23866         * tree-data-ref.h (dr_with_seg_len): Add access_size and align
23867         fields.  Make seg_len the distance travelled, not including the
23868         access size.
23869         (dr_direction_indicator): Declare.
23870         (dr_zero_step_indicator): Likewise.
23871         (dr_known_forward_stride_p): Likewise.
23872         * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
23873         tree-ssanames.h.
23874         (runtime_alias_check_p): Allow runtime alias checks with
23875         variable strides.
23876         (operator ==): Compare access_size and align.
23877         (prune_runtime_alias_test_list): Rework for new distinction between
23878         the access_size and seg_len.
23879         (create_intersect_range_checks_index): Likewise.  Cope with polynomial
23880         segment lengths.
23881         (get_segment_min_max): New function.
23882         (create_intersect_range_checks): Use it.
23883         (dr_step_indicator): New function.
23884         (dr_direction_indicator): Likewise.
23885         (dr_zero_step_indicator): Likewise.
23886         (dr_known_forward_stride_p): Likewise.
23887         * tree-loop-distribution.c (data_ref_segment_size): Return
23888         DR_STEP * (niters - 1).
23889         (compute_alias_check_pairs): Update call to the dr_with_seg_len
23890         constructor.
23891         * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
23892         (vect_preserves_scalar_order_p): New function, split out from...
23893         (vect_analyze_data_ref_dependence): ...here.  Check for zero steps.
23894         (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
23895         (vect_vfa_access_size): New function.
23896         (vect_vfa_align): Likewise.
23897         (vect_compile_time_alias): Take access_size_a and access_b arguments.
23898         (dump_lower_bound): New function.
23899         (vect_check_lower_bound): Likewise.
23900         (vect_small_gap_p): Likewise.
23901         (vectorizable_with_step_bound_p): Likewise.
23902         (vect_prune_runtime_alias_test_list): Ignore cross-iteration
23903         depencies if the vectorization factor is 1.  Convert the checks
23904         for nonzero steps into checks on the bounds of DR_STEP.  Try using
23905         a bunds check for variable steps if the minimum required step is
23906         relatively small. Update calls to the dr_with_seg_len
23907         constructor and to vect_compile_time_alias.
23908         * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
23909         function.
23910         (vect_loop_versioning): Call it.
23911         * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
23912         when retrying.
23913         (vect_estimate_min_profitable_iters): Account for any bounds checks.
23915 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
23916             Alan Hayward  <alan.hayward@arm.com>
23917             David Sherwood  <david.sherwood@arm.com>
23919         * doc/sourcebuild.texi (vect_scatter_store): Document.
23920         * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
23921         optabs.
23922         * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
23923         Document.
23924         * genopinit.c (main): Add supports_vec_scatter_store and
23925         supports_vec_scatter_store_cached to target_optabs.
23926         * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
23927         IFN_MASK_SCATTER_STORE.
23928         * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
23929         functions.
23930         * internal-fn.h (internal_store_fn_p): Declare.
23931         (internal_fn_stored_value_index): Likewise.
23932         * internal-fn.c (scatter_store_direct): New macro.
23933         (expand_scatter_store_optab_fn): New function.
23934         (direct_scatter_store_optab_supported_p): New macro.
23935         (internal_store_fn_p): New function.
23936         (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
23937         IFN_MASK_SCATTER_STORE.
23938         (internal_fn_mask_index): Likewise.
23939         (internal_fn_stored_value_index): New function.
23940         (internal_gather_scatter_fn_supported_p): Adjust operand numbers
23941         for scatter stores.
23942         * optabs-query.h (supports_vec_scatter_store_p): Declare.
23943         * optabs-query.c (supports_vec_scatter_store_p): New function.
23944         * tree-vectorizer.h (vect_get_store_rhs): Declare.
23945         * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
23946         true for scatter stores.
23947         (vect_gather_scatter_fn_p): Handle scatter stores too.
23948         (vect_check_gather_scatter): Consider using scatter stores if
23949         supports_vec_scatter_store_p.
23950         * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
23951         scatter stores too.
23952         * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
23953         internal_fn_stored_value_index.
23954         (check_load_store_masking): Handle scatter stores too.
23955         (vect_get_store_rhs): Make public.
23956         (vectorizable_call): Use internal_store_fn_p.
23957         (vectorizable_store): Handle scatter store internal functions.
23958         (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
23959         when deciding whether the end of the group has been reached.
23960         * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
23961         * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
23962         (mask_scatter_store<mode>): New insns.
23964 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
23965             Alan Hayward  <alan.hayward@arm.com>
23966             David Sherwood  <david.sherwood@arm.com>
23968         * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
23969         * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
23970         * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
23971         function.
23972         (vect_use_strided_gather_scatters_p): Take a masked_p argument.
23973         Use vect_truncate_gather_scatter_offset if we can't treat the
23974         operation as a normal gather load or scatter store.
23975         (get_group_load_store_type): Take the gather_scatter_info
23976         as argument.  Try using a gather load or scatter store for
23977         single-element groups.
23978         (get_load_store_type): Update calls to get_group_load_store_type
23979         and vect_use_strided_gather_scatters_p.
23981 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
23982             Alan Hayward  <alan.hayward@arm.com>
23983             David Sherwood  <david.sherwood@arm.com>
23985         * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
23986         optional tree argument.
23987         * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
23988         null target hooks.
23989         (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
23990         but continue to use the current value as a fallback.
23991         (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
23992         to compare the updates.
23993         * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
23994         (get_load_store_type): Use it when handling a strided access.
23995         (vect_get_strided_load_store_ops): New function.
23996         (vect_get_data_ptr_increment): Likewise.
23997         (vectorizable_load): Handle strided gather loads.  Always pass
23998         a step to vect_create_data_ref_ptr and bump_vector_ptr.
24000 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
24001             Alan Hayward  <alan.hayward@arm.com>
24002             David Sherwood  <david.sherwood@arm.com>
24004         * doc/md.texi (gather_load@var{m}): Document.
24005         (mask_gather_load@var{m}): Likewise.
24006         * genopinit.c (main): Add supports_vec_gather_load and
24007         supports_vec_gather_load_cached to target_optabs.
24008         * optabs-tree.c (init_tree_optimization_optabs): Use
24009         ggc_cleared_alloc to allocate target_optabs.
24010         * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
24011         * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
24012         functions.
24013         * internal-fn.h (internal_load_fn_p): Declare.
24014         (internal_gather_scatter_fn_p): Likewise.
24015         (internal_fn_mask_index): Likewise.
24016         (internal_gather_scatter_fn_supported_p): Likewise.
24017         * internal-fn.c (gather_load_direct): New macro.
24018         (expand_gather_load_optab_fn): New function.
24019         (direct_gather_load_optab_supported_p): New macro.
24020         (direct_internal_fn_optab): New function.
24021         (internal_load_fn_p): Likewise.
24022         (internal_gather_scatter_fn_p): Likewise.
24023         (internal_fn_mask_index): Likewise.
24024         (internal_gather_scatter_fn_supported_p): Likewise.
24025         * optabs-query.c (supports_at_least_one_mode_p): New function.
24026         (supports_vec_gather_load_p): Likewise.
24027         * optabs-query.h (supports_vec_gather_load_p): Declare.
24028         * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
24029         and memory_type field.
24030         (NUM_PATTERNS): Bump to 15.
24031         * tree-vect-data-refs.c: Include internal-fn.h.
24032         (vect_gather_scatter_fn_p): New function.
24033         (vect_describe_gather_scatter_call): Likewise.
24034         (vect_check_gather_scatter): Try using internal functions for
24035         gather loads.  Recognize existing calls to a gather load function.
24036         (vect_analyze_data_refs): Consider using gather loads if
24037         supports_vec_gather_load_p.
24038         * tree-vect-patterns.c (vect_get_load_store_mask): New function.
24039         (vect_get_gather_scatter_offset_type): Likewise.
24040         (vect_convert_mask_for_vectype): Likewise.
24041         (vect_add_conversion_to_patterm): Likewise.
24042         (vect_try_gather_scatter_pattern): Likewise.
24043         (vect_recog_gather_scatter_pattern): New pattern recognizer.
24044         (vect_vect_recog_func_ptrs): Add it.
24045         * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
24046         internal_fn_mask_index and internal_gather_scatter_fn_p.
24047         (check_load_store_masking): Take the gather_scatter_info as an
24048         argument and handle gather loads.
24049         (vect_get_gather_scatter_ops): New function.
24050         (vectorizable_call): Check internal_load_fn_p.
24051         (vectorizable_load): Likewise.  Handle gather load internal
24052         functions.
24053         (vectorizable_store): Update call to check_load_store_masking.
24054         * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
24055         * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
24056         * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
24057         (aarch64_gather_scale_operand_d): New predicates.
24058         * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
24059         (mask_gather_load<mode>): New insns.
24061 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
24062             Alan Hayward  <alan.hayward@arm.com>
24063             David Sherwood  <david.sherwood@arm.com>
24065         * optabs.def (fold_left_plus_optab): New optab.
24066         * doc/md.texi (fold_left_plus_@var{m}): Document.
24067         * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
24068         * internal-fn.c (fold_left_direct): Define.
24069         (expand_fold_left_optab_fn): Likewise.
24070         (direct_fold_left_optab_supported_p): Likewise.
24071         * fold-const-call.c (fold_const_fold_left): New function.
24072         (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
24073         * tree-parloops.c (valid_reduction_p): New function.
24074         (gather_scalar_reductions): Use it.
24075         * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
24076         (vect_finish_replace_stmt): Declare.
24077         * tree-vect-loop.c (fold_left_reduction_fn): New function.
24078         (needs_fold_left_reduction_p): New function, split out from...
24079         (vect_is_simple_reduction): ...here.  Accept reductions that
24080         forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
24081         (vect_force_simple_reduction): Also store the reduction type in
24082         the assignment's STMT_VINFO_REDUC_TYPE.
24083         (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
24084         (merge_with_identity): New function.
24085         (vect_expand_fold_left): Likewise.
24086         (vectorize_fold_left_reduction): Likewise.
24087         (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION.  Leave the
24088         scalar phi in place for it.  Check for target support and reject
24089         cases that would reassociate the operation.  Defer the transform
24090         phase to vectorize_fold_left_reduction.
24091         * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
24092         * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
24093         (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
24095 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
24097         * tree-if-conv.c (predicate_mem_writes): Remove redundant
24098         call to ifc_temp_var.
24100 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
24101             Alan Hayward  <alan.hayward@arm.com>
24102             David Sherwood  <david.sherwood@arm.com>
24104         * target.def (legitimize_address_displacement): Take the original
24105         offset as a poly_int.
24106         * targhooks.h (default_legitimize_address_displacement): Update
24107         accordingly.
24108         * targhooks.c (default_legitimize_address_displacement): Likewise.
24109         * doc/tm.texi: Regenerate.
24110         * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
24111         as an argument, moving assert of ad->disp == ad->disp_term to...
24112         (process_address_1): ...here.  Update calls to base_plus_disp_to_reg.
24113         Try calling targetm.legitimize_address_displacement before expanding
24114         the address rather than afterwards, and adjust for the new interface.
24115         * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
24116         Match the new hook interface.  Handle SVE addresses.
24117         * config/sh/sh.c (sh_legitimize_address_displacement): Make the
24118         new hook interface.
24120 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
24122         * Makefile.in (OBJS): Add early-remat.o.
24123         * target.def (select_early_remat_modes): New hook.
24124         * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
24125         * doc/tm.texi: Regenerate.
24126         * targhooks.h (default_select_early_remat_modes): Declare.
24127         * targhooks.c (default_select_early_remat_modes): New function.
24128         * timevar.def (TV_EARLY_REMAT): New timevar.
24129         * passes.def (pass_early_remat): New pass.
24130         * tree-pass.h (make_pass_early_remat): Declare.
24131         * early-remat.c: New file.
24132         * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
24133         function.
24134         (TARGET_SELECT_EARLY_REMAT_MODES): Define.
24136 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
24137             Alan Hayward  <alan.hayward@arm.com>
24138             David Sherwood  <david.sherwood@arm.com>
24140         * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
24141         vfm1 with a bound_epilog parameter.
24142         (vect_do_peeling): Update calls accordingly, and move the prologue
24143         call earlier in the function.  Treat the base bound_epilog as 0 for
24144         fully-masked loops and retain vf - 1 for other loops.  Add 1 to
24145         this base when peeling for gaps.
24146         * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
24147         with fully-masked loops.
24148         (vect_estimate_min_profitable_iters): Handle the single peeled
24149         iteration in that case.
24151 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
24152             Alan Hayward  <alan.hayward@arm.com>
24153             David Sherwood  <david.sherwood@arm.com>
24155         * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
24156         single-element interleaving even if the size is not a power of 2.
24157         * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
24158         accesses for single-element interleaving if the group size is
24159         not a power of 2.
24161 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
24162             Alan Hayward  <alan.hayward@arm.com>
24163             David Sherwood  <david.sherwood@arm.com>
24165         * doc/md.texi (fold_extract_last_@var{m}): Document.
24166         * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
24167         * optabs.def (fold_extract_last_optab): New optab.
24168         * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
24169         * internal-fn.c (fold_extract_direct): New macro.
24170         (expand_fold_extract_optab_fn): Likewise.
24171         (direct_fold_extract_optab_supported_p): Likewise.
24172         * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
24173         * tree-vect-loop.c (vect_model_reduction_cost): Handle
24174         EXTRACT_LAST_REDUCTION.
24175         (get_initial_def_for_reduction): Do not create an initial vector
24176         for EXTRACT_LAST_REDUCTION reductions.
24177         (vectorizable_reduction): Leave the scalar phi in place for
24178         EXTRACT_LAST_REDUCTIONs.  Try using EXTRACT_LAST_REDUCTION
24179         ahead of INTEGER_INDUC_COND_REDUCTION.  Do not check for an
24180         epilogue code for EXTRACT_LAST_REDUCTION and defer the
24181         transform phase to vectorizable_condition.
24182         * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
24183         split out from...
24184         (vect_finish_stmt_generation): ...here.
24185         (vect_finish_replace_stmt): New function.
24186         (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
24187         * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
24188         pattern.
24189         * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
24191 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
24192             Alan Hayward  <alan.hayward@arm.com>
24193             David Sherwood  <david.sherwood@arm.com>
24195         * doc/md.texi (extract_last_@var{m}): Document.
24196         * optabs.def (extract_last_optab): New optab.
24197         * internal-fn.def (EXTRACT_LAST): New internal function.
24198         * internal-fn.c (cond_unary_direct): New macro.
24199         (expand_cond_unary_optab_fn): Likewise.
24200         (direct_cond_unary_optab_supported_p): Likewise.
24201         * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
24202         loops using EXTRACT_LAST.
24203         * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
24204         (extract_last_<mode>): ...this optab.
24205         (vec_extract<mode><Vel>): Update accordingly.
24207 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
24208             Alan Hayward  <alan.hayward@arm.com>
24209             David Sherwood  <david.sherwood@arm.com>
24211         * target.def (empty_mask_is_expensive): New hook.
24212         * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
24213         * doc/tm.texi: Regenerate.
24214         * targhooks.h (default_empty_mask_is_expensive): Declare.
24215         * targhooks.c (default_empty_mask_is_expensive): New function.
24216         * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
24217         if the target says that empty masks are expensive.
24218         * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
24219         New function.
24220         (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
24222 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
24223             Alan Hayward  <alan.hayward@arm.com>
24224             David Sherwood  <david.sherwood@arm.com>
24226         * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
24227         (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
24228         (vect_use_loop_mask_for_alignment_p): New function.
24229         (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
24230         * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
24231         niters_skip argument.  Make sure that the first niters_skip elements
24232         of the first iteration are inactive.
24233         (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
24234         Update call to vect_set_loop_masks_directly.
24235         (get_misalign_in_elems): New function, split out from...
24236         (vect_gen_prolog_loop_niters): ...here.
24237         (vect_update_init_of_dr): Take a code argument that specifies whether
24238         the adjustment should be added or subtracted.
24239         (vect_update_init_of_drs): Likewise.
24240         (vect_prepare_for_masked_peels): New function.
24241         (vect_do_peeling): Skip prologue peeling if we're using a mask
24242         instead.  Update call to vect_update_inits_of_drs.
24243         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
24244         mask_skip_niters.
24245         (vect_analyze_loop_2): Allow fully-masked loops with peeling for
24246         alignment.  Do not include the number of peeled iterations in
24247         the minimum threshold in that case.
24248         (vectorizable_induction): Adjust the start value down by
24249         LOOP_VINFO_MASK_SKIP_NITERS iterations.
24250         (vect_transform_loop): Call vect_prepare_for_masked_peels.
24251         Take the number of skipped iterations into account when calculating
24252         the loop bounds.
24253         * tree-vect-stmts.c (vect_gen_while_not): New function.
24255 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
24256             Alan Hayward  <alan.hayward@arm.com>
24257             David Sherwood  <david.sherwood@arm.com>
24259         * doc/sourcebuild.texi (vect_fully_masked): Document.
24260         * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
24261         default value to 0.
24262         * tree-vect-loop.c (vect_analyze_loop_costing): New function,
24263         split out from...
24264         (vect_analyze_loop_2): ...here. Don't check the vectorization
24265         factor against the number of loop iterations if the loop is
24266         fully-masked.
24268 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
24269             Alan Hayward  <alan.hayward@arm.com>
24270             David Sherwood  <david.sherwood@arm.com>
24272         * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
24273         (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
24274         (dump_groups): Update accordingly.
24275         (iv_use::mem_type): New member variable.
24276         (address_p): New function.
24277         (record_use): Add a mem_type argument and initialize the new
24278         mem_type field.
24279         (record_group_use): Add a mem_type argument.  Use address_p.
24280         Remove obsolete null checks of base_object.  Update call to record_use.
24281         (find_interesting_uses_op): Update call to record_group_use.
24282         (find_interesting_uses_cond): Likewise.
24283         (find_interesting_uses_address): Likewise.
24284         (get_mem_type_for_internal_fn): New function.
24285         (find_address_like_use): Likewise.
24286         (find_interesting_uses_stmt): Try find_address_like_use before
24287         calling find_interesting_uses_op.
24288         (addr_offset_valid_p): Use the iv mem_type field as the type
24289         of the addressed memory.
24290         (add_autoinc_candidates): Likewise.
24291         (get_address_cost): Likewise.
24292         (split_small_address_groups_p): Use address_p.
24293         (split_address_groups): Likewise.
24294         (add_iv_candidate_for_use): Likewise.
24295         (autoinc_possible_for_pair): Likewise.
24296         (rewrite_groups): Likewise.
24297         (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
24298         (determine_group_iv_cost): Update after split of USE_ADDRESS.
24299         (get_alias_ptr_type_for_ptr_address): New function.
24300         (rewrite_use_address): Rewrite address uses in calls that were
24301         identified by find_address_like_use.
24303 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
24304             Alan Hayward  <alan.hayward@arm.com>
24305             David Sherwood  <david.sherwood@arm.com>
24307         * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
24308         TARGET_MEM_REFs.
24309         * gimple-expr.h (is_gimple_addressable: Likewise.
24310         * gimple-expr.c (is_gimple_address): Likewise.
24311         * internal-fn.c (expand_call_mem_ref): New function.
24312         (expand_mask_load_optab_fn): Use it.
24313         (expand_mask_store_optab_fn): Likewise.
24315 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
24316             Alan Hayward  <alan.hayward@arm.com>
24317             David Sherwood  <david.sherwood@arm.com>
24319         * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
24320         (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
24321         (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
24322         (cond_umax@var{mode}): Document.
24323         * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
24324         (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
24325         (cond_umin_optab, cond_umax_optab): New optabs.
24326         * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
24327         (COND_IOR, COND_XOR): New internal functions.
24328         * internal-fn.h (get_conditional_internal_fn): Declare.
24329         * internal-fn.c (cond_binary_direct): New macro.
24330         (expand_cond_binary_optab_fn): Likewise.
24331         (direct_cond_binary_optab_supported_p): Likewise.
24332         (get_conditional_internal_fn): New function.
24333         * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
24334         Cope with reduction statements that are vectorized as calls rather
24335         than assignments.
24336         * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
24337         * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
24338         (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
24339         (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
24340         (UNSPEC_COND_EOR): New unspecs.
24341         (optab): Add mappings for them.
24342         (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
24343         (sve_int_op, sve_fp_op): New int attributes.
24345 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
24346             Alan Hayward  <alan.hayward@arm.com>
24347             David Sherwood  <david.sherwood@arm.com>
24349         * optabs.def (while_ult_optab): New optab.
24350         * doc/md.texi (while_ult@var{m}@var{n}): Document.
24351         * internal-fn.def (WHILE_ULT): New internal function.
24352         * internal-fn.h (direct_internal_fn_supported_p): New override
24353         that takes two types as argument.
24354         * internal-fn.c (while_direct): New macro.
24355         (expand_while_optab_fn): New function.
24356         (convert_optab_supported_p): Likewise.
24357         (direct_while_optab_supported_p): New macro.
24358         * wide-int.h (wi::udiv_ceil): New function.
24359         * tree-vectorizer.h (rgroup_masks): New structure.
24360         (vec_loop_masks): New typedef.
24361         (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
24362         and fully_masked_p.
24363         (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
24364         (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
24365         (vect_max_vf): New function.
24366         (slpeel_make_loop_iterate_ntimes): Delete.
24367         (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
24368         (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
24369         (vect_record_loop_mask, vect_get_loop_mask): Likewise.
24370         * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
24371         internal-fn.h, stor-layout.h and optabs-query.h.
24372         (vect_set_loop_mask): New function.
24373         (add_preheader_seq): Likewise.
24374         (add_header_seq): Likewise.
24375         (interleave_supported_p): Likewise.
24376         (vect_maybe_permute_loop_masks): Likewise.
24377         (vect_set_loop_masks_directly): Likewise.
24378         (vect_set_loop_condition_masked): Likewise.
24379         (vect_set_loop_condition_unmasked): New function, split out from
24380         slpeel_make_loop_iterate_ntimes.
24381         (slpeel_make_loop_iterate_ntimes): Rename to..
24382         (vect_set_loop_condition): ...this.  Use vect_set_loop_condition_masked
24383         for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
24384         (vect_do_peeling): Update call accordingly.
24385         (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
24386         loops.
24387         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
24388         mask_compare_type, can_fully_mask_p and fully_masked_p.
24389         (release_vec_loop_masks): New function.
24390         (_loop_vec_info): Use it to free the loop masks.
24391         (can_produce_all_loop_masks_p): New function.
24392         (vect_get_max_nscalars_per_iter): Likewise.
24393         (vect_verify_full_masking): Likewise.
24394         (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
24395         retries, and free the mask rgroups before retrying.  Check loop-wide
24396         reasons for disallowing fully-masked loops.  Make the final decision
24397         about whether use a fully-masked loop or not.
24398         (vect_estimate_min_profitable_iters): Do not assume that peeling
24399         for the number of iterations will be needed for fully-masked loops.
24400         (vectorizable_reduction): Disable fully-masked loops.
24401         (vectorizable_live_operation): Likewise.
24402         (vect_halve_mask_nunits): New function.
24403         (vect_double_mask_nunits): Likewise.
24404         (vect_record_loop_mask): Likewise.
24405         (vect_get_loop_mask): Likewise.
24406         (vect_transform_loop): Handle the case in which the final loop
24407         iteration might handle a partial vector.  Call vect_set_loop_condition
24408         instead of slpeel_make_loop_iterate_ntimes.
24409         * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
24410         (check_load_store_masking): New function.
24411         (prepare_load_store_mask): Likewise.
24412         (vectorizable_store): Handle fully-masked loops.
24413         (vectorizable_load): Likewise.
24414         (supportable_widening_operation): Use vect_halve_mask_nunits for
24415         booleans.
24416         (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
24417         (vect_gen_while): New function.
24418         * config/aarch64/aarch64.md (umax<mode>3): New expander.
24419         (aarch64_uqdec<mode>): New insn.
24421 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
24422             Alan Hayward  <alan.hayward@arm.com>
24423             David Sherwood  <david.sherwood@arm.com>
24425         * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
24426         (reduc_xor_scal_optab): New optabs.
24427         * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
24428         (reduc_xor_scal_@var{m}): Document.
24429         * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
24430         * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
24431         internal functions.
24432         * fold-const-call.c (fold_const_call): Handle them.
24433         * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
24434         internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
24435         * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
24436         (*reduc_<bit_reduc>_scal_<mode>): New patterns.
24437         * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
24438         (UNSPEC_XORV): New unspecs.
24439         (optab): Add entries for them.
24440         (BITWISEV): New int iterator.
24441         (bit_reduc_op): New int attributes.
24443 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
24444             Alan Hayward  <alan.hayward@arm.com>
24445             David Sherwood  <david.sherwood@arm.com>
24447         * doc/md.texi (vec_shl_insert_@var{m}): New optab.
24448         * internal-fn.def (VEC_SHL_INSERT): New internal function.
24449         * optabs.def (vec_shl_insert_optab): New optab.
24450         * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
24451         (duplicate_and_interleave): Likewise.
24452         * tree-vect-loop.c: Include internal-fn.h.
24453         (neutral_op_for_slp_reduction): New function, split out from
24454         get_initial_defs_for_reduction.
24455         (get_initial_def_for_reduction): Handle option 2 for variable-length
24456         vectors by loading the neutral value into a vector and then shifting
24457         the initial value into element 0.
24458         (get_initial_defs_for_reduction): Replace the code argument with
24459         the neutral value calculated by neutral_op_for_slp_reduction.
24460         Use gimple_build_vector for constant-length vectors.
24461         Use IFN_VEC_SHL_INSERT for variable-length vectors if all
24462         but the first group_size elements have a neutral value.
24463         Use duplicate_and_interleave otherwise.
24464         (vect_create_epilog_for_reduction): Take a neutral_op parameter.
24465         Update call to get_initial_defs_for_reduction.  Handle SLP
24466         reductions for variable-length vectors by creating one vector
24467         result for each scalar result, with the elements associated
24468         with other scalar results stubbed out with the neutral value.
24469         (vectorizable_reduction): Call neutral_op_for_slp_reduction.
24470         Require IFN_VEC_SHL_INSERT for double reductions on
24471         variable-length vectors, or SLP reductions that have
24472         a neutral value.  Require can_duplicate_and_interleave_p
24473         support for variable-length unchained SLP reductions if there
24474         is no neutral value, such as for MIN/MAX reductions.  Also require
24475         the number of vector elements to be a multiple of the number of
24476         SLP statements when doing variable-length unchained SLP reductions.
24477         Update call to vect_create_epilog_for_reduction.
24478         * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
24479         and remove initial values.
24480         (duplicate_and_interleave): Make public.
24481         * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
24482         * config/aarch64/aarch64-sve.md (vec_shl_insert_<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         * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
24489         (can_duplicate_and_interleave_p): New function.
24490         (vect_get_and_check_slp_defs): Take the vector of statements
24491         rather than just the current one.  Remove excess parentheses.
24492         Restriction rejectinon of vect_constant_def and vect_external_def
24493         for variable-length vectors to boolean types, or types for which
24494         can_duplicate_and_interleave_p is false.
24495         (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
24496         (duplicate_and_interleave): New function.
24497         (vect_get_constant_vectors): Use gimple_build_vector for
24498         constant-length vectors and suitable variable-length constant
24499         vectors.  Use duplicate_and_interleave for other variable-length
24500         vectors.  Don't defer the update when inserting new statements.
24502 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
24503             Alan Hayward  <alan.hayward@arm.com>
24504             David Sherwood  <david.sherwood@arm.com>
24506         * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
24507         min_profitable_iters doesn't go negative.
24509 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
24510             Alan Hayward  <alan.hayward@arm.com>
24511             David Sherwood  <david.sherwood@arm.com>
24513         * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
24514         (vec_mask_store_lanes@var{m}@var{n}): Likewise.
24515         * optabs.def (vec_mask_load_lanes_optab): New optab.
24516         (vec_mask_store_lanes_optab): Likewise.
24517         * internal-fn.def (MASK_LOAD_LANES): New internal function.
24518         (MASK_STORE_LANES): Likewise.
24519         * internal-fn.c (mask_load_lanes_direct): New macro.
24520         (mask_store_lanes_direct): Likewise.
24521         (expand_mask_load_optab_fn): Handle masked operations.
24522         (expand_mask_load_lanes_optab_fn): New macro.
24523         (expand_mask_store_optab_fn): Handle masked operations.
24524         (expand_mask_store_lanes_optab_fn): New macro.
24525         (direct_mask_load_lanes_optab_supported_p): Likewise.
24526         (direct_mask_store_lanes_optab_supported_p): Likewise.
24527         * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
24528         parameter.
24529         (vect_load_lanes_supported): Likewise.
24530         * tree-vect-data-refs.c (strip_conversion): New function.
24531         (can_group_stmts_p): Likewise.
24532         (vect_analyze_data_ref_accesses): Use it instead of checking
24533         for a pair of assignments.
24534         (vect_store_lanes_supported): Take a masked_p parameter.
24535         (vect_load_lanes_supported): Likewise.
24536         * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
24537         vect_store_lanes_supported and vect_load_lanes_supported.
24538         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
24539         * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
24540         parameter.  Don't allow gaps for masked accesses.
24541         Use vect_get_store_rhs.  Update calls to vect_store_lanes_supported
24542         and vect_load_lanes_supported.
24543         (get_load_store_type): Take a masked_p parameter and update
24544         call to get_group_load_store_type.
24545         (vectorizable_store): Update call to get_load_store_type.
24546         Handle IFN_MASK_STORE_LANES.
24547         (vectorizable_load): Update call to get_load_store_type.
24548         Handle IFN_MASK_LOAD_LANES.
24550 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
24551             Alan Hayward  <alan.hayward@arm.com>
24552             David Sherwood  <david.sherwood@arm.com>
24554         * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
24555         modes for SVE.
24556         * config/aarch64/aarch64-protos.h
24557         (aarch64_sve_struct_memory_operand_p): Declare.
24558         * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
24559         (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
24560         (VPRED, vpred): Handle SVE structure modes.
24561         * config/aarch64/constraints.md (Utx): New constraint.
24562         * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
24563         (aarch64_sve_struct_nonimmediate_operand): New predicates.
24564         * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
24565         * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
24566         (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
24567         structure modes.  Split into pieces after RA.
24568         (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
24569         (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
24570         New patterns.
24571         * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
24572         SVE structure modes.
24573         (aarch64_classify_address): Likewise.
24574         (sizetochar): Move earlier in file.
24575         (aarch64_print_operand): Handle SVE register lists.
24576         (aarch64_array_mode): New function.
24577         (aarch64_sve_struct_memory_operand_p): Likewise.
24578         (TARGET_ARRAY_MODE): Redefine.
24580 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
24581             Alan Hayward  <alan.hayward@arm.com>
24582             David Sherwood  <david.sherwood@arm.com>
24584         * target.def (array_mode): New target hook.
24585         * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
24586         * doc/tm.texi: Regenerate.
24587         * hooks.h (hook_optmode_mode_uhwi_none): Declare.
24588         * hooks.c (hook_optmode_mode_uhwi_none): New function.
24589         * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
24590         targetm.array_mode.
24591         * stor-layout.c (mode_for_array): Likewise.  Support polynomial
24592         type sizes.
24594 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
24595             Alan Hayward  <alan.hayward@arm.com>
24596             David Sherwood  <david.sherwood@arm.com>
24598         * fold-const.c (fold_binary_loc): Check the argument types
24599         rather than the result type when testing for a vector operation.
24601 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
24603         * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
24604         * doc/tm.texi: Regenerate.
24606 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
24607             Alan Hayward  <alan.hayward@arm.com>
24608             David Sherwood  <david.sherwood@arm.com>
24610         * doc/invoke.texi (-msve-vector-bits=): Document new option.
24611         (sve): Document new AArch64 extension.
24612         * doc/md.texi (w): Extend the description of the AArch64
24613         constraint to include SVE vectors.
24614         (Upl, Upa): Document new AArch64 predicate constraints.
24615         * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
24616         enum.
24617         * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
24618         (msve-vector-bits=): New option.
24619         * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
24620         SVE when these are disabled.
24621         (sve): New extension.
24622         * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
24623         modes.  Adjust their number of units based on aarch64_sve_vg.
24624         (MAX_BITSIZE_MODE_ANY_MODE): Define.
24625         * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
24626         aarch64_addr_query_type.
24627         (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
24628         (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
24629         (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
24630         (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
24631         (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
24632         (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
24633         (aarch64_simd_imm_zero_p): Delete.
24634         (aarch64_check_zero_based_sve_index_immediate): Declare.
24635         (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
24636         (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
24637         (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
24638         (aarch64_sve_float_mul_immediate_p): Likewise.
24639         (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
24640         rather than an rtx.
24641         (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
24642         (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
24643         (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
24644         (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
24645         (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
24646         (aarch64_regmode_natural_size): Likewise.
24647         * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
24648         (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
24649         left one place.
24650         (AARCH64_ISA_SVE, TARGET_SVE): New macros.
24651         (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
24652         for VG and the SVE predicate registers.
24653         (V_ALIASES): Add a "z"-prefixed alias.
24654         (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
24655         (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
24656         (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
24657         (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
24658         (REG_CLASS_NAMES): Add entries for them.
24659         (REG_CLASS_CONTENTS): Likewise.  Update ALL_REGS to include VG
24660         and the predicate registers.
24661         (aarch64_sve_vg): Declare.
24662         (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
24663         (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
24664         (REGMODE_NATURAL_SIZE): Define.
24665         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
24666         SVE macros.
24667         * config/aarch64/aarch64.c: Include cfgrtl.h.
24668         (simd_immediate_info): Add a constructor for series vectors,
24669         and an associated step field.
24670         (aarch64_sve_vg): New variable.
24671         (aarch64_dbx_register_number): Handle VG and the predicate registers.
24672         (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
24673         (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
24674         (VEC_ANY_DATA, VEC_STRUCT): New constants.
24675         (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
24676         (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
24677         (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
24678         (aarch64_get_mask_mode): New functions.
24679         (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
24680         and FP_LO_REGS.  Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
24681         (aarch64_hard_regno_mode_ok): Handle VG.  Also handle the SVE
24682         predicate modes and predicate registers.  Explicitly restrict
24683         GPRs to modes of 16 bytes or smaller.  Only allow FP registers
24684         to store a vector mode if it is recognized by
24685         aarch64_classify_vector_mode.
24686         (aarch64_regmode_natural_size): New function.
24687         (aarch64_hard_regno_caller_save_mode): Return the original mode
24688         for predicates.
24689         (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
24690         (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
24691         (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
24692         (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
24693         functions.
24694         (aarch64_add_offset): Add a temp2 parameter.  Assert that temp1
24695         does not overlap dest if the function is frame-related.  Handle
24696         SVE constants.
24697         (aarch64_split_add_offset): New function.
24698         (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
24699         them aarch64_add_offset.
24700         (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
24701         and update call to aarch64_sub_sp.
24702         (aarch64_add_cfa_expression): New function.
24703         (aarch64_expand_prologue): Pass extra temporary registers to the
24704         functions above.  Handle the case in which we need to emit new
24705         DW_CFA_expressions for registers that were originally saved
24706         relative to the stack pointer, but now have to be expressed
24707         relative to the frame pointer.
24708         (aarch64_output_mi_thunk): Pass extra temporary registers to the
24709         functions above.
24710         (aarch64_expand_epilogue): Likewise.  Prevent inheritance of
24711         IP0 and IP1 values for SVE frames.
24712         (aarch64_expand_vec_series): New function.
24713         (aarch64_expand_sve_widened_duplicate): Likewise.
24714         (aarch64_expand_sve_const_vector): Likewise.
24715         (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
24716         Handle SVE constants.  Use emit_move_insn to move a force_const_mem
24717         into the register, rather than emitting a SET directly.
24718         (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
24719         (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
24720         (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
24721         (offset_9bit_signed_scaled_p): New functions.
24722         (aarch64_replicate_bitmask_imm): New function.
24723         (aarch64_bitmask_imm): Use it.
24724         (aarch64_cannot_force_const_mem): Reject expressions involving
24725         a CONST_POLY_INT.  Update call to aarch64_classify_symbol.
24726         (aarch64_classify_index): Handle SVE indices, by requiring
24727         a plain register index with a scale that matches the element size.
24728         (aarch64_classify_address): Handle SVE addresses.  Assert that
24729         the mode of the address is VOIDmode or an integer mode.
24730         Update call to aarch64_classify_symbol.
24731         (aarch64_classify_symbolic_expression): Update call to
24732         aarch64_classify_symbol.
24733         (aarch64_const_vec_all_in_range_p): New function.
24734         (aarch64_print_vector_float_operand): Likewise.
24735         (aarch64_print_operand): Handle 'N' and 'C'.  Use "zN" rather than
24736         "vN" for FP registers with SVE modes.  Handle (const ...) vectors
24737         and the FP immediates 1.0 and 0.5.
24738         (aarch64_print_address_internal): Handle SVE addresses.
24739         (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
24740         (aarch64_regno_regclass): Handle predicate registers.
24741         (aarch64_secondary_reload): Handle big-endian reloads of SVE
24742         data modes.
24743         (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
24744         (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
24745         (aarch64_convert_sve_vector_bits): New function.
24746         (aarch64_override_options): Use it to handle -msve-vector-bits=.
24747         (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
24748         rather than an rtx.
24749         (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
24750         Handle SVE vector and predicate modes.  Accept VL-based constants
24751         that need only one temporary register, and VL offsets that require
24752         no temporary registers.
24753         (aarch64_conditional_register_usage): Mark the predicate registers
24754         as fixed if SVE isn't available.
24755         (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
24756         Return true for SVE vector and predicate modes.
24757         (aarch64_simd_container_mode): Take the number of bits as a poly_int64
24758         rather than an unsigned int.  Handle SVE modes.
24759         (aarch64_preferred_simd_mode): Update call accordingly.  Handle
24760         SVE modes.
24761         (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
24762         if SVE is enabled.
24763         (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
24764         (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
24765         (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
24766         (aarch64_sve_float_mul_immediate_p): New functions.
24767         (aarch64_sve_valid_immediate): New function.
24768         (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
24769         Explicitly reject structure modes.  Check for INDEX constants.
24770         Handle PTRUE and PFALSE constants.
24771         (aarch64_check_zero_based_sve_index_immediate): New function.
24772         (aarch64_simd_imm_zero_p): Delete.
24773         (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
24774         vector modes.  Accept constants in the range of CNT[BHWD].
24775         (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
24776         ask for an Advanced SIMD mode.
24777         (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
24778         (aarch64_simd_vector_alignment): Handle SVE predicates.
24779         (aarch64_vectorize_preferred_vector_alignment): New function.
24780         (aarch64_simd_vector_alignment_reachable): Use it instead of
24781         the vector size.
24782         (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
24783         (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
24784         functions.
24785         (MAX_VECT_LEN): Delete.
24786         (expand_vec_perm_d): Add a vec_flags field.
24787         (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
24788         (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
24789         (aarch64_evpc_ext): Don't apply a big-endian lane correction
24790         for SVE modes.
24791         (aarch64_evpc_rev): Rename to...
24792         (aarch64_evpc_rev_local): ...this.  Use a predicated operation for SVE.
24793         (aarch64_evpc_rev_global): New function.
24794         (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
24795         (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
24796         MAX_VECT_LEN.
24797         (aarch64_evpc_sve_tbl): New function.
24798         (aarch64_expand_vec_perm_const_1): Update after rename of
24799         aarch64_evpc_rev.  Handle SVE permutes too, trying
24800         aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
24801         than aarch64_evpc_tbl.
24802         (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
24803         (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
24804         (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
24805         (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
24806         (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
24807         (aarch64_expand_sve_vcond): New functions.
24808         (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
24809         of aarch64_vector_mode_p.
24810         (aarch64_dwarf_poly_indeterminate_value): New function.
24811         (aarch64_compute_pressure_classes): Likewise.
24812         (aarch64_can_change_mode_class): Likewise.
24813         (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
24814         (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
24815         (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
24816         (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
24817         (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
24818         (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
24819         * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
24820         (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
24821         constraints.
24822         (Dn, Dl, Dr): Accept const as well as const_vector.
24823         (Dz): Likewise.  Compare against CONST0_RTX.
24824         * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
24825         of "vector" where appropriate.
24826         (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
24827         (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
24828         (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
24829         (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
24830         (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
24831         (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
24832         (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
24833         (v_int_equiv): Extend to SVE modes.
24834         (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
24835         mode attributes.
24836         (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
24837         (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
24838         (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
24839         (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
24840         (SVE_COND_FP_CMP): New int iterators.
24841         (perm_hilo): Handle the new unpack unspecs.
24842         (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
24843         attributes.
24844         * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
24845         (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
24846         (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
24847         (aarch64_equality_operator, aarch64_constant_vector_operand)
24848         (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
24849         (aarch64_sve_nonimmediate_operand): Likewise.
24850         (aarch64_sve_general_operand): Likewise.
24851         (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
24852         (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
24853         (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
24854         (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
24855         (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
24856         (aarch64_sve_float_arith_immediate): Likewise.
24857         (aarch64_sve_float_arith_with_sub_immediate): Likewise.
24858         (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
24859         (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
24860         (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
24861         (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
24862         (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
24863         (aarch64_sve_float_arith_operand): Likewise.
24864         (aarch64_sve_float_arith_with_sub_operand): Likewise.
24865         (aarch64_sve_float_mul_operand): Likewise.
24866         (aarch64_sve_vec_perm_operand): Likewise.
24867         (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
24868         (aarch64_mov_operand): Accept const_poly_int and const_vector.
24869         (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
24870         as well as const_vector.
24871         (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
24872         in file.  Use CONST0_RTX and CONSTM1_RTX.
24873         (aarch64_simd_or_scalar_imm_zero): Likewise.  Add match_codes.
24874         (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
24875         Use aarch64_simd_imm_zero.
24876         * config/aarch64/aarch64-sve.md: New file.
24877         * config/aarch64/aarch64.md: Include it.
24878         (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
24879         (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
24880         (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
24881         (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
24882         (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
24883         (sve): New attribute.
24884         (enabled): Disable instructions with the sve attribute unless
24885         TARGET_SVE.
24886         (movqi, movhi): Pass CONST_POLY_INT operaneds through
24887         aarch64_expand_mov_immediate.
24888         (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
24889         CNT[BHSD] immediates.
24890         (movti): Split CONST_POLY_INT moves into two halves.
24891         (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
24892         Split additions that need a temporary here if the destination
24893         is the stack pointer.
24894         (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
24895         (*add<mode>3_poly_1): New instruction.
24896         (set_clobber_cc): New expander.
24898 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
24900         * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
24901         parameter and use it instead of GET_MODE_SIZE (innermode).  Use
24902         inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
24903         Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
24904         GET_MODE_NUNITS (innermode).  Also add a first_elem parameter.
24905         Change innermode from fixed_mode_size to machine_mode.
24906         (simplify_subreg): Update call accordingly.  Handle a constant-sized
24907         subreg of a variable-length CONST_VECTOR.
24909 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
24910             Alan Hayward  <alan.hayward@arm.com>
24911             David Sherwood  <david.sherwood@arm.com>
24913         * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
24914         (add_offset_to_base): New function, split out from...
24915         (create_mem_ref): ...here.  When handling a scale other than 1,
24916         check first whether the address is valid without the offset.
24917         Add it into the base if so, leaving the index and scale as-is.
24919 2018-01-12  Jakub Jelinek  <jakub@redhat.com>
24921         PR c++/83778
24922         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
24923         fold_for_warn before checking if arg2 is INTEGER_CST.
24925 2018-01-12  Segher Boessenkool  <segher@kernel.crashing.org>
24927         * config/rs6000/predicates.md (load_multiple_operation): Delete.
24928         (store_multiple_operation): Delete.
24929         * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
24930         * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
24931         * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
24932         guarded by TARGET_STRING.
24933         (rs6000_output_load_multiple): Delete.
24934         * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
24935         OPTION_MASK_STRING / TARGET_STRING handling.
24936         (print_operand) <'N', 'O'>: Add comment that these are unused now.
24937         (const rs6000_opt_masks) <"string">: Change mask to 0.
24938         * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
24939         (MASK_STRING): Delete.
24940         * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
24941         parts.  Simplify.
24942         (load_multiple): Delete.
24943         (*ldmsi8): Delete.
24944         (*ldmsi7): Delete.
24945         (*ldmsi6): Delete.
24946         (*ldmsi5): Delete.
24947         (*ldmsi4): Delete.
24948         (*ldmsi3): Delete.
24949         (store_multiple): Delete.
24950         (*stmsi8): Delete.
24951         (*stmsi7): Delete.
24952         (*stmsi6): Delete.
24953         (*stmsi5): Delete.
24954         (*stmsi4): Delete.
24955         (*stmsi3): Delete.
24956         (movmemsi_8reg): Delete.
24957         (corresponding unnamed define_insn): Delete.
24958         (movmemsi_6reg): Delete.
24959         (corresponding unnamed define_insn): Delete.
24960         (movmemsi_4reg): Delete.
24961         (corresponding unnamed define_insn): Delete.
24962         (movmemsi_2reg): Delete.
24963         (corresponding unnamed define_insn): Delete.
24964         (movmemsi_1reg): Delete.
24965         (corresponding unnamed define_insn): Delete.
24966         * config/rs6000/rs6000.opt (mno-string): New.
24967         (mstring): Replace by deprecation warning stub.
24968         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
24970 2018-01-12  Jakub Jelinek  <jakub@redhat.com>
24972         * regrename.c (regrename_do_replace): If replacing the same
24973         reg multiple times, try to reuse last created gen_raw_REG.
24975         PR debug/81155
24976         * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
24977         main to workaround a bug in GDB.
24979 2018-01-12  Tom de Vries  <tom@codesourcery.com>
24981         PR target/83737
24982         * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
24984 2018-01-12  Vladimir Makarov  <vmakarov@redhat.com>
24986         PR rtl-optimization/80481
24987         * ira-color.c (get_cap_member): New function.
24988         (allocnos_conflict_by_live_ranges_p): Use it.
24989         (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
24990         (setup_slot_coalesced_allocno_live_ranges): Ditto.
24992 2018-01-12  Uros Bizjak  <ubizjak@gmail.com>
24994         PR target/83628
24995         * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
24996         (*saddl_se_1): Ditto.
24997         (*ssubsi_1): Ditto.
24998         (*ssubl_se_1): Ditto.
25000 2018-01-12  Richard Sandiford  <richard.sandiford@linaro.org>
25002         * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
25003         rather than wi::to_widest for DR_INITs.
25004         * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
25005         wi::to_poly_offset rather than wi::to_offset for DR_INIT.
25006         (vect_analyze_data_ref_accesses): Require both DR_INITs to be
25007         INTEGER_CSTs.
25008         (vect_analyze_group_access_1): Note that here.
25010 2018-01-12  Richard Sandiford  <richard.sandiford@linaro.org>
25012         * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
25013         polynomial type sizes.
25015 2018-01-12  Richard Sandiford  <richard.sandiford@linaro.org>
25017         * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
25018         poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
25019         (gimple_add_tmp_var): Likewise.
25021 2018-01-12  Martin Liska  <mliska@suse.cz>
25023         * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
25024         (gimple_alloc_sizes): Likewise.
25025         (dump_gimple_statistics): Use PRIu64 in printf format.
25026         * gimple.h: Change uint64_t to int.
25028 2018-01-12  Martin Liska  <mliska@suse.cz>
25030         * tree-core.h: Use uint64_t instead of int.
25031         * tree.c (tree_node_counts): Likewise.
25032         (tree_node_sizes): Likewise.
25033         (dump_tree_statistics): Use PRIu64 in printf format.
25035 2018-01-12  Martin Liska  <mliska@suse.cz>
25037         * Makefile.in: As qsort_chk is implemented in vec.c, add
25038         vec.o to linkage of gencfn-macros.
25039         * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
25040         passing the info to record_node_allocation_statistics.
25041         (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
25042         and pass the info.
25043         * ggc-common.c (struct ggc_usage): Add operator== and use
25044         it in operator< and compare function.
25045         * mem-stats.h (struct mem_usage): Likewise.
25046         * vec.c (struct vec_usage): Remove operator< and compare
25047         function. Can be simply inherited.
25049 2018-01-12  Martin Jambor  <mjambor@suse.cz>
25051         PR target/81616
25052         * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
25053         * tree-ssa-math-opts.c: Include domwalk.h.
25054         (convert_mult_to_fma_1): New function.
25055         (fma_transformation_info): New type.
25056         (fma_deferring_state): Likewise.
25057         (cancel_fma_deferring): New function.
25058         (result_of_phi): Likewise.
25059         (last_fma_candidate_feeds_initial_phi): Likewise.
25060         (convert_mult_to_fma): Added deferring logic, split actual
25061         transformation to convert_mult_to_fma_1.
25062         (math_opts_dom_walker): New type.
25063         (math_opts_dom_walker::after_dom_children): New method, body moved
25064         here from pass_optimize_widening_mul::execute, added deferring logic
25065         bits.
25066         (pass_optimize_widening_mul::execute): Moved most of code to
25067         math_opts_dom_walker::after_dom_children.
25068         * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
25069         * config/i386/i386.c (ix86_option_override_internal): Added
25070         maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
25072 2018-01-12  Richard Biener  <rguenther@suse.de>
25074         PR debug/83157
25075         * dwarf2out.c (gen_variable_die): Do not reset old_die for
25076         inline instance vars.
25078 2018-01-12  Oleg Endo  <olegendo@gcc.gnu.org>
25080         PR target/81819
25081         * config/rx/rx.c (rx_is_restricted_memory_address):
25082         Handle SUBREG case.
25084 2018-01-12  Richard Biener  <rguenther@suse.de>
25086         PR tree-optimization/80846
25087         * target.def (split_reduction): New target hook.
25088         * targhooks.c (default_split_reduction): New function.
25089         * targhooks.h (default_split_reduction): Declare.
25090         * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
25091         target requests first reduce vectors by combining low and high
25092         parts.
25093         * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
25094         (get_vectype_for_scalar_type_and_size): Export.
25095         * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
25096         * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
25097         * doc/tm.texi: Regenerate.
25098         * config/i386/i386.c (ix86_split_reduction): Implement
25099         TARGET_VECTORIZE_SPLIT_REDUCTION.
25101 2018-01-12  Eric Botcazou  <ebotcazou@adacore.com>
25103         PR target/83368
25104         * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
25105         in PIC mode except for TARGET_VXWORKS_RTP.
25106         * config/sparc/sparc.c: Include cfgrtl.h.
25107         (TARGET_INIT_PIC_REG): Define.
25108         (TARGET_USE_PSEUDO_PIC_REG): Likewise.
25109         (sparc_pic_register_p): New predicate.
25110         (sparc_legitimate_address_p): Use it.
25111         (sparc_legitimize_pic_address): Likewise.
25112         (sparc_delegitimize_address): Likewise.
25113         (sparc_mode_dependent_address_p): Likewise.
25114         (gen_load_pcrel_sym): Remove 4th parameter.
25115         (load_got_register): Adjust call to above.  Remove obsolete stuff.
25116         (sparc_expand_prologue): Do not call load_got_register here.
25117         (sparc_flat_expand_prologue): Likewise.
25118         (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
25119         (sparc_use_pseudo_pic_reg): New function.
25120         (sparc_init_pic_reg): Likewise.
25121         * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
25122         (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
25124 2018-01-12  Christophe Lyon  <christophe.lyon@linaro.org>
25126         * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
25127         Add item for branch_cost.
25129 2018-01-12  Eric Botcazou  <ebotcazou@adacore.com>
25131         PR rtl-optimization/83565
25132         * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
25133         not extend the result to a larger mode for rotate operations.
25134         (num_sign_bit_copies1): Likewise.
25136 2018-01-12  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
25138         PR target/40411
25139         * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
25140         -symbolic.
25141         Use values-Xc.o for -pedantic.
25142         Link with values-xpg4.o for C90, values-xpg6.o otherwise.
25144 2018-01-12  Martin Liska  <mliska@suse.cz>
25146         PR ipa/83054
25147         * ipa-devirt.c (final_warning_record::grow_type_warnings):
25148         New function.
25149         (possible_polymorphic_call_targets): Use it.
25150         (ipa_devirt): Likewise.
25152 2018-01-12  Martin Liska  <mliska@suse.cz>
25154         * profile-count.h (enum profile_quality): Use 0 as invalid
25155         enum value of profile_quality.
25157 2018-01-12  Chung-Ju Wu  <jasonwucj@gmail.com>
25159         * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
25160         -mext-string options.
25162 2018-01-12  Richard Biener  <rguenther@suse.de>
25164         * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
25165         DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
25166         * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
25167         Likewise.
25168         * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
25170 2018-01-11  Michael Meissner  <meissner@linux.vnet.ibm.com>
25172         * configure.ac (--with-long-double-format): Add support for the
25173         configuration option to change the default long double format on
25174         PowerPC systems.
25175         * config.gcc (powerpc*-linux*-*): Likewise.
25176         * configure: Regenerate.
25177         * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
25178         double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
25179         used without modification.
25181 2018-01-11  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
25183         * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
25184         (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
25185         * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
25186         MISC_BUILTIN_SPEC_BARRIER.
25187         (rs6000_init_builtins): Likewise.
25188         * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
25189         enum value.
25190         (speculation_barrier): New define_insn.
25191         * doc/extend.texi: Document __builtin_speculation_barrier.
25193 2018-01-11  Jakub Jelinek  <jakub@redhat.com>
25195         PR target/83203
25196         * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
25197         is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
25198         * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
25199         iterators.
25200         (ssescalarmodesuffix): Add 512-bit vectors.  Use "d" or "q" for
25201         integral modes instead of "ss" and "sd".
25202         (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
25203         vectors with 32-bit and 64-bit elements.
25204         (vecdupssescalarmodesuffix): New mode attribute.
25205         (vec_dup<mode>): Use it.
25207 2018-01-11  H.J. Lu  <hongjiu.lu@intel.com>
25209         PR target/83330
25210         * config/i386/i386.c (ix86_compute_frame_layout): Align stack
25211         frame if argument is passed on stack.
25213 2018-01-11  Jakub Jelinek  <jakub@redhat.com>
25215         PR target/82682
25216         * ree.c (combine_reaching_defs): Optimize also
25217         reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
25218         reg2=any_extend(exp); reg1=reg2;, formatting fix.
25220 2018-01-11  Jan Hubicka  <hubicka@ucw.cz>
25222         PR middle-end/83189
25223         * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
25225 2018-01-11  Jan Hubicka  <hubicka@ucw.cz>
25227         PR middle-end/83718
25228         * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
25229         after they are computed.
25231 2018-01-11  Bin Cheng  <bin.cheng@arm.com>
25233         PR tree-optimization/83695
25234         * gimple-loop-linterchange.cc
25235         (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
25236         reset cached scev information after interchange.
25237         (pass_linterchange::execute): Remove call to scev_reset_htab.
25239 2018-01-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
25241         * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
25242         vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
25243         vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
25244         vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
25245         vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
25246         vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
25247         * config/arm/arm_neon_builtins.def (vfmal_lane_low,
25248         vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
25249         vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
25250         vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
25251         vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
25252         * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
25253         (V_lane_reg): Likewise.
25254         * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
25255         New define_expand.
25256         (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
25257         (vfmal_lane_low<mode>_intrinsic,
25258         vfmal_lane_low<vfmlsel2><mode>_intrinsic,
25259         vfmal_lane_high<vfmlsel2><mode>_intrinsic,
25260         vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
25261         vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
25262         vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
25263         vfmsl_lane_high<mode>_intrinsic): New define_insns.
25265 2018-01-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
25267         * config/arm/arm-cpus.in (fp16fml): New feature.
25268         (ALL_SIMD): Add fp16fml.
25269         (armv8.2-a): Add fp16fml as an option.
25270         (armv8.3-a): Likewise.
25271         (armv8.4-a): Add fp16fml as part of fp16.
25272         * config/arm/arm.h (TARGET_FP16FML): Define.
25273         * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
25274         when appropriate.
25275         * config/arm/arm-modes.def (V2HF): Define.
25276         * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
25277         vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
25278         vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
25279         * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
25280         vfmsl_low, vfmsl_high): New set of builtins.
25281         * config/arm/iterators.md (PLUSMINUS): New code iterator.
25282         (vfml_op): New code attribute.
25283         (VFMLHALVES): New int iterator.
25284         (VFML, VFMLSEL): New mode attributes.
25285         (V_reg): Define mapping for V2HF.
25286         (V_hi, V_lo): New mode attributes.
25287         (VF_constraint): Likewise.
25288         (vfml_half, vfml_half_selector): New int attributes.
25289         * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
25290         define_expand.
25291         (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
25292         vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
25293         New define_insn.
25294         * config/arm/t-arm-elf (v8_fps): Add fp16fml.
25295         * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
25296         * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
25297         * doc/invoke.texi (ARM Options): Document fp16fml.  Update armv8.4-a
25298         documentation.
25299         * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
25300         Document new effective target and option set.
25302 2018-01-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
25304         * config/arm/arm-cpus.in (armv8_4): New feature.
25305         (ARMv8_4a): New fgroup.
25306         (armv8.4-a): New arch.
25307         * config/arm/arm-tables.opt: Regenerate.
25308         * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
25309         * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
25310         * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
25311         Add matching rules for -march=armv8.4-a and extensions.
25312         * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
25314 2018-01-11  Oleg Endo  <olegendo@gcc.gnu.org>
25316         PR target/81821
25317         * config/rx/rx.md (BW): New mode attribute.
25318         (sync_lock_test_and_setsi): Add mode suffix to insn output.
25320 2018-01-11  Richard Biener  <rguenther@suse.de>
25322         PR tree-optimization/83435
25323         * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
25324         * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
25325         * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
25327 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
25328             Alan Hayward  <alan.hayward@arm.com>
25329             David Sherwood  <david.sherwood@arm.com>
25331         * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
25332         field.
25333         (aarch64_classify_address): Initialize it.  Track polynomial offsets.
25334         (aarch64_print_address_internal): Use it to check for a zero offset.
25336 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
25337             Alan Hayward  <alan.hayward@arm.com>
25338             David Sherwood  <david.sherwood@arm.com>
25340         * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
25341         * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
25342         Return a poly_int64 rather than a HOST_WIDE_INT.
25343         (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
25344         rather than a HOST_WIDE_INT.
25345         * config/aarch64/aarch64.h (aarch64_frame): Protect with
25346         HAVE_POLY_INT_H rather than HOST_WIDE_INT.  Change locals_offset,
25347         hard_fp_offset, frame_size, initial_adjust, callee_offset and
25348         final_offset from HOST_WIDE_INT to poly_int64.
25349         * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
25350         to_constant when getting the number of units in an Advanced SIMD
25351         mode.
25352         (aarch64_builtin_vectorized_function): Check for a constant number
25353         of units.
25354         * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
25355         GET_MODE_SIZE.
25356         (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
25357         attribute instead of GET_MODE_NUNITS.
25358         * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
25359         (aarch64_class_max_nregs): Use the constant_lowest_bound of the
25360         GET_MODE_SIZE for fixed-size registers.
25361         (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
25362         (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
25363         (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
25364         (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
25365         (aarch64_print_operand, aarch64_print_address_internal)
25366         (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
25367         (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
25368         (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
25369         Handle polynomial GET_MODE_SIZE.
25370         (aarch64_hard_regno_caller_save_mode): Likewise.  Return modes
25371         wider than SImode without modification.
25372         (tls_symbolic_operand_type): Use strip_offset instead of split_const.
25373         (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
25374         (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
25375         passing and returning SVE modes.
25376         (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
25377         rather than GEN_INT.
25378         (aarch64_emit_probe_stack_range): Take the size as a poly_int64
25379         rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
25380         (aarch64_allocate_and_probe_stack_space): Likewise.
25381         (aarch64_layout_frame): Cope with polynomial offsets.
25382         (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
25383         start_offset as a poly_int64 rather than a HOST_WIDE_INT.  Track
25384         polynomial offsets.
25385         (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
25386         (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
25387         poly_int64 rather than a HOST_WIDE_INT.
25388         (aarch64_get_separate_components, aarch64_process_components)
25389         (aarch64_expand_prologue, aarch64_expand_epilogue)
25390         (aarch64_use_return_insn_p): Handle polynomial frame offsets.
25391         (aarch64_anchor_offset): New function, split out from...
25392         (aarch64_legitimize_address): ...here.
25393         (aarch64_builtin_vectorization_cost): Handle polynomial
25394         TYPE_VECTOR_SUBPARTS.
25395         (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
25396         GET_MODE_NUNITS.
25397         (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
25398         number of elements from the PARALLEL rather than the mode.
25399         (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
25400         rather than GET_MODE_BITSIZE.
25401         (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
25402         (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
25403         (aarch64_expand_vec_perm_const_1): Handle polynomial
25404         d->perm.length () and d->perm elements.
25405         (aarch64_evpc_tbl): Likewise.  Use nelt rather than GET_MODE_NUNITS.
25406         Apply to_constant to d->perm elements.
25407         (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
25408         polynomial CONST_VECTOR_NUNITS.
25409         (aarch64_move_pointer): Take amount as a poly_int64 rather
25410         than an int.
25411         (aarch64_progress_pointer): Avoid temporary variable.
25412         * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
25413         the mode attribute instead of GET_MODE.
25415 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
25416             Alan Hayward  <alan.hayward@arm.com>
25417             David Sherwood  <david.sherwood@arm.com>
25419         * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
25420         x exists before using it.
25421         (aarch64_add_constant_internal): Rename to...
25422         (aarch64_add_offset_1): ...this.  Replace regnum with separate
25423         src and dest rtxes.  Handle the case in which they're different,
25424         including when the offset is zero.  Replace scratchreg with an rtx.
25425         Use 2 additions if there is no spare register into which we can
25426         move a 16-bit constant.
25427         (aarch64_add_constant): Delete.
25428         (aarch64_add_offset): Replace reg with separate src and dest
25429         rtxes.  Take a poly_int64 offset instead of a HOST_WIDE_INT.
25430         Use aarch64_add_offset_1.
25431         (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
25432         an rtx rather than an int.  Take the delta as a poly_int64
25433         rather than a HOST_WIDE_INT.  Use aarch64_add_offset.
25434         (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
25435         (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
25436         aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
25437         (aarch64_expand_epilogue): Update calls to aarch64_add_offset
25438         and aarch64_add_sp.
25439         (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
25440         aarch64_add_constant.
25442 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
25444         * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
25445         Use scalar_float_mode.
25447 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
25449         * config/aarch64/aarch64-simd.md
25450         (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
25451         (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
25452         (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
25453         (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
25454         (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
25455         (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
25456         (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
25457         (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
25458         (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
25459         (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
25461 2018-01-11  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
25463         PR target/83514
25464         * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
25465         targ_options->x_arm_arch_string is non NULL.
25467 2018-01-11  Tamar Christina  <tamar.christina@arm.com>
25469         * config/aarch64/aarch64.h
25470         (AARCH64_FL_FOR_ARCH8_4): Add  AARCH64_FL_DOTPROD.
25472 2018-01-11  Sudakshina Das  <sudi.das@arm.com>
25474         PR target/82096
25475         * expmed.c (emit_store_flag_force): Swap if const op0
25476         and change VOIDmode to mode of op0.
25478 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
25480         PR rtl-optimization/83761
25481         * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
25482         than bytes to mode_for_size.
25484 2018-01-10  Jan Hubicka  <hubicka@ucw.cz>
25486         PR middle-end/83189
25487         * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
25488         * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
25489         profile.
25491 2018-01-10  Jan Hubicka  <hubicka@ucw.cz>
25493         PR middle-end/83575
25494         * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
25495         when in layout mode.
25496         (cfg_layout_finalize): Do not verify cfg before we are out of layout.
25497         * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
25498         partition fixup.
25500 2018-01-10  Michael Collison  <michael.collison@arm.com>
25502         * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
25503         * config/aarch64/aarch64-option-extension.def: Add
25504         AARCH64_OPT_EXTENSION of 'fp16fml'.
25505         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
25506         (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
25507         * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
25508         * config/aarch64/constraints.md (Ui7): New constraint.
25509         * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
25510         (VFMLA_SEL_W): Ditto.
25511         (f16quad): Ditto.
25512         (f16mac1): Ditto.
25513         (VFMLA16_LOW): New int iterator.
25514         (VFMLA16_HIGH): Ditto.
25515         (UNSPEC_FMLAL): New unspec.
25516         (UNSPEC_FMLSL): Ditto.
25517         (UNSPEC_FMLAL2): Ditto.
25518         (UNSPEC_FMLSL2): Ditto.
25519         (f16mac): New code attribute.
25520         * config/aarch64/aarch64-simd-builtins.def
25521         (aarch64_fmlal_lowv2sf): Ditto.
25522         (aarch64_fmlsl_lowv2sf): Ditto.
25523         (aarch64_fmlalq_lowv4sf): Ditto.
25524         (aarch64_fmlslq_lowv4sf): Ditto.
25525         (aarch64_fmlal_highv2sf): Ditto.
25526         (aarch64_fmlsl_highv2sf): Ditto.
25527         (aarch64_fmlalq_highv4sf): Ditto.
25528         (aarch64_fmlslq_highv4sf): Ditto.
25529         (aarch64_fmlal_lane_lowv2sf): Ditto.
25530         (aarch64_fmlsl_lane_lowv2sf): Ditto.
25531         (aarch64_fmlal_laneq_lowv2sf): Ditto.
25532         (aarch64_fmlsl_laneq_lowv2sf): Ditto.
25533         (aarch64_fmlalq_lane_lowv4sf): Ditto.
25534         (aarch64_fmlsl_lane_lowv4sf): Ditto.
25535         (aarch64_fmlalq_laneq_lowv4sf): Ditto.
25536         (aarch64_fmlsl_laneq_lowv4sf): Ditto.
25537         (aarch64_fmlal_lane_highv2sf): Ditto.
25538         (aarch64_fmlsl_lane_highv2sf): Ditto.
25539         (aarch64_fmlal_laneq_highv2sf): Ditto.
25540         (aarch64_fmlsl_laneq_highv2sf): Ditto.
25541         (aarch64_fmlalq_lane_highv4sf): Ditto.
25542         (aarch64_fmlsl_lane_highv4sf): Ditto.
25543         (aarch64_fmlalq_laneq_highv4sf): Ditto.
25544         (aarch64_fmlsl_laneq_highv4sf): Ditto.
25545         * config/aarch64/aarch64-simd.md:
25546         (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
25547         (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
25548         (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
25549         (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
25550         (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
25551         (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
25552         (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
25553         (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
25554         (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
25555         (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
25556         (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
25557         (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
25558         (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
25559         (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
25560         (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
25561         (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
25562         (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
25563         (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
25564         (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
25565         (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
25566         * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
25567         (vfmlsl_low_u32): Ditto.
25568         (vfmlalq_low_u32): Ditto.
25569         (vfmlslq_low_u32): Ditto.
25570         (vfmlal_high_u32): Ditto.
25571         (vfmlsl_high_u32): Ditto.
25572         (vfmlalq_high_u32): Ditto.
25573         (vfmlslq_high_u32): Ditto.
25574         (vfmlal_lane_low_u32): Ditto.
25575         (vfmlsl_lane_low_u32): Ditto.
25576         (vfmlal_laneq_low_u32): Ditto.
25577         (vfmlsl_laneq_low_u32): Ditto.
25578         (vfmlalq_lane_low_u32): Ditto.
25579         (vfmlslq_lane_low_u32): Ditto.
25580         (vfmlalq_laneq_low_u32): Ditto.
25581         (vfmlslq_laneq_low_u32): Ditto.
25582         (vfmlal_lane_high_u32): Ditto.
25583         (vfmlsl_lane_high_u32): Ditto.
25584         (vfmlal_laneq_high_u32): Ditto.
25585         (vfmlsl_laneq_high_u32): Ditto.
25586         (vfmlalq_lane_high_u32): Ditto.
25587         (vfmlslq_lane_high_u32): Ditto.
25588         (vfmlalq_laneq_high_u32): Ditto.
25589         (vfmlslq_laneq_high_u32): Ditto.
25590         * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
25591         (AARCH64_FL_FOR_ARCH8_4): New.
25592         (AARCH64_ISA_F16FML): New ISA flag.
25593         (TARGET_F16FML): New feature flag for fp16fml.
25594         (doc/invoke.texi): Document new fp16fml option.
25596 2018-01-10  Michael Collison  <michael.collison@arm.com>
25598         * config/aarch64/aarch64-builtins.c:
25599         (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
25600         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
25601         (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
25602         * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
25603         (AARCH64_ISA_SHA3): New ISA flag.
25604         (TARGET_SHA3): New feature flag for sha3.
25605         * config/aarch64/iterators.md (sha512_op): New int attribute.
25606         (CRYPTO_SHA512): New int iterator.
25607         (UNSPEC_SHA512H): New unspec.
25608         (UNSPEC_SHA512H2): Ditto.
25609         (UNSPEC_SHA512SU0): Ditto.
25610         (UNSPEC_SHA512SU1): Ditto.
25611         * config/aarch64/aarch64-simd-builtins.def
25612         (aarch64_crypto_sha512hqv2di): New builtin.
25613         (aarch64_crypto_sha512h2qv2di): Ditto.
25614         (aarch64_crypto_sha512su0qv2di): Ditto.
25615         (aarch64_crypto_sha512su1qv2di): Ditto.
25616         (aarch64_eor3qv8hi): Ditto.
25617         (aarch64_rax1qv2di): Ditto.
25618         (aarch64_xarqv2di): Ditto.
25619         (aarch64_bcaxqv8hi): Ditto.
25620         * config/aarch64/aarch64-simd.md:
25621         (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
25622         (aarch64_crypto_sha512su0qv2di): Ditto.
25623         (aarch64_crypto_sha512su1qv2di): Ditto.
25624         (aarch64_eor3qv8hi): Ditto.
25625         (aarch64_rax1qv2di): Ditto.
25626         (aarch64_xarqv2di): Ditto.
25627         (aarch64_bcaxqv8hi): Ditto.
25628         * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
25629         (vsha512h2q_u64): Ditto.
25630         (vsha512su0q_u64): Ditto.
25631         (vsha512su1q_u64): Ditto.
25632         (veor3q_u16): Ditto.
25633         (vrax1q_u64): Ditto.
25634         (vxarq_u64): Ditto.
25635         (vbcaxq_u16): Ditto.
25636         * config/arm/types.md (crypto_sha512): New type attribute.
25637         (crypto_sha3): Ditto.
25638         (doc/invoke.texi): Document new sha3 option.
25640 2018-01-10  Michael Collison  <michael.collison@arm.com>
25642         * config/aarch64/aarch64-builtins.c:
25643         (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
25644         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
25645         (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
25646         (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
25647         * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
25648         (AARCH64_ISA_SM4): New ISA flag.
25649         (TARGET_SM4): New feature flag for sm4.
25650         * config/aarch64/aarch64-simd-builtins.def
25651         (aarch64_sm3ss1qv4si): Ditto.
25652         (aarch64_sm3tt1aq4si): Ditto.
25653         (aarch64_sm3tt1bq4si): Ditto.
25654         (aarch64_sm3tt2aq4si): Ditto.
25655         (aarch64_sm3tt2bq4si): Ditto.
25656         (aarch64_sm3partw1qv4si): Ditto.
25657         (aarch64_sm3partw2qv4si): Ditto.
25658         (aarch64_sm4eqv4si): Ditto.
25659         (aarch64_sm4ekeyqv4si): Ditto.
25660         * config/aarch64/aarch64-simd.md:
25661         (aarch64_sm3ss1qv4si): Ditto.
25662         (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
25663         (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
25664         (aarch64_sm4eqv4si): Ditto.
25665         (aarch64_sm4ekeyqv4si): Ditto.
25666         * config/aarch64/iterators.md (sm3tt_op): New int iterator.
25667         (sm3part_op): Ditto.
25668         (CRYPTO_SM3TT): Ditto.
25669         (CRYPTO_SM3PART): Ditto.
25670         (UNSPEC_SM3SS1): New unspec.
25671         (UNSPEC_SM3TT1A): Ditto.
25672         (UNSPEC_SM3TT1B): Ditto.
25673         (UNSPEC_SM3TT2A): Ditto.
25674         (UNSPEC_SM3TT2B): Ditto.
25675         (UNSPEC_SM3PARTW1): Ditto.
25676         (UNSPEC_SM3PARTW2): Ditto.
25677         (UNSPEC_SM4E): Ditto.
25678         (UNSPEC_SM4EKEY): Ditto.
25679         * config/aarch64/constraints.md (Ui2): New constraint.
25680         * config/aarch64/predicates.md (aarch64_imm2): New predicate.
25681         * config/arm/types.md (crypto_sm3): New type attribute.
25682         (crypto_sm4): Ditto.
25683         * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
25684         (vsm3tt1aq_u32): Ditto.
25685         (vsm3tt1bq_u32): Ditto.
25686         (vsm3tt2aq_u32): Ditto.
25687         (vsm3tt2bq_u32): Ditto.
25688         (vsm3partw1q_u32): Ditto.
25689         (vsm3partw2q_u32): Ditto.
25690         (vsm4eq_u32): Ditto.
25691         (vsm4ekeyq_u32): Ditto.
25692         (doc/invoke.texi): Document new sm4 option.
25694 2018-01-10  Michael Collison  <michael.collison@arm.com>
25696         * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
25697         * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
25698         (AARCH64_FL_FOR_ARCH8_4): New.
25699         (AARCH64_FL_V8_4): New flag.
25700         (doc/invoke.texi): Document new armv8.4-a option.
25702 2018-01-10  Michael Collison  <michael.collison@arm.com>
25704         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
25705         (__ARM_FEATURE_AES): Define if TARGET_AES is true.
25706         (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
25707         * config/aarch64/aarch64-option-extension.def: Add
25708         AARCH64_OPT_EXTENSION of 'sha2'.
25709         (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
25710         (crypto): Disable sha2 and aes if crypto disabled.
25711         (crypto): Enable aes and sha2 if enabled.
25712         (simd): Disable sha2 and aes if simd disabled.
25713         * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
25714         New flags.
25715         (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
25716         (TARGET_SHA2): New feature flag for sha2.
25717         (TARGET_AES): New feature flag for aes.
25718         * config/aarch64/aarch64-simd.md:
25719         (aarch64_crypto_aes<aes_op>v16qi): Make pattern
25720         conditional on TARGET_AES.
25721         (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
25722         (aarch64_crypto_sha1hsi): Make pattern conditional
25723         on TARGET_SHA2.
25724         (aarch64_crypto_sha1hv4si): Ditto.
25725         (aarch64_be_crypto_sha1hv4si): Ditto.
25726         (aarch64_crypto_sha1su1v4si): Ditto.
25727         (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
25728         (aarch64_crypto_sha1su0v4si): Ditto.
25729         (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
25730         (aarch64_crypto_sha256su0v4si): Ditto.
25731         (aarch64_crypto_sha256su1v4si): Ditto.
25732         (doc/invoke.texi): Document new aes and sha2 options.
25734 2018-01-10  Martin Sebor  <msebor@redhat.com>
25736         PR tree-optimization/83781
25737         * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
25738         as string arrays.
25740 2018-01-11  Martin Sebor  <msebor@gmail.com>
25741             Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
25743         PR tree-optimization/83501
25744         PR tree-optimization/81703
25746         * tree-ssa-strlen.c (get_string_cst): Rename...
25747         (get_string_len): ...to this.  Handle global constants.
25748         (handle_char_store): Adjust.
25750 2018-01-10  Kito Cheng  <kito.cheng@gmail.com>
25751             Jim Wilson  <jimw@sifive.com>
25753         * config/riscv/riscv-protos.h (riscv_output_return): New.
25754         * config/riscv/riscv.c (struct machine_function): New naked_p field.
25755         (riscv_attribute_table, riscv_output_return),
25756         (riscv_handle_fndecl_attribute, riscv_naked_function_p),
25757         (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
25758         (riscv_compute_frame_info): Only compute frame->mask if not a naked
25759         function.
25760         (riscv_expand_prologue): Add early return for naked function.
25761         (riscv_expand_epilogue): Likewise.
25762         (riscv_function_ok_for_sibcall): Return false for naked function.
25763         (riscv_set_current_function): New.
25764         (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
25765         (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
25766         * config/riscv/riscv.md (simple_return): Call riscv_output_return.
25767         * doc/extend.texi (RISC-V Function Attributes): New.
25769 2018-01-10  Michael Meissner  <meissner@linux.vnet.ibm.com>
25771         * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
25772         check for 128-bit long double before checking TCmode.
25773         * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
25774         128-bit long doubles before checking TFmode or TCmode.
25775         (FLOAT128_IBM_P): Likewise.
25777 2018-01-10  Martin Sebor  <msebor@redhat.com>
25779         PR tree-optimization/83671
25780         * builtins.c (c_strlen): Unconditionally return zero for the empty
25781         string.
25782         Use -Warray-bounds for warnings.
25783         * gimple-fold.c (get_range_strlen): Handle non-constant lengths
25784         for non-constant array indices with COMPONENT_REF, arrays of
25785         arrays, and pointers to arrays.
25786         (gimple_fold_builtin_strlen): Determine and set length range for
25787         non-constant character arrays.
25789 2018-01-10  Aldy Hernandez  <aldyh@redhat.com>
25791         PR middle-end/81897
25792         * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
25793         empty blocks.
25795 2018-01-10  Eric Botcazou  <ebotcazou@adacore.com>
25797         * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
25799 2018-01-10  Peter Bergner  <bergner@vnet.ibm.com>
25801         PR target/83399
25802         * config/rs6000/rs6000.c (print_operand) <'y'>: Use
25803         VECTOR_MEM_ALTIVEC_OR_VSX_P.
25804         * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
25805         indexed_or_indirect_operand predicate.
25806         (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
25807         (*vsx_le_perm_load_v8hi): Likewise.
25808         (*vsx_le_perm_load_v16qi): Likewise.
25809         (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
25810         (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
25811         (*vsx_le_perm_store_v8hi): Likewise.
25812         (*vsx_le_perm_store_v16qi): Likewise.
25813         (eight unnamed splitters): Likewise.
25815 2018-01-10  Peter Bergner  <bergner@vnet.ibm.com>
25817         * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
25818         * config/rs6000/emmintrin.h: Likewise.
25819         * config/rs6000/mmintrin.h: Likewise.
25820         * config/rs6000/xmmintrin.h: Likewise.
25822 2018-01-10  David Malcolm  <dmalcolm@redhat.com>
25824         PR c++/43486
25825         * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
25826         "public_flag".
25827         * tree.c (tree_nop_conversion): Return true for location wrapper
25828         nodes.
25829         (maybe_wrap_with_location): New function.
25830         (selftest::check_strip_nops): New function.
25831         (selftest::test_location_wrappers): New function.
25832         (selftest::tree_c_tests): Call it.
25833         * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
25834         (maybe_wrap_with_location): New decl.
25835         (EXPR_LOCATION_WRAPPER_P): New macro.
25836         (location_wrapper_p): New inline function.
25837         (tree_strip_any_location_wrapper): New inline function.
25839 2018-01-10  H.J. Lu  <hongjiu.lu@intel.com>
25841         PR target/83735
25842         * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
25843         stack_realign_offset for the largest alignment of stack slot
25844         actually used.
25845         (ix86_find_max_used_stack_alignment): New function.
25846         (ix86_finalize_stack_frame_flags): Use it.  Set
25847         max_used_stack_alignment if we don't realign stack.
25848         * config/i386/i386.h (machine_function): Add
25849         max_used_stack_alignment.
25851 2018-01-10  Christophe Lyon  <christophe.lyon@linaro.org>
25853         * config/arm/arm.opt (-mbranch-cost): New option.
25854         * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
25855         account.
25857 2018-01-10  Segher Boessenkool  <segher@kernel.crashing.org>
25859         PR target/83629
25860         * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
25861         load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
25863 2018-01-10  Richard Biener  <rguenther@suse.de>
25865         PR debug/83765
25866         * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
25867         early out so it also covers the case where we have a non-NULL
25868         origin.
25870 2018-01-10  Richard Sandiford  <richard.sandiford@linaro.org>
25872         PR tree-optimization/83753
25873         * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
25874         for non-strided grouped accesses if the number of elements is 1.
25876 2018-01-10  Jan Hubicka  <hubicka@ucw.cz>
25878         PR target/81616
25879         * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
25880         * i386.h (TARGET_USE_GATHER): Define.
25881         * x86-tune.def (X86_TUNE_USE_GATHER): New.
25883 2018-01-10  Martin Liska  <mliska@suse.cz>
25885         PR bootstrap/82831
25886         * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
25887         * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
25888         partitioning.
25889         * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
25890         CLEANUP_NO_PARTITIONING is not set.
25892 2018-01-10  Richard Sandiford  <richard.sandiford@linaro.org>
25894         * doc/rtl.texi: Remove documentation of (const ...) wrappers
25895         for vectors, as a partial revert of r254296.
25896         * rtl.h (const_vec_p): Delete.
25897         (const_vec_duplicate_p): Don't test for vector CONSTs.
25898         (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
25899         * expmed.c (make_tree): Likewise.
25901         Revert:
25902         * common.md (E, F): Use CONSTANT_P instead of checking for
25903         CONST_VECTOR.
25904         * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
25905         checking for CONST_VECTOR.
25907 2018-01-09  Jan Hubicka  <hubicka@ucw.cz>
25909         PR middle-end/83575
25910         * predict.c (force_edge_cold): Handle in more sane way edges
25911         with no prediction.
25913 2018-01-09  Carl Love  <cel@us.ibm.com>
25915         * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
25916         V4SI, V4SF types.
25917         (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
25918         * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
25919         VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
25920         VMRGOW_V2DI, VMRGOW_V2DF.  Remove definition for VMRGOW.
25921         * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
25922         P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW):  Add definitions.
25923         * config/rs6000/rs6000-protos.h: Add extern defition for
25924         rs6000_generate_float2_double_code.
25925         * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
25926         function.
25927         * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
25928         (float2_v2df): Add define_expand.
25930 2018-01-09  Uros Bizjak  <ubizjak@gmail.com>
25932         PR target/83628
25933         * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
25934         op_mode in the force_to_mode call.
25936 2018-01-09  Richard Sandiford  <richard.sandiford@linaro.org>
25938         * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
25939         instead of checking each element individually.
25940         (aarch64_evpc_uzp): Likewise.
25941         (aarch64_evpc_zip): Likewise.
25942         (aarch64_evpc_ext): Likewise.
25943         (aarch64_evpc_rev): Likewise.
25944         (aarch64_evpc_dup): Test the encoding for a single duplicated element,
25945         instead of checking each element individually.  Return true without
25946         generating rtl if
25947         (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
25948         whether all selected elements come from the same input, instead of
25949         checking each element individually.  Remove calls to gen_rtx_REG,
25950         start_sequence and end_sequence and instead assert that no rtl is
25951         generated.
25953 2018-01-09  Richard Sandiford  <richard.sandiford@linaro.org>
25955         * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
25956         order of HIGH and CONST checks.
25958 2018-01-09  Richard Sandiford  <richard.sandiford@linaro.org>
25960         * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
25961         if the destination isn't an SSA_NAME.
25963 2018-01-09  Richard Biener  <rguenther@suse.de>
25965         PR tree-optimization/83668
25966         * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
25967         move prologue...
25968         (canonicalize_loop_form): ... here, renamed from ...
25969         (canonicalize_loop_closed_ssa_form): ... this and amended to
25970         swap successor edges for loop exit blocks to make us use
25971         the RPO order we need for initial schedule generation.
25973 2018-01-09  Joseph Myers  <joseph@codesourcery.com>
25975         PR tree-optimization/64811
25976         * match.pd: When optimizing comparisons with Inf, avoid
25977         introducing or losing exceptions from comparisons with NaN.
25979 2018-01-09  Martin Liska  <mliska@suse.cz>
25981         PR sanitizer/82517
25982         * asan.c (shadow_mem_size): Add gcc_assert.
25984 2018-01-09  Georg-Johann Lay  <avr@gjlay.de>
25986         Don't save registers in main().
25988         PR target/83738
25989         * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
25990         * config/avr/avr.opt (-mmain-is-OS_task): New target option.
25991         * config/avr/avr.c (avr_set_current_function): Don't error if
25992         naked, OS_task or OS_main are specified at the same time.
25993         (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
25994         OS_main.
25995         (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
25996         attribute.
25997         * common/config/avr/avr-common.c (avr_option_optimization_table):
25998         Switch on -mmain-is-OS_task for optimizing compilations.
26000 2018-01-09  Richard Biener  <rguenther@suse.de>
26002         PR tree-optimization/83572
26003         * graphite.c: Include cfganal.h.
26004         (graphite_transform_loops): Connect infinite loops to exit
26005         and remove fake edges at the end.
26007 2018-01-09  Jan Hubicka  <hubicka@ucw.cz>
26009         * ipa-inline.c (edge_badness): Revert accidental checkin.
26011 2018-01-09  Jan Hubicka  <hubicka@ucw.cz>
26013         PR ipa/80763
26014         * ipa-comdats.c (set_comdat_group): Only set comdat group of real
26015         symbols; not inline clones.
26017 2018-01-09  Jakub Jelinek  <jakub@redhat.com>
26019         PR target/83507
26020         * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
26021         hard registers.  Formatting fixes.
26023         PR preprocessor/83722
26024         * gcc.c (try_generate_repro): Pass
26025         &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
26026         &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
26027         do_report_bug.
26029 2018-01-08  Monk Chiang  <sh.chiang04@gmail.com>
26030             Kito Cheng  <kito.cheng@gmail.com>
26032         * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
26033         (riscv_leaf_function_p): Delete.
26034         (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
26036 2018-01-08  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
26038         * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
26039         function.
26040         (do_ifelse): New function.
26041         (do_isel): New function.
26042         (do_sub3): New function.
26043         (do_add3): New function.
26044         (do_load_mask_compare): New function.
26045         (do_overlap_load_compare): New function.
26046         (expand_compare_loop): New function.
26047         (expand_block_compare): Call expand_compare_loop() when appropriate.
26048         * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
26049         option description.
26050         (-mblock-compare-inline-loop-limit): New option.
26052 2018-01-08  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
26054         PR target/83677
26055         * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
26056         Reverse order of second and third operands in first alternative.
26057         * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
26058         of first and second elements in UNSPEC_VPERMR vector.
26059         (altivec_expand_vec_perm_le): Likewise.
26061 2018-01-08  Jeff Law  <law@redhat.com>
26063         PR rtl-optimizatin/81308
26064         * tree-switch-conversion.c (cfg_altered): New file scoped static.
26065         (process_switch): If group_case_labels makes a change, then set
26066         cfg_altered.
26067         (pass_convert_switch::execute): If a switch is converted, then
26068         set cfg_altered.  Return TODO_cfg_cleanup if cfg_altered is true.
26070         PR rtl-optimization/81308
26071         * recog.c (split_all_insns): Conditionally cleanup the CFG after
26072         splitting insns.
26074 2018-01-08  Vidya Praveen  <vidyapraveen@arm.com>
26076         PR target/83663 - Revert r255946
26077         * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
26078         generation for cases where splatting a value is not useful.
26079         * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
26080         across a vec_duplicate and a paradoxical subreg forming a vector
26081         mode to a vec_concat.
26083 2018-01-08  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
26085         * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
26086         -march=armv8.3-a variants.
26087         * config/arm/t-multilib: Likewise.
26088         * config/arm/t-arm-elf: Likewise.  Handle dotprod extension.
26090 2018-01-08  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
26092         * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
26093         to generate rtl.
26094         (cceq_ior_compare_complement): Give it a name so I can use it, and
26095         change boolean_or_operator predicate to boolean_operator so it can
26096         be used to generate a crand.
26097         (eqne): New code iterator.
26098         (bd/bd_neg): New code_attrs.
26099         (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
26100         a single define_insn.
26101         (<bd>tf_<mode>): A new insn pattern for the conditional form branch
26102         decrement (bdnzt/bdnzf/bdzt/bdzf).
26103         * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
26104         with the new names of the branch decrement patterns, and added the
26105         names of the branch decrement conditional patterns.
26107 2018-01-08  Richard Biener  <rguenther@suse.de>
26109         PR tree-optimization/83563
26110         * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
26111         cache.
26113 2018-01-08  Richard Biener  <rguenther@suse.de>
26115         PR middle-end/83713
26116         * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
26118 2018-01-08  Richard Biener  <rguenther@suse.de>
26120         PR tree-optimization/83685
26121         * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
26122         references to abnormals.
26124 2018-01-08  Richard Biener  <rguenther@suse.de>
26126         PR lto/83719
26127         * dwarf2out.c (output_indirect_strings): Handle empty
26128         skeleton_debug_str_hash.
26129         (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
26131 2018-01-08  Claudiu Zissulescu  <claziss@synopsys.com>
26133         * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
26134         (emit_store_direct): Likewise.
26135         (arc_trampoline_adjust_address): Likewise.
26136         (arc_asm_trampoline_template): New function.
26137         (arc_initialize_trampoline): Use asm_trampoline_template.
26138         (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
26139         * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
26140         * config/arc/arc.md (flush_icache): Delete pattern.
26142 2018-01-08  Claudiu Zissulescu  <claziss@synopsys.com>
26144         * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
26145         * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
26146         munaligned-access.
26148 2018-01-08  Sebastian Huber  <sebastian.huber@embedded-brains.de>
26150         PR target/83681
26151         * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
26152         by not USED_FOR_TARGET.
26153         (make_pass_resolve_sw_modes): Likewise.
26155 2018-01-08  Sebastian Huber  <sebastian.huber@embedded-brains.de>
26157         * config/nios2/nios2.h (nios2_section_threshold): Guard by not
26158         USED_FOR_TARGET.
26160 2018-01-08  Richard Biener  <rguenther@suse.de>
26162         PR middle-end/83580
26163         * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
26165 2018-01-08  Richard Biener  <rguenther@suse.de>
26167         PR middle-end/83517
26168         * match.pd ((t * 2) / 2) -> t): Add missing :c.
26170 2018-01-06  Aldy Hernandez  <aldyh@redhat.com>
26172         PR middle-end/81897
26173         * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
26174         basic blocks with a small number of successors.
26175         (convert_control_dep_chain_into_preds): Improve handling of
26176         forwarder blocks.
26177         (dump_predicates): Split apart into...
26178         (dump_pred_chain): ...here...
26179         (dump_pred_info): ...and here.
26180         (can_one_predicate_be_invalidated_p): Add debugging printfs.
26181         (can_chain_union_be_invalidated_p): Improve check for invalidation
26182         of paths.
26183         (uninit_uses_cannot_happen): Avoid unnecessary if
26184         convert_control_dep_chain_into_preds yielded nothing.
26186 2018-01-06  Martin Sebor  <msebor@redhat.com>
26188         PR tree-optimization/83640
26189         * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
26190         subtracting negative offset from size.
26191         (builtin_access::overlap): Adjust offset bounds of the access to fall
26192         within the size of the object if possible.
26194 2018-01-06  Richard Sandiford  <richard.sandiford@linaro.org>
26196         PR rtl-optimization/83699
26197         * expmed.c (extract_bit_field_1): Restrict the vector usage of
26198         extract_bit_field_as_subreg to cases in which the extracted
26199         value is also a vector.
26201         * lra-constraints.c (process_alt_operands): Test for the equivalence
26202         substitutions when detecting a possible reload cycle.
26204 2018-01-06  Jakub Jelinek  <jakub@redhat.com>
26206         PR debug/83480
26207         * toplev.c (process_options): Don't enable debug_nonbind_markers_p
26208         by default if flag_selective_schedling{,2}.  Formatting fixes.
26210         PR rtl-optimization/83682
26211         * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
26212         if it has non-VECTOR_MODE element mode.
26213         (vec_duplicate_p): Likewise.
26215         PR middle-end/83694
26216         * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
26217         and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
26219 2018-01-05  Jakub Jelinek  <jakub@redhat.com>
26221         PR target/83604
26222         * config/i386/i386-builtin.def
26223         (__builtin_ia32_vgf2p8affineinvqb_v64qi,
26224         __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
26225         Require also OPTION_MASK_ISA_AVX512F in addition to
26226         OPTION_MASK_ISA_GFNI.
26227         (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
26228         __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
26229         OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
26230         to OPTION_MASK_ISA_GFNI.
26231         (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
26232         OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
26233         OPTION_MASK_ISA_AVX512BW.
26234         (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
26235         OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
26236         addition to OPTION_MASK_ISA_GFNI.
26237         (__builtin_ia32_vgf2p8affineinvqb_v16qi,
26238         __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
26239         Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
26240         to OPTION_MASK_ISA_GFNI.
26241         * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
26242         a requirement for all ISAs rather than any of them with a few
26243         exceptions.
26244         (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
26245         processing.
26246         (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
26247         bitmasks to be enabled with 3 exceptions, instead of requiring any
26248         enabled ISA with lots of exceptions.
26249         * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
26250         vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
26251         Change avx512bw in isa attribute to avx512f.
26252         * config/i386/sgxintrin.h: Add license boilerplate.
26253         * config/i386/vaesintrin.h: Likewise.  Fix macro spelling __AVX512F
26254         to __AVX512F__ and __AVX512VL to __AVX512VL__.
26255         (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
26256         _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
26257         defined.
26258         * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
26259         _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
26260         temporarily sse2 rather than sse if not enabled already.
26262         PR target/83604
26263         * config/i386/sse.md (VI248_VLBW): Rename to ...
26264         (VI248_AVX512VL): ... this.  Don't guard V32HI with TARGET_AVX512BW.
26265         (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
26266         vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
26267         vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
26268         vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
26269         mode iterator instead of VI248_VLBW.
26271 2018-01-05  Jan Hubicka  <hubicka@ucw.cz>
26273         * ipa-fnsummary.c (record_modified_bb_info): Add OP.
26274         (record_modified): Skip clobbers; add debug output.
26275         (param_change_prob): Use sreal frequencies.
26277 2018-01-05  Richard Sandiford  <richard.sandiford@linaro.org>
26279         * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
26280         punt for user-aligned variables.
26282 2018-01-05  Richard Sandiford  <richard.sandiford@linaro.org>
26284         * tree-chrec.c (chrec_contains_symbols): Return true for
26285         POLY_INT_CST.
26287 2018-01-05  Sudakshina Das  <sudi.das@arm.com>
26289         PR target/82439
26290         * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
26291         of (x|y) == x for BICS pattern.
26293 2018-01-05  Jakub Jelinek  <jakub@redhat.com>
26295         PR tree-optimization/83605
26296         * gimple-ssa-strength-reduction.c: Include tree-eh.h.
26297         (find_candidates_dom_walker::before_dom_children): Ignore stmts that
26298         can throw.
26300 2018-01-05  Sebastian Huber  <sebastian.huber@embedded-brains.de>
26302         * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
26303         * config/epiphany/rtems.h: New file.
26305 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
26306             Uros Bizjak  <ubizjak@gmail.com>
26308         PR target/83554
26309         * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
26310         QIreg_operand instead of register_operand predicate.
26311         * config/i386/i386.c (ix86_rop_should_change_byte_p,
26312         set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
26313         comments instead of -fmitigate[-_]rop.
26315 2018-01-04  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
26317         PR bootstrap/81926
26318         * cgraphunit.c (symbol_table::compile): Switch to text_section
26319         before calling assembly_start debug hook.
26320         * run-rtl-passes.c (run_rtl_passes): Likewise.
26321         Include output.h.
26323 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
26325         * tree-vrp.c (extract_range_from_binary_expr_1): Check
26326         range_int_cst_p rather than !symbolic_range_p before calling
26327         extract_range_from_multiplicative_op_1.
26329 2018-01-04  Jeff Law  <law@redhat.com>
26331         * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
26332         redundant test in assertion.
26334 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
26336         * doc/rtl.texi: Document machine_mode wrapper classes.
26338 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
26340         * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
26341         using tree_to_uhwi.
26343 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
26345         * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
26346         the VEC_PERM_EXPR fold to fail.
26348 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
26350         PR debug/83585
26351         * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
26352         to switched_sections.
26354 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
26356         PR target/83680
26357         * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
26358         test for d.testing.
26360 2018-01-04  Peter Bergner  <bergner@vnet.ibm.com>
26362         PR target/83387
26363         * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
26364         allow arguments in FP registers if TARGET_HARD_FLOAT is false.
26366 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
26368         PR debug/83666
26369         * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
26370         is BLKmode and bitpos not zero or mode change is needed.
26372 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
26374         PR target/83675
26375         * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
26376         TARGET_VIS2.
26378 2018-01-04  Uros Bizjak  <ubizjak@gmail.com>
26380         PR target/83628
26381         * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
26382         instead of MULT rtx.  Update all corresponding splitters.
26383         (*saddl_se): Ditto.
26384         (*ssub<modesuffix>): Ditto.
26385         (*ssubl_se): Ditto.
26386         (*cmp_sadd_di): Update split patterns.
26387         (*cmp_sadd_si): Ditto.
26388         (*cmp_sadd_sidi): Ditto.
26389         (*cmp_ssub_di): Ditto.
26390         (*cmp_ssub_si): Ditto.
26391         (*cmp_ssub_sidi): Ditto.
26392         * config/alpha/predicates.md (const23_operand): New predicate.
26393         * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
26394         Look for ASHIFT, not MULT inner operand.
26395         (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
26397 2018-01-04  Martin Liska  <mliska@suse.cz>
26399         PR gcov-profile/83669
26400         * gcov.c (output_intermediate_file): Add version to intermediate
26401         gcov file.
26402         * doc/gcov.texi: Document new field 'version' in intermediate
26403         file format. Fix location of '-k' option of gcov command.
26405 2018-01-04  Martin Liska  <mliska@suse.cz>
26407         PR ipa/82352
26408         * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
26410 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
26412         * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
26414 2018-01-03  Martin Sebor  <msebor@redhat.com>
26416         PR tree-optimization/83655
26417         * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
26418         checking calls with invalid arguments.
26420 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
26422         * tree-vect-stmts.c (vect_get_store_rhs): New function.
26423         (vectorizable_mask_load_store): Delete.
26424         (vectorizable_call): Return false for masked loads and stores.
26425         (vectorizable_store): Handle IFN_MASK_STORE.  Use vect_get_store_rhs
26426         instead of gimple_assign_rhs1.
26427         (vectorizable_load): Handle IFN_MASK_LOAD.
26428         (vect_transform_stmt): Don't set is_store for call_vec_info_type.
26430 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
26432         * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
26433         split out from..,
26434         (vectorizable_mask_load_store): ...here.
26435         (vectorizable_load): ...and here.
26437 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
26439         * tree-vect-stmts.c (vect_build_all_ones_mask)
26440         (vect_build_zero_merge_argument): New functions, split out from...
26441         (vectorizable_load): ...here.
26443 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
26445         * tree-vect-stmts.c (vect_check_store_rhs): New function,
26446         split out from...
26447         (vectorizable_mask_load_store): ...here.
26448         (vectorizable_store): ...and here.
26450 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
26452         * tree-vect-stmts.c (vect_check_load_store_mask): New function,
26453         split out from...
26454         (vectorizable_mask_load_store): ...here.
26456 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
26458         * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
26459         (vect_model_store_cost): Take a vec_load_store_type instead of a
26460         vect_def_type.
26461         * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
26462         (vect_model_store_cost): Take a vec_load_store_type instead of a
26463         vect_def_type.
26464         (vectorizable_mask_load_store): Update accordingly.
26465         (vectorizable_store): Likewise.
26466         * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
26468 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
26470         * tree-vect-loop.c (vect_transform_loop): Stub out scalar
26471         IFN_MASK_LOAD calls here rather than...
26472         * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
26474 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
26475             Alan Hayward  <alan.hayward@arm.com>
26476             David Sherwood  <david.sherwood@arm.com>
26478         * expmed.c (extract_bit_field_1): For vector extracts,
26479         fall back to extract_bit_field_as_subreg if vec_extract
26480         isn't available.
26482 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
26483             Alan Hayward  <alan.hayward@arm.com>
26484             David Sherwood  <david.sherwood@arm.com>
26486         * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
26487         they are variable or constant sized.
26488         (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
26489         slots for constant-sized data.
26491 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
26492             Alan Hayward  <alan.hayward@arm.com>
26493             David Sherwood  <david.sherwood@arm.com>
26495         * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
26496         handling COND_EXPRs with boolean comparisons, try to find a better
26497         basis for the mask type than the boolean itself.
26499 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
26501         * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
26502         is calculated and how it can be overridden.
26503         * genmodes.c (max_bitsize_mode_any_mode): New variable.
26504         (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
26505         if defined.
26506         (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
26507         if nonzero.
26509 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
26510             Alan Hayward  <alan.hayward@arm.com>
26511             David Sherwood  <david.sherwood@arm.com>
26513         * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
26514         Remove the mode argument.
26515         (aarch64_simd_valid_immediate): Remove the mode and inverse
26516         arguments.
26517         * config/aarch64/iterators.md (bitsize): New iterator.
26518         * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
26519         (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
26520         * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
26521         aarch64_simd_valid_immediate.
26522         * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
26523         (aarch64_reg_or_bic_imm): Likewise.
26524         * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
26525         with an insn_type enum and msl with a modifier_type enum.
26526         Replace element_width with a scalar_mode.  Change the shift
26527         to unsigned int.  Add constructors for scalar_float_mode and
26528         scalar_int_mode elements.
26529         (aarch64_vect_float_const_representable_p): Delete.
26530         (aarch64_can_const_movi_rtx_p)
26531         (aarch64_simd_scalar_immediate_valid_for_move)
26532         (aarch64_simd_make_constant): Update call to
26533         aarch64_simd_valid_immediate.
26534         (aarch64_advsimd_valid_immediate_hs): New function.
26535         (aarch64_advsimd_valid_immediate): Likewise.
26536         (aarch64_simd_valid_immediate): Remove mode and inverse
26537         arguments.  Rewrite to use the above.  Use const_vec_duplicate_p
26538         to detect duplicated constants and use aarch64_float_const_zero_rtx_p
26539         and aarch64_float_const_representable_p on the result.
26540         (aarch64_output_simd_mov_immediate): Remove mode argument.
26541         Update call to aarch64_simd_valid_immediate and use of
26542         simd_immediate_info.
26543         (aarch64_output_scalar_simd_mov_immediate): Update call
26544         accordingly.
26546 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
26547             Alan Hayward  <alan.hayward@arm.com>
26548             David Sherwood  <david.sherwood@arm.com>
26550         * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
26551         (mode_nunits): Likewise CONST_MODE_NUNITS.
26552         * machmode.def (ADJUST_NUNITS): Document.
26553         * genmodes.c (mode_data::need_nunits_adj): New field.
26554         (blank_mode): Update accordingly.
26555         (adj_nunits): New variable.
26556         (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
26557         parameter.
26558         (emit_mode_size_inline): Set need_bytesize_adj for all modes
26559         listed in adj_nunits.
26560         (emit_mode_nunits_inline): Set need_nunits_adj for all modes
26561         listed in adj_nunits.  Don't emit case statements for such modes.
26562         (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
26563         and CONST_MODE_PRECISION.  Make CONST_MODE_SIZE expand to
26564         nothing if adj_nunits is nonnull.
26565         (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
26566         (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
26567         (emit_mode_fbit): Update use of print_maybe_const_decl.
26568         (emit_move_size): Likewise.  Treat the array as non-const
26569         if adj_nunits.
26570         (emit_mode_adjustments): Handle adj_nunits.
26572 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
26574         * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
26575         * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
26576         (VECTOR_MODES): Use it.
26577         (make_vector_modes): Take the prefix as an argument.
26579 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
26580             Alan Hayward  <alan.hayward@arm.com>
26581             David Sherwood  <david.sherwood@arm.com>
26583         * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
26584         * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
26585         for MODE_VECTOR_BOOL.
26586         * machmode.def (VECTOR_BOOL_MODE): Document.
26587         * genmodes.c (VECTOR_BOOL_MODE): New macro.
26588         (make_vector_bool_mode): New function.
26589         (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
26590         MODE_VECTOR_BOOL.
26591         * lto-streamer-in.c (lto_input_mode_table): Likewise.
26592         * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
26593         Likewise.
26594         * stor-layout.c (int_mode_for_mode): Likewise.
26595         * tree.c (build_vector_type_for_mode): Likewise.
26596         * varasm.c (output_constant_pool_2): Likewise.
26597         * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
26598         CONSTM1_RTX (BImode) are the same thing.  Initialize const_tiny_rtx
26599         for MODE_VECTOR_BOOL.
26600         * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
26601         of mode class checks.
26602         * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
26603         instead of a list of mode class checks.
26604         (expand_vector_scalar_condition): Likewise.
26605         (type_for_widest_vector_mode): Handle BImode as an inner mode.
26607 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
26608             Alan Hayward  <alan.hayward@arm.com>
26609             David Sherwood  <david.sherwood@arm.com>
26611         * machmode.h (mode_size): Change from unsigned short to
26612         poly_uint16_pod.
26613         (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
26614         (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
26615         or if measurement_type is not polynomial.
26616         (fixed_size_mode::includes_p): Check for constant-sized modes.
26617         * genmodes.c (emit_mode_size_inline): Make mode_size_inline
26618         return a poly_uint16 rather than an unsigned short.
26619         (emit_mode_size): Change the type of mode_size from unsigned short
26620         to poly_uint16_pod.  Use ZERO_COEFFS for the initializer.
26621         (emit_mode_adjustments): Cope with polynomial vector sizes.
26622         * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
26623         for GET_MODE_SIZE.
26624         * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
26625         for GET_MODE_SIZE.
26626         * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
26627         * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
26628         * caller-save.c (setup_save_areas): Likewise.
26629         (replace_reg_with_saved_mem): Likewise.
26630         * calls.c (emit_library_call_value_1): Likewise.
26631         * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
26632         * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
26633         (gen_lowpart_for_combine): Likewise.
26634         * convert.c (convert_to_integer_1): Likewise.
26635         * cse.c (equiv_constant, cse_insn): Likewise.
26636         * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
26637         (cselib_subst_to_values): Likewise.
26638         * dce.c (word_dce_process_block): Likewise.
26639         * df-problems.c (df_word_lr_mark_ref): Likewise.
26640         * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
26641         * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
26642         (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
26643         (rtl_for_decl_location): Likewise.
26644         * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
26645         * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
26646         * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
26647         (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
26648         (expand_expr_real_1): Likewise.
26649         * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
26650         (pad_below): Likewise.
26651         * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
26652         * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
26653         * ira.c (get_subreg_tracking_sizes): Likewise.
26654         * ira-build.c (ira_create_allocno_objects): Likewise.
26655         * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
26656         (ira_sort_regnos_for_alter_reg): Likewise.
26657         * ira-costs.c (record_operand_costs): Likewise.
26658         * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
26659         (resolve_simple_move): Likewise.
26660         * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
26661         (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
26662         (lra_constraints): Likewise.
26663         (CONST_POOL_OK_P): Reject variable-sized modes.
26664         * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
26665         (add_pseudo_to_slot, lra_spill): Likewise.
26666         * omp-low.c (omp_clause_aligned_alignment): Likewise.
26667         * optabs-query.c (get_best_extraction_insn): Likewise.
26668         * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
26669         * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
26670         (expand_mult_highpart, valid_multiword_target_p): Likewise.
26671         * recog.c (offsettable_address_addr_space_p): Likewise.
26672         * regcprop.c (maybe_mode_change): Likewise.
26673         * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
26674         * regrename.c (build_def_use): Likewise.
26675         * regstat.c (dump_reg_info): Likewise.
26676         * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
26677         (find_reloads, find_reloads_subreg_address): Likewise.
26678         * reload1.c (eliminate_regs_1): Likewise.
26679         * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
26680         * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
26681         (simplify_binary_operation_1, simplify_subreg): Likewise.
26682         * targhooks.c (default_function_arg_padding): Likewise.
26683         (default_hard_regno_nregs, default_class_max_nregs): Likewise.
26684         * tree-cfg.c (verify_gimple_assign_binary): Likewise.
26685         (verify_gimple_assign_ternary): Likewise.
26686         * tree-inline.c (estimate_move_cost): Likewise.
26687         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
26688         * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
26689         (get_address_cost_ainc): Likewise.
26690         * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
26691         (vect_supportable_dr_alignment): Likewise.
26692         * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
26693         (vectorizable_reduction): Likewise.
26694         * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
26695         (vectorizable_operation, vectorizable_load): Likewise.
26696         * tree.c (build_same_sized_truth_vector_type): Likewise.
26697         * valtrack.c (cleanup_auto_inc_dec): Likewise.
26698         * var-tracking.c (emit_note_insn_var_location): Likewise.
26699         * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
26700         (ADDR_VEC_ALIGN): Likewise.
26702 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
26703             Alan Hayward  <alan.hayward@arm.com>
26704             David Sherwood  <david.sherwood@arm.com>
26706         * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
26707         unsigned short.
26708         (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
26709         or if measurement_type is polynomial.
26710         * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
26711         * combine.c (make_extraction): Likewise.
26712         * dse.c (find_shift_sequence): Likewise.
26713         * dwarf2out.c (mem_loc_descriptor): Likewise.
26714         * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
26715         (extract_bit_field, extract_low_bits): Likewise.
26716         * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
26717         (optimize_bitfield_assignment_op, expand_assignment): Likewise.
26718         (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
26719         * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
26720         * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
26721         * reload.c (find_reloads): Likewise.
26722         * reload1.c (alter_reg): Likewise.
26723         * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
26724         * targhooks.c (default_secondary_memory_needed_mode): Likewise.
26725         * tree-if-conv.c (predicate_mem_writes): Likewise.
26726         * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
26727         * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
26728         * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
26729         * valtrack.c (dead_debug_insert_temp): Likewise.
26730         * varasm.c (mergeable_constant_section): Likewise.
26731         * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
26733 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
26734             Alan Hayward  <alan.hayward@arm.com>
26735             David Sherwood  <david.sherwood@arm.com>
26737         * expr.c (expand_assignment): Cope with polynomial mode sizes
26738         when assigning to a CONCAT.
26740 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
26741             Alan Hayward  <alan.hayward@arm.com>
26742             David Sherwood  <david.sherwood@arm.com>
26744         * machmode.h (mode_precision): Change from unsigned short to
26745         poly_uint16_pod.
26746         (mode_to_precision): Return a poly_uint16 rather than an unsigned
26747         short.
26748         (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
26749         or if measurement_type is not polynomial.
26750         (HWI_COMPUTABLE_MODE_P): Turn into a function.  Optimize the case
26751         in which the mode is already known to be a scalar_int_mode.
26752         * genmodes.c (emit_mode_precision): Change the type of mode_precision
26753         from unsigned short to poly_uint16_pod.  Use ZERO_COEFFS for the
26754         initializer.
26755         * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
26756         for GET_MODE_PRECISION.
26757         * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
26758         for GET_MODE_PRECISION.
26759         * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
26760         as polynomial.
26761         (try_combine, find_split_point, combine_simplify_rtx): Likewise.
26762         (expand_field_assignment, make_extraction): Likewise.
26763         (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
26764         (get_last_value): Likewise.
26765         * convert.c (convert_to_integer_1): Likewise.
26766         * cse.c (cse_insn): Likewise.
26767         * expr.c (expand_expr_real_1): Likewise.
26768         * lra-constraints.c (simplify_operand_subreg): Likewise.
26769         * optabs-query.c (can_atomic_load_p): Likewise.
26770         * optabs.c (expand_atomic_load): Likewise.
26771         (expand_atomic_store): Likewise.
26772         * ree.c (combine_reaching_defs): Likewise.
26773         * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
26774         * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
26775         * tree.h (type_has_mode_precision_p): Likewise.
26776         * ubsan.c (instrument_si_overflow): Likewise.
26778 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
26779             Alan Hayward  <alan.hayward@arm.com>
26780             David Sherwood  <david.sherwood@arm.com>
26782         * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
26783         polynomial numbers of units.
26784         (SET_TYPE_VECTOR_SUBPARTS): Likewise.
26785         (valid_vector_subparts_p): New function.
26786         (build_vector_type): Remove temporary shim and take the number
26787         of units as a poly_uint64 rather than an int.
26788         (build_opaque_vector_type): Take the number of units as a
26789         poly_uint64 rather than an int.
26790         * tree.c (build_vector_from_ctor): Handle polynomial
26791         TYPE_VECTOR_SUBPARTS.
26792         (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
26793         (uniform_vector_p, vector_type_mode, build_vector): Likewise.
26794         (build_vector_from_val): If the number of units is variable,
26795         use build_vec_duplicate_cst for constant operands and
26796         VEC_DUPLICATE_EXPR otherwise.
26797         (make_vector_type): Remove temporary is_constant ().
26798         (build_vector_type, build_opaque_vector_type): Take the number of
26799         units as a poly_uint64 rather than an int.
26800         (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
26801         VECTOR_CST_NELTS.
26802         * cfgexpand.c (expand_debug_expr): Likewise.
26803         * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
26804         (store_constructor, expand_expr_real_1): Likewise.
26805         (const_scalar_mask_from_tree): Likewise.
26806         * fold-const-call.c (fold_const_reduction): Likewise.
26807         * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
26808         (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
26809         (native_encode_vector, vec_cst_ctor_to_array): Likewise.
26810         (fold_relational_const): Likewise.
26811         (native_interpret_vector): Likewise.  Change the size from an
26812         int to an unsigned int.
26813         * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
26814         TYPE_VECTOR_SUBPARTS.
26815         (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
26816         (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
26817         duplicating a non-constant operand into a variable-length vector.
26818         * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
26819         TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
26820         * ipa-icf.c (sem_variable::equals): Likewise.
26821         * match.pd: Likewise.
26822         * omp-simd-clone.c (simd_clone_subparts): Likewise.
26823         * print-tree.c (print_node): Likewise.
26824         * stor-layout.c (layout_type): Likewise.
26825         * targhooks.c (default_builtin_vectorization_cost): Likewise.
26826         * tree-cfg.c (verify_gimple_comparison): Likewise.
26827         (verify_gimple_assign_binary): Likewise.
26828         (verify_gimple_assign_ternary): Likewise.
26829         (verify_gimple_assign_single): Likewise.
26830         * tree-pretty-print.c (dump_generic_node): Likewise.
26831         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
26832         (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
26833         * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
26834         (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
26835         (vect_shift_permute_load_chain): Likewise.
26836         * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
26837         (expand_vector_condition, optimize_vector_constructor): Likewise.
26838         (lower_vec_perm, get_compute_type): Likewise.
26839         * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
26840         (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
26841         * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
26842         (vect_recog_mask_conversion_pattern): Likewise.
26843         * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
26844         (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
26845         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
26846         (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
26847         (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
26848         (vectorizable_shift, vectorizable_operation, vectorizable_store)
26849         (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
26850         (supportable_widening_operation): Likewise.
26851         (supportable_narrowing_operation): Likewise.
26852         * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
26853         Likewise.
26854         * varasm.c (output_constant): Likewise.
26856 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
26857             Alan Hayward  <alan.hayward@arm.com>
26858             David Sherwood  <david.sherwood@arm.com>
26860         * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
26861         so that both the length == 3 and length != 3 cases set up their
26862         own permute vectors.  Add comments explaining why we know the
26863         number of elements is constant.
26864         (vect_permute_load_chain): Likewise.
26866 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
26867             Alan Hayward  <alan.hayward@arm.com>
26868             David Sherwood  <david.sherwood@arm.com>
26870         * machmode.h (mode_nunits): Change from unsigned char to
26871         poly_uint16_pod.
26872         (ONLY_FIXED_SIZE_MODES): New macro.
26873         (pod_mode::measurement_type, scalar_int_mode::measurement_type)
26874         (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
26875         (complex_mode::measurement_type, fixed_size_mode::measurement_type):
26876         New typedefs.
26877         (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
26878         (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
26879         or if measurement_type is not polynomial.
26880         * genmodes.c (ZERO_COEFFS): New macro.
26881         (emit_mode_nunits_inline): Make mode_nunits_inline return a
26882         poly_uint16.
26883         (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
26884         Use ZERO_COEFFS when emitting initializers.
26885         * data-streamer.h (bp_pack_poly_value): New function.
26886         (bp_unpack_poly_value): Likewise.
26887         * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
26888         for GET_MODE_NUNITS.
26889         * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
26890         for GET_MODE_NUNITS.
26891         * tree.c (make_vector_type): Remove temporary shim and make
26892         the real function take the number of units as a poly_uint64
26893         rather than an int.
26894         (build_vector_type_for_mode): Handle polynomial nunits.
26895         * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
26896         * emit-rtl.c (const_vec_series_p_1): Likewise.
26897         (gen_rtx_CONST_VECTOR): Likewise.
26898         * fold-const.c (test_vec_duplicate_folding): Likewise.
26899         * genrecog.c (validate_pattern): Likewise.
26900         * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
26901         * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
26902         * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
26903         (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
26904         (expand_vec_cond_expr, expand_mult_highpart): Likewise.
26905         * rtlanal.c (subreg_get_info): Likewise.
26906         * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
26907         (vect_grouped_load_supported): Likewise.
26908         * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
26909         * tree-vect-loop.c (have_whole_vector_shift): Likewise.
26910         * simplify-rtx.c (simplify_unary_operation_1): Likewise.
26911         (simplify_const_unary_operation, simplify_binary_operation_1)
26912         (simplify_const_binary_operation, simplify_ternary_operation)
26913         (test_vector_ops_duplicate, test_vector_ops): Likewise.
26914         (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
26915         instead of CONST_VECTOR_NUNITS.
26916         * varasm.c (output_constant_pool_2): Likewise.
26917         * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
26918         explicit-encoded elements in the XVEC for variable-length vectors.
26920 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
26922         * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
26924 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
26925             Alan Hayward  <alan.hayward@arm.com>
26926             David Sherwood  <david.sherwood@arm.com>
26928         * coretypes.h (fixed_size_mode): Declare.
26929         (fixed_size_mode_pod): New typedef.
26930         * builtins.h (target_builtins::x_apply_args_mode)
26931         (target_builtins::x_apply_result_mode): Change type to
26932         fixed_size_mode_pod.
26933         * builtins.c (apply_args_size, apply_result_size, result_vector)
26934         (expand_builtin_apply_args_1, expand_builtin_apply)
26935         (expand_builtin_return): Update accordingly.
26937 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
26939         * cse.c (hash_rtx_cb): Hash only the encoded elements.
26940         * cselib.c (cselib_hash_rtx): Likewise.
26941         * expmed.c (make_tree): Build VECTOR_CSTs directly from the
26942         CONST_VECTOR encoding.
26944 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
26945             Jeff Law  <law@redhat.com>
26947         PR target/83641
26948         * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
26949         noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
26950         only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
26951         and add REG_CFA_ADJUST_CFA notes in that case to both insns.
26953         PR target/83641
26954         * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
26955         explicitly probe *sp in a noreturn function if there were any callee
26956         register saves or frame pointer is needed.
26958 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
26960         PR debug/83621
26961         * cfgexpand.c (expand_debug_expr): Return NULL if mode is
26962         BLKmode for ternary, binary or unary expressions.
26964         PR debug/83645
26965         * var-tracking.c (delete_vta_debug_insn): New inline function.
26966         (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
26967         insns from get_insns () to NULL instead of each bb separately.
26968         Use delete_vta_debug_insn.  No longer static.
26969         (vt_debug_insns_local, variable_tracking_main_1): Adjust
26970         delete_vta_debug_insns callers.
26971         * rtl.h (delete_vta_debug_insns): Declare.
26972         * final.c (rest_of_handle_final): Call delete_vta_debug_insns
26973         instead of variable_tracking_main.
26975 2018-01-03  Martin Sebor  <msebor@redhat.com>
26977         PR tree-optimization/83603
26978         * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
26979         arguments past the endof the argument list in functions declared
26980         without a prototype.
26981         * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
26982         Avoid checking when arguments are null.
26984 2018-01-03  Martin Sebor  <msebor@redhat.com>
26986         PR c/83559
26987         * doc/extend.texi (attribute const): Fix a typo.
26988         * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
26989         issuing -Wsuggest-attribute for void functions.
26991 2018-01-03  Martin Sebor  <msebor@redhat.com>
26993         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
26994         offset_int::from instead of wide_int::to_shwi.
26995         (maybe_diag_overlap): Remove assertion.
26996         Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
26997         * gimple-ssa-sprintf.c (format_directive): Same.
26998         (parse_directive): Same.
26999         (sprintf_dom_walker::compute_format_length): Same.
27000         (try_substitute_return_value): Same.
27002 2018-01-03  Jeff Law  <law@redhat.com>
27004         PR middle-end/83654
27005         * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
27006         non-constant residual for zero at runtime and avoid probing in
27007         that case.  Reorganize code for trailing problem to mirror handling
27008         of the residual.
27010 2018-01-03  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
27012         PR tree-optimization/83501
27013         * tree-ssa-strlen.c (get_string_cst): New.
27014         (handle_char_store): Call get_string_cst.
27016 2018-01-03  Martin Liska  <mliska@suse.cz>
27018         PR tree-optimization/83593
27019         * tree-ssa-strlen.c: Include tree-cfg.h.
27020         (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
27021         (strlen_dom_walker): Add new member variable m_cleanup_cfg.
27022         (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
27023         to false.
27024         (strlen_dom_walker::before_dom_children): Call
27025         gimple_purge_dead_eh_edges. Dump tranformation with details
27026         dump flags.
27027         (strlen_dom_walker::before_dom_children): Update call by adding
27028         new argument cleanup_eh.
27029         (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
27031 2018-01-03  Martin Liska  <mliska@suse.cz>
27033         PR ipa/83549
27034         * cif-code.def (VARIADIC_THUNK): New enum value.
27035         * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
27036         thunks.
27038 2018-01-03  Jan Beulich  <jbeulich@suse.com>
27040         * sse.md (mov<mode>_internal): Tighten condition for when to use
27041         vmovdqu<ssescalarsize> for TI and OI modes.
27043 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
27045         Update copyright years.
27047 2018-01-03  Martin Liska  <mliska@suse.cz>
27049         PR ipa/83594
27050         * ipa-visibility.c (function_and_variable_visibility): Skip
27051         functions with noipa attribure.
27053 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
27055         * gcc.c (process_command): Update copyright notice dates.
27056         * gcov-dump.c (print_version): Ditto.
27057         * gcov.c (print_version): Ditto.
27058         * gcov-tool.c (print_version): Ditto.
27059         * gengtype.c (create_file): Ditto.
27060         * doc/cpp.texi: Bump @copying's copyright year.
27061         * doc/cppinternals.texi: Ditto.
27062         * doc/gcc.texi: Ditto.
27063         * doc/gccint.texi: Ditto.
27064         * doc/gcov.texi: Ditto.
27065         * doc/install.texi: Ditto.
27066         * doc/invoke.texi: Ditto.
27068 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27070         * vector-builder.h (vector_builder::m_full_nelts): Change from
27071         unsigned int to poly_uint64.
27072         (vector_builder::full_nelts): Update prototype accordingly.
27073         (vector_builder::new_vector): Likewise.
27074         (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
27075         (vector_builder::operator ==): Likewise.
27076         (vector_builder::finalize): Likewise.
27077         * int-vector-builder.h (int_vector_builder::int_vector_builder):
27078         Take the number of elements as a poly_uint64 rather than an
27079         unsigned int.
27080         * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
27081         from unsigned int to poly_uint64.
27082         (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
27083         (vec_perm_indices::new_vector): Likewise.
27084         (vec_perm_indices::length): Likewise.
27085         (vec_perm_indices::nelts_per_input): Likewise.
27086         (vec_perm_indices::input_nelts): Likewise.
27087         * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
27088         number of elements per input as a poly_uint64 rather than an
27089         unsigned int.  Use the original encoding for variable-length
27090         vectors, rather than clamping each individual element.
27091         For the second and subsequent elements in each pattern,
27092         clamp the step and base before clamping their sum.
27093         (vec_perm_indices::series_p): Handle polynomial element counts.
27094         (vec_perm_indices::all_in_range_p): Likewise.
27095         (vec_perm_indices_to_tree): Likewise.
27096         (vec_perm_indices_to_rtx): Likewise.
27097         * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
27098         * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
27099         (tree_vector_builder::new_binary_operation): Handle polynomial
27100         element counts.  Return false if we need to know the number
27101         of elements at compile time.
27102         * fold-const.c (fold_vec_perm): Punt if the number of elements
27103         isn't known at compile time.
27105 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27107         * vec-perm-indices.h (vec_perm_builder): Change element type
27108         from HOST_WIDE_INT to poly_int64.
27109         (vec_perm_indices::element_type): Update accordingly.
27110         (vec_perm_indices::clamp): Handle polynomial element_types.
27111         * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
27112         (vec_perm_indices::all_in_range_p): Likewise.
27113         (tree_to_vec_perm_builder): Check for poly_int64 trees rather
27114         than shwi trees.
27115         * vector-builder.h (vector_builder::stepped_sequence_p): Handle
27116         polynomial vec_perm_indices element types.
27117         * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
27118         * fold-const.c (fold_vec_perm): Likewise.
27119         * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
27120         * tree-vect-generic.c (lower_vec_perm): Likewise.
27121         * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
27122         * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
27123         element type to HOST_WIDE_INT.
27125 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27126             Alan Hayward  <alan.hayward@arm.com>
27127             David Sherwood  <david.sherwood@arm.com>
27129         * alias.c (addr_side_effect_eval): Take the size as a poly_int64
27130         rather than an int.  Use plus_constant.
27131         (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
27132         Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
27134 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27135             Alan Hayward  <alan.hayward@arm.com>
27136             David Sherwood  <david.sherwood@arm.com>
27138         * calls.c (emit_call_1, expand_call): Change struct_value_size from
27139         a HOST_WIDE_INT to a poly_int64.
27141 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27142             Alan Hayward  <alan.hayward@arm.com>
27143             David Sherwood  <david.sherwood@arm.com>
27145         * calls.c (load_register_parameters): Cope with polynomial
27146         mode sizes.  Require a constant size for BLKmode parameters
27147         that aren't described by a PARALLEL.  If BLOCK_REG_PADDING
27148         forces a parameter to be padded at the lsb end in order to
27149         fill a complete number of words, require the parameter size
27150         to be ordered wrt UNITS_PER_WORD.
27152 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27153             Alan Hayward  <alan.hayward@arm.com>
27154             David Sherwood  <david.sherwood@arm.com>
27156         * reload1.c (spill_stack_slot_width): Change element type
27157         from unsigned int to poly_uint64_pod.
27158         (alter_reg): Treat mode sizes as polynomial.
27160 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27161             Alan Hayward  <alan.hayward@arm.com>
27162             David Sherwood  <david.sherwood@arm.com>
27164         * reload.c (complex_word_subreg_p): New function.
27165         (reload_inner_reg_of_subreg, push_reload): Use it.
27167 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27168             Alan Hayward  <alan.hayward@arm.com>
27169             David Sherwood  <david.sherwood@arm.com>
27171         * lra-constraints.c (process_alt_operands): Reject matched
27172         operands whose sizes aren't ordered.
27173         (match_reload): Refer to this check here.
27175 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27176             Alan Hayward  <alan.hayward@arm.com>
27177             David Sherwood  <david.sherwood@arm.com>
27179         * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
27180         that the mode size is in the set {1, 2, 4, 8, 16}.
27182 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27183             Alan Hayward  <alan.hayward@arm.com>
27184             David Sherwood  <david.sherwood@arm.com>
27186         * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
27187         Use plus_constant instead of gen_rtx_PLUS.
27189 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27190             Alan Hayward  <alan.hayward@arm.com>
27191             David Sherwood  <david.sherwood@arm.com>
27193         * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
27194         * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
27195         * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
27196         * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
27197         * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
27198         * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
27199         * config/i386/i386-protos.h (ix86_push_rounding): Declare.
27200         * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
27201         * config/i386/i386.c (ix86_push_rounding): ...this new function.
27202         * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
27203         a poly_int64.
27204         * config/m32c/m32c.c (m32c_push_rounding): Likewise.
27205         * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
27206         * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
27207         * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
27208         * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
27209         * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
27210         * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
27211         * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
27212         * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
27213         * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
27214         function.
27215         * expr.c (emit_move_resolve_push): Treat the input and result
27216         of PUSH_ROUNDING as a poly_int64.
27217         (emit_move_complex_push, emit_single_push_insn_1): Likewise.
27218         (emit_push_insn): Likewise.
27219         * lra-eliminations.c (mark_not_eliminable): Likewise.
27220         * recog.c (push_operand): Likewise.
27221         * reload1.c (elimination_effects): Likewise.
27222         * rtlanal.c (nonzero_bits1): Likewise.
27223         * calls.c (store_one_arg): Likewise.  Require the padding to be
27224         known at compile time.
27226 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27227             Alan Hayward  <alan.hayward@arm.com>
27228             David Sherwood  <david.sherwood@arm.com>
27230         * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
27231         Use plus_constant instead of gen_rtx_PLUS.
27233 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27234             Alan Hayward  <alan.hayward@arm.com>
27235             David Sherwood  <david.sherwood@arm.com>
27237         * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
27238         rather than an int.
27240 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27241             Alan Hayward  <alan.hayward@arm.com>
27242             David Sherwood  <david.sherwood@arm.com>
27244         * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
27245         instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
27246         via stack temporaries.  Treat the mode size as polynomial too.
27248 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27249             Alan Hayward  <alan.hayward@arm.com>
27250             David Sherwood  <david.sherwood@arm.com>
27252         * expr.c (expand_expr_real_2): When handling conversions involving
27253         unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
27254         multiplying int_size_in_bytes by BITS_PER_UNIT.  Treat GET_MODE_BISIZE
27255         as a poly_uint64 too.
27257 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27258             Alan Hayward  <alan.hayward@arm.com>
27259             David Sherwood  <david.sherwood@arm.com>
27261         * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
27263 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27264             Alan Hayward  <alan.hayward@arm.com>
27265             David Sherwood  <david.sherwood@arm.com>
27267         * combine.c (can_change_dest_mode): Handle polynomial
27268         REGMODE_NATURAL_SIZE.
27269         * expmed.c (store_bit_field_1): Likewise.
27270         * expr.c (store_constructor): Likewise.
27271         * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
27272         and polynomial REGMODE_NATURAL_SIZE.
27273         (gen_lowpart_common): Likewise.
27274         * reginfo.c (record_subregs_of_mode): Likewise.
27275         * rtlanal.c (read_modify_subreg_p): Likewise.
27277 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27278             Alan Hayward  <alan.hayward@arm.com>
27279             David Sherwood  <david.sherwood@arm.com>
27281         * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
27282         numbers of elements.
27284 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27285             Alan Hayward  <alan.hayward@arm.com>
27286             David Sherwood  <david.sherwood@arm.com>
27288         * match.pd: Cope with polynomial numbers of vector elements.
27290 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27291             Alan Hayward  <alan.hayward@arm.com>
27292             David Sherwood  <david.sherwood@arm.com>
27294         * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
27295         in a POINTER_PLUS_EXPR.
27297 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27298             Alan Hayward  <alan.hayward@arm.com>
27299             David Sherwood  <david.sherwood@arm.com>
27301         * omp-simd-clone.c (simd_clone_subparts): New function.
27302         (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
27303         (ipa_simd_modify_function_body): Likewise.
27305 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27306             Alan Hayward  <alan.hayward@arm.com>
27307             David Sherwood  <david.sherwood@arm.com>
27309         * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
27310         (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
27311         (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
27312         (expand_vector_condition, vector_element): Likewise.
27313         (subparts_gt): New function.
27314         (get_compute_type): Use subparts_gt.
27315         (count_type_subparts): Delete.
27316         (expand_vector_operations_1): Use subparts_gt instead of
27317         count_type_subparts.
27319 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27320             Alan Hayward  <alan.hayward@arm.com>
27321             David Sherwood  <david.sherwood@arm.com>
27323         * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
27324         (vect_compile_time_alias): ...this new function.  Do the calculation
27325         on poly_ints rather than trees.
27326         (vect_prune_runtime_alias_test_list): Update call accordingly.
27328 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27329             Alan Hayward  <alan.hayward@arm.com>
27330             David Sherwood  <david.sherwood@arm.com>
27332         * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
27333         numbers of units.
27334         (vect_schedule_slp_instance): Likewise.
27336 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27337             Alan Hayward  <alan.hayward@arm.com>
27338             David Sherwood  <david.sherwood@arm.com>
27340         * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
27341         constant and extern definitions for variable-length vectors.
27342         (vect_get_constant_vectors): Note that the number of units
27343         is known to be constant.
27345 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27346             Alan Hayward  <alan.hayward@arm.com>
27347             David Sherwood  <david.sherwood@arm.com>
27349         * tree-vect-stmts.c (vectorizable_conversion): Treat the number
27350         of units as polynomial.  Choose between WIDE and NARROW based
27351         on multiple_p.
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         * tree-vect-stmts.c (simd_clone_subparts): New function.
27358         (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
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         * tree-vect-stmts.c (vectorizable_call): Treat the number of
27365         vectors as polynomial.  Use build_index_vector for
27366         IFN_GOMP_SIMD_LANE.
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-stmts.c (get_load_store_type): Treat the number of
27373         units as polynomial.  Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
27374         for variable-length vectors.
27375         (vectorizable_mask_load_store): Treat the number of units as
27376         polynomial, asserting that it is constant if the condition has
27377         already been enforced.
27378         (vectorizable_store, vectorizable_load): Likewise.
27380 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27381             Alan Hayward  <alan.hayward@arm.com>
27382             David Sherwood  <david.sherwood@arm.com>
27384         * tree-vect-loop.c (vectorizable_live_operation): Treat the number
27385         of units as polynomial.  Punt if we can't tell at compile time
27386         which vector contains the final result.
27388 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27389             Alan Hayward  <alan.hayward@arm.com>
27390             David Sherwood  <david.sherwood@arm.com>
27392         * tree-vect-loop.c (vectorizable_induction): Treat the number
27393         of units as polynomial.  Punt on SLP inductions.  Use an integer
27394         VEC_SERIES_EXPR for variable-length integer reductions.  Use a
27395         cast of such a series for variable-length floating-point
27396         reductions.
27398 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27399             Alan Hayward  <alan.hayward@arm.com>
27400             David Sherwood  <david.sherwood@arm.com>
27402         * tree.h (build_index_vector): Declare.
27403         * tree.c (build_index_vector): New function.
27404         * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
27405         of units as polynomial, forcibly converting it to a constant if
27406         vectorizable_reduction has already enforced the condition.
27407         (vect_create_epilog_for_reduction): Likewise.  Use build_index_vector
27408         to create a {1,2,3,...} vector.
27409         (vectorizable_reduction): Treat the number of units as polynomial.
27410         Choose vectype_in based on the largest scalar element size rather
27411         than the smallest number of units.  Enforce the restrictions
27412         relied on above.
27414 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27415             Alan Hayward  <alan.hayward@arm.com>
27416             David Sherwood  <david.sherwood@arm.com>
27418         * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
27419         number of units as polynomial.
27421 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27422             Alan Hayward  <alan.hayward@arm.com>
27423             David Sherwood  <david.sherwood@arm.com>
27425         * target.h (vector_sizes, auto_vector_sizes): New typedefs.
27426         * target.def (autovectorize_vector_sizes): Return the vector sizes
27427         by pointer, using vector_sizes rather than a bitmask.
27428         * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
27429         * targhooks.c (default_autovectorize_vector_sizes): Likewise.
27430         * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
27431         Likewise.
27432         * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
27433         * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
27434         * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
27435         * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
27436         * omp-general.c (omp_max_vf): Likewise.
27437         * omp-low.c (omp_clause_aligned_alignment): Likewise.
27438         * optabs-query.c (can_vec_mask_load_store_p): Likewise.
27439         * tree-vect-loop.c (vect_analyze_loop): Likewise.
27440         * tree-vect-slp.c (vect_slp_bb): Likewise.
27441         * doc/tm.texi: Regenerate.
27442         * tree-vectorizer.h (current_vector_size): Change from an unsigned int
27443         to a poly_uint64.
27444         * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
27445         the vector size as a poly_uint64 rather than an unsigned int.
27446         (current_vector_size): Change from an unsigned int to a poly_uint64.
27447         (get_vectype_for_scalar_type): Update accordingly.
27448         * tree.h (build_truth_vector_type): Take the size and number of
27449         units as a poly_uint64 rather than an unsigned int.
27450         (build_vector_type): Add a temporary overload that takes
27451         the number of units as a poly_uint64 rather than an unsigned int.
27452         * tree.c (make_vector_type): Likewise.
27453         (build_truth_vector_type): Take the number of units as a poly_uint64
27454         rather than an unsigned int.
27456 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27457             Alan Hayward  <alan.hayward@arm.com>
27458             David Sherwood  <david.sherwood@arm.com>
27460         * target.def (get_mask_mode): Take the number of units and length
27461         as poly_uint64s rather than unsigned ints.
27462         * targhooks.h (default_get_mask_mode): Update accordingly.
27463         * targhooks.c (default_get_mask_mode): Likewise.
27464         * config/i386/i386.c (ix86_get_mask_mode): Likewise.
27465         * doc/tm.texi: Regenerate.
27467 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27468             Alan Hayward  <alan.hayward@arm.com>
27469             David Sherwood  <david.sherwood@arm.com>
27471         * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
27472         * omp-general.c (omp_max_vf): Likewise.
27473         * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
27474         (expand_omp_simd): Handle polynomial safelen.
27475         * omp-low.c (omplow_simd_context): Add a default constructor.
27476         (omplow_simd_context::max_vf): Change from int to poly_uint64.
27477         (lower_rec_simd_input_clauses): Update accordingly.
27478         (lower_rec_input_clauses): Likewise.
27480 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27481             Alan Hayward  <alan.hayward@arm.com>
27482             David Sherwood  <david.sherwood@arm.com>
27484         * tree-vectorizer.h (vect_nunits_for_cost): New function.
27485         * tree-vect-loop.c (vect_model_reduction_cost): Use it.
27486         * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
27487         (vect_analyze_slp_cost): Likewise.
27488         * tree-vect-stmts.c (vect_model_store_cost): Likewise.
27489         (vect_model_load_cost): Likewise.
27491 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27492             Alan Hayward  <alan.hayward@arm.com>
27493             David Sherwood  <david.sherwood@arm.com>
27495         * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
27496         (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
27497         from an unsigned int * to a poly_uint64_pod *.
27498         (calculate_unrolling_factor): New function.
27499         (vect_analyze_slp_instance): Use it.  Track polynomial max_nunits.
27501 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27502             Alan Hayward  <alan.hayward@arm.com>
27503             David Sherwood  <david.sherwood@arm.com>
27505         * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
27506         from an unsigned int to a poly_uint64.
27507         (_loop_vec_info::slp_unrolling_factor): Likewise.
27508         (_loop_vec_info::vectorization_factor): Change from an int
27509         to a poly_uint64.
27510         (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
27511         (vect_get_num_vectors): New function.
27512         (vect_update_max_nunits, vect_vf_for_cost): Likewise.
27513         (vect_get_num_copies): Use vect_get_num_vectors.
27514         (vect_analyze_data_ref_dependences): Change max_vf from an int *
27515         to an unsigned int *.
27516         (vect_analyze_data_refs): Change min_vf from an int * to a
27517         poly_uint64 *.
27518         (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
27519         than an unsigned HOST_WIDE_INT.
27520         * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
27521         (vect_analyze_data_ref_dependence): Change max_vf from an int *
27522         to an unsigned int *.
27523         (vect_analyze_data_ref_dependences): Likewise.
27524         (vect_compute_data_ref_alignment): Handle polynomial vf.
27525         (vect_enhance_data_refs_alignment): Likewise.
27526         (vect_prune_runtime_alias_test_list): Likewise.
27527         (vect_shift_permute_load_chain): Likewise.
27528         (vect_supportable_dr_alignment): Likewise.
27529         (dependence_distance_ge_vf): Take the vectorization factor as a
27530         poly_uint64 rather than an unsigned HOST_WIDE_INT.
27531         (vect_analyze_data_refs): Change min_vf from an int * to a
27532         poly_uint64 *.
27533         * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
27534         vfm1 as a poly_uint64 rather than an int.  Make the same change
27535         for the returned bound_scalar.
27536         (vect_gen_vector_loop_niters): Handle polynomial vf.
27537         (vect_do_peeling): Likewise.  Update call to
27538         vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
27539         (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
27540         be constant.
27541         * tree-vect-loop.c (vect_determine_vectorization_factor)
27542         (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
27543         (vect_get_known_peeling_cost): Likewise.
27544         (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
27545         (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
27546         (vect_transform_loop): Likewise.  Use the lowest possible VF when
27547         updating the upper bounds of the loop.
27548         (vect_min_worthwhile_factor): Make static.  Return an unsigned int
27549         rather than an int.
27550         * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
27551         polynomial unroll factors.
27552         (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
27553         (vect_make_slp_decision): Likewise.
27554         (vect_supported_load_permutation_p): Likewise, and polynomial
27555         vf too.
27556         (vect_analyze_slp_cost): Handle polynomial vf.
27557         (vect_slp_analyze_node_operations): Likewise.
27558         (vect_slp_analyze_bb_1): Likewise.
27559         (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
27560         than an unsigned HOST_WIDE_INT.
27561         * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
27562         (vectorizable_load): Handle polynomial vf.
27563         * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
27564         a poly_uint64.
27565         (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
27567 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27568             Alan Hayward  <alan.hayward@arm.com>
27569             David Sherwood  <david.sherwood@arm.com>
27571         * match.pd: Handle bit operations involving three constants
27572         and try to fold one pair.
27574 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27576         * tree-vect-loop-manip.c: Include gimple-fold.h.
27577         (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
27578         niters_maybe_zero parameters.  Handle other cases besides a step of 1.
27579         (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
27580         Add a path that uses a step of VF instead of 1, but disable it
27581         for now.
27582         (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
27583         and niters_no_overflow parameters.  Update calls to
27584         slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
27585         Create a new SSA name if the latter choses to use a ste other
27586         than zero, and return it via niters_vector_mult_vf_var.
27587         * tree-vect-loop.c (vect_transform_loop): Update calls to
27588         vect_do_peeling, vect_gen_vector_loop_niters and
27589         slpeel_make_loop_iterate_ntimes.
27590         * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
27591         (vect_gen_vector_loop_niters): Update declarations after above changes.
27593 2018-01-02  Michael Meissner  <meissner@linux.vnet.ibm.com>
27595         * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
27596         128-bit round to integer instructions.
27597         (ceil<mode>2): Likewise.
27598         (btrunc<mode>2): Likewise.
27599         (round<mode>2): Likewise.
27601 2018-01-02  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
27603         * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
27604         unaligned VSX load/store on P8/P9.
27605         (expand_block_clear): Allow the use of unaligned VSX
27606         load/store on P8/P9.
27608 2018-01-02  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
27610         * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
27611         New function.
27612         (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
27613         swap associated with both a load and a store.
27615 2018-01-02  Andrew Waterman  <andrew@sifive.com>
27617         * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
27618         * config/riscv/riscv.md (clear_cache): Use it.
27620 2018-01-02  Artyom Skrobov  <tyomitch@gmail.com>
27622         * web.c: Remove out-of-date comment.
27624 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
27626         * expr.c (fixup_args_size_notes): Check that any existing
27627         REG_ARGS_SIZE notes are correct, and don't try to re-add them.
27628         (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
27629         (emit_single_push_insn): ...here.
27631 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
27633         * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
27634         (const_vector_encoded_nelts): New function.
27635         (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
27636         (const_vector_int_elt, const_vector_elt): Declare.
27637         * emit-rtl.c (const_vector_int_elt_1): New function.
27638         (const_vector_elt): Likewise.
27639         * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
27640         of CONST_VECTOR_ELT.
27642 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
27644         * expr.c: Include rtx-vector-builder.h.
27645         (const_vector_mask_from_tree): Use rtx_vector_builder and operate
27646         directly on the tree encoding.
27647         (const_vector_from_tree): Likewise.
27648         * optabs.c: Include rtx-vector-builder.h.
27649         (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
27650         sequence of "u" values.
27651         * vec-perm-indices.c: Include rtx-vector-builder.h.
27652         (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
27653         directly on the vec_perm_indices encoding.
27655 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
27657         * doc/rtl.texi (const_vector): Describe new encoding scheme.
27658         * Makefile.in (OBJS): Add rtx-vector-builder.o.
27659         * rtx-vector-builder.h: New file.
27660         * rtx-vector-builder.c: Likewise.
27661         * rtl.h (rtx_def::u2): Add a const_vector field.
27662         (CONST_VECTOR_NPATTERNS): New macro.
27663         (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
27664         (CONST_VECTOR_DUPLICATE_P): Likewise.
27665         (CONST_VECTOR_STEPPED_P): Likewise.
27666         (CONST_VECTOR_ENCODED_ELT): Likewise.
27667         (const_vec_duplicate_p): Check for a duplicated vector encoding.
27668         (unwrap_const_vec_duplicate): Likewise.
27669         (const_vec_series_p): Check for a non-duplicated vector encoding.
27670         Say that the function only returns true for integer vectors.
27671         * emit-rtl.c: Include rtx-vector-builder.h.
27672         (gen_const_vec_duplicate_1): Delete.
27673         (gen_const_vector): Call gen_const_vec_duplicate instead of
27674         gen_const_vec_duplicate_1.
27675         (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
27676         (gen_const_vec_duplicate): Use rtx_vector_builder.
27677         (gen_const_vec_series): Likewise.
27678         (gen_rtx_CONST_VECTOR): Likewise.
27679         * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
27680         (swap_const_vector_halves): Take an rtx pointer rather than rtx.
27681         Build a new vector rather than modifying a CONST_VECTOR in-place.
27682         (handle_special_swappables): Update call accordingly.
27683         * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
27684         (swap_const_vector_halves): Take an rtx pointer rather than rtx.
27685         Build a new vector rather than modifying a CONST_VECTOR in-place.
27686         (handle_special_swappables): Update call accordingly.
27688 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
27690         * simplify-rtx.c (simplify_const_binary_operation): Use
27691         CONST_VECTOR_ELT instead of XVECEXP.
27693 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
27695         * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
27696         the selector elements to be different from the data elements
27697         if the selector is a VECTOR_CST.
27698         * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
27699         ssizetype for the selector.
27701 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
27703         * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
27704         before testing each element individually.
27705         * tree-vect-generic.c (lower_vec_perm): Likewise.
27707 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
27709         * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
27710         * selftest-run-tests.c (selftest::run_tests): Call it.
27711         * vector-builder.h (vector_builder::operator ==): New function.
27712         (vector_builder::operator !=): Likewise.
27713         * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
27714         (vec_perm_indices::all_from_input_p): New function.
27715         * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
27716         (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
27717         * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
27718         instead of reading the VECTOR_CST directly.  Detect whether both
27719         vector inputs are the same before constructing the vec_perm_indices,
27720         and update the number of inputs argument accordingly.  Use the
27721         utility functions added above.  Only construct sel2 if we need to.
27723 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
27725         * optabs.c (expand_vec_perm_var): Use an explicit encoding for
27726         the broadcast of the low byte.
27727         (expand_mult_highpart): Use an explicit encoding for the permutes.
27728         * optabs-query.c (can_mult_highpart_p): Likewise.
27729         * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
27730         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
27731         (vectorizable_bswap): Likewise.
27732         * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
27733         explicit encoding for the power-of-2 permutes.
27734         (vect_permute_store_chain): Likewise.
27735         (vect_grouped_load_supported): Likewise.
27736         (vect_permute_load_chain): Likewise.
27738 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
27740         * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
27741         * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
27742         * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
27743         * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
27744         * tree-vect-stmts.c (vectorizable_bswap): Likewise.
27745         (vect_gen_perm_mask_any): Likewise.
27747 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
27749         * int-vector-builder.h: New file.
27750         * vec-perm-indices.h: Include int-vector-builder.h.
27751         (vec_perm_indices): Redefine as an int_vector_builder.
27752         (auto_vec_perm_indices): Delete.
27753         (vec_perm_builder): Redefine as a stand-alone class.
27754         (vec_perm_indices::vec_perm_indices): New function.
27755         (vec_perm_indices::clamp): Likewise.
27756         * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
27757         (vec_perm_indices::new_vector): New function.
27758         (vec_perm_indices::new_expanded_vector): Update for new
27759         vec_perm_indices class.
27760         (vec_perm_indices::rotate_inputs): New function.
27761         (vec_perm_indices::all_in_range_p): Operate directly on the
27762         encoded form, without computing elided elements.
27763         (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
27764         encoding.  Update for new vec_perm_indices class.
27765         * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
27766         the given vec_perm_builder.
27767         (expand_vec_perm_var): Update vec_perm_builder constructor.
27768         (expand_mult_highpart): Use vec_perm_builder instead of
27769         auto_vec_perm_indices.
27770         * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
27771         vec_perm_indices instead of auto_vec_perm_indices.  Use a single
27772         or double series encoding as appropriate.
27773         * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
27774         vec_perm_indices instead of auto_vec_perm_indices.
27775         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
27776         * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
27777         (vect_permute_store_chain): Likewise.
27778         (vect_grouped_load_supported): Likewise.
27779         (vect_permute_load_chain): Likewise.
27780         (vect_shift_permute_load_chain): Likewise.
27781         * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
27782         (vect_transform_slp_perm_load): Likewise.
27783         (vect_schedule_slp_instance): Likewise.
27784         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
27785         (vectorizable_mask_load_store): Likewise.
27786         (vectorizable_bswap): Likewise.
27787         (vectorizable_store): Likewise.
27788         (vectorizable_load): Likewise.
27789         * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
27790         vec_perm_indices instead of auto_vec_perm_indices.  Use
27791         tree_to_vec_perm_builder to read the vector from a tree.
27792         * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
27793         vec_perm_builder instead of a vec_perm_indices.
27794         (have_whole_vector_shift): Use vec_perm_builder and
27795         vec_perm_indices instead of auto_vec_perm_indices.  Leave the
27796         truncation to calc_vec_perm_mask_for_shift.
27797         (vect_create_epilog_for_reduction): Likewise.
27798         * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
27799         from auto_vec_perm_indices to vec_perm_indices.
27800         (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
27801         instead of changing individual elements.
27802         (aarch64_vectorize_vec_perm_const): Use new_vector to install
27803         the vector in d.perm.
27804         * config/arm/arm.c (expand_vec_perm_d::perm): Change
27805         from auto_vec_perm_indices to vec_perm_indices.
27806         (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
27807         instead of changing individual elements.
27808         (arm_vectorize_vec_perm_const): Use new_vector to install
27809         the vector in d.perm.
27810         * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
27811         Update vec_perm_builder constructor.
27812         (rs6000_expand_interleave): Likewise.
27813         * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
27814         (rs6000_expand_interleave): Likewise.
27816 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
27818         * optabs-query.c (can_vec_perm_var_p): Check whether lowering
27819         to qimode could truncate the indices.
27820         * optabs.c (expand_vec_perm_var): Likewise.
27822 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
27824         * Makefile.in (OBJS): Add vec-perm-indices.o.
27825         * vec-perm-indices.h: New file.
27826         * vec-perm-indices.c: Likewise.
27827         * target.h (vec_perm_indices): Replace with a forward class
27828         declaration.
27829         (auto_vec_perm_indices): Move to vec-perm-indices.h.
27830         * optabs.h: Include vec-perm-indices.h.
27831         (expand_vec_perm): Delete.
27832         (selector_fits_mode_p, expand_vec_perm_var): Declare.
27833         (expand_vec_perm_const): Declare.
27834         * target.def (vec_perm_const_ok): Replace with...
27835         (vec_perm_const): ...this new hook.
27836         * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
27837         (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
27838         * doc/tm.texi: Regenerate.
27839         * optabs.def (vec_perm_const): Delete.
27840         * doc/md.texi (vec_perm_const): Likewise.
27841         (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
27842         * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
27843         expand_vec_perm for constant permutation vectors.  Assert that
27844         the mode of variable permutation vectors is the integer equivalent
27845         of the mode that is being permuted.
27846         * optabs-query.h (selector_fits_mode_p): Declare.
27847         * optabs-query.c: Include vec-perm-indices.h.
27848         (selector_fits_mode_p): New function.
27849         (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
27850         is defined, instead of checking whether the vec_perm_const_optab
27851         exists.  Use targetm.vectorize.vec_perm_const instead of
27852         targetm.vectorize.vec_perm_const_ok.  Check whether the indices
27853         fit in the vector mode before using a variable permute.
27854         * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
27855         vec_perm_indices instead of an rtx.
27856         (expand_vec_perm): Replace with...
27857         (expand_vec_perm_const): ...this new function.  Take the selector
27858         as a vec_perm_indices rather than an rtx.  Also take the mode of
27859         the selector.  Update call to shift_amt_for_vec_perm_mask.
27860         Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
27861         Use vec_perm_indices::new_expanded_vector to expand the original
27862         selector into bytes.  Check whether the indices fit in the vector
27863         mode before using a variable permute.
27864         (expand_vec_perm_var): Make global.
27865         (expand_mult_highpart): Use expand_vec_perm_const.
27866         * fold-const.c: Includes vec-perm-indices.h.
27867         * tree-ssa-forwprop.c: Likewise.
27868         * tree-vect-data-refs.c: Likewise.
27869         * tree-vect-generic.c: Likewise.
27870         * tree-vect-loop.c: Likewise.
27871         * tree-vect-slp.c: Likewise.
27872         * tree-vect-stmts.c: Likewise.
27873         * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
27874         Delete.
27875         * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
27876         * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
27877         (aarch64_vectorize_vec_perm_const_ok): Fuse into...
27878         (aarch64_vectorize_vec_perm_const): ...this new function.
27879         (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
27880         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
27881         * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
27882         * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
27883         * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
27884         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
27885         (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
27886         into...
27887         (arm_vectorize_vec_perm_const): ...this new function.  Explicitly
27888         check for NEON modes.
27889         * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
27890         * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
27891         * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
27892         (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
27893         into...
27894         (ix86_vectorize_vec_perm_const): ...this new function.  Incorporate
27895         the old VEC_PERM_CONST conditions.
27896         * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
27897         * config/ia64/vect.md (vec_perm_const<mode>): Delete.
27898         * config/ia64/ia64.c (ia64_expand_vec_perm_const)
27899         (ia64_vectorize_vec_perm_const_ok): Merge into...
27900         (ia64_vectorize_vec_perm_const): ...this new function.
27901         * config/mips/loongson.md (vec_perm_const<mode>): Delete.
27902         * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
27903         * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
27904         * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
27905         * config/mips/mips.c (mips_expand_vec_perm_const)
27906         (mips_vectorize_vec_perm_const_ok): Merge into...
27907         (mips_vectorize_vec_perm_const): ...this new function.
27908         * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
27909         * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
27910         * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
27911         * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
27912         * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
27913         (rs6000_expand_vec_perm_const): Delete.
27914         * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
27915         Delete.
27916         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
27917         (altivec_expand_vec_perm_const_le): Take each operand individually.
27918         Operate on constant selectors rather than rtxes.
27919         (altivec_expand_vec_perm_const): Likewise.  Update call to
27920         altivec_expand_vec_perm_const_le.
27921         (rs6000_expand_vec_perm_const): Delete.
27922         (rs6000_vectorize_vec_perm_const_ok): Delete.
27923         (rs6000_vectorize_vec_perm_const): New function.
27924         (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
27925         an element count and rtx array.
27926         (rs6000_expand_extract_even): Update call accordingly.
27927         (rs6000_expand_interleave): Likewise.
27928         * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
27929         * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
27930         * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
27931         * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
27932         (rs6000_expand_vec_perm_const): Delete.
27933         * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
27934         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
27935         (altivec_expand_vec_perm_const_le): Take each operand individually.
27936         Operate on constant selectors rather than rtxes.
27937         (altivec_expand_vec_perm_const): Likewise.  Update call to
27938         altivec_expand_vec_perm_const_le.
27939         (rs6000_expand_vec_perm_const): Delete.
27940         (rs6000_vectorize_vec_perm_const_ok): Delete.
27941         (rs6000_vectorize_vec_perm_const): New function.  Remove stray
27942         reference to the SPE evmerge intructions.
27943         (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
27944         an element count and rtx array.
27945         (rs6000_expand_extract_even): Update call accordingly.
27946         (rs6000_expand_interleave): Likewise.
27947         * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
27948         * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
27949         new function.
27950         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
27952 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
27954         * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
27955         vector mode and that that mode matches the mode of the data
27956         being permuted.
27957         (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
27958         out into expand_vec_perm_var.  Do all CONST_VECTOR handling here,
27959         directly using expand_vec_perm_1 when forcing selectors into
27960         registers.
27961         (expand_vec_perm_var): New function, split out from expand_vec_perm.
27963 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
27965         * optabs-query.h (can_vec_perm_p): Delete.
27966         (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
27967         * optabs-query.c (can_vec_perm_p): Split into...
27968         (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
27969         (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
27970         particular selector is valid.
27971         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
27972         * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
27973         (vect_grouped_load_supported): Likewise.
27974         (vect_shift_permute_load_chain): Likewise.
27975         * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
27976         (vect_transform_slp_perm_load): Likewise.
27977         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
27978         (vectorizable_bswap): Likewise.
27979         (vect_gen_perm_mask_checked): Likewise.
27980         * fold-const.c (fold_ternary_loc): Likewise.  Don't take
27981         implementations of variable permutation vectors into account
27982         when deciding which selector to use.
27983         * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
27984         vec_perm_const_optab is supported; instead use can_vec_perm_const_p
27985         with a false third argument.
27986         * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
27987         to test whether the constant selector is valid and can_vec_perm_var_p
27988         to test whether a variable selector is valid.
27990 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
27992         * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
27993         * optabs-query.c (can_vec_perm_p): Likewise.
27994         * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
27995         instead of vec_perm_indices.
27996         * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
27997         (vect_gen_perm_mask_checked): Likewise,
27998         * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
27999         (vect_gen_perm_mask_checked): Likewise,
28001 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
28003         * optabs-query.h (qimode_for_vec_perm): Declare.
28004         * optabs-query.c (can_vec_perm_p): Split out qimode search to...
28005         (qimode_for_vec_perm): ...this new function.
28006         * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
28008 2018-01-02  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
28010         * rtlanal.c (canonicalize_condition): Return 0 if final rtx
28011         does not have a conditional at the top.
28013 2018-01-02  Richard Biener  <rguenther@suse.de>
28015         * ipa-inline.c (big_speedup_p): Fix expression.
28017 2018-01-02  Jan Hubicka  <hubicka@ucw.cz>
28019         PR target/81616
28020         * config/i386/x86-tune-costs.h: Increase cost of integer load costs
28021         for generic 4->6.
28023 2018-01-02  Jan Hubicka  <hubicka@ucw.cz>
28025         PR target/81616
28026         Generic tuning.
28027         * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
28028         cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
28029         and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
28030         cond_taken_branch_cost 3->4.
28032 2018-01-01  Jakub Jelinek  <jakub@redhat.com>
28034         PR tree-optimization/83581
28035         * tree-loop-distribution.c (pass_loop_distribution::execute): Return
28036         TODO_cleanup_cfg if any changes have been made.
28038         PR middle-end/83608
28039         * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
28040         convert_modes if target mode has the right side, but different mode
28041         class.
28043         PR middle-end/83609
28044         * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
28045         last argument when extracting from CONCAT.  If either from_real or
28046         from_imag is NULL, use expansion through memory.  If result is not
28047         a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
28048         the parts directly to inner mode, if even that fails, use expansion
28049         through memory.
28051         PR middle-end/83623
28052         * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
28053         check for bswap in mode rather than HImode and use that in expand_unop
28054         too.
28056 Copyright (C) 2018 Free Software Foundation, Inc.
28058 Copying and distribution of this file, with or without modification,
28059 are permitted in any medium without royalty provided the copyright
28060 notice and this notice are preserved.