re PR target/87156 (ICE building libstdc++ for mips64)
[official-gcc.git] / gcc / ChangeLog
blobf8d7217c363f02c2a23818564c5622f538717855
1 2018-10-11  Jan Hubicka  <hubicka@ucw.cz>
3         PR target/87156
4         * cgraphclones.c (cgraph_node::create_version_clone_with_body):
5         Set new_decl virtual flag to zero.
7 2018-10-11  Martin Sebor  <msebor@redhat.com>
9         PR middle-end/87593
10         * doc/extend.texi (attribute format_arg): Discuss using multiple
11         attributes on a single function.
13 2018-10-11  Giuliano Belinassi  <giuliano.belinassi@usp.br>
15         PR tree-optimization/86829
16         * match.pd (sin (atan (x))): New simplification rules.
17         (cos (atan (x))): Likewise.
18         * real.c (build_sinatan_real): New function.
19         * real.h (build_sinatan_real): Prototype.
21 2018-10-11  Will Schmidt <will_schmidt@vnet.ibm.com>
23         * config/rs6000/rs6000.c (map_to_integral_tree_type): New helper
24         function.
25         (fold_mergeeo_helper): New helper function.
26         (rs6000_gimple_fold_builtin): Add hooks for vec_mergee and vec_mergeo
27         intrinsics.  Correct some whitespace indentation issues.
29 2018-10-11  Wilco Dijkstra  <wdijkstr@arm.com>
31         PR target/87511
32         * config/aarch64/aarch64.c (aarch64_mask_and_shift_for_ubfiz_p):
33         Use HOST_WIDE_INT_1U for shift.
35 2018-10-11  Doug Rupp  <rupp@adacore.com>
36             Olivier Hainque  <hainque@adacore.com>
38         * config/rs6000/vxworks.h (VXWORKS_RELAX_LINK_SPEC): New macro.
39         Pass --relax to the linker for RTPs.
40         (LINK_SPEC): Append VXWORKS_RELAX_LINK_SPEC.
42 2018-10-11  Andrew Stubbs  <ams@codesourcery.com>
43             Jan Hubicka  <jh@suse.cz>
44             Martin Jambor  <mjambor@suse.cz>
46         * print-rtl.c (print_rtx_operand_codes_E_and_V): Print how many times
47         the same elements are repeated rather than printing all of them.
48         * read-rtl.c (rtx_reader::read_rtx_operand): Recognize and expand
49         "repeated" elements.
50         * read-rtl-function.c (test_loading_repeat): New function.
51         (read_rtl_function_c_tests): Call test_loading_repeat.
52         * rtl-tests.c (test_dumping_repeat): New function.
53         (rtl_tests_c_tests): Call test_dumping_repeat.
55 2018-10-11  Richard Biener  <rguenther@suse.de>
57         * config/i386/x86-tune-costs.h (bdver?_memcpy, bdver?_memset,
58         bdver?_cost): Unify to ...
59         (bdver_memcpy, bdver_memset, bdver_cost): ... this.
60         * config/i386/i386.c (processor_cost_table): Adjust.
62 2018-10-10  Eric Botcazou  <ebotcazou@adacore.com>
64         PR middle-end/87574
65         * cgraphunit.c (cgraph_node::expand_thunk): Force DECL_IGNORED_P on
66         the thunk when expanding to GIMPLE.
68 2018-10-10  Bernd Edlinger  <bernd.edlinger@hotmail.de>
70         * varasm.c (mergeable_string_section): Don't try to move zero-length
71         strings to the merge section.
73 2018-10-10  Uros Bizjak  <ubizjak@gmail.com>
75         PR target/87573
76         * config/i386/mmx.md (const_vector 0 -> mem splitter): New splitter.
78 2018-10-10  Jakub Jelinek  <jakub@redhat.com>
80         PR target/87550
81         * config/i386/i386-builtin.def (IX86_BUILTIN_RDPMC): Move from args set
82         to special_args set.
84 2018-10-10  Richard Biener  <rguenther@suse.de>
86         * config/i386/sse.md (reduc_plus_scal_v8df, reduc_plus_scal_v4df,
87         reduc_plus_scal_v2df, reduc_plus_scal_v16sf, reduc_plus_scal_v8sf,
88         reduc_plus_scal_v4sf): Merge into pattern reducing to half width
89         and recursing and pattern terminating the recursion on SSE
90         vector width using ix86_expand_reduc.
91         (reduc_sminmax_scal_<mode>): Split into part reducing to half
92         width and recursing and SSE2 vector variant doing the final
93         reduction with ix86_expand_reduc.
94         (reduc_uminmax_scal_<mode>): Likewise for the AVX512 variants
95         with terminating the recursion at AVX level, splitting that
96         to SSE there.
98 2018-10-09  David Malcolm  <dmalcolm@redhat.com>
100         * genmatch.c (error_cb): Rename to...
101         (diagnostic_cb): ...this, converting int params to enums.
102         (fatal_at): Update for renaming.
103         (warning_at): Likewise.
104         (main): Likewise.
105         * input.c (selftest::ebcdic_execution_charset::apply):
106         Update for renaming of...
107         (selftest::ebcdic_execution_charset::on_error): ...this, renaming
108         to...
109         (selftest::ebcdic_execution_charset::on_diagnostic): ...this,
110         converting level and reason to enums.
111         (class selftest::lexer_error_sink): Rename to...
112         (class selftest::lexer_test_options): ...this, renaming field
113         "m_errors" to "m_diagnostics".
114         (selftest::lexer_test_options::apply): Update for renaming of...
115         (selftest::lexer_test_options::on_error): ...this, renaming to...
116         (selftest::lexer_test_options::on_diagnostic): ...this
117         converting level and reason to enums.
118         (selftest::test_lexer_string_locations_raw_string_unterminated):
119         Update for renamings.
120         * opth-gen.awk (struct cpp_reason_option_codes_t): Use enum for
121         "reason".
123 2018-10-09  Paul A. Clarke  <pc@us.ibm.com>
125         * config.gcc (powerpc*-*-*): Add pmmintrin.h to extra_headers.
126         * config/rs6000/pmmintrin.h: New file.
128 2018-10-09  Eric Botcazou  <ebotcazou@adacore.com>
130         PR tree-optimization/86659
131         * gimple-match.h (gimple_match_op constructors): Initialize reverse.
133 2018-10-09  Richard Biener  <rguenther@suse.de>
135         PR tree-optimization/63155
136         * tree-ssa-structalias.c: Include tree-ssa.h.
137         (get_constraint_for_ssa_var): For undefs return nothing_id.
138         (find_func_aliases): Cleanup PHI handling.
140 2018-10-09  Richard Biener  <rguenther@suse.de>
142         * tree-vectorizer.c (dump_stmt_cost): Fix cut&paste missing
143         replacements.
145 2018-10-09  Martin Liska  <mliska@suse.cz>
147         * asan.c (asan_emit_stack_protection): If a stack variable
148         is located in a same file as current function, then emit
149         line info into variable definition string.
151 2018-10-08  Eric Botcazou  <ebotcazou@adacore.com>
153         * print-rtl.c (rtx_writer::print_rtx_operand_code_i): Print column
154         information.
156 2018-10-08  Eric Botcazou  <ebotcazou@adacore.com>
158         * cgraphunit.c (cgraph_node::expand_thunk): Do not force DECL_IGNORED_P
159         on the thunk.
161 2018-10-08  Eric Botcazou  <ebotcazou@adacore.com>
163         PR tree-optimization/86659
164         * gimple-match.h (struct gimple_match_op): Add reverse field.
165         (gimple_match_op::set_op): New overloaded method.
166         * gimple-match-head.c (maybe_build_generic_op) <BIT_FIELD_REF>: Set
167         the REF_REVERSE_STORAGE_ORDER flag on the value.
168         (gimple_simplify) <GIMPLE_ASSIGN>: For BIT_FIELD_REF, propagate the
169         REF_REVERSE_STORAGE_ORDER flag and avoid simplifying if it is set.
170         
171 2018-10-08  Richard Sandiford  <richard.sandiford@arm.com>
173         PR middle-end/63155
174         * gimple-ssa-backprop.c (backprop::intersect_uses): Use
175         FOR_EACH_IMM_USE_FAST instead of FOR_EACH_IMM_USE_STMT.
177 2018-10-08  H.J. Lu  <hongjiu.lu@intel.com>
179         PR target/87517
180         * config/i386/avx512fintrin.h (_mm512_mask_fmaddsub_round_pd):
181         Defined with __builtin_ia32_vfmaddsubpd512_mask.
183 2018-10-08  Richard Biener  <rguenther@suse.de>
185         * config/i386/x86-tune-costs.h (znver1_cost): Make AVX256 vector loads
186         cost the same as AVX128 ones.
188 2018-10-08  Paul Koning  <ni1d@arrl.net>
190         * config/pdp11/pdp11-protos.h (output_block_move): Remove.
191         (expand_block_move): New function.
192         * config/pdp11/pdp11.c (output_block_move): Remove.
193         (expand_block_move): New function.
194         * config/pdp11/pdp11.h (MOVE_RATIO): New definition.
195         * config/pdp11/pdp11.md (movmemhi): Use expand_block_move.
196         (*movmemhi1): Remove.
198 2018-10-08  Robin Dapp  <rdapp@linux.ibm.com>
200         * config/s390/2827.md: Increase latencies for some FP instructions.
202 2018-10-08  Richard Biener  <rguenther@suse.de>
204         * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
205         Open a dump scope.
206         * tree-vectorizer.c (dump_stmt_cost): Add cost param and dump it.
207         * tree-vectorizer.h (dump_stmt_cost): Adjust.
208         (add_stmt_cost): Dump return value of the hook.
210 2018-10-08  Richard Biener  <rguenther@suse.de>
212         PR tree-optimization/63155
213         * tree-ssa-propagate.c (add_ssa_edge): Do cheap check first.
214         (ssa_propagation_engine::ssa_propagate): Remove redundant
215         bitmap bit clearing.
217 2018-10-05  Peter Bergner  <bergner@linux.ibm.com>
219         PR rtl-optimization/86939
220         PR rtl-optimization/87479
221         * ira.h (non_conflicting_reg_copy_p): New prototype.
222         * ira-lives.c (ignore_reg_for_conflicts): New static variable.
223         (make_hard_regno_dead): Don't add conflicts for register
224         ignore_reg_for_conflicts.
225         (make_object_dead): Likewise.
226         (non_conflicting_reg_copy_p): New function.
227         (process_bb_node_lives): Set ignore_reg_for_conflicts for copies.
228         Remove special conflict handling of REAL_PIC_OFFSET_TABLE_REGNUM.
229         * lra-lives.c (ignore_reg_for_conflicts): New static variable.
230         (make_hard_regno_dead): Don't add conflicts for register
231         ignore_reg_for_conflicts.  Remove special conflict handling of
232         REAL_PIC_OFFSET_TABLE_REGNUM.  Remove now unused argument
233         check_pic_pseudo_p and update callers.
234         (mark_pseudo_dead): Don't add conflicts for register
235         ignore_reg_for_conflicts.
236         (process_bb_lives): Set ignore_reg_for_conflicts for copies.
238 2018-10-05  Andrew Waterman  <andrew@sifive.com>
239             Jim Wilson  <jimw@sifive.com>
241         * config/riscv/riscv.md (f<quiet_pattern>_quiet<ANYF:mode><X:mode>4):
242         Add define_expand.  Add ! HONOR_SNANS check to current pattern.  Add
243         new pattern using HONOR_SNANS that emits one extra instruction.
245 2018-10-05  Segher Boessenkool  <segher@kernel.crashing.org>
247         * config/rs6000/rs6000.md (unnamed mfcr scc_comparison_operator
248         patterns): Merge SI and DI patterns to a GPR pattern.
249         (unnamed define_insn and define_split for record form of that): Merge
250         to a single define_insn_and_split pattern.
252 2018-10-05  David Malcolm  <dmalcolm@redhat.com>
254         PR c++/56856
255         * input.c (expand_location_to_spelling_point): Add param "aspect"
256         and use rather than hardcoding LOCATION_ASPECT_CARET.
257         (get_substring_ranges_for_loc): Handle the case of a single token
258         within a macro expansion.
259         * input.h (expand_location_to_spelling_point): Add "aspect" param,
260         defaulting to LOCATION_ASPECT_CARET.
262 2018-10-05  Paul Koning  <ni1d@arrl.net>
264         * config/pdp11/pdp11.c (TARGET_CXX_GUARD_TYPE): Define.
265         (TARGET_CXX_CLASS_DATA_ALWAYS_COMDAT): Define.
266         (TARGET_CXX_LIBRARY_RTTI_COMDAT): Define.
267         (pdp11_guard_type): New function.
269 2018-10-05  Paul Koning  <ni1d@arrl.net>
271         * config/pdp11/pdp11.h (FLOAT_TYPE_SIZE): Always 32.
272         * config/pdp11/pdp11.opt (mfloat32): Remove.
273         (mfloat64): Remove.
274         * doc/invoke.texi (pdp11 -mfloat32): Remove:
275         (pdp11 -mfloat64): Remove.
277 2018-10-05  Uros Bizjak  <ubizjak@gmail.com>
279         * config/i386/i386.md (*cmpxf_cc_i387): Remove pattern.
280         (*cmp<mode>_cc_i387): Ditto.
281         (*cmpu<mode>_cc_i387): Ditto.
282         (*cmp<X87MODEF:mode>_<SWI24:mode>_cc_i387): Ditto.
283         * config/i386/i386.c (ix86_expand_fp_compare): Remove
284         "scratch" argument.
285         <case IX86_FPCMP_SAHF>: Do not generate pattern with HImode clobber.
286         Emit x86_sahf_1 pattern.
287         (ix86_expand_compare): Update call to ix86_expand_fp_compare.
288         (ix86_expand_carry_flag_compare): Ditto.
290 2018-10-05  Uros Bizjak  <ubizjak@gmail.com>
292         * config/i386/i386.md (*cmpxf_i387): Change operand 2 predicate
293         to reg_or_0_operand.  Add "C" constraint.
294         (*cmpxf_cc_i387): Ditto.
295         (*cmp<mode>_i387): Change operand 2 predicate
296         to nonimm_or_0_operand.  Add "C" constraint.
297         (*cmp<mode>_cc_i387): Ditto.
298         (*cmp<mode>_0_i387): Remove insn pattern.
299         (*cmp<mode>_0_cc_i387): Ditto.
301 2018-10-05  Uros Bizjak  <ubizjak@gmail.com>
303         * config/i386/constraints.md ("C"): Do not depend on TARGET_SSE.
304         * config/i386/predicates.md (nonimm_or_0_operand): Rename
305         from vector_move_operand.  Update all uses.
307 2018-10-05  Martin Sebor  <msebor@redhat.com>
309         PR tree-optimization/87490
310         * builtins.c (expand_builtin_strnlen): Handle a null data.decl
311         consistently.
313 2018-10-05  Richard Biener  <rguenther@suse.de>
315         PR tree-optimization/63155
316         * tree-ssa-ccp.c (ccp_propagate::visit_phi): Avoid excess
317         vertical space in dumpfiles.
318         * tree-ssa-propagate.h
319         (ssa_propagation_engine::process_ssa_edge_worklist): Remove.
320         * tree-ssa-propagate.c (cfg_blocks_back): New global.
321         (ssa_edge_worklist_back): Likewise.
322         (curr_order): Likewise.
323         (cfg_blocks_get): Remove abstraction.
324         (cfg_blocks_add): Likewise.
325         (cfg_blocks_empty_p): Likewise.
326         (add_ssa_edge): Add to current or next worklist based on
327         RPO index.
328         (add_control_edge): Likewise.
329         (ssa_propagation_engine::process_ssa_edge_worklist): Fold
330         into ...
331         (ssa_propagation_engine::ssa_propagate): ... here.  Unify
332         iteration from CFG and SSA edge worklist so we process
333         everything in RPO order, prioritizing forward progress
334         over iteration.
335         (ssa_prop_init): Allocate new worklists, do not dump
336         immediate uses.
337         (ssa_prop_fini): Free new worklists.
339 2018-10-05  Richard Biener  <rguenther@suse.de>
341         * tree-core.h (tree_block::abstract_flag): Remove.
342         (tree_block::block_num): Make full 32bits.
343         * tree.def (BLOCK): Remove docs about BLOCK_ABSTRACT.
344         * tree.h (BLOCK_ABSTRACT): Remove.
345         * dwarf2out.c (gen_lexical_block_die): Remove dead code
346         resulting from BLOCK_ABSTRACT being always false.
347         (gen_inlined_subroutine_die): Likewise.
348         (gen_block_die): Likewise.
349         * tree.c (block_ultimate_origin): Likewise.
350         * tree-pretty-print.c (dump_block_node): Remove code dealing
351         with BLOCK_ABSTRACT.
352         * tree-ssa-live.c (dump_scope_block): Likewise.
353         * tree-streamer-in.c (unpack_ts_block_value_fields): Likewise.
354         * tree-streamer-out.c (pack_ts_block_value_fields): Likewise.
356 2018-10-05   Richard Biener  <rguenther@suse.de>
358         * config/i386/i386.c (ix86_add_stmt_cost): When scalar cost
359         is asked for initialize mode to the component mode of the
360         vector type.
362 2018-10-05  H.J. Lu  <hongjiu.lu@intel.com>
364         PR target/87522
365         * config/i386/gnu-user.h (ASM_SPEC): Don't pass -msse2avx to
366         assembler for -mavx.
367         * config/i386/gnu-user64.h (ASM_SPEC): Likewise.
369 2018-10-05  Segher Boessenkool  <segher@kernel.crashing.org>
371         PR target/87509
372         * config/rs6000/rs6000-builtin.def (RS6000_BUILTIN_SET_FPSCR_DRN): Use
373         RS6000_BTM_DFP.
374         * config/rs6000/rs6000.md (rs6000_set_fpscr_rn): Require the operand
375         to be DImode.  When using mffscrn, force the operand to a register.
377 2018-10-04  Uros Bizjak  <ubizjak@gmail.com>
379         * config/i386/i386.md (*fop_<X87MODEF:mode>_2_i387): Macroize insn
380         from *fop_<MODEF:mode>_2_i387 and *fop_xf_2_i387 using
381         X87MODEF mode iterator.
382         (*fop_<X87MODEF:mode>_3_i387): Macroize insn from
383         *fop_<MODEF:mode>_3_i387 and *fop_xf_3_i387 using
384         X87MODEF mode iterator.
386 2018-10-04  Vinay Kumar  <vinay.kumar@blackfigtech.com>
388         * doc/invoke.texi (-Wno-prio-ctor-dtor): Document new warning
389         -Wno-prio-ctor-dtor.
391 2018-10-04  David Malcolm  <dmalcolm@redhat.com>
393         * Makefile.in (OBJS): Add opt-problem.o.
394         * dump-context.h: Include "selftest.h.
395         (selftest::temp_dump_context): New forward decl.
396         (class dump_context): Make friend of class
397         selftest::temp_dump_context.
398         (dump_context::dump_loc_immediate): New decl.
399         (class dump_pretty_printer): Move here from dumpfile.c.
400         (class temp_dump_context): Move to namespace selftest.
401         (temp_dump_context::temp_dump_context): Add param
402         "forcibly_enable_dumping".
403         (selftest::verify_dumped_text):
404         (ASSERT_DUMPED_TEXT_EQ): Move here from dumpfile.c.
405         (selftest::verify_item):
406         (ASSERT_IS_TEXT): Move here from dumpfile.c.
407         (ASSERT_IS_TREE): Likewise.
408         (ASSERT_IS_GIMPLE): Likewise.
409         * dumpfile.c (dump_context::dump_loc): Move immediate dumping
410         to...
411         (dump_context::dump_loc_immediate): ...this new function.
412         (class dump_pretty_printer): Move to dump-context.h.
413         (dump_switch_p_1): Don't enable MSG_PRIORITY_REEMITTED.
414         (opt_info_switch_p_1): Enable MSG_PRIORITY_REEMITTED.
415         (temp_dump_context::temp_dump_context): Move to "selftest"
416         namespace.  Add param "forcibly_enable_dumping", and use it to
417         conditionalize the use of m_pp;
418         (selftest::verify_dumped_text): Make non-static.
419         (ASSERT_DUMPED_TEXT_EQ): Move to dump-context.h.
420         (selftest::verify_item): Make non-static.
421         (ASSERT_IS_TEXT): Move to dump-context.h.
422         (ASSERT_IS_TREE): Likewise.
423         (ASSERT_IS_GIMPLE): Likewise.
424         (selftest::test_capture_of_dump_calls): Pass "true" for new
425         param of temp_dump_context.
426         * dumpfile.h (enum dump_flag): Add MSG_PRIORITY_REEMITTED, adding
427         it to MSG_ALL_PRIORITIES.  Update values of TDF_COMPARE_DEBUG and
428         TDF_COMPARE_DEBUG.
429         * opt-problem.cc: New file.
430         * opt-problem.h: New file.
431         * optinfo-emit-json.cc
432         (selftest::test_building_json_from_dump_calls): Pass "true" for
433         new param of temp_dump_context.
434         * optinfo.cc (optinfo_kind_to_dump_flag): New function.
435         (optinfo::emit_for_opt_problem): New function.
436         (optinfo::emit): Clarity which emit_item is used.
437         * optinfo.h (optinfo::get_dump_location): New accessor.
438         (optinfo::emit_for_opt_problem): New decl.
439         (optinfo::emit): Make const.
440         * selftest-run-tests.c (selftest::run_tests): Call
441         selftest::opt_problem_cc_tests.
442         * selftest.h (selftest::opt_problem_cc_tests): New decl.
443         * tree-data-ref.c (dr_analyze_innermost): Convert return type from
444         bool to opt_result, converting fprintf messages to
445         opt_result::failure_at calls.  Add "stmt" param for use by the
446         failure_at calls.
447         (create_data_ref): Pass "stmt" to the dr_analyze_innermost call.
448         (runtime_alias_check_p): Convert return type from bool to
449         opt_result, converting dump_printf calls to
450         opt_result::failure_at, using the statement DDR_A for their
451         location.
452         (find_data_references_in_stmt): Convert return type from bool to
453         opt_result, converting "return false" to opt_result::failure_at
454         with a new message.
455         * tree-data-ref.h: Include "opt-problem.h".
456         (dr_analyze_innermost): Convert return type from bool to opt_result,
457         and add a const gimple * param.
458         (find_data_references_in_stmt): Convert return type from bool to
459         opt_result.
460         (runtime_alias_check_p): Likewise.
461         * tree-predcom.c (find_looparound_phi): Pass "init_stmt" to
462         dr_analyze_innermost.
463         * tree-vect-data-refs.c (vect_mark_for_runtime_alias_test):
464         Convert return type from bool to opt_result, adding a message for
465         the PARAM_VECT_MAX_VERSION_FOR_ALIAS_CHECKS zero case.
466         (vect_analyze_data_ref_dependence): Convert return type from bool
467         to opt_result.  Change sense of return type from "false"
468         effectively meaning "no problems" to "false" meaning a problem,
469         so that "return false" becomes "return opt_result::success".
470         Convert "return true" calls to opt_result::failure_at, using
471         the location of statement A rather than vect_location.
472         (vect_analyze_data_ref_dependences): Convert return type from bool
473         to opt_result.
474         (verify_data_ref_alignment): Likewise, converting dump_printf_loc
475         calls to opt_result::failure_at, using the stmt location rather
476         than vect_location.
477         (vect_verify_datarefs_alignment): Convert return type from bool
478         to opt_result.
479         (vect_enhance_data_refs_alignment): Likewise.  Split local "stat"
480         into multiple more-tightly-scoped copies.
481         (vect_analyze_data_refs_alignment): Convert return type from bool
482         to opt_result.
483         (vect_analyze_data_ref_accesses): Likewise, converting a
484         "return false" to a "return opt_result::failure_at", adding a
485         new message.
486         (vect_prune_runtime_alias_test_list): Convert return type from
487         bool to opt_result, converting dump_printf_loc to
488         opt_result::failure_at.  Add a %G to show the pertinent statement,
489         and use the stmt's location rather than vect_location.
490         (vect_find_stmt_data_reference): Convert return type from
491         bool to opt_result, converting dump_printf_loc to
492         opt_result::failure_at, using stmt's location.
493         (vect_analyze_data_refs):  Convert return type from bool to
494         opt_result.  Convert "return false" to "return
495         opt_result::failure_at", adding messages as needed.
496         * tree-vect-loop.c (vect_determine_vf_for_stmt_1): Convert return
497         type from bool to opt_result.
498         (vect_determine_vf_for_stmt): Likewise.
499         (vect_determine_vectorization_factor): Likewise, converting
500         dump_printf_loc to opt_result::failure_at, using location of phi
501         rather than vect_location.
502         (vect_analyze_loop_form_1): Convert return type from bool to
503         opt_result, converting dump_printf_loc calls, retaining the use of
504         vect_location.
505         (vect_analyze_loop_form): Convert return type from loop_vec_info
506         to opt_loop_vec_info.
507         (vect_analyze_loop_operations): Convert return type from bool to
508         opt_result, converting dump_printf_loc calls, using the location
509         of phi/stmt rather than vect_location where available.  Convert
510         various "return false" to "return opt_result::failure_at" with
511         "unsupported phi" messages.
512         (vect_get_datarefs_in_loop): Convert return type from bool to
513         opt_result.  Add a message for the
514         PARAM_LOOP_MAX_DATAREFS_FOR_DATADEPS failure.
515         (vect_analyze_loop_2): Convert return type from bool to
516         opt_result.  Ensure "ok" is set to a opt_result::failure_at before
517         each "goto again;", adding new messages where needed.
518         Add "unsupported grouped {store|load}" messages.
519         (vect_analyze_loop): Convert return type from loop_vec_info to
520         opt_loop_vec_info.
521         * tree-vect-slp.c (vect_analyze_slp): Convert return type from
522         bool to opt_result.
523         * tree-vect-stmts.c (process_use): Likewise, converting
524         dump_printf_loc call and using stmt location, rather than
525         vect_location.
526         (vect_mark_stmts_to_be_vectorized): Likeise.
527         (vect_analyze_stmt): Likewise, adding a %G.
528         (vect_get_vector_types_for_stmt): Convert return type from bool to
529         opt_result, converting dump_printf_loc calls and using stmt
530         location, rather than vect_location.
531         (vect_get_mask_type_for_stmt): Convert return type from tree to
532         opt_tree, converting dump_printf_loc calls and using stmt location.
533         * tree-vectorizer.c: Include "opt-problem.h.
534         (try_vectorize_loop_1): Flag "Analyzing loop at" dump message as
535         MSG_PRIORITY_INTERNALS.  Convert local "loop_vinfo" from
536         loop_vec_info to opt_loop_vec_info.  If if fails, and dumping is
537         enabled, use it to report at the top level "couldn't vectorize
538         loop" followed by the problem.
539         * tree-vectorizer.h (opt_loop_vec_info): New typedef.
540         (vect_mark_stmts_to_be_vectorized): Convert return type from bool
541         to opt_result.
542         (vect_analyze_stmt): Likewise.
543         (vect_get_vector_types_for_stmt): Likewise.
544         (tree vect_get_mask_type_for_stmt): Likewise.
545         (vect_analyze_data_ref_dependences): Likewise.
546         (vect_enhance_data_refs_alignment): Likewise.
547         (vect_analyze_data_refs_alignment): Likewise.
548         (vect_verify_datarefs_alignment): Likewise.
549         (vect_analyze_data_ref_accesses): Likewise.
550         (vect_prune_runtime_alias_test_list): Likewise.
551         (vect_find_stmt_data_reference): Likewise.
552         (vect_analyze_data_refs): Likewise.
553         (vect_analyze_loop): Convert return type from loop_vec_info to
554         opt_loop_vec_info.
555         (vect_analyze_loop_form): Likewise.
556         (vect_analyze_slp): Convert return type from bool to opt_result.
558 2018-10-04  David Malcolm  <dmalcolm@redhat.com>
560         * doc/invoke.texi (-fopt-info): Document new "internals"
561         sub-option.
562         * dump-context.h (dump_context::apply_dump_filter_p): New decl.
563         * dumpfile.c (dump_options): Update for renaming of MSG_ALL to
564         MSG_ALL_KINDS.
565         (optinfo_verbosity_options): Add "internals".
566         (kind_as_string): Update for renaming of MSG_ALL to MSG_ALL_KINDS.
567         (dump_context::apply_dump_filter_p): New member function.
568         (dump_context::dump_loc): Use apply_dump_filter_p rather than
569         explicitly masking the dump_kind.
570         (dump_context::begin_scope): Increment the scope depth first.  Use
571         apply_dump_filter_p rather than explicitly masking the dump_kind.
572         (dump_context::emit_item): Use apply_dump_filter_p rather than
573         explicitly masking the dump_kind.
574         (dump_dec): Likewise.
575         (dump_hex): Likewise.
576         (dump_switch_p_1): Default to MSG_ALL_PRIORITIES.
577         (opt_info_switch_p_1): Default to MSG_PRIORITY_USER_FACING.
578         (opt_info_switch_p): Update handling of default
579         MSG_OPTIMIZED_LOCATIONS to cope with default of
580         MSG_PRIORITY_USER_FACING.
581         (dump_basic_block): Use apply_dump_filter_p rather than explicitly
582         masking the dump_kind.
583         (selftest::test_capture_of_dump_calls): Update test_dump_context
584         instances to use MSG_ALL_KINDS | MSG_PRIORITY_USER_FACING rather
585         than MSG_ALL.  Generalize scope test to be run at all four
586         combinations of with/without MSG_PRIORITY_USER_FACING and
587         MSG_PRIORITY_INTERNALS, adding examples of explicit priority
588         for each of the two values.
589         * dumpfile.h (enum dump_flag): Add comment about the MSG_* flags.
590         Rename MSG_ALL to MSG_ALL_KINDS.  Add MSG_PRIORITY_USER_FACING,
591         MSG_PRIORITY_INTERNALS, and MSG_ALL_PRIORITIES, updating the
592         values for TDF_COMPARE_DEBUG and TDF_ALL_VALUES.
593         (AUTO_DUMP_SCOPE): Add a note to the comment about the interaction
594         with MSG_PRIORITY_*.
595         * tree-vect-loop-manip.c (vect_loop_versioning): Mark versioning
596         dump messages as MSG_PRIORITY_USER_FACING.
597         * tree-vectorizer.h (DUMP_VECT_SCOPE): Add a note to the comment
598         about the interaction with MSG_PRIORITY_*.
600 2018-10-04  Bernd Edlinger  <bernd.edlinger@hotmail.de>
602         * varasm.c (output_constant): Add new parameter merge_strings.
603         Make strings properly zero terminated in merge string sections.
604         (mergeable_string_section): Don't fail if the last char is non-zero.
605         (assemble_variable_contents): Handle merge string sections.
606         (assemble_variable): Likewise.
607         (assemble_constant_contents): Likewise.
608         (output_constant_def_contents): Likewise.
609         (output_constructor_array_range,
610         output_constructor_regular_field): Adjust call to output_constant.
611         (output_object_block): Adjust call to assemble_constant_contents
612         and assemble_variable_contents.
614 2018-10-04  Martin Liska  <mliska@suse.cz>
616         PR c/87483
617         * cgraphunit.c (process_function_and_variable_attributes):
618         Warn about a function with alias attribute and a body.
620 2018-10-04  Martin Liska  <mliska@suse.cz>
622         PR ipa/82625
623         * multiple_target.c (redirect_to_specific_clone): New function.
624         (ipa_target_clone): Use it.
625         * tree-inline.c: Fix comment.
627 2018-10-04  David Malcolm  <dmalcolm@redhat.com>
629         * dumpfile.c (gcc::dump_manager::dump_manager): Initialize new
630         fields.
631         (gcc::dump_manager::~dump_manager): Free m_optinfo_filename.
632         (gcc::dump_manager::register_pass): New member function, adapted
633         from loop body in gcc::pass_manager::register_pass, adding a
634         call to update_dfi_for_opt_info.
635         (gcc::dump_manager::opt_info_enable_passes): Store the
636         -fopt-info options into the new fields.  Move the loop
637         bodies into...
638         (gcc::dump_manager::update_dfi_for_opt_info): ...this new member
639         function.
640         * dumpfile.h (struct opt_pass): New forward decl.
641         (gcc::dump_manager::register_pass): New decl.
642         (gcc::dump_manager::update_dfi_for_opt_info): New decl.
643         (class gcc::dump_manager): Add fields "m_optgroup_flags",
644         "m_optinfo_flags", and "m_optinfo_filename".
645         * passes.c (gcc::pass_manager::register_pass): Move all of the
646         dump-handling code to gcc::dump_manager::register_pass.
648 2018-10-04  Peter Bergner  <bergner@linux.ibm.com>
650         PR rtl-optimization/87466
651         * target.def (setjmp_preserves_nonvolatile_regs_p): New target hook.
652         * doc/tm.texi.in (TARGET_SETJMP_PRESERVES_NONVOLATILE_REGS_P): New hook.
653         * doc/tm.texi: Regenerate.
654         * ira-lives.c (process_bb_node_lives): Use the new target hook.
655         * lra-lives.c (process_bb_lives): Likewise.
656         * config/rs6000/rs6000.c (TARGET_SETJMP_PRESERVES_NONVOLATILE_REGS_P):
657         Define.
659 2018-10-04  Tamar Christina  <tamar.christina@arm.com>
661         * params.c (add_params): Fix initialization.
663 2018-10-04  Martin Liska  <mliska@suse.cz>
665         PR gcov-profile/84107
666         * tree-profile.c (init_ic_make_global_vars):
667         Remove ic_void_ptr_var and ic_gcov_type_ptr_var.
668         Come up with new ic_tuple* variables.  Emit
669         __gcov_indirect_call{,_topn} variables.
670         (gimple_gen_ic_profiler): Access the variable
671         and emit gimple.
672         (gimple_gen_ic_func_profiler): Access
673         __gcov_indirect_call.callee field.
674         (gimple_init_gcov_profiler): Use ptr_type_node.
675         * value-prof.c (gimple_ic): Use ptr_type_node.
677 2018-10-04  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
679         PR tree-optimization/85787
680         * ipa-pure-const.c (malloc_candidate_p_1): Move most of malloc_candidate_p
681         into this function and add support for detecting multiple phis.
682         (DUMP_AND_RETURN): Move from malloc_candidate_p into top-level macro.
684 2018-10-04  Martin Liska  <mliska@suse.cz>
686         PR ipa/87491
687         * ipa-inline.c (inline_to_all_callers_1):
688         Call ultimate_alias_target for node being inlined.
690 2018-10-03  Jeff Law  <law@redhat.com>
692         * gimple-ssa-sprintf.c (format_string): Do not hardcode size of
693         target's wchar_t.
694         * tree.c (get_typenode_from_name): Moved from fortran/trans-types.c.
695         * tree.h (get_typenode_from_name): Prototype.
697 2018-10-03  Uros Bizjak  <ubizjak@gmail.com>
699         * config/i386/i386.md (*cmp<X87MODEF:mode>_<SWI24:mode>_i387):
700         Change operand 2 predicate to nonimmediate_operand.
701         (*cmp<X87MODEF:mode>_<SWI24:mode>_cc_i387): Ditto.
703 2018-10-03  Martin Sebor  <msebor@redhat.com>
704             Jeff Law  <law@redhat.com>
706         * gimple-ssa-sprintf.c (struct fmtresult): Add new member and
707         initialize it.
708         (get_string_length): Detect unterminated arrays.
709         (format_string): Same.
710         (format_directive): Warn about unterminated arrays.
711         (handle_gimple_call): Mark statements with no_warning as needed.
713 2018-10-03  Jim Wilson  <jimw@sifive.com>
715         * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): For ABI_ILP32E,
716         also define __riscv_abi_rve.  Delete trailing white space.
718 2018-10-03  Paul Koning  <ni1d@arrl.net>
720         Enable LRA register allocator for PDP11.
721         * config/pdp11/constraints.md (Q): Use define_memory_constraint.
722         (R): Likewise.
723         (D): Likewise.
724         * config/pdp11/pdp11.c (pdp11_lra_p): New function.
725         * config/pdp11/pdp11.opt (-mlra): New option.
726         * doc/invoke.texi (PDP-11 Options): Document -mlra.
728 2018-10-03  Uros Bizjak  <ubizjak@gmail.com>
730         * config/i386/i386.md (*<absneg:code>extendsfdf2): Remove.
731         (*<absneg:code>extend<mode>xf2): Ditto.
733 2018-10-03  Aldy Hernandez  <aldyh@redhat.com>
735         PR tree-optimization/87415
736         * tree-vrp.c (set_value_range_with_overflow): Special case one bit
737         precision fields.
739 2018-10-02  Jeff Law  <law@redhat.com>
741         * gimple-fold.c (get_range_strlen): Only set *nonstr when
742         an unterminated string is discovered.  Bubble up range
743         even for unterminated strings.
744         (gimple_fold_builtin_strlen): Do not fold if get_range_strlen
745         indicates the string was not terminated via NONSTR.
747 2018-10-03  Aldy Hernandez  <aldyh@redhat.com>
749         * tree-vrp.c (extract_range_from_unary_expr): Special case all
750         pointer conversions.
751         Do not do anything special for anti-ranges.
753 2018-10-03  Jérôme Lambourg  <lambourg@adacore.com>
755         * config/arm/vxworks.h (ARM_TARGET2_DWARF_FORMAT): Adjust to
756         DW_EH_PE_pcrel | DW_EH_PE_indirect for RTPs.
758 2018-10-03  Martin Liska  <mliska@suse.cz>
760         PR gcov-profile/86109
761         * coverage.c (coverage_begin_function): Do not
762         mark lambdas as artificial.
763         * tree-core.h (struct GTY): Remove tm_clone_flag
764         and introduce new lambda_function.
765         * tree.h (DECL_LAMBDA_FUNCTION): New macro.
767 2018-10-02  Aaron Sawdey  <acsawdey@linux.ibm.com>
769         PR target/87474
770         * config/rs6000/rs6000-string.c (expand_strn_compare): Check that both
771         P8_VECTOR and VSX are enabled.
773 2018-10-02  Andreas Krebbel  <krebbel@linux.ibm.com>
775         * config/s390/driver-native.c (s390_host_detect_local_cpu): Add
776         0x3907 as CPU model number.
778 2018-10-02  Andreas Krebbel  <krebbel@linux.ibm.com>
780         * common/config/s390/s390-common.c: Rename PF_ARCH12 to PF_Z14.
781         * config/s390/s390.h (enum processor_flags): Rename PF_ARCH12 to
782         PF_Z14.  Rename TARGET_CPU_ARCH12 to TARGET_CPU_Z14,
783         TARGET_CPU_ARCH12_P to TARGET_CPU_Z14_P, TARGET_ARCH12 to
784         TARGET_Z14, and TARGET_ARCH12_P to TARGET_Z14_P.
785         * config/s390/s390.md: Likewise. Rename also the cpu attribute
786         value from arch12 to z14.
788 2018-10-02  Uros Bizjak  <ubizjak@gmail.com>
790         * config/i386/i386.md (fxam<mode>2_i387_with_temp): Remove.
791         (isinfxf2): Ditto.
792         (isinf<mode>2): Ditto.
794 2018-10-02  Uros Bizjak  <ubizjak@gmail.com>
796         * config/i386/i386.c (ix86_emit_i387_round): Extend op1 to XFmode
797         before emitting fxam.  Perform calculations in XFmode.
799 2018-10-02  Marc Glisse  <marc.glisse@inria.fr>
801         * match.pd (((X /[ex] A) +- B) * A): New transformation.
803 2018-10-02  Marc Glisse  <marc.glisse@inria.fr>
805         PR middle-end/87319
806         * fold-const.c (fold_plusminus_mult_expr): Handle complex and vectors.
807         * tree.c (signed_or_unsigned_type_for): Handle complex.
809 2018-10-02  Jeff Law  <law@redhat.com>
811         * gimple-fold.c (get_range_strlen): Remove dead code.
813 2018-10-02  Martin Sebor  <msebor@redhat.com>
814             Jeff Law  <law@redhat.com>
816         * builtins.c (unterminated_array): Add new arguments.
817         If argument is not terminated, bubble up size and exact
818         state to callers.
819         (expand_builtin_strnlen): Detect, avoid expanding
820         and diagnose unterminated arrays.
821         (c_strlen): Fill in offset of start of unterminated strings.
822         * builtins.h (unterminated_array): Update prototype.
824 2018-10-02  Richard Biener  <rguenther@suse.de>
826         * config/i386/sse.md (reduc_plus_scal_v4df): Avoid the use
827         of haddv4df, first reduce to SSE width and exploit the fact
828         that we only need element zero with the reduction result.
829         (reduc_plus_scal_v2df): Likewise.
831 2018-10-02  Eric Botcazou  <ebotcazou@adacore.com>
833         * dojump.h (do_jump): Delete.
834         (do_jump_1): Likewise.
835         (split_comparison): Move around.
836         * dojump.c (do_jump): Make static.
837         (do_jump_1): Likewise.
838         (jumpifnot): Move around.
839         (jumpifnot_1): Likewise.
840         (jumpif): Likewise.
841         (jumpif_1): Likewise.
842         * expr.c (expand_expr_real_1): Call jumpif[not] instead of do_jump.
844 2018-10-02  Eric Botcazou  <ebotcazou@adacore.com>
846         * reorg.c (make_return_insns): Use emit_copy_of_insn_after for the
847         insns in the delay slot and add_insn_after for the jump insn.
849 2018-10-02  Richard Biener  <rguenther@suse.de>
851         * tree-inline.c (expand_call_inline): Use the location of
852         the callee declaration for the inline-entry marker.
853         * final.c (notice_source_line): Remove special-casing of
854         NOTE_INSN_INLINE_ENTRY.
856 2018-10-01  Carl Love  <cel@us.ibm.com>
858         PR 69431
859         * config/rs6000/rs6000-builtin.def (__builtin_mffsl): New.
860         (__builtin_mtfsb0): New.
861         (__builtin_mtfsb1): New.
862         ( __builtin_set_fpscr_rn): New.
863         (__builtin_set_fpscr_drn): New.
864         * config/rs6000/rs6000.c (rs6000_expand_mtfsb_builtin): Add.
865         (rs6000_expand_set_fpscr_rn_builtin): Add.
866         (rs6000_expand_set_fpscr_drn_builtin): Add.
867         (rs6000_expand_builtin): Add case statement entries for
868         RS6000_BUILTIN_MTFSB0, RS6000_BUILTIN_MTFSB1,
869         RS6000_BUILTIN_SET_FPSCR_RN, RS6000_BUILTIN_SET_FPSCR_DRN,
870         RS6000_BUILTIN_MFFSL.
871         (rs6000_init_builtins): Add ftype initialization and def_builtin
872         calls for __builtin_mffsl, __builtin_mtfsb0, __builtin_mtfsb1,
873         __builtin_set_fpscr_rn, __builtin_set_fpscr_drn.
874         * config/rs6000.md (rs6000_mtfsb0, rs6000_mtfsb1, rs6000_mffscrn,
875         rs6000_mffscdrn): Add define_insn.
876         (rs6000_set_fpscr_rn, rs6000_set_fpscr_drn): Add define_expand.
877         * doc/extend.texi: Add documentation for the builtins.
879 2018-10-01  Richard Biener  <rguenther@suse.de>
881         PR tree-optimization/87465
882         * tree-ssa-loop-ivcanon.c (tree_estimate_loop_size): Fix typo
883         causing branch miscounts.
885 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
887         * common/config/aarch64/aarch64-common.c (TARGET_OPTION_DEFAULT_PARAM,
888         aarch64_option_default_param):  New.
889         (params.h): Include.
890         (TARGET_OPTION_VALIDATE_PARAM, aarch64_option_validate_param): New.
891         * config/aarch64/aarch64.c (aarch64_override_options_internal): Simplify
892         stack-clash protection validation code.
894 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
896         * params.c (validate_param): New.
897         (add_params): Use it.
898         (set_param_value): Refactor param validation into validate_param.
899         (diagnostic.h): Include.
900         * diagnostic.h (diagnostic_ready_p): New.
902 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
904         * params.c (set_param_value):
905         Add index of parameter being validated.
906         * common/common-target.def (option_validate_param): New.
907         * common/common-targhooks.h (default_option_validate_param): New.
908         * common/common-targhooks.c (default_option_validate_param): New.
909         * doc/tm.texi.in (TARGET_OPTION_VALIDATE_PARAM): New.
910         * doc/tm.texi: Regenerate.
912 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
914         PR target/86486
915         * config/aarch64/aarch64.c (aarch64_override_options_internal):
916         Add validation for stack-clash parameters and set defaults.
918 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
920         PR target/86486
921         * configure.ac: Add stack-clash-protection-guard-size.
922         * doc/install.texi: Document it.
923         * config.in (DEFAULT_STK_CLASH_GUARD_SIZE): New.
924         * params.def: Update comment for guard-size.
925         (PARAM_STACK_CLASH_PROTECTION_GUARD_SIZE,
926         PARAM_STACK_CLASH_PROTECTION_PROBE_INTERVAL): Update description.
927         * configure: Regenerate.
929 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
931         PR target/86486
932         * config/aarch64/aarch64.h (STACK_CLASH_MIN_BYTES_OUTGOING_ARGS,
933         STACK_DYNAMIC_OFFSET): New.
934         * config/aarch64/aarch64.c (aarch64_layout_frame):
935         Update outgoing args size.
936         (aarch64_stack_clash_protection_alloca_probe_range,
937         TARGET_STACK_CLASH_PROTECTION_ALLOCA_PROBE_RANGE): New.
939 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
941         PR target/86486
942         * explow.c (anti_adjust_stack_and_probe_stack_clash): Support custom
943         probe ranges.
944         * target.def (stack_clash_protection_alloca_probe_range): New.
945         (stack_clash_protection_final_dynamic_probe): Remove.
946         * targhooks.h (default_stack_clash_protection_alloca_probe_range) New.
947         (default_stack_clash_protection_final_dynamic_probe): Remove.
948         * targhooks.c: Likewise.
949         * doc/tm.texi.in (TARGET_STACK_CLASH_PROTECTION_ALLOCA_PROBE_RANGE): New.
950         (TARGET_STACK_CLASH_PROTECTION_FINAL_DYNAMIC_PROBE): Remove.
951         * doc/tm.texi: Regenerate.
953 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
955         PR target/86486
956         * config/aarch64/aarch64-protos.h (aarch64_output_probe_sve_stack_clash): New.
957         * config/aarch64/aarch64.c (aarch64_output_probe_sve_stack_clash,
958         aarch64_clamp_to_uimm12_shift): New.
959         (aarch64_allocate_and_probe_stack_space): Add SVE specific section.
960         * config/aarch64/aarch64.md (probe_sve_stack_clash): New.
962 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
964         PR target/86486
965         * config/aarch64/aarch64.c (aarch64_layout_frame): Add assert.
967 2018-10-01  Jeff Law  <law@redhat.com>
968             Richard Sandiford <richard.sandiford@linaro.org>
969             Tamar Christina  <tamar.christina@arm.com>
971         PR target/86486
972         * config/aarch64/aarch64.md
973         (probe_stack_range): Add k (SP) constraint.
974         * config/aarch64/aarch64.h (STACK_CLASH_CALLER_GUARD,
975         STACK_CLASH_MAX_UNROLL_PAGES): New.
976         * config/aarch64/aarch64.c (aarch64_output_probe_stack_range): Emit
977         stack probes for stack clash.
978         (aarch64_allocate_and_probe_stack_space): New.
979         (aarch64_expand_prologue): Use it.
980         (aarch64_expand_epilogue): Likewise and update IP regs re-use criteria.
981         (aarch64_sub_sp): Add emit_move_imm optional param.
983 2018-10-01  MCC CS <deswurstes@users.noreply.github.com>
985         PR tree-optimization/87261
986         * match.pd: Remove trailing whitespace.
987         Add (x & y) | ~(x | y) -> ~(x ^ y),
988         (~x | y) ^ (x ^ y) -> x | ~y and (x ^ y) | ~(x | y) -> ~(x & y)
990 2018-10-01  Claudiu Zissulescu  <claziss@synopsys.com>
992         * config/arc/arc.md (*add_n): Clean up pattern, update instruction
993         constraints.
994         (ashlsi3_insn): Update instruction constraints.
995         (ashrsi3_insn): Likewise.
996         (rotrsi3): Likewise.
997         (add_shift): Likewise.
998         * config/arc/constraints.md (Csz): New 32 bit constraint. It
999         avoids placing in the limm field small constants which, otherwise,
1000         could end into a small instruction.
1002 2018-10-01  Claudiu Zissulescu  <claziss@synopsys.com>
1004         * config/arc/arc.md (maddsidi4_split): Don't use dmac if the
1005         destination register is not odd-even.
1006         (umaddsidi4_split): Likewise.
1008 2018-10-01  Richard Biener  <rguenther@suse.de>
1010         * tree-inline.c (expand_call_inline): Store origin of fn
1011         in BLOCK_ABSTRACT_ORIGIN for the inline BLOCK.
1012         * tree.c (block_ultimate_origin): Simplify and do some
1013         checking.
1015 2018-09-30  Uros Bizjak  <ubizjak@gmail.com>
1017         * config/i386/mmx.md (EMMS): New int iterator.
1018         (emms): New int attribute.
1019         (mmx_<emms>): Macroize insn from *mmx_emms and *mmx_femms using
1020         EMMS int iterator.  Explicitly declare clobbers.
1021         (mmx_emms): Remove expander.
1022         (mmx_femms): Ditto.
1023         * config/i386/predicates.md (emms_operation): Remove predicate.
1024         (vzeroall_pattern): New predicate.
1025         (vzeroupper_pattern): Rename from vzeroupper_operation.
1026         * config/i386/i386.c (ix86_avx_u128_mode_after): Use
1027         vzeroupper_pattern and vzeroall_pattern predicates.
1029 2018-09-30  Peter Bergner  <bergner@linux.ibm.com>
1031         PR rtl-optimization/86939
1032         * ira-lives.c (make_hard_regno_born): Rename from this...
1033         (make_hard_regno_live): ... to this.  Remove update to conflict
1034         information.  Update function comment.
1035         (make_hard_regno_dead): Add conflict information update.  Update
1036         function comment.
1037         (make_object_born): Rename from this...
1038         (make_object_live): ... to this.  Remove update to conflict information.
1039         Update function comment.
1040         (make_object_dead):  Add conflict information update.  Update function
1041         comment.
1042         (mark_pseudo_regno_live): Call make_object_live.
1043         (mark_pseudo_regno_subword_live): Likewise.
1044         (mark_hard_reg_dead): Update function comment.
1045         (mark_hard_reg_live): Call make_hard_regno_live.
1046         (process_bb_node_lives): Likewise.
1047         * lra-lives.c (make_hard_regno_born): Rename from this...
1048         (make_hard_regno_live): ... to this.  Remove update to conflict
1049         information.  Remove now uneeded check_pic_pseudo_p argument.
1050         Update function comment.
1051         (make_hard_regno_dead): Add check_pic_pseudo_p argument and add update
1052         to conflict information.  Update function comment.
1053         (mark_pseudo_live): Remove update to conflict information.  Update
1054         function comment.
1055         (mark_pseudo_dead): Add conflict information update.
1056         (mark_regno_live): Call make_hard_regno_live.
1057         (mark_regno_dead): Call make_hard_regno_dead with new arguement.
1058         (process_bb_lives): Call make_hard_regno_live and make_hard_regno_dead.
1060 2018-09-29  H.J. Lu  <hongjiu.lu@intel.com>
1062         PR target/87370
1063         * config/i386/i386.c (construct_container): Use TImode for
1064         BLKmode values in 2 integer registers.
1066 2018-09-29  Jeff Law  <law@redhat.com>
1068         * builtins.c (unterminated_array): Pass in c_strlen_data * to
1069         c_strlen rather than just a tree *.
1070         (c_strlen): Change NONSTR argument to a c_strlen_data pointer.
1071         Update recursive calls appropriately.  If caller did not provide a
1072         suitable data pointer, create a local one.  When a non-terminated
1073         string is discovered, bubble up information about the string via the
1074         c_strlen_data object.
1075         * builtins.h (c_strlen): Update prototype.
1076         (c_strlen_data): New structure.
1077         * gimple-fold.c (get_range_strlen): Update calls to c_strlen.
1078         For a type 2 call, if c_strlen indicates a non-terminated string
1079         use the length of the non-terminated string.
1080         (gimple_fold_builtin_stpcpy): Update calls to c_strlen.
1082 2018-09-29  Jakub Jelinek  <jakub@redhat.com>
1084         PR target/87467
1085         * config/i386/avx512fintrin.h (_mm512_abs_pd, _mm512_mask_abs_pd): Use
1086         __m512d type for __A argument rather than __m512.
1088 2018-09-28  John David Anglin  <danglin@gcc.gnu.org>
1090         * match.pd (simple_comparison): Don't optimize if either operand is
1091         a function pointer when target needs function pointer canonicalization.
1093 2018-09-28  Segher Boessenkool  <segher@kernel.crashing.org>
1095         * config/rs6000/driver-rs6000.c (asm_names): Adjust the entries for
1096         power5 .. power9 to remove indirection.
1097         * config/rs6000/rs6000.h (ASM_CPU_POWER5_SPEC, ASM_CPU_POWER6_SPEC,
1098         ASM_CPU_POWER7_SPEC, ASM_CPU_POWER8_SPEC, ASM_CPU_POWER9_SPEC,
1099         ASM_CPU_476_SPEC): Delete.
1100         (ASM_CPU_SPEC): Adjust.
1101         (EXTRA_SPECS): Delete asm_cpu_power5, asm_cpu_power6, asm_cpu_power7,
1102         asm_cpu_power8, asm_cpu_power9, asm_cpu_476.
1104 2018-09-28  Segher Boessenkool  <segher@kernel.crashing.org>
1106         * config.in: Delete HAVE_AS_DCI.
1107         * config/powerpcspe/powerpcspe.h: Treat HAVE_AS_DCI as always true.
1108         * config/rs6000/rs6000.h: Ditto.
1109         * configure.ac: Delete HAVE_AS_DCI.
1110         * configure: Regenerate.
1112 2018-09-28  Segher Boessenkool  <segher@kernel.crashing.org>
1114         * config.in (HAVE_AS_LWSYNC): Delete.
1115         * config/powerpcspe/powerpcspe.h (TARGET_LWSYNC_INSTRUCTION): Delete.
1116         * config/powerpcspe/sync.md (*lwsync): Always generate lwsync, never
1117         do it as a .long .
1118         * config/rs6000/rs6000.h (TARGET_LWSYNC_INSTRUCTION): Delete.
1119         * config/rs6000/sync.md (*lwsync): Always generate lwsync, never do it
1120         as a .long .
1121         * configure.ac: Delete HAVE_AS_LWSYNC.
1122         * configure: Regenerate.
1124 2018-09-28  Eric Botcazou  <ebotcazou@adacore.com>
1125             Pierre-Marie de Rodat  <derodat@adacore.com>
1127         * calls.c (expand_call): Try to do a tail call for thunks at -O0 too.
1128         * cgraph.h (struct cgraph_thunk_info): Add indirect_offset.
1129         (cgraph_node::create_thunk): Add indirect_offset parameter.
1130         (thunk_adjust): Likewise.
1131         * cgraph.c (cgraph_node::create_thunk): Add indirect_offset parameter
1132         and initialize the corresponding field with it.
1133         (cgraph_node::dump): Dump indirect_offset field.
1134         * cgraphclones.c (duplicate_thunk_for_node): Deal with indirect_offset.
1135         * cgraphunit.c (cgraph_node::analyze): Be prepared for external thunks.
1136         (thunk_adjust): Add indirect_offset parameter and deal with it.
1137         (cgraph_node::expand_thunk): Deal with the indirect_offset field and
1138         pass it to thunk_adjust.  Do not call the target hook if it's non-zero
1139         or if the thunk is external or local.  Fix formatting.  Do not chain
1140         the RESULT_DECL to BLOCK_VARS.  Pass the static chain to the target,
1141         if any, in the GIMPLE representation.
1142         * ipa-icf.c (sem_function::equals_wpa): Deal with indirect_offset.
1143         * lto-cgraph.c (lto_output_node): Write indirect_offset field.
1144         (input_node): Read indirect_offset field.
1145         * tree-inline.c (expand_call_inline): Pass indirect_offset field in the
1146         call to thunk_adjust.
1147         * tree-nested.c (struct nesting_info): Add thunk_p field.
1148         (create_nesting_tree): Set it.
1149         (convert_all_function_calls): Copy static chain from targets to thunks.
1150         (finalize_nesting_tree_1): Return early for thunks.
1151         (unnest_nesting_tree_1): Do not finalize thunks.
1152         (gimplify_all_functions): Do not gimplify thunks.
1154 2018-09-28  David Malcolm  <dmalcolm@redhat.com>
1156         * opt-suggestions.c (option_proposer::build_option_suggestions):
1157         Release "option_values".
1159 2018-09-28  David Malcolm  <dmalcolm@redhat.com>
1161         * coverage.c (get_coverage_counts): Convert problem-reporting dump
1162         messages from MSG_OPTIMIZED_LOCATIONS to MSG_MISSED_OPTIMIZATION.
1163         * dumpfile.c (kind_as_string): New function.
1164         (dump_loc): Rather than a hardcoded prefix of "note: ", use
1165         kind_as_string to vary the prefix based on dump_kind.
1166         (selftest::test_capture_of_dump_calls): Update for above.
1168 2018-09-28  Uros Bizjak  <ubizjak@gmail.com>
1170         * config/i386/i386.h (SSE_REGNO): Fix check for FIRST_REX_SSE_REG.
1171         (GET_SSE_REGNO): Rename from SSE_REGNO.  Update all uses for rename.
1173 2018-09-28  Uros Bizjak  <ubizjak@gmail.com>
1175         * config/i386/i386.h (CC_REGNO): Remove FPSR_REG.
1176         * config/i386/i386.c (ix86_fixed_condition_code_regs): Use
1177         INVALID_REGNUM instead of FPSR_REG.
1178         (ix86_md_asm_adjust): Do not clobber FPSR_REG.
1179         * config/i386/i386.md: Update comment of FP compares.
1180         (fldenv): Do not clobber FPSR_REG.
1182 2018-09-28  Richard Biener  <rguenther@suse.de>
1184         * tree.h (BLOCK_ORIGIN): New.
1185         * omp-expand.c (grid_expand_target_grid_body): Assign
1186         BLOCK_ORIGIN to BLOCK_ABSTRACT_ORIGIN.
1187         * tree-inline.c (remap_block): Likewise.
1188         * auto-profile.c (get_function_decl_from_block): Simplify
1189         by eliding the BLOCK_ABSTRACT_ORIGIN chasing.
1190         * langhooks.c (lhd_print_error_function): Likewise.
1191         * optinfo-emit-json.cc (optrecord_json_writer::inlining_chain_to):
1192         Likewise.
1193         * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
1194         * tree.c (block_nonartificial_location): Likewise.
1195         (block_ultimate_origin): Likewise.
1196         * tree-pretty-print.c (percent_K_format): Likewise.  Remove
1197         no longer needed LTO case.
1199 2018-09-28  Andrew Stubbs  <ams@codesourcery.com>
1200             Jan Hubicka  <jh@suse.cz>
1201             Martin Jambor  <mjambor@suse.cz>
1203         * simplify-rtx.c (simplify_merge_mask): New function.
1204         (simplify_ternary_operation): Use it, also see if VEC_MERGEs with the
1205         same masks are used in op1 or op2.
1206         (test_vec_merge): New function.
1207         (test_vector_ops): Call test_vec_merge.
1209 2018-09-28  Eric Botcazou  <ebotcazou@adacore.com>
1211         * config/sparc/sparc-protos.h (sparc_branch_cost): Declare.
1212         * config/sparc/sparc.h (BRANCH_COST): Call sparc_branch_cost.
1213         * config/sparc/sparc.c (struct processor_costs): Add branch_cost field.
1214         (cypress_costs): Set it.
1215         (supersparc_costs): Likewise.
1216         (hypersparc_costs): Likewise.
1217         (leon_cost): Likewise.
1218         (leon3_costs): Likewise.
1219         (sparclet_costs): Likewise.
1220         (ultrasparc_costs): Likewise.
1221         (ultrasparc_costs): Likewise.
1222         (niagara_costs): Likewise.
1223         (niagara2_costs): Likewise.
1224         (niagara3_costs): Likewise.
1225         (niagara4_costs): Likewise.
1226         (niagara7_costs): Likewise.
1227         (m8_costs): Likewise.
1228         (TARGET_CAN_FOLLOW_JUMP): Define.
1229         (pass_work_around_errata::gate): Minor tweak.
1230         (sparc_option_override): Remove MASK_FSMULD mask for V7 processors.
1231         Do not set both MASK_VIS4 and MASK_VIS4B for M8 processor.
1232         Automaitcally clear MASK_FSMULD mask for V7 processors.
1233         (sparc_can_follow_jump): New static function.
1234         (output_ubranch): Deal with CROSSING_JUMP_P.
1235         (sparc_use_sched_lookahead): Rewrite using switch statement.
1236         (sparc_issue_rate): Reorder.
1237         (sparc_branch_cost): New function.
1239 2018-09-27  Martin Sebor  <msebor@redhat.com>
1241         * tree.h (tree_to_shwi): Add attribute nonnull and pure.
1242         (tree_to_poly_int64, tree_to_uhwi, tree_to_poly_uint64): Same.
1243         (int_fits_type_p): Same.
1245 2018-09-27  Uros Bizjak  <ubizjak@gmail.com>
1247         * config/i386/i386.md (FPCR_REG): Remove.
1248         (UNSPEC_FLDCW): Remove.
1249         (x86_fnstcw_1): Use (const_int 0) instead of FPCR_REG.
1250         (x86_fldcw_1): Remove insn pattern.
1251         (fnstenv): Do not clobber FPCR_REG.
1252         (fldenv): Ditto.
1253         * config/i386/i386.h (FIXED_REGISTERS) Remove fpsr register.
1254         (CALL_USED_REGISTERS): Ditto.
1255         (REG_ALLOC_ORDER): Ditto.
1256         (REG_CLASS_CONTENTS): Ditto.
1257         (HI_REGISTER_NAMES): Ditto.
1258         (ADDITIONAL_REGISTER_NAMES): Use defines instead
1259         of numerical constants.
1260         * config/i386/i386.c (regclass_map): Remove fpsr register.
1261         (dbx_register_map): Ditto.
1262         (dbx64_register_map): Ditto.
1263         (svr4_dbx_register_map): Ditto.
1264         (print_reg): Do not handle FPCR_REG.
1266 2018-09-27  Segher Boessenkool  <segher@kernel.crashing.org>
1268         PR target/87149
1269         * config.in (HAVE_AS_CMPB, HAVE_AS_DFP, HAVE_AS_FPRND, HAVE_AS_MFPGPR,
1270         HAVE_AS_POPCNTB, HAVE_AS_POPCNTD, HAVE_AS_POWER8, HAVE_AS_POWER9):
1271         Delete, always treat as true.
1272         * config/powerpcspe/powerpcspe.c (rs6000_option_override_internal):
1273         Ditto.  Simplify remaining code.
1274         * config/powerpcspe/powerpcspe.h: Ditto.
1275         * config/rs6000/rs6000.c (rs6000_option_override_internal): Ditto.
1276         Simplify remaining code.
1277         (rs6000_expand_builtin): Ditto.
1278         * config/rs6000/rs6000.h: Ditto.
1279         * configure.ac: Ditto.
1280         * configure: Regenerate.
1282 2018-09-27  Martin Liska  <mliska@suse.cz>
1284         * coverage.c (get_coverage_counts): Revert the formatting
1285         of missing profile opt info.
1287 2018-09-27  Richard Biener  <rguenther@suse.de>
1289         PR debug/37801
1290         PR debug/87440
1291         * dwarf2out.c (set_block_origin_self): Do not mark outermost
1292         block as we do not output that.
1293         (gen_inlined_subroutine_die): Elide the originally outermost
1294         block, matching what we do for concrete instances.
1295         (decls_for_scope): Add parameter specifying whether to recurse
1296         to subblocks.
1298 2018-09-27  Andrew Stubbs  <ams@codesourcery.com>
1299             Tom de Vries  <tom@codesourcery.com>
1301         PR 82089
1303         * expmed.c (emit_cstore): Fix handling of result_mode == BImode and
1304         STORE_FLAG_VALUE == 1.
1306 2018-09-27  Andreas Krebbel  <krebbel@linux.ibm.com>
1308         * config/s390/s390.md (PPA_TX_ABORT, PPA_OOO_BARRIER): New
1309         constant definitions.
1310         ("tx_assist"): Replace magic number with PPA_TX_ABORT.
1311         ("*ppa"): Enable pattern also for -march=zEC12 -mno-htm.
1312         ("speculation_barrier"): New expander definition.
1314 2018-09-26  Indu Bhagat  <indu.bhagat@oracle.com>
1316         PR gcov-profile/86957
1317         * common.opt: New warning option -Wmissing-profile.
1318         * coverage.c (get_coverage_counts): Add warning for missing .gcda file.
1319         * doc/invoke.texi: Document -Wmissing-profile.
1321 2018-09-26  Jim Wilson  <jimw@sifive.com>
1323         * config/riscv/riscv.md (subsi3_extended2): Add J constraint.
1324         (negdi2, negsi2, negsi2_extended, negsi2_extended2): New.
1326 2018-09-26  Martin Sebor  <msebor@redhat.com>
1328         * tree.c (zerop): Change return type to bool.
1329         (integer_zerop, integer_onep, integer_each_onep): Same.
1330         (integer_all_onesp, integer_minus_onep, integer_pow2p): Same.
1331         (integer_nonzerop, integer_truep, tree_ctz, real_zerop): Same.
1332         (real_onep, real_minus_onep, chain_index): Same.
1333         (print_type_hash_statistics, type_list_equal): Same.
1334         * tree.h (zerop): Same.
1335         (zerop, integer_zerop, integer_onep, integer_each_onep): Same.
1336         (integer_all_onesp, integer_minus_onep, integer_pow2p): Same.
1337         (integer_nonzerop, integer_truep, tree_ctz, real_zerop): Same.
1338         (real_onep, real_minus_onep, chain_index): Same.
1339         (print_type_hash_statistics, type_list_equal): Same.
1341 2018-09-26  Jim Wilson  <jimw@sifive.com>
1343         * config/riscv/riscv.h (FUNCTION_ARG_REGNO_P): Fix comment.
1345 2018-09-26  Jakub Jelinek  <jakub@redhat.com>
1347         PR target/87414
1348         * config/i386/i386.c: Include debug.h and dwarf2out.h.
1349         (output_indirect_thunk): Emit DW_CFA_def_cfa_offset after the
1350         call.
1352 2018-09-25  Andrew Stubbs  <ams@codesourcery.com>
1354         * builtins.c (get_builtin_sync_mem): Force address mode conversion.
1356 2018-09-26  Uros Bizjak  <ubizjak@gmail.com>
1358         * config/i386/i386.h (enum reg_class): Remove FP_TOP_SSE_REGS
1359         and FP_SECOND_SSE_REGS.
1360         (REG_CLASS_NAMES): Ditto.
1361         (REG_CLASS_CONTENTS): Ditto.
1362         * config/i386/i386.c (ix86_preferred_reload_class) Do not handle
1363         FP_TOP_SSE_REGS and FP_SECOND_SSE_REGS classes.
1364         (ix86_preferred_output_reload_class): Ditto.
1365         * config/i386/i386.md (fix_trunc<mode>_i387_fisttp): Change "=&1f"
1366         clobber constraint to "=&f".
1367         (fix_truncdi_i387): Ditto.
1368         (lrintxfdi2): Ditto.
1369         (fistdi2_<rounding>): Ditto.
1370         (fpremxf4_i387): Change "=u" constraint to "=f".
1371         (fprem1xf4_i387): Ditto.
1372         (sincosxf3): Ditto.
1373         (fptanxf4_i387): Ditto.
1374         (fxtractxf3_i387): Ditto.
1375         (fscalexf4_i387): Ditto.
1376         (atan2xf3): Change "u" constraint to "f".
1377         (fyl2xxf3_i387): Ditto.
1378         (fyl2xp1xf3_i387): Ditto.
1380 2018-09-26  Uros Bizjak  <ubizjak@gmail.com>
1382         PR target/87439
1383         * config/i386/i386.h (NUM_MODES_FOR_MODE_SWITCHING): Update
1384         for removed I387_MASK_PM entity.
1387 2018-09-26  Jeff Law  <law@redhat.com>
1388         Revert
1389         2018-09-26  Alexey Neyman  <stilor@att.net>
1391         * graphite.h: Include <isl/id.h> and <isl/space.h>; these
1392         headers are no longer pulled in by <isl/val.h>.
1394 2018-09-26  Richard Biener  <rguenther@suse.de>
1396         PR debug/87443
1397         * dwarf2out.c (gen_lexical_block_die): Do not equate inline
1398         or concrete instance DIE to the tree.  Create abstract origin
1399         attributes also for concrete instances.
1401 2018-09-26  Alexey Neyman  <stilor@att.net>
1403         * graphite.h: Include <isl/id.h> and <isl/space.h>; these
1404         headers are no longer pulled in by <isl/val.h>.
1406 2018-09-26  Matthew Malcomson  <matthew.malcomson@arm.com>
1408         * config/arm/arm.c (arm_split_compare_and_swap, arm_split_atomic_op):
1409         Use new helper functions.
1410         * config/arm/sync.md (atomic_load<mode>, atomic_store<mode>):
1411         Use new helper functions.
1412         * config/arm/aarch-common-protos.h (aarch_mm_needs_acquire,
1413         aarch_mm_needs_release): New declarations.
1414         * config/arm/aarch-common.c (aarch_mm_needs_acquire,
1415         aarch_mm_needs_release): New.
1417 2018-09-26  Eric Botcazou  <ebotcazou@adacore.com>
1419         * config/arm/arm.c (arm_reorg): Skip Thumb reorg pass for thunks.
1420         (arm32_output_mi_thunk): Deal with long calls.
1422 2018-09-26  Richard Biener  <rguenther@suse.de>
1424         PR debug/87428
1425         PR debug/87362
1426         * tree-inline.c (expand_call_inline): When the location
1427         of the call is UNKNOWN_LOCATION use DECL_SOURCE_LOCATION
1428         or BUILTINS_LOCATION for the BLOCK_SOURCE_LOCATION of
1429         the inserted BLOCK to make inlined_function_outer_scope_p
1430         recognize it.
1431         * dwarf2out.c (add_call_src_coords_attributes): Do not add
1432         coords for reserved locations.
1434 2018-09-25  Segher Boessenkool  <segher@kernel.crashing.org>
1436         * config/rs6000/rs6000.md (*movcc_internal1): Use set_attr_alternative.
1437         (*call_indirect_nonlocal_sysv<mode>): Ditto.
1438         (*call_value_indirect_nonlocal_sysv<mode>): Ditto.
1439         (*sibcall_nonlocal_sysv<mode>): Ditto.
1440         (*sibcall_value_nonlocal_sysv<mode>): Ditto.
1441         (<bd>_<mode>): Ditto.
1442         (<bd>tf_<mode>): Ditto.
1444 2018-09-25  Segher Boessenkool  <segher@kernel.crashing.org>
1446         * config/rs6000/altivec.md (*altivec_mov<mode>): Write the output
1447         control string as a list of templates instead of as C code.
1448         (*altivec_movti): Ditto.
1449         * config/rs6000/darwin.md (movdf_low_di): Ditto.
1451 2018-09-25  Jim Wilson  <jimw@sifive.com>
1453         * config/riscv/riscv.c (riscv_split_symbol): Mark auipc label as weak
1454         when target symbol is weak.
1456 2018-09-25  Bernd Edlinger  <bernd.edlinger@hotmail.de>
1458         PR c/87387
1459         * builtins.c (unterminated_array): Simplify.
1460         * expr.c (string_constant): Handle SSA_NAME.  Add more exceptions
1461         where pointer arithmetic is safe.
1463 2018-09-25  Segher Boessenkool  <segher@kernel.crashing.org>
1465         PR target/86987
1466         * config/rs6000/altivec.md (altivec_vspltb): Use
1467         const_0_to_15_operand instead of u5bit_cint_operand.
1468         (*altivec_vspltb_internal): Ditto.
1469         (altivec_vspltb_direct): Ditto.
1470         (altivec_vsplth): Use const_0_to_7_operand instead of
1471         u5bit_cint_operand.
1472         (*altivec_vsplth_internal): Ditto.
1473         (altivec_vsplth_direct): Ditto.
1474         (altivec_vspltw): Use const_0_to_3_operand instead of
1475         u5bit_cint_operand.
1476         (*altivec_vspltw_internal): Ditto.
1477         (altivec_vspltw_direct): Ditto.
1478         (altivec_vspltsf): Ditto.
1479         (*altivec_vspltsf_internal): Ditto.
1480         * config/rs6000/rs6000.c (rs6000_expand_binop_builtin): Handle the
1481         various splats with the proper size immediate.  Reorder the various
1482         cases by ascending size of immediate, and put all such together.
1484 2018-09-25  Richard Biener  <rguenther@suse.de>
1486         PR debug/83941
1487         * dwarf2out.c (add_AT_external_die_ref): Remove now redundant
1488         GC-ification.
1489         (maybe_create_die_with_external_ref): Do not create
1490         DW_TAG_imported_unit here.
1491         (add_abstract_origin_attribute): Handle external BLOCK refs.
1492         (dwarf2out_abstract_function): Simplify LTO case.
1493         (dwarf2out_early_finish): Create DW_TAG_imported_unit explicitely
1494         rather than using maybe_create_die_with_external_ref.
1496 2018-09-25  Uros Bizjak  <ubizjak@gmail.com>
1498         PR target/71278
1499         * config/i386/i386.md (frndintxf2_mask_pm): Remove.
1500         (frndintxf2_mask_pm_i387): Ditto.
1501         (nearbyintxf2): Rewrite expander pattern to match rintxf2.
1502         Enable for !flag_trapping_math.
1503         (nearbyint<mode>2): Enable x87 modes for !flag_trapping_math.
1504         Enable SSE modes for TARGET_SSE4_1 and expand them with round insn.
1505         Change operand 1 predicate to nonimmediate_operand.
1506         (attr "i387_cw"): Remove mask_pm.
1507         * config/i386/i386.h (enum ix86_stack_slot): Remove SLOT_CW_MASK_PM.
1508         (enum ix86_entity): Remove I387_MASK_PM.
1509         * config/i386/i386.c (ix86_i387_mode_needed): Do not
1510         handle I387_MASK_PM.
1511         (ix86_mode_needed): Ditto.
1512         (ix86_mode_after): Ditto.
1513         (ix86_mode_entry): Ditto.
1514         (ix86_mode_exit): Ditto.
1515         (emit_i387_cw_initialization): Do not handle I387_CW_MASK_PM.
1517 2018-09-25  Jakub Jelinek  <jakub@redhat.com>
1519         * vr-values.c (vr_values::vr_values): Initialize to_remove_edges and
1520         to_update_switch_stmts to vNULL instead of calling create on them
1521         immediately.
1523 2018-09-25  Richard Biener  <rguenther@suse.de>
1525         PR tree-optimization/87402
1526         * tree-ssa-sccvn.c (SSA_VISITED): Remove unused function.
1527         (visit_phi): Re-instantiate handling of supposed to be VARYING
1528         but non-VARYING backedge value.
1530 2018-09-25  Richard Biener  <rguenther@suse.de>
1532         PR debug/83941
1533         * dwarf2out.c (struct sym_off_pair): New.
1534         (external_die_map): New global.
1535         (lookup_decl_die): When in LTO create DIEs lazily from the
1536         external_die_map.
1537         (lookup_block_die): New function, create DIEs lazily in LTO.
1538         (equate_block_to_die): New function.
1539         (dwarf2out_die_ref_for_decl): During WPA get the association
1540         from the external DIE map.
1541         (dwarf2out_register_external_die): Record mapping into the
1542         external DIE map.
1543         (maybe_create_die_with_external_ref): New function split out from
1544         DIE generation part of old dwarf2out_register_external_die.
1545         (add_abstract_origin_attribute): Do not return the DIE.  When
1546         in LTO reference externals directly.
1547         (dwarf2out_abstract_function): When in LTO ignore calls for
1548         decls with external DIEs (already present abstract instances).
1549         (gen_call_site_die): Adjust.
1550         (add_high_low_attributes): Likewise.
1551         (gen_lexical_block_die): Likewise.
1552         (gen_inlined_subroutine_die): Likewie.
1553         (gen_block_die): Likewise.
1554         (dwarf2out_inline_entry): Likewise.
1555         (dwarf2out_early_finish): In LTRANS phase create DW_TAG_imported_unit
1556         DIEs.
1558 2018-09-25  Martin Liska  <mliska@suse.cz>
1560         * ipa-fnsummary.c (estimate_node_size_and_time): Scale by two
1561         integers and not by a float value.
1563 2018-09-25  Martin Liska  <mliska@suse.cz>
1565         PR fortran/87394
1566         * dbgcnt.c (dbg_cnt_process_single_pair): Return false
1567         instead of NULL.
1568         * dumpfile.c (dump_enable_all): Remove extra parenthesis.
1569         * gcov-tool.c: Declare the function with ATTRIBUTE_NORETURN.
1570         * godump.c (go_format_type): Remove extra parenthesis.
1572 2018-09-25  Martin Liska  <mliska@suse.cz>
1574         * alias.c (set_dest_equal_p): Remove unused function.
1575         * config/i386/i386.c (def_builtin_pure2): Likewise.
1576         * diagnostic-show-locus.c (class layout): Remove
1577         unused field.
1578         (layout::layout): Likewise here.
1579         * dump-context.h (class temp_dump_context): Likewise.
1580         * dwarf2out.c (add_AT_fde_ref): Remove unused function.
1581         (add_AT_loclistsptr): Likewise.
1582         (add_AT_offset): Likewise.
1583         (get_AT_hi_pc): Likewise.
1584         (is_comdat_die): Likewise.
1585         (type_is_enum): Likewise.
1586         (ceiling): Likewise.
1587         (add_AT_vms_delta): Likewise.
1588         (is_class_die): Likewise.
1589         * edit-context.c (class line_event): Remove unused field.
1590         * graphite-sese-to-poly.c (tree_int_to_gmp): Remove
1591         unused function.
1592         * ipa-cp.c (ipa_get_vr_lat): Likewise.
1593         * lra-constraints.c (ok_for_index_p_nonstrict): Likewise.
1594         (ok_for_base_p_nonstrict): Likewise.
1595         * tree-chrec.c (is_not_constant_evolution): Likewise.
1596         (chrec_fold_poly_cst): Likewise.
1597         * tree-if-conv.c (has_pred_critical_p): Likewise.
1598         * tree-ssa-coalesce.c (print_exprs): Likewise.
1599         * tree-ssa-pre.c (bitmap_set_contains_expr): Likewise.
1600         * tree-ssa-uninit.c (is_and_or_or_p): Likewise.
1601         * tree-vrp.c (value_ranges_intersect_p): Likewise.
1602         (value_range_nonnegative_p): Likewise.
1604 2018-09-25  Martin Liska  <mliska@suse.cz>
1606         * config/powerpcspe/powerpcspe.c (rs6000_output_function_epilogue):
1607         Do not handle "GNU Pascal".
1608         * config/rs6000/rs6000.c (rs6000_output_function_epilogue):
1609         Likewise.
1610         * config/sparc/sparc.c (sparc_pass_by_reference): Remove Pascal
1611         from documentation. Likewise.
1612         * dbxout.c (dbxout_range_type): Likewise.
1613         * doc/cpp.texi: Likewise.
1614         * doc/extend.texi: Likewise.
1615         * doc/frontends.texi: Likewise.
1616         * doc/invoke.texi: Remove Pascal entry.
1617         * tree.def (CLEANUP_POINT_EXPR): Likewise.
1618         * doc/rtl.texi (MODE_FUNCTION): Remove not used entry.
1620 2018-09-25  Martin Liska  <mliska@suse.cz>
1622         PR middle-end/86078
1623         * doc/invoke.texi: Document all parameters and remove default
1624         of the parameters.
1626 2018-09-25  Ilya Leoshkevich  <iii@linux.ibm.com>
1628         PR bootstrap/87417
1629         * rtl.c (rtx_code_size): Take into account that EQ_ATTR_ALT
1630         contains HOST_WIDE_INTs when computing its size.
1632 2018-09-24  Jim Wilson  <jimw@sifive.com>
1634         PR target/87391
1635         * config/riscv/riscv.h (STACK_BOUNDARY): Test riscv_abi == ABI_ILP32E
1636         not TARGET_RVE.
1637         (ABI_STACK_BOUNDARY, MAX_ARGS_IN_REGISTERS): Likewise.
1639 2018-09-24  Andrew Pinski  <apinski@marvell.com>
1641         *  config/aarch64/aarch64.c (aarch_macro_fusion_pair_p): Don't
1642         access prev before checking it for NULLness in the
1643         AARCH64_FUSE_CMP_BRANCH case.
1645 2018-09-24  H.J. Lu  <hongjiu.lu@intel.com>
1647         PR target/82699
1648         * config/i386/i386.c (rest_of_insert_endbranch): Set
1649         endbr_queued_at_entrance to true and don't insert ENDBR if
1650         x86_function_profiler will be called.
1651         (x86_function_profiler): Insert ENDBR if endbr_queued_at_entrance
1652         is true.
1653         * config/i386/i386.h (machine_function): Add
1654         endbr_queued_at_entrance.
1656 2018-09-24  Ilya Leoshkevich  <iii@linux.ibm.com>
1658         * genattrtab.c (mk_attr_alt): Use alternative_mask.
1659         (attr_rtx_1): Adjust caching to match the new EQ_ATTR_ALT field
1660         types.
1661         (check_attr_test): Use alternative_mask.
1662         (get_attr_value): Likewise.
1663         (compute_alternative_mask): Use alternative_mask and XWINT.
1664         (make_alternative_compare): Use alternative_mask.
1665         (attr_alt_subset_p): Use XWINT.
1666         (attr_alt_subset_of_compl_p): Likewise.
1667         (attr_alt_intersection): Use alternative_mask and XWINT.
1668         (attr_alt_union): Likewise.
1669         (attr_alt_complement): Use HOST_WIDE_INT and XWINT.
1670         (mk_attr_alt): Use alternative_mask and HOST_WIDE_INT.
1671         (simplify_test_exp): Use alternative_mask and XWINT.
1672         (write_test_expr): Use alternative_mask and XWINT, adjust bit
1673         number calculation to support 64 bits.  Generate code that
1674         checks 64-bit masks.
1675         (main): Use alternative_mask.
1676         * rtl.def (EQ_ATTR_ALT): Change field types from ii to ww.
1678 2018-09-24  Ilya Leoshkevich  <iii@linux.ibm.com>
1680         PR target/80080
1681         * config/s390/s390.c (s390_emit_epilogue): Do not use PARALLEL
1682         RETURN+USE when returning via %r14.
1684 2018-09-24  Martin Liska  <mliska@suse.cz>
1686         * gcov.c (output_lines): Print colorization legend
1687         for both flag_use_colors and flag_use_hotness_colors.
1688         Reword the help.
1690 2018-09-24  Martin Liska  <mliska@suse.cz>
1692         * coverage.c (get_coverage_counts): Use warning_at
1693         with current_function_decl location. Use %qD in warning
1694         message.
1696 2018-09-24  Martin Liska  <mliska@suse.cz>
1698         * memory-block.h (memory_block_pool::release): Annotate with
1699         valgrind that the memory is not accessible.
1701 2018-09-24  Martin Liska  <mliska@suse.cz>
1703         PR sanitizer/85774
1704         * asan.c: Make asan_handled_variables extern.
1705         * asan.h: Likewise.
1706         * cfgexpand.c (expand_stack_vars): Make sure
1707         a representative is unpoison if another
1708         variable in the partition is handled by
1709         use-after-scope sanitization.
1711 2018-09-24  Richard Biener  <rguenther@suse.de>
1713         PR tree-optimization/63155
1714         * tree-ssa-propagate.c (add_ssa_edge): Avoid adding PHIs to
1715         the worklist when the edge of the respective argument isn't
1716         executable.
1718 2018-09-23   Uros Bizjak  <ubizjak@gmail.com>
1720         * config/i386/i386.h (enum reg_class): Rename MASK_REGS to
1721         ALL_MASK_REGS and MASK_EVEX_REGS to MASK_REGS.
1722         (MASK_CLASS_P): Update for rename.
1723         (MAYBE_MASK_CLASS_P): Ditto.
1724         (REG_CLASS_NAMES): Update.
1725         (REG_CLASS_CONTENT): Update.
1726         * config/i386/i386.c (regclass_map): Update for MASK_REG
1727         and ALL_MASK_REGS rename.
1728         * config/i386/constraints.md (Yk): Update for rename.
1729         (k): Ditto.
1731 2018-09-23   Uros Bizjak  <ubizjak@gmail.com>
1733         * config/i386/i386.h (enum reg_class): Remove
1734         EVEX_SSE_REGS and MOD4_SSE_REGS.
1735         (REG_CLASS_NAMES): Update.
1736         (REG_CLASS_CONTENT): Update.
1737         * config/i386/i386.c (regclass_map): Declare AVX-512 SSE
1738         registers as ALL_SSE_REGS.
1739         (ix86_additional_allocno_class_p): Remove.
1740         (TARGET_ADDITIONAL_ALLOCNO_CLASS_P): Remove.
1741         (ix86_register_priority): Lower priority of EVEX SSE registers.
1742         Use IN_RANGE macro where appropriate.
1743         (ix86_hard_regno_mode_ok): Merge AVX-5124FMAPS and
1744         AVX-5124VNNIW checks.
1745         (ix86_modes_tieable_p): Tie 512-bit SSE modes.
1746         * config/i386/sse.md (avx5124fmaddps_4fmaddps)
1747         (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddps_maskz)
1748         (avx5124fmaddps_4fmaddss, avx5124fmaddps_4fmaddss_mask)
1749         (avx5124fmaddps_4fmaddss_maskz, avx5124fmaddps_4fnmaddps)
1750         (avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddps_maskz)
1751         (avx5124fmaddps_4fnmaddss, avx5124fmaddps_4fnmaddss_mask)
1752         (avx5124fmaddps_4fnmaddss_maskz, avx5124vnniw_vp4dpwssd)
1753         (avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssd_maskz)
1754         (avx5124vnniw_vp4dpwssds, avx5124vnniw_vp4dpwssds_mask)
1755         (avx5124vnniw_vp4dpwssds_maskz): Use "v" instead of "Yh" constraint.
1756         * config/i386/constraints.md (Yh): Remove.
1758 2018-09-23   Uros Bizjak  <ubizjak@gmail.com>
1760         * config/i386/i386.c (regclass_map): Declare integer REX registers
1761         as GENERAL_REGS.
1763 2018-09-23  Gerald Pfeifer  <gerald@pfeifer.com>
1765         * doc/service.texi (Service): Switch the fsf.org link to https.
1767 2018-09-22  Chung-Ju Wu  <jasonwucj@gmail.com>
1769         PR target/86798
1770         * config/nds32/nds32.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
1771         Define to speculation_safe_value_not_needed.
1773 2018-09-21  Florian Weimer  <fweimer@redhat.com>
1775         PR middle-end/81035
1776         * doc/extend.texi (Common Function Attributes): Mention that
1777         noreturn suppresses tail call optimization.
1779 2018-09-21  Jeff Law  <law@redhat.com>
1781         * gimple-ssa-evrp.c (evrp_dom_walker::cleanup): Call
1782         vr_values::cleanup_edges_and_switches.
1783         * tree-vrp.c (to_remove_edges, to_update_switch_stmts): Moved into
1784         vr_values class.
1785         (identify_jump_threads): Remove EDGE_IGNORE handling.
1786         (execute_vrp): Move handling of to_remove_edges and
1787         to_update_switch_stmts into vr_values class member functions.
1788         * tree-vrp.h (switch_update, to_remove_edges): Remove declarations.
1789         (to_update_switch_stmts): Likewise.
1790         * vr-values.c: Include cfghooks.h.
1791         (vr_values::vr_values): Initialize to_remove_edges and
1792         to_update_switch_stmts.
1793         (vr_values::~vr_values): Verify to_remove_edges and
1794         to_update_switch_stmts are empty.
1795         (vr_values::simplify_switch_using_ranges): Set EDGE_IGNORE as needed.
1796         (vr_values::cleanup_edges_and_switches): New member function.
1797         * vr-values.h (vr_values): Add cleanup_edges_and_switches member
1798         function.  Add new data members.
1800 2018-09-21  David Malcolm  <dmalcolm@redhat.com>
1802         PR tree-optimization/87309
1803         * dumpfile.c (dump_context::begin_scope): Filter the dump_loc
1804         calls with pflags and alt_flags.
1805         (selftest::test_capture_of_dump_calls): Add test of interaction of
1806         MSG_OPTIMIZED_LOCATIONS with AUTO_DUMP_SCOPE.
1808 2018-09-21  Olivier Hainque  <hainque@adacore.com>
1810         * config.gcc: Factorize and comment inclusion of vxworks-dummy.h.
1812 2018-09-21  Olivier Hainque  <hainque@adacore.com>
1814         * config/vxworks.h (CLEAR_INSN_CACHE): #define to 1.
1816 2018-09-21  Olivier Hainque  <hainque@adacore.com>
1818         * config/vxworks.h (VXWORKS_LIBGCC_SPEC): Remove -lc_internal.
1819         Merge block comment with the one ahead of VXWORKS_LIBS_RTP. Then:
1820         (VXWORKS_LIBS_RTP): Minor reordering.
1822 2018-09-21  Olivier Hainque  <hainque@adacore.com>
1824         * config/vxworks.h (STARTFILE_PREFIX_SPEC): Define.
1825         (VXWORKS_LIBS_DIR_RTP): Remove definition and use.
1827 2018-09-21  Olivier Hainque  <hainque@adacore.com>
1829         * config/vxworks.h (SIZE_TYPE): Account for TARGET_VXWORKS64.
1830         (PTRDIFF_TYPE): Likewise.
1832 2018-09-21  Olivier Hainque  <hainque@adacore.com>
1834         * config.gcc: Enforce def of TARGET_VXWORKS64 to 1 from
1835         triplet, similar to support for VxWorks7.
1836         * config/vxworks-dummy.h: Provide a default definition
1837         of TARGET_VXWORKS64 to 0.
1839 2018-09-21  Olivier Hainque  <hainque@adacore.com>
1841         * config/vxworks.h (TARGET_VXWORKS7): Move default definition ...
1842         * config/vxworks-dummy.h: here.
1844 2018-09-21  Olivier Hainque  <hainque@adacore.com>
1846         * config.gcc: Prepend vxworks-dummy.h to tm_file for powerpc*
1848 2018-09-21  Shaokun Zhang  <zhangshaokun@hisilicon.com>
1849             Bo Zhou  <zbo.zhou@hisilicon.com>
1851         * config/aarch64/aarch64-cores.def (tsv110): New CPU.
1852         * config/aarch64/aarch64-tune.md: Regenerated.
1853         * doc/invoke.texi (AArch64 Options/-mtune): Add "tsv110".
1854         * config/aarch64/aarch64.c (tsv110_tunings): New tuning table.
1855         * config/aarch64/aarch64-cost-tables.h: Add "tsv110" extra costs.
1857 2018-09-21  Andrew Stubbs  <ams@codesourcery.com>
1858             Julian Brown  <julian@codesourcery.com>
1860         * builtins.c (get_builtin_sync_mem): Handle address spaces.
1862 2018-09-21  Eric Botcazou  <ebotcazou@adacore.com>
1864         * config/rs6000/rs6000.c (rs6000_function_ok_for_sibcall): Return false
1865         if the call takes a static chain.
1867 2018-09-21  Martin Liska  <mliska@suse.cz>
1869         * auto-profile.c (autofdo_source_profile::read): Do not
1870         set sum_all.
1871         (read_profile): Do not add working sets.
1872         (read_autofdo_file): Remove sum_all.
1873         (afdo_callsite_hot_enough_for_early_inline): Remove const
1874         qualifier.
1875         * coverage.c (struct counts_entry): Remove gcov_summary.
1876         (read_counts_file): Read new GCOV_TAG_OBJECT_SUMMARY,
1877         do not support GCOV_TAG_PROGRAM_SUMMARY.
1878         (get_coverage_counts): Remove summary and expected
1879         arguments.
1880         * coverage.h (get_coverage_counts): Likewise.
1881         * doc/gcov-dump.texi: Remove -w option.
1882         * gcov-dump.c (dump_working_sets): Remove.
1883         (main): Do not support '-w' option.
1884         (print_usage): Likewise.
1885         (tag_summary): Likewise.
1886         * gcov-io.c (gcov_write_summary): Do not dump
1887         histogram.
1888         (gcov_read_summary): Likewise.
1889         (gcov_histo_index): Remove.
1890         (gcov_histogram_merge): Likewise.
1891         (compute_working_sets): Likewise.
1892         * gcov-io.h (GCOV_TAG_OBJECT_SUMMARY): Mark
1893         it not obsolete.
1894         (GCOV_TAG_PROGRAM_SUMMARY): Mark it obsolete.
1895         (GCOV_TAG_SUMMARY_LENGTH): Adjust.
1896         (GCOV_HISTOGRAM_SIZE): Remove.
1897         (GCOV_HISTOGRAM_BITVECTOR_SIZE): Likewise.
1898         (struct gcov_summary): Simplify rapidly just
1899         to runs and sum_max fields.
1900         (gcov_histo_index): Remove.
1901         (NUM_GCOV_WORKING_SETS): Likewise.
1902         (compute_working_sets): Likewise.
1903         * gcov-tool.c (print_overlap_usage_message): Remove
1904         trailing empty line.
1905         * gcov.c (read_count_file): Read GCOV_TAG_OBJECT_SUMMARY.
1906         (output_lines): Remove program related line.
1907         * ipa-profile.c (ipa_profile): Do not consider GCOV histogram.
1908         * lto-cgraph.c (output_profile_summary): Do not stream GCOV
1909         histogram.
1910         (input_profile_summary): Do not read it.
1911         (merge_profile_summaries): And do not merge it.
1912         (input_symtab): Do not call removed function.
1913         * modulo-sched.c (sms_schedule): Do not print sum_max.
1914         * params.def (HOT_BB_COUNT_FRACTION): Reincarnate param that was
1915         removed when histogram method was invented.
1916         (HOT_BB_COUNT_WS_PERMILLE): Mention that it's used only in LTO
1917         mode.
1918         * postreload-gcse.c (eliminate_partially_redundant_load): Fix
1919         GCOV coding style.
1920         * predict.c (get_hot_bb_threshold): Use HOT_BB_COUNT_FRACTION
1921         and dump selected value.
1922         * profile.c (add_working_set): Remove.
1923         (get_working_sets): Likewise.
1924         (find_working_set): Likewise.
1925         (get_exec_counts): Do not work with working sets.
1926         (read_profile_edge_counts): Do not inform as sum_max is removed.
1927         (compute_branch_probabilities): Likewise.
1928         (compute_value_histograms): Remove argument for call of
1929         get_coverage_counts.
1930         * profile.h: Do not make gcov_summary const.
1932 2018-09-21  Monk Chiang  <sh.chiang04@gmail.com>
1934         * config.gcc (nds32*-*-*): Set TARGET_DEFAULT_TLSDESC_TRAMPOLINE=0.
1936 2018-09-21  Eric Botcazou  <ebotcazou@adacore.com>
1938         PR tree-optimization/86990
1939         * gimple-ssa-store-merging.c (imm_store_chain_info:coalesce_immediate):
1940         Check that the entire merged store group is made of constants only for
1941         overlapping stores.
1943 2018-09-20  Allan Sandfeld Jensen  <allan.jensen@qt.io>
1945         * gcc.c (LINK_COMMAND_SPEC): Handle -r like -nostdlib.
1946         (VTABLE_VERIFICATION_SPEC): Likewise.
1947         (SANITIZER_EARLY_SPEC): Likewise.
1948         (SANITIZER_SPEC): Likewise.
1949         * config/darwin.h (LINK_COMMAND_SPEC): Likewise.
1950         * doc/invoke.texi (Link Options): Document -r.
1952 2018-09-20  Richard Biener <rguenther@suse.de>
1954         PR middle-end/87054
1955         * gimplify.c (gimplify_expr): Retain alignment of
1956         addressable lvalue in dereference.
1958 2018-09-20  Alexandre Oliva <aoliva@redhat.com>
1960         PR bootstrap/87013
1961         * configure.ac: Check for .loc is_stmt support.
1962         * configure, config.in: Rebuilt.
1963         * dwarf2out.c (dwarf2out_source_line): Skip is_stmt
1964         if not supported.
1966 2018-09-20  Segher Boessenkool  <segher@kernel.crashing.org>
1968         * config/rs6000/rs6000.opt (misel=no, misel=yes): Delete.
1969         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -misel=yes and
1970         -misel=no.
1972 2018-09-20  Segher Boessenkool  <segher@kernel.crashing.org>
1974         * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
1975         VECTOR_OTHER.
1976         * config/rs6000/rs6000.c (rs6000_debug_vector_unit): Delete
1977         case VECTOR_OTHER.
1979 2018-09-20  Marek Polacek  <polacek@redhat.com>
1981         * doc/invoke.texi: Add -Wno-init-list-lifetime to C++ Language Options.
1983 2018-09-20  Richard Sandiford  <richard.sandiford@arm.com>
1985         PR tree-optimization/87288
1986         * tree-vect-loop.c (vect_analyze_loop_2): Take PEELING_FOR_GAPS
1987         into account when determining PEELING_FOR_NITERS.
1989 2018-09-20  Richard Sandiford  <richard.sandiford@arm.com>
1991         PR tree-optimization/86877
1992         * tree-vect-loop.c (vect_analyze_loop_2): Call
1993         vect_verify_datarefs_alignment.
1995 2018-09-19  Marek Polacek  <polacek@redhat.com>
1997         * doc/invoke.texi: Document -Wclass-conversion.
1999 2018-09-19  John David Anglin  <danglin@gcc.gnu.org>
2001         * config/pa/pa.c (pa_adjust_priority): Delete.
2002         (TARGET_SCHED_ADJUST_PRIORITY): Delete define.
2004         * config/pa/pa.md (atomic_storeqi): Restore deleted expander.
2005         (atomic_storehi): Likewise.
2006         (atomic_storesi): Likewise.
2007         (atomic_loaddi): Restore compare and swap exchange loop code.
2009 2018-09-19  Segher Boessenkool  <segher@kernel.crashing.org>
2011         PR rtl-optimization/86902
2012         * combine.c (try_combine): When changing the CC mode used, don't change
2013         an unrelated mode in other_insn to that new CC mode.
2015 2018-09-19  David Malcolm  <dmalcolm@redhat.com>
2017         * tree-data-ref.c (runtime_alias_check_p): Use formatted printing
2018         with %T in place of calls to dump_generic_expr.
2019         (prune_runtime_alias_test_list): Likewise.
2020         (create_runtime_alias_checks): Likewise.
2021         * tree-vect-data-refs.c (vect_check_nonzero_value): Likewise.
2022         (vect_analyze_data_ref_dependence): Likewise.
2023         (vect_slp_analyze_data_ref_dependence): Likewise.
2024         (vect_record_base_alignment): Likewise.  Use %G in place of call
2025         to dump_gimple_stmt.
2026         (vect_compute_data_ref_alignment): Likewise.
2027         (verify_data_ref_alignment): Likewise.
2028         (vect_find_same_alignment_drs): Likewise.
2029         (vect_analyze_group_access_1): Likewise.
2030         (vect_analyze_data_ref_accesses): Likewise.
2031         (dependence_distance_ge_vf): Likewise.
2032         (dump_lower_bound): Likewise.
2033         (vect_prune_runtime_alias_test_list): Likewise.
2034         (vect_find_stmt_data_reference): Likewise.
2035         (vect_analyze_data_refs): Likewise.
2036         (vect_create_addr_base_for_vector_ref): Likewise.
2037         (vect_create_data_ref_ptr): Likewise.
2038         * tree-vect-loop-manip.c (vect_set_loop_condition): Likewise.
2039         (vect_can_advance_ivs_p): Likewise.
2040         (vect_update_ivs_after_vectorizer): Likewise.
2041         (vect_gen_prolog_loop_niters): Likewise.
2042         (vect_prepare_for_masked_peels): Likewise.
2043         * tree-vect-loop.c (vect_determine_vf_for_stmt): Likewise.
2044         (vect_determine_vectorization_factor): Likewise.
2045         (vect_is_simple_iv_evolution): Likewise.
2046         (vect_analyze_scalar_cycles_1): Likewise.
2047         (vect_analyze_loop_operations): Likewise.
2048         (report_vect_op): Likewise.
2049         (vect_is_slp_reduction): Likewise.
2050         (check_reduction_path): Likewise.
2051         (vect_is_simple_reduction): Likewise.
2052         (vect_create_epilog_for_reduction): Likewise.
2053         (vect_finalize_reduction:): Likewise.
2054         (vectorizable_induction): Likewise.
2055         (vect_transform_loop_stmt): Likewise.
2056         (vect_transform_loop): Likewise.
2057         (optimize_mask_stores): Likewise.
2058         * tree-vect-patterns.c (vect_pattern_detected): Likewise.
2059         (vect_split_statement): Likewise.
2060         (vect_recog_over_widening_pattern): Likewise.
2061         (vect_recog_average_pattern): Likewise.
2062         (vect_determine_min_output_precision_1): Likewise.
2063         (vect_determine_precisions_from_range): Likewise.
2064         (vect_determine_precisions_from_users): Likewise.
2065         (vect_mark_pattern_stmts): Likewise.
2066         (vect_pattern_recog_1): Likewise.
2067         * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
2068         (vect_record_max_nunits): Likewise.
2069         (vect_build_slp_tree_1): Likewise.
2070         (vect_build_slp_tree_2): Likewise.
2071         (vect_print_slp_tree): Likewise.
2072         (vect_analyze_slp_instance): Likewise.
2073         (vect_detect_hybrid_slp_stmts): Likewise.
2074         (vect_detect_hybrid_slp_1): Likewise.
2075         (vect_slp_analyze_operations): Likewise.
2076         (vect_slp_analyze_bb_1): Likewise.
2077         (vect_transform_slp_perm_load): Likewise.
2078         (vect_schedule_slp_instance): Likewise.
2079         * tree-vect-stmts.c (vect_mark_relevant): Likewise.
2080         (vect_mark_stmts_to_be_vectorized): Likewise.
2081         (vect_init_vector_1): Likewise.
2082         (vect_get_vec_def_for_operand): Likewise.
2083         (vect_finish_stmt_generation_1): Likewise.
2084         (vect_check_load_store_mask): Likewise.
2085         (vectorizable_call): Likewise.
2086         (vectorizable_conversion): Likewise.
2087         (vectorizable_operation): Likewise.
2088         (vectorizable_load): Likewise.
2089         (vect_analyze_stmt): Likewise.
2090         (vect_is_simple_use): Likewise.
2091         (vect_get_vector_types_for_stmt): Likewise.
2092         (vect_get_mask_type_for_stmt): Likewise.
2093         * tree-vectorizer.c (increase_alignment): Likewise.
2095 2018-09-19  Andrew Stubbs  <ams@codesourcery.com>
2097         * doc/rtl.texi: Adjust vec_select description.
2098         * simplify-rtx.c (simplify_binary_operation_1): Allow VEC_SELECT to use
2099         non-constant selectors.
2101 2018-09-19  Matthew Malcomson  <matthew.malcomson@arm.com>
2103         * config/aarch64/aarch64-protos.h
2104         (aarch64_offset_9bit_signed_unscaled_p): New declaration.
2105         * config/aarch64/aarch64.md (arches): New "rcpc8_4" attribute value.
2106         (arch_enabled): Add check for "rcpc8_4" attribute value of "arch".
2107         * config/aarch64/aarch64.h (AARCH64_FL_RCPC8_4): New bitfield.
2108         (AARCH64_FL_FOR_ARCH8_4): Include AARCH64_FL_RCPC8_4.
2109         (AARCH64_FL_PROFILE): Move index so flags are ordered.
2110         (AARCH64_ISA_RCPC8_4): New flag.
2111         * config/aarch64/aarch64.c (offset_9bit_signed_unscaled_p): Renamed
2112         to aarch64_offset_9bit_signed_unscaled_p.
2113         * config/aarch64/atomics.md (atomic_store<mode>): Allow offset
2114         and use stlur.
2115         * config/aarch64/constraints.md (Ust): New constraint.
2116         * config/aarch64/predicates.md.
2117         (aarch64_9bit_offset_memory_operand): New predicate.
2118         (aarch64_rcpc_memory_operand): New predicate.
2120 2018-09-19  Eric Botcazou  <ebotcazou@adacore.com>
2122         PR rtl-optimization/87361
2123         * rtlanal.c (nonzero_bits1): Revert accidental change.
2125 2018-09-19  Richard Biener  <rguenther@suse.de>
2127         PR tree-optimization/87349
2128         PR tree-optimization/87342
2129         * tree-ssa-sccvn.c (do_rpo_vn): Iterate max_rpo computation.
2131 2018-09-18  Marek Polacek  <polacek@redhat.com>
2133         P1064R0 - Allowing Virtual Function Calls in Constant Expressions
2134         * gimple-fold.c (gimple_get_virt_method_for_vtable): Adjust assert.
2136 2018-09-18  Segher Boessenkool  <segher@kernel.crashing.org>
2138         * config/rs6000/rs6000.md: Remove old "Cygnus sibcall" comment.
2140 2018-09-18  Segher Boessenkool  <segher@kernel.crashing.org>
2142         PR rtl-optimization/86882
2143         * rtlanal.c (reg_overlap_mentioned_p): Handle CLOBBER.
2145 2018-09-18  Uros Bizjak  <ubizjak@gmail.com>
2147         * config/i386/i386.md (*<code>extend<mode>xf2): Macroize insn from
2148         *<code>extendsfxf2 and *<code>extenddfxf2 using MODEF mode iterator.
2150 2018-09-18  Jonathan Wakely  <jwakely@redhat.com>
2152         PR other/87353
2153         * doc/invoke.texi (Link Options): Fix formatting and grammar.
2155 2018-09-18  Richard Biener  <rguenther@suse.de>
2157         PR middle-end/63155
2158         * tree-ssa-coalesce.c (tree_int_map_hasher): Remove.
2159         (compute_samebase_partition_bases): Likewise.
2160         (coalesce_ssa_name): Always use compute_optimized_partition_bases.
2161         (gimple_can_coalesce_p): Simplify.
2163 2018-09-18  Hans-Peter Nilsson  <hp@bitrange.com>
2165         Handle a library implementation of ffs calling __builtin_ffs.
2166         * config/mmix/mmix.c (TARGET_INIT_LIBFUNCS): Override with...
2167         (mmix_init_libfuncs): New function: make __builtin_ffs expand
2168         to __ffsdi2.
2170 2018-09-17  David Malcolm  <dmalcolm@redhat.com>
2172         * diagnostic-show-locus.c (class layout_range): Add field
2173         "m_original_idx".
2174         (layout_range::layout_range): Add "original_idx" param and use it
2175         to initialize new field.
2176         (make_range): Use 0 for original_idx.
2177         (layout::layout): Pass in index to calls to
2178         maybe_add_location_range.
2179         (layout::maybe_add_location_range): Add param "original_idx" and
2180         pass it on to layout_range.
2181         (layout::print_any_labels): Pass on range->m_original_idx to
2182         get_text call.
2183         (gcc_rich_location::add_location_if_nearby): Use 0 for
2184         original_idx.
2185         * gcc-rich-location.h (text_range_label::get_text): Update for new
2186         param.
2187         (range_label_for_type_mismatch::get_text): Likewise.
2189 2018-09-17  Uros Bizjak  <ubizjak@gmail.com>
2191         * config/i386/i386.c (ix86_emit_i387_log1p): Emit fldln2 earlier.
2193 2018-09-17  David Malcolm  <dmalcolm@redhat.com>
2195         * gimple-ssa-sprintf.c (fmtwarn): Update for introduction of
2196         format_string_diagnostic_t.
2197         (fmtwarn_n): Likewise.
2198         * substring-locations.c
2199         (format_string_diagnostic_t::format_string_diagnostic_t) New ctor.
2200         (format_warning_n_va): Convert to...
2201         (format_string_diagnostic_t::emit_warning_n_va): ...this.
2202         (format_warning_va): Convert to...
2203         (format_string_diagnostic_t::emit_warning_va): ...this.
2204         (format_warning_at_substring): Convert to...
2205         (format_string_diagnostic_t::emit_warning): ...this.
2206         (format_warning_at_substring_n): Convert to...
2207         (format_string_diagnostic_t::emit_warning_n): ...this.
2208         * substring-locations.h (class format_string_diagnostic_t): New
2209         class.
2210         (format_warning_va): Convert to
2211         format_string_diagnostic_t::emit_warning_va.
2212         (format_warning_n_va): Convert to
2213         format_string_diagnostic_t::emit_warning_n_va.
2214         (format_warning_at_substring): Convert to
2215         format_string_diagnostic_t::emit_warning.
2216         (format_warning_at_substring_n): Convert to
2217         format_string_diagnostic_t::emit_warning_n.
2219 2018-09-17  Cesar Philippidis  <cesar@codesourcery.com>
2220             Bernd Schmidt <bernds_cb1@t-online.de>
2222         * config/nvptx/nvptx.md (atomic_fetch_<logic><mode>): Enable with
2223         SImode args.
2225 2018-09-17  Uros Bizjak  <ubizjak@gmail.com>
2227         * config/i386/i386.md (truncxf<mode>2_i387_noop_unspec): Change
2228         operand 0 predicate to nonimmediate operand.
2229         (rint<mode>2_frndint): Remove insn pattern.
2230         (rint<mode>2): Change operand 1 predicate to general_operand.
2231         Extend operand 1 to XFmode and generate rintxf2 insn.
2232         (frndintxf2_<rounding>): Rename from frndint<mode>2_<rounding>.
2233         Do not use X87MODEF mode macro.
2234         (frndintxf2_<rounding>_i387): Rename from
2235         frndint<mode>2_<rounding>_i387.  Do not use X87MODEF mode macro.
2236         (<rounding_insn><mode>2): For non-SSE modes, extend operand 1
2237         to XFmode and generate significandxf3 insn.
2239 2018-09-17  Richard Biener  <rguenther@suse.de>
2241         PR tree-optimization/87328
2242         * tree-ssa-sccvn.c (process_bb): Remove assertion about not
2243         visiting unexecutable backedges when not iterating.
2244         (do_rpo_vn): Mark all edges not executable even when not
2245         iterating.
2247 2018-09-17  Martin Jambor  <mjambor@suse.cz>
2249         PR c/63886
2250         * doc/invoke.texi (Warning Options): Likewise.
2252 2018-09-17  Richard Biener  <rguenther@suse.de>
2254         PR tree-optimization/87301
2255         * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_cleanup): Properly
2256         clean EH info from leftover copy assignments.
2258 2018-09-17  Martin Liska  <mliska@suse.cz>
2260         PR gcov-profile/85871
2261         * gcov.c (output_intermediate_file): Fix out of bounds
2262         access.
2264 2018-09-17  Vineet Gupta  <vgupta@synopsys.com>
2266         * config/arc/arc.c: Object attributes for core4 not reflected
2267         correctly.
2268         * config/arc/arc.h: Don't restrict DBNZ to core3 (core4 includes
2269         core3).
2271 2018-09-17  Alexey Brodkin  <abrodkin@synopsys.com>
2273         * config/arc/linux.h (LINK_EH_SPEC): Add missing space.
2275 2018-09-17  Martin Liska  <mliska@suse.cz>
2277         * doc/gcov.texi: Document new option --use-hotness-colors.
2278         * gcov.c (struct source_info): Declare new field.
2279         (source_info::source_info): Set default for maximum_count.
2280         (print_usage): Add new -q option.
2281         (process_args): Process it.
2282         (accumulate_line_info): Save src->maximum_count.
2283         (output_line_beginning): Make color line number if
2284         flag_use_hotness_colors is set.
2285         (output_line_details): Pass default argument value.
2286         (output_lines): Pass src->maximum_count.
2288 2018-09-17  Martin Liska  <mliska@suse.cz>
2290         * common/config/i386/i386-common.c (ix86_get_valid_option_values):
2291         Use processor_names table.
2292         * config/i386/i386.c (ix86_default_align): Use
2293         processor_cost_table for alignment values.
2294         (ix86_option_override_internal): Use processor_names.
2295         (ix86_function_specific_print): Likewise.
2296         * config/i386/i386.h (struct processor_costs):
2297         Add alignment values.
2298         (struct ptt): Remove and replace with const char *.
2299         * config/i386/x86-tune-costs.h (struct processor_costs):
2300         Declare default alignments for all costs.
2302 2018-09-17  Aldy Hernandez  <aldyh@redhat.com>
2304         * tree-vrp.c (extract_range_from_unary_expr): Do not special case
2305         symbolics or VR_VARYING ranges for ABS_EXPR.
2306         * wide-int-range.cc (wide_int_range_abs): Return positive numbers
2307         when range will wrap.
2309 2018-09-15  Eric Botcazou  <ebotcazou@adacore.com>
2311         PR middle-end/86864
2312         * cfgexpand.c (expand_gimple_basic_block): Be prepared for a BARRIER
2313         before and after a JUMP_TABLE_DATA.
2315 2018-09-14  John David Anglin  <danglin@gcc.gnu.org>
2317         PR middle-end/87188
2318         * dojump.c (do_compare_and_jump): Canonicalize function pointers
2319         when one operand is a function pointer.  Use POINTER_TYPE_P and
2320         FUNC_OR_METHOD_TYPE_P.
2321         * expr.c (do_store_flag): Use POINTER_TYPE_P and FUNC_OR_METHOD_TYPE_P.
2322         * fold-const.c (build_range_check): Likewise.
2323         * match.pd (simple_comparison): Likewise.
2325 2018-09-14  David Malcolm  <dmalcolm@redhat.com>
2327         PR c/82967
2328         * spellcheck.c (get_edit_distance_cutoff): New function.
2329         (selftest::test_edit_distance_unit_test_oneway): Rename to...
2330         (selftest::test_get_edit_distance_one_way): ...this.
2331         (selftest::test_get_edit_distance_unit): Rename to...
2332         (selftest::test_get_edit_distance_both_ways): ...this.
2333         (selftest::test_edit_distances): Move tests to this new function,
2334         and test some more pairs of strings.  Update for above renaming.
2335         (selftest::get_old_cutoff): New function.
2336         (selftest::test_get_edit_distance_cutoff): New function.
2337         (selftest::assert_suggested_for): New function.
2338         (ASSERT_SUGGESTED_FOR): New macro.
2339         (selftest::assert_not_suggested_for): New function.
2340         (ASSERT_NOT_SUGGESTED_FOR): New macro.
2341         (selftest::test_suggestions): New function.
2342         (selftest::spellcheck_c_tests): Move test_get_edit_distance_unit
2343         tests to selftest::test_edit_distances and call it.  Add calls to
2344         selftest::test_get_edit_distance_cutoff and
2345         selftest::test_suggestions.
2346         * spellcheck.h (get_edit_distance_cutoff): New function declaration.
2347         (best_match::consider): Replace hard-coded cutoff calculation with
2348         a call to...
2349         (best_match::get_cutoff): New declaration.
2350         (best_match::get_best_meaningful_candidate): Likewise.
2352 2018-09-14  Bernd Edlinger  <bernd.edlinger@hotmail.de>
2354         * builtins.c (fold_builtin_strlen): Remove TODO comment.
2356 2018-09-14  Bernd Edlinger  <bernd.edlinger@hotmail.de>
2358         revert:
2359         2018-07-30  Bernd Edlinger  <bernd.edlinger@hotmail.de>
2361         * tree-ssa-forwprop.c (simplify_builtin_call): Don't create a not NUL
2362         terminated string literal.
2364 2018-09-14  Martin Sebor  <msebor@redhat.com>
2366         * builtins.c (unterminated_array): Handle ARRAY_REF.
2367         (expand_builtin_stpcpy_1): Detect unterminated char arrays.
2368         * builtins.h (unterminated_array): Declare extern.
2369         * gimple-fold.c (gimple_fold_builtin_stpcpy): Detect unterminated
2370         arrays.
2371         (gimple_fold_builtin_sprintf): Propagate NO_WARNING to transformed
2372         calls.
2374 2018-09-14  Martin Sebor  <msebor@redhat.com>
2375             Jeff Law  <law@redhat.com>
2377         * builtins.c (unterminated_array): New.
2378         (expand_builtin_strcpy): Adjust.
2379         (expand_builtin_strcpy_args): Detect unterminated arrays.
2380         * gimple-fold.c (get_maxval_strlen): Add argument.  Detect
2381         unterminated arrays.
2382         * gimple-fold.h (get_maxval_strlen): Add argument.
2383         (gimple_fold_builtin_strcpy): Detec unterminated arrays.
2385         * gimple-fold.c (get_range_strlen): Add argument.
2386         (get_maxval_strlen): Adjust.
2387         * gimple-fold.h (get_range_strlen): Add argument.
2389 2018-09-14  Wei Xiao  <wei3.xiao@intel.com>
2391         * config/i386/movdirintrin.h: Fix copyright year.
2393 2018-09-14  Uros Bizjak  <ubizjak@gmail.com>
2395         * reg-stack.c: Include regs.h.
2396         (replace_reg): Assert that mode is MODE_FLOAT or MODE_COMPLEX_FLOAT.
2397         (emit_pop_insn): Default pop insn mode to the reg_raw_mode of
2398         FIRST_STACK_REG, not DFmode.
2399         (emit_swap_insn): Default swap insn mode to the reg_raw_mode of
2400         FIRST_STACK_REG, not XFmode.  Explicitly construct swap RTX.
2401         (change stack): Default register mode to the reg_raw_mode of
2402         FIRST_STACK_REG, not DFmode.
2403         * config/i386/i386.md (*swap<mode>): Remove insn pattern.
2404         (*swapxf): Rename from swapxf.
2406 2018-09-14  Carl Love  <cel@us.ibm.com>
2408         * config/rs6000/emmintrin.h: Add _MM_SHUFFLE2.
2409         * config/rs6000/xmmintrin.h: Add _MM_SHUFFLE.
2411 2018-09-14  Segher Boessenkool  <segher@kernel.crashing.org>
2413         PR target/87224
2414         * config/rs6000/rs6000.md (*mov<mode>_hardfloat64): Add Z to the Y
2415         alternatives.
2417 2018-09-14  Sam Tebbs  <sam.tebbs@arm.com>
2419         PR target/85628
2420         * config/aarch64/aarch64.md (*aarch64_bfxilsi_uxtw): Define.
2422 2018-09-14  Jason Merrill  <jason@redhat.com>
2424         Fix --enable-gather-detailed-mem-stats.
2425         * hash-table.c (hash_table_usage): Change from variable to function.
2426         * hash-table.h: Adjust.
2427         * Makefile.in: Add missing dependencies on hash-table.h.
2429 2018-09-14  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
2431         PR tree-optimization/87259
2432         PR lto/87283
2433         (pass_cse_reciprocals::execute): Run optimize_recip_sqrt after
2434         execute_cse_reciprocals_1 has tried transforming.
2436 2018-09-14  Aldy Hernandez  <aldyh@redhat.com>
2438         * tree-vrp.c (extract_range_from_binary_expr_1): Normalize
2439         VR_VARYING for PLUS/MINUS_EXPR.
2441 2018-09-14  Ilya Leoshkevich  <iii@linux.ibm.com>
2443         * config/s390/s390-passes.def (INSERT_PASS_BEFORE): Improve
2444         formatting.
2446 2018-09-14  Richard Biener  <rguenther@suse.de>
2448         PR middle-end/63155
2449         * tree-ssa-coalesce.c (ssa_conflicts_merge): Remove conflict
2450         bits for the merged partition.
2452 2018-09-13  Martin Sebor  <msebor@redhat.com>
2453             Bernd Edlinger  <bernd.edlinger@hotmail.de>
2455         * builtins.h (c_srlen): Add argument.
2456         * builtins.c (warn_string_no_nul): New function.
2457         (c_strlen): Add argument and use it.  Update recursive calls.
2458         Pass DECL argument to string_constant to get info on non
2459         terminated strings.  Update *NONSTR as needed.
2460         (fold_builtin_strlen): Add argument to calls to c_strlen.
2461         Warn for unterminated arrays.
2462         (warn_string_no_null): Add prototype.
2463         * expr.c (string_constant): Update arguments.  Update recursive
2464         calls appropriately.  Detect missing NUL terminator and outermost
2465         declaration its missing in.
2466         Improve checks for arrays with nonzero lower bound or elements
2467         that are not a single byte.  Simplify offset computation.
2468         Simplify checks for non-NUL terminated strings.
2469         * gimple-fold.c (get_range_strlen): Add argument to c_strlen call.
2470         * gimple-ssa-sprintf.c (get_string_length): Remove unnecessary code.
2472 2018-09-13  Bernd Edlinger  <bernd.edlinger@hotmail.de>
2474         * builtins.c (c_strlen): Handle not zero terminated STRING_CSTs
2475         correctly.
2476         * fold-const.c (c_getstr): Fix function comment.  Remove unused third
2477         argument.  Fix range checks.
2478         * fold-const.h (c_getstr): Adjust protoype.
2479         * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid folding when
2480         string is constant but contains no NUL byte.
2482         * expr.c (string_constant): Adjust function comment.
2483         Remove bogus check for zero termination.
2485         * fold-const.c (c_getstr): Clamp STRING_LENGTH to STRING_SIZE.
2487         * varasm.c (compare_constant): Compare type size of STRING_CSTs.
2488         (get_constant_size): Don't make STRING_CSTs larger than they are.
2489         (check_string_literal): New check function for STRING_CSTs.
2490         (output_constant): Use it.
2492 2018-09-13  Eric Botcazou  <ebotcazou@adacore.com>
2494         PR target/86812
2495         * config/visium/visium.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define.
2497 2018-09-13  Richard Biener  <rguenther@suse.de>
2499         PR tree-optimization/87263
2500         * tree-ssa-sccvn.c (visit_phi): Revert some earlier changes.
2501         (struct unwind_state): Add max_rpo field.
2502         (do_rpo_vn): Allow up-to-date loop state to be used when not iterating.
2503         Compute max_rpo, the max RPO number a block can be backwards reached
2504         from.  Re-write non-iterating mode to a RPO ordered worklist approach,
2505         separating it from the iterating mode.
2507 2018-09-13  Vlad Lazar  <vlad.lazar@arm.com>
2509         * haifa-sched.c (rank_for_schedule): Schedule by INSN_COST.
2510         (rfs_decision): New scheduling decision.
2512 2018-09-13  Richard Biener  <rguenther@suse.de>
2514         PR bootstrap/87134
2515         * tree-ssa-sccvn.c (vn_nary_op_insert_into): Fix assert.
2516         (vn_nary_op_insert_pieces_predicated): Do not write useless
2517         valid_dominated_by_p entry outside of the allocated storage.
2519 2018-09-13  Omar Sandoval  <osandov@osandov.com>
2520             Tom de Vries  <tdevries@suse.de>
2522         PR debug/86985
2523         * dwarf2out.c (is_c): New function.
2524         (add_subscript_info): Add DW_AT_count of 0 for C zero-length arrays.
2526 2018-09-13  Sam Tebbs  <sam.tebbs@arm.com>
2528         PR target/85628
2529         * config/aarch64/aarch64.md (*aarch64_bfxil):
2530         Define.
2531         * config/aarch64/constraints.md (Ulc): Define.
2532         * config/aarch64/aarch64-protos.h (aarch64_high_bits_all_ones_p):
2533         Define.
2534         * config/aarch64/aarch64.c (aarch64_high_bits_all_ones_p):
2535         New function.
2537 2018-09-13  Vlad Lazar  <vlad.lazar@arm.com>
2539         * config/aarch64/aarch64.h (TARGET_COMPUTE_FRAME_LAYOUT): Define.
2540         * config/aarch64/aarch64.c (aarch64_expand_prologue): Remove
2541         aarch64_layout_frame call.
2542         (aarch64_expand_epilogue): Likewise.
2543         (aarch64_initial_elimination_offset): Likewise.
2544         (aarch64_get_separate_components): Likewise.
2545         (aarch64_use_return_insn_p): Likewise.
2546         (aarch64_layout_frame): Remove unneeded check.
2548 2018-09-13  Jakub Jelinek  <jakub@redhat.com>
2550         * configure.ac: Only append
2551         " : (reconfigured) $TOPLEVEL_CONFIGURE_ARGUMENTS" to
2552         gcc_config_arguments if it was never reconfigured or last reconfigure
2553         was with different arguments.
2554         * configure: Regenerated.
2556 2018-09-13  Jakub Jelinek  <jakub@redhat.com>
2557             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
2559         PR middle-end/87290
2560         * expr.c (maybe_optimize_pow2p_mod_cmp): New function.
2561         (maybe_optimize_mod_cmp): Use it if integer_pow2p treeop1.
2563 2018-09-13  Jakub Jelinek  <jakub@redhat.com>
2565         PR tree-optimization/87287
2566         * fold-const.c (fold_binary_loc) <case EQ_EXPR>: Move signed modulo
2567         X % C == 0 to X % (unsigned) C == 0 optimization to ...
2568         * match.pd (X % C == 0): ... here.  New optimization.
2570 2018-09-12  Jakub Jelinek  <jakub@redhat.com>
2572         PR middle-end/82853
2573         * expr.h (maybe_optimize_mod_cmp): Declare.
2574         * expr.c (mod_inv): New function.
2575         (maybe_optimize_mod_cmp): New function.
2576         (do_store_flag): Use it.
2577         * cfgexpand.c (expand_gimple_cond): Likewise.
2579 2018-09-09  Cesar Philippidis  <cesar@codesourcery.com>
2580             Julian Brown  <julian@codesourcery.com>
2582         PR middle-end/86336
2583         * gimplify.c (gimplify_scan_omp_clauses): Set
2584         target_firstprivatize_array_bases in OpenACC parallel and kernels
2585         region contexts.  Remove GOMP_MAP_FIRSTPRIVATE_REFERENCE clauses from
2586         OpenACC data regions.
2588 2018-09-12  Uros Bizjak  <ubizjak@gmail.com>
2590         * config/i386/i386.md (sqrt_extend<mode>xf3_i387): Remove.
2591         (sqrt<mode>2): Extend operand 1 to XFmode and generate sqrtxf3 insn.
2593 2018-09-12  Richard Biener  <rguenther@suse.de>
2595         PR tree-optimization/87280
2596         * tree-ssa-sccvn.c (process_bb): Handle the case of executable
2597         edge but unreachable target.
2598         (do_rpo_vn): For conservatively handling a PHI only mark
2599         the backedge executable but not the block reachable.
2601 2018-09-12  Richard Biener  <rguenther@suse.de>
2603         PR tree-optimization/87266
2604         * tree-ssa-sccvn.c (do_rpo_vn): Always iterate to not yet
2605         visited blocks.
2607 2018-09-12  Andreas Krebbel  <krebbel@linux.ibm.com>
2609         * config/s390/s390.md (PFPO_RND_MODE_DFP, PFPO_RND_MODE_BFP): New
2610         constants.
2611         ("trunc<BFP:mode><DFP_ALL:mode>2")
2612         ("trunc<DFP_ALL:mode><BFP:mode>2")
2613         ("extend<BFP:mode><DFP_ALL:mode>2")
2614         ("extend<DFP_ALL:mode><BFP:mode>2"): Set proper rounding mode
2615         according to the target operand type.
2617 2018-09-12  Jakub Jelinek  <jakub@redhat.com>
2618             Andreas Krebbel  <krebbel@linux.ibm.com>
2620         PR tree-optimization/86844
2621         * gimple-ssa-store-merging.c
2622         (imm_store_chain_info::coalesce_immediate): For overlapping stores, if
2623         there are any overlapping stores in between them, make sure they are
2624         also coalesced or we give up completely.
2626 2018-09-12  Jakub Jelinek  <jakub@redhat.com>
2628         PR middle-end/87248
2629         * fold-const.c (fold_ternary_loc) <case COND_EXPR>: Verify also that
2630         BIT_AND_EXPR's second operand is a power of two.  Formatting fix.
2632 2018-09-12  Tom de Vries  <tdevries@suse.de>
2634         * common.opt (gdescribe-dies): Add option.
2635         * dwarf2out.c (add_name_and_src_coords_attributes): Add description
2636         attribute for artifical and nameless decls.
2637         (dwarf2out_register_external_die): Add description attribute to
2638         external reference die.
2639         (add_desc_attribute): New functions.
2640         (gen_subprogram_die): Add description attribute to
2641         DW_TAG_call_site_parameter.
2642         * tree-pretty-print.c (print_generic_expr_to_str): New function.
2643         * tree-pretty-print.h (print_generic_expr_to_str): Declare.
2644         * doc/invoke.texi (@item Debugging Options): Add -gdescribe-dies and
2645         -gno-describe-dies.
2646         (@item -gdescribe-dies): Add.
2648 2018-09-12  Aldy Hernandez  <aldyh@redhat.com>
2650         * tree-vrp.c (vrp_shift_undefined_p): Remove.
2651         (extract_range_from_binary_expr_1: Call
2652         wide_int_range_shift_undefined_p instead of vrp_shift_undefined_p.
2653         * wide-int-range.h (wide_int_range_shift_undefined_p): Do not
2654         depend on sign.
2656 2018-09-12  Aldy Hernandez  <aldyh@redhat.com>
2658         * gimple-ssa-warn-alloca.c
2659         (alloca_type_and_limit::alloca_type_and_limit): Initialize limit
2660         field for ALLOCA_BOUND_*_LARGE.
2662 2018-09-11  Nathan Sidwell  <nathan@acm.org>
2664         * gcc.c (load_specs, execute, run_attempt): Use %qs not '%s'.
2666 2018-09-11  Uros Bizjak  <ubizjak@gmail.com>
2668         * reg-stack.c (subst_asm_stack_regs): Call replace_reg also
2669         for clobbers.  Remove obsolete comment.
2671 2018-09-11  Uros Bizjak  <ubizjak@gmail.com>
2673         * config/i386/i386.md (define_attr "type"): Remove mpxmov, mpxmk,
2674         mpxchk, mpxld and mpxst types.
2675         (define_attr length_immediate): Remove all processing of mpx types.
2676         (define_attr prefix_0f): Ditto.
2677         (define_attr memory): Ditto.
2679 2018-09-11  Uros Bizjak  <ubizjak@gmail.com>
2681         * config/i386/i386.md (fyl2x_extend<mode>xf3_i387): Remove.
2682         (log<mode>2): Change operand 1 predicate to general_operand.
2683         Extend operand 1 to XFmode and generate logxf3 insn.
2684         (log10<mode>2): Change operand 1 predicate to general_operand.
2685         Extend operand 1 to XFmode and generate log10xf3 insn.
2686         (log2<mode>2): Change operand 1 predicate to general_operand.
2687         Extend operand 1 to XFmode and generate log2xf3 insn.
2688         (fyl2xp1_extend<mode>xf3_i387): Remove.
2689         (log1p<mode>2): Change operand 1 predicate to general_operand.
2690         Extend operand 1 to XFmode and generate log1pxf3 insn.
2691         (fxtract_extend<mode>xf3_i387): Remove.
2692         (logb<mode>2): Change operand 1 predicate to general_operand.
2693         Extend operand 1 to XFmode and generate logbxf3 insn.
2694         (ilogb<mode>2): Change operand 1 predicate to general_operand.
2695         Extend operand 1 to XFmode and generate fxtractxf3_i387 insn.
2696         (significand<mode>2): Change operand 1 predicate to general_operand.
2697         Extend operand 1 to XFmode and generate significandxf3 insn.
2699 2018-09-11  Nathan Sidwell  <nathan@acm.org>
2701         * gcc.c (perror_with_name, pfatal_with_name): Delete.
2702         (load_specs): Use fatal_error.
2703         (DELETE_IF_ORDINARY, process_command): Use error.
2704         (execute, run_attempt): Use fatal_error.
2706 2018-09-11  Andrew Stubbs  <ams@codesourcery.com>
2708         * diagnostic-core.h (sorry_at): New prototype.
2709         * diagnostic.c (sorry_at): New function.
2711 2018-09-11  Aldy Hernandez  <aldyh@redhat.com>
2713         * tree-vrp.c (extract_range_from_binary_expr_1): Treat all divisions
2714         by zero as VR_UNDEFINED.
2716 2018-09-10  Uros Bizjak  <ubizjak@gmail.com>
2718         * config/i386/i386.md (<sincos>xf2): Rename from *<sincos>xf2_i387.
2719         (*<sincos>_extend<mode>xf2_i387): Remove insn pattern.
2720         (<sincos>mode2): New expander.
2721         (sincos_extend<mode>xf3_i387): Remove insn pattern.
2722         (sincos -> sin, cos splitters): Remove splitter patterns.
2723         (sincos<mode>3): Change operand 2 predicate to general_operand.
2724         Extend operand 2 to XFmode and generate sincosxf3 insn.
2725         (fptanxf4_i387): Change mode of operands 0 and 3 to SFmode.
2726         Change operand 3 predicate to const1_operand.
2727         (fptan_extend<mode>xf4_i387): Remove insn pattern.
2728         (tanxf2): Update operands in the call to fptanxf4_i387.
2729         (tan<mode>2): Change operand 1 predicate to general_operand.
2730         Extend operand 1 to XFmode and generate tanxf3 insn.
2731         (atan2xf3): Rename from *fpatanxf3_i387.
2732         (fpatan_extend<mode>xf3_i387): Remove insn pattern.
2733         (atan2xf3): Remove expander.
2734         (atan2<mode<3):  Change operand 1 and 2 predicates to general_operand.
2735         Extend operands 1 and 2 to XFmode and generate atan2xf3 insn.
2736         (atan<mode>2): Change operand 1 predicate to general_operand.
2737         Extend operand 1 to XFmode and generate atanxf3 insn.
2739 2018-09-10  Uros Bizjak  <ubizjak@gmail.com>
2741         * config/i386/i386.md (x87/SSE constant load splitter): Use
2742         memory_operand instead of nonimmediate_operand for input operand
2743         predicate.
2745 2018-09-09  Uros Bizjak  <ubizjak@gmail.com>
2747         * config/i386/i386.md (float partial SSE register stall splitter): Move
2748         splitter near its instruction pattern.
2749         (float_extend partial SSE register stall splitter): Ditto.
2750         (float_truncate partial SSE register stall splitter): Ditto.
2752 2018-09-09  Hans-Peter Nilsson  <hp@bitrange.com>
2754         PR target/86794
2755         * config/mmix/mmix.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
2756         to speculation_safe_value_not_needed.
2758         PR target/85666
2759         * config/mmix/mmix.c (mmix_assemble_integer): Handle byte-size
2760         non-CONST_INT rtx:es using assemble_integer_with_op ".byte".
2761         (MMIX_CFUN_NEEDS_SAVED_EH_RETURN_ADDRESS): Don't call
2762         leaf_function_p, instead use has_hard_reg_initial_val.
2764 2018-09-09  Nathan Sidwell  <nathan@acm.org>
2766         * gcc.h (pfatal_with_name): Don't declare here.
2767         * gcc.c (pfatal_with_name): Make static.
2769 2018-09-09  Xianmiao Qu  <xianmiao_qu@c-sky.com>
2771         * config/csky/csky.md (*cskyv2_adddi3, *cskyv2_subdi3): Add
2772         earlyclobber.
2774 2018-09-08  John David Anglin  <danglin@gcc.gnu.org>
2776         PR rtl-optimization/85458
2777         * sel-sched.c (sel_target_adjust_priority): Allow backend adjust
2778         priority hook to reduce the priority of EXPR.
2780 2018-09-07  Uros Bizjak  <ubizjak@gmail.com>
2782         * config/i386/i386.md (float<SWI48x:mode><MODEF:mode>2) Enable
2783         DImode for x87 on 32bit targets.  Conditionally disable x87 modes
2784         with X87_ENABLE_FLOAT.  Remove preparation code.
2785         (*float<SWI48:mode><MODEF:mode>2): Rename from
2786         *float<SWI48:mode><MODEF:mode>2_mixed.  Handle x87, SSE and mixed
2787         math using "enabled" attribute.
2788         (*floatdi<MODEF:mode>2_i387): Rename from
2789         *float<SWI48x:mode><MODEF:mode>2_i387.  Handle only DImode and
2790         enable for 32bit targets only.
2791         (floatdi<X87MODEF:mode>2_i387_with_xmm pre-reload splitter): New
2792         splitter.
2793         (floatdi<X87MODEF:mode>2_i387_with_xmm): Use register_operand
2794         as operand 1 predicate.  Rewrite as define_insn_and_split.
2795         (floatdi<X87MODEF:mode>2_i387_with_xmm memory input splitter): Remove.
2797 2018-09-06  Uros Bizjak  <ubizjak@gmail.com>
2799         * reg-stack.c (get_true_reg) <case FLOAT_TRUNCATE>: Reorder
2800         to fallthru to FLOAT case.
2802 2018-09-06  Will Schmidt  <will_schmidt@vnet.ibm.com>
2804         PR target/86731
2805         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Update logic
2806         around folding of vec_sl to handle out of range shift values.
2808 2018-09-06  Uros Bizjak  <ubizjak@gmail.com>
2810         * config/i386/i386.md (fix_trunc<mode>_fisttp_i387_1): Remove.
2811         Update callers to gen_fix_trunc<mode>_i387_fisttp
2812         (fix_trunc<mode>_i387_fisttp): Change operand 0 predicate to
2813         nonimmediate_operand.
2814         (fix_trunc<mode>_i387_fisttp_with_temp): Remove insn pattern
2815         and corresponding splitters.
2816         (*fix_trunc<mode>_i387_1): Always expand with fix_trunc<mode>_i387.
2817         (fix_truncdi_i387): Change operand 0 predicate to nonimmediate_operand.
2818         (fix_truncdi_i387_with_temp): Remove insn pattern
2819         and corresponding splitters.
2820         (fix_trunc<mode>_i387): Change operand 0 predicate to
2821         nonimmediate_operand.
2822         (fix_trunc<mode>_i387_with_temp): Remove insn pattern
2823         and corresponding splitters.
2824         (*fistdi2_1): Remove.
2825         (fistdi2): Ditto.
2826         (fistdi2_with_temp): Remove insn pattern and corresponding splitters.
2827         (lrintxfdi2): Remove expander.  Reimplement as define_insn.
2828         (*fist<mode>2_1): Remove.
2829         (fist<mode>2): Ditto.
2830         (fist<mode>2_with_temp): Remove insn pattern and corresponding
2831         splitters.
2832         (lrintxf<mode>2): Remove expander.  Reimplement as define_insn.
2833         (*fist<mode>2_<rounding>_1): Always expand with fist<mode>2_<rounding>.
2834         (fistdi2_<rounding>): Change operand 0 predicate to
2835         nonimmediate_operand.
2836         (fistdi2_<rounding>_with_temp): Remove insn pattern
2837         and corresponding splitters.
2838         (fist<mode>2_<rounding>): Change operand 0 predicate to
2839         nonimmediate_operand.
2840         (fist<mode>2_<rounding>_with_temp): Remove insn pattern
2841         and corresponding splitters.
2843         (*fixuns_trunc<mode>si2_avx512f_zext): Depend on TARGET_SSE_MATH.
2845 2018-09-06  Bernd Edlinger  <bernd.edlinger@hotmail.de>
2847         * varasm.c (output_constructor_regular_field): Check TYPE_SIZE_UNIT of
2848         the init value.
2850 2018-09-06  Will Schmidt  <will_schmidt@vnet.ibm.com>
2852         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support for
2853         early gimple folding of vec_splat().
2854         * tree-vect-generic.c: Remove static from tree_vec_extract() definition.
2855         * gimple-fold.h: Add an extern define for tree_vec_extract().
2857 2018-09-06  Will Schmidt  <will_schmidt@vnet.ibm.com>
2859         * config/rs6000/rs6000.c (fold_mergehl_helper): Add types_compatible_p
2860         wrappers around TREE_TYPE comparisons.
2862 2018-09-06  Ilya Leoshkevich  <iii@linux.ibm.com>
2864         PR target/80080
2865         * config/s390/predicates.md: Add nonsym_memory_operand.
2866         * config/s390/s390.c (s390_legitimize_cs_operand): If operand
2867         contains a SYMBOL_REF, load it into an intermediate pseudo.
2868         (s390_emit_compare_and_swap): Legitimize operand.
2869         * config/s390/s390.md: Use the new nonsym_memory_operand
2870         with UNSPECV_CAS patterns.
2872 2018-09-06  Ilya Leoshkevich  <iii@linux.ibm.com>
2874         PR target/80080
2875         * config/s390/s390-passes.def: New file.
2876         * config/s390/s390-protos.h (class rtl_opt_pass): Add forward
2877         declaration.
2878         (make_pass_s390_early_mach): Add declaration.
2879         * config/s390/s390.c (make_pass_s390_early_mach):
2880         (s390_option_override): Remove dynamic registration.
2881         * config/s390/t-s390: Add s390-passes.def.
2883 2018-09-06  Ilya Leoshkevich  <iii@linux.ibm.com>
2885         * config/s390/s390.c (s390_decompose_constant_pool_ref):
2886         Remove UNSPEC_LTREL_BASE check.
2887         (annotate_constant_pool_refs): Likewise.
2888         (find_constant_pool_ref): Likewise.
2889         (find_ltrel_base): Removed.
2890         (replace_ltrel_base): Removed.
2891         (s390_mainpool_finish): Remove replace_ltrel_base call.
2892         (s390_chunkify_start): Remove pending LTREL_BASE logic.
2893         (s390_chunkify_finish): Remove replace_ltrel_base call.
2894         * config/s390/s390.md: Remove UNSPEC_LTREL_BASE.
2896 2018-09-06  Hans-Peter Nilsson  <hp@axis.com>
2898         PR target/86779
2899         * config/cris/cris.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
2900         to speculation_safe_value_not_needed.
2902 2018-09-05  Cesar Philippidis  <cesar@codesourcery.com>
2903             Bernd Schmidt  <bernds_cb1@t-online.de>
2905         * config/nvptx/nvptx-opts.h: New file.
2906         * config/nvptx/nvptx.c (nvptx_file_start): Print the correct .target.
2907         * config/nvptx/nvptx.h: Include "nvptx-opts.h".
2908         (ASM_SPEC): Define.
2909         (TARGET_SM35): New macro.
2910         * config/nvptx/nvptx.md (atomic_fetch_<logic><mode>): Enable with the
2911         correct predicate.
2912         * config/nvptx/nvptx.opt (ptx_isa, sm_30, sm_35): New enum and its
2913         values.
2914         (misa=): New option.
2915         * doc/invoke.texi (Nvidia PTX Options): Document -misa.
2917 2018-09-05  Uros Bizjak  <ubizjak@gmail.com>
2919         * config/i386/i386.md (truncdfsf2): Remove expander.
2920         (truncdfsf2_with_temp): Ditto.
2921         (truncxf<mode>2): Ditto.
2922         (*truncdfsf_fast_mixed): Remove insn pattern.
2923         (*truncdfsf_fast_i387): Ditto.
2924         (*truncdfsf_mixed): Ditto.
2925         (*truncdfsf_i387): Ditto.
2926         (*truncdfsf2_i387_1): Ditto.
2927         (*truncxfsf2_mixed): Ditto.
2928         (*truncxfdf2_mixed): Ditto.
2929         (*truncxf<mode>2_i387_noop): Ditto. Update callers
2930         to call gen_truncxf<mode>2 instead.
2931         (*truncxf<mode>2_i387): Remove.
2932         (reg->reg splitters): Remove splitter pattern.
2933         (reg->mem splitters): Ditto.
2935         (truncdfsf2): New insn pattern.
2936         (truncxf<mode>2): Ditto.
2938 2018-09-05  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
2940         * tree-ssa-math-opts.c (is_mult_by): New function.
2941         (is_square_of): Use the above.
2942         (optimize_recip_sqrt): New function.
2943         (pass_cse_reciprocals::execute): Use the above.
2945 2018-09-05  Richard Biener  <rguenther@suse.de>
2947         PR bootstrap/87134
2948         * tree-ssa-sccvn.c (rpo_elim::eliminate_push_avail): Make sure
2949         to zero-init the emplaced vec.
2951 2018-09-05  Martin Liska  <mliska@suse.cz>
2953         PR tree-optimization/87205
2954         * tree-switch-conversion.c (pass_lower_switch::execute):
2955         Group cases for switch statements.
2957 2018-09-05  Richard Biener  <rguenther@suse.de>
2959         PR tree-optimization/87217
2960         * tree-ssa-sccvn.c (vuse_valueize): New.
2961         (vn_reference_lookup_pieces): Use it.
2962         (vn_reference_lookup): Likewise.
2964 2018-09-05  Nathan Sidwell  <nathan@acm.org>
2966         PR c++/87137
2967         * stor-layout.c (place_field): Scan forwards to check last
2968         bitfield when ms_bitfield_placement is in effect.
2970 2018-09-05  Richard Biener  <rguenther@suse.de>
2972         PR bootstrap/87225
2973         * tree-vect-stmts.c (vectorizable_simd_clone_call): Fix bogus
2974         return.
2976 2018-09-05  Siddhesh Poyarekar  <siddhesh@sourceware.org>
2977             Kugan Vivekanandarajah  <kugan.vivekanandarajah@linaro.org>
2979         * config/aarch64/falkor-tag-collision-avoidance.c: New file.
2980         * config.gcc (extra_objs): Build it.
2981         * config/aarch64/t-aarch64 (falkor-tag-collision-avoidance.o):
2982         Likewise.
2983         * config/aarch64/aarch64-passes.def
2984         (pass_tag_collision_avoidance): New pass.
2985         * config/aarch64/aarch64.c (qdf24xx_tunings): Add
2986         AARCH64_EXTRA_TUNE_RENAME_LOAD_REGS to tuning_flags.
2987         (aarch64_classify_address): Remove static qualifier.
2988         (aarch64_address_info, aarch64_address_type): Move to...
2989         * config/aarch64/aarch64-protos.h: ... here.
2990         (make_pass_tag_collision_avoidance): New function.
2991         * config/aarch64/aarch64-tuning-flags.def (rename_load_regs):
2992         New tuning flag.
2994 2018-09-05  Martin Liska  <mliska@suse.cz>
2996         * doc/gcov.texi: Update documentation of humar
2997         readable mode.
2998         * gcov.c (format_count): Print one decimal place, it provides
2999         more fine number of situations like '1G' vs. '1.4G'.
3001 2018-09-05  Martin Liska  <mliska@suse.cz>
3003         PR target/87164
3004         * config/rs6000/rs6000.opt: Mark the option as Deprecated.
3005         * optc-gen.awk: Allow 'Var' for Deprecated options in order
3006         to generate a MASK value.
3008 2018-09-04  H.J. Lu  <hongjiu.lu@intel.com>
3010         PR debug/86593
3011         * dwarf2out.c (based_loc_descr): Allow hard frame pointer even
3012         if frame pointer isn't used.
3013         (compute_frame_pointer_to_fb_displacement): Likewise.
3015 2018-09-04  Jakub Jelinek  <jakub@redhat.com>
3017         PR target/87198
3018         * common/config/i386/i386-common.c (OPTION_MASK_ISA_XSAVEOPT_SET,
3019         OPTION_MASK_ISA_XSAVES_SET, OPTION_MASK_ISA_XSAVEC_SET): Use
3020         OPTION_MASK_ISA_XSAVE_SET instead of OPTION_MASK_ISA_XSAVE.
3021         (OPTION_MASK_ISA_XSAVE_UNSET): Add OPTION_MASK_ISA_XSAVES_UNSET
3022         and OPTION_MASK_ISA_XSAVEC_UNSET.
3024 2018-09-04  Max Filippov  <jcmvbkbc@gmail.com>
3026         * config/xtensa/xtensa.c (xtensa_expand_atomic): Reorder AND and
3027         XOR operations in NAND case.
3029 2018-09-04  Aldy Hernandez  <aldyh@redhat.com>
3031         * wide-int-range.cc (wide_int_range_convert): New.
3032         * wide-int-range.h (wide_int_range_convert): New.
3033         * tree-vrp.c (extract_range_from_unary_expr): Abstract wide int
3034         code into wide_int_range_convert.
3035         (extract_range_into_wide_ints): Do not munge anti range constants
3036         into the entire domain.  Just return the range back.
3038 2018-09-04  Martin Liska  <mliska@suse.cz>
3040         * genmatch.c (output_line_directive): Add new argument
3041         fnargs.
3042         (dt_simplify::gen_1): Encapsulate dump within __builtin_expect.
3044 2018-09-04  Jonathan Wakely  <jwakely@redhat.com>
3046         * doc/invoke.texi (Option Summary): Add whitespace.
3048         * doc/invoke.texi (Option Summary): Add -Waligned-new.
3050 2018-09-04  Richard Biener  <rguenther@suse.de>
3052         PR tree-optimization/87211
3053         * tree-ssa-sccvn.c (visit_phi): When value-numbering to a
3054         backedge value we're supposed to treat as VARYING also number
3055         the PHI to VARYING in case it got a different value-number already.
3057 2018-09-04  Aldy Hernandez  <aldyh@redhat.com>
3059         * tree-vrp.c (vrp_can_optimize_bit_op): Remove.
3060         (extract_range_from_binary_expr_1): Do not call
3061         vrp_can_optimize_bit_op.
3062         * wide-int-range.cc (wide_int_range_can_optimize_bit_op): Make
3063         static.
3064         (wide_int_range_get_mask_and_bounds): New.
3065         (wide_int_range_optimize_bit_op): New.
3066         (wide_int_range_bit_ior): Call wide_int_range_optimize_bit_op.
3067         (wide_int_range_bit_and): Same.
3068         * wide-int-range.h (wide_int_range_can_optimize_bit_op): Remove.
3069         (wide_int_range_optimize_bit_op): New.
3070         (wide_int_range_get_mask_and_bounds): New.
3072 2018-09-04  Richard Biener  <rguenther@suse.de>
3074         PR tree-optimization/87176
3075         * tree-ssa-sccvn.c (visit_phi): Remove redundant allsame
3076         variable.  When value-numbering a virtual PHI node make sure
3077         to not value-number to the backedge value.
3079 2018-09-04  Jonathan Wakely  <jwakely@redhat.com>
3081         * doc/extend.texi (Long Long, Hex Floats): Document support for
3082         long long and hex floats in more recent versions of ISO C++.
3084 2018-09-03  Richard Biener  <rguenther@suse.de>
3086         PR tree-optimization/87177
3087         * tree-ssa-sccvn.c (vuse_ssa_val): Revert previous change, keep
3088         cleanup.
3090 2018-09-03  Alexander Monakov  <amonakov@ispras.ru>
3092         * bb-reorder.c (edge_order): Convert to C-qsort-style
3093         tri-state comparator.
3094         (reorder_basic_blocks_simple): Change std::stable_sort to
3095         gcc_stablesort.
3097 2018-09-03  Alexander Monakov  <amonakov@ispras.ru>
3099         * tree-loop-distribution.c (offset_cmp): Convert to C-qsort-style
3100         tri-state comparator.
3101         (fuse_memset_builtins): Change std::stable_sort to gcc_stablesort.
3103 2018-09-03  Alexander Monakov  <amonakov@ispras.ru>
3105         * sort.cc (struct sort_ctx): New field 'nlim'.  Use it...
3106         (mergesort): ... here as maximum count for using netsort.
3107         (gcc_qsort): Set nlim to 3 if stable sort is requested.
3108         (gcc_stablesort): New.
3109         * system.h (gcc_stablesort): Declare.
3111 2018-09-03  Alexander Monakov  <amonakov@ispras.ru>
3113         * sort.cc (gcc_qsort) [CHECKING_P]: Call qsort_chk.
3114         * system.h (qsort): Always redirect to gcc_qsort.  Update comment.
3115         * vec.c (qsort_chk): Do not call gcc_qsort.  Update comment.
3117 2018-09-03  Segher Boessenkool  <segher@kernel.crashing.org>
3119         * config/rs6000/rs6000.md (*mov<mode>_hardfloat32): Remove %U from the
3120         lxsdx and stxsdx alternatives.
3121         (*mov<mode>_hardfloat64): Ditto.
3122         * config/rs6000/vsx.md (*vsx_extract_<mode>_store): Ditto.
3124 2018-09-03  Richard Biener  <rguenther@suse.de>
3126         PR tree-optimization/87200
3127         * tree-ssa-sccvn.c (vn_nary_build_or_lookup_1): Valueize a
3128         simplify result.
3130 2018-09-03  Martin Liska  <mliska@suse.cz>
3132         PR tree-optimization/87201
3133         * tree-switch-conversion.c (switch_decision_tree::balance_case_nodes):
3134         Fix parenthesis in an expression.
3136 2018-09-03  Richard Biener  <rguenther@suse.de>
3138         PR tree-optimization/87197
3139         * tree-ssa-sccvn.c (vn_nary_build_or_lookup_1): Mark the new def
3140         visited.  CSE the VN_INFO hashtable lookup.
3142         PR tree-optimization/87169
3143         * tree-ssa-sccvn.c (do_rpo_vn): When marking loops for not
3144         iterating make sure there's no extra backedges from irreducible
3145         regions feeding the header.  Mark the destination block
3146         executable.
3148 2018-09-03  Martin Liska  <mliska@suse.cz>
3150         PR driver/83193
3151         * common/common-target.def: Add TARGET_GET_VALID_OPTION_VALUES.
3152         * common/common-targhooks.c (default_get_valid_option_values):
3153         New function.
3154         * common/common-targhooks.h (default_get_valid_option_values):
3155         Likewise.
3156         * common/config/i386/i386-common.c: Move processor_target_table
3157         from i386.c.
3158         (ix86_get_valid_option_values): New function.
3159         (TARGET_GET_VALID_OPTION_VALUES): New macro.
3160         * config/i386/i386.c (struct ptt): Move to i386-common.c.
3161         (PTA_*): Move all defined masks into i386-common.c.
3162         (ix86_function_specific_restore): Use new processor_cost_table.
3163         * config/i386/i386.h (struct ptt): Moved from i386.c.
3164         (struct pta): Likewise.
3165         * doc/tm.texi: Document new TARGET_GET_VALID_OPTION_VALUES.
3166         * doc/tm.texi.in: Likewise.
3167         * opt-suggestions.c (option_proposer::suggest_option):
3168         Pass prefix to build_option_suggestions.
3169         (option_proposer::get_completions): Likewise.
3170         (option_proposer::build_option_suggestions): Use the new target
3171         hook.
3172         * opts.c (struct option_help_tuple): New struct.
3173         (print_filtered_help): Use the new target hook.
3175 2018-09-03  Martin Liska  <mliska@suse.cz>
3177         PR middle-end/59521
3178         * predict.c (set_even_probabilities): Add likely_edges
3179         argument and handle cases where we have precisely one
3180         likely edge.
3181         (combine_predictions_for_bb): Catch also likely_edges.
3182         (tree_predict_by_opcode): Handle gswitch statements.
3183         * tree-cfg.h (find_case_label_for_value): New declaration.
3184         (find_taken_edge_switch_expr): Likewise.
3185         * tree-switch-conversion.c (switch_decision_tree::balance_case_nodes):
3186         Find pivot in decision tree based on probabily, not by number of
3187         nodes.
3189 2018-09-02  Gerald Pfeifer  <gerald@pfeifer.com>
3191         * doc/standards.texi (Standards): Update Objective-C reference.
3193 2018-09-01  Gerald Pfeifer  <gerald@pfeifer.com>
3195         * doc/install.texi (Prerequisites): Update link for MPC.
3197 2018-09-01  Michael Matz  <matz@suse.de>
3199         PR tree-optimization/87074
3200         * gimple-loop-jam.c (unroll_jam_possible_p): Check loop exit
3201         PHIs for outer-loop uses.
3203 2018-09-01  Gerald Pfeifer  <gerald@pfeifer.com>
3205         * doc/generic.texi (OpenMP): Adjust link to openmp.org.
3206         * doc/invoke.texi (C Dialect Options): Ditto.
3208 2018-09-01  Gerald Pfeifer  <gerald@pfeifer.com>
3210         * doc/install.texi (Prerequisites): Adjust link mpfr.org.
3212 2018-08-31  Richard Biener  <rguenther@suse.de>
3214         PR tree-optimization/87168
3215         * tree-ssa-sccvn.c (SSA_VAL): Add visited output parameter.
3216         (rpo_elim::eliminate_avail): When OP was not visited it must
3217         be available.
3219 2018-08-31  David Malcolm  <dmalcolm@redhat.com>
3221         * tree-vrp.c (copy_value_range): Convert param "from" from
3222         "value_range *" to "const value_range *".
3223         (range_is_null): Likewise for param "vr".
3224         (range_int_cst_p): Likewise.
3225         (range_int_cst_singleton_p): Likewise.
3226         (symbolic_range_p): Likewise.
3227         (value_ranges_intersect_p): Likewise for both params.
3228         (value_range_nonnegative_p): Likewise for param "vr".
3229         (value_range_constant_singleton): Likewise.
3230         (vrp_set_zero_nonzero_bits): Likewise for param "ar".
3231         (extract_range_into_wide_ints): Likewise for param "vr".
3232         (extract_range_from_multiplicative_op): Likewise for params "vr0"
3233         and "vr1".
3234         (vrp_can_optimize_bit_op): Likewise.
3235         (extract_range_from_binary_expr_1): Likewise for params "vr0_" and
3236         "vr1_".
3237         (extract_range_from_unary_expr): Likewise.
3238         (debug_value_range): Likewise for param "vr".
3239         (value_range::dump): Add "const" qualifier.
3240         (vrp_prop::check_array_ref): Convert local "vr" from
3241         "value_range *" to "const value_range *".
3242         (vrp_prop::check_mem_ref): Likewise.
3243         (vrp_prop::visit_stmt): Likewise for local "old_vr".
3244         (vrp_intersect_ranges_1): Likewise for param "vr_1".
3245         (vrp_intersect_ranges): Likewise.
3246         (simplify_stmt_for_jump_threading): Likewise for local "vr".
3247         (vrp_prop::vrp_finalize): Likewise.
3248         * tree-vrp.h (value_range::dump): Add "const" qualifier.
3249         (vrp_intersect_ranges): Add "const" qualifier to params as above.
3250         (extract_range_from_unary_expr): Likewise.
3251         (value_range_constant_singleton): Likewise.
3252         (symbolic_range_p): Likewise.
3253         (copy_value_range): Likewise.
3254         (extract_range_from_binary_expr_1): Likewise.
3255         (range_int_cst_p): Likewise.
3256         (vrp_set_zero_nonzero_bits): Likewise.
3257         (range_int_cst_singleton_p): Likewise.
3259 2018-08-31  Vlad Lazar  <vlad.lazar@arm.com>
3261         * config/aarch64/arm_neon.h (vabsd_s64): New.
3262         (vnegd_s64): Likewise.
3264 2018-08-31  Martin Jambor  <mjambor@suse.cz>
3266         * ipa-cp.c (estimate_local_effects): Replace wrong MAX with MIN.
3268 2018-08-31  Martin Liska  <mliska@suse.cz>
3270         * ipa-icf.c (sem_item::add_type): Use
3271         sem_item::m_type_hash_cache.
3272         * ipa-icf.h: Move the cache from sem_item_optimizer
3273         to sem_item.
3275 2018-08-31  Nathan Sidwell  <nathan@acm.org>
3277         * doc/extend.texi (Backwards Compatibility): Remove implicit
3278         extern C leeway of () being (...).
3280 2018-08-31  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
3282         * ipa-inline.c (can_inline_edge_by_limits_p): Fix typos in comment.
3284 2018-08-31  Segher Boessenkool  <segher@kernel.crashing.org>
3286         PR target/86684
3287         PR target/87149
3288         * config/rs6000/rs6000.md (lround<mode>di2): Gate on TARGET_FPRND.
3290 2018-08-31  Jakub Jelinek  <jakub@redhat.com>
3292         PR middle-end/87138
3293         * expmed.c (expand_mult_const): Use immed_wide_int_const instead of
3294         gen_int_mode.  Formatting fixes.
3296 2018-08-30  Sandra Loosemore  <sandra@codesourcery.com>
3298         * target.def (custom_function_descriptors): Improve documentation.
3299         * doc/tm.texi.in (Trampolines): Expand discussion of function
3300         descriptors and move TARGET_CUSTOM_FUNCTION_DESCRIPTORS to the
3301         beginning of the section.
3302         * doc/tm.texi: Regenerated.
3304 2018-08-30  Jose E. Marchesi  <jose.marchesi@oracle.com>
3306         * cfg.h (class auto_edge_flag): Spell out the template-id of the
3307         base class in the initializer list.  This is a workaround for
3308         building with older GCC.
3309         (class auto_bb_flag): Likewise.
3311 2018-08-30  Aaron Sawdey  <acsawdey@linux.ibm.com>
3313         * config/rs6000/altivec.md (altivec_eq<mode>): Remove star.
3314         (altivec_vcmpequ<VI_char>_p): Remove star.
3315         * config/rs6000/rs6000-string.c (do_load_for_compare): Support
3316         vector load modes.
3317         (expand_strncmp_vec_sequence): New function.
3318         (emit_final_str_compare_vec): New function.
3319         (expand_strn_compare): Add support for vector strncmp.
3320         * config/rs6000/rs6000.opt (-mstring-compare-inline-limit): Change
3321         length specification to bytes.
3322         * config/rs6000/vsx.md (vsx_ld_elemrev_v16qi_internal): Remove star.
3323         (vcmpnezb_p): New pattern.
3324         * doc/invoke.texi (RS/6000 and PowerPC Options): Update documentation
3325         for option -mstring-compare-inline-limit.
3327 2018-08-30  Thiago Macieira  <thiago.macieira@intel.com>
3329         * config/i386/i386.c (PTA_WESTMERE): Remove PTA_AES.
3330         (PTA_SKYLAKE): Add PTA_AES.
3331         (PTA_GOLDMONT): Likewise.
3333 2018-08-29  Jan Hubicka  <jh@suse.cz>
3335         PR lto/86517
3336         * lto-opts.c (lto_write_options): Always stream PIC/PIE mode.
3337         * lto-wrapper.c (merge_and_complain): Fix merging of PIC/PIE.
3339 2018-08-29  Jan Hubicka  <jh@suse.cz>
3341         * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not follow
3342         TYPE_STUB_DECL.
3343         (hash_tree): Do not visit TYPE_STUB_DECL.
3344         * tree-streamer-out.c (write_ts_type_common_tree_pointers): Do not
3345         stream TYPE_STUB_DECL.
3346         * tree-streamer-in.c (lto_input_ts_type_common_tree_pointers): Likewise.
3347         * ipa-utils.h (type_with_linkage_p): Do not rely on TYPE_STUB_DECL
3348         after free_lang_data.
3349         (type_in_anonymous_namespace_p): Likewise.
3351 2018-08-29  Jan Hubicka  <jh@suse.cz>
3353         * sreal.h (SREAL_PART_BITS): Change to 31; remove seemingly unnecessary
3354         comment that it has to be even number.
3355         (class sreal): Change m_sig type to int32_t.
3356         * sreal.c (sreal::dump, sreal::to_int, opreator+, operator-): Use
3357         int64_t for temporary calculations.
3358         (sreal_verify_basics): Drop one bit from minimum and maximum.
3360 2018-08-30  Richard Biener  <rguenther@suse.de>
3362         PR tree-optimization/87147
3363         * tree-ssa-sccvn.c (SSA_VISITED): New function.
3364         (visit_phi): When the degenerate result is from the backedge and
3365         we didn't visit its definition yet drop to VARYING.
3366         (do_rpo_vn): Properly mark blocks with incoming backedges as executable.
3368 2018-08-29  Jan Hubicka  <jh@suse.cz>
3370         * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not walk
3371         DECL_VINDEX.
3372         (hash_tree): Likewise.
3374 2018-08-29  Jan Hubicka  <jh@suse.cz>
3376         * tree.c (find_decls_types_r): Walk also TYPE_NEXT_PTR_TO
3377         and TYPE_NEXT_REF_TO.
3379 2018-08-29  Jan Hubicka  <jh@suse.cz>
3381         * sreal.h (SREAL_PART_BITS): Change to 31; remove seemingly unnecessary
3382         comment that it has to be even number.
3383         (class sreal): Change m_sig type to int32_t.
3384         * sreal.c (sreal::dump, sreal::to_int, opreator+, operator-): Use
3385         int64_t for temporary calculations.
3386         (sreal_verify_basics): Drop one bit from minimum and maximum.
3388 2018-08-30  Tamar Christina  <tamar.christina@arm.com>
3390         * config/aarch64/aarch64.c (aarch64_expand_movmem): Set TImode max.
3392 2018-08-30  Vlad Lazar  <vlad.lazar@arm.com>
3394         PR middle-end/86995
3395         * expmed.c (canonicalize_comparison): Use wi::sub instead of wi::add
3396         if to_add is negative.
3398 2018-08-29  Bernd Edlinger  <bernd.edlinger@hotmail.de>
3400         PR middle-end/87053
3401         * builtins.c (c_strlen): Improve range checks.
3403 2018-08-29  Martin Sebor  <msebor@redhat.com>
3404             Jeff Law  <law@redhat.com>
3406         PR tree-optimization/86714
3407         PR tree-optimization/86711
3408         * builtins.c (c_strlen): Add arguments to call to string_constant.
3409         * expr.c (string_constant): Add argument.  Detect missing nul
3410         terminator and outermost declaration it's missing in.
3411         * expr.h (string_constant): Add argument.
3412         * fold-const.c (read_from_constant_string): Add arguments to call to
3413         string_constant.
3414         (c_getstr): Likewise.
3415         * tree-ssa-forwprop.c (simplify_builtin_call): Likewise.
3416         to string_constant.
3417         * tree-ssa-strlen.c (get_stridx): Likewise.
3419 2018-08-29  Jan Hubicka  <jh@suse.cz>
3421         * tree-streamer-in.c (lto_input_ts_function_decl_tree_pointers):
3422         Do not stream DECL_VINDEX.
3423         * tree-streamer-out.c (write_ts_function_decl_tree_pointers): Likewise.
3424         * tree.c (free_lang_data_in_decl): Clear DECL_VINDEX.
3425         (decl_function_context): Use DECL_VIRTUAL_P rather than DECL_VINDEX.
3427 2018-08-29  Richard Biener  <rguenther@suse.de>
3429         * tree-ssa-sccvn.c (vuse_ssa_val): Return NULL for unvisited
3430         virtual operands that are not default defs to honor region
3431         boundaries.
3432         (rpo_vn_valueize): Remove ineffective code here.
3434 2018-08-29  Richard Biener  <rguenther@suse.de>
3436         PR tree-optimization/87132
3437         * tree-ssa-alias.c (get_continuation_for_phi): Do not translate
3438         when skipping defs reachable over backedges.
3440 2018-08-29  Richard Biener  <rguenther@suse.de>
3442         * tree-core.h: Document use of deprecated_flag in SSA_NAME.
3443         * tree.h (SSA_NAME_POINTS_TO_READONLY_MEMORY): Define.
3444         * tree-into-ssa.c (pass_build_ssa::execute): Initialize
3445         function parameters SSA_NAME_POINTS_TO_READONLY_MEMORY from fnspec.
3446         * tree-ssa-sccvn.c (const_parms, init_const_parms): Remove.
3447         (vn_reference_lookup_3): Remove use of const_parms.
3448         (free_rpo_vn): Do not free const_parms.
3449         (do_rpo_vn): Do not call init_const_parms.
3450         * tree-ssa-alias.c (refs_may_alias_p_1): Honor
3451         SSA_NAME_POINTS_TO_READONLY_MEMORY.
3452         (call_may_clobber_ref_p_1): Likewise.
3454 2018-08-29  Alexander Monakov  <amonakov@ispras.ru>
3456         PR other/86726
3457         * invoke.texi (Optimization Options): List -ftree-scev-cprop.
3458         (-O): Ditto.
3459         (-ftree-scev-cprop): Document.
3461 2018-08-29  Jan Hubicka  <jh@suse.cz>
3463         * sreal.h (normalize, normalize_up, normalize_down): Add new_sig/new_exp
3464         parameters.
3465         (sreal constructor): Update.
3466         * sreal.c (sreal:operator+, sreal:operator-, sreal:operator*,
3467         sreal:operator/): Update.
3469 2018-08-29  Martin Liska  <mliska@suse.cz>
3471         * tree-switch-conversion.c (switch_conversion::expand):
3472         Strenghten assumption about gswitch statements.
3474 2018-08-29  Richard Biener  <rguenther@suse.de>
3476         PR tree-optimization/87117
3477         * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_stmt): Only
3478         re-value-number released SSA VDEFs.
3480 2018-08-29  Richard Biener  <rguenther@suse.de>
3482         PR tree-optimization/87126
3483         * tree-ssa-sccvn.c (vn_reference_insert): Remove assert.
3485 2018-08-28  Jim Wilson  <jimw@sifive.com>
3487         * config/riscv/pic.md: Rewrite.
3488         * config/riscv/riscv.c (riscv_address_insns): Return cost of 3 for
3489         invalid address.
3490         * config/riscv/riscv.md (ZERO_EXTEND_LOAD): Delete.
3491         (SOFTF, default_load, softload, softstore): New.
3493 2018-08-28  Jeff Law  <law@redhat.com>
3495         * fold-const.c (fold_binary_loc): Remove recently added assert.
3497 2018-08-28  Joern Rennecke  <joern.rennecke@riscy-ip.com>
3499         * genpreds.c (write_predicate_subfunction): Also add ATTRIBUTE_UNUSED
3500         to OP parmeter of generated function.
3502 2018-08-28  MCC CS  <deswurstes@users.noreply.github.com>
3504         PR tree-optimization/87009
3505         * match.pd: Add boolean optimizations.
3507 2018-08-28  Martin Sebor  <msebor@redhat.com>
3509         PR middle-end/86631
3510         * calls.c (alloc_max_size): Treat HOST_WIDE_INT special.
3511         * gimple-ssa-warn-alloca.c (adjusted_warn_limit): New function.
3512         (pass_walloca::gate): Use it.
3513         (alloca_call_type): Same.
3514         (pass_walloca::execute): Same.
3515         * stor-layout.c (layout_decl): Treat HOST_WIDE_INT special.
3517 2018-08-28  David Malcolm  <dmalcolm@redhat.com>
3519         * dumpfile.h (ATTRIBUTE_GCC_DUMP_PRINTF): Change version check on
3520         GCC_VERSION for usage of "__gcc_dump_printf__" format from
3521         >= 3005 to >= 9000.
3523 2018-08-28  Richard Biener  <rguenther@suse.de>
3525         PR tree-optimization/87124
3526         * tree-ssa-sccvn.c (vn_lookup_simplify_result): Guard against
3527         constants before looking up avail.
3529 2018-08-28  Jakub Jelinek  <jakub@redhat.com>
3531         PR middle-end/87099
3532         * calls.c (maybe_warn_nonstring_arg): Punt early if
3533         warn_stringop_overflow is zero.  Don't call get_range_strlen
3534         on 3rd argument, keep iterating until lenrng[1] is INTEGER_CST.
3535         Swap comparison operands to have constants on rhs.  Only use
3536         lenrng[1] if non-NULL and INTEGER_CST.  Don't uselessly
3537         increment lenrng[0].
3539 2018-08-28  Richard Sandiford  <richard.sandiford@arm.com>
3541         * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Fix unguarded
3542         use of tree_to_shwi.  Remove duplicated test for the size being
3543         a whole number of bytes.
3545 2018-08-28  Richard Biener  <rguenther@suse.de>
3547         PR tree-optimization/87117
3548         * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_cleanup):
3549         Handle removed stmt without LHS (GIMPLE_NOP).
3551 2018-08-28  Richard Biener  <rguenther@suse.de>
3553         PR tree-optimization/87117
3554         * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Exclude
3555         void which is is_gimple_reg_type by checking for COMPLETE_TYPE_P.
3557 2018-08-28  Richard Biener  <rguenther@suse.de>
3559         PR tree-optimization/87117
3560         * tree-ssa-pre.c (compute_avail): Do not make expressions
3561         with predicated values available.
3562         (get_expr_value_id): Assert we do not run into predicated value
3563         expressions.
3565 2018-08-28  Richard Biener  <rguenther@suse.de>
3567         PR tree-optimization/87117
3568         * tree-ssa-operands.c (add_stmt_operand): STRING_CST may
3569         get virtual operands.
3570         (get_expr_operands): Handle STRING_CST like other decls.
3572 2018-08-28  Martin Liska  <mliska@suse.cz>
3574         * tree.h: Update documentation of fndecl_built_in_p
3575         functions.
3578 2018-08-27  Jeff Law  <law@redhat.com>
3579         PR tree-optimization/87110
3580         * tree-ssa-dse.c (compute_trims): Handle non-constant
3581         TYPE_SIZE_UNIT.
3583 2018-08-27  Martin Sebor  <msebor@redhat.com>
3585         PR tree-optimization/86914
3586         * tree-ssa-strlen.c (maybe_set_strlen_range): Avoid MEM_REF.
3588 2018-08-27  Martin Sebor  <msebor@redhat.com>
3590         PR tree-optimization/87112
3591         * builtins.c (expand_builtin_strnlen): Convert c_strlen result to
3592         the type of the bound argument.
3594 2018-08-27  Jeff Law  <law@redhat.com>
3596         * tree-ssa-dse.c (compute_trims): Handle case where the reference's
3597         type does not have a TYPE_SIZE_UNIT.
3599 2018-08-27  Steve Ellcey  <sellcey@cavium.com>
3601         * config/aarch64/aarch64-speculation.cc: Replace include of cfg.h
3602         with include of backend.h.
3604 2018-08-27  Richard Biener  <rguenther@suse.de>
3606         PR tree-optimization/86927
3607         * tree-vect-loop.c (vect_create_epilog_for_reduction): Properly
3608         use const cond reduction code.
3610 2018-08-27  Alexander Monakov  <amonakov@ispras.ru>
3612         PR tree-optimization/85758
3613         * match.pd ((X & Y) ^ Y): Add :s qualifier to inner expression.
3615 2018-08-27  David Malcolm  <dmalcolm@redhat.com>
3617         PR c++/87091
3618         * diagnostic-show-locus.c (class layout_range): Update for
3619         conversion of show_caret_p to a tri-state.
3620         (layout_range::layout_range): Likewise.
3621         (make_range): Likewise.
3622         (layout::maybe_add_location_range): Likewise.
3623         (layout::should_print_annotation_line_p): Don't show annotation
3624         lines for ranges that are SHOW_LINES_WITHOUT_RANGE.
3625         (layout::get_state_at_point): Update for conversion of
3626         show_caret_p to a tri-state.  Bail out early for
3627         SHOW_LINES_WITHOUT_RANGE, so that such ranges don't affect
3628         underlining or source colorization.
3629         (gcc_rich_location::add_location_if_nearby): Update for conversion
3630         of show_caret_p to a tri-state.
3631         (selftest::test_one_liner_multiple_carets_and_ranges): Likewise.
3632         (selftest::test_one_liner_fixit_replace_equal_secondary_range):
3633         Likewise.
3634         (selftest::test_one_liner_labels): Likewise.
3635         * gcc-rich-location.c (gcc_rich_location::add_expr): Update for
3636         conversion of show_caret_p to a tri-state.
3637         * pretty-print.c (text_info::set_location): Likewise.
3638         * pretty-print.h (text_info::set_location): Likewise.
3639         * substring-locations.c (format_warning_n_va): Likewise.
3640         * tree-diagnostic.c (default_tree_printer): Likewise.
3641         * tree-pretty-print.c (newline_and_indent): Likewise.
3643 2018-08-27  David Malcolm  <dmalcolm@redhat.com>
3645         PR c++/87091
3646         * diagnostic-show-locus.c (get_line_span_for_fixit_hint): Show the
3647         line above for line-insertion fix-it hints.
3648         (selftest::test_fixit_insert_containing_newline): Update the
3649         expected results, and add a test with line-numbering enabled.
3651 2018-08-27  Martin Liska  <mliska@suse.cz>
3653         PR sanitizer/86962
3654         * sanopt.c (sanitize_rewrite_addressable_params): Ignore
3655         params with DECL_HAS_VALUE_EXPR_P.
3657 2018-08-27  Martin Liska  <mliska@suse.cz>
3659         * config/i386/i386.c (ix86_expand_set_or_movmem): Dump
3660         selected expansion strategy.
3662 2018-08-27  Martin Liska  <mliska@suse.cz>
3664         * builtins.h (is_builtin_fn): Remove and fndecl_built_in_p.
3665         * builtins.c (is_builtin_fn): Likewise.
3666         * attribs.c (diag_attr_exclusions): Use new function
3667         fndecl_built_in_p and remove check for FUNCTION_DECL if
3668         possible.
3669         (builtin_mathfn_code): Likewise.
3670         (fold_builtin_expect): Likewise.
3671         (fold_call_expr): Likewise.
3672         (fold_builtin_call_array): Likewise.
3673         (fold_call_stmt): Likewise.
3674         (set_builtin_user_assembler_name): Likewise.
3675         (is_simple_builtin): Likewise.
3676         * calls.c (gimple_alloca_call_p): Likewise.
3677         (maybe_warn_nonstring_arg): Likewise.
3678         * cfgexpand.c (expand_call_stmt): Likewise.
3679         * cgraph.c (cgraph_update_edges_for_call_stmt_node): Likewise.
3680         (cgraph_edge::verify_corresponds_to_fndecl): Likewise.
3681         (cgraph_node::verify_node): Likewise.
3682         * cgraphclones.c (build_function_decl_skip_args): Likewise.
3683         (cgraph_node::create_clone): Likewise.
3684         * config/arm/arm.c (arm_insert_attributes): Likewise.
3685         * config/i386/i386.c (ix86_gimple_fold_builtin): Likewise.
3686         * dse.c (scan_insn): Likewise.
3687         * expr.c (expand_expr_real_1): Likewise.
3688         * fold-const.c (operand_equal_p): Likewise.
3689         (fold_binary_loc): Likewise.
3690         * gimple-fold.c (gimple_fold_stmt_to_constant_1): Likewise.
3691         * gimple-low.c (lower_stmt): Likewise.
3692         * gimple-pretty-print.c (dump_gimple_call): Likewise.
3693         * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
3694         Likewise.
3695         * gimple.c (gimple_build_call_from_tree): Likewise.
3696         (gimple_call_builtin_p): Likewise.
3697         (gimple_call_combined_fn): Likewise.
3698         * gimplify.c (gimplify_call_expr): Likewise.
3699         (gimple_boolify): Likewise.
3700         (gimplify_modify_expr): Likewise.
3701         (gimplify_addr_expr): Likewise.
3702         * hsa-gen.c (gen_hsa_insns_for_call): Likewise.
3703         * ipa-cp.c (determine_versionability): Likewise.
3704         * ipa-fnsummary.c (compute_fn_summary): Likewise.
3705         * ipa-param-manipulation.c (ipa_modify_formal_parameters): Likewise.
3706         * ipa-split.c (visit_bb): Likewise.
3707         (split_function): Likewise.
3708         * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
3709         * lto-cgraph.c (input_node): Likewise.
3710         * lto-streamer-out.c (write_symbol): Likewise.
3711         * omp-low.c (setjmp_or_longjmp_p): Likewise.
3712         (lower_omp_1): Likewise.
3713         * predict.c (strip_predict_hints): Likewise.
3714         * print-tree.c (print_node): Likewise.
3715         * symtab.c (symtab_node::output_to_lto_symbol_table_p): Likewise.
3716         * trans-mem.c (is_tm_irrevocable): Likewise.
3717         (is_tm_load): Likewise.
3718         (is_tm_simple_load): Likewise.
3719         (is_tm_store): Likewise.
3720         (is_tm_simple_store): Likewise.
3721         (is_tm_abort): Likewise.
3722         (tm_region_init_1): Likewise.
3723         * tree-call-cdce.c (gen_shrink_wrap_conditions): Likewise.
3724         * tree-cfg.c (verify_gimple_call): Likewise.
3725         (move_stmt_r): Likewise.
3726         (stmt_can_terminate_bb_p): Likewise.
3727         * tree-eh.c (lower_eh_constructs_2): Likewise.
3728         * tree-if-conv.c (if_convertible_stmt_p): Likewise.
3729         * tree-inline.c (remap_gimple_stmt): Likewise.
3730         (copy_bb): Likewise.
3731         (estimate_num_insns): Likewise.
3732         (fold_marked_statements): Likewise.
3733         * tree-sra.c (scan_function): Likewise.
3734         * tree-ssa-ccp.c (surely_varying_stmt_p): Likewise.
3735         (optimize_stack_restore): Likewise.
3736         (pass_fold_builtins::execute): Likewise.
3737         * tree-ssa-dce.c (mark_stmt_if_obviously_necessary): Likewise.
3738         (mark_all_reaching_defs_necessary_1): Likewise.
3739         * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Likewise.
3740         * tree-ssa-forwprop.c (simplify_builtin_call): Likewise.
3741         (pass_forwprop::execute): Likewise.
3742         * tree-ssa-loop-im.c (stmt_cost): Likewise.
3743         * tree-ssa-math-opts.c (pass_cse_reciprocals::execute): Likewise.
3744         * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Likewise.
3745         * tree-ssa-strlen.c (get_string_length): Likewise.
3746         * tree-ssa-structalias.c (handle_lhs_call): Likewise.
3747         (find_func_aliases_for_call): Likewise.
3748         * tree-ssa-ter.c (find_replaceable_in_bb): Likewise.
3749         * tree-stdarg.c (optimize_va_list_gpr_fpr_size): Likewise.
3750         * tree-tailcall.c (find_tail_calls): Likewise.
3751         * tree.c (need_assembler_name_p): Likewise.
3752         (free_lang_data_in_decl): Likewise.
3753         (get_call_combined_fn): Likewise.
3754         * ubsan.c (is_ubsan_builtin_p): Likewise.
3755         * varasm.c (incorporeal_function_p): Likewise.
3756         * tree.h (DECL_BUILT_IN): Remove and replace with
3757         fndecl_built_in_p.
3758         (DECL_BUILT_IN_P): Transfort to fndecl_built_in_p.
3759         (fndecl_built_in_p): New.
3761 2018-08-27  Martin Liska  <mliska@suse.cz>
3763         PR tree-optimization/86847
3764         * tree-switch-conversion.c (switch_decision_tree::dump_case_nodes):
3765         Dump also subtree probability.
3766         (switch_decision_tree::do_jump_if_equal): New function.
3767         (switch_decision_tree::emit_case_nodes): Handle special
3768         situations in balanced tree that can be emitted much simpler.
3769         Fix calculation of probabilities that happen in tree expansion.
3770         * tree-switch-conversion.h (struct cluster): Add
3771         is_single_value_p.
3772         (struct simple_cluster): Likewise.
3773         (struct case_tree_node): Add new function has_child.
3774         (do_jump_if_equal): New.
3776 2018-08-27  Martin Liska  <mliska@suse.cz>
3778         * tree-switch-conversion.c (bit_test_cluster::find_bit_tests):
3779         Add new argument to bit_test_cluster constructor.
3780         (bit_test_cluster::emit): Set bits really number of values
3781         handlel by a test.
3782         (bit_test_cluster::hoist_edge_and_branch_if_true): Add
3783         probability argument.
3784         * tree-switch-conversion.h (struct bit_test_cluster):
3785         Add m_handles_entire_switch.
3787 2018-08-27  Martin Liska  <mliska@suse.cz>
3789         PR tree-optimization/86702
3790         * tree-switch-conversion.c (jump_table_cluster::emit):
3791         Make probabilities even for values in jump table
3792         according to number of cases handled.
3793         (switch_decision_tree::compute_cases_per_edge): Pass
3794         argument to reset_out_edges_aux function.
3795         (switch_decision_tree::analyze_switch_statement): Likewise.
3796         * tree-switch-conversion.h (switch_decision_tree::reset_out_edges_aux):
3797         Make it static.
3799 2018-08-27  Martin Liska  <mliska@suse.cz>
3801         * cfgexpand.c (expand_asm_stmt): Use label_to_block and pass
3802         cfun argument explicitly.
3803         * gimple-pretty-print.c (dump_gimple_switch): Likewise.
3804         * hsa-gen.c (gen_hsa_insns_for_switch_stmt): Use new
3805         function gimple_switch_default_bb.
3806         (convert_switch_statements):
3807         (expand_builtins):
3808         * ipa-fnsummary.c (set_switch_stmt_execution_predicate):
3809         * stmt.c (label_to_block_fn): Use label_to_block and pass
3810         cfun argument explicitly and use gimple_switch_label_bb.
3811         (expand_case): Likewise.
3812         * tree-cfg.c (lower_phi_internal_fn): Use label_to_block and pass
3813         cfun argument explicitly. Likewise.
3814         (make_edges_bb): Likewise.
3815         (make_cond_expr_edges): Likewise.
3816         (get_cases_for_edge): Likewise.
3817         (make_gimple_switch_edges): Likewise.
3818         (label_to_block_fn): Likewise.
3819         (label_to_block): Likewise.
3820         (make_goto_expr_edges): Likewise.
3821         (make_gimple_asm_edges): Likewise.
3822         (main_block_label): Likewise.
3823         (group_case_labels_stmt): Likewise.
3824         (find_taken_edge_computed_goto): Likewise.
3825         (find_taken_edge_switch_expr): Likewise.
3826         (gimple_verify_flow_info): Likewise.
3827         (gimple_redirect_edge_and_branch): Likewise.
3828         (gimple_switch_label_bb): New function.
3829         (gimple_switch_default_bb): Likewise.
3830         (gimple_switch_edge): Likewise.
3831         (gimple_switch_default_edge): Likewise.
3832         * tree-cfg.h (label_to_block_fn): Remove and replace ...
3833         (label_to_block): ... with this.
3834         (gimple_switch_label_bb): New.
3835         (gimple_switch_default_bb): Likewise.
3836         (gimple_switch_edge): Likewise.
3837         (gimple_switch_default_edge): Likewise.
3838         * tree-cfgcleanup.c (convert_single_case_switch): Use
3839         new gimple functions and pass new argument to label_to_block.
3840         (cleanup_control_flow_bb):
3841         * tree-eh.c (make_eh_dispatch_edges): Use label_to_block and pass
3842         cfun argument explicitly.
3843         (make_eh_edges): Likewise.
3844         (redirect_eh_dispatch_edge): Likewise.
3845         (lower_resx): Likewise.
3846         (lower_eh_dispatch): Likewise.
3847         (maybe_remove_unreachable_handlers): Likewise.
3848         (unsplit_eh): Likewise.
3849         (cleanup_empty_eh): Likewise.
3850         (verify_eh_edges): Likewise.
3851         (verify_eh_dispatch_edge): Likewise.
3852         * tree-ssa-dom.c (record_edge_info): Likewise.
3853         * tree-ssa-forwprop.c (simplify_gimple_switch_label_vec): Likewise.
3854         * tree-ssa-threadedge.c (thread_around_empty_blocks): Likewise.
3855         (thread_through_normal_block): Likewise.
3856         * tree-ssa-uncprop.c (associate_equivalences_with_edges): Likewise.
3857         * tree-ssa-uninit.c (convert_control_dep_chain_into_preds):
3858         * tree-switch-conversion.c (switch_conversion::collect): Use new
3859         gimple functions.
3860         (switch_conversion::check_final_bb): Likewise.
3861         (switch_conversion::gather_default_values): Pass new argument
3862         to label_to_block.
3863         (switch_conversion::build_constructors): Likewise.
3864         (switch_decision_tree::compute_cases_per_edge): Use new
3865         gimple_switch_edge function.
3866         (switch_decision_tree::analyze_switch_statement): Pass new argument
3867         to label_to_block.
3868         (switch_decision_tree::try_switch_expansion): Use
3869         gimple_switch_default_edge.
3870         * tree-vrp.c (find_switch_asserts): Pass new argument
3871         to label_to_block.
3872         * vr-values.c (vr_values::vrp_visit_switch_stmt): Likewise.
3873         (vr_values::simplify_switch_using_ranges): Likewise.
3875 2018-08-27  Richard Biener  <rguenther@suse.de>
3877         * cfganal.h (rev_post_order_and_mark_dfs_back_seme): Declare.
3878         * cfganal.c (rev_post_order_and_mark_dfs_back_seme): New function.
3880         * tree-ssa-sccvn.h (struct vn_pval): New structure.
3881         (struct vn_nary_op_s): Add unwind_to member.  Add
3882         predicated_values flag and put result into a union together
3883         with a linked list of vn_pval.
3884         (struct vn_ssa_aux): Add name member to make maintaining
3885         a map of SSA name to vn_ssa_aux possible.  Remove no longer
3886         needed info, dfsnum, low, visited, on_sccstack, use_processed
3887         and range_info_anti_range_p members.
3888         (run_scc_vn, vn_eliminate, free_scc_vn, vn_valueize): Remove.
3889         (do_rpo_vn, run_rpo_vn, eliminate_with_rpo_vn, free_rpo_vn):
3890         New functions.
3891         (vn_valueize): New global.
3892         (vn_context_bb): Likewise.
3893         (VN_INFO_RANGE_INFO, VN_INFO_ANTI_RANGE_P, VN_INFO_RANGE_TYPE,
3894         VN_INFO_PTR_INFO): Remove.
3895         * tree-ssa-sccvn.c: ... (rewrite)
3896         (pass_fre::execute): For -O2+ initialize loops and run
3897         RPO VN in optimistic mode (iterating).  For -O1 and -Og
3898         run RPO VN in non-optimistic mode.
3899         * params.def (PARAM_SCCVN_MAX_SCC_SIZE): Remove.
3900         (PARAM_RPO_VN_MAX_LOOP_DEPTH): Add.
3901         * doc/invoke.texi (sccvn-max-scc-size): Remove.
3902         (rpo-vn-max-loop-depth): Document.
3903         * tree-ssa-alias.c (walk_non_aliased_vuses): Stop walking
3904         when valuezing the VUSE signals we walked out of the region.
3905         * tree-ssa-pre.c (phi_translate_1): Ignore predicated values.
3906         (phi_translate): Set VN context block to use for availability
3907         lookup.
3908         (compute_avail): Likewise.
3909         (pre_valueize): New function.
3910         (pass_pre::execute): Adjust to the RPO VN API.
3912         * tree-ssa-loop-ivcanon.c: Include tree-ssa-sccvn.h.
3913         (propagate_constants_for_unrolling): Remove.
3914         (tree_unroll_loops_completely): Perform value-numbering
3915         on the unrolled bodies loop parent.
3917 2018-08-27  Richard Biener  <rguenther@suse.de>
3919         * tree-ssa-pre.c (compute_antic): Re-use inverted postorder
3920         for partial antic compute.
3922 2018-08-27  Jakub Jelinek  <jakub@redhat.com>
3924         PR rtl-optimization/87065
3925         * combine.c (simplify_if_then_else): Formatting fix.
3926         (if_then_else_cond): Guard MULT optimization with SCALAR_INT_MODE_P
3927         check.
3928         (known_cond): Don't return const_true_rtx for vector modes.  Use
3929         CONST0_RTX instead of const0_rtx.  Formatting fixes.
3931 2018-08-27  Martin Liska  <mliska@suse.cz>
3933         PR gcov-profile/87069
3934         * gcov.c (process_file): Record files already processed
3935         and warn about a file being processed multiple times.
3937 2018-08-27  Martin Liska  <mliska@suse.cz>
3939         PR driver/83193
3940         * config/aarch64/aarch64.c (aarch64_override_options_internal):
3941         Set default values for x_aarch64_*_string strings.
3942         * config/aarch64/aarch64.opt: Remove --{march,mcpu,mtune}==
3943         prefix.  For -mabi do not print '=ABI' in help and use
3944         <option_value> format for -msve-vector-bits and -moverride
3945         options.
3947 2018-08-26  Jeff Law  <law@redhat.com>
3949         * config/mips/frame-header-opt.c: Include "backend.h" rather than
3950         "cfg.h"
3952 2018-08-26  Marek Polacek  <polacek@redhat.com>
3954         PR c++/87029, Implement -Wredundant-move.
3955         * doc/invoke.texi: Document -Wredundant-move.
3957 2018-08-25  Martin Sebor  <msebor@redhat.com>
3959         PR tree-optimization/87059
3960         * builtins.c (expand_builtin_strncmp): Convert MIN_EXPR operand
3961         to the same type as the other.
3962         * fold-const.c (fold_binary_loc): Assert expectation.
3964 2018-08-25  Iain Sandoe  <iain@sandoe.co.uk>
3966         * config/darwin.c (machopic_legitimize_pic_address): Clean up
3967         extraneous parentheses, dead code section and formatting.
3969 2018-08-24  David Malcolm  <dmalcolm@redhat.com>
3971         PR c++/87091
3972         * diagnostic-show-locus.c (layout::layout): Ensure the margin is
3973         wide enough for jumps in the line-numbering to be visible.
3974         (layout::print_gap_in_line_numbering): New member function.
3975         (layout::calculate_line_spans): When using line numbering, merge
3976         line spans that are only 1 line apart.
3977         (diagnostic_show_locus): When printing line numbers, show gaps in
3978         line numbering directly, rather than printing headers.
3979         (selftest::test_diagnostic_show_locus_fixit_lines): Add test of
3980         line-numbering with multiple line spans.
3981         (selftest::test_fixit_insert_containing_newline_2): Add test of
3982         line-numbering, in which the spans are close enough to be merged.
3984 2018-08-24  Aldy Hernandez  <aldyh@redhat.com>
3986         * gimple-ssa-evrp-analyze.c (set_ssa_range_info): Pass value_range
3987         to range_includes_zero_p.  Do not special case VR_ANTI_RANGE.
3988         * tree-vrp.c (range_is_nonnull): Remove.
3989         (range_includes_zero_p): Accept value_range instead of min/max.
3990         (extract_range_from_binary_expr_1): Do not early bail on
3991         POINTER_PLUS_EXPR.
3992         Use range_includes_zero_p instead of range_is_nonnull.
3993         (extract_range_from_unary_expr): Use range_includes_zero_p instead
3994         of range_is_nonnull.
3995         (vrp_meet_1): Pass value_range to range_includes_zero_p.  Do not
3996         special case VR_ANTI_RANGE.
3997         (vrp_finalize): Same.
3998         * tree-vrp.h (range_includes_zero_p): Pass value_range as argument
3999         instead of min/max.
4000         (range_is_nonnull): Remove.
4001         * vr-values.c (vrp_stmt_computes_nonzero): Use
4002         range_includes_zero_p instead of range_is_nonnull.
4003         (extract_range_basic): Pass value_range to range_includes_zero_p
4004         instead of range_is_nonnull.
4006 2018-08-24  Uros Bizjak  <ubizjak@gmail.com>
4008         * emit-rtl.c (init_emit_once): Do not emit MODE_POINTER_BOUNDS RTXes.
4009         * emit-rtl.h (rtl_data): Remove return_bnd.
4010         * explow.c (trunc_int_for_mode): Do not handle POINTER_BOUNDS_MODE_P.
4011         * function.c (diddle_return_value): Do not handle crtl->return_bnd.
4012         * genmodes.c (complete_mode): Do not handle MODE_POINTER_BOUNDS.
4013         (POINTER_BOUNDS_MODE): Remove definition.
4014         (make_pointer_bounds_mode): Remove.
4015         (get_mode_class): Do not handle MODE_POINTER_BOUNDS.
4016         * machmode.h (POINTER_BOUNDS_MODE_P): Remove definition.
4017         (scalare_mode::includes_p): Do not handle MODE_POINTER_BOUNDS.
4018         * mode-classes.def: Do not define MODE_POINTER_BOUNDS.
4019         * stor-layout.c (int_mode_for_mode): Do not handle MODE_POINTER_BOUNDS.
4020         * tree-core.h (enum tree_index): Remove TI_POINTER_BOUNDS_TYPE.
4021         * varasm.c (output_constant_pool_2): Do not handle MODE_POINTER_BOUNDS.
4023         * config/i386/i386-modes.def (BND32, BND64): Remove.
4024         * config/i386/i386.c (dbx_register_map): Remove bound registers.
4025         (dbx64_register_map): Ditto.
4026         (svr4_dbx_register_map): Ditto.
4027         (indirect_thunk_bnd_needed): Remove.
4028         (indirect_thunks_bnd_used): Ditto.
4029         (indirect_return_bnd_needed): Ditto.
4030         (indirect_return_via_cx_bnd): Ditto.
4031         (enum indirect_thunk_prefix): Remove indirect_thunk_prefix_bnd.
4032         (indirect_thunk_name): Remove handling of indirect_thunk_prefix_bnd.
4033         (output_indirect_thunk): Ditto.  Remove need_prefix argument.
4034         (output_indirect_thunk_function): Remove handling of
4035         indirect_return_bnd_needed, indirect_return_via_cx_bnd,
4036         indirect_thunk_bnd_needed and indirect_thunks_bnd_used variables.
4037         (ix86_save_reg): Remove handling of crtl->return_bnd.
4038         (ix86_legitimate_constant_p): Remove handling of POINTER_BOUNDS_MODE_P.
4039         (ix86_print_operand_address_as): Remove handling of UNSPEC_BNDMK_ADDR
4040         and UNSPEC_BNDLX_ADDR.
4041         (ix86_output_indirect_branch_via_reg): Remove handling of
4042         indirect_thunk_prefix_bnd.
4043         (ix86_output_indirect_branch_via_push): Ditto.
4044         (ix86_output_function_return): Ditto.
4045         (ix86_output_indirect_function_return): Ditto.
4046         (avoid_func_arg_motion): Do not handle UNSPEC_BNDSTX.
4047         * config/i386/i386.h (FIXED_REGISTERS): Remove bound registers.
4048         (CALL_USED_REGISTERS): Ditto.
4049         (REG_ALLOC_ORDER): Update for removal of bound registers.
4050         (HI_REGISTER_NAMES): Ditto.
4051         * config/i386/i386.md (UNSPEC_BNDMK, UNSPEC_BNDMK_ADDR, UNSPEC_BNDSTX)
4052         (UNSPEC_BNDLDX, UNSPEC_BNDLDX_ADDR, UNSPEC_BNDCL, UNSPEC_BNDCU)
4053         (UNSPEC_BNDCN, UNSPEC_MPX_FENCE): Remove.
4054         (BND0_REG, BND1_REG, BND2_REG, BND3_REG): Remove
4055         (FIRST_PSEUDO_REG): Update.
4056         (BND): Remove mode iterator.
4057         * config/i386/predicates.md (bnd_mem_operator): Remove.
4059 2018-08-24  Richard Sandiford  <richard.sandiford@arm.com>
4061         * tree-vect-stmts.c (vectorizable_bswap): Handle variable-length
4062         vectors.
4064 2018-08-24  Richard Sandiford  <richard.sandiford@arm.com>
4066         * tree-vect-slp.c (vect_transform_slp_perm_load): Separate out
4067         the case in which the permute needs only a single element and
4068         repeats for every vector of the result.  Extend that case to
4069         handle variable-length vectors.
4070         * tree-vect-stmts.c (vectorizable_load): Update accordingly.
4072 2018-08-24  H.J. Lu  <hongjiu.lu@intel.com>
4074         PR debug/79342
4075         * dwarf2out.c (save_macinfo_strings): Call set_indirect_string
4076         on DW_MACINFO_start_file for -gsplit-dwarf -g3.
4078 2018-08-24  Richard Biener  <rguenther@suse.de>
4080         * cfg.h (struct control_flow_graph): Add edge_flags_allocated and
4081         bb_flags_allocated members.
4082         (auto_flag): New RAII class for allocating flags.
4083         (auto_edge_flag): New RAII class for allocating edge flags.
4084         (auto_bb_flag): New RAII class for allocating bb flags.
4085         * cfgloop.c (verify_loop_structure): Allocate temporary edge
4086         flag dynamically.
4087         * cfganal.c (dfs_enumerate_from): Remove use of visited sbitmap
4088         in favor of temporarily allocated BB flag.
4089         * hsa-brig.c: Re-order includes.
4090         * hsa-dump.c: Likewise.
4091         * hsa-regalloc.c: Likewise.
4092         * print-rtl.c: Likewise.
4093         * profile-count.c: Likewise.
4095 2018-08-24  Segher Boessenkool  <segher@kernel.crashing.org>
4097         PR target/86989
4098         * config/rs6000/rs6000.c (toc_relative_expr_p): Check that the base is
4099         the TOC register.
4101 2018-08-24  Aldy Hernandez  <aldyh@redhat.com>
4103         PR 87073/bootstrap
4104         * wide-int-range.cc (wide_int_range_div): Do not ignore result
4105         from wide_int_range_multiplicative_op.
4107 2018-08-23  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
4109         * tree-vect-data-refs.c (vect_grouped_store_supported): Fix typo
4110         "permutaion".
4112 2018-08-23  Giuliano Belinassi  <giuliano.belinassi@usp.br>
4114         * genmatch.c (parser::parse_operation): Fix typo 'exapnded'
4115         to 'expanded'.
4117 2018-08-23  Alexander Monakov  <amonakov@ispras.ru>
4119         * tree-scalar-evolution.c (final_value_replacement_loop): Dump
4120         full GENERIC expression used for replacement.
4122 2018-08-23  Aldy Hernandez  <aldyh@redhat.com>
4124         * tree-vrp.c (abs_extent_range): Remove.
4125         (extract_range_into_wide_ints): Pass wide ints by reference.
4126         (extract_range_from_binary_expr_1): Rewrite the *DIV_EXPR code.
4127         Pass wide ints by reference in all calls to
4128         extract_range_into_wide_ints.
4129         * wide-int-range.cc (wide_int_range_div): New.
4130         * wide-int-range.h (wide_int_range_div): New.
4131         (wide_int_range_includes_zero_p): New.
4132         (wide_int_range_zero_p): New.
4134 2018-08-23  Matthew Malcomson  <matthew.malcomson@arm.com>
4136         * config/aarch64/aarch64.md (arches): New enum.
4137         (arch): New enum attr.
4138         (arch_enabled): New attr.
4139         (enabled): Now uses arch_enabled only.
4140         (simd, sve, fp16): Removed attribute.
4141         (fp): Attr now defined in terms of 'arch'.
4142         (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64, *movti_aarch64,
4143         *movhf_aarch64, <optab><fcvt_target><GPF:mode>2,
4144         <FCVT_F2FIXED:fcvt_fixed_insn><GPF:mode>3,
4145         <FCVT_FIXED2F:fcvt_fixed_insn><GPI:mode>3): Merge 'fp' and 'simd'
4146         attributes into 'arch'.
4147         (*movsf_aarch64, *movdf_aarch64, *movtf_aarch64, *add<mode>3_aarch64,
4148         subdi3, neg<mode>2, <optab><mode>3, one_cmpl<mode>2,
4149         *<NLOGICAL:optab>_one_cmpl<mode>3, *xor_one_cmpl<mode>3,
4150         *aarch64_ashl_sisd_or_int_<mode>3, *aarch64_lshr_sisd_or_int_<mode>3,
4151         *aarch64_ashr_sisd_or_int_<mode>3, *aarch64_sisd_ushl): Convert use of
4152         'simd' attribute into 'arch'.
4153         (load_pair_sw_<SX:mode><SX2:mode>, load_pair_dw_<DX:mode><DX2:mode>,
4154         store_pair_sw_<SX:mode><SX2:mode>, store_pair_dw_<DX:mode><DX2:mode>):
4155         Convert use of 'fp' attribute to 'arch'.
4156         * config/aarch64/aarch64-simd.md (move_lo_quad_internal_<mode>,
4157         move_lo_quad_internal_<mode>): (different modes) Merge 'fp' and 'simd'
4158         into 'arch'.
4159         (move_lo_quad_internal_be_<mode>, move_lo_quad_internal_be_<mode>):
4160         (different modes) Merge 'fp' and 'simd' into 'arch'.
4161         (*aarch64_combinez<mode>, *aarch64_combinez_be<mode>): Merge 'fp' and
4162         'simd' into 'arch'.
4164 2018-08-23  Segher Boessenkool  <segher@kernel.crashing.org>
4166         PR rtl-optimization/87026
4167         * expmed.c (canonicalize_comparison): If we can no longer create
4168         pseudoregisters, don't.
4170 2018-08-23  Richard Earnshaw  <rearnsha@arm.com>
4172         PR target/86951
4173         * config/arm/arm-protos.h (arm_emit_speculation_barrier): New
4174         prototype.
4175         * config/arm/arm.c (speculation_barrier_libfunc): New static
4176         variable.
4177         (arm_init_libfuncs): Initialize it.
4178         (arm_emit_speculation_barrier): New function.
4179         * config/arm/arm.md (speculation_barrier): Call
4180         arm_emit_speculation_barrier for architectures that do not have
4181         DSB or ISB.
4182         (speculation_barrier_insn): Only match on Armv7 or later.
4184 2018-08-23  Richard Biener  <rguenther@suse.de>
4186         PR middle-end/87024
4187         * tree-inline.c (copy_bb): Drop unused __builtin_va_arg_pack_len
4188         calls.
4190 2018-08-23  Richard Sandiford  <richard.sandiford@arm.com>
4192         * config/aarch64/aarch64.c (aarch64_evpc_sve_tbl): Fix handling
4193         of single-vector TBLs.
4194         (aarch64_vectorize_vec_perm_const): Set one_vector_p when only
4195         one input is given.
4197 2018-08-23  Richard Sandiford  <richard.sandiford@arm.com>
4199         PR target/85910
4200         * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const_1): Fix
4201         aarch64_evpc_tbl guard.
4203 2018-08-22  Bernd Edlinger  <bernd.edlinger@hotmail.de>
4205         * tree-ssa-dse.c (compute_trims): Avoid folding away undefined
4206         behaviour.
4208 2018-08-22  Martin Sebor  <msebor@redhat.com>
4210         PR middle-end/87052
4211         * tree-pretty-print.c (pretty_print_string): Add argument.
4212         (dump_generic_node): Call to pretty_print_string with string size.
4214 2018-08-22  Segher Boessenkool  <segher@kernel.crashing.org>
4216         PR rtl-optimization/86771
4217         * combine.c (try_combine): Do not allow splitting a resulting PARALLEL
4218         of two SETs into those two SETs, one to be placed at i2, if that SETs
4219         destination is modified between i2 and i3.
4221 2018-08-22  Richard Sandiford  <richard.sandiford@arm.com>
4223         PR tree-optimization/86725
4224         * tree-vect-loop.c (vect_inner_phi_in_double_reduction_p): New
4225         function.
4226         (vect_analyze_scalar_cycles_1): Check it.
4228 2018-08-22  Richard Sandiford  <richard.sandiford@arm.com>
4230         PR tree-optimization/86725
4231         * tree-vect-loop.c (vect_is_simple_reduction): When treating
4232         an outer loop phi as a double reduction, make sure that the
4233         single user of the phi result is an inner loop phi.
4235 2018-08-22  Richard Sandiford  <richard.sandiford@arm.com>
4237         * tree-vect-data-refs.c (vect_analyze_group_access_1): Convert
4238         grouped stores with gaps to a strided group.
4240 2018-08-22  Richard Sandiford  <richard.sandiford@arm.com>
4242         * tree-vect-stmts.c (get_group_load_store_type)
4243         (get_load_store_type): Only test STMT_VINFO_STRIDED_P for the
4244         first statement in a group.
4246 2018-08-22  Iain Sandoe  <iain@sandoe.co.uk>
4248         * config/darwin.h (LINK_COMMAND_SPEC_A): Sync LTO options with
4249         the sequence used in gcc/gcc.c.
4251 2018-08-22  Iain Sandoe  <iain@sandoe.co.uk>
4253         PR other/704
4254         * gcc-ar.c (main): Don’t try to invoke the plug-in if we’re not
4255         building it.
4257 2018-08-22  Iain Sandoe  <iain@sandoe.co.uk>
4259         * config/darwin10.h (LINK_GCC_C_SEQUENCE_SPEC): Adjust to use the
4260         Darwin10-specific unwinder-shim.
4261         * config/darwin12.h (LINK_GCC_C_SEQUENCE_SPEC): Remove.
4262         * config/rs6000/darwin.h (DARWIN_CRT1_SPEC, DARWIN_DYLIB1_SPEC):
4263         New to cater for Darwin10 Rosetta.
4265 2018-08-22  Iain Sandoe  <iain@sandoe.co.uk>
4267         * config/i386/i386.c (ix86_output_addr_diff_elt): Move the MACH-O
4268         specific test before the one for HAVE_AS_GOTOFF_IN_DATA.
4270 2018-08-22  Iain Sandoe  <iain@sandoe.co.uk>
4272         PR bootstrap/81033
4273         PR target/81733
4274         PR target/52795
4275         * gcc/dwarf2out.c (FUNC_SECOND_SECT_LABEL): New.
4276         (dwarf2out_switch_text_section): Generate a local label for the
4277         second function sub-section and apply it as the second FDE start
4278         label.
4279         * gcc/final.c (final_scan_insn_1): Emit second FDE label after the
4280         second sub-section start.
4282 2018-08-22  Richard Biener  <rguenther@suse.de>
4284         PR tree-optimization/86988
4285         * tree-vrp.c (vrp_prop::check_mem_ref): Bail out on VLAs.
4287 2018-08-22  Richard Biener  <rguenther@suse.de>
4289         PR tree-optimization/86945
4290         * tree-cfg.c (generate_range_test): Use unsigned arithmetic.
4292 2018-08-22  Alexandre Oliva <oliva@adacore.com>
4294         * config/rs6000/rs6000.c (SMALL_DATA_RELOC, SMALL_DATA_REG): Add
4295         a comment about how uses of r2 for .sdata2 come about.
4297 2018-08-22  Alexandre Oliva <aoliva@redhat.com>
4299         * tree-ssa-reassoc.c (is_reassociable_op): Fix cut&pasto.
4301 2018-08-21  Marek Polacek  <polacek@redhat.com>
4303         PR c++/86981, Implement -Wpessimizing-move.
4304         * doc/invoke.texi: Document -Wpessimizing-move.
4306 2018-08-21  Jan Hubicka  <jh@suse.cz>
4308         * tree.c (find_decls_types_r): Do not check for redundant typedefs.
4309         * tree.h (is_redundant_typedef): Remove.
4310         * dwarf2out.c (is_redundant_typedef): Turn into static function.
4312 2018-08-21  Jan Hubicka  <jh@suse.cz>
4314         * tree.c (free_lang_data_in_decl): Remove types from DECL_CONTEXT
4315         when possible.
4317 2018-08-21  Tamar Christina  <tamar.christina@arm.com>
4319         * expmed.c (extract_low_bits): Reject invalid subregs early.
4321 2018-08-21  Bernd Edlinger  <bernd.edlinger@hotmail.de>
4323         PR middle-end/86121
4324         * tree-ssa-strlen.c (adjust_last_stmt): Avoid folding away undefined
4325         behaviour.
4327 2018-08-21  Rasmus Villemoes  <rv@rasmusvillemoes.dk>
4329         * config/vxworks.h: Guard vxworks_asm_out_constructor and
4330         vxworks_asm_out_destructor by !HAVE_INITFINI_ARRAY_SUPPORT
4331         * config/vxworks.c: Likewise.
4333 2018-08-21  Rasmus Villemoes  <rv@rasmusvillemoes.dk>
4335         * config/vxworks.c: Set targetm.have_ctors_dtors
4336         if HAVE_INITFINI_ARRAY_SUPPORT.
4337         * config/vxworks.h: Set SUPPORTS_INIT_PRIORITY
4338         if HAVE_INITFINI_ARRAY_SUPPORT.
4340 2018-08-21  Rasmus Villemoes  <rv@rasmusvillemoes.dk>
4342         * config/vxworks.h: Add $(WIND_BASE)/target/h/wrn/coreip to
4343         default search path for VxWorks < 7.
4345 2018-08-21  Bernd Edlinger  <bernd.edlinger@hotmail.de>
4347         * gimple-ssa-sprintf.c (decl_constant_value): Remove.
4348         (get_format_string): Refer to c_getstr.
4350 2018-08-21  Tom de Vries  <tdevries@suse.de>
4352         * cgraph.h (debuginfo_early_init, debuginfo_init, debuginfo_fini)
4353         (debuginfo_start, debuginfo_stop, debuginfo_early_start)
4354         (debuginfo_early_stop): Declare.
4355         * cgraphunit.c (debuginfo_early_init, debuginfo_init, debuginfo_fini)
4356         (debuginfo_start, debuginfo_stop, debuginfo_early_start)
4357         (debuginfo_early_stop): New function.
4358         (symbol_table::finalize_compilation_unit): Call debuginfo_early_start
4359         and debuginfo_early_stop.
4360         * dwarf2out.c (dwarf2out_finish, dwarf2out_early_finish): Dump dwarf.
4361         * toplev.c (compile_file): Call debuginfo_start and debuginfo_stop.
4362         (general_init): Call debuginfo_early_init.
4363         (finalize): Call debuginfo_fini.
4364         (do_compile): Call debuginfo_init.
4365         * doc/invoke.texi (@gccoptlist): Add -fdump-debug and
4366         -fdump-early-debug.
4367         (@item -fdump-debug, @item -fdump-earlydebug): Add.
4369 2018-08-21  Tom de Vries  <tdevries@suse.de>
4371         * dwarf2out.c (print_dw_val, print_loc_descr, print_die): Handle
4372         flag_dump_noaddr and flag_dump_unnumbered.
4374 2018-08-21  Aldy Hernandez  <aldyh@redhat.com>
4376         * wide-int-range.cc (wide_int_range_abs): New.
4377         (wide_int_range_order_set): Rename from wide_int_range_min_max.
4378         * wide-int-range.h (wide_int_range_abs): New.
4379         (wide_int_range_min_max): New.
4380         * tree-vrp.c (extract_range_from_unary_expr): Rewrite ABS_EXPR
4381         case to call wide_int_range_abs.
4382         Rewrite MIN/MAX_EXPR to call wide_int_range_min_max.
4383         (extract_range_from_abs_expr): Delete.
4385 2018-08-20  Michael Meissner  <meissner@linux.ibm.com>
4387         PR target/87033
4388         * config/rs6000/rs6000.md (extendsi<mode>2): Change constraints
4389         from 'Y' to 'YZ' to enable the LWAX instruction to be generated
4390         for indexed loads.
4392 2018-08-20  Nathan Sidwell  <nathan@acm.org>
4393             Jeff Law <law@redhat.com>
4395         * config/s390/s390-c (s390_macro_to_expand): Use cpp_macro_p.
4396         * config/spu/spu-c.c (spu_macro_to_expand): Likewise.
4398 2018-08-20  David Malcolm  <dmalcolm@redhat.com>
4400         PR other/84889
4401         * attribs.c (diag_attr_exclusions): Add auto_diagnostic_group instance.
4402         (decl_attributes): Likewise.
4403         * calls.c (maybe_warn_nonstring_arg): Add auto_diagnostic_group
4404         instance.
4405         * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
4406         * diagnostic-core.h (class auto_diagnostic_group): New class.
4407         * diagnostic.c (diagnostic_initialize): Initialize the new fields.
4408         (diagnostic_report_diagnostic): Handle the first diagnostics within
4409         a group.
4410         (emit_diagnostic): Add auto_diagnostic_group instance.
4411         (inform): Likewise.
4412         (inform_n): Likewise.
4413         (warning): Likewise.
4414         (warning_at): Likewise.
4415         (warning_n): Likewise.
4416         (pedwarn): Likewise.
4417         (permerror): Likewise.
4418         (error): Likewise.
4419         (error_n): Likewise.
4420         (error_at): Likewise.
4421         (sorry): Likewise.
4422         (fatal_error): Likewise.
4423         (internal_error): Likewise.
4424         (internal_error_no_backtrace): Likewise.
4425         (auto_diagnostic_group::auto_diagnostic_group): New ctor.
4426         (auto_diagnostic_group::~auto_diagnostic_group): New dtor.
4427         * diagnostic.h (struct diagnostic_context): Add fields
4428         "diagnostic_group_nesting_depth",
4429         "diagnostic_group_emission_count", "begin_group_cb",
4430         "end_group_cb".
4431         * gimple-ssa-isolate-paths.c (find_implicit_erroneous_behavior):
4432         Add auto_diagnostic_group instance(s).
4433         (find_explicit_erroneous_behavior): Likewise.
4434         * gimple-ssa-warn-alloca.c (pass_walloca::execute): Likewise.
4435         * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Likewise.
4436         * gimplify.c (warn_implicit_fallthrough_r): Likewise.
4437         (gimplify_va_arg_expr): Likewise.
4438         * hsa-gen.c (HSA_SORRY_ATV): Likewise.
4439         (HSA_SORRY_AT): Likewise.
4440         * ipa-devirt.c (compare_virtual_tables): Likewise.
4441         (warn_odr): Likewise.
4442         * multiple_target.c (expand_target_clones): Likewise.
4443         * opts-common.c (cmdline_handle_error): Likewise.
4444         * reginfo.c (globalize_reg): Likewise.
4445         * substring-locations.c (format_warning_n_va): Likewise.
4446         * tree-inline.c (expand_call_inline): Likewise.
4447         * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Likewise.
4448         * tree-ssa-loop-niter.c
4449         (do_warn_aggressive_loop_optimizations): Likewise.
4450         * tree-ssa-uninit.c (warn_uninit): Likewise.
4451         * tree.c (warn_deprecated_use): Likewise.
4453 2018-08-20  H.J. Lu  <hongjiu.lu@intel.com>
4455         PR target/87014
4456         * config/i386/i386.md (eh_return): Always update EH return
4457         address in word_mode.
4459 2018-08-20  Chung-Lin Tang  <cltang@codesourcery.com>
4461         * targhooks.c (std_gimplify_va_arg_expr): Properly handle case of when
4462         TARGET_SPLIT_COMPLEX_ARG is defined.
4464 2018-08-20  Bernd Edlinger  <bernd.edlinger@hotmail.de>
4466         * expr.c (store_field): Change gcc_assert to gcc_checking_assert.
4468 2018-08-20  Bernd Edlinger  <bernd.edlinger@hotmail.de>
4470         PR target/86984
4471         * expr.c (expand_assignment): Assert that bitpos is positive.
4472         (store_field): Likewise
4473         (expand_expr_real_1): Make sure that bitpos is positive.
4474         * config/alpha/alpha.h (CONSTANT_ADDRESS_P): Avoid signed
4475         integer overflow.
4477 2018-08-20  Nathan Sidwell  <nathan@acm.org>
4479         * Makefile.in (CPP_ID_DATA_H): Delete.
4480         (CPP_INTERNAL_H): Don't add it.
4481         (GTFILES): Replace CPP_ID_DATA_H with CPPLIB_H.
4482         * gengtype.c (open_base_files): Replace cpp-id-data.h with cpplib.h
4484 2018-08-20  Richard Biener  <rguenther@suse.de>
4486         PR tree-optimization/78655
4487         * tree-vrp.c (extract_range_from_binary_expr_1): Make
4488         pointer + offset nonnull if either operand is nonnull work.
4490 2018-08-20  Tom de Vries  <tdevries@suse.de>
4492         * dwarf2out.c (add_scalar_info): Don't add reference to existing die
4493         unless the referenced die describes the added property using
4494         DW_AT_location or DW_AT_const_value.  Fall back to exprloc case.
4495         Otherwise, add a DW_AT_location to the referenced die.
4497 2018-08-19  Uros Bizjak  <ubizjak@gmail.com>
4499         PR target/86994
4500         * config/i386/i386.c (ix86_rtx_costs) [case SET]: Check source for
4501         register_operand when calling ix86_set_reg_reg_cost.
4502         [case CONST_INT, case CONST, case LABEL_REF, case SYMBOL_REF]:
4503         Set *total to 0 for operands that satisfy x86_64_immediate_operand
4504         predicate and to 1 otherwise.
4506 2018-08-18  Iain Sandoe  <iain@sandoe.co.uk>
4508         * config/darwin.c (darwin_override_options): If -gsplit-dwarf is set,
4509         emit a diagnostic that it is not supported and reset the option.
4510         * config/darwin.h (DRIVER_SELF_SPECS): Note that gsplit-dwarf is not
4511         supported and consume the option.  (ASM_FINAL_SPEC): New.
4513 2018-08-17  Segher Boessenkool  <segher@kernel.crashing.org>
4515         * doc/md.texi (Patterns): Use @ref instead of @xref in the middle of
4516         a sentence.
4518 2018-08-17  Sandra Loosemore  <sandra@codesourcery.com>
4520         C-SKY port: Documentation
4522         * doc/extend.texi (C-SKY Function Attributes): New section.
4523         * doc/invoke.texi (Option Summary): Add C-SKY options.
4524         (C-SKY Options): New section.
4525         * doc/md.texi (Machine Constraints): Document C-SKY constraints.
4527 2018-08-17  Jojo  <jijie_rong@c-sky.com>
4528             Huibin Wang  <huibin_wang@c-sky.com>
4529             Sandra Loosemore  <sandra@codesourcery.com>
4530             Chung-Lin Tang  <cltang@codesourcery.com>
4532         C-SKY port: Backend implementation
4534         * config/csky/*: New.
4535         * common/config/csky/*: New.
4537 2018-08-17  Jojo  <jijie_rong@c-sky.com>
4538             Huibin Wang  <huibin_wang@c-sky.com>
4539             Sandra Loosemore  <sandra@codesourcery.com>
4540             Chung-Lin Tang  <cltang@codesourcery.com>
4541             Andrew Jenner  <andrew@codesourcery.com>
4543         C-SKY port: Configury
4545         * config.gcc (csky-*-*): New.
4546         * configure.ac: Add csky to targets for dwarf2 debug_line support.
4547         * configure: Regenerated.
4549 2018-08-17  David Malcolm  <dmalcolm@redhat.com>
4551         * dump-context.h: Include "dumpfile.h".
4552         (dump_context::dump_printf_va): Convert final param from va_list
4553         to va_list *.  Convert from ATTRIBUTE_PRINTF to
4554         ATTRIBUTE_GCC_DUMP_PRINTF.
4555         (dump_context::dump_printf_loc_va): Likewise.
4556         * dumpfile.c: Include "stringpool.h".
4557         (make_item_for_dump_printf_va): Delete.
4558         (make_item_for_dump_printf): Delete.
4559         (class dump_pretty_printer): New class.
4560         (dump_pretty_printer::dump_pretty_printer): New ctor.
4561         (dump_pretty_printer::emit_items): New member function.
4562         (dump_pretty_printer::emit_any_pending_textual_chunks): New member
4563         function.
4564         (dump_pretty_printer::emit_item): New member function.
4565         (dump_pretty_printer::stash_item): New member function.
4566         (dump_pretty_printer::format_decoder_cb): New member function.
4567         (dump_pretty_printer::decode_format): New member function.
4568         (dump_context::dump_printf_va): Reimplement in terms of
4569         dump_pretty_printer.
4570         (dump_context::dump_printf_loc_va): Convert final param from va_list
4571         to va_list *.
4572         (dump_context::begin_scope): Reimplement call to
4573         make_item_for_dump_printf.
4574         (dump_printf): Update for change to dump_printf_va.
4575         (dump_printf_loc): Likewise.
4576         (selftest::test_capture_of_dump_calls): Convert "stmt" from
4577         greturn * to gimple *.  Add a test_decl.  Add tests of dump_printf
4578         with %T, %E, and %G.
4579         * dumpfile.h (ATTRIBUTE_GCC_DUMP_PRINTF): New macro.
4580         (dump_printf): Replace ATTRIBUTE_PRINTF_2 with
4581         ATTRIBUTE_GCC_DUMP_PRINTF (2, 3).
4582         (dump_printf_loc): Replace ATTRIBUTE_PRINTF_3 with
4583         ATTRIBUTE_GCC_DUMP_PRINTF (3, 0).
4584         * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Convert
4585         use of HOST_WIDE_INT_PRINT_DEC on unsigned HOST_WIDE_INT "count"
4586         within a dump_printf_loc call to "%wu".
4587         (vector_alignment_reachable_p): Merge two dump_printf[_loc] calls,
4588         converting a use of HOST_WIDE_INT_PRINT_DEC to "%wd".  Add a
4589         missing space after "=".
4590         * tree-vect-loop.c (vect_analyze_loop_2) Within a dump_printf
4591         call, convert use of HOST_WIDE_INT_PRINT_DEC to "%wd".
4592         * tree-vect-slp.c (vect_slp_bb): Within a dump_printf_loc call,
4593         convert use of HOST_WIDE_INT_PRINT_UNSIGNED to "%wu".
4594         * tree-vectorizer.c (try_vectorize_loop_1): Likewise.  Remove
4595         duplicate "vectorized" from message.
4597 2018-08-17  Szabolcs Nagy  <szabolcs.nagy@arm.com>
4599         * config/arm/arm-builtins.c (arm_init_simd_builtin_types): Clear
4600         polyNxK_t element's TYPE_STRING_FLAG.
4602 2018-08-17  Segher Boessenkool  <segher@kernel.crashing.org>
4604         * config/rs6000/rs6000.md (*cbranch, *creturn): Name these patterns
4605         (they were unnamed before).  Fix comments.
4607 2018-08-17  Nathan Sidwell  <nathan@acm.org>
4609         * cppbuiltin.c: Include "cpplib.h", not "cpp-id-data.h".
4611 2018-08-17  Richard Biener  <rguenther@suse.de>
4613         PR tree-optimization/86841
4614         * wide-int-range.cc (wide_int_range_lshift): Use to_uhwi.
4616 2018-08-17  Martin Liska  <mliska@suse.cz>
4618         * common.opt: Remove Warn, Init and Report for options with
4619         Ignore/Deprecated flag. Warning is done automatically for
4620         Deprecated flags.
4621         * config/i386/i386.opt: Likewise.
4622         * config/ia64/ia64.opt: Likewise.
4623         * config/rs6000/rs6000.opt: Likewise.
4624         * cppbuiltin.c (define_builtin_macros_for_compilation_flags):
4625         Remove usage of flag_check_pointer_bounds.
4626         * lto-wrapper.c (merge_and_complain): Do not handle
4627         OPT_fcheck_pointer_bounds.
4628         (append_compiler_options): Likewise.
4629         * opt-functions.awk: Do not handle Deprecated.
4630         * optc-gen.awk: Check that Var, Report and Init are not
4631         used for an option with Ignore/Deprecated flag.
4632         * opts-common.c (decode_cmdline_option): Do not report
4633         CL_ERR_DEPRECATED.
4634         (read_cmdline_option): Report warning for OPT_SPECIAL_deprecated
4635         options.
4636         * opts.h (struct cl_option): Remove cl_deprecated flag.
4637         (CL_ERR_DEPRECATED): Remove error enum value.
4639 2018-08-17  Richard Biener  <rguenther@suse.de>
4641         PR middle-end/86505
4642         * tree-inline.c (copy_bb): When inlining __builtin_va_arg_pack_len ()
4643         across a va-arg-pack using call adjust its return value accordingly.
4645 2018-08-16  Martin Sebor  <msebor@redhat.com>
4647         PR tree-optimization/86853
4648         * gimple-ssa-sprintf.c (struct format_result): Rename member.
4649         (struct fmtresult): Add member and initialize it in ctors.
4650         (format_character): Handle %C.  Extend range to NUL.  Set MAYFAIL.
4651         (format_string): Handle %S the same as %ls.  Set MAYFAIL.
4652         (format_directive): Set POSUNDER4K when MAYFAIL is set.
4653         (parse_directive): Handle %C same as %c.
4654         (sprintf_dom_walker::compute_format_length): Adjust.
4655         (is_call_safe): Adjust.
4657 2018-08-16  Bernd Edlinger  <bernd.edlinger@hotmail.de>
4659         * builtins.c (c_strlen): Add new parameter eltsize.  Use it
4660         for determining how to count the elements.
4661         * builtins.h (c_strlen): Adjust prototype.
4662         * expr.c (string_constant): Add new parameter mem_size.
4663         Set *mem_size appropriately.
4664         * expr.h (string_constant): Adjust protoype.
4665         * gimple-fold.c (get_range_strlen): Add new parameter eltsize.
4666         * gimple-fold.h (get_range_strlen): Adjust prototype.
4667         * gimple-ssa-sprintf.c (get_string_length): Add new parameter eltsize.
4668         (format_string): Call get_string_length with eltsize.
4670 2018-08-16  David Malcolm  <dmalcolm@redhat.com>
4672         * diagnostic.c (default_diagnostic_start_span_fn): Call pp_string
4673         to emit the span, rather than setting it as the prefix.
4675 2018-08-16  David Malcolm  <dmalcolm@redhat.com>
4677         * diagnostic-show-locus.c (layout::start_annotation_line): Add
4678         "margin_char" parameter, defaulting to space.  Use it in place
4679         of pp_space for the initial part of the margin.
4680         (layout::print_leading_fixits): Use '+' when filling the margin
4681         of line-insertion fix-it hints.
4683 2018-08-16  Segher Boessenkool  <segher@kernel.crashing.org>
4685         * config/rs6000/rs6000.md (two unnamed define_insn and define_split):
4686         Delete.
4688 2018-08-16  Segher Boessenkool  <segher@kernel.crashing.org>
4690         * config/rs6000/altivec.md: Don't set length attribute to the default
4691         value.
4692         * config/rs6000/darwin.md: Ditto.
4693         * config/rs6000/dfp.md: Ditto.
4694         * config/rs6000/htm.md: Ditto.
4695         * config/rs6000/rs6000.md: Ditto.
4696         * config/rs6000/sync.md: Ditto.
4697         * config/rs6000/vsx.md: Ditto.
4699 2018-08-16  Segher Boessenkool  <segher@kernel.crashing.org>
4701         * config/rs6000/altivec.md: Don't set length attribute to the default
4702         value, for branch instructions.
4703         * config/rs6000/darwin.md: Ditto.
4704         * config/rs6000/rs6000.md: Ditto.
4706 2018-08-16  Segher Boessenkool  <segher@kernel.crashing.org>
4708         * config/rs6000/rs6000.md (length): Always define as const_int 4.
4709         (unnamed conditional branch define_insn): Set length to 4 or 8
4710         depending on offset.
4711         (<bd>_<mode>): Similar, for alternative 0.
4712         (<bd>tf_<mode>): Ditto.
4714 2018-08-16  Tamar Christina  <tamar.christina@arm.com>
4716         * expr.c (copy_blkmode_to_reg): Perform larger copies when safe.
4718 2018-08-16  Matthew Malcomson  <matthew.malcomson@arm.com>
4720         * doc/rtl.texi: Replace old RTX class names with new names.
4723 2018-08-16  Vlad Lazar  <vlad.lazar@arm.com>
4725         * expmed.h (canonicalize_comparison): New declaration.
4726         * expmed.c (canonicalize_comparison, equivalent_cmp_code): New function.
4727         * expmed.c (emit_store_flag_1): Add call to canonicalize_comparison.
4728         * optabs.c (prepare_cmp_insn): Likewise.
4729         * rtl.h (unsigned_condition_p): New function which checks if a
4730         comparison operator is unsigned.
4732 2018-08-16  Nathan Sidwell  <nathan@acm.org>
4734         * config/rs6000/rs6000-c.c (rs6000_macro_to_expend): Use cpp_macro_p.
4735         * config/powerpcspc/powerpcspe-c.c (rs6000_macro_to_expend): Likewise.
4737 2018-08-16  Tamar Christina  <tamar.christina@arm.com>
4739         PR target/84711
4740         * config/arm/arm.c (arm_can_change_mode_class): Disallow subreg.
4741         * config/arm/neon.md (movv4hf, movv8hf): Refactored to..
4742         (mov<mov>): ..this and enable unconditionally.
4744 2018-08-16  Tamar Christina  <tamar.christina@arm.com>
4746         * config/arm/neon.md (*neon_mov<mode>): Remove reg-to-reg alternative.
4748 2018-08-16  Sam Tebbs  <sam.tebbs@arm.com>
4750         * config/aarch64/aarch64.opt (mlow-precision-recip-sqrt)
4751         (mlow-precision-sqrt, mlow-precision-div, mverbose-cost-dump): Replace
4752         "Common" with "Target".
4754 2018-08-15  Uros Bizjak  <ubizjak@gmail.com>
4756         * config/i386/i386.opt (mmitigate-rop): Mark as deprecated.
4757         * doc/invoke.texi (mmitigate-rop): Remove.
4758         * config/i386/i386.c: Do not include "regrename.h".
4759         (ix86_rop_should_change_byte_p, reg_encoded_number)
4760         (ix86_get_modrm_for_rop, set_rop_modrm_reg_bits, ix86_mitigate_rop):
4761         Remove.
4762         (ix86_reorg): Remove call to ix86_mitigate_rop.
4763         * config/i386/i386.md (attr "modrm_class"): Remove.
4764         (cmp<mode>_ccno_1, mov<mode>_xor, movstrict<mode>_xor)
4765         (x86_mov<mode>cc_0_m1. x86_mov<mode>cc_0_m1_se)
4766         (x86_mov<mode>cc_0_m1_neg): Remove modrm_class attribute override.
4768 2018-08-15  Will Schmidt  <will_schmidt@vnet.ibm.com>
4770         * config/rs6000/rs600.c (rs6000_gimple_fold_builtin): Add entries to
4771         allow folding of mergeh() and mergel() for the float and double types.
4772         (fold_mergehl_helper): Rework to handle building a permute tree
4773         for float vectors.
4775 2018-08-15  Uros Bizjak  <ubizjak@gmail.com>
4777         * config/i386/i386.c (expand_vec_perm_movs): Enable V4SFmode
4778         for TARGET_SSE.
4780 2018-08-15  David Malcolm  <dmalcolm@redhat.com>
4782         * common.opt (fdiagnostics-show-labels): New option.
4783         * diagnostic-show-locus.c (class layout_range): Add field
4784         "m_label".
4785         (class layout): Add field "m_show_labels_p".
4786         (layout_range::layout_range): Add param "label" and use it to
4787         initialize m_label.
4788         (make_range): Pass in NULL for new "label" param of layout_range's
4789         ctor.
4790         (layout::layout): Initialize m_show_labels_p.
4791         (layout::maybe_add_location_range): Pass in loc_range->m_label
4792         when constructing layout_range instances.
4793         (struct line_label): New struct.
4794         (layout::print_any_labels): New member function.
4795         (layout::print_line): Call it if label-printing is enabled.
4796         (selftest::test_one_liner_labels): New test.
4797         (selftest::test_diagnostic_show_locus_one_liner): Call it.
4798         * diagnostic.c (diagnostic_initialize): Initialize
4799         context->show_labels_p.
4800         * diagnostic.h (struct diagnostic_context): Add field
4801         "show_labels_p".
4802         * doc/invoke.texi (Diagnostic Message Formatting Options): Add
4803         -fno-diagnostics-show-labels.
4804         * dwarf2out.c (gen_producer_string): Add
4805         OPT_fdiagnostics_show_labels to the ignored options.
4806         * gcc-rich-location.c (gcc_rich_location::add_expr): Add "label"
4807         param.
4808         (gcc_rich_location::maybe_add_expr): Likewise.
4809         * gcc-rich-location.h (gcc_rich_location::gcc_rich_location): Add
4810         label" param, defaulting to NULL.
4811         (gcc_rich_location::add_expr): Add "label" param.
4812         (gcc_rich_location::maybe_add_expr): Likewise.
4813         (class text_range_label): New class.
4814         (class range_label_for_type_mismatch): New class.
4815         * gimple-ssa-sprintf.c (fmtwarn): Pass NULL for new label params
4816         of format_warning_va.
4817         (fmtwarn_n): Likewise for new params of format_warning_n_va.
4818         * lto-wrapper.c (merge_and_complain): Add
4819         OPT_fdiagnostics_show_labels to the "pick one setting" options.
4820         (append_compiler_options): Likewise to the dropped options.
4821         (append_diag_options): Likewise to the passed-on options.
4822         * opts.c (common_handle_option): Handle the new option.
4823         * selftest-diagnostic.c
4824         (test_diagnostic_context::test_diagnostic_context): Enable
4825         show_labels_p.
4826         * substring-locations.c: Include "gcc-rich-location.h".
4827         (format_warning_n_va): Add "fmt_label" and "param_label" params
4828         and use them as appropriate.
4829         (format_warning_va): Add "fmt_label" and "param_label" params,
4830         passing them on to format_warning_n_va.
4831         (format_warning_at_substring): Likewise.
4832         (format_warning_at_substring_n): Likewise.
4833         * substring-locations.h (format_warning_va): Add "fmt_label" and
4834         "param_label" params.
4835         (format_warning_n_va): Likewise.
4836         (format_warning_at_substring): Likewise.
4837         (format_warning_at_substring_n): Likewise.
4838         * toplev.c (general_init): Initialize global_dc->show_labels_p.
4840 2018-08-15  Qing Zhao  <qing.zhao@oracle.com>
4842         PR testsuite/86519
4843         * builtins.c (expand_builtin_memcmp): Do not expand the call
4844         when overflow is detected.
4846 2018-08-15  Martin Sebor  <msebor@redhat.com>
4848         PR tree-optimization/71625
4849         * config/aarch64/aarch64-builtins.c
4850         (aarch64_init_simd_builtin_types): Clear Poly8_t's TYPE_STRING_FLAG.
4852 2018-08-15  Ilya Leoshkevich  <iii@linux.ibm.com>
4854         * config/s390/s390.c (s390_reorg): Remove loop.
4856 2018-08-15  Iain Sandoe  <iain@sandoe.co.uk>
4858         * config/darwin.c
4859          (darwin_function_switched_text_sections): Delete.
4860         * gcc/config/darwin.h
4861          (TARGET_ASM_FUNCTION_SWITCHED_TEXT_SECTIONS): Likewise.
4863 2018-08-15  Iain Sandoe  <iain@sandoe.co.uk>
4865         PR target/81685
4866         * config/darwin.h: (DEBUG_STR_OFFSETS_SECTION, DEBUG_LOCLISTS_SECTION,
4867         DEBUG_RNGLISTS_SECTION) new macros.  (DEBUG_PUBNAMES_SECTION,
4868         DEBUG_PUBTYPES_SECTION) update to include GNU variant.
4870 2018-08-15  Martin Liska  <mliska@suse.cz>
4872         PR tree-optimization/86925
4873         * predict.c (expr_expected_value_1): When taking
4874         later predictor, assign also probability.
4875         Use fold_build2_initializer_loc in order to fold
4876         the expression in -frounding-math.
4878 2018-08-14  Allan Sandfeld Jensen <allan.jensen@qt.io>
4880         * config/i386/i386.c (expand_vec_perm_movs): New method matching movs
4881         patterns.
4882         (expand_vec_perm_1): Try the new method.
4884 2018-08-14  Ilya Leoshkevich  <iii@linux.ibm.com>
4886         PR target/86547
4887         * lra-lives.c (remove_some_program_points_and_update_live_ranges):
4888         Check whether lra_live_max_point is 0 before dividing.
4890 2018-08-14  Martin Sebor  <msebor@redhat.com>
4892         PR tree-optimization/86650
4893         * tree-vrp.c (vrp_prop::check_array_ref): Print an inform message.
4894         (vrp_prop::check_mem_ref): Same.
4896 2018-08-13  Liu Hao <lh_mouse@126.com>
4898         * pretty-print.c (eat_esc_sequence): Swap the foreground and
4899         background colors if the COMMON_LVB_REVERSE_VIDEO flag is set,
4900         and clear it thereafter, as it only works for DBCS.
4902 2018-08-13  Liu Hao <lh_mouse@126.com>
4904         * pretty-print.c (mingw_ansi_fputs): Do not call _close() on the
4905         handle returned by _get_osf_handle().
4907 2018-08-13  Will Schmidt  <will_schmidt@vnet.ibm.com>
4909         * gcc/config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support
4910         for folding vec_perm.
4912 2018-08-13  Will Schmidt  <will_schmidt@vnet.ibm.com>
4914         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin):
4915         Add support for gimple-folding of vec_pack() and vec_unpack()
4916         intrinsics.
4918 2018-08-13  Will Schmidt <will_schmidt@vnet.ibm.com>
4920         * config/rs6000/rs6000.c (rs6000_builtin_valid_without_lhs): Add
4921         vec_xst variants to the list.
4922         (rs6000_gimple_fold_builtin): Add support for folding unaligned
4923         vector loads and stores.
4925 2018-08-13  David Edelsohn  <dje.gcc@gmail.com>
4927         * config.gcc (rs6000-ibm-aix4.x): Delete.
4928         (rs6000-ibm-aix5.1): Delete.
4929         (rs6000-ibm-aix5.2): Delete.
4930         (rs6000-ibm-aix5.3): Delete.
4931         * config/rs6000/aix43.h: Delete.
4932         * config/rs6000/aix51.h: Delete.
4933         * config/rs6000/aix52.h: Delete.
4934         * config/rs6000/t-aix43: Delete.
4936 2018-08-13  Ilya Leoshkevich  <iii@linux.ibm.com>
4938         * config/s390/s390.c (s390_decompose_constant_pool_ref):
4939         New function.
4940         (s390_decompose_address): Factor out constant pool ref
4941         decomposition.
4943 2018-08-12  Chung-Ju Wu  <jasonwucj@gmail.com>
4945         * config/nds32/nds32-predicates.c
4946         (nds32_can_use_bclr_p): Change return type as bool.
4947         (nds32_can_use_bset_p): Ditto.
4948         (nds32_can_use_btgl_p): Ditto.
4949         (nds32_can_use_bitci_p): Ditto.
4950         * config/nds32/nds32-protos.h
4951         (nds32_can_use_bclr_p): Change declaration.
4952         (nds32_can_use_bset_p): Ditto.
4953         (nds32_can_use_btgl_p): Ditto.
4954         (nds32_can_use_bitci_p): Ditto.
4956 2018-08-12  Chung-Ju Wu  <jasonwucj@gmail.com>
4958         * config/nds32/nds32.c (nds32_expand_prologue, nds32_expand_epilogue):
4959         Support -msched-prolog-epilog option.
4960         * config/nds32/nds32.opt (msched-prolog-epilog): New option.
4962 2018-08-12  Chung-Ju Wu  <jasonwucj@gmail.com>
4964         * common/config/nds32/nds32-common.c
4965         (nds32_option_optimization_table): Enalbe -malways-align.
4967 2018-08-12  Chung-Ju Wu  <jasonwucj@gmail.com>
4969         * config.gcc (nds32*): Add nds32_isr.h and nds32_init.inc in
4970         extra_headers.
4971         * common/config/nds32/nds32-common.c (nds32_handle_option): Handle
4972         OPT_misr_secure_ case.
4973         * config/nds32/nds32-isr.c: Implementation of backward compatibility.
4974         * config/nds32/nds32-protos.h (nds32_isr_function_critical_p): New.
4975         * config/nds32/nds32.c (nds32_attribute_table): Add critical and
4976         secure attribute.
4977         * config/nds32/nds32.h (nds32_isr_nested_type): Add NDS32_CRITICAL.
4978         (nds32_isr_info): New field security_level.
4979         (TARGET_ISR_VECTOR_SIZE_4_BYTE): New macro.
4980         * config/nds32/nds32.md (return_internal): Consider critical attribute.
4981         * config/nds32/nds32.opt (misr-secure): New option.
4982         * config/nds32/nds32_init.inc: New file.
4983         * config/nds32/nds32_isr.h: New file.
4985 2018-08-11  John David Anglin  <danglin@gcc.gnu.org>
4987         * config/pa/pa.md (UNSPEC_MEMORY_BARRIER): New unspec enum.
4988         Update comment for atomic instructions.
4989         (atomic_storeqi, atomic_storehi, atomic_storesi, atomic_storesf,
4990         atomic_loaddf, atomic_loaddf_1, atomic_storedf, atomic_storedf_1):
4991         Remove.
4992         (atomic_loaddi): Revise fence expansion to only emit fence prior to
4993         load for __ATOMIC_SEQ_CST model.
4994         (atomic_loaddi_1): Remove float register target.
4995         (atomic_storedi): Handle CONST_INT values.
4996         (atomic_storedi_1): Remove float register source.  Add special case
4997         for zero value.
4998         (memory_barrier): New expander and insn.
5000 2018-08-11  Jakub Jelinek  <jakub@redhat.com>
5002         PR tree-optimization/86835
5003         * tree-ssa-math-opts.c (insert_reciprocals): Even when inserting
5004         new_stmt after def_gsi, make sure to insert new_square_stmt after
5005         that stmt, not 2 stmts before it.
5007 2018-08-10  Alexander Monakov  <amonakov@ispras.ru>
5009         PR target/82418
5010         * config/i386/i386.md (<s>mul<mode>3_highpart): Use DWIH mode iterator
5011         instead of SWI48.
5013 2018-08-10  Martin Liska  <mliska@suse.cz>
5015         PR target/83610
5016         * builtin-types.def (BT_FN_LONG_LONG_LONG_DOUBLE): Add new
5017         function type.
5018         * builtins.c (expand_builtin_expect_with_probability):
5019         New function.
5020         (expand_builtin_expect_with_probability): New function.
5021         (build_builtin_expect_predicate): Add new argumnet probability
5022         for BUILT_IN_EXPECT_WITH_PROBABILITY.
5023         (fold_builtin_expect):
5024         (fold_builtin_2):
5025         (fold_builtin_3):
5026         * builtins.def (BUILT_IN_EXPECT_WITH_PROBABILITY):
5027         * builtins.h (fold_builtin_expect): Set new argument.
5028         * doc/extend.texi: Document __builtin_expect_with_probability.
5029         * doc/invoke.texi: Likewise.
5030         * gimple-fold.c (gimple_fold_call): Pass new argument.
5031         * ipa-fnsummary.c (find_foldable_builtin_expect): Handle
5032         also BUILT_IN_EXPECT_WITH_PROBABILITY.
5033         * predict.c (get_predictor_value): New function.
5034         (expr_expected_value): Add new argument probability. Assume
5035         that predictor and probability are always non-null.
5036         (expr_expected_value_1): Likewise.  For __builtin_expect and
5037         __builtin_expect_with_probability set probability.  Handle
5038         combination in binary expressions.
5039         (tree_predict_by_opcode): Simplify code by simply calling
5040         get_predictor_value.
5041         (pass_strip_predict_hints::execute): Add handling of
5042         BUILT_IN_EXPECT_WITH_PROBABILITY.
5043         * predict.def (PRED_BUILTIN_EXPECT_WITH_PROBABILITY): Add
5044         new predictor.
5045         * tree.h (DECL_BUILT_IN_P): New function.
5047 2018-08-10  Martin Liska  <mliska@suse.cz>
5049         PR tree-optimization/85799
5050         * passes.def: Add argument for pass_strip_predict_hints.
5051         * predict.c (class pass_strip_predict_hints): Add new argument
5052         early_p.
5053         (strip_predictor_early): New function.
5054         (pass_strip_predict_hints::execute): Call the function to
5055         strip predictors.
5056         (strip_predict_hints): New function.
5057         * predict.def: Fix comment.
5059 2018-08-10  Thomas Preud'homme  <thomas.preudhomme@linaro.org>
5061         * Makefile.in: Clarify which tm.texi to copy over to assert the
5062         right to grant a GFDL license for all.
5064 2018-08-09  Jeff Law  <law@redhat.com>
5066         * config/m68k/m68k.c (m68k_adjust_decorated_operand): Remove
5067         unused variable.
5069 2018-08-09  Andreas Schwab  <schwab@linux-m68k.org>
5071         * config/m68k/m68k-protos.h (m68k_final_prescan_insn): Remove
5072         prototype.
5074 2018-08-09  Richard Sandiford  <richard.sandiford@arm.com>
5076         * tree-vect-loop.c (vectorizable_reduction): Allow inner-loop
5077         reductions for variable-length vectors.
5079 2018-08-09  David Malcolm  <dmalcolm@redhat.com>
5081         PR other/84889
5082         * common.opt (fdiagnostics-show-line-numbers): New option.
5083         * diagnostic-show-locus.c (class layout): Add fields
5084         "m_show_line_numbers_p" and "m_linenum_width";
5085         (num_digits): New function.
5086         (test_num_digits): New function.
5087         (layout::layout): Initialize new fields.  Update m_x_offset
5088         logic to handle any left margin.
5089         (layout::print_source_line): Print line number when requested.
5090         (layout::start_annotation_line): New member function.
5091         (layout::print_annotation_line): Call it.
5092         (layout::print_leading_fixits): Likewise.
5093         (layout::print_trailing_fixits): Likewise.  Update calls to
5094         move_to_column for new parameter.
5095         (layout::get_x_bound_for_row): Add "add_left_margin" param and use
5096         it to potentially call start_annotation_line.
5097         (layout::show_ruler): Call start_annotation_line.
5098         (selftest::test_line_numbers_multiline_range): New selftest.
5099         (selftest::diagnostic_show_locus_c_tests): Call test_num_digits
5100         and selftest::test_line_numbers_multiline_range.
5101         * diagnostic.c (diagnostic_initialize): Initialize
5102         show_line_numbers_p.
5103         * diagnostic.h (struct diagnostic_context): Add field
5104         "show_line_numbers_p".
5105         * doc/invoke.texi (Diagnostic Message Formatting Options): Add
5106         -fno-diagnostics-show-line-numbers.
5107         * dwarf2out.c (gen_producer_string): Add
5108         OPT_fdiagnostics_show_line_numbers to the ignored options.
5109         * lto-wrapper.c (merge_and_complain): Likewise to the "pick
5110         one setting" options.
5111         (append_compiler_options): Likewise to the dropped options.
5112         (append_diag_options): Likewise to the passed-on options.
5113         * opts.c (common_handle_option): Handle the new option.
5114         * toplev.c (general_init): Set up global_dc->show_line_numbers_p.
5116 2018-08-09  Kelvin Nilsen  <kelvin@gcc.gnu.org>
5118         * doc/extend.texi (PowerPC AltiVec Built-in Functions Available on
5119         ISA 2.07): Correct spelling of bcdsub to be __builtin_bcdsub.  Add
5120         third argument of type "const signed char" to descriptions of
5121         __builtin_bcdadd, __builtin_bcdadd_lt, __builtin_bcdadd_eq,
5122         __builtin_bcdadd_gt, __builtin_bcdadd_ov, __builtin_bcdsub,
5123         __builtin_bcdsub_lt, __builtin_bcdsub_eq, __builtin_bcdsub_gt,
5124         __builtin_bcdsub_ov functions.
5126 2018-08-09  Richard Sandiford  <richard.sandiford@arm.com>
5128         PR tree-optimization/86858
5129         * tree-vect-loop.c (vect_is_simple_reduction): Restore
5130         flow_bb_inside_loop_p calls.
5132 2018-08-09  Richard Sandiford  <richard.sandiford@arm.com>
5134         PR tree-optimization/86871
5135         * tree-vect-stmts.c (vect_transform_stmt): Use gimple_get_lhs
5136         instead of gimple_assign_lhs.
5138 2018-08-09  Richard Earnshaw  <rearnsha@arm.com>
5140         PR target/86887
5141         * config/aarch64/aarch64.md (add<mode>3_carryinC_zero): Add missing
5142         register constraint to operand 0.
5143         (add<mode>3_carryinC): Likewise.
5144         (add<mode>3_carryinV_zero, add<mode>3_carryinV): Likewise.
5146 2018-08-09  Martin Liska  <mliska@suse.cz>
5148         PR c/86895
5149         * common.opt: Remove extra line.
5151 2018-08-09  Martin Liska  <mliska@suse.cz>
5153         * params.def (PARAM_ALIGN_LOOP_ITERATIONS): Remove double dots
5154         at the end of a line, make first letter capital and end up
5155         a sentence with a dot.
5156         (PARAM_LOOP_INTERCHANGE_STRIDE_RATIO): Likewise.
5157         (PARAM_LOOP_BLOCK_TILE_SIZE): Likewise.
5158         (PARAM_GRAPHITE_MAX_NB_SCOP_PARAMS): Likewise.
5159         (PARAM_GRAPHITE_MAX_ARRAYS_PER_SCOP): Likewise.
5160         (PARAM_MAX_ISL_OPERATIONS): Likewise.
5161         (PARAM_GRAPHITE_ALLOW_CODEGEN_ERRORS): Likewise.
5162         (PARAM_PROFILE_FUNC_INTERNAL_ID): Likewise.
5163         (PARAM_INDIR_CALL_TOPN_PROFILE): Likewise.
5164         (PARAM_SLP_MAX_INSNS_IN_BB): Likewise.
5165         (PARAM_IPA_CP_EVAL_THRESHOLD): Likewise.
5166         (PARAM_IPA_CP_RECURSION_PENALTY): Likewise.
5167         (PARAM_IPA_CP_SINGLE_CALL_PENALTY): Likewise.
5168         (PARAM_IPA_CP_LOOP_HINT_BONUS): Likewise.
5169         (PARAM_IPA_CP_ARRAY_INDEX_HINT_BONUS): Likewise.
5170         (PARAM_TREE_REASSOC_WIDTH): Likewise.
5171         (PARAM_HSA_GEN_DEBUG_STORES): Likewise.
5172         (PARAM_MAX_SPECULATIVE_DEVIRT_MAYDEFS): Likewise.
5173         (PARAM_MAX_VRP_SWITCH_ASSERTIONS): Likewise.
5175 2018-08-09  Andreas Krebbel  <krebbel@linux.ibm.com>
5177         PR target/84332
5178         * config/s390/s390.c (s390_option_override_internal): Reduce the
5179         stack-clash-protection-probe-interval param if it would be too big
5180         for z900.
5182 2018-08-08  Andreas Schwab  <schwab@linux-m68k.org>
5184         PR target/46179
5185         * config/m68k/m68k.h (FINAL_PRESCAN_INSN): Don't define.
5186         * config/m68k/m68k.c (handle_move_double): Don't call
5187         m68k_final_prescan_insn.
5188         (m68k_adjust_decorated_operand): Renamed from
5189         m68k_final_prescan_insn, remove first and third operand and
5190         simplify.
5191         (print_operand): Call it.
5192         (print_operand_address): Call it.
5194 2018-08-08  Nathan Sidwell  <nathan@acm.org>
5196         * diagnostic.c (diagnostic_report_current_module): Use
5197         linemap_included_from & linemap_included_from_linemap.
5199 2018-08-08  Hongbo Zhang  <hongbo.zhang@linaro.org>
5201         * config/aarch64/aarch64-cores.def: Add phecda core.
5202         * config/aarch64/aarch64-tune.md: Regenerate.
5203         * doc/invoke.texi: Add phecda core.
5205 2018-08-08  Andreas Krebbel  <krebbel@linux.ibm.com>
5207         PR target/85295
5208         * config/s390/constraints.md ("NxHD0", "NxSD0"): New constraint
5209         definitions.
5210         * config/s390/s390.md ("movti"): Add more alternatives for
5211         constant to GPR copies.
5213 2018-08-08  Andreas Krebbel  <krebbel@linux.ibm.com>
5215         * config/s390/s390.c: Fix whitespace damage throughout the file.
5216         * config/s390/s390.h: Likewise.
5217         * config/s390/tpf.h: Likewise.
5219 2018-08-08  Ilya Leoshkevich  <iii@linux.ibm.com>
5221         * config/s390/s390.c (s390_loadrelative_operand_p):
5222         Remove TARGET_CPU_ZARCH usages.
5223         (s390_rtx_costs): Likewise.
5224         (s390_legitimate_constant_p): Likewise.
5225         (s390_cannot_force_const_mem): Likewise.
5226         (legitimate_reload_constant_p): Likewise.
5227         (s390_preferred_reload_class): Likewise.
5228         (legitimize_pic_address): Likewise.
5229         (legitimize_tls_address): Likewise.
5230         (s390_split_branches): Removed.
5231         (s390_add_execute): Removed.
5232         (s390_dump_pool): Remove TARGET_CPU_ZARCH usages.
5233         (s390_mainpool_start): Likewise.
5234         (s390_mainpool_finish): Likewise.
5235         (s390_mainpool_cancel): Removed.
5236         (s390_chunkify_start): Remove TARGET_CPU_ZARCH usages.
5237         (s390_chunkify_cancel): Likewise.
5238         (s390_return_addr_rtx): Likewise.
5239         (s390_register_info): Remove split_branches_pending_p uages.
5240         (s390_optimize_register_info): Likewise.
5241         (s390_init_frame_layout): Remove TARGET_CPU_ZARCH and
5242         split_branches_pending_p usages.
5243         (s390_can_eliminate): Remove TARGET_CPU_ZARCH usages.
5244         (s390_load_got): Likewise.
5245         (s390_expand_split_stack_prologue): Likewise.
5246         (output_asm_nops): Likewise.
5247         (s390_function_profiler): Likewise.
5248         (s390_emit_call): Likewise.
5249         (s390_conditional_register_usage): Likewise.
5250         (s390_optimize_prologue): Likewise.
5251         (s390_reorg): Remove TARGET_CPU_ZARCH and
5252         split_branches_pending_p usages.
5253         (s390_option_override_internal): Remove TARGET_CPU_ZARCH
5254         usages.
5255         (s390_output_indirect_thunk_function): Likewise.
5256         * config/s390/s390.h (TARGET_CPU_ZARCH): Removed.
5257         (TARGET_CPU_ZARCH_P): Removed.
5258         (struct machine_function): Remove split_branches_pending_p.
5259         * config/s390/s390.md: Remove TARGET_CPU_ZARCH usages.
5261 2018-08-08  Ilya Leoshkevich  <iii@linux.ibm.com>
5263         * common/config/s390/s390-common.c (processor_flags_table):
5264         Remove flags.
5265         * config.gcc: Remove with_arch/with_tune support.
5266         * config/s390/2064.md: Remove cpu attribute comparisons.
5267         * config/s390/driver-native.c (s390_host_detect_local_cpu):
5268         Remove MTN.
5269         * config/s390/linux.h (ASM_SPEC):
5270         Remove -march support.
5271         * config/s390/s390-c.c (s390_cpu_cpp_builtins_internal):
5272         Use a table to get an arch level.
5273         * config/s390/s390-opts.h (enum processor_type):
5274         Remove enum values.
5275         * config/s390/s390.c
5276         (processor_table): Remove entries, add arch_level values.
5277         (s390_issue_rate): Remove cases.
5278         (s390_option_override): Adjust
5279         s390_option_override_internal() call.
5280         (s390_option_override_internal): Remove deprecation warning.
5281         (s390_valid_target_attribute_tree): Adjust
5282         s390_option_override_internal() call.
5283         * config/s390/s390.h (struct s390_processor):
5284         Share with s390-c.c, add arch_level field.
5285         * config/s390/s390.md:
5286         Remove occurrences in cpu attribute.
5287         * config/s390/s390.opt: Remove -march/-mtune support.
5288         * config/s390/tpf.h (ASM_SPEC): Remove -march support.
5289         * doc/invoke.texi: Remove deprecation warning.
5291 2018-08-08  Luis Machado  <luis.machado@linaro.org>
5293         * config/aarch64/aarch64.c (qdf24xx_vector_cost): New static
5294         global.
5295         (qdf24xx_tunings): Set vector cost structure to
5296         qdf24xx_vector_cost.
5298         * config/aarch64/aarch64.c (qdf24xx_addrcost_table)
5299         <register_sextend>: Set to 3.
5301 2018-08-07  Richard Sandiford  <richard.sandiford@arm.com>
5303         PR target/86838
5304         * config/aarch64/iterators.md (FRECP, frecp_suffix): Delete.
5305         * config/aarch64/aarch64-simd.md
5306         (aarch64_frecp<FRECP:frecp_suffix><mode>): Fold FRECPE into...
5307         (@aarch64_frecpe<mode>): ...here and the move FRECPX to...
5308         (aarch64_frecpx<mode>): ...this new pattern.
5309         * config/aarch64/aarch64-simd-builtins.def: Remove comment
5310         about aarch64_frecp<FRECP:frecp_suffix><mode>.
5312 2018-08-07  Martin Liska  <mliska@suse.cz>
5314         PR middle-end/83023
5315         * predict.c (expr_expected_value_1): Handle DECL_IS_MALLOC,
5316         BUILT_IN_REALLOC and DECL_IS_OPERATOR_NEW.
5317         * predict.def (PRED_MALLOC_NONNULL): New predictor.
5318         * doc/extend.texi: Document that malloc attribute adds
5319         hit to compiler.
5321 2018-08-06  John David Anglin  <danglin@gcc.gnu.org>
5323         PR target/86785
5324         * config/pa/pa.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
5325         Define to speculation_safe_value_not_needed.
5327 2018-08-06  Jeff Law  <law@redhat.com>
5329         * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Pass down
5330         the vr_values instance to cprop_into_stmt.
5331         (cprop_into_stmt): Pass vr_values instance down to cprop_operand.
5332         (cprop_operand): Also query EVRP to determine if OP is a constant.
5334 2018-08-06  Nathan Sidwell  <nathan@acm.org>
5336         * diagnostic.c (diagnostic_report_current_module): Reroll
5337         included-at loop.  Translate text.
5339 2018-08-06  David Malcolm  <dmalcolm@redhat.com>
5341         * function-tests.c (selftest::test_expansion_to_rtl): Call
5342         free_after_compilation.
5344 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
5346         * config/aarch64/aarch64.md: Add clobber highs to tls_desc.
5348 2018-08-06  Andreas Krebbel  <krebbel@linux.ibm.com>
5350         * config/s390/s390.c (s390_loop_unroll_adjust): Prevent small
5351         loops with memory block operations from getting unrolled.
5353 2018-08-06  Ulrich Weigand  <uweigand@de.ibm.com>
5355         PR target/86807
5356         * config/spu/spu.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
5357         Define to speculation_safe_value_not_needed.
5359 2018-08-06  Jeff Law  <law@redhat.com>
5361         * reload1.c (forget_old_reloads_1): Adjust CLOBBER_HIGH
5362         assert.
5364 2018-08-06  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
5366         PR target/86662
5367         * gcc/tree.c (build_common_tree_nodes): Initialize integer_types array
5368         with all enabled __intN types.
5370         * gcc/testsuite/gcc.target/msp430/pr86662.c: New test.
5372 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
5374         * alias.c (record_set): Check for clobber high.
5375         * cfgexpand.c (expand_gimple_stmt): Likewise.
5376         * combine-stack-adj.c (single_set_for_csa): Likewise.
5377         * combine.c (find_single_use_1): Likewise.
5378         (set_nonzero_bits_and_sign_copies): Likewise.
5379         (get_combine_src_dest): Likewise.
5380         (is_parallel_of_n_reg_sets): Likewise.
5381         (try_combine): Likewise.
5382         (record_dead_and_set_regs_1): Likewise.
5383         (reg_dead_at_p_1): Likewise.
5384         (reg_dead_at_p): Likewise.
5385         * dce.c (deletable_insn_p): Likewise.
5386         (mark_nonreg_stores_1): Likewise.
5387         (mark_nonreg_stores_2): Likewise.
5388         * df-scan.c (df_find_hard_reg_defs): Likewise.
5389         (df_uses_record): Likewise.
5390         (df_get_call_refs): Likewise.
5391         * dwarf2out.c (mem_loc_descriptor): Likewise.
5392         * haifa-sched.c (haifa_classify_rtx): Likewise.
5393         * ira-build.c (create_insn_allocnos): Likewise.
5394         * ira-costs.c (scan_one_insn): Likewise.
5395         * ira.c (equiv_init_movable_p): Likewise.
5396         (rtx_moveable_p): Likewise.
5397         (interesting_dest_for_shprep): Likewise.
5398         * jump.c (mark_jump_label_1): Likewise.
5399         * postreload-gcse.c (record_opr_changes): Likewise.
5400         * postreload.c (reload_cse_simplify): Likewise.
5401         (struct reg_use): Add source expr.
5402         (reload_combine): Check for clobber high.
5403         (reload_combine_note_use): Likewise.
5404         (reload_cse_move2add): Likewise.
5405         (move2add_note_store): Likewise.
5406         * print-rtl.c (print_pattern): Likewise.
5407         * recog.c (decode_asm_operands): Likewise.
5408         (store_data_bypass_p): Likewise.
5409         (if_test_bypass_p): Likewise.
5410         * regcprop.c (kill_clobbered_value): Likewise.
5411         (kill_set_value): Likewise.
5412         * reginfo.c (reg_scan_mark_refs): Likewise.
5413         * reload1.c (maybe_fix_stack_asms): Likewise.
5414         (eliminate_regs_1): Likewise.
5415         (elimination_effects): Likewise.
5416         (mark_not_eliminable): Likewise.
5417         (scan_paradoxical_subregs): Likewise.
5418         (forget_old_reloads_1): Likewise.
5419         * reorg.c (find_end_label): Likewise.
5420         (try_merge_delay_insns): Likewise.
5421         (redundant_insn): Likewise.
5422         (own_thread_p): Likewise.
5423         (fill_simple_delay_slots): Likewise.
5424         (fill_slots_from_thread): Likewise.
5425         (dbr_schedule): Likewise.
5426         * resource.c (update_live_status): Likewise.
5427         (mark_referenced_resources): Likewise.
5428         (mark_set_resources): Likewise.
5429         * rtl.c (copy_rtx): Likewise.
5430         * rtlanal.c (reg_referenced_p): Likewise.
5431         (single_set_2): Likewise.
5432         (noop_move_p): Likewise.
5433         (note_stores): Likewise.
5434         * sched-deps.c (sched_analyze_reg): Likewise.
5435         (sched_analyze_insn): Likewise.
5437 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
5439         * cse.c (invalidate_reg): New function extracted from...
5440         (invalidate): ...here.
5441         (canonicalize_insn): Check for clobber high.
5442         (invalidate_from_clobbers): invalidate clobber highs.
5443         (invalidate_from_sets_and_clobbers): Likewise.
5444         (count_reg_usage): Check for clobber high.
5445         (insn_live_p): Likewise.
5446         * cselib.c (cselib_expand_value_rtx_1):Likewise.
5447         (cselib_invalidate_regno): Check for clobber in setter.
5448         (cselib_invalidate_rtx): Pass through setter.
5449         (cselib_invalidate_rtx_note_stores):
5450         (cselib_process_insn): Check for clobber high.
5451         * cselib.h (cselib_invalidate_rtx): Add operand.
5453 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
5455         * lra-eliminations.c (lra_eliminate_regs_1): Check for clobber high.
5456         (mark_not_eliminable): Likewise.
5457         * lra-int.h (struct lra_insn_reg): Add clobber high marker.
5458         * lra-lives.c (process_bb_lives): Check for clobber high.
5459         * lra.c (new_insn_reg): Remember clobber highs.
5460         (collect_non_operand_hard_regs): Check for clobber high.
5461         (lra_set_insn_recog_data): Likewise.
5462         (add_regs_to_insn_regno_info): Likewise.
5463         (lra_update_insn_regno_info): Likewise.
5465 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
5467         * rtl.h (reg_is_clobbered_by_clobber_high): Add declarations.
5468         * rtlanal.c (reg_is_clobbered_by_clobber_high): Add function.
5470 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
5472         * emit-rtl.c (verify_rtx_sharing): Check for CLOBBER_HIGH.
5473         (copy_insn_1): Likewise.
5474         (gen_hard_reg_clobber_high): New gen function.
5475         * genconfig.c (walk_insn_part): Check for CLOBBER_HIGH.
5476         * genemit.c (gen_exp): Likewise.
5477         (gen_emit_seq): Pass through info.
5478         (gen_insn): Check for CLOBBER_HIGH.
5479         (gen_expand): Pass through info.
5480         (gen_split): Likewise.
5481         (output_add_clobbers): Likewise.
5482         * genrecog.c (validate_pattern): Check for CLOBBER_HIGH.
5483         (remove_clobbers): Likewise.
5484         * rtl.h (gen_hard_reg_clobber_high): New declaration.
5486 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
5488         * doc/rtl.texi (clobber_high): Add.
5489         (parallel): Add in clobber high
5490         * rtl.c (rtl_check_failed_code3): Add function.
5491         * rtl.def (CLOBBER_HIGH): Add expression.
5492         * rtl.h (RTL_CHECKC3): Add macro.
5493         (rtl_check_failed_code3): Add declaration.
5494         (XC3EXP): Add macro.
5496 2018-08-05  H.J. Lu  <hongjiu.lu@intel.com>
5498         PR target/86386
5499         * config/i386/i386.c (ix86_finalize_stack_frame_flags): Set
5500         cfun->machine->max_used_stack_alignment if needed.
5502 2018-08-04  Martin Sebor  <msebor@redhat.com>
5504         PR tree-optimization/86571
5505         * gimple-ssa-sprintf.c (format_floating): Extend upper bound of
5506         NaN output to 4.
5508 2018-08-03  Sandra Loosemore  <sandra@codesourcery.com>
5510         PR target/86799
5511         * config/nios2/nios2.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
5512         Define.
5514 2018-08-03  Jeff Law  <law@redhat.com>
5516         PR target/86795
5517         * config/mn10300/mn10300.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
5518         Define to speculation_safe_value_not_needed.
5520 2018-08-03  David Malcolm  <dmalcolm@redhat.com>
5522         * doc/gcov.texi (-x): Remove duplicate "to".
5523         * doc/invoke.texi (-Wnoexcept-type): Remove duplicate "calls".
5524         (-Wif-not-aligned): Remove duplicate "is".
5525         (-flto): Remove duplicate "the".
5526         (MicroBlaze Options): In examples of "-mcpu=cpu-type", remove
5527         duplicate "v5.00.b".
5528         (MSP430 Options): Remove duplicate "and" from the description
5529         of "-mgprel-sec=regexp".
5530         (x86 Options): Remove duplicate copies of "vmldLog102" and
5531         vmlsLog104 from description of "-mveclibabi=type".
5533 2018-08-03  Richard Sandiford  <richard.sandiford@arm.com>
5535         * internal-fn.h (first_commutative_argument): Declare.
5536         * internal-fn.c (first_commutative_argument): New function.
5537         * tree-vect-slp.c (vect_get_and_check_slp_defs): Remove extra
5538         restrictions for pattern statements.  Use first_commutative_argument
5539         to look for commutative operands in calls to internal functions.
5541 2018-08-03  Aldy Hernandez  <aldyh@redhat.com>
5543         * Makefile.in (wide-int-range.o): New.
5544         * tree-vrp.c: Move all the wide_int_* functions to...
5545         * wide-int-range.cc: ...here.
5546         * tree-vrp.h: Move all the wide_int_* prototypes to...
5547         * wide-int-range.h: ...here.
5549 2018-08-03  Tom de Vries  <tdevries@suse.de>
5551         * common/config/nvptx/nvptx-common.c (nvptx_except_unwind_info): Return
5552         UI_NONE.
5553         * config/nvptx/nvptx.c (TARGET_ASM_BYTE_OP): Remove define.
5554         * except.c (output_function_exception_table): Do early exit if
5555         targetm_common.except_unwind_info (&global_options) == UI_NONE.
5557 2018-08-03  Martin Liska  <mliska@suse.cz>
5559         * predict.c (dump_prediction): Change to 2 digits
5560         in fraction part.
5562 2018-08-03  Siddhesh Poyarekar  <siddhesh@sourceware.org>
5564         * config/aarch64/falkor.md (falkor_am_1_vxvy_vxvy): Move
5565         neon_dup_q to...
5566         (falkor_am_1_gtov_gtov): ... a new insn reservation.
5568 2018-07-19  Ilya Leoshkevich  <iii@linux.ibm.com>
5570         * config/nds32/nds32.c (nds32_hard_regno_mode_ok): Replace > with >=.
5571         * df-problems.c (df_remove_dead_eq_notes): Replace > with >=.
5572         * dwarf2out.c (mem_loc_descriptor): Replace > with >=.
5573         * lra-constraints.c (spill_hard_reg_in_range): Replace <= with <.
5574         * lra-remat.c (call_used_input_regno_present_p): Replace <= with <.
5576 2018-08-02  David Malcolm  <dmalcolm@redhat.com>
5578         * diagnostic-show-locus.c (diagnostic_show_locus): Use
5579         pp_take_prefix when saving the existing prefix.
5580         * diagnostic.c (diagnostic_append_note): Likewise.
5581         * langhooks.c (lhd_print_error_function): Likewise.
5582         * pretty-print.c (pp_set_prefix): Drop the "const" from "prefix"
5583         param's type.  Free the existing prefix.
5584         (pp_take_prefix): New function.
5585         (pretty_printer::pretty_printer): Drop the prefix parameter.
5586         Rename the length parameter to match the comment.
5587         (pretty_printer::~pretty_printer): Free the prefix.
5588         * pretty-print.h (pretty_printer::pretty_printer): Drop the prefix
5589         parameter.
5590         (struct pretty_printer): Drop the "const" from "prefix" field's
5591         type and clarify memory management.
5592         (pp_set_prefix): Drop the "const" from the 2nd param.
5593         (pp_take_prefix): New decl.
5595 2018-08-02  Aaron Sawdey  <acsawdey@linux.ibm.com>
5597         * config/rs6000/rs6000-string.c (select_block_compare_mode): Move test
5598         for word_mode_ok here instead of passing as argument.
5599         (expand_block_compare): Change select_block_compare_mode() call.
5600         (expand_strncmp_gpr_sequence): New function.
5601         (expand_strn_compare): Make use of expand_strncmp_gpr_sequence.
5603 2018-08-02  Jeff Law  <law@redhat.com>
5605         PR target/86790
5606         * config/m68k/m68k.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
5607         Define to speculation_safe_value_not_needed.
5609         PR target/86784
5610         * config/h8300/h8300.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
5611         Define to speculation_safe_value_not_needed.
5613 2018-08-02  Tom de Vries  <tdevries@suse.de>
5615         PR target/86660
5616         * common/config/nvptx/nvptx-common.c (nvptx_except_unwind_info): New
5617         function.  Return UI_TARGET unconditionally.
5618         (TARGET_EXCEPT_UNWIND_INFO): Redefine to nvptx_except_unwind_info.
5619         * config/nvptx/nvptx.c (TARGET_ASM_BYTE_OP): Emit commented out '.byte'.
5621 2018-08-02  Richard Sandiford  <richard.sandiford@arm.com>
5623         * genemit.c (print_overload_test): Fix typo.
5625 2018-08-02  Richard Biener  <rguenther@suse.de>
5627         PR tree-optimization/86816
5628         * tree-ssa-tail-merge.c (tail_merge_valueize): New function
5629         which checks for value availability before querying it.
5630         (gvn_uses_equal): Use it.
5631         (same_succ_hash): Likewise.
5632         (gimple_equal_p): Likewise.
5634 2018-08-02  Nick Clifton  <nickc@redhat.com>
5636         PR target/86813
5637         * config/stormy16/stormy16.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
5638         Define to speculation_safe_value_not_needed.
5640         PR target/86810
5641         * config/v850/v850.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
5642         Define to speculation_safe_value_not_needed.
5644         PR target/86810
5645         * config/v850/v850.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
5646         Define to speculation_safe_value_not_needed.
5648         PR target/86803
5649         * config/rx/rx.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
5650         Define to speculation_safe_value_not_needed.
5652         PR target/86797
5653         * config/msp430/msp430.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
5654         Define to speculation_safe_value_not_needed.
5656         PR target/86791
5657         * config/mcore/mcore.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
5658         Define to speculation_safe_value_not_needed.
5660         PR target/86789
5661         * config/m32r/m32r.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
5662         Define to speculation_safe_value_not_needed.
5664         PR target/86787
5665         * config/iq2000/iq2000.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
5666         Define to speculation_safe_value_not_needed.
5668         PR target/86782
5669         * config/frv/frv.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define to
5670         speculation_safe_value_not_needed.
5672         PR target/86781
5673         * config/fr30/fr30.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define
5674         to speculation_safe_value_not_needed.
5676 2018-08-02  Richard Sandiford  <richard.sandiford@arm.com>
5678         * doc/md.texi: Expand the documentation of instruction names
5679         to mention port-local uses.  Document '@' in pattern names.
5680         * read-md.h (overloaded_instance, overloaded_name): New structs.
5681         (mapping): Declare.
5682         (md_reader::handle_overloaded_name): New member function.
5683         (md_reader::get_overloads): Likewise.
5684         (md_reader::m_first_overload): New member variable.
5685         (md_reader::m_next_overload_ptr): Likewise.
5686         (md_reader::m_overloads_htab): Likewise.
5687         * read-md.c (md_reader::md_reader): Initialize m_first_overload,
5688         m_next_overload_ptr and m_overloads_htab.
5689         * read-rtl.c (iterator_group): Add "type" and "get_c_token" fields.
5690         (get_mode_token, get_code_token, get_int_token): New functions.
5691         (map_attr_string): Add an optional argument that passes back
5692         the associated iterator.
5693         (overloaded_name_hash, overloaded_name_eq_p, named_rtx_p):
5694         (md_reader::handle_overloaded_name, add_overload_instance): New
5695         functions.
5696         (apply_iterators): Handle '@' names.  Report an error if '@'
5697         is used without iterators.
5698         (initialize_iterators): Initialize the new iterator_group fields.
5699         * genopinit.c (handle_overloaded_code_for)
5700         (handle_overloaded_gen): New functions.
5701         (main): Use them to print declarations of maybe_code_for_* and
5702         maybe_gen_* functions, and inline definitions of code_for_* and gen_*.
5703         * genemit.c (print_overload_arguments, print_overload_test)
5704         (handle_overloaded_code_for, handle_overloaded_gen): New functions.
5705         (main): Use it to print definitions of maybe_code_for_* and
5706         maybe_gen_* functions.
5707         * config/aarch64/aarch64.c (aarch64_split_128bit_move): Use
5708         gen_aarch64_mov{low,high}_di and gen_aarch64_movdi_{low,high}
5709         instead of explicit mode checks.
5710         (aarch64_split_simd_combine): Likewise gen_aarch64_simd_combine.
5711         (aarch64_split_simd_move): Likewise gen_aarch64_split_simd_mov.
5712         (aarch64_emit_load_exclusive): Likewise gen_aarch64_load_exclusive.
5713         (aarch64_emit_store_exclusive): Likewise gen_aarch64_store_exclusive.
5714         (aarch64_expand_compare_and_swap): Likewise
5715         gen_aarch64_compare_and_swap and gen_aarch64_compare_and_swap_lse
5716         (aarch64_gen_atomic_cas): Likewise gen_aarch64_atomic_cas.
5717         (aarch64_emit_atomic_swap): Likewise gen_aarch64_atomic_swp.
5718         (aarch64_constant_pool_reload_icode): Delete.
5719         (aarch64_secondary_reload): Use code_for_aarch64_reload_movcp
5720         instead of aarch64_constant_pool_reload_icode.  Use
5721         code_for_aarch64_reload_mov instead of explicit mode checks.
5722         (rsqrte_type, get_rsqrte_type, rsqrts_type, get_rsqrts_type): Delete.
5723         (aarch64_emit_approx_sqrt): Use gen_aarch64_rsqrte instead of
5724         get_rsqrte_type and gen_aarch64_rsqrts instead of gen_rqrts_type.
5725         (recpe_type, get_recpe_type, recps_type, get_recps_type): Delete.
5726         (aarch64_emit_approx_div): Use gen_aarch64_frecpe instead of
5727         get_recpe_type and gen_aarch64_frecps instead of get_recps_type.
5728         (aarch64_atomic_load_op_code): Delete.
5729         (aarch64_emit_atomic_load_op): Likewise.
5730         (aarch64_gen_atomic_ldop): Use UNSPECV_ATOMIC_* instead of
5731         aarch64_atomic_load_op_code.  Use gen_aarch64_atomic_load
5732         instead of aarch64_emit_atomic_load_op.
5733         * config/aarch64/aarch64.md (aarch64_reload_movcp<GPF_TF:mode><P:mode>)
5734         (aarch64_reload_movcp<VALL:mode><P:mode>, aarch64_reload_mov<mode>)
5735         (aarch64_movdi_<mode>low, aarch64_movdi_<mode>high)
5736         (aarch64_mov<mode>high_di, aarch64_mov<mode>low_di): Add a '@'
5737         character before the pattern name.
5738         * config/aarch64/aarch64-simd.md (aarch64_split_simd_mov<mode>)
5739         (aarch64_rsqrte<mode>, aarch64_rsqrts<mode>)
5740         (aarch64_simd_combine<mode>, aarch64_frecpe<mode>)
5741         (aarch64_frecps<mode>): Likewise.
5742         * config/aarch64/atomics.md (atomic_compare_and_swap<mode>)
5743         (aarch64_compare_and_swap<mode>, aarch64_compare_and_swap<mode>_lse)
5744         (aarch64_load_exclusive<mode>, aarch64_store_exclusive<mode>)
5745         (aarch64_atomic_swp<mode>, aarch64_atomic_cas<mode>)
5746         (aarch64_atomic_load<atomic_ldop><mode>): Likewise.
5748 2018-08-02  Richard Sandiford  <richard.sandiford@arm.com>
5750         * config/aarch64/aarch64.c (aarch64_float_const_representable_p):
5751         Allow HFmode constants if TARGET_FP_F16INST.
5753 2018-08-02  Jackson Woodruff  <jackson.woodruff@arm.com>
5755         PR target/86014
5756         * config/aarch64/aarch64.c (aarch64_operands_adjust_ok_for_ldpstp):
5757         No longer check last store for clobber of address register.
5759 2018-08-02  Martin Liska  <mliska@suse.cz>
5761         PR gcov-profile/86817
5762         * gcov.c (process_all_functions): New function.
5763         (main): Call it.
5764         (process_file): Move functions processing to
5765         process_all_functions.
5767 2018-08-02  David Malcolm  <dmalcolm@redhat.com>
5769         * dumpfile.c (dump_user_location_t::dump_user_location_t): Add
5770         "const" to the "gimple *" and "rtx_insn *" parameters.
5771         * dumpfile.h (dump_user_location_t::dump_user_location_t):
5772         Likewise.
5773         (dump_location_t::dump_location_t): Likewise.
5775 2018-08-01  Martin Sebor  <msebor@redhat.com>
5777         PR tree-optimization/86650
5778         * gimple-pretty-print.c (percent_G_format): Accept a "gimple *"
5779         rather than a "gcall *".  Directly pass the data of interest
5780         to percent_K_format, rather than building a temporary CALL_EXPR
5781         to hold it.
5782         * gimple-fold.c (gimple_fold_builtin_strncpy): Adjust.
5783         (gimple_fold_builtin_strncat): Adjust.
5784         * gimple-ssa-warn-restrict.h (check_bounds_or_overlap): Replace
5785         gcall* argument with gimple*.
5786         * gimple-ssa-warn-restrict.c (check_call): Same.
5787         (wrestrict_dom_walker::before_dom_children): Same.
5788         (builtin_access::builtin_access): Same.
5789         (check_bounds_or_overlap): Same
5790         (maybe_diag_overlap): Same.
5791         (maybe_diag_offset_bounds): Same.
5792         * tree-diagnostic.c (default_tree_printer): Move usage of
5793         EXPR_LOCATION (t) and TREE_BLOCK (t) from within percent_K_format
5794         to this callsite.
5795         * tree-pretty-print.c (percent_K_format): Add argument.
5796         * tree-pretty-print.h: Add argument.
5797         * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Adjust.
5798         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Adjust.
5799         (maybe_diag_stxncpy_trunc): Same.
5800         (handle_builtin_stxncpy): Same.
5801         (handle_builtin_strcat): Same.
5803 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
5805         * match.pd: Optimise pointer range checks.
5807 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
5809         PR tree-optimization/86758
5810         * tree-vect-stmts.c (vectorizable_simd_clone_call): Don't try
5811         to remove pattern statements.
5813 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
5815         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Use the
5816         result of dfs_enumerate_from when constructing stmt_vec_infos,
5817         instead of additionally calling get_loop_body.
5819 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
5821         * tree-vectorizer.h (vect_create_data_ref_ptr): Remove inv_p
5822         parameter.
5823         * tree-vect-data-refs.c (vect_create_data_ref_ptr): Likewise.
5824         When creating an iv, assert that the step is not known to be zero.
5825         (vect_setup_realignment): Update call accordingly.
5826         * tree-vect-stmts.c (vectorizable_store): Likewise.
5827         (vectorizable_load): Likewise.  Handle VMAT_INVARIANT separately.
5829 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
5831         * tree-vectorizer.h (vect_stmt_to_vectorize): New function.
5832         * tree-vect-loop.c (vect_update_vf_for_slp): Use it.
5833         (vectorizable_reduction): Likewise.
5834         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
5835         (vect_detect_hybrid_slp_stmts): Likewise.
5836         * tree-vect-stmts.c (vect_is_simple_use): Likewise.
5838 2018-08-01  Aldy Hernandez  <aldyh@redhat.com>
5840         * tree-vrp (zero_nonzero_bits_from_bounds): Rename to...
5841         (wide_int_set_zero_nonzero_bits): ...this.
5842         (zero_nonzero_bits_from_vr): Rename to...
5843         (vrp_set_zero_nonzero_bits): ...this.
5844         (extract_range_from_multiplicative_op_1): Abstract wide int
5845         code...
5846         (wide_int_range_multiplicative_op): ...here.
5847         (extract_range_from_binary_expr_1): Extract wide int binary
5848         operations into their own functions.
5849         (wide_int_range_lshift): New.
5850         (wide_int_range_can_optimize_bit_op): New.
5851         (wide_int_range_shift_undefined_p): New.
5852         (wide_int_range_bit_xor): New.
5853         (wide_int_range_bit_ior): New.
5854         (wide_int_range_bit_and): New.
5855         (wide_int_range_trunc_mod): New.
5856         (extract_range_into_wide_ints): New.
5857         (vrp_shift_undefined_p): New.
5858         (extract_range_from_multiplicative_op): New.
5859         (vrp_can_optimize_bit_op): New.
5860         * tree-vrp.h (value_range::dump): New.
5861         (wide_int_range_multiplicative_op): New.
5862         (wide_int_range_lshift):New.
5863         (wide_int_range_shift_undefined_p): New.
5864         (wide_int_range_bit_xor): New.
5865         (wide_int_range_bit_ior): New.
5866         (wide_int_range_bit_and): New.
5867         (wide_int_range_trunc_mod): New.
5868         (zero_nonzero_bits_from_bounds): Rename to...
5869         (wide_int_set_zero_nonzero_bits): ...this.
5870         (zero_nonzero_bits_from_vr): Rename to...
5871         (vrp_set_zero_nonzero_bits): ...this.
5872         (range_easy_mask_min_max): Rename to...
5873         (wide_int_range_can_optimize_bit_op): this.
5874         * vr-values.c (simplify_bit_ops_using_ranges): Rename
5875         zero_nonzero_bits_from_vr into vrp_set_zero_nonzero_bits.
5877 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
5879         * tree-vectorizer.h (vect_orig_stmt): New function.
5880         * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Use it.
5881         * tree-vect-loop.c (vect_model_reduction_cost): Likewise.
5882         (vect_create_epilog_for_reduction): Likewise.
5883         (vectorizable_live_operation): Likewise.
5884         * tree-vect-slp.c (vect_find_last_scalar_stmt_in_slp): Likewise.
5885         (vect_detect_hybrid_slp_stmts, vect_schedule_slp): Likewise.
5886         * tree-vect-stmts.c (vectorizable_call): Likewise.
5887         (vectorizable_simd_clone_call, vect_remove_stores): Likewise.
5889 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
5891         * tree-vectorizer.h (vect_transform_stmt): Remove grouped_store
5892         argument.
5893         * tree-vect-stmts.c (vect_transform_stmt): Likewise.
5894         * tree-vect-loop.c (vect_transform_loop_stmt): Update call accordingly.
5895         (vect_transform_loop): Likewise.
5896         * tree-vect-slp.c (vect_schedule_slp_instance): Likewise.
5898 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
5900         * tree-vectorizer.h (vect_schedule_slp): Return void.
5901         * tree-vect-slp.c (vect_schedule_slp_instance): Likewise.
5902         (vect_schedule_slp): Likewise.
5904 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
5906         * tree-vect-loop.c (vect_transform_loop_stmt): Remove slp_scheduled
5907         argument.
5908         (vect_transform_loop): Update calls accordingly.  Schedule SLP
5909         instances before the main loop, if any exist.
5911 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
5913         PR tree-optimization/86749
5914         * tree-vect-patterns.c (vect_determine_min_output_precision_1):
5915         If the lhs is used in a COND_EXPR, check that it is being used
5916         as the "then" or "else" value.
5918 2018-08-01  Tom de Vries  <tdevries@suse.de>
5920         PR target/86800
5921         * config/nvptx/nvptx.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define to
5922         speculation_safe_value_not_needed.
5924 2018-08-01  Richard Biener  <rguenther@suse.de>
5926         * tree-ssa-sccvn.c (visit_phi): Compare invariant addresses
5927         as base and offset.
5929 2018-08-01  Martin Liska  <mliska@suse.cz>
5931         * value-prof.c (gimple_divmod_fixed_value_transform): Unify
5932         format how successful transformation is dumped.
5933         (gimple_mod_pow2_value_transform): Likewise.
5934         (gimple_mod_subtract_transform): Likewise.
5935         (gimple_stringops_transform): Likewise.
5937 2018-08-01  Martin Liska  <mliska@suse.cz>
5939         PR value-prof/35543
5940         * value-prof.c (interesting_stringop_to_profile_p):
5941         Simplify the code and add BUILT_IN_MEMMOVE.
5942         (gimple_stringops_transform): Likewise.
5944 2018-08-01  Sam Tebbs  <sam.tebbs@arm.com>
5946         * config/aarch64/aarch64-simd.md
5947         (*aarch64_get_lane_zero_extendsi<mode>): Rename to...
5948         (*aarch64_get_lane_zero_extend<GPI:mode><VDQQH:mode>): ... This and
5949         use GPI iterator instead of SI mode.
5951 2018-08-01  Richard Earnshaw  <rearnsha@arm.com>
5953         * config/rs6000/rs6000.md (speculation_barrier): Renamed from
5954         rs6000_speculation_barrier.
5955         * config/rs6000/rs6000.c (rs6000_expand_builtin): Adjust for
5956         new barrier pattern name.
5958 2018-08-01  Richard Earnshaw  <rearnsha@arm.com>
5960         * config/i386/i386.md (unspecv): Add UNSPECV_SPECULATION_BARRIER.
5961         (speculation_barrier): New insn.
5963 2018-08-01  Richard Biener  <rguenther@suse.de>
5965         PR bootstrap/86724
5966         * graphite.h: Include isl/id.h and isl/space.h to allow build
5967         with ISL 0.20.
5969 2018-08-01  Jan Willem Jagersma  <jwjagersma@gmail.com>
5971         PR target/86651
5972         * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
5973         mode for COFF targets.
5974         * defaults.h (TARGET_COFF): Define.
5975         * config/i386/djgpp.h (TARGET_ASM_LTO_START, TARGET_ASM_LTO_END,
5976         TARGET_COFF): Define.
5977         (i386_djgpp_asm_lto_start, i386_djgpp_asm_lto_end): Declare.
5978         * config/i386/djgpp.c (saved_debug_info_level): New static variable.
5979         (i386_djgpp_asm_lto_start, i386_djgpp_asm_lto_end): New functions.
5981 2018-07-31  Alexandre Oliva <oliva@adacore.com>
5983         * gimple-streamer-in.c (input_bb): Restore BB discriminator.
5984         * gimple-streamer-out.c (output_bb): Save it.
5985         * lto-streamer-in.c (input_struct_function_base): Restore
5986         instance discriminator if available.  Create map on demand.
5987         * lto-streamer-out.c (output_struct_function_base): Save it if
5988         available.
5989         * final.c (decl_to_instance_map): Document LTO strategy.
5991 2018-07-31  Alexandre Oliva  <oliva@adacore.com>
5992             Olivier Hainque  <hainque@adacore.com>
5994         * debug.h (decl_to_instance_map_t): New type.
5995         (decl_to_instance_map): Declare.
5996         (maybe_create_decl_to_instance_map): New inline function.
5997         * final.c (bb_discriminator, last_bb_discriminator): New statics,
5998         to track basic block discriminators.
5999         (final_start_function_1): Initialize them.
6000         (final_scan_insn_1): On NOTE_INSN_BASIC_BLOCK, track
6001         bb_discriminator.
6002         (decl_to_instance_map): New variable.
6003         (map_decl_to_instance, maybe_set_discriminator): New functions.
6004         (notice_source_line): Set discriminator.
6006 2018-07-31  Ian Lance Taylor  <iant@golang.org>
6008         * targhooks.c (default_have_speculation_safe_value): Add
6009         ATTRIBUTE_UNUSED.
6011 2018-07-31  David Malcolm  <dmalcolm@redhat.com>
6013         * dump-context.h: Include "pretty-print.h".
6014         (dump_context::refresh_dumps_are_enabled): New decl.
6015         (dump_context::emit_item): New decl.
6016         (class dump_context): Add fields "m_test_pp" and
6017         "m_test_pp_flags".
6018         (temp_dump_context::temp_dump_context): Add param "test_pp_flags".
6019         (temp_dump_context::get_dumped_text): New decl.
6020         (class temp_dump_context): Add field "m_pp".
6021         * dumpfile.c (refresh_dumps_are_enabled): Convert to...
6022         (dump_context::refresh_dumps_are_enabled): ...and add a test for
6023         m_test_pp.
6024         (set_dump_file): Update for above change.
6025         (set_alt_dump_file): Likewise.
6026         (dump_loc): New overload, taking a pretty_printer *.
6027         (dump_context::dump_loc): Call end_any_optinfo.  Dump the location
6028         to any test pretty-printer.
6029         (make_item_for_dump_gimple_stmt): New function, adapted from
6030         optinfo::add_gimple_stmt.
6031         (dump_context::dump_gimple_stmt): Call it, and use the result,
6032         eliminating the direct usage of dump_file and alt_dump_file in
6033         favor of indirectly using them via emit_item.
6034         (make_item_for_dump_gimple_expr): New function, adapted from
6035         optinfo::add_gimple_expr.
6036         (dump_context::dump_gimple_expr): Call it, and use the result,
6037         eliminating the direct usage of dump_file and alt_dump_file in
6038         favor of indirectly using them via emit_item.
6039         (make_item_for_dump_generic_expr): New function, adapted from
6040         optinfo::add_tree.
6041         (dump_context::dump_generic_expr): Call it, and use the result,
6042         eliminating the direct usage of dump_file and alt_dump_file in
6043         favor of indirectly using them via emit_item.
6044         (make_item_for_dump_printf_va): New function, adapted from
6045         optinfo::add_printf_va.
6046         (make_item_for_dump_printf): New function.
6047         (dump_context::dump_printf_va): Call make_item_for_dump_printf_va,
6048         and use the result, eliminating the direct usage of dump_file and
6049         alt_dump_file in favor of indirectly using them via emit_item.
6050         (make_item_for_dump_dec): New function.
6051         (dump_context::dump_dec): Call it, and use the result,
6052         eliminating the direct usage of dump_file and alt_dump_file in
6053         favor of indirectly using them via emit_item.
6054         (make_item_for_dump_symtab_node): New function, adapted from
6055         optinfo::add_symtab_node.
6056         (dump_context::dump_symtab_node): Call it, and use the result,
6057         eliminating the direct usage of dump_file and alt_dump_file in
6058         favor of indirectly using them via emit_item.
6059         (dump_context::begin_scope): Reimplement, avoiding direct usage
6060         of dump_file and alt_dump_file in favor of indirectly using them
6061         via emit_item.
6062         (dump_context::emit_item): New member function.
6063         (temp_dump_context::temp_dump_context): Add param "test_pp_flags".
6064         Set up test pretty-printer on the underlying context.  Call
6065         refresh_dumps_are_enabled.
6066         (temp_dump_context::~temp_dump_context): Call
6067         refresh_dumps_are_enabled.
6068         (temp_dump_context::get_dumped_text): New member function.
6069         (selftest::verify_dumped_text): New function.
6070         (ASSERT_DUMPED_TEXT_EQ): New macro.
6071         (selftest::test_capture_of_dump_calls): Run all tests twice, with
6072         and then without optinfo enabled.  Add uses of
6073         ASSERT_DUMPED_TEXT_EQ to all tests.  Add test of nested scopes.
6074         * dumpfile.h: Update comment for the dump_* API.
6075         * optinfo-emit-json.cc
6076         (selftest::test_building_json_from_dump_calls): Update for new
6077         param for temp_dump_context ctor.
6078         * optinfo.cc (optinfo_item::optinfo_item): Remove "owned" param
6079         and "m_owned" field.
6080         (optinfo_item::~optinfo_item): Likewise.
6081         (optinfo::add_item): New member function.
6082         (optinfo::emit): Update comment.
6083         (optinfo::add_string): Delete.
6084         (optinfo::add_printf): Delete.
6085         (optinfo::add_printf_va): Delete.
6086         (optinfo::add_gimple_stmt): Delete.
6087         (optinfo::add_gimple_expr): Delete.
6088         (optinfo::add_tree): Delete.
6089         (optinfo::add_symtab_node): Delete.
6090         (optinfo::add_dec): Delete.
6091         * optinfo.h (class dump_context): New forward decl.
6092         (optinfo::add_item): New decl.
6093         (optinfo::add_string): Delete.
6094         (optinfo::add_printf): Delete.
6095         (optinfo::add_printf_va): Delete.
6096         (optinfo::add_gimple_stmt): Delete.
6097         (optinfo::add_gimple_expr): Delete.
6098         (optinfo::add_tree): Delete.
6099         (optinfo::add_symtab_node): Delete.
6100         (optinfo::add_dec): Delete.
6101         (optinfo::add_poly_int): Delete.
6102         (optinfo_item::optinfo_item): Remove "owned" param.
6103         (class optinfo_item): Remove field "m_owned".
6105 2018-07-31  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
6107         PR middle-end/86705
6108         * gcc/cfgexpand.c (set_parm_rtl): Use the alignment of Pmode when
6109         MAX_SUPPORTED_STACK_ALIGNMENT would otherwise be exceeded by the
6110         requested variable alignment.
6111         (expand_one_ssa_partition): Likewise.
6112         (expand_one_var): Likewise.
6114 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
6116         * config/pdp11/pdp11.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
6117         to speculation_safe_value_not_needed.
6119 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
6121         * targhooks.h (speculation_safe_value_not_needed): New prototype.
6122         * targhooks.c (speculation_safe_value_not_needed): New function.
6123         * target.def (have_speculation_safe_value): Update documentation.
6124         * doc/tm.texi: Regenerated.
6126 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
6128         * config/aarch64/iterators.md (ALLI_TI): New iterator.
6129         * config/aarch64/aarch64.md (despeculate_copy<ALLI_TI:mode>): New
6130         expand.
6131         (despeculate_copy<ALLI:mode>_insn): New insn.
6132         (despeculate_copyti_insn): New insn.
6133         (despeculate_simple<ALLI:mode>): New insn
6134         (despeculate_simpleti): New insn.
6135         * config/aarch64/aarch64.c (aarch64_speculation_safe_value): New
6136         function.
6137         (TARGET_SPECULATION_SAFE_VALUE): Redefine to
6138         aarch64_speculation_safe_value.
6139         (aarch64_print_operand): Handle const0_rtx in modifier 'H'.
6141 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
6143         * config/aarch64/aarch64-speculation.cc: New file.
6144         * config/aarch64/aarch64-passes.def (pass_track_speculation): Add
6145         before pass_reorder_blocks.
6146         * config/aarch64/aarch64-protos.h (make_pass_track_speculation): Add
6147         prototype.
6148         * config/aarch64/aarch64.c (aarch64_conditional_register_usage): Fix
6149         X14 and X15 when tracking speculation.
6150         * config/aarch64/aarch64.md (register name constants): Add
6151         SPECULATION_TRACKER_REGNUM and SPECULATION_SCRATCH_REGNUM.
6152         (unspec): Add UNSPEC_SPECULATION_TRACKER.
6153         (speculation_barrier): New insn attribute.
6154         (cmp<mode>): Allow SP in comparisons.
6155         (speculation_tracker): New insn.
6156         (speculation_barrier): Add speculation_barrier attribute.
6157         * config/aarch64/t-aarch64: Add make rule for aarch64-speculation.o.
6158         * config.gcc (aarch64*-*-*): Add aarch64-speculation.o to extra_objs.
6159         * doc/invoke.texi (AArch64 Options): Document -mtrack-speculation.
6161 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
6163         * config/aarch64/aarch64.md (cb<optab><mode>1): Disable when
6164         aarch64_track_speculation is true.
6165         (tb<optab><mode>1): Likewise.
6166         * config/aarch64/aarch64.c (aarch64_split_compare_regs): Do not
6167         generate CB[N]Z when tracking speculation.
6168         (aarch64_split_compare_and_swap): Likewise.
6169         (aarch64_split_atomic_op): Likewise.
6171 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
6173         * config/aarch64/aarch64.opt (mtrack-speculation): New target option.
6175 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
6177         * config/aarch64.md (unspecv): Add UNSPECV_SPECULAION_BARRIER.
6178         (speculation_barrier): New insn.
6180 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
6182         * config/arm/unspecs.md (unspecv): Add VUNSPEC_SPECULATION_BARRIER.
6183         * config/arm/arm.md (speculation_barrier): New expand.
6184         (speculation_barrier_insn): New pattern.
6186 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
6188         * builtin-types.def (BT_FN_PTR_PTR_VAR): New function type.
6189         (BT_FN_I1_I1_VAR, BT_FN_I2_I2_VAR, BT_FN_I4_I4_VAR): Likewise.
6190         (BT_FN_I8_I8_VAR, BT_FN_I16_I16_VAR): Likewise.
6191         * builtin-attrs.def (ATTR_NOVOPS_NOTHROW_LEAF_LIST): New attribute
6192         list.
6193         * builtins.def (BUILT_IN_SPECULATION_SAFE_VALUE_N): New builtin.
6194         (BUILT_IN_SPECULATION_SAFE_VALUE_PTR): New internal builtin.
6195         (BUILT_IN_SPECULATION_SAFE_VALUE_1): Likewise.
6196         (BUILT_IN_SPECULATION_SAFE_VALUE_2): Likewise.
6197         (BUILT_IN_SPECULATION_SAFE_VALUE_4): Likewise.
6198         (BUILT_IN_SPECULATION_SAFE_VALUE_8): Likewise.
6199         (BUILT_IN_SPECULATION_SAFE_VALUE_16): Likewise.
6200         * builtins.c (expand_speculation_safe_value): New function.
6201         (expand_builtin): Call it.
6202         * doc/cpp.texi: Document predefine __HAVE_SPECULATION_SAFE_VALUE.
6203         * doc/extend.texi: Document __builtin_speculation_safe_value.
6204         * doc/md.texi: Document "speculation_barrier" pattern.
6205         * doc/tm.texi.in: Pull in TARGET_SPECULATION_SAFE_VALUE and
6206         TARGET_HAVE_SPECULATION_SAFE_VALUE.
6207         * doc/tm.texi: Regenerated.
6208         * target.def (have_speculation_safe_value, speculation_safe_value): New
6209         hooks.
6210         * targhooks.c (default_have_speculation_safe_value): New function.
6211         (default_speculation_safe_value): New function.
6212         * targhooks.h (default_have_speculation_safe_value): Add prototype.
6213         (default_speculation_safe_value): Add prototype.
6215 2018-07-31  David Malcolm  <dmalcolm@redhat.com>
6217         * dump-context.h (dump_context::dump_loc): New decl.
6218         * dumpfile.c (dump_context::dump_loc): New member function.
6219         (dump_context::dump_gimple_stmt_loc): Reimplement using dump_loc
6220         and dump_gimple_stmt.
6221         (dump_context::dump_gimple_expr_loc): Likewise, using
6222         dump_gimple_expr.
6223         (dump_context::dump_generic_expr_loc): Likewise, using
6224         dump_generic_expr.
6225         (dump_context::dump_printf_loc_va): Likewise, using
6226         dump_printf_va.
6227         (dump_context::begin_scope): Explicitly using the global function
6228         "dump_loc", rather than the member function.
6230 2018-07-31  Martin Sebor  <msebor@redhat.com>
6232         PR tree-optimization/86741
6233         * tree-vrp.c (vrp_prop::check_mem_ref): Avoid incomplete types.
6235 2018-07-31  Andreas Krebbel  <krebbel@linux.ibm.com>
6237         * config/s390/s390.c (s390_expand_setmem): Make the unrolling to
6238         depend on whether prefetch instructions will be emitted or not.
6239         Use TARGET_SETMEM_PFD for checking whether prefetch instructions
6240         will be emitted or not.
6241         * config/s390/s390.h (TARGET_SETMEM_PREFETCH_DISTANCE)
6242         (TARGET_SETMEM_PFD): New macros.
6244 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
6246         * tree-vectorizer.h (stmt_vec_info): Turn back into a typedef.
6247         (NULL_STMT_VEC_INFO): Delete.
6248         (stmt_vec_info::operator*): Likewise.
6249         (stmt_vec_info::operator gimple *): Likewise.
6250         * tree-vect-loop.c (vectorizable_reduction): Use NULL instead
6251         of NULL_STMT_VEC_INFO.
6252         * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
6253         (vect_reassociating_reduction_p): Likewise.
6254         * tree-vect-stmts.c (vect_build_gather_load_calls): Likewise.
6255         (vectorizable_store): Likewise.
6256         * tree-vectorizer.c (vec_info::set_vinfo_for_stmt): Likewise.
6257         (vec_info::free_stmt_vec_infos): Likewise.
6259 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
6261         * tree-vectorizer.h (vect_stmt_in_region_p): Delete.
6262         * tree-vectorizer.c (vect_stmt_in_region_p): Likewise.
6264 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
6266         * tree-vectorizer.h (vec_info::new_vinfo_for_stmt)
6267         (vec_info::set_vinfo_for_stmt, vec_info::free_stmt_vec_infos)
6268         (vec_info::free_stmt_vec_info): New private member functions.
6269         (set_stmt_vec_info_vec, free_stmt_vec_infos, vinfo_for_stmt)
6270         (set_vinfo_for_stmt, new_stmt_vec_info, free_stmt_vec_info): Delete.
6271         * tree-parloops.c (gather_scalar_reductions): Remove calls to
6272         set_stmt_vec_info_vec and free_stmt_vec_infos.
6273         * tree-vect-loop.c (_loop_vec_info): Remove call to
6274         set_stmt_vec_info_vec.
6275         * tree-vect-stmts.c (new_stmt_vec_info, set_stmt_vec_info_vec)
6276         (free_stmt_vec_infos, free_stmt_vec_info): Delete in favor of...
6277         * tree-vectorizer.c (vec_info::new_stmt_vec_info)
6278         (vec_info::set_vinfo_for_stmt, vec_info::free_stmt_vec_infos)
6279         (vec_info::free_stmt_vec_info): ...these new functions.  Remove
6280         assignments in {vec_info::,}new_stmt_vec_info that are redundant
6281         with the clearing in the xcalloc.
6282         (stmt_vec_info_vec): Delete.
6283         (vec_info::vec_info): Don't call set_stmt_vec_info_vec.
6284         (vectorize_loops): Likewise.
6285         (vec_info::~vec_info): Remove argument from call to
6286         free_stmt_vec_infos.
6287         (vec_info::add_stmt): Remove vinfo argument from call to
6288         new_stmt_vec_info.
6290 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
6292         * tree-vectorizer.h (free_stmt_vec_info): Take a stmt_vec_info
6293         rather than a gimple stmt.
6294         * tree-vect-stmts.c (free_stmt_vec_info): Likewise.  Don't free
6295         information for pattern statements when passed the original
6296         statement; instead wait to be passed the pattern statement itself.
6297         Don't call set_vinfo_for_stmt here.
6298         (free_stmt_vec_infos): Update call to free_stmt_vec_info.
6299         * tree-vect-loop.c (_loop_vec_info::~loop_vec_info): Don't free
6300         stmt_vec_infos here.
6301         * tree-vect-slp.c (_bb_vec_info::~bb_vec_info): Likewise.
6302         * tree-vectorizer.c (vec_info::remove_stmt): Nullify the statement's
6303         stmt_vec_infos entry.
6305 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
6307         * tree-vectorizer.h (vec_info::replace_stmt): Declare.
6308         * tree-vectorizer.c (vec_info::replace_stmt): New function.
6309         * tree-vect-slp.c (vect_remove_slp_scalar_calls): Use it.
6310         * tree-vect-stmts.c (vectorizable_call): Likewise.
6311         (vectorizable_simd_clone_call): Likewise.
6313 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
6315         * tree-vectorizer.h (vec_info::remove_stmt): Declare.
6316         * tree-vectorizer.c (vec_info::remove_stmt): New function.
6317         * tree-vect-loop-manip.c (vect_set_loop_condition): Use it.
6318         * tree-vect-loop.c (vect_transform_loop): Likewise.
6319         * tree-vect-slp.c (vect_schedule_slp): Likewise.
6320         * tree-vect-stmts.c (vect_remove_stores): Likewise.
6322 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
6324         * tree-vectorizer.h (vec_info::lookup_dr): New member function.
6325         (vect_dr_stmt): Delete.
6326         * tree-vectorizer.c (vec_info::lookup_dr): New function.
6327         * tree-vect-loop-manip.c (vect_update_inits_of_drs): Use it instead
6328         of DR_VECT_AUX.
6329         * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
6330         (vect_analyze_data_ref_dependence, vect_record_base_alignments)
6331         (vect_verify_datarefs_alignment, vect_peeling_supportable)
6332         (vect_analyze_data_ref_accesses, vect_prune_runtime_alias_test_list)
6333         (vect_analyze_data_refs): Likewise.
6334         (vect_slp_analyze_data_ref_dependence): Likewise.  Take a vec_info
6335         argument.
6336         (vect_find_same_alignment_drs): Likewise.
6337         (vect_slp_analyze_node_dependences): Update calls accordingly.
6338         (vect_analyze_data_refs_alignment): Likewise.  Use vec_info::lookup_dr
6339         instead of DR_VECT_AUX.
6340         (vect_get_peeling_costs_all_drs): Take a loop_vec_info instead
6341         of a vector data references.  Use vec_info::lookup_dr instead of
6342         DR_VECT_AUX.
6343         (vect_peeling_hash_get_lowest_cost): Update calls accordingly.
6344         (vect_enhance_data_refs_alignment): Likewise.  Use vec_info::lookup_dr
6345         instead of DR_VECT_AUX.
6347 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
6349         * tree-vectorizer.h (_loop_vec_info::unaligned_dr): Change to
6350         dr_vec_info.
6351         * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Update
6352         accordingly.
6353         * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
6354         * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
6355         (vect_gen_prolog_loop_niters): Likewise.
6357 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
6359         * tree-vectorizer.h (set_dr_misalignment, dr_misalignment)
6360         (DR_TARGET_ALIGNMENT, aligned_access_p, known_alignment_for_access_p)
6361         (vect_known_alignment_in_bytes, vect_dr_behavior)
6362         (vect_get_scalar_dr_size): Take references as dr_vec_infos
6363         instead of data_references.  Update calls to other routines for
6364         which the same change has been made.
6365         * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Take
6366         dr_vec_infos instead of stmt_vec_infos.
6367         (vect_analyze_data_ref_dependence): Update call accordingly.
6368         (vect_slp_analyze_data_ref_dependence)
6369         (vect_record_base_alignments): Use DR_VECT_AUX.
6370         (vect_calculate_target_alignment, vect_compute_data_ref_alignment)
6371         (vect_update_misalignment_for_peel, verify_data_ref_alignment)
6372         (vector_alignment_reachable_p, vect_get_data_access_cost)
6373         (vect_peeling_supportable, vect_analyze_group_access_1)
6374         (vect_analyze_group_access, vect_analyze_data_ref_access)
6375         (vect_vfa_segment_size, vect_vfa_access_size, vect_vfa_align)
6376         (vect_compile_time_alias, vect_small_gap_p)
6377         (vectorizable_with_step_bound_p, vect_duplicate_ssa_name_ptr_info):
6378         (vect_supportable_dr_alignment): Take references as dr_vec_infos
6379         instead of data_references.  Update calls to other routines for
6380         which the same change has been made.
6381         (vect_verify_datarefs_alignment, vect_get_peeling_costs_all_drs)
6382         (vect_find_same_alignment_drs, vect_analyze_data_refs_alignment)
6383         (vect_slp_analyze_and_verify_node_alignment)
6384         (vect_analyze_data_ref_accesses, vect_prune_runtime_alias_test_list)
6385         (vect_create_addr_base_for_vector_ref, vect_create_data_ref_ptr)
6386         (vect_setup_realignment): Use dr_vec_infos.  Update calls after
6387         above changes.
6388         (_vect_peel_info::dr): Replace with...
6389         (_vect_peel_info::dr_info): ...this new field.
6390         (vect_peeling_hash_get_most_frequent)
6391         (vect_peeling_hash_choose_best_peeling): Update accordingly.
6392         (vect_peeling_hash_get_lowest_cost):
6393         (vect_enhance_data_refs_alignment): Likewise.  Update calls to other
6394         routines for which the same change has been made.
6395         (vect_peeling_hash_insert): Likewise.  Take a dr_vec_info instead of a
6396         data_reference.
6397         * tree-vect-loop-manip.c (get_misalign_in_elems)
6398         (vect_gen_prolog_loop_niters): Use dr_vec_infos.  Update calls after
6399         above changes.
6400         * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
6401         * tree-vect-stmts.c (vect_get_store_cost, vect_get_load_cost)
6402         (vect_truncate_gather_scatter_offset, compare_step_with_zero)
6403         (get_group_load_store_type, get_negative_load_store_type)
6404         (vect_get_data_ptr_increment, vectorizable_store)
6405         (vectorizable_load): Likewise.
6406         (ensure_base_align): Take a dr_vec_info instead of a data_reference.
6407         Update calls to other routines for which the same change has been made.
6409 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
6411         * tree-vectorizer.h (vec_info::move_dr): New member function.
6412         (dataref_aux): Rename to...
6413         (dr_vec_info): ...this and add "dr" and "stmt" fields.
6414         (_stmt_vec_info::dr_aux): Update accordingly.
6415         (_stmt_vec_info::data_ref_info): Delete.
6416         (STMT_VINFO_GROUPED_ACCESS, DR_GROUP_FIRST_ELEMENT)
6417         (DR_GROUP_NEXT_ELEMENT, DR_GROUP_SIZE, DR_GROUP_STORE_COUNT)
6418         (DR_GROUP_GAP, DR_GROUP_SAME_DR_STMT, REDUC_GROUP_FIRST_ELEMENT):
6419         (REDUC_GROUP_NEXT_ELEMENT, REDUC_GROUP_SIZE): Use dr_aux.dr instead
6420         of data_ref.
6421         (STMT_VINFO_DATA_REF): Likewise.  Turn into an lvalue.
6422         (STMT_VINFO_DR_INFO): New macro.
6423         (DR_VECT_AUX): Use STMT_VINFO_DR_INKFO and vect_dr_stmt.
6424         (set_dr_misalignment): Update after rename of dataref_aux.
6425         (vect_dr_stmt): Move earlier in file.  Return dr_aux.stmt.
6426         * tree-vect-stmts.c (new_stmt_vec_info): Remove redundant
6427         initialization of STMT_VINFO_DATA_REF.
6428         * tree-vectorizer.c (vec_info::move_dr): New function.
6429         * tree-vect-patterns.c (vect_recog_bool_pattern)
6430         (vect_recog_mask_conversion_pattern)
6431         (vect_recog_gather_scatter_pattern): Use it.
6432         * tree-vect-data-refs.c (vect_analyze_data_refs): Initialize
6433         the "dr" and "stmt" fields of dr_vec_info instead of
6434         STMT_VINFO_DATA_REF.
6436 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
6438         * tree-vectorizer.h (_stmt_vec_info::pattern_stmt_p): New field.
6439         (is_pattern_stmt_p): Use it.
6440         * tree-vect-patterns.c (vect_init_pattern_stmt): Set pattern_stmt_p
6441         on pattern statements.
6443 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
6445         * tree-vect-patterns.c (vect_mark_pattern_stmts): Take the
6446         original stmt as a stmt_vec_info rather than a gimple stmt.
6447         (vect_pattern_recog_1): Take the statement directly as a
6448         stmt_vec_info, rather than via a gimple_stmt_iterator.
6449         Update call to vect_mark_pattern_stmts.
6450         (vect_pattern_recog): Update calls accordingly.
6452 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
6454         * tree-vectorizer.h (vect_get_vec_defs_for_stmt_copy)
6455         (vect_get_vec_def_for_stmt_copy): Take a vec_info rather than
6456         a vect_def_type for the first argument.
6457         * tree-vect-stmts.c (vect_get_vec_defs_for_stmt_copy): Likewise.
6458         (vect_get_vec_def_for_stmt_copy): Likewise.  Return the original
6459         operand if it isn't defined by a vectorized statement.
6460         (vect_build_gather_load_calls): Remove the mask_dt argument and
6461         update calls to vect_get_vec_def_for_stmt_copy.
6462         (vectorizable_bswap): Likewise the dt argument.
6463         (vectorizable_call): Update calls to vectorizable_bswap and
6464         vect_get_vec_def_for_stmt_copy.
6465         (vectorizable_simd_clone_call, vectorizable_assignment)
6466         (vectorizable_shift, vectorizable_operation, vectorizable_condition)
6467         (vectorizable_comparison): Update calls to
6468         vect_get_vec_def_for_stmt_copy.
6469         (vectorizable_store): Likewise.  Remove now-unnecessary calls to
6470         vect_is_simple_use.
6471         (vect_get_loop_based_defs): Remove dt argument and update call
6472         to vect_get_vec_def_for_stmt_copy.
6473         (vectorizable_conversion): Update calls to vect_get_loop_based_defs
6474         and vect_get_vec_def_for_stmt_copy.
6475         (vectorizable_load): Update calls to vect_build_gather_load_calls
6476         and vect_get_vec_def_for_stmt_copy.
6477         * tree-vect-loop.c (vect_create_epilog_for_reduction)
6478         (vectorizable_reduction, vectorizable_live_operation): Update calls
6479         to vect_get_vec_def_for_stmt_copy.
6481 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
6483         * tree-vect-data-refs.c (vect_record_base_alignment): Replace vec_info
6484         and gimple stmt arguments with a stmt_vec_info.
6485         (vect_record_base_alignments): Update calls accordingly.
6486         * tree-vect-slp.c (vect_record_max_nunits): Replace vec_info
6487         and gimple stmt arguments with a stmt_vec_info.
6488         (vect_build_slp_tree_1): Remove vinfo argument and update call
6489         to vect_record_max_nunits.
6490         (vect_build_slp_tree_2): Update calls to vect_build_slp_tree_1
6491         and vect_record_max_nunits.
6493 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
6495         * tree-vectorizer.h (nested_in_vect_loop_p): Move further down
6496         file and take a stmt_vec_info instead of a gimple stmt.
6497         (supportable_widening_operation, vect_finish_replace_stmt)
6498         (vect_finish_stmt_generation, vect_get_store_rhs)
6499         (vect_get_vec_def_for_operand_1, vect_get_vec_def_for_operand)
6500         (vect_get_vec_defs, vect_init_vector, vect_transform_stmt)
6501         (vect_remove_stores, vect_analyze_stmt, vectorizable_condition)
6502         (vect_get_smallest_scalar_type, vect_check_gather_scatter)
6503         (vect_create_data_ref_ptr, bump_vector_ptr)
6504         (vect_permute_store_chain, vect_setup_realignment)
6505         (vect_transform_grouped_load, vect_record_grouped_load_vectors)
6506         (vect_create_addr_base_for_vector_ref, vectorizable_live_operation)
6507         (vectorizable_reduction, vectorizable_induction)
6508         (get_initial_def_for_reduction, is_simple_and_all_uses_invariant)
6509         (vect_get_place_in_interleaving_chain): Take stmt_vec_infos rather
6510         than gimple stmts as arguments.
6511         * tree-vect-data-refs.c (vect_get_smallest_scalar_type)
6512         (vect_preserves_scalar_order_p, vect_slp_analyze_node_dependences)
6513         (can_group_stmts_p, vect_check_gather_scatter)
6514         (vect_create_addr_base_for_vector_ref, vect_create_data_ref_ptr)
6515         (bump_vector_ptr, vect_permute_store_chain, vect_setup_realignment)
6516         (vect_permute_load_chain, vect_shift_permute_load_chain)
6517         (vect_transform_grouped_load)
6518         (vect_record_grouped_load_vectors): Likewise.
6519         * tree-vect-loop.c (vect_fixup_reduc_chain)
6520         (get_initial_def_for_reduction, vect_create_epilog_for_reduction)
6521         (vectorize_fold_left_reduction, is_nonwrapping_integer_induction)
6522         (vectorizable_reduction, vectorizable_induction)
6523         (vectorizable_live_operation, vect_loop_kill_debug_uses): Likewise.
6524         * tree-vect-patterns.c (type_conversion_p, adjust_bool_stmts)
6525         (vect_get_load_store_mask): Likewise.
6526         * tree-vect-slp.c (vect_get_place_in_interleaving_chain)
6527         (vect_analyze_slp_instance, vect_mask_constant_operand_p): Likewise.
6528         * tree-vect-stmts.c (vect_mark_relevant)
6529         (is_simple_and_all_uses_invariant)
6530         (exist_non_indexing_operands_for_use_p, process_use)
6531         (vect_init_vector_1, vect_init_vector, vect_get_vec_def_for_operand_1)
6532         (vect_get_vec_def_for_operand, vect_get_vec_defs)
6533         (vect_finish_stmt_generation_1, vect_finish_replace_stmt)
6534         (vect_finish_stmt_generation, vect_truncate_gather_scatter_offset)
6535         (compare_step_with_zero, vect_get_store_rhs, get_group_load_store_type)
6536         (get_negative_load_store_type, get_load_store_type)
6537         (vect_check_load_store_mask, vect_check_store_rhs)
6538         (vect_build_gather_load_calls, vect_get_strided_load_store_ops)
6539         (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
6540         (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
6541         (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
6542         (get_group_alias_ptr_type, vectorizable_store, hoist_defs_of_uses)
6543         (vectorizable_load, vectorizable_condition, vectorizable_comparison)
6544         (vect_analyze_stmt, vect_transform_stmt, vect_remove_stores)
6545         (supportable_widening_operation): Likewise.
6547 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
6549         * tree-vect-data-refs.c (vect_describe_gather_scatter_call): Take
6550         a stmt_vec_info instead of a gcall.
6551         (vect_check_gather_scatter): Update call accordingly.
6552         * tree-vect-loop-manip.c (iv_phi_p): Take a stmt_vec_info instead
6553         of a gphi.
6554         (vect_can_advance_ivs_p, vect_update_ivs_after_vectorizer)
6555         (slpeel_update_phi_nodes_for_loops):): Update calls accordingly.
6556         * tree-vect-loop.c (vect_transform_loop_stmt): Take a stmt_vec_info
6557         instead of a gimple stmt.
6558         (vect_transform_loop): Update calls accordingly.
6559         * tree-vect-slp.c (vect_split_slp_store_group): Take and return
6560         stmt_vec_infos instead of gimple stmts.
6561         (vect_analyze_slp_instance): Update use accordingly.
6562         * tree-vect-stmts.c (read_vector_array, write_vector_array)
6563         (vect_clobber_variable, vect_stmt_relevant_p, permute_vec_elements)
6564         (vect_use_strided_gather_scatters_p, vect_build_all_ones_mask)
6565         (vect_build_zero_merge_argument, vect_get_gather_scatter_ops)
6566         (vect_gen_widened_results_half, vect_get_loop_based_defs)
6567         (vect_create_vectorized_promotion_stmts, can_vectorize_live_stmts):
6568         Take a stmt_vec_info instead of a gimple stmt and pass stmt_vec_infos
6569         down to subroutines.
6571 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
6573         * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Change the type
6574         of the worklist from a vector of gimple stmts to a vector of
6575         stmt_vec_infos.
6576         * tree-vect-stmts.c (vect_mark_relevant, process_use)
6577         (vect_mark_stmts_to_be_vectorized): Likewise
6579 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
6581         * tree-vect-loop.c (vect_analyze_loop_operations): Look up the
6582         statement before passing it to vect_analyze_stmt.
6583         (vect_create_epilog_for_reduction): Use a stmt_vec_info to walk
6584         the chain of phi vector definitions.  Track the exit phi via its
6585         stmt_vec_info.
6586         (vectorizable_reduction): Set cond_stmt_vinfo directly from the
6587         STMT_VINFO_REDUC_DEF.
6588         * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Use
6589         stmt_vec_infos to handle the statement chains.
6590         (vect_get_slp_defs): Record the first statement in the node
6591         using a stmt_vec_info.
6592         * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Look up
6593         statements here and pass their stmt_vec_info down to subroutines.
6594         (vect_init_vector_1): Hoist call to vinfo_for_stmt and pass it
6595         down to vect_finish_stmt_generation.
6596         (vect_init_vector, vect_get_vec_defs, vect_finish_replace_stmt)
6597         (vect_finish_stmt_generation): Call vinfo_for_stmt and pass
6598         stmt_vec_infos to subroutines.
6599         (vect_remove_stores): Use stmt_vec_infos to handle the statement
6600         chains.
6602 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
6604         * tree-vect-data-refs.c (vect_slp_analyze_node_dependences):
6605         (vect_check_gather_scatter, vect_create_data_ref_ptr, bump_vector_ptr)
6606         (vect_permute_store_chain, vect_setup_realignment)
6607         (vect_permute_load_chain, vect_shift_permute_load_chain)
6608         (vect_transform_grouped_load): Use stmt_vec_info rather than gimple
6609         stmts internally, and when passing values to other vectorizer routines.
6610         * tree-vect-loop-manip.c (vect_can_advance_ivs_p): Likewise.
6611         * tree-vect-loop.c (vect_analyze_scalar_cycles_1)
6612         (vect_analyze_loop_operations, get_initial_def_for_reduction)
6613         (vect_create_epilog_for_reduction, vectorize_fold_left_reduction)
6614         (vectorizable_reduction, vectorizable_induction)
6615         (vectorizable_live_operation, vect_transform_loop_stmt)
6616         (vect_transform_loop): Likewise.
6617         * tree-vect-patterns.c (vect_reassociating_reduction_p)
6618         (vect_recog_widen_op_pattern, vect_recog_mixed_size_cond_pattern)
6619         (vect_recog_bool_pattern, vect_recog_gather_scatter_pattern): Likewise.
6620         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
6621         (vect_slp_analyze_node_operations_1): Likewise.
6622         * tree-vect-stmts.c (vect_mark_relevant, process_use)
6623         (exist_non_indexing_operands_for_use_p, vect_init_vector_1)
6624         (vect_mark_stmts_to_be_vectorized, vect_get_vec_def_for_operand)
6625         (vect_finish_stmt_generation_1, get_group_load_store_type)
6626         (get_load_store_type, vect_build_gather_load_calls)
6627         (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
6628         (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
6629         (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
6630         (vectorizable_store, vectorizable_load, vectorizable_condition)
6631         (vectorizable_comparison, vect_analyze_stmt, vect_transform_stmt)
6632         (supportable_widening_operation): Likewise.
6633         (vect_get_vector_types_for_stmt): Likewise.
6634         * tree-vectorizer.h (vect_dr_behavior): Likewise.
6636 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
6638         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
6639         (vect_slp_analyze_node_dependences, vect_analyze_data_ref_accesses)
6640         (vect_permute_store_chain, vect_permute_load_chain)
6641         (vect_shift_permute_load_chain, vect_transform_grouped_load): Avoid
6642         repeated stmt_vec_info lookups.
6643         * tree-vect-loop-manip.c (vect_can_advance_ivs_p): Likewise.
6644         (vect_update_ivs_after_vectorizer): Likewise.
6645         * tree-vect-loop.c (vect_is_simple_reduction): Likewise.
6646         (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
6647         * tree-vect-patterns.c (adjust_bool_stmts): Likewise.
6648         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
6649         (vect_bb_slp_scalar_cost): Likewise.
6650         * tree-vect-stmts.c (get_group_alias_ptr_type): Likewise.
6652 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
6654         * tree-vect-data-refs.c (vect_check_gather_scatter): Pass the
6655         gcall rather than the generic gimple stmt to gimple_call_internal_fn.
6656         (vect_get_smallest_scalar_type, can_group_stmts_p): Use dyn_cast
6657         to get gassigns and gcalls, rather than operating on generc gimple
6658         stmts.
6659         * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p)
6660         (vect_mark_stmts_to_be_vectorized, vectorizable_store)
6661         (vectorizable_load, vect_analyze_stmt): Likewise.
6662         * tree-vect-loop.c (vectorizable_reduction): Likewise gphi.
6664 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
6666         * tree-vectorizer.h (get_earlier_stmt, get_later_stmt): Take and
6667         return stmt_vec_infos rather than gimple stmts.  Do not accept
6668         null arguments.
6669         (vect_find_last_scalar_stmt_in_slp): Return a stmt_vec_info instead
6670         of a gimple stmt.
6671         * tree-vect-slp.c (vect_find_last_scalar_stmt_in_slp): Likewise.
6672         Update use of get_later_stmt.
6673         (vect_get_constant_vectors): Update call accordingly.
6674         (vect_schedule_slp_instance): Likewise
6675         * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Likewise.
6676         (vect_slp_analyze_instance_dependence): Likewise.
6677         (vect_preserves_scalar_order_p): Update use of get_earlier_stmt.
6679 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
6681         * tree-vectorizer.h (stmt_info_for_cost::stmt): Replace with...
6682         (stmt_info_for_cost::stmt_info): ...this new field.
6683         (add_stmt_costs): Update accordingly.
6684         * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost)
6685         (vect_get_known_peeling_cost): Likewise.
6686         (vect_estimate_min_profitable_iters): Likewise.
6687         * tree-vect-stmts.c (record_stmt_cost): Likewise.
6689 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
6691         * tree-vectorizer.h (_loop_vec_info::may_misalign_stmts): Change
6692         from an auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
6693         * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Update
6694         accordingly.
6695         * tree-vect-loop-manip.c (vect_create_cond_for_align_checks): Likewise.
6697 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
6699         * tree-vectorizer.h (_stmt_vec_info::same_dr_stmt): Change from
6700         a gimple stmt to a stmt_vec_info.
6701         * tree-vect-stmts.c (vectorizable_load): Update accordingly.
6703 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
6705         * tree-vectorizer.h (vec_info::grouped_stores): Change from
6706         an auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
6707         (_loop_vec_info::reduction_chains): Likewise.
6708         * tree-vect-loop.c (vect_fixup_scalar_cycles_with_patterns): Update
6709         accordingly.
6710         * tree-vect-slp.c (vect_analyze_slp): Likewise.
6712 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
6714         * tree-vectorizer.h (_stmt_vec_info::first_element): Change from
6715         a gimple stmt to a stmt_vec_info.
6716         (_stmt_vec_info::next_element): Likewise.
6717         * tree-vect-data-refs.c (vect_update_misalignment_for_peel)
6718         (vect_slp_analyze_and_verify_node_alignment)
6719         (vect_analyze_group_access_1, vect_analyze_group_access)
6720         (vect_small_gap_p, vect_prune_runtime_alias_test_list)
6721         (vect_create_data_ref_ptr, vect_record_grouped_load_vectors)
6722         (vect_supportable_dr_alignment): Update accordingly.
6723         * tree-vect-loop.c (vect_fixup_reduc_chain): Likewise.
6724         (vect_fixup_scalar_cycles_with_patterns, vect_is_slp_reduction)
6725         (vect_is_simple_reduction, vectorizable_reduction): Likewise.
6726         * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
6727         * tree-vect-slp.c (vect_build_slp_tree_1)
6728         (vect_attempt_slp_rearrange_stmts, vect_supported_load_permutation_p)
6729         (vect_split_slp_store_group, vect_analyze_slp_instance)
6730         (vect_analyze_slp, vect_transform_slp_perm_load): Likewise.
6731         * tree-vect-stmts.c (vect_model_store_cost, vect_model_load_cost)
6732         (get_group_load_store_type, get_load_store_type)
6733         (get_group_alias_ptr_type, vectorizable_store, vectorizable_load)
6734         (vect_transform_stmt, vect_remove_stores): Likewise.
6736 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
6738         * tree-vectorizer.h (vect_dr_stmt): Return a stmt_vec_info rather
6739         than a gimple stmt.
6740         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
6741         (vect_slp_analyze_data_ref_dependence, vect_record_base_alignments)
6742         (vect_calculate_target_alignmentm, vect_compute_data_ref_alignment)
6743         (vect_update_misalignment_for_peel, vect_verify_datarefs_alignment)
6744         (vector_alignment_reachable_p, vect_get_data_access_cost)
6745         (vect_get_peeling_costs_all_drs, vect_peeling_hash_get_lowest_cost)
6746         (vect_peeling_supportable, vect_enhance_data_refs_alignment)
6747         (vect_find_same_alignment_drs, vect_analyze_data_refs_alignment)
6748         (vect_analyze_group_access_1, vect_analyze_group_access)
6749         (vect_analyze_data_ref_access, vect_analyze_data_ref_accesses)
6750         (vect_vfa_access_size, vect_small_gap_p, vect_analyze_data_refs)
6751         (vect_supportable_dr_alignment): Remove vinfo_for_stmt from the
6752         result of vect_dr_stmt and use the stmt_vec_info instead of
6753         the associated gimple stmt.
6754         * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
6755         (vect_gen_prolog_loop_niters): Likewise.
6756         * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
6758 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
6760         * tree-vectorizer.h (_slp_tree::stmts): Change from a vec<gimple *>
6761         to a vec<stmt_vec_info>.
6762         * tree-vect-slp.c (vect_free_slp_tree): Update accordingly.
6763         (vect_create_new_slp_node): Take a vec<gimple *> instead of a
6764         vec<stmt_vec_info>.
6765         (_slp_oprnd_info::def_stmts): Change from a vec<gimple *>
6766         to a vec<stmt_vec_info>.
6767         (bst_traits::value_type, bst_traits::value_type): Likewise.
6768         (bst_traits::hash): Update accordingly.
6769         (vect_get_and_check_slp_defs): Change the stmts parameter from
6770         a vec<gimple *> to a vec<stmt_vec_info>.
6771         (vect_two_operations_perm_ok_p, vect_build_slp_tree_1): Likewise.
6772         (vect_build_slp_tree): Likewise.
6773         (vect_build_slp_tree_2): Likewise.  Update uses of
6774         SLP_TREE_SCALAR_STMTS.
6775         (vect_print_slp_tree): Update uses of SLP_TREE_SCALAR_STMTS.
6776         (vect_mark_slp_stmts, vect_mark_slp_stmts_relevant)
6777         (vect_slp_rearrange_stmts, vect_attempt_slp_rearrange_stmts)
6778         (vect_supported_load_permutation_p, vect_find_last_scalar_stmt_in_slp)
6779         (vect_detect_hybrid_slp_stmts, vect_slp_analyze_node_operations_1)
6780         (vect_slp_analyze_node_operations, vect_slp_analyze_operations)
6781         (vect_bb_slp_scalar_cost, vect_slp_analyze_bb_1)
6782         (vect_get_constant_vectors, vect_get_slp_defs)
6783         (vect_transform_slp_perm_load, vect_schedule_slp_instance)
6784         (vect_remove_slp_scalar_calls, vect_schedule_slp): Likewise.
6785         (vect_analyze_slp_instance): Build up a vec of stmt_vec_infos
6786         instead of gimple stmts.
6787         * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Change
6788         the stores parameter for a vec<gimple *> to a vec<stmt_vec_info>.
6789         (vect_slp_analyze_instance_dependence): Update uses of
6790         SLP_TREE_SCALAR_STMTS.
6791         (vect_slp_analyze_and_verify_node_alignment): Likewise.
6792         (vect_slp_analyze_and_verify_instance_alignment): Likewise.
6793         * tree-vect-loop.c (neutral_op_for_slp_reduction): Likewise.
6794         (get_initial_defs_for_reduction): Likewise.
6795         (vect_create_epilog_for_reduction): Likewise.
6796         (vectorize_fold_left_reduction): Likewise.
6797         * tree-vect-stmts.c (vect_prologue_cost_for_slp_op): Likewise.
6798         (vect_model_simple_cost, vectorizable_shift, vectorizable_load)
6799         (can_vectorize_live_stmts): Likewise.
6801 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
6803         * tree-vectorizer.h (_loop_vec_info::reductions): Change from an
6804         auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
6805         (vect_force_simple_reduction): Take and return stmt_vec_infos rather
6806         than gimple stmts.
6807         * tree-parloops.c (valid_reduction_p): Take a stmt_vec_info instead
6808         of a gimple stmt.
6809         (gather_scalar_reductions): Update after above interface changes.
6810         * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Likewise.
6811         (vect_is_simple_reduction): Take and return stmt_vec_infos rather
6812         than gimple stmts.
6813         (vect_force_simple_reduction): Likewise.
6814         * tree-vect-patterns.c (vect_pattern_recog_1): Update use of
6815         LOOP_VINFO_REDUCTIONS.
6816         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
6818 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
6820         * tree-vectorizer.h (_stmt_vec_info::reduc_def): Change from
6821         a gimple stmt to a stmt_vec_info.
6822         * tree-vect-loop.c (vect_active_double_reduction_p)
6823         (vect_force_simple_reduction, vectorizable_reduction): Update
6824         accordingly.
6826 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
6828         * tree-vectorizer.h (_slp_tree::vec_stmts): Change from a
6829         vec<gimple *> to a vec<stmt_vec_info>.
6830         * tree-vect-loop.c (vect_create_epilog_for_reduction): Change
6831         the reduction_phis argument from a vec<gimple *> to a
6832         vec<stmt_vec_info>.
6833         (vectorizable_reduction): Likewise the phis local variable that
6834         is passed to vect_create_epilog_for_reduction.  Update for new type
6835         of SLP_TREE_VEC_STMTS.
6836         (vectorizable_induction): Update for new type of SLP_TREE_VEC_STMTS.
6837         (vectorizable_live_operation): Likewise.
6838         * tree-vect-slp.c (vect_get_slp_vect_defs): Likewise.
6839         (vect_transform_slp_perm_load, vect_schedule_slp_instance): Likewise.
6841 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
6843         * tree-vectorizer.h (_stmt_vec_info::vectorized_stmt): Change from
6844         a gimple stmt to a stmt_vec_info.
6845         (vectorizable_condition, vectorizable_live_operation)
6846         (vectorizable_reduction, vectorizable_induction): Pass back the
6847         vectorized statement as a stmt_vec_info.
6848         * tree-vect-data-refs.c (vect_record_grouped_load_vectors): Update
6849         use of STMT_VINFO_VEC_STMT.
6850         * tree-vect-loop.c (vect_create_epilog_for_reduction): Likewise,
6851         accumulating the inner phis that feed the STMT_VINFO_VEC_STMT
6852         as stmt_vec_infos rather than gimple stmts.
6853         (vectorize_fold_left_reduction): Change vec_stmt from a gimple stmt
6854         to a stmt_vec_info.
6855         (vectorizable_live_operation): Likewise.
6856         (vectorizable_reduction, vectorizable_induction): Likewise,
6857         updating use of STMT_VINFO_VEC_STMT.
6858         * tree-vect-stmts.c (vect_get_vec_def_for_operand_1): Update use
6859         of STMT_VINFO_VEC_STMT.
6860         (vect_build_gather_load_calls, vectorizable_bswap, vectorizable_call)
6861         (vectorizable_simd_clone_call, vectorizable_conversion)
6862         (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
6863         (vectorizable_store, vectorizable_load, vectorizable_condition)
6864         (vectorizable_comparison, can_vectorize_live_stmts): Change vec_stmt
6865         from a gimple stmt to a stmt_vec_info.
6866         (vect_transform_stmt): Update use of STMT_VINFO_VEC_STMT.  Pass a
6867         pointer to a stmt_vec_info to the vectorizable_* routines.
6869 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
6871         * tree-vectorizer.h (_stmt_vec_info::related_stmt): Change from
6872         a gimple stmt to a stmt_vec_info.
6873         (is_pattern_stmt_p): Update accordingly.
6874         * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Likewise.
6875         (vect_record_grouped_load_vectors): Likewise.
6876         * tree-vect-loop.c (vect_determine_vf_for_stmt): Likewise.
6877         (vect_fixup_reduc_chain, vect_update_vf_for_slp): Likewise.
6878         (vect_model_reduction_cost): Likewise.
6879         (vect_create_epilog_for_reduction): Likewise.
6880         (vectorizable_reduction, vectorizable_induction): Likewise.
6881         * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
6882         Return the stmt_vec_info for the pattern statement.
6883         (vect_set_pattern_stmt): Update use of STMT_VINFO_RELATED_STMT.
6884         (vect_split_statement, vect_mark_pattern_stmts): Likewise.
6885         * tree-vect-slp.c (vect_detect_hybrid_slp_stmts): Likewise.
6886         (vect_detect_hybrid_slp, vect_get_slp_defs): Likewise.
6887         * tree-vect-stmts.c (vect_mark_relevant): Likewise.
6888         (vect_get_vec_def_for_operand_1, vectorizable_call): Likewise.
6889         (vectorizable_simd_clone_call, vect_analyze_stmt, new_stmt_vec_info)
6890         (free_stmt_vec_info, vect_is_simple_use): Likewise.
6892 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
6894         * tree-vectorizer.h (vect_finish_replace_stmt): Return a stmt_vec_info
6895         (vect_finish_stmt_generation): Likewise.
6896         * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
6897         (vect_finish_replace_stmt, vect_finish_stmt_generation): Likewise.
6898         (vect_build_gather_load_calls): Use the return value of the above
6899         functions instead of a separate call to vinfo_for_stmt.  Use narrow
6900         scopes for the input gimple stmt and wider scopes for the associated
6901         stmt_vec_info.  Use vec_info::lookup_def when setting these
6902         stmt_vec_infos from an SSA_NAME definition.
6903         (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
6904         (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
6905         (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
6906         (vectorizable_store, vectorizable_load, vectorizable_condition)
6907         (vectorizable_comparison): Likewise.
6908         * tree-vect-loop.c (vectorize_fold_left_reduction): Likewise.
6909         (vectorizable_reduction): Likewise.
6911 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
6913         * tree-vectorizer.h (vect_is_simple_use): Add an optional
6914         stmt_vec_info * parameter before the optional gimple **.
6915         * tree-vect-stmts.c (vect_is_simple_use): Likewise.
6916         (process_use, vect_get_vec_def_for_operand_1): Update callers.
6917         (vect_get_vec_def_for_operand, vectorizable_shift): Likewise.
6918         * tree-vect-loop.c (vectorizable_reduction): Likewise.
6919         (vectorizable_live_operation): Likewise.
6920         * tree-vect-patterns.c (type_conversion_p): Likewise.
6921         (vect_look_through_possible_promotion): Likewise.
6922         (vect_recog_rotate_pattern): Likewise.
6923         * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
6925 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
6927         * tree-vectorizer.h (stmt_vec_info): Temporarily change from
6928         a typedef to a wrapper class.
6929         (NULL_STMT_VEC_INFO): New macro.
6930         (vec_info::stmt_infos): Change to vec<stmt_vec_info>.
6931         (stmt_vec_info::operator*): New function.
6932         (stmt_vec_info::operator gimple *): Likewise.
6933         (set_vinfo_for_stmt): Use NULL_STMT_VEC_INFO.
6934         (add_stmt_costs): Likewise.
6935         * tree-vect-loop-manip.c (iv_phi_p): Likewise.
6936         * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost)
6937         (vect_get_known_peeling_cost): Likewise.
6938         (vect_estimate_min_profitable_iters): Likewise.
6939         * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
6940         * tree-vect-slp.c (vect_remove_slp_scalar_calls): Likewise.
6941         * tree-vect-stmts.c (vect_build_gather_load_calls): Likewise.
6942         (vectorizable_store, free_stmt_vec_infos): Likewise.
6943         (new_stmt_vec_info): Change return type of xcalloc to
6944         _stmt_vec_info *.
6946 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
6948         * tree-vectorizer.h (vec_info::lookup_single_use): Declare.
6949         * tree-vectorizer.c (vec_info::lookup_single_use): New function.
6950         * tree-vect-loop.c (vectorizable_reduction): Use it instead of
6951         a single_imm_use-based sequence.
6952         * tree-vect-stmts.c (supportable_widening_operation): Likewise.
6954 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
6956         * tree-vectorizer.h (vec_info::lookup_def): Declare.
6957         * tree-vectorizer.c (vec_info::lookup_def): New function.
6958         * tree-vect-patterns.c (vect_get_internal_def): Use it.
6959         (vect_widened_op_tree): Likewise.
6960         * tree-vect-stmts.c (vect_is_simple_use): Likewise.
6961         * tree-vect-loop.c (vect_analyze_loop_operations): Likewise.
6962         (vectorizable_reduction): Likewise.
6963         (vect_valid_reduction_input_p): Take a stmt_vec_info instead
6964         of a gimple *.
6965         (vect_is_slp_reduction): Update calls accordingly.  Use
6966         vec_info::lookup_def.
6967         (vect_is_simple_reduction): Likewise
6968         * tree-vect-slp.c (vect_detect_hybrid_slp_1): Use vec_info::lookup_def.
6970 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
6972         * tree-vectorizer.h (vec_info::lookup_stmt): Declare.
6973         * tree-vectorizer.c (vec_info::lookup_stmt): New function.
6974         * tree-vect-loop.c (vect_determine_vf_for_stmt): Use it instead
6975         of vinfo_for_stmt.
6976         (vect_determine_vectorization_factor, vect_analyze_scalar_cycles_1)
6977         (vect_compute_single_scalar_iteration_cost, vect_analyze_loop_form)
6978         (vect_update_vf_for_slp, vect_analyze_loop_operations)
6979         (vect_is_slp_reduction, vectorizable_induction)
6980         (vect_transform_loop_stmt, vect_transform_loop): Likewise.
6981         * tree-vect-patterns.c (vect_init_pattern_stmt):
6982         (vect_determine_min_output_precision_1, vect_determine_precisions)
6983         (vect_pattern_recog): Likewise.
6984         * tree-vect-stmts.c (vect_analyze_stmt, vect_transform_stmt): Likewise.
6985         * config/powerpcspe/powerpcspe.c (rs6000_density_test): Likewise.
6986         * config/rs6000/rs6000.c (rs6000_density_test): Likewise.
6987         * tree-vect-slp.c (vect_detect_hybrid_slp_stmts): Likewise.
6988         (vect_detect_hybrid_slp_1, vect_detect_hybrid_slp_2)
6989         (vect_detect_hybrid_slp): Likewise.  Change the walk_stmt_info
6990         info field from a loop to a loop_vec_info.
6992 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
6994         * tree-vectorizer.h (stmt_vec_info): Move typedef earlier in file.
6995         (vec_info::add_stmt): Declare.
6996         * tree-vectorizer.c (vec_info::add_stmt): New function.
6997         * tree-vect-data-refs.c (vect_create_data_ref_ptr): Use it.
6998         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Likewise.
6999         (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
7000         (vectorizable_induction): Likewise.
7001         * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Likewise.
7002         * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
7003         (vectorizable_simd_clone_call, vectorizable_store): Likewise.
7004         (vectorizable_load): Likewise.
7005         * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
7006         (vect_recog_bool_pattern, vect_recog_mask_conversion_pattern)
7007         (vect_recog_gather_scatter_pattern): Likewise.
7008         (append_pattern_def_seq): Likewise.  Remove a check that is
7009         performed by add_stmt itself.
7011 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
7013         * tree-vect-loop.c (vectorizable_reduction): Fix an instance in
7014         which make_ssa_name was called with new_stmt before new_stmt
7015         had been created.
7017 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
7019         * tree-vect-loop.c (vect_valid_reduction_input_p): New function,
7020         split out from...
7021         (vect_is_slp_reduction): ...here...
7022         (vect_is_simple_reduction): ...and here.  Remove repetition of tests
7023         that are already known to be false.
7025 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
7027         * tree-vectorizer.h (vect_free_slp_instance): Add a final_p parameter.
7028         * tree-vect-slp.c (vect_free_slp_tree): Likewise.  Don't update
7029         STMT_VINFO_NUM_SLP_USES when it's true.
7030         (vect_free_slp_instance): Add a final_p parameter and pass it to
7031         vect_free_slp_tree.
7032         (vect_build_slp_tree_2): Update call to vect_free_slp_instance.
7033         (vect_analyze_slp_instance): Likewise.
7034         (vect_slp_analyze_operations): Likewise.
7035         (vect_slp_analyze_bb_1): Likewise.
7036         * tree-vectorizer.c (vec_info): Likewise.
7037         * tree-vect-loop.c (vect_transform_loop): Likewise.
7039 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
7041         * tree-vect-loop.c (vectorizable_reduction): Assert that the
7042         function is not called for second and subsequent members of
7043         a reduction group.
7045 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
7047         * tree-vect-loop.c (get_initial_def_for_reduction): Move special
7048         cases for nested loops from here to ...
7049         (vect_create_epilog_for_reduction): ...here.  Only call
7050         vect_is_simple_use for inner-loop reductions.
7052 2018-07-31  Martin Liska  <mliska@suse.cz>
7054         PR gcov-profile/85338
7055         PR gcov-profile/85350
7056         PR gcov-profile/85372
7057         * profile.c (struct location_triplet): New.
7058         (struct location_triplet_hash): Likewise.
7059         (output_location): Do not output a BB that
7060         is already recorded for a line.
7061         (branch_prob): Use streamed_locations.
7063 2018-07-31  Martin Liska  <mliska@suse.cz>
7065         PR gcov-profile/85370
7066         * coverage.c (coverage_begin_function): Do not mark target
7067         clones as artificial functions.
7069 2018-07-31  Martin Liska  <mliska@suse.cz>
7071         PR gcov-profile/83813
7072         PR gcov-profile/84758
7073         PR gcov-profile/85217
7074         PR gcov-profile/85332
7075         * profile.c (branch_prob): Do not record GOTO expressions
7076         for GIMPLE statements which locations are already streamed.
7078 2018-07-31  Olivier Hainque  <hainque@adacore.com>
7080         * gcc.c (handle_spec_function): Accept a soft_matched_part
7081         argument, as do_spec_1.  Pass it down to ...
7082         (eval_spec_function): Accept a soft_matched_part argument,
7083         and pass it down to ...
7084         (do_spec_2): Accept a soft_matched_part argument, and pass
7085         it down to do_spec_1.
7086         (do_spec_1): Pass soft_matched_part to handle_spec_function.
7087         (handle_braces): Update call to handle_spec_function.
7088         (driver::set_up_specs): Update calls to do_spec_2.
7089         (compare_debug_dump_opt_spec_function): Likewise.
7090         (compare_debug_self_opt_spec_function): Likewise.
7092 2018-07-31  Olivier Hainque  <hainque@adacore.com>
7094         * common.opt (nolibc): New option.
7095         * doc/invoke.texi (Link Options): Document it.
7096         * gcc.c (LINK_GCC_C_SEQUENCE_SPEC): Honor nolibc.
7097         * config/alpha/linux.h: Likewise.
7098         * config/arc/elf.h: Likewise.
7099         * config/arm/uclinux-elf.h: Likewise.
7100         * config/arm/unknown-elf.h: Likewise.
7101         * config/avr/avrlibc.h: Likewise.
7102         * config/bfin/bfin.h: Likewise.
7103         * config/bfin/linux.h: Likewise.
7104         * config/bfin/uclinux.h: Likewise.
7105         * config/darwin.h: Likewise.
7106         * config/darwin10.h: Likewise.
7107         * config/darwin12.h: Likewise.
7108         * config/gnu-user.h: Likewise.
7109         * config/lm32/uclinux-elf.h: Likewise.
7110         * config/pa/pa-hpux11.h: Likewise.
7111         * config/pa/pa64-hpux.h: Likewise.
7112         * config/sparc/sparc.h: Likewise.
7114 2018-07-31  Olivier Hainque  <hainque@adacore.com>
7116         * gcc.c (getenv_spec_function): Prepend '/' to value for allowed
7117         undefined variables.
7119 2018-07-30  Segher Boessenkool  <segher@kernel.crashing.org>
7121         PR target/86640
7122         * config/arm/arm.c (arm_block_set_aligned_vect): Use gen_int_mode
7123         instead of GEN_INT.
7125 2018-07-30  Bernd Edlinger  <bernd.edlinger@hotmail.de>
7127         * tree-ssa-forwprop.c (simplify_builtin_call): Don't create a not NUL
7128         terminated string literal.
7130 2018-07-30  Segher Boessenkool  <segher@kernel.crashing.org>
7132         PR rtl-optimization/85160
7133         * combine.c (is_just_move): New function.
7134         (try_combine): Allow combining two instructions into two if neither of
7135         the original instructions was a move.
7137 2018-07-30  Alexander Monakov  <amonakov@ispras.ru>
7139         PR target/86673
7140         * doc/extend.texi (Global Register Variables): Discourage use of type
7141         qualifiers.
7142         (Local Register Variables): Likewise.
7144 2018-07-30  Richard Sandiford  <richard.sandiford@arm.com>
7146         PR tree-optimization/86506
7147         * hwint.h (ceil_log2): Resync with hwint.c implementation.
7149 2018-07-30  Ilya Leoshkevich  <iii@linux.ibm.com>
7151         PR target/86547
7152         * lra-constraints.c (spill_hard_reg_in_range): When selecting the
7153         hard_regno, make sure no insn between `from` and `to` clobbers it.
7155 2018-07-30  Cesar Philippidis  <cesar@codesourcery.com>
7156             Tom de Vries  <tdevries@suse.de>
7158         * config/nvptx/nvptx.c (PTX_GANG_DEFAULT): Rename to ...
7159         (PTX_DEFAULT_RUNTIME_DIM): ... this.
7160         (nvptx_goacc_validate_dims): Set default worker and gang dims to
7161         PTX_DEFAULT_RUNTIME_DIM.
7162         (nvptx_dim_limit): Ignore GOMP_DIM_WORKER.
7164 2018-07-29  John David Anglin  <danglin@gcc.gnu.org>
7166         * config/pa/pa.c (pa_output_addr_vec): Align address table.
7167         * config/pa/pa.h (JUMP_TABLES_IN_TEXT_SECTION): Revise comment.
7168         * config/pa/pa32-linux.h (JUMP_TABLES_IN_TEXT_SECTION): Define.
7170 2018-07-27  Michael Meissner  <meissner@linux.ibm.com>
7172         * config/rs6000/constraints.md (wG constraint): Delete, no longer
7173         used.
7174         * config/rs6000/predicates.md (p9_fusion_reg_operand): Rename
7175         predicate to reflect toc fusion has been deleted.
7176         (toc_fusion_mem_raw): Delete, no longer used.
7177         (toc_fusion_mem_wrapped): Likewise.
7178         * config/rs6000/rs6000-cpus.def (POWERPC_MASKS): Delete toc
7179         fusion mask bit.
7180         * config/rs6000/rs6000-protos.h (fusion_wrap_memory_address):
7181         Delete, no longer used.
7182         * config/rs6000/rs6000.c (struct rs6000_reg_addr): Delete fields
7183         meant to be used for toc fusion.
7184         (rs6000_debug_print_mode): Delete toc fusion debugging.
7185         (rs6000_debug_reg_global): Likewise.
7186         (rs6000_init_hard_regno_mode_ok): Delete setting up fields for toc
7187         fusion and secondary reload support that were never used.
7188         (rs6000_option_override_internal): Delete TOC fusion, that was only
7189         partially defined, and it did not work unless you also used the
7190         -mcmodel= switch.
7191         (rs6000_legitimate_address_p): Delete TOC fusion support.
7192         (rs6000_opt_masks): Likewise.
7193         (fusion_wrap_memory_address): Delete function, no longer used.
7194         (fusion_split_address); Delete TOC fusion support.
7195         * config/rs6000/rs6000.h (TARGET_TOC_FUSION_INT): Delete, no
7196         longer used with toc fusion being deleted.
7197         (TARGET_TOC_FUSION_FP): Likewise.
7198         * config/rs6000/rs6000.md (UNSPEC_FUSION_ADDIS): Delete TOC fusion
7199         UNSPEC.
7200         (toc fusion spliter): Delete TOC fusion support.
7201         (toc_fusionload_<mode>): Likewise.
7202         (toc_fusionload_di): Likewise.
7203         (fusion_gpr_load_<mode>): Delete generator function, this insn no
7204         longer needs to be named.  Rename predicate to delete TOC fusion.
7205         (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_load): Likewise.
7206         (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_store): Likewise.
7207         (fusion_vsx_<P:mode>_<GPR_FUSION:mode>_load): Likewise.
7208         (fusion_vsx_<P:mode>_<GPR_FUSION:mode>_store): Likewise.
7209         (p9 fusion peephole2s): Rename predicate to delete TOC fusion.
7211 2018-07-27  Kelvin Nilsen  <kelvin@gcc.gnu.org>
7213         * doc/extend.texi (Basic PowerPC Built-in Functions Available on
7214         ISA 2.05):  Replace __uint128_t with __uint128 and __int128_t with
7215         __int128 in built-in function prototypes.
7216         (PowerPC AltiVec Built-in Functions on ISA 2.07): Likewise.
7217         (PowerPC AltiVec Built-in Functions on ISA 3.0): Likewise.
7219 2018-07-27  Martin Sebor  <msebor@redhat.com>
7221         PR tree-optimization/86696
7222         * tree-ssa-strlen.c (get_min_string_length): Handle all integer
7223         types, including enums.
7224         (handle_char_store): Be prepared for the above function to fail.
7226 2018-07-26  Qing Zhao  <qing.zhao@oracle.com>
7228         * builtins.c (inline_expand_builtin_string_cmp): Disable inlining
7229         when optimization level is lower than 2 or optimize for size.
7231 2018-07-26  Martin Sebor  <msebor@redhat.com>
7233         PR tree-optimization/86043
7234         PR tree-optimization/86042
7235         * tree-ssa-strlen.c (handle_builtin_memcpy): Handle strict overlaps.
7236         (get_string_cst_length): Rename...
7237         (get_min_string_length): ...to this.  Add argument.
7238         (handle_char_store): Extend to handle multi-character stores by
7239         MEM_REF.
7240         * tree.c (initializer_zerop): Use new argument.  Handle MEM_REF.
7241         * tree.h (initializer_zerop): Add argument.
7243 2018-07-26  Jakub Jelinek  <jakub@redhat.com>
7245         PR middle-end/86660
7246         * omp-low.c (scan_sharing_clauses): Don't ignore map clauses for
7247         declare target to variables if they have always,{to,from,tofrom} map
7248         kinds.
7250 2018-07-26  Martin Liska  <mliska@suse.cz>
7252         PR lto/86548
7253         * lto-wrapper.c: Add linker_output as prefix
7254         for ltrans_output_file.
7256 2018-07-26  Segher Boessenkool  <segher@kernel.crashing.org>
7258         PR rtl-optimization/85805
7259         * combine.c (reg_nonzero_bits_for_combine): Only use the last set
7260         value for hard registers if that was written in the same mode.
7262 2018-07-26  Martin Liska  <mliska@suse.cz>
7264         PR gcov-profile/86536
7265         * gcov.c (format_gcov): Use printf format %.*f directly
7266         and do not handle special values.
7268 2018-07-25  Claudiu Zissulescu  <claziss@synopsys.com>
7270         * common/config/arc/arc-common.c (arc_option_optimization_table):
7271         Update default optimizations for size.
7273 2018-07-25  Claudiu Zissulescu  <claziss@synopsys.com>
7275         * config/arc/arc.md (movsf_insn): Add short instruction selection.
7276         * config/arc/constraints.md (CfZ): New constraint.
7277         * config/arc/fpu.md (addssf3_fpu): Use CfZ constraint.
7278         (subsf3_fpu): Likewise.
7279         (cmpsf_fpu): Likewise.
7280         (cmpsf_fpu_uneq): Likewise.
7282 2018-07-25  Claudiu Zissulescu  <claziss@synopsys.com>
7284         * config/arc/arc.c (compact_memory_operand_p): Check for uncached
7285         accesses as well.
7286         (arc_is_uncached_mem_p): uncached applies to both the variable and
7287         the pointer.
7289 2018-07-25  Claudiu Zissulescu  <claziss@synopsys.com>
7291         * config/arc/arc.h (ADDITIONAL_REGISTER_NAMES): Add additional
7292         register names.
7294 2018-07-25  David Malcolm  <dmalcolm@redhat.com>
7296         * optinfo-emit-json.cc (class optrecord_json_writer): Convert
7297         field "m_scopes" from vec to auto_vec.
7299 2018-07-25  Martin Liska  <mliska@suse.cz>
7301         * config/powerpcspe/powerpcspe-protos.h (rs6000_loop_align): Fix
7302         return type.
7304 2018-07-25  Richard Biener  <rguenther@suse.de>
7306         PR debug/86654
7307         * dwarf2out.c (dwarf2out_decl): Do not handle nested functions
7308         special wrt context_die late.
7309         (gen_subprogram_die): Re-use DIEs in local scope.
7311 2018-07-25  Richard Sandiford  <richard.sandiford@arm.com>
7313         PR tree-optimization/86644
7314         * hwint.c (ceil_log2): Fix comment.  Return 0 for 0.
7316 2018-07-25  Martin Liska  <mliska@suse.cz>
7318         PR middle-end/86645
7319         * dumpfile.c: And excluded values with TDF_ALL_VALUES.
7320         * dumpfile.h (enum dump_flag): Defince TDF_ALL_VALUES.
7322 2018-07-25  Martin Liska  <mliska@suse.cz>
7324         PR sanitizer/79635
7325         * params.def: Explain ASan abbreviation and provide
7326         a documentation link.
7328 2018-07-24  Martin Sebor  <msebor@redhat.com>
7330         PR tree-optimization/86622
7331         PR tree-optimization/86532
7332         * builtins.h (string_length): Declare.
7333         * builtins.c (c_strlen): Correct handling of non-constant offsets.
7334         (check_access): Be prepared for non-constant length ranges.
7335         (string_length): Make extern.
7336         * expr.c (string_constant): Only handle the minor non-constant
7337         array index.  Use string_constant to compute the length of
7338         a generic string constant.
7340 2018-07-24  Richard Sandiford  <richard.sandiford@arm.com>
7342         PR tree-optimization/86618
7343         * tree-vect-stmts.c (vectorizable_call): Don't take the address
7344         of LOOP_VINFO_MASKS (loop_vinfo) when loop_vinfo is null.
7346 2018-07-24  David Malcolm  <dmalcolm@redhat.com>
7348         PR tree-optimization/86636
7349         * json.cc (json::object::set): Fix comment.  Add assertions.
7350         (json::array::append): Move here from json.h.  Add comment and an
7351         assertion.
7352         (json::string::string): Likewise.
7353         * json.h (json::array::append): Move to json.cc.
7354         (json::string::string): Likewise.
7355         * optinfo-emit-json.cc
7356         (optrecord_json_writer::impl_location_to_json): Assert that we
7357         aren't attempting to write out UNKNOWN_LOCATION, or an ad-hoc
7358         wrapper around it.  Expand the location once, rather than three
7359         times.
7360         (optrecord_json_writer::inlining_chain_to_json): Fix the check for
7361         UNKNOWN_LOCATION, to use LOCATION_LOCUS to look through ad-hoc
7362         wrappers.
7363         (optrecord_json_writer::optinfo_to_json): Likewise, in four
7364         places.  Fix some overlong lines.
7366 2018-07-24  Matthew Malcomson  <matthew.malcomson@arm.com>
7368         * config/aarch64/aarch64-simd.md
7369         (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w<mode>): Split into...
7370         (aarch64_<ANY_EXTEND:su>subw<mode>): ... This...
7371         (aarch64_<ANY_EXTEND:su>addw<mode>): ... And this.
7372         (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w<mode>_internal): Split into...
7373         (aarch64_<ANY_EXTEND:su>subw<mode>_internal): ... This...
7374         (aarch64_<ANY_EXTEND:su>addw<mode>_internal): ... And this.
7375         (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w2<mode>_internal): Split into...
7376         (aarch64_<ANY_EXTEND:su>subw2<mode>_internal): ... This...
7377         (aarch64_<ANY_EXTEND:su>addw2<mode>_internal): ... And this.
7379 2018-07-24  Jakub Jelinek  <jakub@redhat.com>
7381         PR middle-end/86627
7382         * expmed.c (expand_divmod): Punt if d == HOST_WIDE_INT_MIN
7383         and size > HOST_BITS_PER_WIDE_INT.  For size > HOST_BITS_PER_WIDE_INT
7384         and abs_d == d, do the power of two handling if profitable.
7386 2018-07-24  Richard Biener  <rguenther@suse.de>
7388         * match.pd: Add BIT_FIELD_REF canonicalizations.
7390 2018-07-23  Bernd Edlinger  <bernd.edlinger@hotmail.de>
7392         PR c/86617
7393         * genmatch.c (dt_operand::gen_match_op): Avoid folding volatile values.
7395 2018-07-23  Bernd Edlinger  <bernd.edlinger@hotmail.de>
7397         * gimple-fold.c (gimple_fold_builtin_printf): Don't create a not NUL
7398         terminated STRING_CST object.
7400 2018-07-23  Bernd Edlinger  <bernd.edlinger@hotmail.de>
7402         hsa-dump.c (dump_hsa_symbol): Avoid out of scope access to buf.
7404 2018-07-23  Segher Boessenkool  <segher@kernel.crashing.org>
7406         * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Adjust.
7407         * config/rs6000/rs6000-protos.h (rs6000_split_v4si_init): Delete.
7408         * config/rs6000/rs6000.c (rs6000_expand_vector_init): Always force
7409         the elements into a register.
7410         (rs6000_split_v4si_init_di_reg): Delete.
7411         (rs6000_split_v4si_init): Delete.
7412         * config/rs6000/vsx.md (unspec): Delete UNSPEC_VSX_VEC_INIT.
7413         (vsx_init_v4si): Rewrite as a define_expand.
7415 2018-07-23  Segher Boessenkool  <segher@kernel.crashing.org>
7417         * config/rs6000/rs6000.md (splitters for rldimi and rlwimi with the
7418         zero_extend argument from memory): New.
7420 2018-07-22  Martin Sebor  <msebor@redhat.com>
7422         PR bootstrap/86621
7423         * gimple-ssa-warn-alloca.c (alloca_call_type_by_arg): Avoid
7424         diagnosing calls with unknown arguments unless -Walloca-larger-than
7425         is restricted to less than PTRDIFF_MAX bytes.
7427 2018-07-22  Gerald Pfeifer  <gerald@pfeifer.com>
7429         * doc/gcov.texi (Invoking Gcov): Editorial changes.
7431 2018-07-20  David Malcolm  <dmalcolm@redhat.com>
7433         * pretty-print.c (text_info::set_location): Remove redundant
7434         "line_table" parameter from call to rich_location::set_range.
7436 2018-07-20  Martin Sebor  <msebor@redhat.com>
7438         PR middle-end/82063
7439         * builtins.c (expand_builtin_alloca): Adjust.
7440         * calls.c (alloc_max_size): Simplify.
7441         * cgraphunit.c (cgraph_node::expand): Adjust.
7442         * common.opt (larger_than_size, warn_frame_larger_than): Remove
7443         variables.
7444         (frame_larger_than_size): Same.
7445         (-Wframe-larger-than, -Wlarger-than, -Wstack-usage): Change options
7446         to take a HOST_WIDE_INT argument and accept a byte-size suffix.
7447         Initialize.
7448         * doc/invoke.texi (GCC Command Options): Document option arguments.
7449         Explain byte-size arguments and suffixes.
7450         (-Wvla-larger-than, -Wno-alloc-size-larger-than): Update.
7451         (-Wno-alloca-larger-than, -Wno-vla-larger-than): Same.
7452         (-Wframe-larger-than, -Wlarger-than, -Wstack-usage): Same.
7453         * doc/options.texi (UInteger): Expand.
7454         (Host_Wide_Int, ByteSize): Document new properties.
7455         * final.c (final_start_function_1): Include sizes in an error message.
7456         * function.c (frame_offset_overflow): Same.
7457         * gimple-ssa-warn-alloca.c (pass_walloca::gate): Adjust.
7458         (alloca_call_type_by_arg): Change function argument to HOST_WIDE_INT.
7459         Diagnose unbounded alloca calls only for limits of less than
7460         PTRDIFF_MAX.
7461         (alloca_call_type): Adjust.  Diagnose possibly out-of-bounds alloca
7462         calls and VLA size only for limits of less than PTRDIFF_MAX.  Same
7463         for alloca(0).
7464         (pass_walloca::execute): Adjust.  Diagnose alloca calls in loops
7465         only for limits of less than PTRDIFF_MAX.
7466         * langhooks-def.h (lhd_handle_option): Change function argument
7467         to HOST_WIDE_INT.
7468         * langhooks.c (lhd_handle_option): Same.
7469         * langhooks.h (handle_option): Same.
7470         * opt-functions.awk (switch_bit_fields): Handle Host_Wide_Int and
7471         ByteSize flags.
7472         (var_type, var_type_struct): Same.
7473         (var_set): Handle ByteSize flag.
7474         * optc-gen.awk: Add comments to output to ease debugging.  Make
7475         use of HOST_WIDE_INT where appropriate.
7476         * opts-gen-save.awk:  Use %lx to format unsigned long.
7477         * opth-gen.awk: Change function argument to HOST_WIDE_INT.
7478         * opts-common.c (integral_argument): Return HOST_WIDE_INT and add
7479         arguments.  Parse bytes-size suffixes.
7480         (enum_arg_to_value): Change function argument to HOST_WIDE_INT.
7481         (enum_value_to_arg): Same.
7482         (decode_cmdline_option): Handle cl_host_wide_int.  Adjust.
7483         (handle_option): Adjust.
7484         (generate_option): Change function argument to HOST_WIDE_INT.
7485         (cmdline_handle_error): Adjust.
7486         (read_cmdline_option): Change function argument to HOST_WIDE_INT.
7487         (set_option): Change function argument to HOST_WIDE_INT.
7488         (option_enabled): Handle cl_host_wide_int.
7489         (get_option_state): Handle CLVC_SIZE.
7490         (control_warning_option): Same.
7491         * opts.c (common_handle_option): Change function argument to
7492         HOST_WIDE_INT.  Remove handling of OPT_Walloca_larger_than_ and
7493         OPT_Wvla_larger_than_.
7494         * opts.h (enum cl_var_type): Add an enumerator.
7495         * stor-layout.c (layout_decl): Print a more meaningful warning.
7496         * toplev.c (output_stack_usage): Adjust.
7498 2018-07-20  Qing Zhao  <qing.zhao@oracle.com>
7500         * builtins.c (expand_builtin_memcmp): Delete the last parameter for
7501         call to inline_expand_builtin_string_cmp.
7502         (expand_builtin_strcmp): Likewise.
7503         (expand_builtin_strncmp): Likewise.
7504         (inline_string_cmp): Delete the last parameter, change char_type_node
7505         to unsigned_char_type_node for strcmp/strncmp, add conversions to the
7506         two operands.
7507         (inline_expand_builtin_string_cmp): Delete the last parameter, give up
7508         the inlining expansion on target where the type of the call has same or
7509         narrower precision than unsigned char.
7511 2018-07-20  David Malcolm  <dmalcolm@redhat.com>
7513         * Makefile.in (OBJS): Add json.o and optinfo-emit-json.o.
7514         (CFLAGS-optinfo-emit-json.o): Define TARGET_NAME.
7515         * common.opt (fsave-optimization-record): New option.
7516         * coretypes.h (struct kv_pair): Move here from dumpfile.c.
7517         * doc/invoke.texi (-fsave-optimization-record): New option.
7518         * dumpfile.c: Include "optinfo-emit-json.h".
7519         (struct kv_pair): Move to coretypes.h.
7520         (optgroup_options): Make non-static.
7521         (dump_context::end_scope): Call
7522         optimization_records_maybe_pop_dump_scope.
7523         * dumpfile.h (optgroup_options): New decl.
7524         * json.cc: New file.
7525         * json.h: New file.
7526         * optinfo-emit-json.cc: New file.
7527         * optinfo-emit-json.h: New file.
7528         * optinfo.cc: Include "optinfo-emit-json.h".
7529         (optinfo::emit): Call optimization_records_maybe_record_optinfo.
7530         (optinfo_enabled_p): Check optimization_records_enabled_p.
7531         (optinfo_wants_inlining_info_p): Likewise.
7532         * optinfo.h: Update comment.
7533         * profile-count.c (profile_quality_as_string): New function.
7534         * profile-count.h (profile_quality_as_string): New decl.
7535         (profile_count::quality): New accessor.
7536         * selftest-run-tests.c (selftest::run_tests): Call json_cc_tests
7537         and optinfo_emit_json_cc_tests.
7538         * selftest.h (selftest::json_cc_tests): New decl.
7539         (selftest::optinfo_emit_json_cc_tests): New decl.
7540         * toplev.c: Include "optinfo-emit-json.h".
7541         (compile_file): Call optimization_records_finish.
7542         (do_compile): Call optimization_records_start.
7543         * tree-ssa-live.c: Include optinfo.h.
7544         (remove_unused_scope_block_p): Retain inlining information if
7545         optinfo_wants_inlining_info_p returns true.
7547 2018-07-20  Richard Biener  <rguenther@suse.de>
7549         PR debug/86585
7550         * dwarf2out.c (dwarf2out_die_ref_for_decl): Test in_lto_p
7551         to cover -flto-partition=none.
7553 2018-07-20  Martin Liska  <mliska@suse.cz>
7555         * tree.h (DECL_LOCATION_RANGE): Remove unused macro.
7556         (get_decl_source_range): Remove unused function.
7558 2018-07-20  Richard Biener  <rguenther@suse.de>
7560         * tree-ssa-sccvn.h (struct vn_nary_op_s): Add next member.
7561         (struct vn_phi_s): Likewise.
7562         (struct vn_reference_s): Likewise.
7563         * tree-ssa-sccvn.c (vn_nary_op_hasher::equal): Add shortcut
7564         for searching the slot of an entry known to be in the hash itself.
7565         (vn_phi_hasher::equal): Likewise.
7566         (vn_reference_hasher::equal): Likewise.
7567         (last_inserted_ref, last_inserted_phi, last_inserted_nary): New
7568         globals.
7569         (optimistic_info, current_info): Remove, keeping only valid_info.
7570         (vn_reference_lookup_1): Remove fallback lookup.
7571         (vn_reference_lookup_2): Likewise.
7572         (vn_nary_op_lookup_1): Likewise.
7573         (vn_phi_lookup): Likewise.
7574         (vn_nary_build_or_lookup_1): Make sure to not chain the built
7575         hash element.
7576         (vn_reference_insert): Adjust, chain the inserted hash element
7577         at last_inserted_ref.
7578         (vn_reference_insert_pieces): Likewise.
7579         (visit_reference_op_call): Likewise.
7580         (vn_nary_op_insert_into): Chain the inserted hash element at
7581         last_inserted_nary.
7582         (vn_nary_op_insert_pieces): Adjust.
7583         (vn_nary_op_insert): Likewise.
7584         (vn_nary_op_insert_stmt): Likewise.
7585         (vn_phi_insert): Adjust, chain the inserted hash element at
7586         last_inserted_phi.
7587         (process_scc): Remove clearing and copying the optimistic
7588         table.  Instead remove elements inserted during an optimistic
7589         iteration from the single table we maintain.
7590         (init_scc_vn): Adjust.
7591         (free_scc_vn): Likewise.
7592         (sccvn_dom_walker::record_cond): Likewise.
7593         (sccvn_dom_walker::after_dom_children): Likewise.
7595 2018-07-19  Martin Sebor  <msebor@redhat.com>
7597         PR tree-optimization/84047
7598         PR tree-optimization/83776
7599         * tree-vrp.c (vrp_prop::check_mem_ref): New function.
7600         (check_array_bounds): Call it.
7602 2018-07-19  Martin Sebor  <msebor@redhat.com>
7604         * align.h (align_flags): Use member initialization.
7606 2018-07-19  David Malcolm  <dmalcolm@redhat.com>
7608         * Makefile.in (OBJS): Add optinfo.o.
7609         * coretypes.h (class symtab_node): New forward decl.
7610         (struct cgraph_node): New forward decl.
7611         (class varpool_node): New forward decl.
7612         * dump-context.h: New file.
7613         * dumpfile.c: Include "optinfo.h", "dump-context.h", "cgraph.h",
7614         "tree-pass.h".
7615         (refresh_dumps_are_enabled): Use optinfo_enabled_p.
7616         (set_dump_file): Call dumpfile_ensure_any_optinfo_are_flushed.
7617         (set_alt_dump_file): Likewise.
7618         (dump_context::~dump_context): New dtor.
7619         (dump_gimple_stmt): Move implementation to...
7620         (dump_context::dump_gimple_stmt): ...this new member function.
7621         Add the stmt to any pending optinfo, creating one if need be.
7622         (dump_gimple_stmt_loc): Move implementation to...
7623         (dump_context::dump_gimple_stmt_loc): ...this new member function.
7624         Start a new optinfo and add the stmt to it.
7625         (dump_gimple_expr): Move implementation to...
7626         (dump_context::dump_gimple_expr): ...this new member function.
7627         Add the stmt to any pending optinfo, creating one if need be.
7628         (dump_gimple_expr_loc): Move implementation to...
7629         (dump_context::dump_gimple_expr_loc): ...this new member function.
7630         Start a new optinfo and add the stmt to it.
7631         (dump_generic_expr): Move implementation to...
7632         (dump_context::dump_generic_expr): ...this new member function.
7633         Add the tree to any pending optinfo, creating one if need be.
7634         (dump_generic_expr_loc): Move implementation to...
7635         (dump_context::dump_generic_expr_loc): ...this new member
7636         function.  Add the tree to any pending optinfo, creating one if
7637         need be.
7638         (dump_printf): Move implementation to...
7639         (dump_context::dump_printf_va): ...this new member function.  Add
7640         the text to any pending optinfo, creating one if need be.
7641         (dump_printf_loc): Move implementation to...
7642         (dump_context::dump_printf_loc_va): ...this new member function.
7643         Start a new optinfo and add the stmt to it.
7644         (dump_dec): Move implementation to...
7645         (dump_context::dump_dec): ...this new member function.  Add the
7646         value to any pending optinfo, creating one if need be.
7647         (dump_context::dump_symtab_node): New member function.
7648         (dump_context::get_scope_depth): New member function.
7649         (dump_context::begin_scope): New member function.
7650         (dump_context::end_scope): New member function.
7651         (dump_context::ensure_pending_optinfo): New member function.
7652         (dump_context::begin_next_optinfo): New member function.
7653         (dump_context::end_any_optinfo): New member function.
7654         (dump_context::s_current): New global.
7655         (dump_context::s_default): New global.
7656         (dump_scope_depth): Delete global.
7657         (dumpfile_ensure_any_optinfo_are_flushed): New function.
7658         (dump_symtab_node): New function.
7659         (get_dump_scope_depth): Reimplement in terms of dump_context.
7660         (dump_begin_scope): Likewise.
7661         (dump_end_scope): Likewise.
7662         (selftest::temp_dump_context::temp_dump_context): New ctor.
7663         (selftest::temp_dump_context::~temp_dump_context): New dtor.
7664         (selftest::verify_item): New function.
7665         (ASSERT_IS_TEXT): New macro.
7666         (ASSERT_IS_TREE): New macro.
7667         (ASSERT_IS_GIMPLE): New macro.
7668         (selftest::test_capture_of_dump_calls): New test.
7669         (selftest::dumpfile_c_tests): Call it.
7670         * dumpfile.h (dump_printf, dump_printf_loc, dump_basic_block)
7671         (dump_generic_expr_loc, dump_generic_expr, dump_gimple_stmt_loc)
7672         (dump_gimple_stmt, dump_dec): Gather these related decls and add a
7673         descriptive comment.
7674         (dump_function, print_combine_total_stats, enable_rtl_dump_file)
7675         (dump_node, dump_bb): Move these unrelated decls.
7676         (class dump_manager): Add leading comment.
7677         * optinfo.cc: New file.
7678         * optinfo.h: New file.
7680 2018-07-19  Michael Collison  <michael.collison@arm.com>
7681             Richard Henderson <rth@redhat.com>
7683         * config/aarch64/aarch64.md (subv<GPI>4, usubv<GPI>4): New patterns.
7684         (subti): Handle op1 zero.
7685         (subvti4, usub4ti4): New.
7686         (*sub<GPI>3_compare1_imm): New.
7687         (sub<GPI>3_carryinCV): New.
7688         (*sub<GPI>3_carryinCV_z1_z2, *sub<GPI>3_carryinCV_z1): New.
7689         (*sub<GPI>3_carryinCV_z2, *sub<GPI>3_carryinCV): New.
7691 2018-07-19  Michael Collison  <michael.collison@arm.com>
7692             Richard Henderson <rth@redhat.com>
7694         * config/aarch64/aarch64.md: (addv<GPI>4, uaddv<GPI>4): New.
7695         (addti3): Create simpler code if low part is already known to be 0.
7696         (addvti4, uaddvti4): New.
7697         (*add<GPI>3_compareC_cconly_imm): New.
7698         (*add<GPI>3_compareC_cconly): New.
7699         (*add<GPI>3_compareC_imm): New.
7700         (*add<GPI>3_compareC): Rename from add<GPI>3_compare1; do not
7701         handle constants within this pattern..
7702         (*add<GPI>3_compareV_cconly_imm): New.
7703         (*add<GPI>3_compareV_cconly): New.
7704         (*add<GPI>3_compareV_imm): New.
7705         (add<GPI>3_compareV): New.
7706         (add<GPI>3_carryinC, add<GPI>3_carryinV): New.
7707         (*add<GPI>3_carryinC_zero, *add<GPI>3_carryinV_zero): New.
7708         (*add<GPI>3_carryinC, *add<GPI>3_carryinV): New.
7709         ((*add<GPI>3_compareC_cconly_imm): Replace 'ne' operator
7710         with 'comparison' operator.
7711         (*add<GPI>3_compareV_cconly_imm): Ditto.
7712         (*add<GPI>3_compareV_cconly): Ditto.
7713         (*add<GPI>3_compareV_imm): Ditto.
7714         (add<GPI>3_compareV): Ditto.
7715         (add<mode>3_carryinC): Ditto.
7716         (*add<mode>3_carryinC_zero): Ditto.
7717         (*add<mode>3_carryinC): Ditto.
7718         (add<mode>3_carryinV): Ditto.
7719         (*add<mode>3_carryinV_zero): Ditto.
7720         (*add<mode>3_carryinV): Ditto.
7722 2018-07-19  Michael Collison  <michael.collison@arm.com>
7723             Richard Henderson <rth@redhat.com>
7725         * config/aarch64/aarch64-modes.def (CC_V): New.
7726         * config/aarch64/aarch64-protos.h
7727         (aarch64_addti_scratch_regs): Declare
7728         (aarch64_subvti_scratch_regs): Declare.
7729         (aarch64_expand_subvti): Declare.
7730         (aarch64_gen_unlikely_cbranch): Declare
7731         * config/aarch64/aarch64.c (aarch64_select_cc_mode): Test
7732         for signed overflow using CC_Vmode.
7733         (aarch64_get_condition_code_1): Handle CC_Vmode.
7734         (aarch64_gen_unlikely_cbranch): New function.
7735         (aarch64_addti_scratch_regs): New function.
7736         (aarch64_subvti_scratch_regs): New function.
7737         (aarch64_expand_subvti): New function.
7739 2018-07-19  Andre Vieira  <andre.simoesdiasvieira@arm.com>
7741         * config/aarch64/aarch64-option-extensions.def: New entry for profile
7742         extension.
7743         * config/aarch64/aarch64.h (AARCH64_FL_PROFILE): New.
7744         * doc/invoke.texi (aarch64-feature-modifiers): New entry for profile
7745         extension.
7747 2018-07-19  Andre Vieira  <andre.simoesdiasvieira@arm.com>
7749         PR target/83009
7750         * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
7751         address check not strict.
7753 2018-07-19  Andre Vieira  <andre.simoesdiasvieira@arm.com>
7755         * config/aarch64/aarch64-simd.md (aarch64_simd_mov<VQ:mode>): Replace
7756         Umq with Umn.
7757         (store_pair_lanes<mode>): Likewise.
7758         * config/aarch64/aarch64-protos.h (aarch64_addr_query_type): Add new
7759         enum value 'ADDR_QUERY_LDP_STP_N'.
7760         * config/aarch64/aarch64.c (aarch64_addr_query_type): Likewise.
7761         (aarch64_print_address_internal): Add declaration.
7762         (aarch64_print_ldpstp_address): Remove.
7763         (aarch64_classify_address): Adapt mode for 'ADDR_QUERY_LDP_STP_N'.
7764         (aarch64_print_operand): Change printing of 'y'.
7765         * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Use
7766         new enum value 'ADDR_QUERY_LDP_STP_N', don't hardcode mode and use
7767         'true' rather than '1'.
7768         * config/aarch64/constraints.md (Uml): Likewise.
7769         (Uml): Rename to Umn.
7770         (Umq): Remove.
7772 2018-07-19  Richard Biener  <rguenther@suse.de>
7774         * tree-ssa-sccvn.h (struct vn_phi_s): Make phiargs member
7775         a trailing array.
7776         * tree-ssa-sccvn.c: Remove alloc-pool.h use.
7777         (vn_phi_hasher): Derive from nofree_ptr_hash and remove remove method.
7778         (vn_reference_hasher): Likewise.
7779         (struct vn_tables_s): Remove obstack and alloc-pool members.
7780         (vn_tables_obstack, vn_tables_insert_obstack): New global obstacks.
7781         (vn_nary_build_or_lookup_1): Manually build in vn_tables_insert_obstack.
7782         (vn_reference_insert): Allocate from obstack instead of from alloc-pool.
7783         (vn_reference_insert_pieces): Likewise.
7784         (alloc_vn_nary_op_noinit): Adjust.
7785         (vn_nary_op_insert_stmt): Allocate phiargs in-place.
7786         (vn_phi_eq): Adjust.
7787         (shared_lookup_phiargs): Remove.
7788         (vn_phi_lookup): Allocate temporary vn_phi_s on the stack.
7789         (vn_phi_insert): Allocate from obstack instead of from alloc-pool.
7790         (visit_reference_op_call): Likewise.
7791         (copy_nary, copy_phi, copy_reference): Remove.
7792         (process_scc): Rewind the obstack when iterating.  Do not
7793         copy the elements to valid_info but just move them from one
7794         hashtable to the other.
7795         (allocate_vn_table): Adjust.
7796         (free_vn_table): Likewise.
7797         (init_scc_vn): Likewise.
7798         (free_scc_vn): Likewise.
7800 2018-07-19  H.J. Lu  <hongjiu.lu@intel.com>
7802         PR target/86560
7803         * config/i386/i386.c (rest_of_insert_endbranch): Lookup
7804         indirect_return as function type attribute.
7805         (ix86_attribute_table): Change indirect_return to function
7806         type attribute.
7807         * doc/extend.texi: Update indirect_return attribute.
7809 2018-07-19  Aldy Hernandez  <aldyh@redhat.com>
7811         * wide-int.h (widest2_int): New.
7812         * gimple-fold.c (arith_overflowed_p): Use it.
7813         * tree.h (widest2_int_cst): New.
7814         * tree-vrp.c (wide_int_binop_overflow): Rename from
7815         vrp_int_const_binop.
7816         Rewrite to work on trees.
7817         (extract_range_from_multiplicative_op_1): Abstract code to...
7818         (wide_int_range_min_max): ...here.
7819         (wide_int_range_cross_product): ...and here.
7820         (extract_range_from_binary_expr_1): Abstract overflow code to...
7821         (wide_int_range_mult_wrapping): ...here.
7822         * tree-vrp.h (wide_int_range_cross_product): New.
7823         (wide_int_range_mult_wrapping): New.
7825 2018-07-19  Andrew Senkevich  <andrew.senkevich@intel.com>
7826             Julia Koval  <julia.koval@intel.com>
7828         * config/i386/x86-tune-costs.h (skylake_memcpy,
7829         skylake_memset): Replace rep_prefix with unrolling for size 512.
7831 2018-07-18  Kugan Vivekanandarajah  <kuganv@linaro.org>
7833         PR middle-end/86544
7834         * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): Handle
7835         comparision with EQ_EXPR in last stmt.
7837 2018-07-18  Kelvin Nilsen  <kelvin@gcc.gnu.org>
7839         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Rename
7840         this subsection to "PowerPC AltiVec/VSX Built-in Functions".
7841         (PowerPC AltiVec/VSX Built-in Functions): New name for subsection
7842         previously known as "PowerPC AltiVec Built-in Functions".  Move
7843         some material to new subsubsections "PowerPC AltiVec Built-in
7844         Functions on ISA 2.06" and "PowerPC AltiVec Built-in Functions on
7845         ISA 2.07".
7846         (PowerPC Altivec Built-in Functions on ISA 2.05): New subsubsection.
7847         (PowerPC Altivec Built-in Functions on ISA 2.06): Likewise.
7848         (PowerPC Altivec Built-in Functions on ISA 2.07): Likewise.
7849         (PowerPC Altivec Built-in Functions on ISA 3.0): Likewise.
7851 2018-07-18  Richard Biener  <rguenther@suse.de>
7853         PR tree-optimization/86557
7854         * tree-vect-patterns.c (vect_recog_divmod_pattern): Also handle
7855         EXACT_DIV_EXPR.
7857 2018-07-18  Ilya Leoshkevich  <iii@linux.ibm.com>
7859         * config/s390/s390.c (s390_function_profiler): Generate CFI.
7861 2018-07-17  Jeff Law  <law@redhat.com>
7863         * config/arm/arm.c (get_label_padding): Update for recent
7864         changes to label_to_alignment.
7866         PR tree-optimization/86010
7867         * tree-ssa-dse.c (compute_trims): Fix typo/thinko.
7869         * config/mips/mips.c (vr4130_align_insns): Update for recent
7870         changes to label_to_alignment.
7872         * config/frv/frv.c (frv_label_align): Update for recent changes
7873         to label_to_alignment.
7875         * config/nios2/nios2.c (nios2_label_align): Update for recent
7876         changes which dropped ALIGN_LABELS_LOG.
7878 2018-07-17  Andreas Schwab  <schwab@linux-m68k.org>
7880         * config/m68k/m68k.md (umulsi3_highpart+1, const_umulsi3_highpart)
7881         (smulsi3_highpart+1, const_smulsi3_highpart): Add CC_STATUS_INIT.
7883 2018-07-17  Claudiu Zissulescu  <claziss@synopsys.com>
7885         * config/arc/arc.c (arc_label_align): Use align_labels instead of
7886         deprecated align_labels_log.
7888 2018-07-17  Richard Biener  <rguenther@suse.de>
7890         PR lto/86456
7891         * dwarf2out.c (init_sections_and_labels): Always generate
7892         a debug_line_str_section for early LTO debug.
7893         (dwarf2out_finish): Reset debug_line_str_hash output early.
7894         Bump counter for extra dwarf5 .debug_loc labels to not conflict
7895         with fat LTO part.
7896         (dwarf2out_early_finish): Output debug_line_str.
7898 2018-07-17  Robin Dapp  <rdapp@linux.ibm.com>
7900         * config/s390/s390.c (preferred_la_operand_p): Do not use LA with
7901         index register on z196 or later.
7903 2018-07-17  Robin Dapp  <rdapp@linux.ibm.com>
7905         * config/s390/s390.c (s390_default_align): Set default function
7906         alignment to 16.
7907         (s390_override_options_after_change): Call s390_default align.
7908         (s390_option_override_internal): Call s390_default_align.
7909         (TARGET_OVERRIDE_OPTIONS_AFTER_CHANGE): Define.
7911 2018-07-17  Jakub Jelinek  <jakub@redhat.com>
7913         PR middle-end/86542
7914         * omp-low.c (create_task_copyfn): Copy over also fields corresponding
7915         to _looptemp_ clauses, other than the first two.
7917 2018-07-17  Martin Liska  <mliska@suse.cz>
7919         * opts.c: Do not enable OPT_falign_* for -Os.
7921 2018-07-17  Martin Liska  <mliska@suse.cz>
7923         * align.h (MAX_CODE_ALIGN): New.
7924         (MAX_CODE_ALIGN_VALUE): New.
7925         * common/config/i386/i386-common.c (ix86_handle_option):
7926         (MAX_CODE_ALIGN): Moved to align.h.
7927         * final.c (MAX_CODE_ALIGN): Likewise.
7928         * opts.c (parse_and_check_align_values):
7929         (MAX_CODE_ALIGN): Likewise.
7930         (MAX_CODE_ALIGN_VALUE): Likewise.
7932 2018-07-17  Martin Liska  <mliska@suse.cz>
7934         * config/i386/att.h (ASM_OUTPUT_ALIGN): Fix spacing
7935         in order to fulfil coding style.
7936         * config/i386/cygming.h (ASM_OUTPUT_ALIGN): Likewise.
7937         * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
7938         * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
7939         * config/iq2000/iq2000.h (ASM_OUTPUT_ALIGN): Likewise.
7940         * config/pa/pa.h (ASM_OUTPUT_ALIGN): Likewise.
7941         * config/sparc/sol2.h (ASM_OUTPUT_ALIGN_WITH_NOP): Likewise.
7942         * config/sparc/sparc.h (ASM_OUTPUT_ALIGN): Likewise.
7943         * config/visium/visium.h (ASM_OUTPUT_ALIGN): Likewise.
7944         (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
7946 2018-07-17  Martin Liska  <mliska@suse.cz>
7948         * align.h: New file.
7949         * config/alpha/alpha.c (alpha_align_insns_1): Use align_functions
7950         directly.
7951         * config/i386/i386.c (ix86_avoid_jump_mispredicts): Use new return type
7952         align_flags of label_to_alignment.
7953         * config/m32r/m32r.h (LOOP_ALIGN): Wrap returned values into
7954         align_flags class.
7955         * config/m68k/m68k.c: Do not use removed align_labels_value and
7956         align_loops_value.
7957         * config/nds32/nds32.h (JUMP_ALIGN): Wrap result into align_flags class.
7958         (LOOP_ALIGN): Likewise.
7959         (LABEL_ALIGN): Likewise.
7960         * config/powerpcspe/powerpcspe.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
7961         Remove not used macro.
7962         (rs6000_loop_align): Change return type to align_flags.
7963         (rs6000_loop_align_max_skip): Remove.
7964         * config/rs6000/rs6000-protos.h (rs6000_loop_align):
7965         Change return type to align_flags.
7966         * config/rs6000/rs6000.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
7967         Remove not used macro.
7968         (rs6000_loop_align):  Change return type to align_flags.
7969         (rs6000_loop_align_max_skip): Remove.
7970         * config/rx/rx.h (JUMP_ALIGN): Wrap integer values
7971         * config/rx/rx-protos.h (rx_align_for_label): Make it
7972         static function.
7973         * config/rx/rx.c (rx_align_for_label): Change return type
7974         to align_flags.
7975         (rx_max_skip_for_label): Remove TARGET_ASM_*_ALIGN_MAX_SKIP
7976         macro definitions.
7977         into align_flags class.
7978         (LABEL_ALIGN): Likewise.
7979         (LOOP_ALIGN): Likewise.
7980         * config/s390/s390.c (s390_label_align): Use align_flags
7981         class member.
7982         (s390_asm_output_function_label): Likewise.
7983         * config/sh/sh.c (sh_override_options_after_change):
7984         Use align_flags class directly without macros.
7985         (find_barrier): Likewise.
7986         (barrier_align): Likewise.
7987         (sh_loop_align): Likewise.
7988         * config/spu/spu.c (spu_option_override):
7989         Use align_flags_tuple::get_value instead of removed macros.
7990         (spu_sched_init): Likewise.
7991         * config/spu/spu.h (GTY): Likewise.
7992         * config/visium/visium.c (visium_option_override):
7993         Set "8" as default secondary alignment.
7994         * config/visium/visium.h (SUBALIGN_LOG): Define to 3
7995         in order to guarantee secondary alignment of 8.
7996         * coretypes.h: Include align.h header file.
7997         * doc/tm.texi: Remove TARGET_ASM_JUMP_ALIGN_MAX_SKIP,
7998         TARGET_ASM_LOOP_ALIGN_MAX_SKIP, TARGET_ASM_LABEL_ALIGN_MAX_SKIP
7999         and TARGET_ASM_LABEL_ALIGN_AFTER_BARRIER_MAX_SKIP macros.
8000         * doc/tm.texi.in: Likewise.
8001         * final.c (struct label_alignment): Remove not used structure.
8002         (LABEL_ALIGN): Change type to align_flags.
8003         (LOOP_ALIGN): Likewise.
8004         (JUMP_ALIGN): Likewise.
8005         (default_loop_align_max_skip): Remove.
8006         (default_label_align_max_skip): Likewise.
8007         (default_jump_align_max_skip): Likewise.
8008         (default_label_align_after_barrier_max_skip):
8009         (LABEL_TO_ALIGNMENT): Change to access label_align vector.
8010         (LABEL_TO_MAX_SKIP): Remove.
8011         (label_to_alignment): Return align_flags type instead of integer.
8012         (label_to_max_skip): Remove.
8013         (align_fuzz): Use align_flags type.
8014         (compute_alignments): Use align_flags type and use align_flags::max
8015         to combine multiple alignments.
8016         (grow_label_align): Grow vec instead of C array.
8017         (update_alignments): Assign just LABEL_TO_ALIGNMENT.
8018         (shorten_branches):  Use align_flags type and use align_flags::max
8019         to combine multiple alignments.
8020         (final_scan_insn_1): Remove usage of secondary alignment that comes
8021         from label alignment, but instead use proper secondary alignment
8022         which is computed in grow_label_align.
8023         * flags.h (struct align_flags_tuple): Move to align.h.
8024         (struct align_flags): Likewise.
8025         (state_align_loops): Rename to align_loops.
8026         (state_align_jumps): Rename to align_jumps.
8027         (state_align_labels): Rename to align_labels.
8028         (state_align_functions): Rename to align_functions.
8029         (align_loops_log): Remove.
8030         (align_jumps_log): Remove.
8031         (align_labels_log): Remove.
8032         (align_functions_log): Remove.
8033         (align_loops_max_skip): Remove.
8034         (align_jumps_max_skip): Remove.
8035         (align_labels_max_skip): Remove.
8036         (align_functions_max_skip): Remove.
8037         (align_loops_value): Remove.
8038         (align_jumps_value): Remove.
8039         (align_labels_value): Remove.
8040         (align_functions_value): Remove.
8041         * output.h (label_to_alignment): Change return type to align_flags.
8042         (label_to_max_skip): Remove.
8043         * target.def: Remove loop_align_max_skip, label_align_max_skip,
8044         jump_align_max_skip macros.
8045         * targhooks.h (default_loop_align_max_skip): Remove.
8046         (default_label_align_max_skip): Likewise.
8047         (default_jump_align_max_skip): Likewise.
8048         (default_label_align_after_barrier_max_skip): Remove.
8049         * toplev.c (read_log_maxskip): Use ::normalize function.
8050         (parse_N_M): Remove not used argument and also call ::normalize.
8051         (parse_alignment_opts): Do not pass unused arguments.
8052         * varasm.c (assemble_start_function): Use directly align_functions
8053         instead of removed macros.
8054         * system.h: Do not poison removed macros.
8056 2018-07-17  Jakub Jelinek  <jakub@redhat.com>
8058         PR middle-end/86539
8059         * gimplify.c (gimplify_omp_for): Ensure taskloop firstprivatized init
8060         and cond temporaries don't have reference type if iterator has
8061         pointer type.  For init use &for_pre_body instead of pre_p if
8062         for_pre_body is non-empty.
8064 2018-07-16  Segher Boessenkool  <segher@kernel.crashing.org>
8066         * config/rs6000/rs6000.md (trunc<mode>sf2): Expand truncates of
8067         double-double modes to SFmode directly directly.
8068         (trunc<mode>sf2_fprs): Delete.
8070 2018-07-16  Segher Boessenkool  <segher@kernel.crashing.org>
8072         * config/rs6000/rs6000.c (init_float128_ibm): Use the correct names
8073         for conversions between IFmode and the decimal floating point modes.
8074         (init_float128_ieee): Use the correct names for conversions between
8075         KFmode and the decimal floating point modes.
8077 2018-07-16  Segher Boessenkool  <segher@kernel.crashing.org>
8079         * config/rs6000/rs6000.c (init_float128_ibm): Use more correct names
8080         for the conversions between TDmode and IFmode.
8081         (init_float128_ieee): Use more correct names for the conversions
8082         between TDmode and KFmode.
8084 2018-07-16  Jakub Jelinek  <jakub@redhat.com>
8086         PR tree-optimization/86526
8087         * builtins.c (expand_builtin_memcmp): Formatting fixes.
8088         (inline_expand_builtin_string_cmp): Likewise.
8089         (inline_string_cmp): Likewise.  Use c_readstr instead of
8090         builtin_memcpy_read_str.  Add unit_mode temporary.
8092 2018-07-16  Bernd Edlinger  <bernd.edlinger@hotmail.de>
8094         PR middle-end/86528
8095         * builtins.c (check_access): Bail out if range[0] is no INTEGER_CST.
8096         * expr.c (string_constant): Fix the element size of ARRAY_TYPE.
8098 2018-07-16  Kelvin Nilsen  <kelvin@gcc.gnu.org>
8100         * doc/extend.texi (PowerPC AltiVec Built-in Functions):
8101         Alphabetize prototypes of built-in functions, separating out
8102         built-in functions that are listed in this section but should be
8103         described elsewhere.
8105 2018-07-16  Uros Bizjak  <ubizjak@gmail.com>
8107         PR target/86511
8108         * expmed.c (emit_store_flag): Do not emit setcc followed by a
8109         conditional move when trapping comparison was split to a
8110         non-trapping one (and vice versa).
8112 2018-07-16  Ilya Leoshkevich  <iii@linux.ibm.com>
8114         * config/s390/s390.c (s390_function_profiler): Generate nops
8115         instead of profiler call sequences.
8116         * config/s390/s390.opt: Add the new option.
8118 2018-07-16  Ilya Leoshkevich  <iii@linux.ibm.com>
8120         * config/s390/s390.c (s390_function_profiler): Generate
8121         __mcount_loc section.
8122         * config/s390/s390.opt: Add the new option.
8124 2018-07-16  Ilya Leoshkevich  <iii@linux.ibm.com>
8126         * common.opt: Add the new warning.
8127         * config/s390/s390.c (s390_function_profiler): Emit "brasl
8128         %r0,__fentry__" when -mfentry is specified.
8129         (s390_option_override_internal): Disallow -mfentry for 31-bit
8130         CPUs.
8131         * config/s390/s390.opt: Add the new option.
8133 2018-07-16  Richard Biener  <rguenther@suse.de>
8135         PR lto/86523
8136         * dwarf2out.c (dwarf2out_register_external_die): Assign DIE parents
8137         for function-local FUNCTION_DECL and RESULT_DECL immediately.
8139 2018-07-16  Martin Liska  <mliska@suse.cz>
8141         PR ipa/86529
8142         * ipa-pure-const.c (malloc_candidate_p): Revert ::get
8143         to ::get_create.
8145 2017-07-16  Claudiu Zissulescu  <claziss@synopsys.com>
8147         * config/arc/arcHS.md: Update ARCHS scheduling rules.
8149 2017-07-16  Claudiu Zissulescu  <claziss@synopsys.com>
8151         * config/arc/arc-arch.h (arc_tune_attr): Add new tune parameters
8152         for ARCHS4x.
8153         * config/arc/arc-cpus.def (hs4x): New cpu.
8154         (hs4xd): Likewise.
8155         * config/arc/arc-tables.opt: Regenerate.
8156         * config/arc/arc.c (arc_sched_issue_rate): New function.
8157         (TARGET_SCHED_ISSUE_RATE): Define.
8158         (TARGET_SCHED_EXPOSED_PIPELINE): Likewise.
8159         * config/arc/arc.md (attr type): Add fpu_fuse, fpu_sdiv, fpu_ddiv,
8160         fpu_cvt.
8161         (attr tune): Add ARCHS4x tune values.
8162         (attr tune_dspmpy): Define.
8163         (*tst): Correct instruction type.
8164         * config/arc/arcHS.md: Don't use this automaton for ARCHS4x cpus.
8165         * config/arc/arcHS4x.md: New file.
8166         * config/arc/fpu.md: Update instruction type attributes.
8167         * config/arc/t-multilib: Regenerate.
8169 2018-07-16  Tom de Vries  <tdevries@suse.de>
8171         PR debug/86455
8172         * var-tracking.c (vt_initialize): Fix pre_dec handling.
8174 2018-07-16  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
8176         * config/aarch64/atomics.md (aarch64_store_execlusive<mode>): Add
8177         early clobber.
8179 2018-07-16  Eric Botcazou  <ebotcazou@adacore.com>
8181         PR tree-optimization/86514
8182         * tree-ssa-reassoc.c (init_range_entry) <CASE_CONVERT>: Return for a
8183         conversion to a boolean type from a type with greater precision.
8185 2018-07-16  Tom de Vries  <tdevries@suse.de>
8187         * var-tracking.c (vt_initialize): Print adjusted insn slim if
8188         dump_flags request TDF_SLIM.
8190 2018-07-16  Aldy Hernandez  <aldyh@redhat.com>
8192         * fold-const.c (int_const_binop_1): Abstract...
8193         (wide_int_binop): ...wide int code here.
8194         (poly_int_binop): ...poly int code here.
8195         Abstract the rest of int_const_binop_1 into int_const_binop.
8196         * fold-const.h (wide_int_binop): New.
8197         * tree-vrp.c (vrp_int_const_binop): Call wide_int_binop.
8198         Remove useless PLUS/MINUS_EXPR case.
8199         (zero_nonzero_bits_from_vr): Move wide int code...
8200         (zero_nonzero_bits_from_bounds): ...here.
8201         (extract_range_from_binary_expr_1): Move mask optimization code...
8202         (range_easy_mask_min_max): ...here.
8203         * tree-vrp.h (zero_nonzero_bits_from_bounds): New.
8204         (range_easy_mask_min_max): New.
8206 2018-07-15  Jeff Law  <law@redhat.com>
8208         PR target/85993
8209         * config/sh/sh.c (output_mi_thunk): Remove dead conditional
8210         block.
8212 2018-07-14  Jim Wilson  <jimw@sifive.com>
8214         * config/riscv/linux.h (TARGET_ASM_FILE_END): New.
8216 2018-07-14  Paul Koning  <ni1d@arrl.net>
8218         * config/pdp11/pdp11.c (pdp11_rtx_costs): Bugfixes.
8220 2018-07-13  Jan Hubicka  <hubicka@ucw.cz>
8222         * lto-streamer-out.c (copy_function_or_variable): Dump info about
8223         copying section.
8225 2018-07-13  Bill Schmidt  <wschmidt@linux.ibm.com>
8226             Steve Munroe  <munroesj52@gmail.com>
8228         * config/rs6000/emmintrin.h (_mm_and_si128): New function.
8229         (_mm_andnot_si128): Likewise.
8230         (_mm_or_si128): Likewise.
8231         (_mm_xor_si128): Likewise.
8233 2018-07-13  Qing Zhao  <qing.zhao@oracle.com>
8235         PR middle-end/78809
8236         * builtins.c (expand_builtin_memcmp): Inline the calls first
8237         when result_eq is false.
8238         (expand_builtin_strcmp): Inline the calls first.
8239         (expand_builtin_strncmp): Likewise.
8240         (inline_string_cmp): New routine. Expand a string compare
8241         call by using a sequence of char comparison.
8242         (inline_expand_builtin_string_cmp): New routine. Inline expansion
8243         a call to str(n)cmp/memcmp.
8244         * doc/invoke.texi (--param builtin-string-cmp-inline-length):
8245         New option.
8246         * params.def (BUILTIN_STRING_CMP_INLINE_LENGTH): New.
8248 2018-07-13  Richard Earnshaw  <rearnsha@arm.com>
8250         * config/arm/driver-arm.c: Include arm-native.h.
8251         (host_detect_local_cpu): Use auto-generated data tables.
8252         (vendors, arm_cpu_table): Delete.  Move part information to ...
8253         * config/arm/arm-cpus.in: ... here.
8254         * config/arm/parsecpu.awk (gen_native): New function.
8255         (vendor, part): New CPU fields.
8256         (END): Add support for building the native CPU detection tables.
8257         * config/arm/t-arm (arm-native.h): Add build rule.
8258         (driver-arm.o): Add dependency on arm-native.h.
8260 2018-07-13  Richard Biener  <rguenther@suse.de>
8262         PR middle-end/85974
8263         * match.pd (addr1 - addr2): Allow either of the operand to
8264         have a conversion.
8266 2018-07-13  Tom de Vries  <tdevries@suse.de>
8268         * tree-inline.c (remap_ssa_name): Save and reuse debug exprs generated
8269         in remap_ssa_name.
8271 2018-07-13  Jackson Woodruff  <jackson.woodruff@arm.com>
8273         * config/aarch64/aarch64.c (aarch64_operands_adjust_ok_for_ldpstp): Use
8274         arrays instead of numbered variables.
8276 2018-07-13  Eric Botcazou  <ebotcazou@adacore.com>
8278         * config/sparc/sparc-protos.h (sparc_compute_frame_size): Delete.
8279         * config/sparc/sparc.c (sparc_compute_frame_size): Make static.
8281 2018-07-13  Richard Biener  <rguenther@suse.de>
8283         PR debug/86452
8284         * dwarf2out.c (gen_type_die_with_usage): Use scope_die_for
8285         instead of get_context_die.
8287 2018-07-13  Kugan Vivekanandarajah  <kuganv@linaro.org>
8288             Richard Biener  <rguenther@suse.de>
8290         PR middle-end/86489
8291         * tree-ssa-loop-niter.c (number_of_iterations_popcount): Check
8292         that the loop latch destination where phi is defined.
8294 2018-07-12  Kito Cheng  <kito.cheng@gmail.com>
8296         * config/riscv/riscv.c (enum riscv_privilege_levels): Add UNKNOWN_MODE.
8297         (riscv_expand_epilogue): Add assertion to check interrupt mode.
8298         (riscv_set_current_function): Extract getting interrupt type to new
8299         function.
8300         (riscv_get_interrupt_type): New function.
8301         (riscv_merge_decl_attributes): New function, checking interrupt type is
8302         same.
8303         (TARGET_MERGE_DECL_ATTRIBUTES): Define.
8305 2018-07-12  Paul Koning  <ni1d@arrl.net>
8307         * config/pdp11/pdp11.c (pdp11_output_def): Fix typo in .set
8308         directive.
8310 2018-07-12  Paul Koning  <ni1d@arrl.net>
8312         * doc/rtl.texi (REG_NONNEG): Remove decrement and branch until
8313         zero reference, add doloop_end instead.
8314         * doc/md.texi (decrement_and_branch_until_zero): Remove.
8315         (Looping patterns): Remove decrement_and_branch_until_zero.  Add
8316         detail for doloop_end.
8318 2018-07-12  Martin Sebor  <msebor@redhat.com>
8320         PR c/86453
8321         * attribs.c (decl_attributes): Reject conflicting attributes before
8322         calling attribute handlers.
8324 2018-07-12  Jan Hubicka  <hubicka@ucw.cz>
8326         * dumpfile.c (gcc::dump_manager::get_dump_file_name): Add PART
8327          parameter.
8328         (gcc::dump_manager::get_dump_file_name): likewise.
8329         (dump_begin): Likewise.
8330         * dumpfile.h (dump_begin): Update prototype.
8331         (gcc::dump_manager::get_dump_file_name,
8332         gcc::dump_manager::get_dump_file_name): Update prototype.
8334 2018-07-12  Richard Sandiford  <richard.sandiford@linaro.org>
8336         * internal-fn.h (vectorizable_internal_fn_p): New function.
8337         * tree-vect-slp.c (compatible_calls_p): Likewise.
8338         (vect_build_slp_tree_1): Remove nops argument.  Handle calls
8339         to internal functions.
8340         (vect_build_slp_tree_2): Update call to vect_build_slp_tree_1.
8342 2018-07-12  Richard Sandiford  <richard.sandiford@linaro.org>
8344         * fold-const.h (inverse_conditions_p): Declare.
8345         * fold-const.c (inverse_conditions_p): New function.
8346         * match.pd: Use inverse_conditions_p.  Add folds of view_converts
8347         that test the inverse condition of a conditional internal function.
8348         * internal-fn.h (vectorized_internal_fn_supported_p): Declare.
8349         * internal-fn.c (internal_fn_mask_index): Handle conditional
8350         internal functions.
8351         (vectorized_internal_fn_supported_p): New function.
8352         * tree-if-conv.c: Include internal-fn.h and fold-const.h.
8353         (any_pred_load_store): Replace with...
8354         (need_to_predicate): ...this new variable.
8355         (redundant_ssa_names): New variable.
8356         (ifcvt_can_use_mask_load_store): Move initial checks to...
8357         (ifcvt_can_predicate): ...this new function.  Handle tree codes
8358         for which a conditional internal function exists.
8359         (if_convertible_gimple_assign_stmt_p): Use ifcvt_can_predicate
8360         instead of ifcvt_can_use_mask_load_store.  Update after variable
8361         name change.
8362         (predicate_load_or_store): New function, split out from
8363         predicate_mem_writes.
8364         (check_redundant_cond_expr): New function.
8365         (value_available_p): Likewise.
8366         (predicate_rhs_code): Likewise.
8367         (predicate_mem_writes): Rename to...
8368         (predicate_statements): ...this.  Use predicate_load_or_store
8369         and predicate_rhs_code.
8370         (combine_blocks, tree_if_conversion): Update after above name changes.
8371         (ifcvt_local_dce): Handle redundant_ssa_names.
8372         * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): Handle
8373         general conditional functions.
8374         * tree-vect-stmts.c (vectorizable_call): Likewise.
8376 2018-07-12  Richard Sandiford  <richard.sandiford@linaro.org>
8377             Alan Hayward  <alan.hayward@arm.com>
8378             David Sherwood  <david.sherwood@arm.com>
8380         * internal-fn.h (can_interpret_as_conditional_op_p): Declare.
8381         * internal-fn.c (can_interpret_as_conditional_op_p): New function.
8382         * tree-ssa-math-opts.c (convert_mult_to_fma_1): Handle conditional
8383         plus and minus and convert them into IFN_COND_FMA-based sequences.
8384         (convert_mult_to_fma): Handle conditional plus and minus.
8386 2018-07-12  Richard Sandiford  <richard.sandiford@linaro.org>
8388         * doc/md.texi (cond_fma, cond_fms, cond_fnma, cond_fnms): Document.
8389         * optabs.def (cond_fma_optab, cond_fms_optab, cond_fnma_optab)
8390         (cond_fnms_optab): New optabs.
8391         * internal-fn.def (COND_FMA, COND_FMS, COND_FNMA, COND_FNMS): New
8392         internal functions.
8393         (FMA): Use DEF_INTERNAL_FLT_FN rather than DEF_INTERNAL_FLT_FLOATN_FN.
8394         * internal-fn.h (get_conditional_internal_fn): Declare.
8395         (get_unconditional_internal_fn): Likewise.
8396         * internal-fn.c (cond_ternary_direct): New macro.
8397         (expand_cond_ternary_optab_fn): Likewise.
8398         (direct_cond_ternary_optab_supported_p): Likewise.
8399         (FOR_EACH_COND_FN_PAIR): Likewise.
8400         (get_conditional_internal_fn): New function.
8401         (get_unconditional_internal_fn): Likewise.
8402         * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 5.
8403         (gimple_match_op::gimple_match_op): Add a new overload for 5
8404         operands.
8405         (gimple_match_op::set_op): Likewise.
8406         (gimple_resimplify5): Declare.
8407         * genmatch.c (decision_tree::gen): Generate simplifications for
8408         5 operands.
8409         * gimple-match-head.c (gimple_simplify): Define an overload for
8410         5 operands.  Handle calls with 5 arguments in the top-level overload.
8411         (convert_conditional_op): Handle conversions from unconditional
8412         internal functions to conditional ones.
8413         (gimple_resimplify5): New function.
8414         (build_call_internal): Pass a fifth operand.
8415         (maybe_push_res_to_seq): Likewise.
8416         (try_conditional_simplification): Try converting conditional
8417         internal functions to unconditional internal functions.
8418         Handle 3-operand unconditional forms.
8419         * match.pd (UNCOND_TERNARY, COND_TERNARY): Operator lists.
8420         Define ternary equivalents of the current rules for binary conditional
8421         internal functions.
8422         * config/aarch64/aarch64.c (aarch64_preferred_else_value): Handle
8423         ternary operations.
8424         * config/aarch64/iterators.md (UNSPEC_COND_FMLA, UNSPEC_COND_FMLS)
8425         (UNSPEC_COND_FNMLA, UNSPEC_COND_FNMLS): New unspecs.
8426         (optab): Handle them.
8427         (SVE_COND_FP_TERNARY): New int iterator.
8428         (sve_fmla_op, sve_fmad_op): New int attributes.
8429         * config/aarch64/aarch64-sve.md (cond_<optab><mode>)
8430         (*cond_<optab><mode>_2, *cond_<optab><mode_4)
8431         (*cond_<optab><mode>_any): New SVE_COND_FP_TERNARY patterns.
8433 2018-07-12  Richard Sandiford  <richard.sandiford@linaro.org>
8435         * target.def (preferred_else_value): New target hook.
8436         * doc/tm.texi.in (TARGET_PREFERRED_ELSE_VALUE): New hook.
8437         * doc/tm.texi: Regenerate.
8438         * targhooks.h (default_preferred_else_value): Declare.
8439         * targhooks.c (default_preferred_else_value): New function.
8440         * internal-fn.h (conditional_internal_fn_code): Declare.
8441         * internal-fn.c (FOR_EACH_CODE_MAPPING): New macro.
8442         (get_conditional_internal_fn): Use it.
8443         (conditional_internal_fn_code): New function.
8444         * gimple-match.h (gimple_match_cond): New struct.
8445         (gimple_match_op): Add a cond member function.
8446         (gimple_match_op::gimple_match_op): Update all forms to take a
8447         gimple_match_cond.
8448         * genmatch.c (expr::gen_transform): Use the same condition as res_op
8449         for the suboperation, but don't specify a particular else_value.
8450         * tree-ssa-sccvn.c (vn_nary_simplify, vn_reference_lookup_3)
8451         (visit_nary_op, visit_reference_op_load): Pass
8452         gimple_match_cond::UNCOND to the gimple_match_op constructor.
8453         * gimple-match-head.c: Include tree-eh.h
8454         (convert_conditional_op): New function.
8455         (maybe_resimplify_conditional_op): Likewise.
8456         (gimple_resimplify1): Call maybe_resimplify_conditional_op.
8457         (gimple_resimplify2): Likewise.
8458         (gimple_resimplify3): Likewise.
8459         (gimple_resimplify4): Likewise.
8460         (maybe_push_res_to_seq): Return null for conditional operations.
8461         (try_conditional_simplification): New function.
8462         (gimple_simplify): Call it.  Pass conditions to the gimple_match_op
8463         constructor.
8464         * match.pd: Fold VEC_COND_EXPRs of an IFN_COND_* call to a new
8465         IFN_COND_* call.
8466         * config/aarch64/aarch64.c (aarch64_preferred_else_value): New
8467         function.
8468         (TARGET_PREFERRED_ELSE_VALUE): Redefine.
8470 2018-07-12  Jan Hubicka  <hubicka@ucw.cz>
8472         * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
8473         DECL_FCONTEXT
8474         (hash_tree): Do not hash DECL_FCONTEXT
8475         * tree-streamer-in.c (lto_input_ts_field_decl_tree_pointers):
8476         Do not stream DECL_FCONTEXT.
8477         * tree-streamer-out.c (write_ts_field_decl_tree_pointers): Likewise.
8478         * tree.c (free_lang_data_in_decl): Free DECL_FCONTEXT.
8480 2018-07-12  Richard Biener  <rguenther@suse.de>
8482         PR debug/86462
8483         * dwarf2out.c (gen_block_die): Only output blocks when they have
8484         at least one !DECL_IGNORED_P variable.
8486 2018-07-12  Richard Biener  <rguenther@suse.de>
8488         PR target/84829
8489         * config/gnu-user.h (GNU_USER_TARGET_NO_PTHREADS_LIB_SPEC):
8490         Remove -mieee-fp handling.
8492 2018-07-12  Richard Biener  <rguenther@suse.de>
8494         * tree-ssa-sccvn.c (vn_lookup_simplify_result): Remove bogus
8495         left-over from last patch.
8497 2018-07-12  Jakub Jelinek  <jakub@redhat.com>
8499         PR tree-optimization/86492
8500         * gimple-ssa-store-merging.c
8501         (imm_store_chain_info::coalesce_immediate_stores): Call
8502         check_no_overlap even for the merge_overlapping case.  Formatting fix.
8504 2018-07-12  Richard Biener  <rguenther@suse.de>
8506         PR middle-end/86479
8507         * fold-const.c (fold_binary_op_with_conditional_arg): Do not
8508         move possibly trapping operations into the conditional.
8510 2018-07-12  Richard Biener  <rguenther@suse.de>
8512         * tree-ssa-sccvn.c (mprts_hook_cnt): Remove.
8513         (vn_lookup_simplify_result): Remove recursion limit applied
8514         here.
8515         (vn_nary_build_or_lookup_1): Adjust.
8516         (try_to_simplify): Likewise.
8517         * gimple-match-head.c (gimple_resimplify1): Instead apply one
8518         here.
8519         (gimple_resimplify2): Likewise.
8520         (gimple_resimplify3): Likewise.
8521         (gimple_resimplify4): Likewise.
8523 2018-07-11  Jakub Jelinek  <jakub@redhat.com>
8525         * config/i386/avx512bitalgintrin.h (_mm512_mask_bitshuffle_epi64_mask):
8526         Use __mmask64 type instead of __mmask8 for __M argument.
8527         * config/i386/avx512fintrin.h (_mm512_mask_xor_epi64,
8528         _mm512_maskz_xor_epi64): Use __mmask8 type instead of __mmask16 for
8529         __U argument.
8530         (_mm512_mask_cmpneq_epi64_mask): Use __mmask8 type instead of
8531         __mmask16 for __M argument.
8532         (_mm512_maskz_insertf32x4, _mm512_maskz_inserti32x4,
8533         _mm512_mask_insertf32x4, _mm512_mask_inserti32x4): Cast last argument
8534         to __mmask16 instead of __mmask8.
8535         * config/i386/avx512vlintrin.h (_mm_mask_add_ps, _mm_maskz_add_ps,
8536         _mm256_mask_add_ps, _mm256_maskz_add_ps, _mm_mask_sub_ps,
8537         _mm_maskz_sub_ps, _mm256_mask_sub_ps, _mm256_maskz_sub_ps,
8538         _mm256_maskz_cvtepi32_ps, _mm_maskz_cvtepi32_ps): Use __mmask8 type
8539         instead of __mmask16 for __U argument.
8540         * config/i386/avx512vlbwintrin.h (_mm_mask_cmp_epi8_mask): Use
8541         __mmask16 instead of __mmask8 for __U argument.
8542         (_mm256_mask_cmp_epi8_mask): Use __mmask32 instead of __mmask16 for
8543         __U argument.
8544         (_mm256_cmp_epi8_mask): Use __mmask32 return type instead of
8545         __mmask16.
8546         (_mm_mask_cmp_epu8_mask): Use __mmask16 instead of __mmask8 for __U
8547         argument.
8548         (_mm256_mask_cmp_epu8_mask): Use __mmask32 instead of __mmask16 for
8549         __U argument.
8550         (_mm256_cmp_epu8_mask): Use __mmask32 return type instead of
8551         __mmask16.
8552         (_mm_mask_cmp_epi16_mask): Cast last argument to __mmask8 instead
8553         of __mmask16.
8554         (_mm256_mask_cvtepi8_epi16): Use __mmask16 instead of __mmask32 for
8555         __U argument.
8556         (_mm_mask_cvtepi8_epi16): Use __mmask8 instead of __mmask32 for
8557         __U argument.
8558         (_mm256_mask_cvtepu8_epi16): Use __mmask16 instead of __mmask32 for
8559         __U argument.
8560         (_mm_mask_cvtepu8_epi16): Use __mmask8 instead of __mmask32 for
8561         __U argument.
8562         (_mm256_mask_cmpneq_epu8_mask, _mm256_mask_cmplt_epu8_mask,
8563         _mm256_mask_cmpge_epu8_mask, _mm256_mask_cmple_epu8_mask): Change
8564         return type as well as __M argument type and all casts from __mmask8
8565         to __mmask32.
8566         (_mm256_mask_cmpneq_epu16_mask, _mm256_mask_cmplt_epu16_mask,
8567         _mm256_mask_cmpge_epu16_mask, _mm256_mask_cmple_epu16_mask): Change
8568         return type as well as __M argument type and all casts from __mmask8
8569         to __mmask16.
8570         (_mm256_mask_cmpneq_epi8_mask, _mm256_mask_cmplt_epi8_mask,
8571         _mm256_mask_cmpge_epi8_mask, _mm256_mask_cmple_epi8_mask): Change
8572         return type as well as __M argument type and all casts from __mmask8
8573         to __mmask32.
8574         (_mm256_mask_cmpneq_epi16_mask, _mm256_mask_cmplt_epi16_mask,
8575         _mm256_mask_cmpge_epi16_mask, _mm256_mask_cmple_epi16_mask): Change
8576         return type as well as __M argument type and all casts from __mmask8
8577         to __mmask16.
8578         * config/i386/avx512vbmi2vlintrin.h (_mm_mask_shrdi_epi32,
8579         _mm_mask_shldi_epi32): Cast last argument to __mmask8 instead of
8580         __mmask16.
8582 2018-07-11  Grazvydas Ignotas  <notasas@gmail.com>
8584         * config/i386/avx512bwintrin.h: (_mm512_mask_cmp_epi8_mask,
8585         _mm512_mask_cmp_epu8_mask): Use __mmask64 type instead of __mmask32
8586         for __U argument.
8588 2018-07-11  Paul Koning  <ni1d@arrl.net>
8590         * doc/md.texi (define_subst): Document how multiple occurrences of
8591         the same argument in the replacement pattern are handled.
8593 2018-07-11  Paul Koning  <ni1d@arrl.net>
8595         * doc/extend.texi (Common Variable Attributes): Move "mode" into
8596         alphabetical order.
8597         (Common Type Attributes): Add "mode" attribute.
8599 2018-07-11  Jan Hubicka  <hubicka@ucw.cz>
8601         * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not
8602         stream DECL_ORIGINAL_TYPE.
8603         (DFS::DFS_write_tree_body): Drop hack handling local external decls.
8604         (hash_tree): Do not walk DECL_ORIGINAL_TYPE.
8605         * tree-streamer-in.c (lto_input_ts_decl_non_common_tree_pointers):
8606         Do not walk original type.
8607         * tree-streamer-out.c (streamer_write_chain): Drop hack handling
8608         external decls.
8609         (write_ts_decl_non_common_tree_pointers): Do not stream
8610         DECL_ORIGINAL_TYPE
8611         * tree.c (free_lang_data_in_decl): Clear DECL_ORIGINAL_TYPE.
8612         (find_decls_types_r): Do not walk DEC_ORIGINAL_TYPE.
8614 2018-07-11  Aldy Hernandez  <aldyh@redhat.com>
8616         * tree-ssa-threadupdate.c (thread_through_all_blocks): Do not jump
8617         thread twice from the same starting edge.
8619 2018-07-11  Aldy Hernandez  <aldyh@redhat.com>
8621         * vr-values.c (gimple_stmt_nonzero_p): Abstract common code to...
8622         * gimple.c (gimple_call_nonnull_result_p): ...here...
8623         (gimple_call_nonnull_arg): ...and here.
8624         * gimple.h (gimple_call_nonnull_result_p): New.
8625         (gimple_call_nonnull_arg): New.
8627 2018-07-11  Richard Earnshaw  <rearnsha@arm.com>
8629         * config/arm/arm-cpus.in: Move information from fpu field of each
8630         cpu definition to the isa field.
8631         * config/arm/parsecpu.awk (fpu): Delete match rule.
8632         (gen_comm_data): Don't add bits from the CPU's FPU entry.
8634 2018-07-11  Richard Biener  <rguenther@suse.de>
8636         PR debug/86457
8637         * dwarf2out.c (init_sections_and_labels): Use
8638         output_asm_line_debug_info consistently.
8639         (dwarf2out_early_finish): Likewise.
8640         (dwarf2out_finish): Remove DW_AT_stmt_list from early generated
8641         type units.
8643 2018-07-11  Richard Biener  <rguenther@suse.de>
8645         * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1):
8646         Rework father_bb setting in a way to avoid propagating constants
8647         multiple times on a loop body.
8649 2018-07-10  Mark Wielaard  <mark@klomp.org>
8651         PR debug/86459
8652         * dwarf2out.c (output_macinfo_op): Fix dwarf_FORM typo in gcc_assert.
8654 2018-07-10  Richard Biener  <rguenther@suse.de>
8656         * hash-map.h (hash_map::iterator::operator*): Return
8657         references to key and value.
8659 2018-07-10  Jakub Jelinek  <jakub@redhat.com>
8661         PR c++/86443
8662         * gimplify.c (find_combined_omp_for): Add DATA argument, in addition
8663         to finding the inner OMP_FOR/OMP_SIMD stmt find non-trivial wrappers,
8664         BLOCKs with BLOCK_VARs, OMP_PARALLEL in between, OMP_FOR in between.
8665         (gimplify_omp_for): For composite loops, move outer
8666         OMP_{DISTRIBUTE,TASKLOOP,FOR,PARALLEL} right around innermost
8667         OMP_FOR/OMP_SIMD if there are any non-trivial wrappers.  For class
8668         iterators add any needed clauses.  Allow OMP_FOR_ORIG_DECLS to contain
8669         TREE_LIST for both the original class iterator and the "last" helper
8670         var.  Gimplify OMP_FOR_PRE_BODY before the outermost composite
8671         loop, remember has_decl_expr from outer composite loops for the
8672         innermost OMP_SIMD in TREE_PRIVATE bit on OMP_FOR_INIT.
8674 2018-07-09  Martin Sebor  <msebor@redhat.com>
8676         PR middle-end/77357
8677         PR middle-end/86428
8678         * builtins.c (c_strlen): Avoid out-of-bounds warnings when
8679         accessing implicitly initialized array elements.
8680         * expr.c (string_constant): Handle string initializers of
8681         character arrays within aggregates.
8682         * gimple-fold.c (fold_array_ctor_reference): Add argument.
8683         Store element offset.  As a special case, handle zero size.
8684         (fold_nonarray_ctor_reference): Same.
8685         (fold_ctor_reference): Add argument.  Store subobject offset.
8686         * gimple-fold.h (fold_ctor_reference): Add argument.
8688 2018-07-09  Paul Koning  <ni1d@arrl.net>
8690         * config/pdp11/pdp11.c (pdp11_addr_cost): New function.
8691         (pdp11_insn_cost): New function.
8692         (pdp11_md_asm_adjust): New function.
8693         (TARGET_INVALID_WITHIN_DOLOOP): Define.
8694         (pdp11_rtx_costs): Update to match machine better.
8695         (output_addr_const_pdp11): Correct format mismatch warnings.
8696         * config/pdp11/pdp11.h (SLOW_BYTE_ACCESS): Correct definition.
8697         * config/pdp11/pdp11.md: General change to add base_cost and/or
8698         length attributes for use by new pdp11_insn_cost function.
8699         (MIN_BRANCH): Correct definition.
8700         (MIN_SOB): Ditto.
8701         (doloop_end): Use standard pattern name for looping pattern.
8702         (doloop_end_nocc): New.
8703         (movsf): Add another constraint alternative.
8704         (zero_extendqihi2): Add constraint alternatives for not in place
8705         extend.
8706         (zero_extendhisi2): Remove.
8707         (shift patterns): Add CC handling variants.
8708         (bswaphi2): New.
8709         (bswapsi2): New.
8710         (rothi3): New.
8711         (define_peephole2): New peephole to recognize mov that sets CC for
8712         subsequent test.
8714 2018-07-09  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
8716         * config/sparc/sparc.c (sparc_fold_builtin) <SPARC_BUILTIN_PDIST,
8717         SPARC_BUILTIN_PDISTN>: Adapt for signature change of wi::neg,
8718         wi::add.
8720 2018-07-09  Jakub Jelinek  <jakub@redhat.com>
8722         PR c/86420
8723         * real.c (real_nextafter): Return true if result is denormal.
8725 2018-07-09  Martin Liska  <mliska@suse.cz>
8727         * common.opt: Add back wrongly removed attribute.
8729 2018-07-09  Richard Biener  <rguenther@suse.de>
8731         PR debug/86413
8732         * dwarf2out.c (gen_block_die): For an early generated DIE
8733         always output high/low PC attributes.
8735 2018-07-09  Tom de Vries  <tdevries@suse.de>
8737         * cfgexpand.c (expand_debug_source_expr): Handle VAR_DECL.
8738         * tree-inline.c (remap_ssa_name): Handle default def ssa_name mapping
8739         onto VAR_DECL with abstract origin.
8741 2018-07-07  Jim Wilson  <jimw@sifive.com>
8743         * config/riscv/riscv.c (TARGET_CUSTOM_FUNCTION_DESCRIPTORS): New.
8745 2018-07-07  Tom de Vries  <tdevries@suse.de>
8747         * tree-dfa.c (dump_enumerated_decls): Handle cfun->cfg == NULL.
8749 2018-07-07  Aldy Hernandez  <aldyh@redhat.com>
8751         * tree-vrp.c (vrp_int_const_binop): Change overflow type to
8752         overflow_type.
8753         (combine_bound): Use wide-int overflow calculation instead of
8754         rolling our own.
8755         * calls.c (maybe_warn_alloc_args_overflow): Change overflow type to
8756         overflow_type.
8757         * fold-const.c (int_const_binop_2): Same.
8758         (extract_muldiv_1): Same.
8759         (fold_div_compare): Same.
8760         (fold_abs_const): Same.
8761         * match.pd: Same.
8762         * poly-int.h (add): Same.
8763         (sub): Same.
8764         (neg): Same.
8765         (mul): Same.
8766         * predict.c (predict_iv_comparison): Same.
8767         * profile-count.c (slow_safe_scale_64bit): Same.
8768         * simplify-rtx.c (simplify_const_binary_operation): Same.
8769         * tree-chrec.c (tree_fold_binomial): Same.
8770         * tree-data-ref.c (split_constant_offset_1): Same.
8771         * tree-if-conv.c (idx_within_array_bound): Same.
8772         * tree-scalar-evolution.c (iv_can_overflow_p): Same.
8773         * tree-ssa-phiopt.c (minmax_replacement): Same.
8774         * tree-vect-loop.c (is_nonwrapping_integer_induction): Same.
8775         * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Same.
8776         * vr-values.c (vr_values::adjust_range_with_scev): Same.
8777         * wide-int.cc (wi::add_large): Same.
8778         (wi::mul_internal): Same.
8779         (wi::sub_large): Same.
8780         (wi::divmod_internal): Same.
8781         * wide-int.h: Change overflow type to overflow_type for neg, add,
8782         mul, smul, umul, div_trunc, div_floor, div_ceil, div_round,
8783         mod_trunc, mod_ceil, mod_round, add_large, sub_large,
8784         mul_internal, divmod_internal.
8785         (overflow_type): New enum.
8786         (accumulate_overflow): New.
8788 2018-07-06  Kugan Vivekanandarajah  <kugan.vivekanandarajah@linaro.org>
8790         * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): New.
8791         (tree_ssa_phiopt_worker): Call cond_removal_in_popcount_pattern.
8793 2018-07-06  Kugan Vivekanandarajah  <kuganv@linaro.org>
8795         * tree-ssa-loop-niter.c (number_of_iterations_popcount): If popcount
8796         argument is checked for zero before entering loop, avoid checking again.
8798 2018-07-06  Kugan Vivekanandarajah  <kuganv@linaro.org>
8800         * gimplify.h (generic_expr_could_trap_p): Set as global function.
8801         * gimplify.h (generic_expr_could_trap_p): Likwise.
8802         * tree-scalar-evolution.c (expression_expensive_p): Handle COND_EXPR.
8804 2018-07-06  Jakub Jelinek  <jakub@redhat.com>
8806         PR tree-optimization/86401
8807         * fold-const.c (fold_binary_loc) <case BIT_AND_EXPR>: Move the
8808         ((A & N) + B) & M -> (A + B) & M etc. optimization into ...
8809         (fold_bit_and_mask): ... here.  New helper function for match.pd.
8810         * fold-const.h (fold_bit_and_mask): Declare.
8811         * match.pd (((A & N) + B) & M -> (A + B) & M): New optimization.
8813 2018-07-06  Peter Bergner  <bergner@linux.ibm.com>
8815         PR target/86324
8816         * target.def (translate_mode_attribute): New hook.
8817         * targhooks.h (default_translate_mode_attribute): Declare.
8818         * targhooks.c (default_translate_mode_attribute): New function.
8819         * doc/tm.texi.in (TARGET_TRANSLATE_MODE_ATTRIBUTE): New hook.
8820         * doc/tm.texi: Regenerate.
8821         * config/rs6000/rs6000.c (TARGET_TRANSLATE_MODE_ATTRIBUTE): Define.
8822         (rs6000_translate_mode_attribute): New function.
8824 2018-07-06  Paul Koning  <ni1d@arrl.net>
8826         * doc/md.texi (define_split): Document DONE and FAIL.
8827         (define_peephole2): Ditto.
8829 2018-07-05  Jeff Law  <law@redhat.com>
8831         PR tree-optimization/86010
8832         * tree-ssa-dse.c (compute_trims): More aggressively trim at
8833         both the head and tail of mem* and str* calls.
8835 2018-07-05  Jim Wilson  <jimw@sifive.com>
8837         * config.gcc (riscv*-*-*): When setting xlen, handle riscv-*.
8839 2018-07-05  Indu Bhagat  <indu.bhagat@oracle.com>
8841         * config/aarch64/aarch64-simd.md: correct flags text for
8842         MIN_EXPR replacement.
8844 2018-07-05  James Clarke  <jrtc27@jrtc27.com>
8846         * configure: Regenerated.
8848 2018-07-05  Carl Love  <cel@us.ibm.com>
8850         * config/rs6000/rs6000-c.c: Map ALTIVEC_BUILTIN_VEC_UNPACKH for
8851         float argument to VSX_BUILTIN_DOUBLEH_V4SF.
8852         Map ALTIVEC_BUILTIN_VEC_UNPACKL for float argument to
8853         VSX_BUILTIN_DOUBLEL_V4SF.
8855 2018-07-05  Martin Sebor  <msebor@redhat.com>
8857         PR c++/86400
8858         * tree-ssa-strlen.c (maybe_set_strlen_range): Use type size rather
8859         than its domain to compute its the upper bound of a char array.
8861 2018-07-05  Nathan Sidwell  <nathan@acm.org>
8863         Replace NO_IMPLICIT_EXTERN_C with SYSTEM_IMPLICIT_EXTERN_C.
8864         * doc/cpp.texi: Update comment.
8865         * doc/tm.texi: Rebuilt.
8866         * doc/tm.texi.in (NO_IMPLICIT_EXTERN_C): Replace with ...
8867         (SYSTEM_IMPLICIT_EXTERN_C): ... this, opposite sense.
8868         * doc/extend.texi (Backwards Compatibility): Clarify it is system
8869         headers affected by extern "C".
8870         * system.h: Poison NO_IMPLICIT_EXTERN_C.
8871         * config/alpha/alpha.h, config/arm/uclinux-elf.h,
8872         config/bfin/elf.h, config/cris/cris.h, config/darwin.h,
8873         config/dragonfly.h, config/freebsd.h, config/gnu-user.h,
8874         config/i386/cygming.h, config/i386/djgpp.h, config/i386/nto.h,
8875         config/ia64/hpux.h, config/lm32/lm32.h, config/lm32/uclinux-elf.h,
8876         config/lynx.h, config/mips/elf.h, config/mmix/mmix.h,
8877         config/netbsd.h, config/pa/pa-hpux.h, config/powerpcspe/sysv4.h,
8878         config/riscv/elf.h, config/rs6000/sysv4.h, config/rtems.h,
8879         config/s390/tpf.h, config/sh/newlib.h, config/sol2.h,
8880         config/sparc/openbsd64.h, config/sparc/sp-elf.h,
8881         config/sparc/sp64-elf.h, config/spu/spu.h,
8882         config/stormy16/stormy16.h, config/v850/v850.h,
8883         config/visium/visium.h, config/vx-common.h, config/xtensa/elf.h: Don't
8884         define NO_IMPLICIT_EXTERN_C.
8885         * config/rs6000/aix.h: Set SYSTEM_IMPLICIT_EXTERN_C.
8887 2018-07-05  Tamar Christina  <tamar.christina@arm.com>
8889         PR target/84711
8890         * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
8891         instead of GET_MODE_SIZE when comparing Units.
8893 2018-07-05  Tamar Christina  <tamar.christina@arm.com>
8895         PR target/84711
8896         * rtlanal.c (set_noop_p): Constrain on mode change,
8897         include hard-reg-set.h
8899 2018-07-05  Tamar Christina  <tamar.christina@arm.com>
8901         * config/aarch64/aarch64.c (aarch64_expand_movmem): Fix mode size.
8903 2018-07-05  Jakub Jelinek  <jakub@redhat.com>
8905         Revert
8906         2018-07-04  Maxim Ostapenko  <m.ostapenko@samsung.com>
8908         PR sanitizer/84250
8909         * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
8910         libasan.
8911         * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
8913 2018-07-04  Maxim Ostapenko  <m.ostapenko@samsung.com>
8915         PR sanitizer/84250
8916         * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
8917         libasan.
8918         * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
8920 2018-07-04  Eric Botcazou  <ebotcazou@adacore.com>
8922         PR middle-end/86380
8923         * expmed.c (choose_multiplier): Fix incorrect comparison with mask.
8925 2018-07-04  Aldy Hernandez  <aldyh@redhat.com>
8927         * tree-vrp.c (extract_range_from_binary_expr_1): Initialize
8928         neg_*_op* variables.
8930 2018-07-04  Martin Liska  <mliska@suse.cz>
8932         * tree-switch-conversion.c: Define
8933         max_ratio_for_speed and max_ratio_for_size constants.
8935 2018-07-04  Denys Vlasenko  <dvlasenk@redhat.com>
8936             Martin Liska  <mliska@suse.cz>
8938         PR middle-end/66240
8939         PR target/45996
8940         PR c/84100
8941         * common.opt: Rename align options with 'str_' prefix.
8942         * common/config/i386/i386-common.c (set_malign_value): New
8943         function.
8944         (ix86_handle_option): Use it to set -falign-* options/
8945         * config/aarch64/aarch64-protos.h (struct tune_params): Change
8946         type from int to string.
8947         * config/aarch64/aarch64.c: Update default values from int
8948         to string.
8949         * config/alpha/alpha.c (alpha_override_options_after_change):
8950         Likewise.
8951         * config/arm/arm.c (arm_override_options_after_change_1): Likewise.
8952         * config/i386/dragonfly.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
8953         max skip conditionally.
8954         * config/i386/freebsd.h (SUBALIGN_LOG): New.
8955         (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
8956         max skip conditionally.
8957         * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
8958         max skip conditionally.
8959         * config/i386/gnu-user.h (SUBALIGN_LOG): New.
8960         (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
8961         max skip conditionally.
8962         * config/i386/i386.c (struct ptt): Change type from int to
8963         string.
8964         (ix86_default_align): Set default values.
8965         * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Print
8966         max skip conditionally.
8967         * config/i386/iamcu.h (SUBALIGN_LOG): New.
8968         (ASM_OUTPUT_MAX_SKIP_ALIGN):
8969         * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN):
8970         * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
8971         max skip conditionally.
8972         * config/i386/openbsdelf.h (SUBALIGN_LOG): New.
8973         (ASM_OUTPUT_MAX_SKIP_ALIGN) Print max skip conditionally.:
8974         * config/i386/x86-64.h (SUBALIGN_LOG): New.
8975         (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
8976         max skip conditionally.
8977         (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
8978         * config/ia64/ia64.c (ia64_option_override): Set default values
8979         for alignment options.
8980         * config/m68k/m68k.c: Handle new str_align_* options.
8981         * config/mips/mips.c (mips_set_compression_mode): Change
8982         type of constants.
8983         (mips_option_override): Set default values for options.
8984         * config/powerpcspe/powerpcspe.c (rs6000_option_override_internal):
8985         Likewise.
8986         * config/rs6000/rs6000.c (rs6000_option_override_internal):
8987         Likewise.
8988         * config/rx/rx.c (rx_option_override): Likewise.
8989         * config/rx/rx.h (JUMP_ALIGN): Use align_jumps_log.
8990         (LABEL_ALIGN): Use align_labels_log.
8991         (LOOP_ALIGN): Use align_loops_align.
8992         * config/s390/s390.c (s390_asm_output_function_label): Use new
8993         macros.
8994         * config/sh/sh.c (sh_override_options_after_change):
8995         Change type of constants.
8996         * config/spu/spu.c (spu_sched_init): Likewise.
8997         * config/sparc/sparc.c (sparc_option_override): Set default
8998         values for options.
8999         * config/visium/visium.c (visium_option_override): Likewise.
9000         * config/visium/visium.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Do not
9001         emit p2align format with last argument if it's not needed.
9002         * doc/invoke.texi: Document extended format of -falign-*.
9003         * final.c: Use align_labels alignment.
9004         * flags.h (struct target_flag_state): Change type to use
9005         align_flags.
9006         (struct align_flags_tuple): New.
9007         (struct align_flags): Likewise.
9008         (align_loops_log): Redefine macro to use new types.
9009         (align_loops_max_skip): Redefine macro to use new types.
9010         (align_jumps_log): Redefine macro to use new types.
9011         (align_jumps_max_skip): Redefine macro to use new types.
9012         (align_labels_log): Redefine macro to use new types.
9013         (align_labels_max_skip): Redefine macro to use new types.
9014         (align_functions_log): Redefine macro to use new types.
9015         (align_loops): Redefine macro to use new types.
9016         (align_jumps): Redefine macro to use new types.
9017         (align_labels): Redefine macro to use new types.
9018         (align_functions): Redefine macro to use new types.
9019         (align_functions_max_skip): Redefine macro to use new types.
9020         (align_loops_value): New macro.
9021         (align_jumps_value): New macro.
9022         (align_labels_value): New macro.
9023         (align_functions_value): New macro.
9024         * function.c (invoke_set_current_function_hook): Propagate
9025         alignment values from flags to global variables default in
9026         topleev.h.
9027         * ipa-icf.c (sem_function::equals_wpa): Use
9028         cl_optimization_option_eq instead of memcmp.
9029         * lto-streamer.h (cl_optimization_stream_out): Support streaming
9030         of string types.
9031         (cl_optimization_stream_in): Likewise.
9032         * optc-save-gen.awk: Support strings in cl_optimization.
9033         * opth-gen.awk: Likewise.
9034         * opts.c (finish_options): Remove error checking of invalid
9035         value ranges.
9036         (MAX_CODE_ALIGN): Remove.
9037         (MAX_CODE_ALIGN_VALUE): Likewise.
9038         (parse_and_check_align_values): New function.
9039         (check_alignment_argument): Likewise.
9040         (common_handle_option): Use check_alignment_argument.
9041         * opts.h (parse_and_check_align_values): Declare.
9042         * toplev.c (init_alignments): Remove.
9043         (read_log_maxskip): New.
9044         (parse_N_M): Likewise.
9045         (parse_alignment_opts): Likewise.
9046         (backend_init_target): Remove usage of init_alignments.
9047         * toplev.h (parse_alignment_opts): Declare.
9048         * tree-streamer-in.c (streamer_read_tree_bitfields): Add new
9049         argument.
9050         * tree-streamer-out.c (streamer_write_tree_bitfields): Likewise.
9051         * tree.c (cl_option_hasher::equal): New.
9052         * varasm.c: Use new global macros.
9054 2018-07-04  Denys Vlasenko  <dvlasenk@redhat.com>
9056         * config/i386/dragonfly.h: (ASM_OUTPUT_MAX_SKIP_ALIGN):
9057         Use a simpler align directive also if MAXSKIP = ALIGN-1.
9058         * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
9059         * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
9060         * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
9061         * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
9062         * config/i386/freebsd.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Remove "If N
9063         is large, do at least 8 byte alignment" code. Add SUBALIGN_LOG
9064         define. Use a simpler align directive also if MAXSKIP = ALIGN-1.
9065         * config/i386/gnu-user.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
9066         * config/i386/iamcu.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
9067         * config/i386/openbsdelf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
9068         * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
9070 2018-07-04  Martin Liska  <mliska@suse.cz>
9071             Jonathan Wakely  <jwakely@redhat.com>
9073         * coverage.c: Use correct type.
9074         * doc/invoke.texi: Language correction.
9076 2018-07-03  H.J. Lu  <hongjiu.lu@intel.com>
9078         PR target/85620
9079         * config/i386/i386.c (rest_of_insert_endbranch): Also generate
9080         ENDBRANCH for non-tail call which may return via indirect branch.
9081         * doc/extend.texi: Document indirect_return attribute.
9083 2018-07-03  Martin Sebor  <msebor@redhat.com>
9085         PR tree-optimization/86274
9086         * gimple-ssa-sprintf.c (fmtresult::type_max_digits): Verify
9087         precondition.
9088         (format_floating): Correct handling of infinities and NaNs.
9090 2018-07-03  Martin Sebor  <msebor@redhat.com>
9092         * print-tree.c (print_real_cst): New function.
9093         (print_node_brief): Call it.
9094         (print_node): Ditto.
9096 2018-07-03  Jeff Law  <law@redhat.com>
9098         * config/h8300/h8300.md (logical<mode>3_sn, logical<mode>3): Merge
9099         into a single pattern.
9101         * config/h8300/h8300.md (ors code_iterator): New.
9102         (bsetqi_msx, bnotqi_msx patterns and splitters): Consolidate into
9103         a single pattern and single splitter.
9104         (bsethi_msx, bnothi_msx patterns): Consolidate into a single pattern.
9105         (iorqi3_1, xorqi3_1): Likewise.
9106         (iorqi3, xorqi3 expanders): Similarly.
9108         * config/h8300/h8300.md (movmd_internal_normal): Consolidated with
9109         (movmd_internal) into a single pattern using the P mode iterator.
9110         (movmd splitters): Similarly.
9111         (stpcpy_internal_normal, stpcpy_internal): Similarly for thes patterns.
9112         (movsd splitters): Similarly.
9114         * config/h8300/h8300.c (h8300_insn_length_from_table): Consolidate
9115         ADDB, ADDW and ADDL into a single ADD attribute which selects the
9116         right table based on the size of the operand.
9117         * config/h8300/h8300.md (length_table): Corresponding changes. All
9118         references to "addb", "addw" and "addl" changed to "add".
9119         (btst patterns): Merge two variants into a single pattern.
9120         (tstqi, tsthi): Likewise.
9121         (addhi3_incdec, addsi3_incdec): Likewise.
9122         (subhi3_h8300hs, subsi3_h8300hs): Likewise.
9123         (mulhi3, mulsi3): Likewise.
9124         (udivhi3, udivsi3): Likewise.
9125         (divhi3, divsi3): Likewise.
9126         (andorqi3, andorhi3, andorsi3): Likewise.
9128 2018-07-03  Uros Bizjak  <ubizjak@gmail.com>
9130         PR target/85694
9131         * config/i386/sse.md (uavg<mode>3_ceil): New expander.
9132         (<sse2_avx2>_uavg<mode>3<mask_name>): Simplify expander.
9134 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
9136         PR tree-optimization/85694
9137         * config/aarch64/iterators.md (HADD, RHADD): New int iterators.
9138         (u): Handle UNSPEC_SHADD, UNSPEC_UHADD, UNSPEC_SRHADD and
9139         UNSPEC_URHADD.
9140         * config/aarch64/aarch64-simd.md (<u>avg<mode>3_floor)
9141         (<u>avg<mode>3_ceil): New patterns.
9143 2018-07-03  David Malcolm  <dmalcolm@redhat.com>
9145         * gcc.dg/vect/slp-perm-1.c: Remove "note: " prefix from
9146         scan-tree-dump directive.
9147         * gcc.dg/vect/slp-perm-2.c: Likewise.
9148         * gcc.dg/vect/slp-perm-3.c: Likewise.
9149         * gcc.dg/vect/slp-perm-5.c: Likewise.
9150         * gcc.dg/vect/slp-perm-6.c: Likewise.
9151         * gcc.dg/vect/slp-perm-7.c: Likewise.
9152         * gcc.dg/vect/slp-perm-8.c: Likewise.
9154 2018-07-03  Marek Polacek  <polacek@redhat.com>
9156         PR middle-end/86202
9157         * gimple-fold.c (size_must_be_zero_p): Check the type of the size.
9159 2018-07-03  Richard Biener  <rguenther@suse.de>
9161         PR ipa/86389
9162         * tree-ssa-structalias.c (find_func_clobbers): Properly
9163         handle indirect calls.
9165 2018-07-03  Jeff Law  <law@redhat.com>
9167         * config/h8300/h8300.md (HSI, QHSI, QHSIF): New mode iterators.
9168         (shifts): New code iterator.
9169         (movqi, movhi, movsi, movsf expanders): Consolidate into a single
9170         expander.  Fix HImode handling on H8/SX.
9171         (addqi3, addhi3, addsi3 expanders): Consolidate into a single expander.
9172         (subqi3, subhi3, subsi3 expanders): Likewise.
9173         (andqi3, andhi3, andsi3 expanders): Likewise.
9174         (iorqi3, iorhi3, iorsi3 expanders): Likewise.
9175         (xorqi3, xorhi3, xorsi3 expanders): Likewise.
9176         (negqi2, neghi2, negsi2, negsf2 expanders): Likewise.
9177         (one_cmplqi2, one_cmplhi2, one_cmplsi2): Likewise.
9178         (zero_extendqihi2, zero_extendqisi2): Likewise.
9179         (extendqihi2, extendqisi2): Likewise.
9180         (rotlqi3, rotlhi3, rotlsi3): Likewise.
9181         (neghi2_h8300, negsi2_h8300): Likewise for these patterns.
9182         (rotlqi3_1, rotlhi3_1): Likewise.
9183         (logicalhi3_sn, logicalsi3_sn): Likewise.
9184         (logicalhi3, logicalsi3): Likewise.
9186 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
9188         * tree-vect-patterns.c (vect_recog_rotate_pattern)
9189         (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
9190         (vect_recog_mixed_size_cond_pattern, adjust_bool_pattern_cast)
9191         (adjust_bool_pattern, vect_recog_bool_pattern): Pass the vector
9192         type to append_pattern_def_seq instead of creating a stmt_vec_info
9193         directly.
9194         (build_mask_conversion): Likewise.  Remove vinfo argument.
9195         (vect_add_conversion_to_patterm): Likewise, renaming to...
9196         (vect_add_conversion_to_pattern): ...this.
9197         (vect_recog_mask_conversion_pattern): Update call to
9198         build_mask_conversion.  Pass the vector type to
9199         append_pattern_def_seq here too.
9200         (vect_recog_gather_scatter_pattern): Update call to
9201         vect_add_conversion_to_pattern.
9203 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
9205         * tree-vect-patterns.c (new_pattern_def_seq): Delete.
9206         (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
9207         (vect_recog_widen_op_pattern, vect_recog_over_widening_pattern)
9208         (vect_recog_rotate_pattern, vect_synth_mult_by_constant): Don't set
9209         STMT_VINFO_PATTERN_DEF_SEQ to null here.
9210         (vect_recog_pow_pattern, vect_recog_vector_vector_shift_pattern)
9211         (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern): Use
9212         append_pattern_def_seq instead of new_pattern_def_seq.
9213         (vect_recog_divmod_pattern): Do both of the above.
9214         (vect_pattern_recog_1): Assert that STMT_VINO_PATTERN_DEF_SEQ
9215         is null.
9217 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
9219         * tree-vect-patterns.c (vect_recog_dot_prod_pattern):
9220         (vect_recog_sad_pattern, vect_recog_widen_op_pattern)
9221         (vect_recog_widen_mult_pattern, vect_recog_pow_pattern):
9222         (vect_recog_widen_sum_pattern, vect_recog_over_widening_pattern)
9223         (vect_recog_average_pattern, vect_recog_cast_forwprop_pattern)
9224         (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
9225         (vect_recog_vector_vector_shift_pattern, vect_synth_mult_by_constant)
9226         (vect_recog_mult_pattern, vect_recog_divmod_pattern)
9227         (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
9228         (vect_recog_mask_conversion_pattern): Replace vec<gimple *>
9229         parameter with a single stmt_vec_info.
9230         (vect_recog_func_ptr): Likewise.
9231         (vect_recog_gather_scatter_pattern): Likewise, folding in...
9232         (vect_try_gather_scatter_pattern): ...this.
9233         (vect_pattern_recog_1): Remove stmts_to_replace and just pass
9234         the stmt_vec_info of the statement to be matched.  Don't clear
9235         STMT_VINFO_RELATED_STMT.
9236         (vect_pattern_recog): Update call accordingly.
9238 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
9240         PR tree-optimization/85694
9241         * doc/md.texi (avgM3_floor, uavgM3_floor, avgM3_ceil)
9242         (uavgM3_ceil): Document new optabs.
9243         * doc/sourcebuild.texi (vect_avg_qi): Document new target selector.
9244         * internal-fn.def (IFN_AVG_FLOOR, IFN_AVG_CEIL): New internal
9245         functions.
9246         * optabs.def (savg_floor_optab, uavg_floor_optab, savg_ceil_optab)
9247         (savg_ceil_optab): New optabs.
9248         * tree-vect-patterns.c (vect_recog_average_pattern): New function.
9249         (vect_vect_recog_func_ptrs): Add it.
9250         * tree-vect-stmts.c (vectorizable_call): Get the type of the zero
9251         constant directly from the associated lhs.
9253 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
9255         * tree-vect-patterns.c (vect_split_statement): New function.
9256         (vect_convert_input): Use it to try to split an existing cast.
9258 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
9260         * poly-int.h (print_hex): New function.
9261         * dumpfile.h (dump_dec, dump_hex): Declare.
9262         * dumpfile.c (dump_dec, dump_hex): New poly_wide_int functions.
9263         * tree-vectorizer.h (_stmt_vec_info): Add min_output_precision,
9264         min_input_precision, operation_precision and operation_sign.
9265         * tree-vect-patterns.c (vect_get_range_info): New function.
9266         (vect_same_loop_or_bb_p, vect_single_imm_use)
9267         (vect_operation_fits_smaller_type): Delete.
9268         (vect_look_through_possible_promotion): Add an optional
9269         single_use_p parameter.
9270         (vect_recog_over_widening_pattern): Rewrite to use new
9271         stmt_vec_info infomration.  Handle one operation at a time.
9272         (vect_recog_cast_forwprop_pattern, vect_narrowable_type_p)
9273         (vect_truncatable_operation_p, vect_set_operation_type)
9274         (vect_set_min_input_precision): New functions.
9275         (vect_determine_min_output_precision_1): Likewise.
9276         (vect_determine_min_output_precision): Likewise.
9277         (vect_determine_precisions_from_range): Likewise.
9278         (vect_determine_precisions_from_users): Likewise.
9279         (vect_determine_stmt_precisions, vect_determine_precisions): Likewise.
9280         (vect_vect_recog_func_ptrs): Put over_widening first.
9281         Add cast_forwprop.
9282         (vect_pattern_recog): Call vect_determine_precisions.
9284 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
9286         * tree-vect-patterns.c (vect_mark_pattern_stmts): Remove pattern
9287         statements that have been replaced by further pattern statements.
9288         (vect_pattern_recog_1): Clear STMT_VINFO_PATTERN_DEF_SEQ on failure.
9290 2018-07-03  Richard Biener  <rguenther@suse.de>
9292         * tree-vect-stmts.c (vect_is_simple_use): Consolidate dumping,
9293         always set *dt.  Dump vectype in vectype overload.
9294         * dumpfile.h (dump_gimple_expr): New function.
9295         (dump_gimple_expr_loc): Likewise.
9296         * dumpfile.c (dump_gimple_expr): New function.
9297         (dump_gimple_expr_loc): Likewise.
9299 2018-07-02  Jeff Law  <law@redhat.com>
9301         * config/h8300/h8300.md (movqi_h8300, movqi_h8300hs): Consolidate
9302         the H8/300, H8/300H and H8/S variants into a single pattern.
9303         (movhi_h8300, movqi_h8300hs): Similarly.
9304         (pushqi_h8300hs, pushhi_h8300hs): Consolidate into a single pattern.
9305         (QHI mode iterator): New.
9307         * config/h8300/h8300.md: Remove trailing whitespace.
9309 2018-07-02  Jim Wilson  <jimw@sifive.com>
9311         * config/riscv/riscv.c (riscv_expand_epilogue): Use emit_jump_insn
9312         instead of emit_insn for interrupt returns.
9313         * config/riscv/riscv.md (riscv_met): Add (return) to rtl.
9314         (riscv_sret, riscv_uret): Likewise.
9316 2018-07-02  David Malcolm  <dmalcolm@redhat.com>
9318         * pretty-print.c (selftest::test_pp_format): Move save and restore
9319         of quotes to class auto_fix_quotes, and add an instance.
9320         * selftest.c: Include "intl.h".
9321         (selftest::auto_fix_quotes::auto_fix_quotes): New ctor.
9322         (selftest::auto_fix_quotes::~auto_fix_quotes): New dtor.
9323         * selftest.h (selftest::auto_fix_quotes): New class.
9325 2018-07-02  Richard Henderson  <richard.henderson@linaro.org>
9327         * config/aarch64/aarch64-protos.h, config/aarch64/aarch64.c
9328         (aarch64_sve_prepare_conditional_op): Remove.
9329         * config/aarch64/aarch64-sve.md (cond_<SVE_INT_BINARY><SVE_I>):
9330         Allow aarch64_simd_reg_or_zero as select operand; remove
9331         the aarch64_sve_prepare_conditional_op call.
9332         (cond_<SVE_INT_BINARY_SD><SVE_SDI>): Likewise.
9333         (cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
9334         (*cond_<SVE_INT_BINARY><SVE_I>_z): New pattern.
9335         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_z): New pattern.
9336         (*cond_<SVE_COND_FP_BINARY><SVE_F>_z): New pattern.
9337         (*cond_<SVE_INT_BINARY><SVE_I>_any): New pattern.
9338         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_any): New pattern.
9339         (*cond_<SVE_COND_FP_BINARY><SVE_F>_any): New pattern
9340         and a splitters to match all of the *_any patterns.
9341         * config/aarch64/predicates.md (aarch64_sve_any_binary_operator): New.
9343         * config/aarch64/iterators.md (SVE_INT_BINARY_REV): Remove.
9344         (SVE_COND_FP_BINARY_REV): Remove.
9345         (sve_int_op_rev, sve_fp_op_rev): New.
9346         * config/aarch64/aarch64-sve.md (*cond_<SVE_INT_BINARY><SVE_I>_0): New.
9347         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_0): New.
9348         (*cond_<SVE_COND_FP_BINARY><SVE_F>_0): New.
9349         (*cond_<SVE_INT_BINARY><SVE_I>_2): Rename, add movprfx alternative.
9350         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_2): Similarly.
9351         (*cond_<SVE_COND_FP_BINARY><SVE_F>_2): Similarly.
9352         (*cond_<SVE_INT_BINARY><SVE_I>_3): Similarly; use sve_int_op_rev.
9353         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_3): Similarly.
9354         (*cond_<SVE_COND_FP_BINARY><SVE_F>_3): Similarly; use sve_fp_op_rev.
9356         * config/aarch64/aarch64-sve.md (cond_<SVE_COND_FP_BINARY><SVE_F>):
9357         Remove match_dup 1 from the inner unspec.
9358         (*cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
9360         * config/aarch64/aarch64.md (movprfx): New attr.
9361         (length): Default movprfx to 8.
9362         * config/aarch64/aarch64-sve.md (*mul<SVE_I>3): Add movprfx alt.
9363         (*madd<SVE_I>, *msub<SVE_I): Likewise.
9364         (*<su>mul<SVE_I>3_highpart): Likewise.
9365         (*<SVE_INT_BINARY_SD><SVE_SDI>3): Likewise.
9366         (*v<ASHIFT><SVE_I>3): Likewise.
9367         (*<su><MAXMIN><SVE_I>3): Likewise.
9368         (*<su><MAXMIN><SVE_F>3): Likewise.
9369         (*fma<SVE_F>4, *fnma<SVE_F>4): Likewise.
9370         (*fms<SVE_F>4, *fnms<SVE_F>4): Likewise.
9371         (*div<SVE_F>4): Likewise.
9373 2018-07-02  Richard Sandiford  <richard.sandiford@arm.com>
9375         * tree-vect-patterns.c (vect_recog_widen_shift_pattern): Fix typo
9376         in dump string.
9378 2018-07-02  Richard Biener  <rguenther@suse.de>
9380         PR tree-optimization/86363
9381         * tree-ssa-sccvn.c (vn_reference_lookup_3): Check the
9382         memset argument refers to a non-variable address.
9384 2018-07-02  Aldy Hernandez  <aldyh@redhat.com>
9386         * tree-vrp.c (extract_range_from_binary_expr_1): Abstract a lot of the
9387         {PLUS,MINUS}_EXPR code to...
9388         (adjust_symbolic_bound): ...here,
9389         (combine_bound): ...here,
9390         (set_value_range_with_overflow): ...and here.
9392 2018-07-02  Aldy Hernandez  <aldyh@redhat.com>
9394         * tree-vrp.c (extract_range_from_unary_expr): Abstract ABS_EXPR
9395         code...
9396         (extract_range_from_abs_expr): ...here.
9398 2018-07-02  Eric Botcazou  <ebotcazou@adacore.com>
9400         * config/i386/i386.c (ix86_finalize_stack_frame_flags): Do not overrule
9401         -fno-omit-frame-pointer when not optimizing.
9403 2018-07-02  Martin Liska  <mliska@suse.cz>
9405         PR ipa/86279
9406         * ipa-pure-const.c (malloc_candidate_p): Revert usage of ::get.
9407         (propagate_nothrow): Likewise.
9409 2018-07-02  Martin Liska  <mliska@suse.cz>
9411         PR ipa/86323
9412         * ipa-inline.c (early_inliner): Revert wrongly added ::get call.
9414 2018-07-02  David Malcolm  <dmalcolm@redhat.com>
9416         * dumpfile.c (dump_generic_expr_loc): Undo removal of this
9417         function in r262149, changing "loc" param from source_location to
9418         const dump_location_t &.
9419         * dumpfile.h (dump_generic_expr_loc): Undo removal of this
9420         declaration, as above.
9422 2018-07-01  Paul Koning  <ni1d@arrl.net>
9424         * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
9425         -munit-asm, -mgnu-asm, -mdec-asm.
9426         * config/pdp11/pdp11-protos.h (pdp11_gen_int_label): New.
9427         (pdp11_output_labelref): New.
9428         (pdp11_output_def): New.
9429         (pdp11_output_addr_vec_elt): New.
9430         * config/pdp11/pdp11.c: Use tab between opcode and operands.  Use
9431         %# and %@ format codes.
9432         (pdp11_option_override): New.
9433         (TARGET_ASM_FILE_START_FILE_DIRECTIVE): Define.
9434         (pdp11_output_ident): New.
9435         (pdp11_asm_named_section): New.
9436         (pdp11_asm_init_sections): New.
9437         (pdp11_file_start): New.
9438         (pdp11_file_end): New.
9439         (output_ascii): Use .ascii/.asciz for -mdec-asm.
9440         (pdp11_asm_print_operand): Update %# and %$ for -mdec-asm.  Add
9441         %o, like %c but octal.
9442         (pdp11_option_override): New.
9443         * config/pdp11/pdp11.h (TEXT_SECTION_ASM_OP): Update for
9444         -mdec-asm.
9445         (DATA_SECTION_ASM_OP): Ditto.
9446         (READONLY_DATA_SECTION_ASM_OP): New.
9447         (IS_ASM_LOGICAL_LINE_SEPARATOR): New.
9448         (ASM_GENERATE_INTERNAL_LABEL): Use new function.
9449         (ASM_OUTPUT_LABELREF): Ditto.
9450         (ASM_OUTPUT_DEF): Ditto.
9451         (ASM_OUTPUT_EXTERNAL): New.
9452         (ASM_OUTPUT_SOURCE_FILENAME): New.
9453         (ASM_OUTPUT_ADDR_VEC_ELT): Use new function.
9454         (ASM_OUTPUT_SKIP): Update for -mdec-asm.
9455         * config/pdp11/pdp11.md: Use tab between opcode and operands.  Use
9456         %# and %@ format codes.
9457         * config/pdp11/pdp11.opt (mgnu-asm): New.
9458         (mdec-asm): Conflicts with -mgnu-asm and -munix-asm.
9459         (munix-asm): Conflicts with -mdec-asm and -mgnu-asm.
9460         * doc/invoke.txt (PDP-11 Options): Add -mgnu-asm.
9462 2018-07-01  Aldy Hernandez  <aldyh@redhat.com>
9464         * tree-ssa-threadupdate.c (mark_threaded_blocks): Avoid
9465         dereferencing path[] beyond its length.
9466         (debug_path): New.
9467         (debug_all_paths): New.
9468         (rewire_first_differing_edge): New.
9469         (adjust_paths_after_duplication): New.
9470         (duplicate_thread_path): Call adjust_paths_after_duplication.
9471         Add new argument.
9472         (thread_through_all_blocks): Add new argument to
9473         duplicate_thread_path.
9475 2018-06-30  Jim Wilson  <jimw@sifive.com>
9477         * config/riscv/predicates.md (p2m1_shift_operand): New.
9478         (high_mask_shift_operand): New.
9479         * config/riscv/riscv.md (lshrsi3_zero_extend_3+1): New combiner
9480         pattern using p2m1_shift_operand.
9481         (lshsi3_zero_extend_3+2): New combiner pattern using
9482         high_mask_shift_operand.
9484 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
9486         * tree-vect-patterns.c (vect_get_external_def_edge): New function,
9487         split out from...
9488         (vect_recog_rotate_pattern): ...here.
9489         (vect_convert_input): Try to insert casts of invariants in the
9490         preheader.
9491         * tree-vect-loop-manip.c (vect_loop_versioning): Don't require the
9492         preheader to be empty.
9494 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
9496         * tree-vect-patterns.c (append_pattern_def_seq): Take an optional
9497         vector type.  If given, install it in the new statement's
9498         STMT_VINFO_VECTYPE.
9499         (vect_element_precision): New function.
9500         (vect_unpromoted_value): New struct.
9501         (vect_unpromoted_value::vect_unpromoted_value): New function.
9502         (vect_unpromoted_value::set_op): Likewise.
9503         (vect_look_through_possible_promotion): Likewise.
9504         (vect_joust_widened_integer, vect_joust_widened_type): Likewise.
9505         (vect_widened_op_tree, vect_convert_input): Likewise.
9506         (vect_convert_inputs, vect_convert_output): Likewise.
9507         (vect_recog_dot_prod_pattern): Use vect_look_through_possible_promotion
9508         to handle the optional cast of the multiplication result and
9509         vect_widened_op_tree to detect the widened multiplication itself.
9510         Do not require the input and output of promotion casts to have
9511         the same sign, but base the signedness of the operation on the
9512         input rather than the result.  If the pattern includes two
9513         promotions, check that those promotions have the same sign.
9514         Do not restrict the MULT_EXPR handling to a double-width result;
9515         handle quadruple-width results and wider.  Use vect_convert_inputs
9516         to convert the inputs to the common type.
9517         (vect_recog_sad_pattern):  Use vect_look_through_possible_promotion
9518         to handle the optional cast of the ABS result.  Also allow a sign
9519         change or a sign extension between the ABS and MINUS.
9520         Use vect_widened_op_tree to detect the widened subtraction and use
9521         vect_convert_inputs to convert the inputs to the common type.
9522         (vect_handle_widen_op_by_const): Delete.
9523         (vect_recog_widen_op_pattern): New function.
9524         (vect_recog_widen_mult_pattern): Use it.
9525         (vect_recog_widen_shift_pattern): Likewise.
9526         (vect_recog_widen_sum_pattern): Use
9527         vect_look_through_possible_promotion to handle the promoted
9528         PLUS_EXPR operand.
9530 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
9532         * gimple-iterator.c (gsi_for_stmt): Add a new overload that takes
9533         the containing gimple_seq *.
9534         * gimple-iterator.h (gsi_for_stmt): Declare it.
9535         * tree-vect-patterns.c (vect_recog_dot_prod_pattern)
9536         (vect_recog_sad_pattern, vect_recog_widen_sum_pattern)
9537         (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
9538         (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
9539         (vect_recog_mask_conversion_pattern): Remove STMT_VINFO_IN_PATTERN_P
9540         checks.
9541         (vect_init_pattern_stmt, vect_set_pattern_stmt): New functions,
9542         split out from...
9543         (vect_mark_pattern_stmts): ...here.  Handle cases in which the
9544         statement being replaced is part of an existing pattern
9545         definition sequence, inserting the new pattern statements before
9546         the original one.
9547         (vect_pattern_recog_1): Don't return a bool.  If the statement
9548         is already part of a pattern, instead apply pattern matching
9549         to the pattern definition statements.  Don't clear the
9550         STMT_VINFO_RELATED_STMT if is_pattern_stmt_p.
9551         (vect_pattern_recog): Don't break after the first match;
9552         continue processing the pattern definition statements instead.
9553         Don't bail out for STMT_VINFO_IN_PATTERN_P here.
9555 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
9557         * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
9558         (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
9559         (vect_recog_widen_sum_pattern): Use it.
9561 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
9563         * tree-vect-loop.c (vectorizable_reduction): Assert that the
9564         phi is not a pattern statement and has not been replaced by
9565         a pattern statement.
9566         * tree-vect-patterns.c (type_conversion_p): Don't check
9567         STMT_VINFO_IN_PATTERN_P.
9568         (vect_recog_vector_vector_shift_pattern): Likewise.
9569         (vect_recog_dot_prod_pattern): Expect vect_is_simple_use to return
9570         the pattern statement rather than the original statement; check
9571         directly for a WIDEN_MULT_EXPR here.
9572         * tree-vect-slp.c (vect_get_and_check_slp_defs): Expect
9573         vect_is_simple_use to return the pattern statement rather
9574         than the original statement; use is_pattern_stmt_p to check
9575         for such a pattern statement.
9576         * tree-vect-stmts.c (process_use): Expect vect_is_simple_use
9577         to return the pattern statement rather than the original statement;
9578         don't do the same transformation here.
9579         (vect_is_simple_use): If the defining statement has been replaced
9580         by a pattern statement, return the pattern statement instead.
9581         Remove the corresponding (local) transformation from the vectype
9582         overload.
9584 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
9586         * tree-vectorizer.h (vect_is_simple_use): Move the gimple ** to the
9587         end and default to null.
9588         * tree-vect-loop.c (vect_create_epilog_for_reduction)
9589         (vectorizable_reduction): Update calls accordingly, dropping the
9590         gimple ** argument if the passed-back statement isn't needed.
9591         * tree-vect-patterns.c (vect_get_internal_def, type_conversion_p)
9592         (vect_recog_rotate_pattern): Likewise.
9593         (vect_recog_mask_conversion_pattern): Likewise.
9594         * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
9595         (vect_mask_constant_operand_p): Likewise.
9596         * tree-vect-stmts.c (is_simple_and_all_uses_invariant, process_use):
9597         (vect_model_simple_cost, vect_get_vec_def_for_operand): Likewise.
9598         (get_group_load_store_type, get_load_store_type): Likewise.
9599         (vect_check_load_store_mask, vect_check_store_rhs): Likewise.
9600         (vectorizable_call, vectorizable_simd_clone_call): Likewise.
9601         (vectorizable_conversion, vectorizable_assignment): Likewise.
9602         (vectorizable_shift, vectorizable_operation): Likewise.
9603         (vectorizable_store, vect_is_simple_cond): Likewise.
9604         (vectorizable_condition, vectorizable_comparison): Likewise.
9605         (get_same_sized_vectype, vect_get_mask_type_for_stmt): Likewise.
9606         (vect_is_simple_use): Rename the def_stmt argument to def_stmt_out
9607         and move it to the end.  Cope with null def_stmt_outs.
9609 2018-06-30  Bernd Edlinger  <bernd.edlinger@hotmail.de>
9611         * Makefile.in (FLAGS_TO_PASS): Add RANLIB_FOR_TARGET.
9613 2018-06-29  Jeff Law  <law@redhat.com>
9615         * config/v850/v850.c (v850_legitimate_address_p): Handle large
9616         displacements for TARGET_V850E2V3 and newer.
9617         (TARGET_LRA_P): Remove.  Defaults to LRA now.
9618         * config/v850/v850.md (sign23byte_load): Remove.
9619         (unsign23byte_load, sign23hword_load, unsign23hword_load): Likewise.
9620         (23word_load, 23byte_store, 23hword_store, 23word_store): Likewise.
9622 2018-06-29  Martin Liska  <mliska@suse.cz>
9624         PR lto/85759
9625         * coverage.c (coverage_init): Mangle full path name.
9626         * doc/invoke.texi: Document the change.
9627         * gcov-io.c (mangle_path): New.
9628         * gcov-io.h (mangle_path): Likewise.
9629         * gcov.c (mangle_name): Use mangle_path for path mangling.
9631 2018-06-29  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
9633         * config/arm/arm.c (output_move_double): Don't allow STRD instructions
9634         if starting source register is not even.
9636 2018-06-29  Martin Liska  <mliska@suse.cz>
9638         PR tree-optimization/86263
9639         * tree-switch-conversion.c (switch_decision_tree::try_switch_expansion):
9640         Make edge redirection.
9642 2018-06-29  David Malcolm  <dmalcolm@redhat.com>
9644         * dumpfile.c (dump_loc): Add indentation based on scope depth.
9645         (dump_scope_depth): New variable.
9646         (get_dump_scope_depth): New function.
9647         (dump_begin_scope): New function.
9648         (dump_end_scope): New function.
9649         * dumpfile.h (get_dump_scope_depth): New declaration.
9650         (dump_begin_scope): New declaration.
9651         (dump_end_scope): New declaration.
9652         (class auto_dump_scope): New class.
9653         (AUTO_DUMP_SCOPE): New macro.
9654         * tree-vectorizer.h (DUMP_VECT_SCOPE): Reimplement in terms of
9655         AUTO_DUMP_SCOPE.
9657 2018-06-29  Richard Biener  <rguenther@suse.de>
9659         * tree-vect-data-refs.c (vect_analyze_data_ref_dependences): Assert
9660         compute_all_dependences succeeds.
9661         * tree-vect-loop.c (vect_get_datarefs_in_loop): Fail early if we
9662         exceed --param loop-max-datarefs-for-datadeps.
9664 2018-06-29  Jakub Jelinek  <jakub@redhat.com>
9666         * config/rs6000/t-rs6000: Append rs6000-modes.h to TM_H.
9668 2018-06-28  Uros Bizjak  <ubizjak@gmail.com>
9670         PR target/86348
9671         * config/i386/sse.md (*vec_extractv4si_0_zext_sse4): Use
9672         alternative 0 in preferred_for_speed attribute.
9674 2018-06-28  Paul Koning  <ni1d@arrl.net>
9676         * config/pdp11/pdp11-protos.h (pdp11_shift_length): New function.
9677         * config/pdp11/pdp11.c (pdp11_shift_length): New function.
9678         * config/pdp11/pdp11.h (ADJUST_INSN_LENGTH): Remove.
9679         * config/pdp11/pdp11.md: Correct "length" attribute calculation
9680         for shift insn patterns.
9682 2018-06-28  David Malcolm  <dmalcolm@redhat.com>
9684         * cgraph.c (cgraph_node::get_body): Replace assignments to
9685         "dump_file" with calls to set_dump_file.
9686         * dumpfile.c (alt_dump_file): Make static, and group with...
9687         (alt_flags): ...this definition.
9688         (dumps_are_enabled): New variable.
9689         (refresh_dumps_are_enabled): New function.
9690         (set_dump_file): New function.
9691         (set_alt_dump_file): New function.
9692         (gcc::dump_manager::dump_start): Replace assignments to
9693         "dump_file" and "alt_dump_file" with calls to set_dump_file and
9694         set_alt_dump_file.
9695         (gcc::dump_manager::dump_finish): Likewise.
9696         * dumpfile.h (alt_dump_file): Delete decl.
9697         (dumps_are_enabled): New variable decl.
9698         (set_dump_file): New function decl.
9699         (dump_enabled_p): Rewrite in terms of new "dumps_are_enabled"
9700         global.
9701         * tree-nested.c (lower_nested_functions): Replace assignments to
9702         "dump_file" with calls to set_dump_file.
9704 2018-06-28  Eric Botcazou  <ebotcazou@adacore.com>
9706         * tree-cfg.c (verify_gimple_in_cfg): Call verify_location on the
9707         goto_locus of each outgoing edge of each basic block.
9709 2018-06-28  Richard Biener  <rguenther@suse.de>
9711         * dwarf2out.c (decl_scope_table): Remove.
9712         (push_decl_scope): Likewise.
9713         (pop_decl_scope): Likewise.
9714         (gen_type_die_for_member): Do not call push/pop_decl_scope.
9715         (gen_struct_or_union_type_die): Likewise.
9716         (gen_tagged_type_die): Likewise.
9717         (dwarf2out_init): Do not initialize decl_scope_table.
9718         (dwarf2out_c_finalize): Do not free it.
9720 2018-06-28  Richard Biener  <rguenther@suse.de>
9722         * dwarf2out.c (gen_subprogram_die): Use is_unit_die when
9723         deciding whether to not re-use a DIE.
9725 2018-06-28  Richard Biener  <rguenther@suse.de>
9727         * dwarf2out.c (gen_subprogram_die): Always re-use DIEs with an
9728         DW_AT_abstract_origin attribute.
9730 2018-06-28  Martin Liska  <mliska@suse.cz>
9732         * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
9733         Use newly introduced constants.
9734         * tree-switch-conversion.h (struct jump_table_cluster):
9735         Define max_ratio_for_size and max_ratio_for_speed.
9737 2018-06-28  Martin Liska  <mliska@suse.cz>
9739         * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
9740         Add new checking assert to catch invalid state.
9741         (jump_table_cluster::can_be_handled): Handle single case
9742         clusters.
9743         (jump_table_cluster::is_beneficial): Bail out for such case.
9744         (bit_test_cluster::find_bit_tests):
9745         Add new checking assert to catch invalid state.
9746         (bit_test_cluster::can_be_handled): Handle single case
9747         clusters.
9748         (bit_test_cluster::is_beneficial): Bail out for such case.
9749         (switch_decision_tree::analyze_switch_statement):
9750         Fix comment.
9752 2018-06-28  Martin Liska  <mliska@suse.cz>
9754         * common.opt: Introduce -completion option.
9755         * gcc.c (driver_handle_option): Handle it.
9756         (driver::main): Print completions if completion
9757         is set.
9758         * opt-suggestions.c (option_proposer::get_completions):
9759         New function.
9760         (option_proposer::suggest_completion): Likewise.
9761         (option_proposer::find_param_completions): Likewise.
9762         (verify_autocompletions): Likewise.
9763         (test_completion_valid_options): Likewise.
9764         (test_completion_valid_params): Likewise.
9765         (in_completion_p): Likewise.
9766         (empty_completion_p): Likewise.
9767         (test_completion_partial_match): Likewise.
9768         (test_completion_garbage): Likewise.
9769         (opt_proposer_c_tests): Likewise.
9770         * opt-suggestions.h: Declare new functions.
9771         * opts.c (common_handle_option): Handle OPT__completion_.
9772         * selftest-run-tests.c (selftest::run_tests): Add
9773         opt_proposer_c_tests.
9774         * selftest.c (assert_str_startswith): New.
9775         * selftest.h (assert_str_startswith): Likewise.
9776         (opt_proposer_c_tests): New.
9777         (ASSERT_STR_STARTSWITH): Likewise.
9779 2018-06-28  Martin Liska  <mliska@suse.cz>
9781         * Makefile.in: Add opt-suggestions.o.
9782         * gcc-main.c: Include opt-suggestions.h.
9783         * gcc.c (driver::driver): Likewise.
9784         (driver::~driver): Remove m_option_suggestions.
9785         (driver::build_option_suggestions): Moved to option_proposer.
9786         (driver::suggest_option): Likewise.
9787         (driver::handle_unrecognized_options): Use option_proposer.
9788         * gcc.h (class driver): Add new memver m_option_proposer.
9789         * opt-suggestions.c: New file.
9790         * opt-suggestions.h: New file.
9792 2018-06-28  Martin Liska  <mliska@suse.cz>
9794         * vec.h (class auto_string_vec): New (moved from auto_argvec).
9795         (auto_string_vec::~auto_string_vec): Likewise.
9797 2018-06-28  Eric Botcazou  <ebotcazou@adacore.com>
9799         * tree-inline.h (struct copy_body_data): Move remapping_type_depth and
9800         prevent_decl_creation_for_types fields up and add reset_location field.
9801         * tree-inline.c (remap_gimple_stmt): Force input_location on the new
9802         statement if id->reset_location is true.
9803         (copy_edges_for_bb): Do not set goto_locus on the new edges if
9804         id->reset_location is true.
9805         (copy_phis_for_bb): Force input_location on the arguments if
9806         id->reset_location is true.
9807         (expand_call_inline): Set id->reset_location if DECL_IGNORED_P
9808         is set on the function to be inlined.
9810 2018-06-27  Stephan Bergmann  <sbergman@redhat.com>
9812         * doc/invoke.texi (Debugging Options): Add -gsplit-dwarf.
9814 2018-06-27  Dimitar Dimitrov  <dimitar@dinux.eu>
9816         * lra-eliminations.c (update_reg_eliminate): Mark all spanning hard
9817         registers for Pmode.
9818         * lra-lives.c (check_pseudos_live_through_calls): Mark all spanning
9819         hard registers for the clobbered pseudo.
9821 2018-06-27  Paul Koning  <ni1d@arrl.net>
9823         * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
9824         mutually exclusive options.
9825         * config/pdp11/constraints.md (h): New constraint.
9826         (O): Update definition to match shift code generation.
9827         (D): New constraint.
9828         * config/pdp11/pdp11-modes.def (CCNZ): Define mode.
9829         (CCFP): Remove.
9830         * config/pdp11/pdp11-protos.h (int_no_side_effect_operand): New
9831         function.
9832         (output_jump): Change arguments.
9833         (pdp11_fixed_cc_regs): New function.
9834         (pdp11_cc_mode): Ditto.
9835         (pdp11_expand_shift): Ditto.
9836         (pdp11_assemble_shift): Ditto.
9837         (pdp11_small_shift): Ditto.
9838         (pdp11_branch_cost): Remove.
9839         * config/pdp11/pdp11.c (pdp11_assemble_integer): Remove comments
9840         from output.
9841         (pdp11_register_move_cost): Update for CC registers.
9842         (pdp11_rtx_costs): Add case for LSHIFTRT.
9843         (pdp11_output_jump): Add CCNZ mode conditional branches.
9844         (notice_update_cc_on_set): Remove.
9845         (pdp11_cc_mode): New function.
9846         (simple_memory_operand): Correct pre/post decrement case.
9847         (no_side_effect_operand): New function.
9848         (pdp11_regno_reg_class): Add CC_REGS class.
9849         (pdp11_fixed_cc_regs): New function.
9850         (pdp11_small_shift): New function.
9851         (pdp11_expand_shift): New function to expand shift insns.
9852         (pdp11_assemble_shift): New function to output shifts.
9853         (pdp11_branch_cost): Remove.
9854         (pdp11_modes_tieable_p): Make QI/HI modes tieable.
9855         * config/pdp11/pdp11.h (SIZE_TYPE): Ensure 16-bit type.
9856         (WCHAR_TYPE): Ditto.
9857         (PTRDIFF_TYPE): Ditto.
9858         (ADJUST_INSN_LENGTH): New macro.
9859         (FIXED_REGISTERS): Add CC registers.
9860         (CALL_USED_REGISTERS): Ditto.
9861         (reg_class): Ditto.
9862         (REG_CLASS_NAMES): Ditto.
9863         (REG_CLASS_CONTENTS): Ditto.
9864         (SELECT_CC_MODE): Use new function.
9865         (TARGET_FLAGS_REGNUM): New macro.
9866         (TARGET_FIXED_CONDITION_CODE_REGS): Ditto.
9867         (cc0_reg_rtx): Remove.
9868         (CC_STATUS_MDEP): Remove.
9869         (CC_STATUS_MDEFP_INIT): Remove.
9870         (CC_IN_FPU): Remove.
9871         (NOTICE_UPDATE_CC): Remove.
9872         (REGISTER_NAMES): Add CC registers.
9873         (BRANCH_COST): Change to constant 1.
9874         * config/pdp11/pdp11.md: Rewrite for CCmode condition code
9875         handling.
9876         * config/pdp11/pdp11.opt (mbcopy): Remove.
9877         (mbcopy-builtin): Remove.
9878         (mbranch-cheap): Remove.
9879         (mbranch-expensive): Remove.
9880         * config/pdp11/predicates.md (expand_shift_operand): Update to
9881         match shift code generation.
9882         (ccnz_operator): New predicate.
9883         * doc/invoke.texi (PDP-11 Options): Remove deleted options
9884         -mbcopy, -mbcopy-builtin, -mbranch-cheap, -mbranch-expensive.
9885         Remove non-existent option -mabshi, -mno-abshi.  Document mutually
9886         exclusive options.
9887         * doc/md.texi (PDP-11): Document new D and h constraints.  Update
9888         description of O constraint.
9890 2018-06-27  Jeff Law  <law@redhat.com>
9891             Austin Law  <austinklaw@gmail.com>
9893         * config/v850/v850.md (addsi3_set_flags): New pattern.
9894         (subsi3_set_flags, negsi2_set_flags, andsi3_set_flags): Likewise.
9895         (iorsi3_set_flags, xorsi3_set_flags, one_cmplsi2_set_flags): Likewise.
9896         (zero_extendhisi2_v850_set_flags): Likewise.
9897         (zero_extendqisi2_v850_set_flags): Likewise.
9898         (ashlsi3_set_flags, ashlsi3_v850e2_set_flags): Likewise.
9899         (lshrsi3_set_flags, lshrsi3_v850e2_set_flags): Likewise.
9900         (ashrsi3_set_flags, ashrsi3_v850e2_set_flags): Likewise.
9902         * config/v850/v850-protos.h (notice_update_cc): Remove.
9903         * config/v850/v850.c (v850_compare_op0, v850_compare_op1): Remove.
9904         (v850_print_operand): Handle 'D' and "d".
9905         (v850_select_cc_mode): Remove ATTRIBUTE_UNUSED for last argument.
9906         Add handling of arithmetic/logical operations compared against zero.
9907         (v850_gen_float_compare): Remove ATTRIBUTE_UNUSED for last argument.
9908         Do not look at v850_compare_op, instead get mode from last argument.
9909         (v850_gen_compare): Remove
9910         (increment_stack): Use addsi3_clobber_flags to avoid splitting failure
9911         after reload for prologue insns.
9912         (expand_prologue): Account for CLOBBER of CC_REGNUM in various
9913         patterns.
9914         (construct_save_jarl): Likewise.
9915         (TARGET_FLAGS_REGNUM): Define.
9916         * config/v850/v850.h (v850_compare_op0, v850_compare_op1): Remove.
9917         (NOTICE_UPDATE_CC): Remove.
9918         * config/v850/v850.md (v850_tst1): Use (reg:CCZ CC_REGNUM) rather
9919         than cc0.  Conditionalize on reload_completed.
9920         (cmpsi_insn, setfcc_insn): Likewise.
9921         (tst1 splitter): Turn into define_and_split which sets the flags
9922         after reload.
9923         (cstoresi4, cbranchsf4, cbranchdf4, cbranchsi4_insn): Likewise.
9924         (cbranchsi4, branch_normal, branch_invert): Do not expose cc0 here.
9925         (cstoresf4, cstoredf4): Clobber the flags.
9926         (cmpsi, cmpsf, cmpdf): Remove expanders.
9927         (setf_insn): Remove pattern.
9928         (addsi3): Turn into define_and_split which clobbers the flags after
9929         reload and a suitable pattern (addsi3_clobber_flags) for use after
9930         reload.
9931         (subsi3, negsi2, andsi3, iorsi3, xorsi3, one_cmplsi2) Likewise.
9932         (ashlsi3, ashlsi3_v850e2, lshrsi3, lsh4si3_v850e2): Likewise.
9933         (ashrsi3, ashrsi3_v850e2): Likewise.
9934         (bins): Clobber the flags.
9935         (movsicc_normal_cc, movsicc_normal, movsicc_tst1): Likewise.
9936         (movsicc_tst1_revesed, sasf, swap and rotate patterns): Likewise.
9937         (fix_loop_counter, call_internal_short, call_internal_long): Likewise.
9938         (call_value_internal_short, call_value_internal_long): Likewise.
9939         (callt_save_interrupt, callt_return_interrupt): Likewise.
9940         (save_interrupt, return_interrupt): Likewise.
9941         (callt_save_all_interrupt, save_all_interrupt): Likewise.
9942         (_save_all_interrupt, callt_restore_all_interrupt): Likewise.
9943         (restore_all_interrupt, _restore_all_interrupt): Likewise.
9944         (All FP comparisons): Only allow after reload has completed.
9945         (trfsr): Likewise.
9946         (divh, divhu): Tweak output template.
9947         (branch_z_normal, branch_z_invert): Remove
9948         (branch_nz_normal, branch_nz_invert): Likewise.
9949         (extendhisi_insn, extendqisi_insn): Do not clobber flags.
9951         * config/v850/v850-modes.def (CCZ, CCNZ): Add new modes.
9952         * config/v850/v850.c (notice_update_cc): Remove.
9953         * config/v850/v850.h  (CC_OVERFLOW_UNUSABLE): Remove
9954         (CC_NO_CARRY): Likewise.
9955         (NOTICE_UPDATE_CC): Define to nothing.
9956         * config/v850/v850.md: Remove block comment on cc0 handling
9957         Remove "cc" attribute from all patterns.  Remove cc_status handling
9958         from all patterns.  Minor formatting fixes.
9960 2018-06-27  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
9962         * config/aarch64/aarch64-cores.def (cortex-a76): New entry.
9963         (cortex-a76.cortex-a55): Likewise.
9964         * config/aarch64/aarch64-tune.md: Regenerate.
9965         * doc/invoke.texi (AArch64 Options): Document cortex-a76 and
9966         cortex-a76.cortex-a55.
9968 2018-06-27  Jeff Law  <law@redhat.com>
9970         * config/v850/t-v850 (MULTILIB_OPTIONS): Remove 8byte-align.
9971         (MULTILIB_DIRNAMES): Similarly.
9973 2018-06-27  Eric Botcazou  <ebotcazou@adacore.com>
9975         * gimple.h (gimple_return_retbnd): Delete.
9976         (gimple_return_set_retbnd): Likewise.
9977         * cgraphunit.c (cgraph_node::expand_thunk): Remove call to
9978         gimple_return_set_retbnd.
9979         * gimple-pretty-print.c (dump_gimple_return): Remove call to
9980         gimple_return_retbnd and adjust.
9981         * tree-inline.h (struct copy_body_data): Remove retbnd field.
9982         * tree-inline.c (remap_gimple_stmt): Remove handling of retbnd.
9983         Explicitly return NULL in a couple more cases.  Move assertion
9984         on debug statements and remove unreachable code.
9985         (reset_debug_binding): Do not test id->retbnd.
9986         (expand_call_inline): Do not set it.
9988 2018-06-27  Rasmus Villemoes  <rv@rasmusvillemoes.dk>
9990         * configure.ac: Add --disable-gcov option.
9991         * configure: Regenerate.
9992         * Makefile.in: Honour @enable_gcov@.
9993         * doc/install.texi: Document --disable-gcov.
9995 2018-06-27  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
9997         * config/arm/arm-cpus.in (cortex-a76): New entry.
9998         (cortex-a76.cortex-a55): Likewise.
9999         * config/arm/arm-tables.opt: Regenerate.
10000         * config/arm/arm-tune.md: Likewise.
10001         * config/arm/driver-arm.c (arm_cpu_table): Add Cortex-A76 entry.
10002         * doc/invoke.texi (ARM Options): Document cortex-a76 and
10003         cortex-a76.cortex-a55.
10005 2018-06-27  Tamar Christina  <tamar.christina@arm.com>
10007         PR target/85769
10008         * config/aarch64/aarch64.md (*movhf_aarch64): Add dup v0.4h pattern.
10010 2018-06-27  Siddhesh Poyarekar  <siddhesh@sourceware.org>
10012         * config/aarch64/aarch64.h (CALL_USE_REGISTERS): Fix obsolete
10013         comment.
10014         (EPILOGUE_USES): Likewise.
10016 2018-06-26  Eric Botcazou  <ebotcazou@adacore.com>
10018         * tree-inline.c (remap_location): New function extracted from...
10019         (copy_edges_for_bb): Add ID parameter.  Remap goto_locus.
10020         (copy_phis_for_bb): ...here.  Call remap_location.
10021         (copy_cfg_body): Adjust call to copy_edges_for_bb.
10023 2018-06-26  Aaron Sawdey  <acsawdey@linux.ibm.com>
10025         * config/rs6000/rs6000-string.c (expand_block_clear): Don't use
10026         unaligned vsx for 16B memset.
10028 2018-06-26  Segher Boessenkool  <segher@kernel.crashing.org>
10030         PR target/86285
10031         * config/rs6000/rs6000.c (rs6000_init_builtins): Do not set
10032         ieee128_float_type_node to long_double_type_node unless
10033         TARGET_LONG_DOUBLE_128 is set.
10035 2018-06-26  David Malcolm  <dmalcolm@redhat.com>
10037         * cfgloop.c (get_loop_location): Convert return type from
10038         location_t to dump_user_location_t, replacing INSN_LOCATION lookups
10039         by implicit construction from rtx_insn *, and using
10040         dump_user_location_t::from_function_decl for the fallback case.
10041         * cfgloop.h (get_loop_location): Convert return type from
10042         location_t to dump_user_location_t.
10043         * cgraphunit.c (walk_polymorphic_call_targets): Update call to
10044         dump_printf_loc to pass in a dump_location_t rather than a
10045         location_t, via the gimple stmt.
10046         * coverage.c (get_coverage_counts): Update calls to
10047         dump_printf_loc to pass in dump_location_t rather than a
10048         location_t.
10049         * doc/optinfo.texi (Dump types): Convert example of
10050         dump_printf_loc from taking "locus" to taking "insn".  Update
10051         description of the "_loc" calls to cover dump_location_t.
10052         * dumpfile.c: Include "backend.h", "gimple.h", "rtl.h", and
10053         "selftest.h".
10054         (dump_user_location_t::dump_user_location_t): New constructors,
10055         from gimple *stmt and rtx_insn *.
10056         (dump_user_location_t::from_function_decl): New function.
10057         (dump_loc): Make static.
10058         (dump_gimple_stmt_loc): Convert param "loc" from location_t to
10059         const dump_location_t &.
10060         (dump_generic_expr_loc): Delete.
10061         (dump_printf_loc): Convert param "loc" from location_t to
10062         const dump_location_t &.
10063         (selftest::test_impl_location): New function.
10064         (selftest::dumpfile_c_tests): New function.
10065         * dumpfile.h: Include "profile-count.h".
10066         (class dump_user_location_t): New class.
10067         (struct dump_impl_location_t): New struct.
10068         (class dump_location_t): New class.
10069         (dump_printf_loc): Convert 2nd param from source_location to
10070         const dump_location_t &.
10071         (dump_generic_expr_loc): Delete.
10072         (dump_gimple_stmt_loc): Convert 2nd param from source_location to
10073         const dump_location_t &.
10074         * gimple-fold.c (fold_gimple_assign): Update call to
10075         dump_printf_loc to pass in a dump_location_t rather than a
10076         location_t, via the gimple stmt.
10077         (gimple_fold_call): Likewise.
10078         * gimple-loop-interchange.cc
10079         (loop_cand::analyze_iloop_reduction_var): Update for change to
10080         check_reduction_path.
10081         (tree_loop_interchange::interchange): Update for change to
10082         find_loop_location.
10083         * graphite-isl-ast-to-gimple.c (scop_to_isl_ast): Update for
10084         change in return-type of find_loop_location.
10085         (graphite_regenerate_ast_isl): Likewise.
10086         * graphite-optimize-isl.c (optimize_isl): Likewise.
10087         * graphite.c (graphite_transform_loops): Likewise.
10088         * ipa-devirt.c (ipa_devirt): Update call to dump_printf_loc to
10089         pass in a dump_location_t rather than a location_t, via the
10090         gimple stmt.
10091         * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
10092         * ipa.c (walk_polymorphic_call_targets): Likewise.
10093         * loop-unroll.c (report_unroll): Convert "locus" param from
10094         location_t to dump_location_t.
10095         (decide_unrolling): Update for change to get_loop_location's
10096         return type.
10097         * omp-grid.c (struct grid_prop): Convert field "target_loc" from
10098         location_t to dump_user_location_t.
10099         (grid_find_single_omp_among_assignments_1): Updates calls to
10100         dump_printf_loc to pass in a dump_location_t rather than a
10101         location_t, via the gimple stmt.
10102         (grid_parallel_clauses_gridifiable): Convert "tloc" from
10103         location_t to dump_location_t.  Updates calls to dump_printf_loc
10104         to pass in a dump_location_t rather than a location_t, via the
10105         gimple stmt.
10106         (grid_inner_loop_gridifiable_p): Likewise.
10107         (grid_dist_follows_simple_pattern): Likewise.
10108         (grid_gfor_follows_tiling_pattern): Likewise.
10109         (grid_target_follows_gridifiable_pattern): Likewise.
10110         (grid_attempt_target_gridification): Convert initialization
10111         of local "grid" from memset to zero-initialization; FIXME: does
10112         this require C++11?  Update call to dump_printf_loc to pass in a
10113         optinfo_location rather than a location_t, via the gimple stmt.
10114         * profile.c (read_profile_edge_counts): Updates call to
10115         dump_printf_loc to pass in a dump_location_t rather than a
10116         location_t
10117         (compute_branch_probabilities): Likewise.
10118         * selftest-run-tests.c (selftest::run_tests): Call
10119         dumpfile_c_tests.
10120         * selftest.h (dumpfile_c_tests): New decl.
10121         * tree-loop-distribution.c (pass_loop_distribution::execute):
10122         Update for change in return type of find_loop_location.
10123         * tree-parloops.c (parallelize_loops): Likewise.
10124         * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Convert
10125         "locus" from location_t to dump_user_location_t.
10126         (canonicalize_loop_induction_variables): Likewise.
10127         * tree-ssa-loop-ivopts.c (tree_ssa_iv_optimize_loop): Update
10128         for change in return type of find_loop_location.
10129         * tree-ssa-loop-niter.c (number_of_iterations_exit): Update call
10130         to dump_printf_loc to pass in a dump_location_t rather than a
10131         location_t, via the stmt.
10132         * tree-ssa-sccvn.c (eliminate_dom_walker::before_dom_children):
10133         Likewise.
10134         * tree-vect-loop-manip.c (find_loop_location): Convert return
10135         type from source_location to dump_user_location_t.
10136         (vect_do_peeling): Update for above change.
10137         (vect_loop_versioning): Update for change in type of
10138         vect_location.
10139         * tree-vect-loop.c (check_reduction_path): Convert "loc" param
10140         from location_t to dump_user_location_t.
10141         (vect_estimate_min_profitable_iters): Update for change in type
10142         of vect_location.
10143         * tree-vect-slp.c (vect_print_slp_tree): Convert param "loc" from
10144         location_t to dump_location_t.
10145         (vect_slp_bb): Update for change in type of vect_location.
10146         * tree-vectorizer.c (vect_location): Convert from source_location
10147         to dump_user_location_t.
10148         (try_vectorize_loop_1): Update for change in vect_location's type.
10149         (vectorize_loops): Likewise.
10150         (increase_alignment): Likewise.
10151         * tree-vectorizer.h (vect_location): Convert from source_location
10152         to dump_user_location_t.
10153         (find_loop_location): Convert return type from source_location to
10154         dump_user_location_t.
10155         (check_reduction_path): Convert 1st param from location_t to
10156         dump_user_location_t.
10157         * value-prof.c (check_counter): Update call to dump_printf_loc to
10158         pass in a dump_user_location_t rather than a location_t; update
10159         call to error_at for change in type of "locus".
10160         (check_ic_target): Update call to dump_printf_loc to
10161         pass in a dump_user_location_t rather than a location_t, via the
10162         call_stmt.
10164 2018-06-26  Robin Dapp  <rdapp@linux.vnet.ibm.com>
10166         * config/s390/s390.h (enum processor_flags): Do not use
10167         default tune parameter when -march was specified.
10169 2018-06-26  Jakub Jelinek  <jakub@redhat.com>
10171         PR target/86314
10172         * config/i386/i386.md (setcc + movzbl to xor + setcc peephole2s):
10173         Check reg_overlap_mentioned_p in addition to reg_set_p with the same
10174         operands.
10176 2018-06-26  Richard Biener  <rguenther@suse.de>
10178         PR tree-optimization/86287
10179         PR bootstrap/86316
10180         * tree-vect-loop.c (vect_transform_loop_stmt): Fix read-after-free.
10181         (vect_analyze_loop): Initialize n_stmts.
10183 2018-06-26  Richard Biener  <rguenther@suse.de>
10185         PR middle-end/86271
10186         * fold-const.c (fold_convertible_p): Pointer extension
10187         isn't valid.
10189 2018-06-26  Alexandre Oliva <aoliva@redhat.com>
10191         PR debug/86064
10192         * dwarf2out.c (loc_list_has_views): Adjust comments.
10193         (dw_loc_list): Split single cross-partition range with
10194         nonzero locview.
10196 2018-06-25  Jeff Law  <law@redhat.com>
10198         * common/config/v850/v850-common.c (TARGET_DEFAULT_TARGET_FLAGS): Turn
10199         on -mbig-switch by default.
10201         * config/v850/predicates.md (const_float_1_operand): Fix match_code
10202         test.
10203         (const_float_0_operand): Remove unused predicate.
10204         * config/v850/v850.md (define_constants): Remove UNSPEC_LOOP.
10205         (define_c_enum unspec): Add LOOP, RCP and RSQRT constants.
10206         (recipsf2): New expander.  Original pattern now called
10207         (recipsf2_insn).
10208         (recipdf2, recipdf2_insn): Similarly.
10209         (rsqrtsf2, rsqrtsf2_insn): Similarly
10210         (rsqrtdf2, rsqrtdf2_insn): Similarly
10212 2018-06-26  Gerald Pfeifer  <gerald@pfeifer.com>
10214         * ginclude/stddef.h: Remove an obsolete comment on FreeBSD 5.
10215         Simplify logic for FreeBSD (twice).
10217 2018-06-25  Martin Sebor  <msebor@redhat.com>
10219         PR tree-optimization/86204
10220         * tree-ssa-strlen.c (handle_builtin_strlen): Avoid storing
10221         a strnlen result if it's less than the length of the string.
10223 2018-06-25  Martin Sebor  <msebor@redhat.com>
10225         PR tree-optimization/85700
10226         * gimple-fold.c (gimple_fold_builtin_strncat): Adjust comment.
10227         * tree-ssa-strlen.c (is_strlen_related_p): Handle integer subtraction.
10228         (maybe_diag_stxncpy_trunc): Distinguish strncat from strncpy.
10230 2018-06-25  Martin Sebor  <msebor@redhat.com>
10232         * doc/extend.texi (Zero-length arrays): Update and clarify.
10234 2018-06-25  Michael Meissner  <meissner@linux.ibm.com>
10236         * config.gcc (powerpc64le*): Revert January 16th, 2018 patch that
10237         added IEEE/IBM long double multilib support on PowerPC little
10238         endian Linux systems.
10239         * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Likewise.
10240         (MULTILIB_DEFAULTS): Likewise.
10241         * config/rs6000/rs6000.c (rs6000_option_override_internal):
10242         Likewise.
10243         * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Likewise.
10244         * config/rs6000/t-ldouble-linux64le-ibm: Delete, no longer used.
10245         * config/rs6000/t-ldouble-linux64le-ieee: Delete, no longer used.
10247 2018-06-25  Alexander Monakov  <amonakov@ispras.ru>
10249         PR middle-end/86311
10250         * sort.cc (REORDER_23): Avoid memcpy with same destination and source.
10251         (REORDER_45): Likewise.
10253 2018-06-25  Jeff Law  <law@redhat.com>
10255         * config/v850/v850.md (divmodhi4): Make sure to sign extend the
10256         dividend to 32 bits.  Adjust length.
10257         (udivmodhi4): Cleanup output template.  Fix length.
10259 2018-06-25  Carl Love  <cel@us.ibm.com>
10261         * config/rs6000/vsx.md: Change word selector to prefered location.
10263 2018-06-25  Richard Biener  <rguenther@suse.de>
10265         PR tree-optimization/86304
10266         * tree-vectorizer.c (vectorize_loops): Walk over new possibly
10267         epilogue-if-converted loops as well.
10269 2018-06-25  Jan Hubicka  <hubicka@ucw.cz>
10271         * lto-section-out.c (lto_begin_section): Do not print section
10272         name for noaddr and unnumbered dumps.
10274 2018-06-25  Richard Biener  <rguenther@suse.de>
10276         * tree-vectorizer.h (struct vec_info_shared): New structure
10277         with parts split out from struct vec_info and loop_nest from
10278         struct _loop_vec_info.
10279         (struct vec_info): Adjust accordingly.
10280         (struct _loop_vec_info): Likewise.
10281         (LOOP_VINFO_LOOP_NEST): Adjust.
10282         (LOOP_VINFO_DATAREFS): Likewise.
10283         (LOOP_VINFO_DDRS): Likewise.
10284         (struct _bb_vec_info): Likewise.
10285         (BB_VINFO_DATAREFS): Likewise.
10286         (BB_VINFO_DDRS): Likewise.
10287         (struct _stmt_vec_info): Add dr_aux member.
10288         (DR_VECT_AUX): Adjust to refer to member of DR_STMTs vinfo.
10289         (DR_MISALIGNMENT_UNINITIALIZED): New.
10290         (set_dr_misalignment): Adjust.
10291         (dr_misalignment): Assert misalign isn't DR_MISALIGNMENT_UNINITIALIZED.
10292         (vect_analyze_loop): Adjust prototype.
10293         (vect_analyze_loop_form): Likewise.
10294         * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
10295         Compute dependences lazily.
10296         (vect_record_base_alignments): Use shared datarefs/ddrs.
10297         (vect_verify_datarefs_alignment): Likewise.
10298         (vect_analyze_data_refs_alignment): Likewise.
10299         (vect_analyze_data_ref_accesses): Likewise.
10300         (vect_analyze_data_refs): Likewise.
10301         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Add
10302         constructor parameter for shared part.
10303         (vect_analyze_loop_form): Pass in shared part and adjust.
10304         (vect_analyze_loop_2): Pass in storage for the number of
10305         stmts.  Move loop nest finding to the caller.  Compute
10306         datarefs lazily.
10307         (vect_analyze_loop): Pass in shared part.
10308         (vect_transform_loop): Verify shared datarefs are unchanged.
10309         * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Add
10310         constructor parameter for shared part.
10311         (vect_slp_analyze_bb_1): Pass in shared part and adjust.
10312         (vect_slp_bb): Verify shared datarefs are unchanged before
10313         transform.
10314         * tree-vect-stmts.c (ensure_base_align): Adjust for DR_AUX
10315         change.
10316         (new_stmt_vec_info): Initialize DR_AUX misalignment to
10317         DR_MISALIGNMENT_UNINITIALIZED.
10318         * tree-vectorizer.c (vec_info::vec_info): Add constructor
10319         parameter for shared part.
10320         (vec_info::~vec_info): Adjust.
10321         (vec_info_shared::vec_info_shared): New.
10322         (vec_info_shared::~vec_info_shared): Likewise.
10323         (vec_info_shared::save_datarefs): Likewise.
10324         (vec_info_shared::check_datarefs): Likewise.
10325         (try_vectorize_loop_1): Construct shared part live for analyses
10326         of a single loop for multiple vector sizes.
10327         * tree-parloops.c (gather_scalar_reductions): Adjust.
10329 2018-06-25  Richard Biener  <rguenther@suse.de>
10331         * tree-vect-data-refs.c (vect_find_stmt_data_reference): Modify
10332         DR for SIMD lane accesses here and mark DR with (void *)-1 aux.
10333         (vect_analyze_data_refs): Remove similar code from here and
10334         simplify accordingly.
10336 2018-06-25  Richard Biener  <rguenther@suse.de>
10338         * tree-vect-data-refs.c (vect_check_gather_scatter): Fail
10339         for reverse storage order accesses rather than asserting
10340         they cannot happen here.
10342 2018-06-25  Tom de Vries  <tdevries@suse.de>
10344         PR debug/86257
10345         * config/i386/i386.md (define_insn "*tls_global_dynamic_64_<mode>"):
10346         Use data16 instead of .byte for insn prefix.
10348 2018-06-25  Andreas Krebbel  <krebbel@linux.ibm.com>
10350         PR C++/86082
10351         * parser.c (make_char_string_pack): Pass this literal chars
10352         through cpp_interpret_string.
10353         (cp_parser_userdef_numeric_literal): Check the result of
10354         make_char_string_pack.
10356 2018-06-24  Maya Rashish  <coypu@sdf.org>
10358         * ginclude/stddef.h: Simplify conditions around avoiding
10359         re-definition of __size_t.
10361 2018-06-22  Jan Hubicka  <hubicka@ucw.cz>
10363         * lto-streamer-out.c (tree_is_indexable): Make LABEL_DECL nonindexable
10364         unless it is forced or nonlocal; assert that we stream no IMPORTED_DECL.
10366 2018-06-22  Maya Rashish  <coypu@sdf.org>
10368         * doc/invoke.texi (mno-fancy-math-387): Update for changes
10369         made to OpenBSD and NetBSD through the years.
10371 2018-06-22  Kelvin Nilsen  <kelvin@gcc.gnu.org>
10373         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
10374         behavior of vec_pack (vector double, vector double) to match
10375         behavior of vec_float2 (vector double, vector double).
10377 2018-06-22  Olivier Hainque  <hainque@adacore.com>
10379         * gimplify.c (gimplify_function_tree): Prevent creation
10380         of a trampoline for the address of the current function
10381         passed to entry/exit instrumentation hooks.
10383 2018-06-22  Aaron Sawdey  <acsawdey@linux.ibm.com>
10385         PR target/86222
10386         * config/rs6000/rs6000-string.c (expand_strn_compare): Handle -m32
10387         correctly.
10389 2018-06-22  Martin Liska  <mliska@suse.cz>
10391         PR tree-optimization/86263
10392         * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
10393         Bail out if is_enabled is false.
10394         * tree-switch-conversion.h (jump_table_cluster::is_enabled):
10395         New declaration.
10396         (jump_table_cluster::is_enabled): New function.
10398 2018-06-22  Jan Hubicka  <hubicka@ucw.cz>
10400         * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
10401         BINFO_BASE_ACCESSES and BINFO_VPTR_FIELD.
10402         * tree-streamer-in.c (streamer_read_tree_bitfields): Likewise.
10403         (lto_input_ts_binfo_tree_pointers): Likewise.
10404         * tree-streamer-out.c (streamer_write_tree_bitfields,
10405         write_ts_binfo_tree_pointers): Likewise.
10406         * tree.c (free_lang_data_in_binfo): Clear BINFO_VPTR_FIELD.
10408 2018-06-22  Jan Hubicka  <hubicka@ucw.cz>
10410         * tree.c (free_lang_data_in_type): Free all TYPE_VFIELDs.
10412 2018-06-22  Martin Liska  <mliska@suse.cz>
10414         * symbol-summary.h (get): Make it pure and inline move
10415         functionality from ::get function.
10416         (get): Remove and inline into ::get and ::get_create.
10417         (get_create): Move code from ::get function.
10419 2018-06-22  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
10421         PR target/85994
10422         * config/i386/sol2.h (CPP_SPEC): Don't pass -P for
10423         -x assembler-with-cpp.
10425 2018-06-22  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
10427         * config/sol2.h (TARGET_OS_CPP_BUILTINS): Define
10428         _FILE_OFFSET_BITS=64 for C++.
10430 2018-06-21  Michael Meissner  <meissner@linux.ibm.com>
10432         * config/rs6000/rs6000.md (extendtfif2): Add missing 128-bit
10433         conversion insn that shows up when pr85657-3.c is compiled using
10434         IEEE 128-bit long double.
10435         (neg<mode>2_internal): Use the correct mode to check whether the
10436         mode is IBM extended.
10437         * config/rs6000/rs6000.c (init_float128_ieee): Prevent complex
10438         multiply and divide external functions from being created more
10439         than once.
10441 2018-06-21  Eric Botcazou  <ebotcazou@adacore.com>
10443         * cfgrtl.c (fixup_reorder_chain): Do not emit NOPs in DECL_IGNORED_P
10444         functions.
10445         (rtl_merge_blocks): Likewise.  Do not emit a NOP if the location of
10446         the edge can be forwarded.
10447         (cfg_layout_merge_blocks): Likewise.
10449 2018-06-21  Eric Botcazou  <ebotcazou@adacore.com>
10451         * except.c (finish_eh_generation): Commit edge insertions only after
10452         the EH edges have been redirected from post-landing to landing pads.
10454 2018-06-21  Eric Botcazou  <ebotcazou@adacore.com>
10456         * tree-nested.c (get_frame_type): Use create_tmp_var_raw instead of
10457         create_tmp_var_for to create the FRAME decl.
10458         (finalize_nesting_tree_1): Do not unchain the FRAME decl.
10460 2018-06-21  Eric Botcazou  <ebotcazou@adacore.com>
10462         * tree-inline.c (copy_edges_for_bb): Minor tweak.
10463         (maybe_move_debug_stmts_to_successors): Also reset the locus of the
10464         debug statement when resetting its value.
10465         (expand_call_inline): Copy the locus of the call onto the assignment
10466         of the return value, if any.  Use local variable in more cases.
10468 2018-06-21  Martin Liska  <mliska@suse.cz>
10470         * ipa-pure-const.c (propagate_nothrow): Use
10471         funct_state_summaries->get.
10472         (dump_malloc_lattice): Likewise.
10473         (propagate_malloc): Likewise.
10475 2018-06-21  Richard Biener  <rguenther@suse.de>
10477         * lto-streamer-out.c (DFS::DFS_write_tree_body): Update outdated
10478         comment.  Follow BLOCK_ABSTRACT_ORIGIN unconditionally.
10479         * tree-streamer-in.c (lto_input_ts_block_tree_pointers): Update
10480         comment.
10481         * tree-streamer-out.c (write_ts_block_tree_pointers): Stream
10482         BLOCK_ABSTRACT_ORIGIN unconditionally.
10484 2018-06-21  David Malcolm  <dmalcolm@redhat.com>
10486         * ipa-cp.c (ipcp_driver): Set edge_clone_summaries to NULL after
10487         deleting it.
10488         * ipa-reference.c (ipa_reference_c_finalize): Delete
10489         ipa_ref_opt_sum_summaries and set it to NULL.
10491 2018-06-21  Tom de Vries  <tdevries@suse.de>
10493         PR tree-optimization/85859
10494         * tree-ssa-tail-merge.c (stmt_local_def): Copy gimple_is_call
10495         test with comment from bb_no_side_effects_p.
10497 2018-06-21  Richard Biener  <rguenther@suse.de>
10499         PR tree-optimization/86232
10500         * tree-ssa-loop-niter.c (number_of_iterations_popcount): Adjust
10501         max for constant niter.
10503 2018-06-21  Andre Vieira  <andre.simoesdiasvieira@arm.com>
10505         * config/aarch64/aarch64-simd.md
10506         (*aarch64_crypto_aes<aes_op>v16qi_xor_combine): New.
10508 2018-06-21  Andre Vieira  <andre.simoesdiasvieira@arm.com>
10510         * config/aarch64/aarch64-simd.md (aarch64_crypto_aes<aes_op>v16qi):
10511         Make opernads of the unspec commutative.
10513 2018-06-21  Richard Biener  <rguenther@suse.de>
10515         * tree-data-ref.c (dr_step_indicator): Handle NULL DR_STEP.
10516         * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr):
10517         Avoid calling vect_mark_for_runtime_alias_test with gathers or scatters.
10518         (vect_analyze_data_ref_dependence): Re-order checks to deal with
10519         NULL DR_STEP.
10520         (vect_record_base_alignments): Do not record base alignment
10521         for gathers or scatters.
10522         (vect_compute_data_ref_alignment): Drop return value that is always
10523         true.  Bail out early for gathers or scatters.
10524         (vect_enhance_data_refs_alignment): Bail out early for gathers
10525         or scatters.
10526         (vect_find_same_alignment_drs): Likewise.
10527         (vect_analyze_data_refs_alignment): Remove dead code.
10528         (vect_slp_analyze_and_verify_node_alignment): Likewise.
10529         (vect_analyze_data_refs): For possible gathers or scatters do
10530         not create an alternate DR, just check their possible validity
10531         and mark them.  Adjust DECL_NONALIASED handling to not rely
10532         on DR_BASE_ADDRESS.
10533         * tree-vect-loop-manip.c (vect_update_inits_of_drs): Do not
10534         update inits of gathers or scatters.
10535         * tree-vect-patterns.c (vect_recog_mask_conversion_pattern):
10536         Also copy gather/scatter flag to pattern vinfo.
10538 2018-06-20  Kelvin Nilsen  <kelvin@gcc.gnu.org>
10540         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
10541         behavior of vec_packsu (vector unsigned long long, vector unsigned
10542         long long) to match behavior of vec_packs with same signature.
10544 2018-06-20  Chung-Lin Tang <cltang@codesourcery.com>
10545             Thomas Schwinge <thomas@codesourcery.com>
10546             Cesar Philippidis  <cesar@codesourcery.com>
10548         * gimplify.c (gimplify_scan_omp_clauses): Add support for
10549         OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
10550         (gimplify_adjust_omp_clauses): Likewise.
10551         (gimplify_oacc_declare_1): Add support for GOMP_MAP_RELEASE, remove
10552         support for GOMP_MAP_FORCE_{ALLOC,TO,FROM,TOFROM}.
10553         (gimplify_omp_target_update): Update handling of acc update and
10554         enter/exit data.
10555         * omp-low.c (install_var_field): Remove unused parameter
10556         base_pointers_restrict.
10557         (scan_sharing_clauses): Remove base_pointers_restrict parameter.
10558         Update call to install_var_field. Handle OMP_CLAUSE_{IF_PRESENT,
10559         FINALIZE}
10560         (omp_target_base_pointers_restrict_p): Delete.
10561         (scan_omp_target): Update call to scan_sharing_clauses.
10562         * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_{IF_PRESENT,
10563         FINALIZE}.
10564         * tree-nested.c (convert_nonlocal_omp_clauses): Handle
10565         OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
10566         (convert_local_omp_clauses): Likewise.
10567         * tree-pretty-print.c (dump_omp_clause): Likewise.
10568         * tree.c (omp_clause_num_ops): Add entries for  OMP_CLAUSE_{IF_PRESENT,
10569         FINALIZE}.
10570         (omp_clause_code_name): Likewise.
10572 2018-06-20  Jakub Jelinek  <jakub@redhat.com>
10574         PR debug/86194
10575         * var-tracking.c (use_narrower_mode_test): Check if shift amount can
10576         be narrowed.
10578         PR tree-optimization/86231
10579         * tree-vrp.c (union_ranges): For (  [  )  ] or (   )[   ] range and
10580         anti-range don't overwrite *vr0min before using it to compute *vr0max.
10582 2018-06-20  Tom de Vries  <tdevries@suse.de>
10584         PR tree-optimization/86097
10585         * tree-ssa-loop-manip.c (canonicalize_loop_ivs): Also convert *nit to
10586         iv type if signedness of iv type is not the same as that of *nit.
10588 2018-06-20  Jakub Jelinek  <jakub@redhat.com>
10590         * cfgrtl.c (rtl_verify_edges): Formatting fix.  If bb->preds has any
10591         EDGE_EH edges, verify they are all EDGE_EH.
10593 2018-06-20  Maya Rashish  <coypu@sdf.org>
10595         * ginclude/stddef.h: Limit #include <machine/ansi.h> to NetBSD.
10597 2018-06-20  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
10599         * config/aarch64/aarch64-tuning-flags.def (no_ldp_stp_qregs): New.
10600         * config/aarch64/aarch64.c (xgene1_tunings): Add
10601         AARCH64_EXTRA_TUNE_NO_LDP_STP_QREGS to tune_flags.
10602         (aarch64_mode_valid_for_sched_fusion_p):
10603         Allow 16-byte modes.
10604         (aarch64_classify_address): Allow 16-byte modes for load_store_pair_p.
10605         * config/aarch64/aarch64-ldpstp.md: Add peepholes for LDP STP of
10606         128-bit modes.
10607         * config/aarch64/aarch64-simd.md (load_pair<VQ:mode><VQ2:mode>):
10608         New pattern.
10609         (vec_store_pair<VQ:mode><VQ2:mode>): Likewise.
10610         * config/aarch64/iterators.md (VQ2): New mode iterator.
10612 2018-06-20  Martin Liska  <mliska@suse.cz>
10614         * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
10615         Change default ratio from 10 to 8.
10617 2018-06-20  Martin Liska  <mliska@suse.cz>
10619         * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
10620         New.
10621         (bit_test_cluster::find_bit_tests): Likewise.
10622         (switch_decision_tree::analyze_switch_statement): Find clusters.
10623         * tree-switch-conversion.h (struct jump_table_cluster): Document
10624         hierarchy.
10626 2018-06-20  Martin Liska  <mliska@suse.cz>
10628         * tree-switch-conversion.c (switch_conversion::collect):
10629         Record m_uniq property.
10630         (switch_conversion::expand): Bail out for special conditions.
10631         (group_cluster::~group_cluster): New.
10632         (group_cluster::group_cluster): Likewise.
10633         (group_cluster::dump): Likewise.
10634         (jump_table_cluster::emit): New.
10635         (switch_decision_tree::fix_phi_operands_for_edges): New.
10636         (struct case_node): Remove struct.
10637         (jump_table_cluster::can_be_handled): New.
10638         (case_values_threshold): Moved to header.
10639         (reset_out_edges_aux): Likewise.
10640         (jump_table_cluster::is_beneficial): New.
10641         (bit_test_cluster::can_be_handled): Likewise.
10642         (add_case_node): Remove.
10643         (bit_test_cluster::is_beneficial): New.
10644         (case_bit_test::cmp): New.
10645         (bit_test_cluster::emit): New.
10646         (expand_switch_as_decision_tree_p): Remove.
10647         (bit_test_cluster::hoist_edge_and_branch_if_true): New.
10648         (fix_phi_operands_for_edge): Likewise.
10649         (switch_decision_tree::analyze_switch_statement): New.
10650         (compute_cases_per_edge): Move ...
10651         (switch_decision_tree::compute_cases_per_edge): ... here.
10652         (try_switch_expansion): Likewise.
10653         (switch_decision_tree::try_switch_expansion): Likewise.
10654         (record_phi_operand_mapping): Likewise.
10655         (switch_decision_tree::record_phi_operand_mapping): Likewise.
10656         (emit_case_decision_tree): Likewise.
10657         (switch_decision_tree::emit): Likewise.
10658         (balance_case_nodes): Likewise.
10659         (switch_decision_tree::balance_case_nodes): Likewise.
10660         (dump_case_nodes): Likewise.
10661         (switch_decision_tree::dump_case_nodes): Likewise.
10662         (emit_jump): Likewise.
10663         (switch_decision_tree::emit_jump): Likewise.
10664         (emit_cmp_and_jump_insns): Likewise.
10665         (switch_decision_tree::emit_cmp_and_jump_insns): Likewise.
10666         (emit_case_nodes): Likewise.
10667         (switch_decision_tree::emit_case_nodes): Likewise.
10668         (conditional_probability): Remove.
10669         * tree-switch-conversion.h (enum cluster_type): New.
10670         (PRINT_CASE): New.
10671         (struct cluster): Likewise.
10672         (cluster::cluster): Likewise.
10673         (struct simple_cluster): Likewise.
10674         (simple_cluster::simple_cluster): Likewise.
10675         (struct group_cluster): Likewise.
10676         (struct jump_table_cluster): Likewise.
10677         (struct bit_test_cluster): Likewise.
10678         (struct min_cluster_item): Likewise.
10679         (struct case_tree_node): Likewise.
10680         (case_tree_node::case_tree_node): Likewise.
10681         (jump_table_cluster::case_values_threshold): Likewise.
10682         (struct case_bit_test): Likewise.
10683         (struct switch_decision_tree): Likewise.
10684         (struct switch_conversion): Likewise.
10685         (switch_decision_tree::reset_out_edges_aux): Likewise.
10687 2018-06-20  Martin Liska  <mliska@suse.cz>
10689         * tree-switch-conversion.c (MAX_CASE_BIT_TESTS): Remove.
10690         (hoist_edge_and_branch_if_true): Likewise.
10691         (expand_switch_using_bit_tests_p): Likewise.
10692         (struct case_bit_test): Likewise.
10693         (case_bit_test_cmp): Likewise.
10694         (emit_case_bit_tests): Likewise.
10695         (switch_conversion::switch_conversion): New class.
10696         (struct switch_conv_info): Remove old struct.
10697         (collect_switch_conv_info): More to ...
10698         (switch_conversion::collect): ... this.
10699         (check_range): Likewise.
10700         (switch_conversion::check_range): Likewise.
10701         (check_all_empty_except_final): Likewise.
10702         (switch_conversion::check_all_empty_except_final): Likewise.
10703         (check_final_bb): Likewise.
10704         (switch_conversion::check_final_bb): Likewise.
10705         (create_temp_arrays): Likewise.
10706         (switch_conversion::create_temp_arrays): Likewise.
10707         (free_temp_arrays): Likewise.
10708         (gather_default_values): Likewise.
10709         (switch_conversion::gather_default_values): Likewise.
10710         (build_constructors): Likewise.
10711         (switch_conversion::build_constructors): Likewise.
10712         (constructor_contains_same_values_p): Likewise.
10713         (switch_conversion::contains_same_values_p): Likewise.
10714         (array_value_type): Likewise.
10715         (switch_conversion::array_value_type): Likewise.
10716         (build_one_array): Likewise.
10717         (switch_conversion::build_one_array): Likewise.
10718         (build_arrays): Likewise.
10719         (switch_conversion::build_arrays): Likewise.
10720         (gen_def_assigns): Likewise.
10721         (switch_conversion::gen_def_assigns): Likewise.
10722         (prune_bbs): Likewise.
10723         (switch_conversion::prune_bbs): Likewise.
10724         (fix_phi_nodes): Likewise.
10725         (switch_conversion::fix_phi_nodes): Likewise.
10726         (gen_inbound_check): Likewise.
10727         (switch_conversion::gen_inbound_check): Likewise.
10728         (process_switch): Use the newly created class.
10729         (switch_conversion::expand): New.
10730         (switch_conversion::~switch_conversion): New.
10731         * tree-switch-conversion.h: New file.
10733 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
10735         * tree-vectorizer.h (NUM_PATTERNS, vect_recog_func_ptr): Move to
10736         tree-vect-patterns.c.
10737         * tree-vect-patterns.c (vect_supportable_direct_optab_p): New function.
10738         (vect_recog_dot_prod_pattern): Use it.  Remove the type_in argument.
10739         (vect_recog_sad_pattern): Likewise.
10740         (vect_recog_widen_sum_pattern): Likewise.
10741         (vect_recog_pow_pattern): Likewise.  Check for a null vectype.
10742         (vect_recog_widen_shift_pattern): Remove the type_in argument.
10743         (vect_recog_rotate_pattern): Likewise.
10744         (vect_recog_mult_pattern): Likewise.
10745         (vect_recog_vector_vector_shift_pattern): Likewise.
10746         (vect_recog_divmod_pattern): Likewise.
10747         (vect_recog_mixed_size_cond_pattern): Likewise.
10748         (vect_recog_bool_pattern): Likewise.
10749         (vect_recog_mask_conversion_pattern): Likewise.
10750         (vect_try_gather_scatter_pattern): Likewise.
10751         (vect_recog_widen_mult_pattern): Likewise.  Check for a null vectype.
10752         (vect_recog_over_widening_pattern): Likewise.
10753         (vect_recog_gather_scatter_pattern): Likewise.
10754         (vect_recog_func_ptr): Move from tree-vectorizer.h
10755         (vect_vect_recog_func_ptrs): Move further down the file.
10756         (vect_recog_func): Likewise.  Remove the third argument.
10757         (NUM_PATTERNS): Define based on vect_vect_recog_func_ptrs.
10758         (vect_pattern_recog_1): Expect the pattern function to do any
10759         necessary target tests.  Also expect it to provide a vector type.
10760         Remove the type_in handling.
10762 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
10764         * tree-vect-patterns.c (vect_pattern_detected): New function.
10765         (vect_recog_dot_prod_patternm, vect_recog_sad_pattern)
10766         (vect_recog_widen_mult_pattern, vect_recog_widen_sum_pattern)
10767         (vect_recog_over_widening_pattern, vect_recog_widen_shift_pattern
10768         (vect_recog_rotate_pattern, vect_recog_vector_vector_shift_pattern)
10769         (vect_recog_mult_pattern, vect_recog_divmod_pattern)
10770         (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
10771         (vect_recog_mask_conversion_pattern)
10772         (vect_try_gather_scatter_pattern): Likewise.
10774 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
10776         * tree-vect-patterns.c (vect_get_internal_def): New function.
10777         (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
10778         (vect_recog_vector_vector_shift_pattern, check_bool_pattern)
10779         (search_type_for_mask_1): Use it.
10781 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
10783         * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
10784         redundant WIDEN_SUM_EXPR handling.
10785         (vect_recog_sad_pattern): Likewise.
10787 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
10789         * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
10790         redundant check that the types of a PLUS_EXPR or MULT_EXPR agree.
10791         (vect_recog_sad_pattern): Likewise PLUS_EXPR, ABS_EXPR and MINUS_EXPR.
10792         (vect_recog_widen_mult_pattern): Likewise MULT_EXPR.
10793         (vect_recog_widen_sum_pattern): Likewise PLUS_EXPR.
10795 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
10797         * tree-vect-stmts.c (vectorizable_call): Make sure that we
10798         use the stmt_vec_info of the original bb statement for the
10799         new zero assignment, even if the call is part of a pattern.
10801 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
10803         * tree-vectorizer.h (_stmt_vec_info): Note above pattern_def_seq
10804         that the sequence is attached to the original statement rather
10805         than the pattern statement.
10806         * tree-vect-loop.c (vect_determine_vf_for_stmt): Take the
10807         PATTERN_DEF_SEQ from the original statement rather than
10808         the main pattern statement.
10809         * tree-vect-stmts.c (free_stmt_vec_info): Likewise.
10810         * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Likewise.
10811         (vect_mark_pattern_stmts): Don't copy the PATTERN_DEF_SEQ.
10813 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
10815         * tree-vect-stmts.c (vect_analyze_stmt): Move the handling of pattern
10816         definition statements before the early exit for statements that aren't
10817         live or relevant.
10818         * tree-vect-loop.c (vect_transform_loop_stmt): New function,
10819         split out from...
10820         (vect_transform_loop): ...here.  Process pattern definition
10821         statements without first checking whether the main pattern
10822         statement is live or relevant.
10824 2018-06-19  Eric Botcazou  <ebotcazou@adacore.com>
10826         * tree-cfgcleanup.c (tree_forwarder_block_p): Do not return false at
10827         -O0 if the locus represent UNKNOWN_LOCATION but have different values.
10829 2018-06-19  Aaron Sawdey  <acsawdey@linux.ibm.com>
10831         * config/rs6000/rs6000-string.c (select_block_compare_mode): Check
10832         TARGET_EFFICIENT_OVERLAPPING_UNALIGNED here instead of in caller.
10833         (do_and3, do_and3_mask, do_compb3, do_rotl3): New functions.
10834         (expand_block_compare): Change select_block_compare_mode call.
10835         (expand_strncmp_align_check): Use new functions, fix comment.
10836         (emit_final_str_compare_gpr): New function.
10837         (expand_strn_compare): Refactor and clean up code.
10838         * config/rs6000/vsx.md (vsx_mov<mode>_64bit): Remove *.
10840 2018-06-19  Tony Reix  <tony.reix@atos.com>
10841             Damien Bergamini  <damien.bergamini@atos.com>
10842             David Edelsohn  <dje.gcc@gmail.com>
10844         * collect2.c (static_obj): New variable.
10845         (static_libs): New variable.
10846         (is_in_list): Uncomment declaration.
10847         (main): Track AIX libraries linked statically.
10848         (is_in_list): Uncomment definition.
10849         (scan_prog_file): Don't add AIX shared libraries initializer
10850         to constructor list if linking statically.
10852 2018-06-19  Max Filippov  <jcmvbkbc@gmail.com>
10854         * config/xtensa/xtensa.md (UNSPEC_FRAME_BLOCKAGE): New unspec
10855         constant.
10856         (allocate_stack, frame_blockage, *frame_blockage): New patterns.
10858 2018-06-19  Jan Hubicka  <hubicka@ucw.cz>
10860         * tree.c (find_decls_types_r): Remove all non-VAR_DECLs from
10861         blocks.
10863 2018-06-19  Martin Liska  <mliska@suse.cz>
10865         * config/i386/i386.c (ix86_can_inline_p): Do not use
10866         ipa_fn_summaries::get_create.
10867         * ipa-cp.c (ipcp_cloning_candidate_p): Replace get_create with
10868         get.
10869         (devirtualization_time_bonus): Likewise.
10870         (ipcp_propagate_stage): Likewise.
10871         * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
10872         (edge_set_predicate): Likewise.
10873         (evaluate_conditions_for_known_args): Likewise.
10874         (evaluate_properties_for_edge): Likewise.
10875         (ipa_call_summary::reset): Tranform to ...
10876         (ipa_call_summary::~ipa_call_summary): ... this.
10877         (ipa_fn_summary::reset): Transform to ...
10878         (ipa_fn_summary::~ipa_fn_summary): ... this.
10879         (ipa_fn_summary_t::remove): Rename to ...
10880         (ipa_fn_summary_t::remove_callees): ... this.
10881         (ipa_fn_summary_t::duplicate): Use placement new
10882         instead of memory copy.
10883         (ipa_call_summary_t::duplicate): Likewise.
10884         (ipa_call_summary_t::remove): Remove.
10885         (dump_ipa_call_summary): Change get_create to get.
10886         (ipa_dump_fn_summary): Dump only when summary exists.
10887         (analyze_function_body): Use symbol_summary::get instead
10888         of get_create.
10889         (compute_fn_summary): Likewise.
10890         (estimate_edge_devirt_benefit): Likewise.
10891         (estimate_edge_size_and_time): Likewise.
10892         (inline_update_callee_summaries): Likewise.
10893         (remap_edge_change_prob): Likewise.
10894         (remap_edge_summaries): Likewise.
10895         (ipa_merge_fn_summary_after_inlining): Likewise.
10896         (write_ipa_call_summary): Likewise.
10897         (ipa_fn_summary_write): Likewise.
10898         (ipa_free_fn_summary): Likewise.
10899         * ipa-fnsummary.h (struct GTY): Add new ctor and copy ctor.
10900         (struct ipa_call_summary): Likewise.
10901         * ipa-icf.c (sem_function::merge): Use symbol_summary::get instead
10902         of get_create.
10903         * ipa-inline-analysis.c (do_estimate_edge_time): Likewise.
10904         (estimate_size_after_inlining): Likewise.
10905         (estimate_growth): Likewise.
10906         (growth_likely_positive): Likewise.
10907         * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
10908         (inline_call): Likewise.
10909         * ipa-inline.c (caller_growth_limits): Likewise.
10910         (can_inline_edge_p): Likewise.
10911         (can_inline_edge_by_limits_p): Likewise.
10912         (compute_uninlined_call_time): Likewise.
10913         (compute_inlined_call_time): Likewise.
10914         (want_inline_small_function_p): Likewise.
10915         (edge_badness): Likewise.
10916         (update_caller_keys): Likewise.
10917         (update_callee_keys): Likewise.
10918         (inline_small_functions): Likewise.
10919         (inline_to_all_callers_1): Likewise.
10920         (dump_overall_stats): Likewise.
10921         (early_inline_small_functions): Likewise.
10922         (early_inliner): Likewise.
10923         * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
10924         * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
10925         * ipa-pure-const.c (malloc_candidate_p): Likewise.
10926         * ipa-split.c (execute_split_functions): Likewise.
10927         * symbol-summary.h: Likewise.
10928         * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
10930 2018-06-19  Richard Biener  <rguenther@suse.de>
10932         * tree-vectorizer.c (try_vectorize_loop_1): Split out of ...
10933         (vectorize_loops): ... here.  Fix dbgcnt handling.
10934         (try_vectorize_loop): Wrap try_vectorize_loop_1.
10936 2018-06-19  Segher Boessenkool  <segher@kernel.crashing.org>
10938         PR target/86197
10939         * config/rs6000/rs6000.md (rs6000_discover_homogeneous_aggregate): An
10940         ieee128 argument takes up only one (vector) register, not two (floating
10941         point) registers.
10943 2018-06-19  Eric Botcazou  <ebotcazou@adacore.com>
10945         * gimplify.c (gimplify_init_constructor): Really never clear for an
10946         incomplete constructor if CONSTRUCTOR_NO_CLEARING is set.
10948 2018-06-19  Richard Biener  <rguenther@suse.de>
10950         PR tree-optimization/86179
10951         * tree-vect-patterns.c (vect_pattern_recog_1): Clean up
10952         after failed recognition.
10954 2018-06-18  Martin Sebor  <msebor@redhat.com>
10956         PR middle-end/85602
10957         * calls.c (maybe_warn_nonstring_arg): Handle strncat.
10958         * tree-ssa-strlen.c (is_strlen_related_p): Make extern.
10959         Handle integer subtraction.
10960         (maybe_diag_stxncpy_trunc): Handle nonstring source arguments.
10961         * tree-ssa-strlen.h (is_strlen_related_p): Declare.
10963 2018-06-18  David Malcolm  <dmalcolm@redhat.com>
10965         * config/frv/frv-protos.h (frv_ifcvt_modify_insn): Strengthen 3rd
10966         param from rtx to rtx_insn *.
10967         * config/frv/frv.c (frv_ifcvt_add_insn): Likewise for "insn"
10968         param.
10969         (frv_ifcvt_modify_insn): Likwise.
10970         (frv_ifcvt_modify_final): Likwise for local "existing_insn",
10971         adding an as_a <rtx_insn *> cast.  Likewise for local "insn".
10972         * config/mips/mips.c (r10k_insert_cache_barriers): Add an
10973         as_a <rtx_insn *> cast to local "unprotected_region" once
10974         it's been established that it's not NULL or pc_rtx.
10975         * config/nds32/nds32-relax-opt.c (nds32_group_insns): Strengthen
10976         param "sethi" from rtx to rtx_insn *.
10977         (nds32_group_float_insns): Likewise for param "insn".
10978         * config/vax/vax-protos.h (vax_output_int_add): Likewise for 1st
10979         param.
10980         (vax_output_int_subtract): Likewise.
10981         * config/vax/vax.c (vax_output_int_add): Likewise for param
10982         "insn".
10983         (vax_output_int_subtract): Likewise.
10984         * emit-rtl.c (set_insn_deleted): Likewise, removing cast.
10985         (emit_pattern_after): Likewise for param "after".
10986         (emit_insn_after): Likewise.
10987         (emit_jump_insn_after): Likewise.
10988         (emit_call_insn_after): Likewise.
10989         (emit_debug_insn_after): Likewise.
10990         (emit_pattern_before): Likewise for param "before".
10991         (emit_insn_before): Likewise.
10992         (emit_jump_insn_before): Likewise.
10993         * final.c (get_insn_template): Likewise for param "insn", removing
10994         a cast.
10995         * output.h (get_insn_template): Likewise for 2nd param.
10996         * rtl.h (emit_insn_before): Likewise.
10997         (emit_jump_insn_before): Likewise.
10998         (emit_debug_insn_before_noloc): Likewise.
10999         (emit_insn_after): Likewise.
11000         (emit_jump_insn_after): Likewise.
11001         (emit_call_insn_after): Likewise.
11002         (emit_debug_insn_after): Likewise.
11003         (set_insn_deleted): Likewise for param.
11005 2018-06-18  Michael Meissner  <meissner@linux.ibm.com>
11007         PR target/85358
11008         * config/rs6000/rs6000-modes.def (toplevel): Rework the 128-bit
11009         floating point modes, so that IFmode is numerically greater than
11010         TFmode, which is greater than KFmode using FRACTIONAL_FLOAT_MODE
11011         to declare the ordering.  This prevents IFmode from being
11012         converted to TFmode when long double is IEEE 128-bit on an ISA 3.0
11013         machine.  Include rs6000-modes.h to share the fractional values
11014         between genmodes* and the rest of the compiler.
11015         (IFmode): Likewise.
11016         (KFmode): Likewise.
11017         (TFmode): Likewise.
11018         * config/rs6000/rs6000-modes.h: New file.
11019         * config/rs6000/rs6000.c (rs6000_debug_reg_global): Change the
11020         meaning of rs6000_long_double_size so that 126..128 selects an
11021         appropriate 128-bit floating point type.
11022         (rs6000_option_override_internal): Likewise.
11023         * config/rs6000/rs6000.h (toplevel): Include rs6000-modes.h.
11024         (TARGET_LONG_DOUBLE_128): Change the meaning of
11025         rs6000_long_double_size so that 126..128 selects an appropriate
11026         128-bit floating point type.
11027         (LONG_DOUBLE_TYPE_SIZE): Update comment.
11028         * config/rs6000/rs6000.md (trunciftf2): Correct the modes of the
11029         source and destination to match the standard usage.
11030         (truncifkf2): Likewise.
11031         (copysign<mode>3, IEEE iterator): Rework copysign of float128 on
11032         ISA 2.07 to use an explicit clobber, instead of passing in a
11033         temporary.
11034         (copysign<mode>3_soft): Likewise.
11036 2018-06-18  David Malcolm  <dmalcolm@redhat.com>
11038         * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
11039         Replace dump_printf_loc call with DUMP_VECT_SCOPE.
11040         (vect_slp_analyze_instance_dependence): Likewise.
11041         (vect_enhance_data_refs_alignment): Likewise.
11042         (vect_analyze_data_refs_alignment): Likewise.
11043         (vect_slp_analyze_and_verify_instance_alignment
11044         (vect_analyze_data_ref_accesses): Likewise.
11045         (vect_prune_runtime_alias_test_list): Likewise.
11046         (vect_analyze_data_refs): Likewise.
11047         * tree-vect-loop-manip.c (vect_update_inits_of_drs): Likewise.
11048         * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
11049         (vect_analyze_scalar_cycles_1): Likewise.
11050         (vect_get_loop_niters): Likewise.
11051         (vect_analyze_loop_form_1): Likewise.
11052         (vect_update_vf_for_slp): Likewise.
11053         (vect_analyze_loop_operations): Likewise.
11054         (vect_analyze_loop): Likewise.
11055         (vectorizable_induction): Likewise.
11056         (vect_transform_loop): Likewise.
11057         * tree-vect-patterns.c (vect_pattern_recog): Likewise.
11058         * tree-vect-slp.c (vect_analyze_slp): Likewise.
11059         (vect_make_slp_decision): Likewise.
11060         (vect_detect_hybrid_slp): Likewise.
11061         (vect_slp_analyze_operations): Likewise.
11062         (vect_slp_bb): Likewise.
11063         * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Likewise.
11064         (vectorizable_bswap): Likewise.
11065         (vectorizable_call): Likewise.
11066         (vectorizable_simd_clone_call): Likewise.
11067         (vectorizable_conversion): Likewise.
11068         (vectorizable_assignment): Likewise.
11069         (vectorizable_shift): Likewise.
11070         (vectorizable_operation): Likewise.
11071         * tree-vectorizer.h (DUMP_VECT_SCOPE): New macro.
11073 2018-06-18  Martin Sebor  <msebor@redhat.com>
11075         PR tree-optimization/81384
11076         * builtin-types.def (BT_FN_SIZE_CONST_STRING_SIZE): New.
11077         * builtins.c (expand_builtin_strnlen): New function.
11078         (expand_builtin): Call it.
11079         (fold_builtin_n): Avoid setting TREE_NO_WARNING.
11080         * builtins.def (BUILT_IN_STRNLEN): New.
11081         * calls.c (maybe_warn_nonstring_arg): Handle BUILT_IN_STRNLEN.
11082         Warn for bounds in excess of maximum object size.
11083         * tree-ssa-strlen.c (maybe_set_strlen_range): Return tree representing
11084         single-value ranges.  Handle strnlen.
11085         (handle_builtin_strlen): Handle strnlen.
11086         (strlen_check_and_optimize_stmt): Same.
11087         * doc/extend.texi (Other Builtins): Document strnlen.
11089 2018-06-18  Maya Rashish  <coypu@sdf.org>
11091         * config/alpha/openbsd.h (TARGET_DEFAULT): Define.
11092         (LINK_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): Likewise.
11093         (INTMAX_TYPE, UINTMAX_TYPE, WINT_TYPE): Likewise.
11095         * config/alpha/elf.h (STARTFILE_SPEC, ENDFILE_SPEC): Move from
11096         here to ...
11097         * config/alpha/linux.h (STARTFILE_SPEC, ENDFILE_SPEC): Here.
11099 2018-06-18  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
11101         * tree.c (escaped_string::escape): Replace cast to char * by
11102         const_cast<char *> (unescaped).
11104 2018-06-18  Nick Clifton  <nickc@redhat.com>
11106         PR 84195
11107         * tree.c (escaped_string): New class.  Converts an unescaped
11108         string into its escaped equivalent.
11109         (warn_deprecated_use): Use the new class to convert the
11110         deprecation message, if present.
11111         (test_escaped_strings): New self test.
11112         (test_c_tests): Add test_escaped_strings.
11113         * doc/extend.texi (deprecated): Add a note that the
11114         deprecation message is affected by the -fmessage-length
11115         option, and that control characters will be escaped.
11116         (#pragma GCC error): Document this pragma.
11117         (#pragma GCC warning): Likewise.
11118         * doc/invoke.texi (-fmessage-length): Document this option's
11119         effect on the #warning and #error preprocessor directives and
11120         the deprecated attribute.
11122 2018-06-18  Eric Botcazou  <ebotcazou@adacore.com>
11124         * tree.c (decl_value_expr_lookup): Revert latest change.
11125         (decl_value_expr_insert): Likewise.
11127 2018-06-17  Eric Botcazou  <ebotcazou@adacore.com>
11129         * gimplify.c (nonlocal_vlas): Delete.
11130         (nonlocal_vla_vars): Likewise.
11131         (gimplify_var_or_parm_decl): Do not add debug VAR_DECLs for non-local
11132         referenced VLAs.
11133         (gimplify_body): Do not create and destroy nonlocal_vlas.
11134         * tree-nested.c: Include diagnostic.h.
11135         (use_pointer_in_frame): Tweak.
11136         (lookup_field_for_decl): Add assertion and declare the transformation.
11137         (convert_nonlocal_reference_op) <PARM_DECL>: Rework and issue an
11138         internal error when the reference is in a wrong context.  Do not
11139         create a debug decl by default.
11140         (note_nonlocal_block_vlas): Delete.
11141         (convert_nonlocal_reference_stmt) <GIMPLE_BIND>: Do not call it.
11142         (convert_local_reference_op) <PARM_DECL>: Skip the frame decl.  Do not
11143         create a debug decl by default.
11144         (convert_gimple_call) <GIMPLE_CALL>: Issue an internal error when the
11145         call is in a wrong context.
11146         (fixup_vla_decls): New function.
11147         (finalize_nesting_tree_1): Adjust comment.  Call fixup_vla_decls if no
11148         debug variables were created.
11149         * tree.c (decl_value_expr_lookup): Add checking assertion.
11150         (decl_value_expr_insert): Likewise.
11152 2018-06-16  Kugan Vivekanandarajah  <kuganv@linaro.org>
11154         PR middle-end/82479
11155         * ipa-fnsummary.c (will_be_nonconstant_expr_predicate): Handle CALL_EXPR.
11156         * tree-scalar-evolution.c (interpret_expr): Likewise.
11157         (expression_expensive_p): Likewise.
11158         * tree-ssa-loop-ivopts.c (contains_abnormal_ssa_name_p): Likewise.
11159         * tree-ssa-loop-niter.c (number_of_iterations_popcount): New.
11160         (number_of_iterations_exit_assumptions): Use number_of_iterations_popcount.
11161         (ssa_defined_by_minus_one_stmt_p): New.
11163 2018-06-16  Kugan Vivekanandarajah  <kuganv@linaro.org>
11165         PR middle-end/64946
11166         * cfgexpand.c (expand_debug_expr): Hande ABSU_EXPR.
11167         * config/i386/i386.c (ix86_add_stmt_cost): Likewise.
11168         * dojump.c (do_jump): Likewise.
11169         * expr.c (expand_expr_real_2): Check operand type's sign.
11170         * fold-const.c (const_unop): Handle ABSU_EXPR.
11171         (fold_abs_const): Likewise.
11172         * gimple-pretty-print.c (dump_unary_rhs): Likewise.
11173         * gimple-ssa-backprop.c (backprop::process_assign_use): Likesie.
11174         (strip_sign_op_1): Likesise.
11175         * match.pd: Add new pattern to generate ABSU_EXPR.
11176         * optabs-tree.c (optab_for_tree_code): Handle ABSU_EXPR.
11177         * tree-cfg.c (verify_gimple_assign_unary): Likewise.
11178         * tree-eh.c (operation_could_trap_helper_p): Likewise.
11179         * tree-inline.c (estimate_operator_cost): Likewise.
11180         * tree-pretty-print.c (dump_generic_node): Likewise.
11181         * tree-vect-patterns.c (vect_recog_sad_pattern): Likewise.
11182         * tree.def (ABSU_EXPR): New.
11184 2018-06-16  Jakub Jelinek  <jakub@redhat.com>
11186         PR middle-end/86095
11187         * common.opt (Wunsafe-loop-optimizations): Add Ignore, remove Var,
11188         documented as preserved for backward compatibility only.
11189         * doc/invoke.texi: Remove -Wunsafe-loop-optimizations documentation.
11191         PR rtl-optimization/86108
11192         * bb-reorder.c (create_forwarder_block): Renamed to ...
11193         (create_eh_forwarder_block): ... this.  Split OLD_BB after labels and
11194         jump from new landing pad to the second part.
11195         (sjlj_fix_up_crossing_landing_pad, dw2_fix_up_crossing_landing_pad):
11196         Adjust callers.
11198 2018-06-15  Jakub Jelinek  <jakub@redhat.com>
11200         PR middle-end/85878
11201         * expr.c (expand_assignment): Remove now redundant COMPLEX_MODE_P
11202         check from first store_expr, use to_mode instead of GET_MODE (to_rtx).
11203         Only call store_expr for halves if the mode is the same.
11205         PR middle-end/86123
11206         * match.pd ((X / Y) == 0 -> X < Y): Don't transform complex divisions.
11207         Fix up comment formatting.
11209 2018-06-15  Bernd Edlinger  <bernd.edlinger@hotmail.de>
11211         * typed-splay-tree.h (typed_splay_tree::remove): New function.
11212         (typed_splay_tree::closure,
11213         typed_splay_tree::inner_foreach_fn, typed_splay_tree::m_inner): Deleted.
11214         (typed_splay_tree::typed_splay_tree,
11215         typed_splay_tree::operator =): Declared private.
11216         (typed_splay_tree::splay_tree_key, typed_splay_tree::splay_tree_value,
11217         typed_splay_tree::splay_tree_node_s, typed_splay_tree::KDEL,
11218         typed_splay_tree::VDEL, typed_splay_tree::splay_tree_delete_helper,
11219         typed_splay_tree::rotate_left, typed_splay_tree::rotate_right,
11220         typed_splay_tree::splay_tree_splay,
11221         typed_splay_tree::splay_tree_foreach_helper,
11222         typed_splay_tree::splay_tree_insert,
11223         typed_splay_tree::splay_tree_remove,
11224         typed_splay_tree::splay_tree_lookup,
11225         typed_splay_tree::splay_tree_predecessor,
11226         typed_splay_tree::splay_tree_successor,
11227         typed_splay_tree::splay_tree_min,
11228         typed_splay_tree::splay_tree_max): Took over from splay-tree.c/.h.
11229         (typed_splay_tree::root, typed_splay_tree::comp,
11230         typed_splay_tree::delete_key,
11231         typed_splay_tree::delete_value): New data members.
11232         * typed-splay-tree.c (selftest::test_str_to_int): Add a test for
11233         typed_splay_tree::remove.
11235 2018-06-15  Matthew Fortune  <matthew.fortune@mips.com>
11237         * config/mips/mips.h (ASM_SPEC): Pass through -mcrc, -mno-crc,
11238         -mginv and -mno-ginv to the assembler.
11239         * config/mips/mips.opt (-mcrc): New option.
11240         (-mginv): Likewise.
11241         * doc/invoke.text (-mcrc): Document.
11242         (-mginv): Likewise.
11244 2018-06-15  Nick Clifton  <nickc@redhat.com>
11246         PR 84195
11247         * tree.c (escaped_string): New class.  Converts an unescaped
11248         string into its escaped equivalent.
11249         (warn_deprecated_use): Use the new class to convert the
11250         deprecation message, if present.
11251         (test_escaped_strings): New self test.
11252         (test_c_tests): Add test_escaped_strings.
11253         * doc/extend.texi (deprecated): Add a note that the
11254         deprecation message is affected by the -fmessage-length
11255         option, and that control characters will be escaped.
11256         (#pragma GCC error): Document this pragma.
11257         (#pragma GCC warning): Likewise.
11258         * doc/invoke.texi (-fmessage-length): Document this option's
11259         effect on the #warning and #error preprocessor directives and
11260         the deprecated attribute.
11262 2018-06-15  Richard Biener  <rguenther@suse.de>
11264         * tree-vect-slp.c (vect_slp_bb): Dump MSG_OPTIMIZED_LOCATIONS
11265         here, also noting vector size used.
11266         * tree-vectorizer.c (vectorize_loops): Adjust.  Note vector
11267         size used in MSG_OPTIMIZED_LOCATIONS dump.
11268         (pass_slp_vectorize::execute): Adjust.
11270 2018-06-15  Claudiu Zissulescu  <claziss@synopsys.com>
11272         PR target/85968
11273         * config/arc/arc.c (arc_return_address_register): Fix
11274         if-condition.
11276 2018-06-15  Richard Biener  <rguenther@suse.de>
11278         PR middle-end/86159
11279         * tree-cfg.c (gimplify_build3): Do not strip sign conversions,
11280         leave useless conversion stripping to force_gimple_operand_gsi.
11281         (gimplify_build2): Likewise.
11282         (gimplify_build1): Likewise.
11284 2018-06-15  Richard Biener  <rguenther@suse.de>
11286         PR middle-end/86076
11287         * tree-cfg.c (move_stmt_op): unshare invariant addresses
11288         before adjusting their block.
11290 2018-06-15  Sebastian Huber  <sebastian.huber@embedded-brains.de>
11292         * config.gcc (riscv*-*-elf* | riscv*-*-rtems*): Use custom
11293         multilibs for *-*-rtems*.
11294         * config/riscv/t-rtems: New file.
11296 2018-06-14  Jakub Jelinek  <jakub@redhat.com>
11298         PR middle-end/86122
11299         * match.pd ((A +- CST1) +- CST2): Punt if last resort
11300         unsigned_type_for returns NULL.
11302         PR target/85945
11303         * lower-subreg.c (find_decomposable_subregs): Don't decompose float
11304         subregs of multi-word pseudos unless the float mode has word size.
11306 2018-06-14  Richard Biener  <rguenther@suse.de>
11308         PR middle-end/86139
11309         * tree-vect-generic.c (build_word_mode_vector_type): Remove
11310         duplicate and harmful type_hash_canon.
11311         * tree.c (type_hash_canon): Assert we didn't find ourselves.
11313 2018-06-14  Richard Biener  <rguenther@suse.de>
11315         PR ipa/86124
11316         * tree-ssa-struct-alias.c (create_variable_info_for): Handle
11317         NULL cgraph_node.
11319 2018-06-14  Sebastian Huber  <sebastian.huber@embedded-brains.de>
11321         * config/rtems.h (STDINT_LONG32): Define.
11323 2018-06-13  Matthew Fortune  <matthew.fortune@mips.com>
11324             Prachi Godbole  <prachi.godbole@imgtec.com>
11326         * config/mips/mips-cpus.def: Define P6600.
11327         * config/mips/mips-tables.opt: Regenerate.
11328         * config/mips/mips.c (mips_ucbranch_type): New enum.
11329         (mips_rtx_cost_data): Add support for P6600.
11330         (mips_issue_rate): Likewise.
11331         (mips_multipass_dfa_lookahead): Likewise.
11332         (mips_avoid_hazard): Likewise.
11333         (mips_reorg_process_insns): Likewise.
11334         (mips_classify_branch_p6600): New function.
11335         * config/mips/mips.h (TUNE_P6600): New define.
11336         (MIPS_ISA_LEVEL_SPEC): Infer mips64r6 from p6600.
11337         (ENABLE_LD_ST_PAIRS): Enable load/store bonding for p6600.
11338         * config/mips/mips.md: Include p6600.md.
11339         (processor): Add p6600.
11340         * config/mips/p6600.md: New file.
11341         * doc/invoke.texi: Add p6600 to supported architectures.
11343 2018-06-13  Martin Sebor  <msebor@redhat.com>
11345         PR tree-optimization/86114
11346         * gimple-fold.c (gimple_fold_builtin_strlen): Only handle LHS
11347         of integer types.
11348         * tree-ssa-strlen.c (maybe_set_strlen_range): Same.
11350 2018-06-13  Richard Biener  <rguenther@suse.de>
11352         * tree-vect-patterns.c (vect_recog_vector_vector_shift_pattern):
11353         Properly set vector type of the intermediate stmt.
11354         * tree-vect-stmts.c (vectorizable_operation): The destination
11355         var always has vectype_out type.
11357 2018-06-13  Jeff Law  <law@redhat.com>
11359         * config/rl78/rl78.c (move_elim_pass): Use TDF_NONE rather than
11360         integer 0 for argument to print_rtl_with_bb.
11361         (rl78_reorg): Likewise.
11363 2018-06-13  David Malcolm  <dmalcolm@redhat.com>
11365         * config/arc/arc.c (hwloop_optimize): Strengthen local "end_label"
11366         from rtx to rtx_insn *.
11367         * config/bfin/bfin.c (hwloop_optimize): Likewise for local
11368         "label".
11369         (add_sched_insns_for_speculation): Likewise for local "target",
11370         converting usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
11371         * config/c6x/c6x.c (reorg_split_calls): Strengthen param "call_labels"
11372         from rtx_insn ** to rtx_code_label **.
11373         (reorg_emit_nops): Likewise.
11374         (c6x_reorg): Likewise for local "call_labels".
11375         * config/sh/sh-protos.h (get_dest_uid): Strengthen 1st param from
11376         rtx to rtx_insn *.
11377         * config/sh/sh.c (dump_table): Strengthen local "lab" from rtx to
11378         rtx_code_label *, adding safe_as_a <rtx_code_label *> casts to
11379         the loops over LABEL_REFS.
11380         (fixup_addr_diff_vecs): Add as_a <rtx_insn *> to usage of
11381         braf_label.
11382         (barrier_align): Convert usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
11383         (get_dest_uid): Strengthen param "label" from rtx to rtx_insn *.
11384         (split_branches): Strengthen local "olabel" from rtx to
11385         rtx_insn *, adding a safe_as_a cast.
11386         * emit-rtl.c (next_real_insn): Strengthen param from "rtx"
11387         to "rtx_insn *".
11388         (add_insn_after): Likewise for first two params.
11389         (add_insn_before): Likewise.
11390         (remove_insn): Likewise for param.
11391         (emit_pattern_before_noloc): Likewise for second and third params.
11392         (emit_jump_insn_before_noloc): Convert NULL_RTX to NULL.
11393         (emit_call_insn_before_noloc): Likewise.
11394         (emit_debug_insn_before_noloc): Strengthen "before" param from "rtx"
11395         to "rtx_insn *".
11396         (emit_barrier_before): Likewise.
11397         (emit_label_before): Strengthen "label" param from "rtx" to
11398         "rtx_code_label *".  Strengthen "before" param from "rtx" to
11399         "rtx_insn *".
11400         (emit_insn_after_1): Strengthen "after" param from "rtx" to
11401         "rtx_insn *".
11402         (emit_pattern_after_noloc): Likewise.
11403         (emit_insn_after_noloc): Likewise.
11404         (emit_jump_insn_after_noloc): Likewise.
11405         (emit_call_insn_after_noloc): Likewise.
11406         (emit_debug_insn_after_noloc): Likewise.
11407         (emit_barrier_after): Likewise.
11408         (emit_label_after): Likewise for both params.
11409         (emit_pattern_after_setloc): Likewise for "after" param.  Convert
11410         "loc" param from "int" to "location_t".
11411         (emit_insn_after_setloc): Likewise.
11412         (emit_jump_insn_after_setloc): Likewise.
11413         (emit_call_insn_after_setloc): Likewise.
11414         (emit_debug_insn_after_setloc): Likewise.
11415         (emit_pattern_before_setloc): Likewise for "before" param.  Convert
11416         "loc" param from "int" to "location_t".
11417         (emit_pattern_before): Convert NULL_RTX to NULL.
11418         (emit_insn_before_setloc): Convert "loc" param from "int" to
11419         "location_t".
11420         (emit_jump_insn_before_setloc): Likewise.
11421         (emit_call_insn_before_setloc): Likewise.
11422         (emit_debug_insn_before_setloc): Strengthen "before" param from rtx to
11423         rtx_insn *.  Convert "loc" param from "int" to "location_t".
11424         * rtl.h (emit_insn_before_setloc, emit_jump_insn_before_setloc,
11425         emit_call_insn_before_setloc, emit_debug_insn_before_setloc):
11426         Convert 3rd param from "int" to "location_t".
11427         (emit_barrier_before, emit_barrier_after, next_real_insn):
11428         Strengthen param from rtx to rtx_insn *.
11429         (emit_label_before): Strengthen 1st param from "rtx" to
11430         "rtx_code_label *".  Strengthen 2nd param from "rtx" to
11431         "rtx_insn *".
11432         (emit_insn_after_noloc, emit_jump_insn_after_noloc,
11433         emit_call_insn_after_noloc, emit_debug_insn_after_noloc):
11434         Strengthen 2nd param from "rtx" to "rtx_insn *".
11435         (emit_insn_after_setloc, emit_jump_insn_after_setloc)
11436         emit_call_insn_after_setloc, emit_debug_insn_after_setloc):
11437         Likewise. Convert 3rd param from "int" to "location_t".
11438         (emit_label_after): Strengthen 1st param from "rtx" to
11439         "rtx_code_label *".
11440         (next_real_insn, remove_insn): Strengthen param from "rtx" to
11441         "rtx_insn *".
11442         (add_insn_before, add_insn_after): Strengthen 1st and 2nd params
11443         from "rtx" to "rtx_insn *".
11445 2018-06-13  Jan Hubicka  <hubicka@gcc.gnu.org>
11447         * cgraph.c (cgraph_node::get_untransformed_body): Dump function
11448         bodies streamed in with -Q.
11449         * dumpfile.c (dump_files): Add lto-stream-out dump file.
11450         * dumpfile.h (tree_dump_index): Add lto_stream_out.
11451         * gimple-streamer-out.c: Include gimple-pretty-print.h
11452         (output_bb): Dump stmts streamed.
11453         * lto-section-out.c: Include print-tree.h
11454         (lto_begin_section): Dump sections created.
11455         (lto_output_decl_index): Dump decl encoded.
11456         * lto-streamer-out.c: Include print-tree.h
11457         (create_output_block): Dump output block created.
11458         (DFS::DFS_write_tree_body): Dump DFS SCCs streamed.
11459         (output_function): Dump function output.
11460         (output_constructor): Dump constructor streamed.
11461         (write_global_stream): Output indexes encoded.
11462         (produce_asm_for_decls): Dump streams encoded.
11463         * lto-streamer.c (streamer_dump_file): New global var.
11464         * lto-streamer.h (streamer_dump_file): Declare.
11465         * passes.c (ipa_write_summaries): Initialize streamer dump.
11466         * varpool.c (varpool_node::get_constructor): Dump constructors streamed
11467         in.
11469 2018-06-13  Eric Botcazou  <ebotcazou@adacore.com>
11471         PR target/86048
11472         * config/i386/winnt.c (i386_pe_seh_cold_init): Do not emit negative
11473         offsets for register save directives.  Emit a second batch of save
11474         directives, if need be, when the function accesses prior frames.
11476 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
11478         * config/arc/fpu.md (fmasf4): Force operand to register.
11479         (fnmasf4): Likewise.
11481 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
11483         * config/arc/arc-protos.h (arc_pad_return): Remove.
11484         * config/arc/arc.c (machine_function): Remove force_short_suffix
11485         and size_reason.
11486         (arc_print_operand): Adjust printing of '&'.
11487         (arc_verify_short): Remove conditional printing of short suffix.
11488         (arc_final_prescan_insn): Remove reference to size_reason.
11489         (pad_return): New function.
11490         (arc_reorg): Call pad_return.
11491         (arc_pad_return): Remove.
11492         (arc_init_machine_status): Remove reference to force_short_suffix.
11493         * config/arc/arc.md (vunspec): Add VUNSPEC_ARC_BLOCKAGE.
11494         (attr length): When attribute iscompact is true force to 2
11495         regardless; in the case of maybe check if we want to force the
11496         instruction to have 4 bytes length.
11497         (nopv): Change it to generate 4 byte long nop as well.
11498         (blockage): New pattern.
11499         (simple_return): Remove call to arc_pad_return.
11500         (p_return_i): Likewise.
11502 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
11504         * config/arc/elf.h (LINK_GCC_C_SEQUENCE_SPEC): Define.
11506 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
11508         * config/arc/builtins.def (SYNC): SYNC instruction is valid on all
11509         ARC cores.
11511 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
11513         * config/arc/arc.c (atomic_exchangesi): EX instruction is default
11514         for ARC700 and ARCv2.
11516 2018-06-13  Chenghua Xu <paul.hua.gm@gmail.com>
11518         PR target/86076
11519         * config/mips/loongson.md (vec_setv4hi): Gen_lowpart for
11520         operands[2] instead of operands[1].
11523 2018-06-12  Richard Sandiford  <richard.sandiford@linaro.org>
11525         * lra-constraints.c (simplify_operand_subreg): In the paradoxical
11526         case, check whether the outer register overlaps an unallocatable
11527         register, not just whether it fits the required class.
11529 2018-06-12  Richard Sandiford  <richard.sandiford@linaro.org>
11531         * poly-int.h (can_div_trunc_p): Add new overload in which all values
11532         are poly_ints.
11533         * alias.c (get_addr): Extend CONST_INT handling to poly_int_rtx_p.
11534         (memrefs_conflict_p): Likewise.
11535         (init_alias_analysis): Likewise.
11536         * cfgexpand.c (expand_debug_expr): Likewise.
11537         * combine.c (combine_simplify_rtx, force_int_to_mode): Likewise.
11538         * cse.c (fold_rtx): Likewise.
11539         * explow.c (adjust_stack, anti_adjust_stack): Likewise.
11540         * expr.c (emit_block_move_hints): Likewise.
11541         (clear_storage_hints, push_block, emit_push_insn): Likewise.
11542         (store_expr_with_bounds, reduce_to_bit_field_precision): Likewise.
11543         (emit_group_load_1): Use rtx_to_poly_int64 for group offsets.
11544         (emit_group_store): Likewise.
11545         (find_args_size_adjust): Use strip_offset.  Use rtx_to_poly_int64
11546         to read the PRE/POST_MODIFY increment.
11547         * calls.c (store_one_arg): Use strip_offset.
11548         * rtlanal.c (rtx_addr_can_trap_p_1): Extend CONST_INT handling to
11549         poly_int_rtx_p.
11550         (set_noop_p): Use rtx_to_poly_int64 for the elements selected
11551         by a VEC_SELECT.
11552         * simplify-rtx.c (avoid_constant_pool_reference): Use strip_offset.
11553         (simplify_binary_operation_1): Extend CONST_INT handling to
11554         poly_int_rtx_p.
11555         * var-tracking.c (compute_cfa_pointer): Take a poly_int64 rather
11556         than a HOST_WIDE_INT.
11557         (hard_frame_pointer_adjustment): Change from HOST_WIDE_INT to
11558         poly_int64.
11559         (adjust_mems, add_stores): Update accodingly.
11560         (vt_canonicalize_addr): Track polynomial offsets.
11561         (emit_note_insn_var_location): Likewise.
11562         (vt_add_function_parameter): Likewise.
11563         (vt_initialize): Likewise.
11565 2018-06-12  Jeff Law  <law@redhat.com>
11567         * config.gcc (alpha*-*-freebsd*): Remove.
11568         * config/alpha/freebsd.h: Remove.
11570 2018-06-12  David Malcolm  <dmalcolm@redhat.com>
11572         PR other/69968
11573         * spellcheck-tree.c (levenshtein_distance): Rename to...
11574         (get_edit_distance): ...this, and update for underlying renaming.
11575         * spellcheck-tree.h (levenshtein_distance): Rename to...
11576         (get_edit_distance): ...this.
11577         * spellcheck.c (levenshtein_distance): Rename to...
11578         (get_edit_distance): ...this.  Convert from Levenshtein distance
11579         to Damerau-Levenshtein distance by supporting transpositions of
11580         adjacent characters.  Rename "v1" to "v_next" and "v0" to
11581         "v_one_ago".
11582         (selftest::levenshtein_distance_unit_test_oneway): Rename to...
11583         (selftest::test_edit_distance_unit_test_oneway): ...this, and
11584         update for underlying renaming.
11585         (selftest::levenshtein_distance_unit_test): Rename to...
11586         (selftest::test_get_edit_distance_unit): ...this, and update for
11587         underlying renaming.
11588         (selftest::test_find_closest_string): Add example from PR 69968
11589         where transposition helps
11590         (selftest::test_metric_conditions): Update for renaming.
11591         (selftest::test_metric_conditions): Likewise.
11592         (selftest::spellcheck_c_tests): Likewise.
11593         * spellcheck.h (levenshtein_distance): Rename both overloads to...
11594         (get_edit_distance): ...this.
11595         (best_match::consider): Update for renaming.
11597 2018-06-12  Martin Sebor  <msebor@redhat.com>
11599         PR tree-optimization/85259
11600         * builtins.c (compute_objsize): Handle constant offsets.
11601         * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Return
11602         true iff a warning has been issued.
11603         * gimple.h (gimple_nonartificial_location): New function.
11604         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Call
11605         gimple_nonartificial_location and handle -Wno-system-headers.
11606         (handle_builtin_stxncpy): Same.
11608 2018-06-12  Martin Sebor  <msebor@redhat.com>
11610         PR c/85931
11611         * fold-const.c (operand_equal_p): Handle SAVE_EXPR.
11613 2018-06-12  Will Schmidt  <will_schmidt@vnet.ibm.com>
11615         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
11616         BUILTIN_VEC_XST entries for pointer to double and long long.
11618 2018-06-12  H.J. Lu  <hongjiu.lu@intel.com>
11620         PR target/85990
11621         * config/i386/gnu-user.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
11622         Update comments.
11623         * config/i386/gnu-user64.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
11624         Likewise.
11626 2018-06-12  Martin Liska  <mliska@suse.cz>
11628         * doc/options.texi: Document IntegerRange.
11630 2018-06-12  Martin Liska  <mliska@suse.cz>
11632         * config/i386/i386.opt: Make MPX-related options as Deprecated.
11633         * opt-functions.awk: Handle Deprecated flag.
11634         * opts-common.c (decode_cmdline_option): Handle cl_deprecated
11635         and report error.
11636         (read_cmdline_option): Report warning for a deprecated option.
11637         * opts.h (struct cl_option): Add new field cl_deprecated.
11638         (CL_ERR_DEPRECATED): New.
11640 2018-06-12  Martin Liska  <mliska@suse.cz>
11642         * doc/options.texi: Document Deprecated option flag.
11644 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
11646         * config/arc/arc-arch.h (arc_extras): New enum.
11647         (arc_cpu_t):Add field extra.
11648         (arc_cpu_types): Consider the extras.
11649         * config/arc/arc-cpus.def: Add extras info.
11650         * config/arc/arc-opts.h (processor_type): Consider extra field.
11651         * config/arc/arc.c (arc_override_options): Handle extra field.
11653 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
11655         * config/arc/arc-arch.h: Update ARC_OPTX macro.
11656         * config/arc/arc-options.def (ARC_OPTX): Introduce a new doc
11657         field.
11658         * config/arc/arc.c (arc_init): Update pic warning.
11659         (irq_range): Update irq range parsing warnings.
11660         (arc_override_options): Update various warning messages.
11661         (arc_handle_aux_attribute): Likewise.
11663 2018-06-12  Robert Suchanek  <robert.suchanek@mips.com>
11665         * config/mips/i6400.md (i6400_fpu_fadd): Remove frint.
11667 2018-06-12  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
11669         * doc/sourcebuild.texi: Document usage of line number 0 in verify
11670         compiler messages directives.
11672 2018-06-12  Matthew Fortune  <mfortune@gmail.com>
11674         * config/mips/mips-cpus.def: New MIPS_CPU for i6500.
11675         * config/mips/mips-tables.opt: Regenerate.
11676         * config/mips/mips.h (MIPS_ISA_LEVEL_SPEC): Mark i6500 as
11677         mips64r6.
11678         * doc/invoke.texi: Document -march=i6500.
11680 2018-06-12  Prachi Godbole  <prachi.godbole@imgtec.com>
11682         * config/mips/i6400.md (i6400_gpmuldiv): Remove cpu_unit.
11683         (i6400_gpmul): Add cpu_unit.
11684         (i6400_gpdiv): Likewise.
11685         (i6400_msa_add_d): Update reservations.
11686         (i6400_msa_int_add) Likewise.
11687         (i6400_msa_short_logic3) Likewise.
11688         (i6400_msa_short_logic2) Likewise.
11689         (i6400_msa_short_logic) Likewise.
11690         (i6400_msa_move) Likewise.
11691         (i6400_msa_cmp) Likewise.
11692         (i6400_msa_short_float2) Likewise.
11693         (i6400_msa_div_d) Likewise.
11694         (i6400_msa_long_logic1) Likewise.
11695         (i6400_msa_long_logic2) Likewise.
11696         (i6400_msa_mult) Likewise.
11697         (i6400_msa_long_float2) Likewise.
11698         (i6400_msa_long_float4) Likewise.
11699         (i6400_msa_long_float5) Likewise.
11700         (i6400_msa_long_float8) Likewise.
11701         (i6400_fpu_fadd): Include frint type.
11702         (i6400_fpu_store): New define_insn_reservation.
11703         (i6400_fpu_load): Likewise.
11704         (i6400_fpu_move): Likewise.
11705         (i6400_fpu_fcmp): Likewise.
11706         (i6400_fpu_fmadd): Likewise.
11707         (i6400_int_mult): Include imul3nc type and update reservation.
11708         (i6400_int_div): Include idiv3 type and update reservation.
11709         (i6400_int_load): Update to check type not move_type.
11710         (i6400_int_store): Likewise.
11711         (i6400_int_prefetch): Set zero latency.
11713 2018-06-12  Eric Botcazou  <ebotcazou@adacore.com>
11715         * gcc.c: Document new %@{...} sequence.
11716         (LINK_COMMAND_SPEC): Use it for the -L switches.
11717         (cpp_unique_options): Use it for the -I switches.
11718         (at_file_argbuf): New global variable.
11719         (in_at_file): Likewise.
11720         (alloc_args): Create at_file_argbuf.
11721         (clear_args): Truncate at_file_argbuf.
11722         (store_arg): If in_at_file, push the argument onto at_file_argbuf.
11723         (open_at_file): New function.
11724         (close_at_file): Likewise.
11725         (create_at_file): Delete.
11726         (do_spec_1) <'i'>: Use open_at_file/close_at_file.
11727         <'o'>: Likewise.
11728         <'@'>: New case.
11729         (validate_switches_from_spec): Deal with %@{...} sequence.
11730         (validate_switches): Likewise.
11731         (driver::finalize): Call clear_args.
11733 2018-06-11  Rasmus Villemoes  <rasmus.villemoes@prevas.dk>
11735         * config/vx-common.h (USE_TM_CLONE_REGISTRY): #define to 0.
11737 2018-06-11  Martin Sebor  <msebor@redhat.com>
11739         * doc/invoke.texi (-Wall): List -Wc++17-compat.
11740         (Wno-class-memaccess): Add @opindex.
11741         (Wno-templates, Wno-multiple-inheritance): Same.
11742         (Wno-virtual-inheritance, Wno-namespaces): Same.
11743         (Wno-pedantic, Wno-chkp, Wcoverage-mismatch): Same.
11744         (Wno-format-overflow, Wno-format-truncation): Same.
11745         (Wno-unused-local-typedefs, Walloc-size-larger-than): Same.
11746         (Wno-alloc-size-larger-than, Wframe-larger-than): Same
11747         (Wno-stack-usage, Wno-c++-compat, Wno-c++11-compat): Same.
11748         (Wno-c++14-compat, Wno-c++17-compat, Wno-openmp-simd): Same.
11749         (Wno-unsuffixed-float-constants, Wno-addr-space-convert): Same.
11750         (Wno-misspelled-isr): Same.
11752 2018-06-11  Martin Sebor  <msebor@redhat.com>
11754         * PR tree-optimization/86083
11755         * tree-ssa-strlen.c (handle_char_store): Use tree_expr_nonzero_p.
11757 2018-06-11  Zhouyi Zhou <zhouzhouyi@gmail.com>
11759         * tree-eh.c (lower_eh_constructs_2): Add a comma to comment.
11761 2018-06-11  Segher Boessenkool  <segher@kernel.crashing.org>
11763         PR target/85755
11764         * config/rs6000/rs6000.md (*movdi_internal32): Put constraint modifiers
11765         on the correct operand.
11766         (*movdi_internal64): Ditto.
11768 2018-06-11  Martin Liska  <mliska@suse.cz>
11770         PR tree-optimization/86089
11771         * tree-ssa-strlen.c (get_string_length): Move back removed hunk.
11773 2018-06-11  Julia Koval  <julia.koval@intel.com>
11775         * config/i386/vaesintrin.h (_mm_aesdec_epi128, _mm_aesdeclast_epi128)
11776         _mm_aesenc_epi128, _mm_aesenclast_epi128: Remove.
11777         * config/i386/vpclmulqdqintrin.h (_mm_clmulepi64_epi128): Remove.
11779 2018-06-11  Olivier Hainque  <hainque@adacore.com>
11781         * dwarf2out.c (gen_compile_unit_die): Fallout to DW_LANG_Ada83
11782         for Ada with strict dwarf2.
11784 2018-06-08  Peter Bergner  <bergner@vnet.ibm.com>
11786         PR target/85755
11787         * config/rs6000/rs6000.c (mem_operand_gpr): Enable PRE_INC and PRE_DEC
11788         addresses.
11790 2018-06-08  Jan Hubicka  <hubicka@ucw.cz>
11792         * dumpfile.c (FIRST_ME_AUTO_NUMBERED_DUMP): Bump to 4.
11794 2018-06-08  David Edelsohn  <dje.gcc@gmail.com>
11796         * config/rs6000/rs6000.c (rs6000_passes_ieee128): Protect with #if
11797         TARGET_ELF.
11799 2018-06-08  Martin Liska  <mliska@suse.cz>
11801         * tree-cfg.h (debug_function): Fix argument type to match
11802         implementation.
11804 2018-06-08  Martin Liska  <mliska@suse.cz>
11806         * config/powerpcspe/powerpcspe.c (rs6000_xcoff_visibility):
11807         Remove usage of MPX-related (and removed) fields.
11808         * config/rs6000/rs6000.c (rs6000_xcoff_visibility): Likewise.
11810 2018-06-08  David Malcolm  <dmalcolm@redhat.com>
11812         * cfg.c (debug): Use TDF_NONE rather than 0.
11813         * cfghooks.c (debug): Likewise.
11814         * dumpfile.c (DUMP_FILE_INFO): Likewise; also for OPTGROUP.
11815         (struct dump_option_value_info): Convert to...
11816         (struct kv_pair): ...this template type.
11817         (dump_options): Convert to kv_pair<dump_flags_t>; use TDF_NONE
11818         rather than 0.
11819         (optinfo_verbosity_options): Likewise.
11820         (optgroup_options): Convert to kv_pair<optgroup_flags_t>; use
11821         OPTGROUP_NONE.
11822         (gcc::dump_manager::dump_register): Use optgroup_flags_t rather
11823         than int for "optgroup_flags" param.
11824         (dump_generic_expr_loc): Use dump_flags_t rather than int for
11825         "dump_kind" param.
11826         (dump_dec): Likewise.
11827         (dump_finish): Use TDF_NONE rather than 0.
11828         (gcc::dump_manager::opt_info_enable_passes): Use optgroup_flags_t
11829         rather than int for "optgroup_flags" param.  Use TDF_NONE rather
11830         than 0.  Update for change to option_ptr.
11831         (opt_info_switch_p_1): Convert "optgroup_flags" param from int *
11832         to optgroup_flags_t *.  Use TDF_NONE and OPTGROUP_NONE rather than
11833         0.  Update for changes to optinfo_verbosity_options and
11834         optgroup_options.
11835         (opt_info_switch_p): Convert optgroup_flags from int to
11836         optgroup_flags_t.
11837         (dump_basic_block): Use dump_flags_t rather than int
11838         for "dump_kind" param.
11839         * dumpfile.h (TDF_ADDRESS, TDF_SLIM, TDF_RAW, TDF_DETAILS,
11840         TDF_STATS, TDF_BLOCKS, TDF_VOPS, TDF_LINENO, TDF_UID)
11841         TDF_STMTADDR, TDF_GRAPH, TDF_MEMSYMS, TDF_RHS_ONLY, TDF_ASMNAME,
11842         TDF_EH, TDF_NOUID, TDF_ALIAS, TDF_ENUMERATE_LOCALS, TDF_CSELIB,
11843         TDF_SCEV, TDF_GIMPLE, TDF_FOLDING, MSG_OPTIMIZED_LOCATIONS,
11844         MSG_MISSED_OPTIMIZATION, MSG_NOTE, MSG_ALL, TDF_COMPARE_DEBUG,
11845         TDF_NONE): Convert from macros to...
11846         (enum dump_flag): ...this new enum.
11847         (dump_flags_t): Update to use enum.
11848         (operator|, operator&, operator~, operator|=, operator&=):
11849         Implement for dump_flags_t.
11850         (OPTGROUP_NONE, OPTGROUP_IPA, OPTGROUP_LOOP, OPTGROUP_INLINE,
11851         OPTGROUP_OMP, OPTGROUP_VEC, OPTGROUP_OTHER, OPTGROUP_ALL):
11852         Convert from macros to...
11853         (enum optgroup_flag): ...this new enum.
11854         (optgroup_flags_t): New typedef.
11855         (operator|, operator|=): Implement for optgroup_flags_t.
11856         (struct dump_file_info): Convert field "alt_flags" to
11857         dump_flags_t.  Convert field "optgroup_flags" to
11858         optgroup_flags_t.
11859         (dump_basic_block): Use dump_flags_t rather than int for param.
11860         (dump_generic_expr_loc): Likewise.
11861         (dump_dec): Likewise.
11862         (dump_register): Convert param "optgroup_flags" to
11863         optgroup_flags_t.
11864         (opt_info_enable_passes): Likewise.
11865         * early-remat.c (early_remat::dump_edge_list): Use TDF_NONE rather
11866         than 0.
11867         * gimple-pretty-print.c (debug): Likewise.
11868         * gimple-ssa-store-merging.c (bswap_replace): Likewise.
11869         (merged_store_group::apply_stores): Likewise.
11870         * gimple-ssa-strength-reduction.c (insert_initializers): Likewise.
11871         * gimple.c (verify_gimple_pp): Likewise.
11872         * graphite-poly.c (print_pbb_body): Likewise.
11873         * passes.c (pass_manager::register_one_dump_file): Convert
11874         local "optgroup_flags" to optgroup_flags_t.
11875         * print-tree.c (print_node): Use TDF_NONE rather than 0.
11876         (debug): Likewise.
11877         (debug_body): Likewise.
11878         * tree-pass.h (struct pass_data): Convert field "optgroup_flags"
11879         to optgroup_flags_t.
11880         * tree-pretty-print.c (print_struct_decl): Use TDF_NONE rather
11881         than 0.
11882         * tree-ssa-math-opts.c (convert_mult_to_fma_1): Likewise.
11883         (convert_mult_to_fma): Likewise.
11884         * tree-ssa-reassoc.c (undistribute_ops_list): Likewise.
11885         * tree-ssa-sccvn.c (vn_eliminate): Likewise.
11886         * tree-vect-data-refs.c (dump_lower_bound): Convert param
11887         "dump_kind" to dump_flags_t.
11889 2018-06-08  Segher Boessenkool  <segher@kernel.crashing.org>
11891         * config/rs6000/rs6000.c (min, max): Delete.
11893 2018-06-08  Segher Boessenkool  <segher@kernel.crashing.org>
11895         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete mention of
11896         -mabi=spe and -mabi=no-spe.
11898 2018-06-08  Martin Liska  <mliska@suse.cz>
11900         * ipa-pure-const.c (propagate_pure_const): Use ::get at places
11901         where we expect an existing summary.
11903 2018-06-08  Martin Liska  <mliska@suse.cz>
11905         * ipa-inline-analysis.c (simple_edge_hints): Use ::get method.
11906         * ipa-inline.h (estimate_edge_growth): Likewise.
11908 2018-06-08  Martin Liska  <mliska@suse.cz>
11910         * cgraph.c (function_version_hasher::hash): Use
11911         cgraph_node::get_uid ().
11912         (function_version_hasher::equal):
11913         * cgraph.h (cgraph_node::get_uid): New method.
11914         * ipa-inline.c (update_caller_keys): Use
11915         cgraph_node::get_uid ().
11916         (update_callee_keys): Likewise.
11917         * ipa-utils.c (searchc): Likewise.
11918         (ipa_reduced_postorder): Likewise.
11919         * lto-cgraph.c (input_node): Likewise.
11920         * passes.c (is_pass_explicitly_enabled_or_disabled): Likewise.
11921         * symbol-summary.h (symtab_insertion): Likewise.
11922         (symtab_removal): Likewise.
11923         (symtab_duplication): Likewise.
11924         * tree-pretty-print.c (dump_function_header): Likewise.
11925         * tree-sra.c (convert_callers_for_node): Likewise.
11927 2018-06-08  Martin Liska  <mliska@suse.cz>
11929         * cgraph.c (symbol_table::create_edge): Always assign a new
11930         unique number.
11931         (symbol_table::free_edge): Do not recycle numbers.
11932         * cgraph.h (cgraph_edge::get): New method.
11933         * symbol-summary.h (symtab_removal): Use it.
11934         (symtab_duplication): Likewise.
11935         (call_summary::hashable_uid): Remove.
11937 2018-06-08  Martin Liska  <mliska@suse.cz>
11939         * ipa-inline-analysis.c (inline_edge_removal_hook): Remove.
11940         (initialize_growth_caches): Remove.
11941         (free_growth_caches): Likewise.
11942         (do_estimate_edge_time): Use edge_growth_cache.
11943         (do_estimate_edge_size): Likewise.
11944         (do_estimate_edge_hints): Likewise.
11945         * ipa-inline.c (reset_edge_caches): Likewise.
11946         (recursive_inlining): Likewise.
11947         (inline_small_functions): Likewise.
11948         * ipa-inline.h (initialize_growth_caches): Remove.
11949         (estimate_edge_size): Likewise.
11950         (estimate_edge_time): Likewise.
11951         (estimate_edge_hints): Likewise.
11952         (reset_edge_growth_cache): Likewise.
11953         * symbol-summary.h (call_summary::remove): New method.
11955 2018-06-08  Martin Liska  <mliska@suse.cz>
11957         * ipa-cp.c (class edge_clone_summary): New summary.
11958         (grow_edge_clone_vectors): Remove.
11959         (ipcp_edge_duplication_hook): Remove.
11960         (class edge_clone_summary_t): New call_summary class.
11961         (ipcp_edge_removal_hook): Remove.
11962         (edge_clone_summary_t::duplicate): New function.
11963         (get_next_cgraph_edge_clone): Use edge_clone_summaries.
11964         (create_specialized_node): Likewise.
11965         (ipcp_driver): Initialize edge_clone_summaries and do not
11966         register hooks.
11968 2018-06-08  Martin Liska  <mliska@suse.cz>
11970         * symbol-summary.h (get): New function.
11971         (call_summary::m_initialize_when_cloning): New class member.
11973 2018-06-08  Martin Liska  <mliska@suse.cz>
11975         * cgraph.c (cgraph_node::remove): Do not recycle uid.
11976         * cgraph.h (symbol_table::release_symbol): Do not pass uid.
11977         (symbol_table::allocate_cgraph_symbol): Do not set uid.
11978         * passes.c (uid_hash_t): Record removed_nodes by their uids.
11979         (remove_cgraph_node_from_order): Use the removed_nodes set.
11980         (do_per_function_toporder): Likwise.
11981         * symbol-summary.h (symtab_insertion): Use cgraph_node::uid
11982         instead of summary_uid.
11983         (symtab_removal): Likewise.
11984         (symtab_duplication): Likewise.
11986 2018-06-08  Martin Liska  <mliska@suse.cz>
11988         * ipa-cp.c (ipcp_store_bits_results): Use
11989         ipcp_transformation_sum.
11990         (ipcp_store_vr_results): Likewise.
11991         * ipa-prop.c (ipcp_grow_transformations_if_necessary): Renamed
11992         to ...
11993         (ipcp_transformation_initialize): ... this.
11994         (ipa_set_node_agg_value_chain):
11995         (ipa_node_params_t::duplicate): Use ipcp_transformation_sum.
11996         (write_ipcp_transformation_info): Likewise.
11997         (read_ipcp_transformation_info): Likewise.
11998         (ipcp_update_bits): Likewise.
11999         (ipcp_update_vr): Likewise.
12000         (ipcp_transform_function): Likewise.
12001         * ipa-prop.h: Rename ipcp_transformation_summary to
12002         ipcp_transformation.
12003         (class ipcp_transformation_t): New function summary.
12004         (ipcp_get_transformation_summary): Use ipcp_transformation_sum.
12005         (ipa_get_agg_replacements_for_node): Likewise.
12007 2018-06-08  Martin Liska  <mliska@suse.cz>
12009         * ipa-pure-const.c (struct funct_state_d): Do it class instead
12010         of struct.
12011         (class funct_state_summary_t): New function_summary class.
12012         (has_function_state): Remove.
12013         (get_function_state): Likewise.
12014         (set_function_state): Likewise.
12015         (add_new_function): Likewise.
12016         (funct_state_summary_t::insert): New function.
12017         (duplicate_node_data): Remove.
12018         (remove_node_data): Remove.
12019         (funct_state_summary_t::duplicate): New function.
12020         (register_hooks): Create new funct_state_summaries.
12021         (pure_const_generate_summary): Use it.
12022         (pure_const_write_summary): Likewise.
12023         (pure_const_read_summary): Likewise.
12024         (propagate_pure_const): Likewise.
12025         (propagate_nothrow): Likewise.
12026         (dump_malloc_lattice): Likewise.
12027         (propagate_malloc): Likewise.
12028         (execute): Do not register hooks, just remove summary
12029         instead.
12030         (pass_ipa_pure_const::pass_ipa_pure_const): Simplify
12031         constructor.
12033 2018-06-08  Martin Liska  <mliska@suse.cz>
12035         * ipa-reference.c (remove_node_data): Remove.
12036         (duplicate_node_data): Likewise.
12037         (class ipa_ref_var_info_summary_t): New class.
12038         (class ipa_ref_opt_summary_t): Likewise.
12039         (get_reference_vars_info): Use ipa_ref_var_info_summaries.
12040         (get_reference_optimization_summary): Use
12041         ipa_ref_opt_sum_summaries.
12042         (set_reference_vars_info): Remove.
12043         (set_reference_optimization_summary): Likewise.
12044         (ipa_init): Create summaries.
12045         (init_function_info): Use function summary.
12046         (ipa_ref_opt_summary_t::duplicate): New function.
12047         (ipa_ref_opt_summary_t::remove): New function.
12048         (get_read_write_all_from_node): Fix GNU coding style.
12049         (propagate): Use function summary.
12050         (write_node_summary_p): Fix GNU coding style.
12051         (stream_out_bitmap): Likewise.
12052         (ipa_reference_read_optimization_summary): Use function summary.
12053         (ipa_reference_c_finalize): Do not release hooks.
12055 2018-06-08  Martin Liska  <mliska@suse.cz>
12057         * ipa-fnsummary.c (dump_ipa_call_summary): Use ::get method.
12058         (analyze_function_body): Extract multiple calls of get_create.
12059         * ipa-inline-analysis.c (simple_edge_hints): Likewise.
12060         * ipa-inline.c (recursive_inlining): Use ::get method.
12061         * ipa-inline.h (estimate_edge_growth): Likewise.
12063 2018-06-08  Martin Liska  <mliska@suse.cz>
12065         * hsa-common.h (enum hsa_function_kind): Rename HSA_NONE to
12066         HSA_INVALID.
12067         (hsa_function_summary::hsa_function_summary): Use the new enum
12068         value.
12069         (hsa_gpu_implementation_p): Use hsa_summaries::get.
12070         * hsa-gen.c (hsa_get_host_function): Likewise.
12071         (get_brig_function_name): Likewise.
12072         * ipa-hsa.c (process_hsa_functions): Likewise.
12073         (ipa_hsa_write_summary): Likewise.
12074         * symbol-summary.h (symtab_duplication): Use ::get function/
12075         (get): New function.
12077 2018-06-08  Martin Liska  <mliska@suse.cz>
12079         * config/i386/i386.c (ix86_can_inline_p): Use get_create instead
12080         of get.
12081         * hsa-common.c (hsa_summary_t::link_functions): Likewise.
12082         (hsa_register_kernel): Likewise.
12083         * hsa-common.h (hsa_gpu_implementation_p): Likewise.
12084         * hsa-gen.c (hsa_get_host_function): Likewise.
12085         (get_brig_function_name): Likewise.
12086         (generate_hsa): Likewise.
12087         (pass_gen_hsail::execute): Likewise.
12088         * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
12089         (devirtualization_time_bonus): Likewise.
12090         (ipcp_propagate_stage): Likewise.
12091         * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
12092         (edge_set_predicate): Likewise.
12093         (evaluate_conditions_for_known_args): Likewise.
12094         (evaluate_properties_for_edge): Likewise.
12095         (ipa_fn_summary::reset): Likewise.
12096         (ipa_fn_summary_t::duplicate): Likewise.
12097         (dump_ipa_call_summary): Likewise.
12098         (ipa_dump_fn_summary): Likewise.
12099         (analyze_function_body): Likewise.
12100         (compute_fn_summary): Likewise.
12101         (estimate_edge_devirt_benefit): Likewise.
12102         (estimate_edge_size_and_time): Likewise.
12103         (estimate_calls_size_and_time): Likewise.
12104         (estimate_node_size_and_time): Likewise.
12105         (inline_update_callee_summaries): Likewise.
12106         (remap_edge_change_prob): Likewise.
12107         (remap_edge_summaries): Likewise.
12108         (ipa_merge_fn_summary_after_inlining): Likewise.
12109         (ipa_update_overall_fn_summary): Likewise.
12110         (read_ipa_call_summary): Likewise.
12111         (inline_read_section): Likewise.
12112         (write_ipa_call_summary): Likewise.
12113         (ipa_fn_summary_write): Likewise.
12114         (ipa_free_fn_summary): Likewise.
12115         * ipa-hsa.c (process_hsa_functions): Likewise.
12116         (ipa_hsa_write_summary): Likewise.
12117         (ipa_hsa_read_section): Likewise.
12118         * ipa-icf.c (sem_function::merge): Likewise.
12119         * ipa-inline-analysis.c (simple_edge_hints): Likewise.
12120         (do_estimate_edge_time): Likewise.
12121         (estimate_size_after_inlining): Likewise.
12122         (estimate_growth): Likewise.
12123         (growth_likely_positive): Likewise.
12124         * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
12125         (inline_call): Likewise.
12126         * ipa-inline.c (caller_growth_limits): Likewise.
12127         (can_inline_edge_p): Likewise.
12128         (can_inline_edge_by_limits_p): Likewise.
12129         (compute_uninlined_call_time): Likewise.
12130         (compute_inlined_call_time): Likewise.
12131         (want_inline_small_function_p): Likewise.
12132         (edge_badness): Likewise.
12133         (update_caller_keys): Likewise.
12134         (update_callee_keys): Likewise.
12135         (recursive_inlining): Likewise.
12136         (inline_small_functions): Likewise.
12137         (inline_to_all_callers_1): Likewise.
12138         (dump_overall_stats): Likewise.
12139         (early_inline_small_functions): Likewise.
12140         (early_inliner): Likewise.
12141         * ipa-inline.h (estimate_edge_growth): Likewise.
12142         * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
12143         * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
12144         * ipa-prop.h (IPA_NODE_REF): Likewise.
12145         (IPA_EDGE_REF): Likewise.
12146         * ipa-pure-const.c (malloc_candidate_p): Likewise.
12147         (propagate_malloc): Likewise.
12148         * ipa-split.c (execute_split_functions): Likewise.
12149         * symbol-summary.h: Rename get to get_create.
12150         (get): Likewise.
12151         (get_create): Likewise.
12152         * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
12154 2018-06-08  Martin Liska  <mliska@suse.cz>
12156         * symbol-summary.h (release): Move definition out of class
12157         declaration.
12158         (symtab_removal): Likewise.
12159         (symtab_duplication): Likewise.
12161 2018-06-08  Martin Liska  <mliska@suse.cz>
12163         * symbol-summary.h (function_summary): Move constructor
12164         implementation out of class declaration.
12165         (release): Likewise.
12166         (symtab_insertion): Likewise.
12167         (symtab_removal): Likewise.
12168         (symtab_duplication): Likewise.
12169         (get): Likewise.
12171 2018-06-08  Martin Liska  <mliska@suse.cz>
12173         * Makefile.in: Remove support for MPX (macros, related functions,
12174         fields in cgraph_node, ...).
12175         * builtin-types.def (BT_BND): Likewise.
12176         (BT_FN_BND_CONST_PTR): Likewise.
12177         (BT_FN_CONST_PTR_BND): Likewise.
12178         (BT_FN_VOID_PTR_BND): Likewise.
12179         (BT_FN_BND_CONST_PTR_SIZE): Likewise.
12180         (BT_FN_VOID_CONST_PTR_BND_CONST_PTR): Likewise.
12181         * builtins.c (expand_builtin_memcpy_with_bounds): Likewise.
12182         (expand_builtin_mempcpy_with_bounds): Likewise.
12183         (expand_builtin_memset_with_bounds): Likewise.
12184         (expand_builtin_memset_args): Likewise.
12185         (std_expand_builtin_va_start): Likewise.
12186         (expand_builtin): Likewise.
12187         (expand_builtin_with_bounds): Likewise.
12188         * builtins.def (DEF_BUILTIN_CHKP): Likewise.
12189         (DEF_LIB_BUILTIN_CHKP): Likewise.
12190         (DEF_EXT_LIB_BUILTIN_CHKP): Likewise.
12191         (DEF_CHKP_BUILTIN): Likewise.
12192         (BUILT_IN_MEMCPY): Likewise.
12193         (BUILT_IN_MEMMOVE): Likewise.
12194         (BUILT_IN_MEMPCPY): Likewise.
12195         (BUILT_IN_MEMSET): Likewise.
12196         (BUILT_IN_STPCPY): Likewise.
12197         (BUILT_IN_STRCAT): Likewise.
12198         (BUILT_IN_STRCHR): Likewise.
12199         (BUILT_IN_STRCPY): Likewise.
12200         (BUILT_IN_STRLEN): Likewise.
12201         (BUILT_IN_MEMCPY_CHK): Likewise.
12202         (BUILT_IN_MEMMOVE_CHK): Likewise.
12203         (BUILT_IN_MEMPCPY_CHK): Likewise.
12204         (BUILT_IN_MEMSET_CHK): Likewise.
12205         (BUILT_IN_STPCPY_CHK): Likewise.
12206         (BUILT_IN_STRCAT_CHK): Likewise.
12207         (BUILT_IN_STRCPY_CHK): Likewise.
12208         * calls.c (store_bounds): Likewise.
12209         (emit_call_1): Likewise.
12210         (special_function_p): Likewise.
12211         (maybe_warn_nonstring_arg): Likewise.
12212         (initialize_argument_information): Likewise.
12213         (finalize_must_preallocate): Likewise.
12214         (compute_argument_addresses): Likewise.
12215         (expand_call): Likewise.
12216         * cfgexpand.c (expand_call_stmt): Likewise.
12217         (expand_return): Likewise.
12218         (expand_gimple_stmt_1): Likewise.
12219         (pass_expand::execute): Likewise.
12220         * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
12221         (cgraph_node::remove): Likewise.
12222         (cgraph_node::dump): Likewise.
12223         (cgraph_node::verify_node): Likewise.
12224         * cgraph.h (chkp_function_instrumented_p): Likewise.
12225         (symtab_node::get_alias_target): Likewise.
12226         (cgraph_node::can_remove_if_no_direct_calls_and_refs_p): Likewise.
12227         (cgraph_local_p): Likewise.
12228         * cgraphbuild.c (cgraph_edge::rebuild_edges): Likewise.
12229         (cgraph_edge::rebuild_references): Likewise.
12230         * cgraphunit.c (varpool_node::finalize_decl): Likewise.
12231         (walk_polymorphic_call_targets): Likewise.
12232         (cgraph_node::expand_thunk): Likewise.
12233         (symbol_table::output_weakrefs): Likewise.
12234         * common/config/i386/i386-common.c (OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET): Likewise.
12235         (ix86_handle_option): Likewise.
12236         * config/i386/constraints.md: Likewise.
12237         * config/i386/i386-builtin-types.def (BND): Likewise.
12238         (VOID): Likewise.
12239         (PVOID): Likewise.
12240         (ULONG): Likewise.
12241         * config/i386/i386-builtin.def (BDESC_END): Likewise.
12242         (BDESC_FIRST): Likewise.
12243         (BDESC): Likewise.
12244         * config/i386/i386-c.c (ix86_target_macros_internal): Likewise.
12245         * config/i386/i386-protos.h (ix86_bnd_prefixed_insn_p): Likewise.
12246         * config/i386/i386.c (enum reg_class): Likewise.
12247         (ix86_target_string): Likewise.
12248         (ix86_option_override_internal): Likewise.
12249         (ix86_conditional_register_usage): Likewise.
12250         (ix86_valid_target_attribute_inner_p): Likewise.
12251         (ix86_set_indirect_branch_type): Likewise.
12252         (ix86_set_current_function): Likewise.
12253         (ix86_function_arg_regno_p): Likewise.
12254         (init_cumulative_args): Likewise.
12255         (ix86_function_arg_advance): Likewise.
12256         (ix86_function_arg): Likewise.
12257         (ix86_pass_by_reference): Likewise.
12258         (ix86_function_value_regno_p): Likewise.
12259         (ix86_function_value_1): Likewise.
12260         (ix86_function_value_bounds): Likewise.
12261         (ix86_return_in_memory): Likewise.
12262         (ix86_setup_incoming_vararg_bounds): Likewise.
12263         (ix86_va_start): Likewise.
12264         (indirect_thunk_need_prefix): Likewise.
12265         (print_reg): Likewise.
12266         (ix86_print_operand): Likewise.
12267         (ix86_expand_call): Likewise.
12268         (ix86_output_function_return): Likewise.
12269         (reg_encoded_number): Likewise.
12270         (BDESC_VERIFYS): Likewise.
12271         (ix86_init_mpx_builtins): Likewise.
12272         (ix86_init_builtins): Likewise.
12273         (ix86_emit_cmove): Likewise.
12274         (ix86_emit_move_max): Likewise.
12275         (ix86_expand_builtin): Likewise.
12276         (ix86_builtin_mpx_function): Likewise.
12277         (ix86_get_arg_address_for_bt): Likewise.
12278         (ix86_load_bounds): Likewise.
12279         (ix86_store_bounds): Likewise.
12280         (ix86_load_returned_bounds): Likewise.
12281         (ix86_store_returned_bounds): Likewise.
12282         (ix86_class_likely_spilled_p): Likewise.
12283         (ix86_hard_regno_mode_ok): Likewise.
12284         (x86_order_regs_for_local_alloc): Likewise.
12285         (ix86_mitigate_rop): Likewise.
12286         (ix86_bnd_prefixed_insn_p): Likewise.
12287         (ix86_mpx_bound_mode): Likewise.
12288         (ix86_make_bounds_constant): Likewise.
12289         (ix86_initialize_bounds): Likewise.
12290         (TARGET_LOAD_BOUNDS_FOR_ARG): Likewise.
12291         (TARGET_STORE_BOUNDS_FOR_ARG): Likewise.
12292         (TARGET_LOAD_RETURNED_BOUNDS): Likewise.
12293         (TARGET_STORE_RETURNED_BOUNDS): Likewise.
12294         (TARGET_CHKP_BOUND_MODE): Likewise.
12295         (TARGET_BUILTIN_CHKP_FUNCTION): Likewise.
12296         (TARGET_CHKP_FUNCTION_VALUE_BOUNDS): Likewise.
12297         (TARGET_CHKP_MAKE_BOUNDS_CONSTANT): Likewise.
12298         (TARGET_CHKP_INITIALIZE_BOUNDS): Likewise.
12299         * config/i386/i386.h (TARGET_MPX): Likewise.
12300         (TARGET_MPX_P): Likewise.
12301         (VALID_BND_REG_MODE): Likewise.
12302         (FIRST_BND_REG): Likewise.
12303         (LAST_BND_REG): Likewise.
12304         (enum reg_class): Likewise.
12305         (BND_REG_P): Likewise.
12306         (BND_REGNO_P): Likewise.
12307         (BNDmode): Likewise.
12308         (ADJUST_INSN_LENGTH): Likewise.
12309         * config/i386/i386.md: Likewise.
12310         * config/i386/i386.opt: Likewise.
12311         * config/i386/linux-common.h (LIBMPX_LIBS): Likewise.
12312         (defined): Likewise.
12313         (LINK_MPX): Likewise.
12314         (MPX_SPEC): Likewise.
12315         (LIBMPX_SPEC): Likewise.
12316         (LIBMPXWRAPPERS_SPEC): Likewise.
12317         (CHKP_SPEC): Likewise.
12318         * config/i386/predicates.md: Likewise.
12319         * dbxout.c (dbxout_type): Likewise.
12320         * doc/extend.texi: Likewise.
12321         * doc/invoke.texi: Likewise.
12322         * doc/md.texi: Likewise.
12323         * doc/tm.texi: Likewise.
12324         * doc/tm.texi.in: Likewise.
12325         * dwarf2out.c (is_base_type): Likewise.
12326         (gen_formal_types_die): Likewise.
12327         (gen_subprogram_die): Likewise.
12328         (gen_type_die_with_usage): Likewise.
12329         (gen_decl_die): Likewise.
12330         (dwarf2out_late_global_decl): Likewise.
12331         * expr.c (expand_assignment): Likewise.
12332         (emit_storent_insn): Likewise.
12333         (store_expr_with_bounds): Likewise.
12334         (store_expr): Likewise.
12335         (expand_expr_real_1): Likewise.
12336         * expr.h (store_expr_with_bounds): Likewise.
12337         * function.c (use_register_for_decl): Likewise.
12338         (struct bounds_parm_data): Likewise.
12339         (assign_parms_augmented_arg_list): Likewise.
12340         (assign_parm_find_entry_rtl): Likewise.
12341         (assign_parm_is_stack_parm): Likewise.
12342         (assign_parm_load_bounds): Likewise.
12343         (assign_bounds): Likewise.
12344         (assign_parms): Likewise.
12345         (expand_function_start): Likewise.
12346         * gcc.c (CHKP_SPEC): Likewise.
12347         * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
12348         * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Likewise.
12349         (wrestrict_dom_walker::check_call): Likewise.
12350         * gimple.c (gimple_build_call_from_tree): Likewise.
12351         * gimple.h (enum gf_mask): Likewise.
12352         (gimple_call_with_bounds_p): Likewise.
12353         (gimple_call_set_with_bounds): Likewise.
12354         * gimplify.c (gimplify_init_constructor): Likewise.
12355         * ipa-cp.c (initialize_node_lattices): Likewise.
12356         (propagate_constants_across_call): Likewise.
12357         (find_more_scalar_values_for_callers_subset): Likewise.
12358         * ipa-hsa.c (process_hsa_functions): Likewise.
12359         * ipa-icf-gimple.c (func_checker::compare_gimple_call): Likewise.
12360         * ipa-icf.c (sem_function::merge): Likewise.
12361         * ipa-inline.c (early_inliner): Likewise.
12362         * ipa-pure-const.c (warn_function_noreturn): Likewise.
12363         (warn_function_cold): Likewise.
12364         (propagate_pure_const): Likewise.
12365         * ipa-ref.h (enum GTY): Likewise.
12366         * ipa-split.c (find_retbnd): Likewise.
12367         (consider_split): Likewise.
12368         (split_function): Likewise.
12369         * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
12370         * ipa.c (walk_polymorphic_call_targets): Likewise.
12371         (symbol_table::remove_unreachable_nodes): Likewise.
12372         (process_references): Likewise.
12373         (cgraph_build_static_cdtor_1): Likewise.
12374         * lto-cgraph.c (lto_output_node): Likewise.
12375         (output_refs): Likewise.
12376         (compute_ltrans_boundary): Likewise.
12377         (input_overwrite_node): Likewise.
12378         (input_node): Likewise.
12379         (input_cgraph_1): Likewise.
12380         * params.def (PARAM_CHKP_MAX_CTOR_SIZE): Likewise.
12381         * passes.c (pass_manager::execute_early_local_passes): Likewise.
12382         (class pass_chkp_instrumentation_passes): Likewise.
12383         (make_pass_chkp_instrumentation_passes): Likewise.
12384         * passes.def: Likewise.
12385         * rtl.h (struct GTY): Likewise.
12386         (CALL_EXPR_WITH_BOUNDS_P): Likewise.
12387         * stor-layout.c (layout_type): Likewise.
12388         * symtab.c: Likewise.
12389         * target.def: Likewise.
12390         * targhooks.c (default_chkp_bound_type): Likewise.
12391         (default_chkp_bound_mode): Likewise.
12392         (default_builtin_chkp_function): Likewise.
12393         (default_chkp_function_value_bounds): Likewise.
12394         (default_chkp_make_bounds_constant): Likewise.
12395         (default_chkp_initialize_bounds): Likewise.
12396         * targhooks.h (default_chkp_bound_type): Likewise.
12397         (default_chkp_bound_mode): Likewise.
12398         (default_builtin_chkp_function): Likewise.
12399         (default_chkp_function_value_bounds): Likewise.
12400         (default_chkp_make_bounds_constant): Likewise.
12401         (default_chkp_initialize_bounds): Likewise.
12402         * toplev.c (compile_file): Likewise.
12403         (process_options): Likewise.
12404         * tree-core.h (DEF_BUILTIN): Likewise.
12405         (DEF_BUILTIN_CHKP): Likewise.
12406         * tree-inline.c (declare_return_variable): Likewise.
12407         (remap_gimple_stmt): Likewise.
12408         (copy_bb): Likewise.
12409         (initialize_inlined_parameters): Likewise.
12410         (expand_call_inline): Likewise.
12411         * tree-pass.h (make_pass_ipa_chkp_versioning): Likewise.
12412         (make_pass_ipa_chkp_early_produce_thunks): Likewise.
12413         (make_pass_ipa_chkp_produce_thunks): Likewise.
12414         (make_pass_chkp): Likewise.
12415         (make_pass_chkp_opt): Likewise.
12416         (make_pass_chkp_instrumentation_passes): Likewise.
12417         * tree-pretty-print.c (dump_generic_node): Likewise.
12418         * tree-ssa-ccp.c (insert_clobber_before_stack_restore): Likewise.
12419         * tree-ssa-dce.c (propagate_necessity): Likewise.
12420         (eliminate_unnecessary_stmts): Likewise.
12421         * tree-ssa-pre.c (create_expression_by_pieces): Likewise.
12422         * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
12423         * tree-ssa-sccvn.h: Likewise.
12424         * tree-ssa-strlen.c (get_string_length): Likewise.
12425         (valid_builtin_call): Likewise.
12426         (adjust_last_stmt): Likewise.
12427         (handle_builtin_strchr): Likewise.
12428         (handle_builtin_strcpy): Likewise.
12429         (handle_builtin_stxncpy): Likewise.
12430         (handle_builtin_memcpy): Likewise.
12431         (handle_builtin_strcat): Likewise.
12432         (strlen_check_and_optimize_stmt): Likewise.
12433         * tree-stdarg.c (expand_ifn_va_arg_1): Likewise.
12434         * tree-streamer-in.c: Likewise.
12435         * tree-streamer.c (record_common_node): Likewise.
12436         * tree.c (tree_code_size): Likewise.
12437         (wide_int_to_tree_1): Likewise.
12438         (type_contains_placeholder_1): Likewise.
12439         (build_common_tree_nodes): Likewise.
12440         * tree.def (POINTER_BOUNDS_TYPE): Likewise.
12441         * tree.h (POINTER_BOUNDS_TYPE_P): Likewise.
12442         (POINTER_BOUNDS_P): Likewise.
12443         (BOUNDED_TYPE_P): Likewise.
12444         (BOUNDED_P): Likewise.
12445         (CALL_WITH_BOUNDS_P): Likewise.
12446         (pointer_bounds_type_node): Likewise.
12447         * value-prof.c (gimple_ic): Likewise.
12448         * var-tracking.c (vt_add_function_parameters): Likewise.
12449         * varasm.c (make_decl_rtl): Likewise.
12450         (assemble_start_function): Likewise.
12451         (output_constant): Likewise.
12452         (maybe_assemble_visibility): Likewise.
12453         * varpool.c (ctor_for_folding): Likewise.
12454         * chkp-builtins.def: Remove.
12455         * ipa-chkp.c: Remove.
12456         * ipa-chkp.h: Remove.
12457         * rtl-chkp.c: Remove.
12458         * rtl-chkp.h: Remove.
12459         * tree-chkp-opt.c: Remove.
12460         * tree-chkp.c: Remove.
12461         * tree-chkp.h: Remove.
12463 2018-06-07  Carl Love  <cel@us.ibm.com>
12465         * config/rs6000/vsx.md (vextract_fp_from_shorth,
12466         vextract_fp_from_shortl): Add BE support.
12468 2018-06-07  Paul Koning  <ni1d@arrl.net>
12470         * compare-elim.c (try_merge_compare): Don't merge compare if
12471         address contains a side effect.
12472         (try_eliminate_compare): Likewise.
12474 2018-06-07  Olga Makhotina  <olga.makhotina@intel.com>
12476         * config.gcc: Support "tremont".
12477         * config/i386/driver-i386.c (host_detect_local_cpu): Detect "tremont".
12478         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
12479         PROCESSOR_TREMONT.
12480         * config/i386/i386.c (m_TREMONT): Define.
12481         (processor_target_table): Add "tremont".
12482         (PTA_TREMONT): Define.
12483         (ix86_lea_outperforms): Add TARGET_TREMONT.
12484         (get_builtin_code_for_version): Handle PROCESSOR_TREMONT.
12485         (fold_builtin_cpu): Add M_INTEL_TREMONT, replace M_INTEL_GOLDMONT
12486         and M_INTEL_GOLDMONT_PLUS.
12487         (fold_builtin_cpu): Add "tremont".
12488         (ix86_add_stmt_cost): Add TARGET_TREMONT.
12489         (ix86_option_override_internal): Add "tremont".
12490         * config/i386/i386.h (processor_costs): Define TARGET_TREMONT.
12491         (processor_type): Add PROCESSOR_TREMONT.
12492         * config/i386/x86-tune.def: Add m_TREMONT.
12493         * doc/invoke.texi: Add tremont as x86 -march=/-mtune= CPU type.
12495 2018-06-07  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
12497         * config/msp430/msp430.c (msp430_mcu_name): Set the "i" in the
12498         symbol defined for msp430i* devices to be lower case.
12500 2018-06-07  Richard Biener  <rguenther@suse.de>
12502         * graphite-sese-to-poly.c (extract_affine): Avoid unneded
12503         wrapping.  Properly wrap the result of a BIT_NOT_EXPR.
12504         Properly wrap signed arithmetic if overflow wraps.
12506 2018-06-07  Jakub Jelinek  <jakub@redhat.com>
12508         PR tree-optimization/69615
12509         * tree-ssa-reassoc.c (optimize_range_tests_var_bound): If rhs2 is lhs
12510         of a cast from a same precision integral SSA_NAME in a bb dominated
12511         by first_bb, retry with rhs2 set to the rhs1 of the cast.  Don't emit
12512         cast to utype if rhs2 has already a compatible type.
12514 2018-06-07  Richard Biener  <rguenther@suse.de>
12516         PR tree-optimization/85935
12517         * graphite-scop-detection.c (find_params_in_bb): Analyze
12518         condition operands with respect to the correct loop.  Assert
12519         the analysis doesn't fail.
12521 2018-06-04  Carl Love  <cel@us.ibm.com>
12523         * config/rs6000/vsx.md (first_match_index_<mode>): Calculate index
12524         using natural element order.  Use gen_lshrsi3 instead of gen_ashrsi3
12525         as it is slightly cheaper.
12526         (first_match_or_eos_index_<mode>):
12527         Calculate index using natural element order.
12528         (first_match_index_<mode>):
12529         Calculate index using natural element order.
12530         (first_match_or_eos_index_<mode>):
12531         Calculate index using natural order.
12532         (define_insn vclzlsbb): Change to define_insn vclzlsbb_<mode>.
12533         for BE and LE modes.
12534         * config/rs6000/rs6000-c.c: Rename P9V_BUILTIN_VCLZLSBB,
12535         P9V_BUILTIN_VCLZLSBB_V16QI.
12536         * config/rs6000/rs6000-builtin.def: Make VCLZLSBB mode
12537         specific.
12539 2018-06-06  Kelvin Nilsen  <kelvin@gcc.gnu.org>
12541         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Adjust
12542         indentation and line wrap for many prototypes.  Add missing
12543         @smallexample directives around block of prototypes for vec_xl and
12544         vec_xst.
12546 2018-06-05  Michael Meissner  <meissner@linux.ibm.com>
12548         * config/rs6000/rs6000.c (rs6000_passes_ieee128): New boolean to
12549         track if we pass or return IEEE 128-bit floating point.
12550         (ieee128_mangling_gcc_8_1): New boolean to say whether to generate
12551         C++ mangling that is compatible with GCC 8.1.
12552         (TARGET_ASM_GLOBALIZE_DECL_NAME): Override target hook.
12553         (init_cumulative_args): Note if we pass or return IEEE 128-bit
12554         floating point types.
12555         (rs6000_function_arg_advance_1): Likewise.
12556         (rs6000_mangle_type): Optionally generate mangled names that match
12557         what GCC 8.1 generated for IEEE 128-bit floating point types.
12558         (rs6000_globalize_decl_name): If we have an external function that
12559         passes or returns IEEE 128-bit types, generate a weak reference
12560         from the mangled name used in GCC 8.1 to the current mangled
12561         name.
12562         (rs6000_init_builtins): Make __ibm128 use the long double type if
12563         long double is IBM extended double.  Make __float128 use the long
12564         double type if long double is IEEE 128-bit.
12566         PR target/85657
12567         * config/rs6000/rs6000-builtin.def (BU_IBM128_2): New helper
12568         macro for __ibm128 built-in functions.
12569         (PACK_IF): Add __ibm128 pack/unpack functions.
12570         (UNPACK_IF): Likewise.
12571         * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Do not
12572         enable long double built-in functions if long double is IEEE
12573         128-bit floating point.
12574         (rs6000_invalid_builtin): Update long double built-in function
12575         error message.
12576         (rs6000_expand_builtin): For PACK_IF and UNPACK_IF built-in
12577         functions, adjust the built-in function to use the long double
12578         built-in function if __ibm128 and long double are the same type.
12579         * doc/extend.texi (PowerPC builtins): Update documention for
12580         __builtin_{,un}pack_longdouble.  Add documentation for
12581         __builtin_{,un}pack_ibm128.
12583 2018-06-06  Jim Wilson  <jimw@sifive.com>
12585         * config/riscv/riscv.c (enum riscv_privilege_levels): New.
12586         (struct machine_function): New field interrupt_mode.
12587         (riscv_handle_type_attribute): New function.  Add forward declaration.
12588         (riscv_attribute_table) <interrupt>: Use riscv_handle_type_attribute.
12589         (riscv_expand_epilogue): Check interrupt_mode field.
12590         (riscv_set_current_function): Check interrupt attribute args and
12591         set interrupt_mode field.
12592         * config/riscv/riscv.md (UNSPECV_SRET, UNSPECV_URET): New.
12593         (riscv_sret, riscv_uret): New.
12594         * doc/extend.texi (RISC-V Function Attributes) <interrupt>: Document
12595         new arguments to interrupt attribute.
12597 2018-06-06  Peter Bergner  <bergner@vnet.ibm.com>
12599         PR target/63177
12600         * /config/rs6000/rs6000.h (ASM_CPU_SPEC): Add support for -mpower9.
12601         Don't handle -mcpu=power8 if -mpower9-vector is also used.
12603 2018-06-06  Kelvin Nilsen  <kelvin@gcc.gnu.org>
12605         * config/rs6000/rs6000-builtin.def (VSX_BUILTIN_VEC_LD,
12606         VSX_BUILTIN_VEC_ST): Add comment to explain non-traditional uses.
12607         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
12608         several redundant entries.
12610 2018-06-06  David Malcolm  <dmalcolm@redhat.com>
12612         * config/i386/i386-protos.h (ix86_expand_call): Strengthen return
12613         type from "rtx" to "rtx_insn *".
12614         * config/i386/i386.c (ix86_expand_split_stack_prologue): Likewise
12615         for local "call_insn", removing cast.
12616         (ix86_expand_call): Likewise, introducing a "call_insn" local.
12618 2018-06-06  Eric Botcazou  <ebotcazou@adacore.com>
12620         PR tree-optimization/86066
12621         * gimple-ssa-store-merging.c (process_store): Do not bypass BIT_NOT_EXPR
12622         for BIT_INSERT_EXPR stores.
12624 2018-06-06  Richard Biener  <rguenther@suse.de>
12626         PR tree-optimization/86062
12627         * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
12628         component refs ontop
12629         of to be offsetted base.
12631 2018-06-06  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
12633         * config/msp430/msp430.c (msp430_attr): Allow interrupt handlers
12634         to be static and remove check on interrupt attribute name.
12636 2018-06-05  Kelvin Nilsen  <kelvin@gcc.gnu.org>
12638         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
12639         volatile qualifier from vec_lvsl and vec_lvsr argument prototypes.
12641 2018-06-05  Steve Ellcey  <sellcey@cavium.com>
12643         PR target/79924
12644         * config/aarch64/aarch64-protos.h (aarch64_err_no_fpadvsimd): Remove
12645         second argument.
12646         * config/aarch64/aarch64-protos..c (aarch64_err_no_fpadvsimd):
12647         Remove second argument, change how error is called.
12648         (aarch64_layout_arg): Remove second argument from
12649         aarch64_err_no_fpadvsimd call.
12650         (aarch64_init_cumulative_args): Ditto.
12651         (aarch64_gimplify_va_arg_expr): Ditto.
12652         * config/aarch64/aarch64.md (mov<mode>): Ditto.
12654 2018-06-05  Uros Bizjak  <ubizjak@gmail.com>
12656         * config/i386/i386.md (simple_return_indirect_internal): New expander.
12657         (*simple_return_indirect_internal<mode>): Rename from
12658         simple_return_indirect_internal.  Use W mode iterator.
12659         (rstorssp): New expander.
12660         (*rstorssp<mode>): Rename from rstorssp.  Use P mode iterator.
12661         (clrssbsy): New expander.
12662         (*clrssbsy<mode>): Rename from clrssbsy.  Use P mode iterator.
12664 2018-06-05  Andre Vieira  <andre.simoesdiasvieira@arm.com>
12666         * config/arm/arm_cmse.h (cmse_nsfptr_create): Change typeof to
12667         __typeof__.
12668         (cmse_check_pointed_object): Likewise.
12670 2018-06-05  Martin Liska  <mliska@suse.cz>
12672         PR gcov-profile/47618
12673         * doc/invoke.texi: Document how -fprofile-dir format
12674         is extended.
12676 2018-06-05  Richard Biener  <rguenther@suse.de>
12678         * tree-cfgcleanup.c (cleanup_control_flow_pre): For edge
12679         removal pretend DOM info isn't available so we do not update
12680         it and only remove edges, not dominated blocks.  Actually free
12681         DOM info in case we removed something.  Remove unreachable blocks.
12682         (mfb_keep_latches): Work with either DOM info or marked backedges.
12683         (cleanup_tree_cfg_noloop): Do not remove unreachable blocks
12684         first.  Mark backedges if DOM info isn't available.
12685         (Re-)compute DOM info after cleanup_control_flow_pre.
12687 2018-06-05  Richard Biener  <rguenther@suse.de>
12689         * tree-cfg.c (struct locus_discrim_map): Store line, not location.
12690         (locus_discrim_hasher::hash): Adjust.
12691         (locus_discrim_hasher::equal): Likewise.
12692         (next_discriminator_for_locus): Work on line directly.
12693         (same_line_p): Pass in expanded locus1 as well.
12694         (assign_discriminators): Avoid redundant location expansions.
12696 2018-06-05  Richard Biener  <rguenther@suse.de>
12698         PR tree-optimization/86046
12699         * tree-ssa.c (maybe_optimize_var): Clear DECL_GIMPLE_REG_P
12700         if required after clearing TREE_ADDRESSABLE.
12702 2018-06-05  Richard Biener  <rguenther@suse.de>
12704         PR tree-optimization/86047
12705         * tree-ssa-loop.c (for_each_index): Glob handling of all
12706         decls and constants and really handle all of them.
12708 2018-06-05  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
12710         PR target/81497
12711         * config/arm/arm-builtins.c (arm_type_qualifiers): Add
12712         qualifier_void_pointer and qualifier_const_void_pointer.
12713         (arm_ldc_qualifiers, arm_stc_qualifiers): Use the above.
12714         (arm_init_builtins): Handle the above.
12715         * config/arm/arm_acle.h (__arm_cdp, __arm_ldc, __arm_ldcl, __arm_stc,
12716         __arm_stcl, __arm_mcr, __arm_cdp2, __arm_ldc2, __arm_ldcl2, __arm_stc2,
12717         __arm_stcl2,__arm_mcr2, __arm_mcrr, __arm_mcrr2): Remove return for
12718         void intrinsics.
12720 2018-06-05  Martin Liska  <mliska@suse.cz>
12722         * auto-profile.c (read_autofdo_file): Do not use
12723         gcov_ctr_summary struct.
12724         (afdo_callsite_hot_enough_for_early_inline): Likewise.
12725         * coverage.c (struct counts_entry): Likewise.
12726         (read_counts_file): Read just single summary entry.
12727         (get_coverage_counts): Use gcov_summary struct.
12728         * coverage.h (get_coverage_counts): Likewise.
12729         * gcov-dump.c (dump_working_sets): Likewise.
12730         (tag_summary): Dump just single summary.
12731         * gcov-io.c (gcov_write_summary): Write just histogram
12732         summary.
12733         (gcov_read_summary): Read just single summary.
12734         (compute_working_sets): Use gcov_summary struct.
12735         * gcov-io.h (GCOV_TAG_SUMMARY_LENGTH): Remove usage
12736         of GCOV_COUNTERS_SUMMABLE.
12737         (GCOV_COUNTERS_SUMMABLE): Remove.
12738         (GCOV_FIRST_VALUE_COUNTER): Replace with
12739         GCOV_COUNTER_V_INTERVAL.
12740         (struct gcov_ctr_summary): Remove.
12741         (struct gcov_summary): Directly use fields of former
12742         gcov_ctr_summary.
12743         (compute_working_sets): Use gcov_summary struct.
12744         * gcov.c (read_count_file): Do not use ctrs fields.
12745         * lto-cgraph.c (merge_profile_summaries): Use gcov_summary
12746         struct.
12747         * lto-streamer.h (struct GTY): Make profile_info gcov_summary
12748         struct.
12749         * profile.c: Likewise.
12750         * profile.h: Likewise.
12752 2018-06-05  Martin Liska  <mliska@suse.cz>
12754         PR gcov-profile/84846
12755         * gcov.c (output_lines): Print working directory only
12756         in intermediate format.
12758 2018-06-05  Andreas Krebbel  <krebbel@linux.ibm.com>
12760         * config/s390/s390-builtin-types.def: Add void function type.
12761         * config/s390/s390-builtins.def: Use the function type for the
12762         tbeginc builtin.
12764 2018-06-04  Jim Wilson  <jimw@sifive.com>
12766         * config/riscv/riscv-protos.h (riscv_expand_epilogue): Change bool arg
12767         to int.
12768         * config/riscv/riscv.c (riscv_for_each_saved_reg): New args epilogue
12769         and maybe_eh_return.  Change regno to unsigned int.  Use new args to
12770         handle EH_RETURN_DATA_REGNO registers properly.
12771         (riscv_expand_prologue): Pass new args to riscv_for_each_saved_reg.
12772         (riscv_expand_epilogue): Update comment.  Change argument name and
12773         type.  Update code to use new name and type.  Pass new args to
12774         riscv_for_each_saved_reg.  Only use EH_RETURN_STACKADJ_RTX when
12775         EXCEPTION_RETURN.
12776         * config/riscv/riscv.md (NORMAL_RETURN): New.
12777         (SIBCALL_RETURN, EXCEPTION_RETURN): New.
12778         (epilogue, sibcall_epilogue): Update riscv_expand_epilogue arg.
12779         (eh_return): Call gen_eh_return_internal and emit barrier.
12780         (eh_return_internal): Call riscv_expand_epilogue.
12782 2018-06-04  Eric Botcazou  <ebotcazou@adacore.com>
12784         * gimple-ssa-store-merging.c (struct merged_store_group): Move up
12785         bit_insertion field and declare can_be_merged_into method.
12786         (merged_store_group::can_be_merged_into): New method.
12787         (imm_store_chain_info::coalesce_immediate): Call it to decide whether
12788         consecutive non-overlapping stores can be merged.  Turn MEM_REF stores
12789         into BIT_INSERT_EXPR stores if the group contains a non-MEM_REF store.
12791 2018-06-04  Richard Biener  <rguenther@suse.de>
12793         PR tree-optimization/85955
12794         * builtins.c (fold_builtin_sincos): Convert pointers to
12795         destination to appropriate type before dereferencing.
12797 2018-06-04  Segher Boessenkool  <segher@kernel.crashing.org>
12799         * config/rs6000/rs6000.md (abs<mode>2 for FLOAT128): Handle IFmode.
12801 2018-06-04  Richard Sandiford  <richard.sandiford@linaro.org>
12803         * expr.c (expand_expr_real_1): Force the operand into memory if
12804         its TYPE_MODE is BLKmode and if there is no integer mode for
12805         the number of bits being extracted.
12807 2018-06-04  Jakub Jelinek  <jakub@redhat.com>
12809         PR target/85832
12810         PR target/86036
12811         * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
12812         Use vptestnm rather than vptestm in (=Yc,v,C) variant.
12814 2018-06-04  Richard Biener  <rguenther@suse.de>
12816         * tree-cfgcleanup.c (cleanup_tree_cfg_1): Fold into...
12817         (cleanup_tree_cfg_noloop): ... single caller.  Do
12818         start_recording_case_labels later.
12820 2018-06-04  Sebastian Peryt  <sebastian.peryt@intel.com>
12822         * config/i386/cldemoteintrin.h: Change define from _X86INTRIN_H_INCLUDED
12823         to _IMMINTRIN_H_INCLUDED.
12824         * config/i386/pconfigintrin.h: Ditto.
12825         * config/i386/waitpkgintrin.h: Ditto.
12826         * config/i386/immintrin.h: Add includes for sgxintrin.h,
12827         pconfigintrin.h, waitpkgintrin.h and cldemoteintrin.h.
12828         * config/i386/x86intrin.h: Remove includes for mintrin.h, xmmintrin.h,
12829         emmintrin.h, pmmintrin.h, tmmintrin.h, smmintrin.h, wmmintrin.h,
12830         bmiintrin.h, bmi2intrin.h, lzcntintrin.h, sgxintrin.h, pconfigintrin.h,
12831         waitpkgintrin.h and cldemoteintrin.h.
12833 2018-06-04  Richard Biener  <rguenther@suse.de>
12835         PR tree-optimization/86038
12836         * tracer.c (find_best_successor): Check probability for
12837         being initialized, bail out if not.
12839 2018-06-04  Richard Earnshaw  <rearnsha@arm.com>
12841         PR target/86003
12842         * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
12843         of bits to ignore when comparing architectures.
12845 2018-06-04  Jakub Jelinek  <jakub@redhat.com>
12847         PR tree-optimization/69615
12848         * fold-const.c (merge_ranges): If range1 is - [x, x] and x is the
12849         maximum or minimum of the type, try to merge it also as if
12850         range1 is + [-, x - 1] or + [x + 1, -].
12852         PR c++/86025
12853         * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
12855 2018-06-03  Eric Botcazou  <ebotcazou@adacore.com>
12857         PR tree-optimization/86034
12858         * gimple-ssa-store-merging.c (output_merged_store): Convert the RHS to
12859         the unsigned bitfield type in a bit insertion sequence if it does not
12860         have a larger precision than the bitfield size.
12861         (process_store): Also bypass widening conversions for BIT_INSERT_EXPR.
12863 2018-06-03  Kito Cheng  <kito.cheng@gmail.com>
12865         * config/nds32/nds32-peephole2.md: Add new patterns for code size.
12867 2018-06-03  Chung-Ju Wu  <jasonwucj@gmail.com>
12869         * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3J.
12870         * config/nds32/nds32.c (nds32_option_override): Consider ARCH_V3J.
12871         * config/nds32/nds32.h (TARGET_ISA_V3): Add ARCH_V3J checking.
12872         * config/nds32/nds32.opt (march): Add enum value Value(ARCH_V3J).
12874 2018-06-03  Chung-Ju Wu  <jasonwucj@gmail.com>
12876         * common/config/nds32/nds32-common.c (nds32_option_optimization_table):
12877         Disable -fdelete-null-pointer-checks for ELF toolchain.
12879 2018-06-02  Chung-Ju Wu  <jasonwucj@gmail.com>
12880             Kito Cheng  <kito.cheng@gmail.com>
12882         * config.gcc (nds32*): Use nds32-linux.opt and nds32-elf.opt.
12883         (nds32le-*-*, nds32be-*-*): Integrate checking process.
12884         (nds32*-*-*): Add glibc and uclibc conditions.
12885         * common/config/nds32/nds32-common.c (nds32_except_unwind_info): New.
12886         (TARGET_EXCEPT_UNWIND_INFO): Define.
12887         * config/nds32/elf.h: New file.
12888         * config/nds32/linux.h: New file.
12889         * config/nds32/nds32-elf.opt: New file.
12890         * config/nds32/nds32-linux.opt: New file.
12891         * config/nds32/nds32-fp-as-gp.c
12892         (pass_nds32_fp_as_gp::gate): Consider TARGET_LINUX_ABI.
12893         * config/nds32/nds32.c (nds32_conditional_register_usage): Consider
12894         TARGET_LINUX_ABI.
12895         (nds32_asm_file_end): Ditto.
12896         (nds32_print_operand): Ditto.
12897         (nds32_insert_attributes): Ditto.
12898         (nds32_init_libfuncs): New function.
12899         (TARGET_HAVE_TLS): Define.
12900         (TARGET_INIT_LIBFUNCS): Define.
12901         * config/nds32/nds32.h (TARGET_DEFAULT_RELAX): Apply different relax
12902         spec content.
12903         (TARGET_ELF): Apply different mcmodel setting.
12904         (LINK_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): The content has
12905         been migrated into elf.h and linux.h files.
12906         * config/nds32/nds32.md (add_pc): Consider TARGET_LINUX_ABI.
12907         * config/nds32/nds32.opt (mvh): Consider TARGET_LINUX_ABI.
12908         (mcmodel): The content has been migrated into nds32-elf.opt and
12909         nds32-linux.opt files.
12910         * config/nds32/t-elf: New file.
12911         * config/nds32/t-linux: New file.
12913 2018-06-02  Chung-Ju Wu  <jasonwucj@gmail.com>
12914             Shiva Chen  <shiva0217@gmail.com>
12916         * config/nds32/constants.md (unspec_volatile_element): Add
12917         UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END.
12918         * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp
12919         optimization.
12920         * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare.
12921         (make_pass_nds32_fp_as_gp): Declare.
12922         * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one
12923         optmization pass.
12924         (nds32_asm_function_end_prologue): Remove unused asm output.
12925         (nds32_asm_function_begin_epilogue): Remove unused asm output.
12926         (nds32_asm_file_start): Output necessary fp_as_gp information.
12927         (nds32_option_override): Adjust register usage.
12928         (nds32_expand_prologue): Consider fp_as_gp situation.
12929         (nds32_expand_prologue_v3push): Consider fp_as_gp situation.
12930         * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p.
12931         (epilogue): Ditto.
12932         (return): Ditto.
12933         (simple_return): Ditto.
12934         (omit_fp_begin): Output special directive for fp_as_gp.
12935         (omit_fp_end): Output special directive for fp_as_gp.
12936         * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp,
12937         mforbid-fp-as-gp): New options.
12939 2018-06-01  Mark Wielaard  <mark@klomp.org>
12941         * dwarf2out.c (dwarf2out_finish): Remove generation of
12942         DW_AT_loclists_base.
12944 2018-06-01  Eric Botcazou  <ebotcazou@adacore.com>
12946         * gimple-ssa-store-merging.c: Include gimple-fold.h.
12947         (struct store_immediate_info): Document BIT_INSERT_EXPR stores.
12948         (struct merged_store_group): Add bit_insertion field.
12949         (dump_char_array): Use standard hexadecimal format.
12950         (merged_store_group::merged_store_group): Set bit_insertion to false.
12951         (merged_store_group::apply_stores): Use optimal buffer size.  Deal
12952         with BIT_INSERT_EXPR stores.  Move up code updating the mask and
12953         also print the mask in the dump file.
12954         (pass_store_merging::gate): Minor tweak.
12955         (imm_store_chain_info::coalesce_immediate): Fix wrong association
12956         of stores with groups in dump.  Allow coalescing of BIT_INSERT_EXPR
12957         stores with INTEGER_CST stores.
12958         (count_multiple_uses) <BIT_INSERT_EXPR>: New case.
12959         (imm_store_chain_info::output_merged_store): Add try_bitpos variable
12960         and use it throughout.  Generate bit insertion sequences if need be.
12961         (pass_store_merging::process_store): Remove redundant condition.
12962         Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR.
12964 2018-06-01  Segher Boessenkool  <segher@kernel.crashing.org>
12966         * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
12967         the 128-bit floating point types.  Fix function comment.
12969 2018-06-01  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
12971         * config/aarch64/aarch64-simd.md
12972         (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler
12973         mnemonics.
12974         (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
12975         mnemonics.
12977 2018-06-01  Richard Sandiford  <richard.sandiford@linaro.org>
12979         PR tree-optimization/85989
12980         * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
12981         variable.
12982         (backprop::intersect_uses): Check it when deciding whether this
12983         is a backedge reference.
12984         (backprop::process_block): Add each phi to m_visited_phis
12985         after visiting it, then clear it at the end.
12987 2018-06-01  Richard Biener  <rguenther@suse.de>
12989         * tree-vectorizer.h (vect_dr_stmt): New function.
12990         (vect_get_load_cost): Adjust.
12991         (vect_get_store_cost): Likewise.
12992         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
12993         Use vect_dr_stmt instead of DR_SMTT.
12994         (vect_record_base_alignments): Likewise.
12995         (vect_calculate_target_alignment): Likewise.
12996         (vect_compute_data_ref_alignment): Likewise and make static.
12997         (vect_update_misalignment_for_peel): Likewise.
12998         (vect_verify_datarefs_alignment): Likewise.
12999         (vector_alignment_reachable_p): Likewise.
13000         (vect_get_data_access_cost): Likewise.  Pass down
13001         vinfo to vect_get_load_cost/vect_get_store_cost instead of DR.
13002         (vect_get_peeling_costs_all_drs): Likewise.
13003         (vect_peeling_hash_get_lowest_cost): Likewise.
13004         (vect_enhance_data_refs_alignment): Likewise.
13005         (vect_find_same_alignment_drs): Likewise.
13006         (vect_analyze_data_refs_alignment): Likewise.
13007         (vect_analyze_group_access_1): Likewise.
13008         (vect_analyze_group_access): Likewise.
13009         (vect_analyze_data_ref_access): Likewise.
13010         (vect_analyze_data_ref_accesses): Likewise.
13011         (vect_vfa_segment_size): Likewise.
13012         (vect_small_gap_p): Likewise.
13013         (vectorizable_with_step_bound_p): Likewise.
13014         (vect_prune_runtime_alias_test_list): Likewise.
13015         (vect_analyze_data_refs): Likewise.
13016         (vect_supportable_dr_alignment): Likewise.
13017         * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
13018         (vect_gen_prolog_loop_niters): Likewise.
13019         * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
13020         * tree-vect-patterns.c (vect_recog_bool_pattern): Do not
13021         modify DR_STMT.
13022         (vect_recog_mask_conversion_pattern): Likewise.
13023         (vect_try_gather_scatter_pattern): Likewise.
13024         * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info
13025         to vect_get_store_cost.
13026         (vect_get_store_cost): Get stmt_info instead of DR.
13027         (vect_model_load_cost): Pass stmt_info to vect_get_load_cost.
13028         (vect_get_load_cost): Get stmt_info instead of DR.
13030 2018-06-01  Richard Biener  <rguenther@suse.de>
13032         PR middle-end/86017
13033         * gimple-fold.c (var_decl_component_p): Also allow offsetted
13034         vars wrapped in MEM_REFs.
13036 2018-06-01  Richard Sandiford  <richard.sandiford@linaro.org>
13038         * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
13039         Fix subreg tests so that we only return a choice between
13040         GENERAL_REGS and FP_REGS if the original classes included both.
13042 2018-06-01  Richard Biener  <rguenther@suse.de>
13044         PR ipa/85960
13045         * tree-ssa-structalias.c (get_function_part_constraint):
13046         Handle NULL fi->decl.
13047         (find_func_aliases_for_call): Properly handle indirect
13048         fi from direct call.
13049         (find_func_clobbers): Likewise.
13050         (ipa_pta_execute): Likewise.
13051         (create_variable_info_for): For functions that are ifunc_resolver
13052         resolve to a varinfo that contains the result of the resolver call.
13053         (associate_varinfo_to_alias): Do not treat ifunc resolvers as
13054         aliases.
13056 2018-05-31  Michael Collison  <michael.collison@arm.com>
13058         * config/aarch64/aarch64.md:
13059         (*fix_to_zero_extenddfdi2): New pattern.
13060         * gcc.target/aarch64/fix_extend1.c: New testcase.
13062 2018-05-31  Qing Zhao <qing.zhao@oracle.com>
13064         PR middle-end/78809
13065         PR middle-end/83026
13066         * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
13067         and BUILT_IN_STRNCMP_EQ.
13068         * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
13069         BUILT_IN_STRNCMP_EQ.
13070         * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
13071         handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
13072         (gimple_fold_builtin): Likewise.
13073         * tree-ssa-strlen.c (compute_string_length): New function.
13074         (determine_min_obsize): New function.
13075         (handle_builtin_string_cmp): New function to handle calls to
13076         string compare functions.
13077         (strlen_optimize_stmt): Add handling to builtin string compare
13078         calls.
13079         * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
13080         Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
13081         * tree.c (build_common_builtin_nodes): Add new defines of
13082         BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
13084 2018-05-31  Jakub Jelinek  <jakub@redhat.com>
13086         PR target/85984
13087         * bb-reorder.c (pass_partition_blocks::gate): Return false for
13088         functions with naked attribute.
13090 2018-05-31  Uros Bizjak  <ubizjak@gmail.com>
13092         * config/i386/sse.md (avx_vec_concat<mode>):
13093         Substitute concat_tg_mode mode attribute with xtg_mode.
13094         (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
13095         (concat_tg_mode): Remove mode attribute.
13097 2018-05-31  Martin Sebor  <msebor@redhat.com>
13099         PR c/82063
13100         * calls.c (alloc_max_size): Correct a logic error/typo.
13101         Treat excessive arguments as infinite.  Warn for invalid arguments.
13102         * doc/invoke.texi (-Walloc-size-larger-than): Update.
13104 2018-05-31  H.J. Lu  <hongjiu.lu@intel.com>
13106         PR target/85829
13107         * config/i386/x86-tune.def: Re-enable partial_reg_dependency
13108         and movx for Haswell.
13110 2018-05-31  Chung-Lin Tang  <cltang@codesourcery.com>
13111             Cesar Philippidis  <cesar@codesourcery.com>
13113         PR middle-end/85879
13114         * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
13115         when emitting error on private/firstprivate reductions.
13116         * omp-low.c (lower_omp_target): Avoid reference-type processing
13117         on pointers for firstprivate clause.
13119 2018-05-31  Sameera Deshpande <sameera.deshpande@linaro.org>
13121         * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
13122         (st1x2): Likewise.
13123         (st1x3): Likewise.
13124         * config/aarch64/aarch64-simd.md
13125         (aarch64_ld1x3<VALLDIF:mode>): New pattern.
13126         (aarch64_ld1_x3_<mode>): Likewise
13127         (aarch64_st1x2<VALLDIF:mode>): Likewise
13128         (aarch64_st1_x2_<mode>): Likewise
13129         (aarch64_st1x3<VALLDIF:mode>): Likewise
13130         (aarch64_st1_x3_<mode>): Likewise
13131         * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
13132         (vld1_s8_x3): Likewise.
13133         (vld1_u16_x3): Likewise.
13134         (vld1_s16_x3): Likewise.
13135         (vld1_u32_x3): Likewise.
13136         (vld1_s32_x3): Likewise.
13137         (vld1_u64_x3): Likewise.
13138         (vld1_s64_x3): Likewise.
13139         (vld1_f16_x3): Likewise.
13140         (vld1_f32_x3): Likewise.
13141         (vld1_f64_x3): Likewise.
13142         (vld1_p8_x3): Likewise.
13143         (vld1_p16_x3): Likewise.
13144         (vld1_p64_x3): Likewise.
13145         (vld1q_u8_x3): Likewise.
13146         (vld1q_s8_x3): Likewise.
13147         (vld1q_u16_x3): Likewise.
13148         (vld1q_s16_x3): Likewise.
13149         (vld1q_u32_x3): Likewise.
13150         (vld1q_s32_x3): Likewise.
13151         (vld1q_u64_x3): Likewise.
13152         (vld1q_s64_x3): Likewise.
13153         (vld1q_f16_x3): Likewise.
13154         (vld1q_f32_x3): Likewise.
13155         (vld1q_f64_x3): Likewise.
13156         (vld1q_p8_x3): Likewise.
13157         (vld1q_p16_x3): Likewise.
13158         (vld1q_p64_x3): Likewise.
13159         (vst1_s64_x2): Likewise.
13160         (vst1_u64_x2): Likewise.
13161         (vst1_f64_x2): Likewise.
13162         (vst1_s8_x2): Likewise.
13163         (vst1_p8_x2): Likewise.
13164         (vst1_s16_x2): Likewise.
13165         (vst1_p16_x2): Likewise.
13166         (vst1_s32_x2): Likewise.
13167         (vst1_u8_x2): Likewise.
13168         (vst1_u16_x2): Likewise.
13169         (vst1_u32_x2): Likewise.
13170         (vst1_f16_x2): Likewise.
13171         (vst1_f32_x2): Likewise.
13172         (vst1_p64_x2): Likewise.
13173         (vst1q_s8_x2): Likewise.
13174         (vst1q_p8_x2): Likewise.
13175         (vst1q_s16_x2): Likewise.
13176         (vst1q_p16_x2): Likewise.
13177         (vst1q_s32_x2): Likewise.
13178         (vst1q_s64_x2): Likewise.
13179         (vst1q_u8_x2): Likewise.
13180         (vst1q_u16_x2): Likewise.
13181         (vst1q_u32_x2): Likewise.
13182         (vst1q_u64_x2): Likewise.
13183         (vst1q_f16_x2): Likewise.
13184         (vst1q_f32_x2): Likewise.
13185         (vst1q_f64_x2): Likewise.
13186         (vst1q_p64_x2): Likewise.
13187         (vst1_s64_x3): Likewise.
13188         (vst1_u64_x3): Likewise.
13189         (vst1_f64_x3): Likewise.
13190         (vst1_s8_x3): Likewise.
13191         (vst1_p8_x3): Likewise.
13192         (vst1_s16_x3): Likewise.
13193         (vst1_p16_x3): Likewise.
13194         (vst1_s32_x3): Likewise.
13195         (vst1_u8_x3): Likewise.
13196         (vst1_u16_x3): Likewise.
13197         (vst1_u32_x3): Likewise.
13198         (vst1_f16_x3): Likewise.
13199         (vst1_f32_x3): Likewise.
13200         (vst1_p64_x3): Likewise.
13201         (vst1q_s8_x3): Likewise.
13202         (vst1q_p8_x3): Likewise.
13203         (vst1q_s16_x3): Likewise.
13204         (vst1q_p16_x3): Likewise.
13205         (vst1q_s32_x3): Likewise.
13206         (vst1q_s64_x3): Likewise.
13207         (vst1q_u8_x3): Likewise.
13208         (vst1q_u16_x3): Likewise.
13209         (vst1q_u32_x3): Likewise.
13210         (vst1q_u64_x3): Likewise.
13211         (vst1q_f16_x3): Likewise.
13212         (vst1q_f32_x3): Likewise.
13213         (vst1q_f64_x3): Likewise.
13214         (vst1q_p64_x3): Likewise.
13216 2018-05-30  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
13218         * config/msp430/msp430.c (msp430_output_labelref): Prepend
13219         user_label_prefix to name.
13221         * tree-core.h: Update comment about the format of NAME string
13222         passed to handler in attribute_spec.
13224         * config/msp430/msp430.md: Remove erroneous subreg expression from
13225         zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
13226         zero_extend{q,h}isi2.
13228 2018-05-30  Borislav Petkov  <bp@suse.de>
13230         * doc/extend.texi: Document some architecture specific
13231         constraints and sort entries.
13233 2018-05-30  Martin Sebor  <msebor@redhat.com>
13235         PR middle-end/85369
13236         * builtins.c (expand_builtin_stpcpy_1): New function.
13237         (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
13238         only if the former succeeds.
13240 2018-05-31  Sameera Deshpande <sameera.deshpande@linaro.org>
13242         * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
13243         in saphira.
13245 2018-05-30  Jan Hubicka  <hubicka@ucw.cz>
13247         * doc/invoke.texi (-flinker-output): Document
13249 2018-05-30  Jan Hubicka  <hubicka@ucw.cz>
13251         * passes.c (ipa_write_summaries): Only modify statements if body
13252         is in memory.
13253         * cgraphunit.c (ipa_passes): Also produce intermeidate code when
13254         incrementally linking.
13255         (ipa_passes): Likewise.
13256         * lto-cgraph.c (lto_output_node): When incrementally linking do not
13257         pass down resolution info.
13258         * common.opt (flag_incremental_link): Update info.
13259         * gcc.c (plugin specs): Turn flinker-output=* to
13260         -plugin-opt=-linker-output-known
13261         * toplev.c (compile_file): Also cut compilation when doing incremental
13262         link.
13263         * flag-types. (enum lto_partition_model): Add
13264         LTO_LINKER_OUTPUT_NOLTOREL.
13265         (invoke.texi): Add -flinker-output docs.
13266         * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
13267         link same way as WPA; do not stream in dead initializers.
13269         * dwarf2out.c (dwarf2out_die_ref_for_decl,
13270         darf2out_register_external_decl): Support incremental link.
13272 2018-05-30  Jan Hubicka  <hubicka@ucw.cz>
13274         * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
13276 2018-05-30  Jan Hubicka  <hubicka@ucw.cz>
13278         * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
13279         it down to simple_object_copy_lto_debug_sections.
13280         (run_gcc): Determine incremental LTO link time and configure
13281         lto1 into non-wpa mode, disable renaming of debug sections.
13283 2018-05-30  Kelvin Nilsen  <kelvin@gcc.gnu.org>
13285         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
13286         descriptions of various incorrectly documented functions.
13288 2018-05-30  Andre Vieira  <andre.simoesdiasvieira@arm.com>
13290         Revert:
13291         * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
13292         address check not strict.
13294 2018-05-30  Richard Biener  <rguenther@suse.de>
13296         PR tree-optimization/85964
13297         * tracer.c (better_p): Drop initialized count check, we only
13298         call the function with initialized counts now.
13299         (find_best_successor): Do find a best edge if one
13300         has uninitialized count.
13301         (find_best_predecessor): Likewise.  Do BB frequency check only
13302         if count is initialized.
13304 2017-05-30  Jackson Woodruff  <jackson.woodruff@arm.com>
13306         * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
13307         (aarch64_ldrstr_offset_compare): New.
13308         (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
13309         load/store orderings.
13310         (aarch64_gen_adjusted_ldpstp): Likewise.
13312 2018-05-30  Wilco Dijkstra  <wdijkstr@arm.com>
13314         * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
13315         Check for subset of GENERAL_REGS and FP_REGS.
13316         * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
13317         r=w alternative.
13319 2018-05-30  Richard Sandiford  <richard.sandiford@linaro.org>
13321         * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
13322         and wi::to_poly_offset.  Add the current offset and then check
13323         whether the sum fits, rather than using an unchecked addition of
13324         a checked term.  Check for a shwi rather than a uhwi.
13325         * expr.c (get_bit_range): Use tree_to_poly_uint64.
13326         (store_constructor): Use poly_int_tree_p.
13327         (expand_expr_real_1): Likewise.
13328         * function.c (assign_temp): Likewise.
13329         * fold-const.c (const_binop): Use poly_int_tree_p and
13330         wi::to_poly_offset.
13331         (fold_indirect_ref_1): Likewise.  Use multiple_p to attempt an exact
13332         division.
13333         * ipa-icf-gimple.c (func_checker::compare_operand): Use
13334         to_poly_offset for MEM offsets.
13335         * ipa-icf.c (sem_variable::equals): Likewise.
13336         * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
13337         * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
13338         wi::to_poly_offset for BIT_FIELD_REF offsets.
13339         (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
13340         wi::to_poly_offset.
13341         * var-tracking.c (emit_note_insn_var_location): Use
13342         tree_to_poly_uint64.
13344 2018-05-29  Jim Wilson  <jimw@sifive.com>
13346         * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
13348 2018-05-29  Uros Bizjak  <ubizjak@gmail.com>
13350         PR target/85950
13351         * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
13352         Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
13353         sequence.
13354         (sse4_1_round<mode>2): Use nonimmediate_operand
13355         for operand 1 predicate.
13357 2018-05-29  Martin Sebor  <msebor@redhat.com>
13358             Richard Biener  <rguenther@suse.de>
13360         PR testsuite/85888
13361         * calls.c (get_size_range): Call determine_value_range instead
13362         of get_value_range..
13363         * tree-vrp.h (determine_value_range): Declared new function.
13364         * tree-vrp.c (determine_value_range_1, determine_value_range): New.
13366 2018-05-29  Richard Biener  <rguenther@suse.de>
13368         * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
13369         sure to use non-pattern stmts for get_earlier_stmt arguments.
13370         * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
13371         called on pattern stmts.
13372         (get_later_stmt): Likewise.
13374 2018-05-29  Martin Liska  <mliska@suse.cz>
13376         PR gcov-profile/85759
13377         * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
13378         env variables.
13380 2018-05-29  Jakub Jelinek  <jakub@redhat.com>
13382         * tree-cfg.c (verify_gimple_assign_unary): Add checking for
13383         VEC_UNPACK_*_EXPR.
13384         (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
13385         VEC_PACK_*_EXPR.
13387         PR target/85918
13388         * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
13389         VEC_PACK_FLOAT_EXPR): New tree codes.
13390         * tree-pretty-print.c (op_code_prio): Handle
13391         VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
13392         (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
13393         VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
13394         * tree-inline.c (estimate_operator_cost): Likewise.
13395         * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
13396         * fold-const.c (const_binop): Likewise.
13397         (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
13398         VEC_UNPACK_FIX_TRUNC_LO_EXPR.
13399         * tree-cfg.c (verify_gimple_assign_unary): Likewise.
13400         (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
13401         * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
13402         VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
13403         * expr.c (expand_expr_real_2): Likewise.
13404         * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
13405         vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
13406         vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
13407         optabs.
13408         * optabs.c (expand_widen_pattern_expr): For
13409         VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
13410         sign from result type rather than operand's type.
13411         (expand_binop_directly): For vec_packu_float_optab and
13412         vec_packs_float_optab allow result type to be different from operand's
13413         type.
13414         * optabs-tree.c (optab_for_tree_code): Handle
13415         VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
13416         VEC_PACK_FLOAT_EXPR.  Formatting fixes.
13417         * tree-vect-generic.c (expand_vector_operations_1):  Handle
13418         VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
13419         VEC_PACK_FLOAT_EXPR.
13420         * tree-vect-stmts.c (supportable_widening_operation): Handle
13421         FIX_TRUNC_EXPR.
13422         (supportable_narrowing_operation): Handle FLOAT_EXPR.
13423         * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
13424         * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
13425         (float<floatunssuffix>v2div2sf2): ... this.  Formatting fix.
13426         (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
13427         mode attributes.
13428         (vec_pack<floatprefix>_float_<mode>): New expander.
13429         (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
13430         attributes.
13431         (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
13432         vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
13433         * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
13434         vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
13435         vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
13436         Document.
13437         * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
13438         VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
13439         (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
13440         VEC_PACK_FLOAT_EXPR): Document.
13442 2018-05-29  Richard Biener  <rguenther@suse.de>
13444         * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
13445         member.
13446         (stmt_vec_info_vec): Make pointer.
13447         (init_stmt_vec_info_vec): Remove.
13448         (free_stmt_vec_info_vec): Likewise.
13449         (set_stmt_vec_info_vec): New function.
13450         (free_stmt_vec_infos): Likewise.
13451         (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
13452         (set_vinfo_for_stmt): Likewise.
13453         (get_earlier_stmt): Likewise.
13454         (get_later_stmt): Likewise.
13455         * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
13456         (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
13457         (vec_info::~vec_info): Free stmt_vec_infos.
13458         (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
13459         Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
13460         (pass_slp_vectorize::execute): Likewise.
13461         * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
13462         (free_stmt_vec_info_vec): Likewise.
13463         (set_stmt_vec_info_vec): New function.
13464         (free_stmt_vec_infos): Likewise.
13465         * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
13466         the global stmt_vec_info_vec.
13467         * tree-parloops.c (gather_scalar_reductions): Use
13468         set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
13469         vector.
13471 2018-05-29  Richard Biener  <rguenther@suse.de>
13473         * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
13475 2018-05-29  Martin Liska  <mliska@suse.cz>
13476             David Malcolm  <dmalcolm@redhat.com>
13478         * vec.c (test_reverse): New.
13479         (vec_c_tests): Add new test.
13480         * vec.h (vl_ptr>::reverse): New function.
13482 2018-05-29  Gerald Pfeifer  <gerald@pfeifer.com>
13484         * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
13486         * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
13487         and later.
13489 2018-05-28  Bernd Edlinger  <bernd.edlinger@hotmail.de>
13491         * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
13493 2018-05-28  Richard Biener  <rguenther@suse.de>
13495         PR tree-optimization/85933
13496         * tree-vect-data-refs.c (vect_record_base_alignments): Only
13497         look at stmts marked as vectorizable.
13499 2018-05-28  Richard Biener  <rguenther@suse.de>
13501         PR tree-optimization/85934
13502         * tree-vect-generic.c (expand_vector_operations_1): Hoist
13503         vector boolean check before scalar optimization.
13505 2018-05-28  Jakub Jelinek  <jakub@redhat.com>
13507         * doc/invoke.texi (ARM Options): Use @item instead of @itemx
13508         for armv5te.
13510 2018-05-28  Mark Wielaard  <mark@klomp.org>
13512         * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
13513         if it is an expression containing a minus sign.
13515 2018-05-27  John David Anglin  <danglin@gcc.gnu.org>
13517         * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
13519 2018-05-27  Paul Koning  <ni1d@arrl.net>
13521         * config/pdp11/pdp11.md (truncsihi2): Remove.
13523 2018-05-27  Monk Chiang  <sh.chiang04@gmail.com>
13524             Chung-Ju Wu  <jasonwucj@gmail.com>
13526         * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
13527         implementation.
13528         (unaligned_store_dw): Ditto.
13529         * config/nds32/nds32-memory-manipulation.c
13530         (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
13531         (nds32_gen_dup_4_byte_to_word_value): Rename to ...
13532         (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
13533         (emit_setmem_word_loop): Rename to ...
13534         (emit_setmem_doubleword_loop): ... this.
13535         (nds32_gen_dup_4_byte_to_word_value): New function.
13536         (nds32_gen_dup_8_byte_to_double_word_value): New function.
13537         (nds32_expand_setmem_loop): Refine implementation.
13538         (nds32_expand_setmem_loop_v3m): Ditto.
13539         * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
13540         pattern.
13542 2018-05-27  Chung-Ju Wu  <jasonwucj@gmail.com>
13544         * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
13546 2018-05-27  Chung-Ju Wu  <jasonwucj@gmail.com>
13548         * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
13549         (nds32_init_machine_status): Initialize machine->attr_naked_p and
13550         machine->attr_no_prologue_p.
13551         (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
13552         (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
13553         (nds32_expand_epilogue): Consider attr_naked_p.
13554         (nds32_expand_epilogue_v3pop): Likewise.
13555         (nds32_can_use_return_insn): Likewise.
13556         * config/nds32/nds32.h (machine_function): Add attr_naked_p and
13557         attr_no_prologue_p fields.
13558         * config/nds32/nds32.opt (mret-in-naked-func): New option.
13560 2018-05-27  Jakub Jelinek  <jakub@redhat.com>
13562         PR target/85918
13563         * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
13564         attributes.
13565         * config/i386/sse.md
13566         (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
13567         Rename to ...
13568         (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
13569         ... this.
13570         (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
13571         Rename to ...
13572         (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
13573         ... this.
13574         (*<floatsuffix>floatv2div2sf2): Rename to ...
13575         (*float<floatunssuffix>v2div2sf2): ... this.
13576         (<floatsuffix>floatv2div2sf2_mask): Rename to ...
13577         (float<floatunssuffix>v2div2sf2_mask): ... this.
13578         (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
13579         (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
13580         (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
13581         to ...
13582         (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
13583         ... this.
13584         (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
13585         Rename to ...
13586         (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
13587         ... this.
13588         (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
13589         Rename to ...
13590         (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
13591         ... this.
13592         (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
13593         (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
13594         (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
13595         gen_ufix_truncv8dfv8si2.
13596         * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
13597         __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
13598         __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
13599         __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
13600         __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
13601         __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
13602         __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
13603         Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
13605 2018-05-24  H.J. Lu  <hongjiu.lu@intel.com>
13607         PR target/85900
13608         PR target/85345
13609         * varasm.c (assemble_alias): Lookup ifunc attribute on error.
13611 2018-05-25  Jim Wilson  <jimw@sifive.com>
13613         * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
13614         * config/riscv/riscv.c (struct machine_function): Add
13615         interrupt_handler_p and attribute_checked_p fields.
13616         (riscv_attribute_table): Add interrupt.
13617         (riscv_interrupt_type_p): New.
13618         (riscv_save_reg_p): Save extra regs for interrupt handler.
13619         (riscv_use_save_libcall): Return false  for interrupt handler.
13620         (riscv_first_stack_step): Add forward declaration.
13621         (riscv_compute_frame_info): New local interrupt_save_t1.  Set it
13622         for interrupt handler with large frame.  Use it for saved reg list.
13623         (riscv_expand_prologue): Move flag_stack_usage_info support to
13624         eliminate duplication.
13625         (riscv_expand_epilogue): Generate mret for interrupt handler.
13626         (riscv_epilogue_uses): New.
13627         (riscv_can_use_return_insn): Return false for interrupt handler.
13628         (riscv_function_ok_for_sibcall): Likewise.
13629         (riscv_set_current_function): Add interrupt handler support.
13630         * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
13631         * config/riscv/riscv.md (UNSPECV_MRET): New.
13632         (GP_REGNUM): New.
13633         (riscv_frflags, riscv_fsflags): Use tab after opcode.
13634         (riscv_mret): New.
13635         * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
13637 2018-05-25  Bill Schmidt  <wschmidt@linux.ibm.com>
13639         PR tree-optimization/85712
13640         * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
13641         this candidate has already been replaced in-situ by a copy.
13643 2018-05-25  Jason Merrill  <jason@redhat.com>
13645         PR c++/80485 - inline function non-zero address.
13646         * symtab.c (nonzero_address): Check DECL_COMDAT.
13648 2018-05-25  Uros Bizjak  <ubizjak@gmail.com>
13650         PR target/83628
13651         * config/alpha/alpha.md (ashlsi3): New insn pattern.
13652         (*ashlsi_se): Rename from *ashldi_se.  Define as sign
13653         extension of SImode operation.  Use const123_operand predicate.
13654         (*saddsi_1): Remove.
13655         (*saddl_se_1): Ditto.
13656         (*ssubsi_1): Ditto.
13657         (*ssubl_se_1): Ditto.
13658         * config/alpha/predicates.md (const123_operand): New predicate.
13659         * config/alpha/constraints.md (P): Use IN_RANGE.
13661 2018-05-25  Richard Biener  <rguenther@suse.de>
13663         * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
13664         defaulted to true.
13665         (ref_maybe_used_by_stmt_p): Likewise.
13666         (stmt_may_clobber_ref_p): Likewise.
13667         (stmt_may_clobber_ref_p_1): Likewise.
13668         * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
13669         and pass it along.
13670         (ref_maybe_used_by_stmt_p): Likewise.
13671         (stmt_may_clobber_ref_p): Likewise.
13672         (stmt_may_clobber_ref_p_1): Likewise.
13673         * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
13674         the alias oracle to disambiguate DRs with stmts DR analysis
13675         couldn't handle.
13676         (vect_analyze_data_refs): Do not give up on not analyzable
13677         DRs for BB vectorization.  Remove code truncating the dataref
13678         vector.
13680 2018-05-25  Jakub Jelinek  <jakub@redhat.com>
13682         PR target/85832
13683         * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
13684         Add (=Yk,v,C) variant using vptestm insn.  Use TARGET_AVX512BW
13685         in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
13687 2018-05-25  Richard Biener  <rguenther@suse.de>
13689         * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
13690         function, combining stmt data ref gathering and fatal analysis
13691         parts.
13692         (vect_analyze_data_refs): Remove now redudnant code and simplify.
13693         * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
13694         vect_analyze_loop_2 and use vect_find_stmt_data_reference.
13695         * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
13696         * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
13698 2018-05-25  Bin Cheng  <bin.cheng@arm.com>
13700         PR tree-optimization/85720
13701         * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
13702         SCC if all partitions are builtins.
13703         (version_loop_by_alias_check): New parameter.  Generate cancelable
13704         runtime alias check if all partitions are builtins.
13705         (distribute_loop): Update call to above function.
13707 2018-05-25  Bin Cheng  <bin.cheng@arm.com>
13709         * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
13710         (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
13711         (parm_default_def_partition_arg): Ditto.
13712         (set_parm_default_def_partition): Ditto.
13713         (get_parm_default_def_partitions): Ditto and make it static.
13714         (get_undefined_value_partitions): Ditto and make it static.
13715         (remove_ssa_form): Refactor call to init_var_map here.
13716         * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
13717         computation for loop region.
13718         (coalesce_partitions, compute_optimized_partition_bases): Ditto.
13719         (register_default_def): Delete.
13720         (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
13721         (parm_default_def_partition_arg): Ditto.
13722         (set_parm_default_def_partition): Ditto.
13723         (get_parm_default_def_partitions): Ditto and make it static.
13724         (get_undefined_value_partitions): Ditto and make it static.
13725         (coalesce_with_default, coalesce_with_default): Update comment.
13726         (create_coalesce_list_for_region): New func factored out from
13727         create_outofssa_var_map.
13728         (populate_coalesce_list_for_outofssa): New func factored out from
13729         create_outofssa_var_map and coalesce_ssa_name.
13730         (create_outofssa_var_map): Delete.
13731         (coalesce_ssa_name): Refactor to support live range computation.
13732         * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
13733         (get_parm_default_def_partitions): Delete.
13734         (get_undefined_value_partitions): Ditto.
13735         * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
13736         computation for loop region.
13737         (new_tree_live_info, loe_visit_block): Ditto.
13738         (live_worklist, set_var_live_on_entry): Ditto.
13739         (calculate_live_on_exit, verify_live_on_entry): Ditto.
13740         * tree-ssa-live.h (struct _var_map): New fields.
13741         (init_var_map): Change decl.
13742         (region_contains_p): New.
13744 2018-05-25  Bin Cheng  <bin.cheng@arm.com>
13746         * tree-ssa-live.h (live_merge_and_clear): Delete.
13748 2018-05-25  Richard Biener  <rguenther@suse.de>
13750         PR c++/85912
13751         * tree-dump.c (dequeue_and_dump): Remove access to removed
13752         operand 2 of a SWITCH_EXPR.
13754 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
13756         * doc/sourcebuild.texi (vect_double_cond_arith): Include
13757         multiplication and division.
13758         * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
13759         (cond_udiv@var{m}, cond_umod@var{m}): Document.
13760         * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
13761         (cond_udiv_optab, cond_umod_optab): New optabs.
13762         * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
13763         (IFN_COND_RDIV): New internal functions.
13764         * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
13765         TRUNC_MOD_EXPR and RDIV_EXPR.
13766         * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
13767         * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
13768         New unspecs.
13769         (SVE_INT_BINARY): Include mult.
13770         (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
13771         (optab, sve_int_op): Handle mult.
13772         (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
13773         UNSPEC_COND_DIV.
13774         * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
13775         for SVE_INT_BINARY_SD.
13777 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
13779         * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
13780         (optab, sve_int_op): Handle div and udiv.
13781         * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
13782         for SVE_INT_BINARY_SD.
13783         (*<optab><mode>3): New insn for the same.
13785 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
13787         * tree-vect-patterns.c: Include predict.h.
13788         (vect_recog_divmod_pattern): Restrict check for division support
13789         to when optimizing for size.
13791 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
13793         * doc/sourcebuild.texi (vect_double_cond_arith: Document.
13794         * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
13795         (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
13796         (gimple_match_op::set_op): Likewise.
13797         (gimple_resimplify4): Declare.
13798         * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
13799         (expr::gen_transform): Likewise.
13800         (decision_tree::gen): Generate a simplification routine for 4 operands.
13801         * gimple-match-head.c (gimple_simplify): Add an overload for
13802         4 operands.  In the top-level function, handle up to 4 call
13803         arguments and call gimple_resimplify4.
13804         (gimple_resimplify4): New function.
13805         (build_call_internal): Pass a fourth operand.
13806         (maybe_push_to_seq): Likewise.
13807         * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
13808         Fold VEC_COND_EXPRs of an operation and a default value into
13809         an IFN_COND_* function if possible.
13810         * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
13811         New unspecs.
13812         (SVE_COND_FP_BINARY): Include them.
13813         (optab, sve_fp_op): Handle them.
13814         (SVE_INT_BINARY_REV): New code iterator.
13815         (SVE_COND_FP_BINARY_REV): New int iterator.
13816         (commutative): New int attribute.
13817         * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
13818         Declare.
13819         * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
13820         function.
13821         * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
13822         (*cond_<optab><mode>): New patterns for reversed operands.
13824 2018-05-25  Richard Biener  <rguenther@suse.de>
13826         * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
13827         (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
13828         (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
13829         (STMT_VINFO_GROUPED_ACCESS): Adjust.
13830         * tree-vect-data-refs.c (everywhere): Adjust users.
13831         * tree-vect-loop.c (everywhere): Likewise.
13832         * tree-vect-slp.c (everywhere): Likewise.
13833         * tree-vect-stmts.c (everywhere): Likewise.
13834         * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
13836 2018-05-25  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
13838         * configure.ac (gcc_cv_as_section_has_e): Move to common section.
13839         Rename to...
13840         (gcc_cv_as_section_exclude): ... this.
13841         Try Solaris as #exclude syntax.
13842         * configure: Regenerate.
13843         * config.in: Regenerate.
13844         * config/i386/i386.c (i386_solaris_elf_named_section): Handle
13845         SECTION_EXCLUDE.
13846         * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
13847         [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
13849         * varasm.c (default_elf_asm_named_section): Don't check if
13850         HAVE_GAS_SECTION_EXCLUDE is defined.
13852 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
13854         * doc/md.texi: Update the documentation of the cond_* optabs
13855         to mention the new final operand.  Fix GET_MODE_NUNITS call.
13856         Describe the scalar case too.
13857         * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
13858         * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
13859         instead of 2.
13860         (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
13861         (get_conditional_internal_fn): Update comment.
13862         * tree-vect-loop.c (vectorizable_reduction): Pass the original
13863         accumulator value as a final argument to conditional functions.
13864         * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
13865         a define_expand and add an "else" operand.  Assert for now that
13866         the else operand is equal to operand 2.  Use SVE_INT_BINARY and
13867         SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
13868         (*cond_<optab><mode>): New patterns.
13869         * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
13870         (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
13871         (UNSPEC_COND_EOR): Delete.
13872         (optab): Remove associated mappings.
13873         (SVE_INT_BINARY): New code iterator.
13874         (sve_int_op): Remove int attribute and add "minus" to the code
13875         attribute.
13876         (SVE_COND_INT_OP): Delete.
13877         (SVE_COND_FP_OP): Rename to...
13878         (SVE_COND_FP_BINARY): ...this.
13880 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
13882         * optabs.c (can_reuse_operands_p): New function.
13883         (maybe_legitimize_operands): Try to reuse the results for
13884         earlier operands.
13886 2018-05-24  Uros Bizjak  <ubizjak@gmail.com>
13888         * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
13889         Add {q} suffix to insn mnemonic.
13891 2018-05-23  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
13893         * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
13894         (msp430_warn_func_return): New.
13896 2018-05-24  Roger Sayle  <roger@nextmovesoftware.com>
13898         * fold-const.c (tree_nonzero_bits): New function.
13899         * fold-const.h (tree_nonzero_bits): Likewise.
13900         * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
13901         friends.  POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
13903 2018-05-24  H.J. Lu  <hongjiu.lu@intel.com>
13905         PR target/85900
13906         PR target/85345
13907         * varasm.c (assemble_alias): Check ifunc_resolver only on
13908         FUNCTION_DECL.
13910 2018-05-24  Uros Bizjak  <ubizjak@gmail.com>
13912         PR target/85903
13913         * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
13914         when memory input operand is handled.
13916 2018-05-24  Luis Machado  <luis.machado@linaro.org>
13918         * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
13919         global.
13920         (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
13922 2018-05-24  Richard Sandiford  <richard.sandiford@linaro.org>
13924         * match.pd: Delay FMA folds until after vectorization.
13926 2018-05-24  Andre Vieira  <andre.simoesdiasvieira@arm.com>
13928         PR target/83009
13929         * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
13930         address check not strict.
13932 2018-05-24  Richard Sandiford  <richard.sandiford@linaro.org>
13934         * gimple-match.h (gimple_match_op): New class.
13935         (mprts_hook): Replace parameters with a gimple_match_op *.
13936         (maybe_build_generic_op): Likewise.
13937         (gimple_simplified_result_is_gimple_val): Replace parameters with
13938         a const gimple_match_op *.
13939         (gimple_simplify): Replace code_helper * and tree * parameters with
13940         a gimple_match_op * parameter.
13941         (gimple_resimplify1): Replace code_helper *, tree and tree *
13942         parameters with a gimple_match_op * parameter.
13943         (gimple_resimplify2): Likewise.
13944         (gimple_resimplify3): Likewise.
13945         (maybe_push_res_to_seq): Replace code_helper, tree and tree *
13946         parameters with a gimple_match_op * parameter.
13947         * gimple-match-head.c (gimple_simplify): Change prototypes of
13948         auto-generated functions to take a gimple_match_op * instead of
13949         separate code_helper * and tree * parameters.  Make the same
13950         change in the top-level overload and update calls to the
13951         gimple_resimplify routines.  Update calls to the auto-generated
13952         functions and to maybe_push_res_to_seq in the publicly-facing
13953         operation-specific gimple_simplify overloads.
13954         (gimple_match_op::MAX_NUM_OPS): Define.
13955         (gimple_resimplify1): Replace rcode and ops with a single res_op
13956         parameter.  Update call to gimple_simplify.
13957         (gimple_resimplify2): Likewise.
13958         (gimple_resimplify3): Likewise.
13959         (mprts_hook): Replace parameters with a gimple_match_op *.
13960         (maybe_build_generic_op): Likewise.
13961         (build_call_internal): Replace type, nargs and ops with
13962         a gimple_match_op *.
13963         (maybe_push_res_to_seq): Replace res_code, type and ops parameters
13964         with a single gimple_match_op *.  Update calls to mprts_hook,
13965         build_call_internal and gimple_simplified_result_is_gimple_val.
13966         Factor out code that is common to the tree_code and combined_fn cases.
13967         * genmatch.c (expr::gen_transform): Replace tem_code and
13968         tem_ops with a gimple_match_op called tem_op.  Update calls
13969         to the gimple_resimplify functions and maybe_push_res_to_seq.
13970         (dt_simplify::gen_1): Manipulate res_op instead of res_code and
13971         res_ops.  Update call to the gimple_resimplify functions.
13972         (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
13973         (decision_tree::gen): Make the functions take a gimple_match_op *
13974         called res_op instead of separate res_code and res_ops parameters.
13975         Update call accordingly.
13976         * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
13977         and ops with a single res_op parameter.  Update calls to
13978         maybe_build_generic_op and maybe_push_res_to_seq.
13979         (fold_stmt_1): Update calls to gimple_simplify and
13980         replace_stmt_with_simplification.
13981         (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
13982         and gimple_simplified_result_is_gimple_val.
13983         * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
13984         gimple_simplify.
13985         * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
13986         with a gimple_match_op *.
13987         (vn_nary_build_or_lookup): Likewise.  Update call to
13988         vn_nary_build_or_lookup_1.
13989         (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
13990         gimple_match_op *.  Update calls to the gimple_resimplify routines
13991         and to gimple_simplified_result_is_gimple_val.
13992         (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
13993         Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
13994         (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
13995         (visit_nary_op): Likewise.
13996         (visit_reference_op_load): Likewise.
13998 2018-05-23  Luis Machado  <luis.machado@linaro.org>
14000         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
14001         modifier for printing the step amount.
14003 2018-05-23  Jozef Lawrynowicz  <jozef.l@somniumtech.com>
14005         PR target/78849
14006         * tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
14007         types.
14009 2018-05-23  Segher Boessenkool  <segher@kernel.crashing.org>
14011         * doc/sourcebuild.texi (Endianness): New subsubsection.
14013 2018-05-23  Luis Machado  <luis.machado@linaro.org>
14015         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
14016         <prefetch_dynamic_strides>: New const bool field.
14017         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
14018         prefetch_dynamic_strides.
14019         (exynosm1_prefetch_tune): Likewise.
14020         (thunderxt88_prefetch_tune): Likewise.
14021         (thunderx_prefetch_tune): Likewise.
14022         (thunderx2t99_prefetch_tune): Likewise.
14023         (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
14024         false.
14025         (aarch64_override_options_internal): Update to set
14026         PARAM_PREFETCH_DYNAMIC_STRIDES.
14027         * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
14028         * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
14029         * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
14030         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
14031         prefetch-dynamic-strides setting.
14033 2018-05-23  Luis Machado  <luis.machado@linaro.org>
14035         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
14036         <minimum_stride>: New const int field.
14037         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
14038         minimum_stride field defaulting to -1.
14039         (exynosm1_prefetch_tune): Likewise.
14040         (thunderxt88_prefetch_tune): Likewise.
14041         (thunderx_prefetch_tune): Likewise.
14042         (thunderx2t99_prefetch_tune): Likewise.
14043         (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
14044         <default_opt_level>: Set to 3.
14045         (aarch64_override_options_internal): Update to set
14046         PARAM_PREFETCH_MINIMUM_STRIDE.
14047         * doc/invoke.texi (prefetch-minimum-stride): Document new option.
14048         * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
14049         * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
14050         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
14051         stride is constant and is below the minimum stride threshold.
14053 2018-05-23  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
14055         * config/arm/arm-cpus.in (mode26): Delete.
14056         (armv4): Delete mode26 reference.
14057         * config/arm/arm.c (arm_configure_build_target): Delete use of
14058         isa_bit_mode26.
14060 2018-05-23  Uros Bizjak  <ubizjak@gmail.com>
14062         * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
14063         New insn pattern.
14064         (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
14065         Rewrite expander pattern.  Emit gen_floatunssi<mode>2_i387_with_xmm
14066         for non-SSE modes.
14067         (floatunsdisf2): Rewrite expander pattern.  Hanlde TARGET_AVX512F.
14068         (floatunsdidf2): Ditto.
14070 2018-05-23  Uros Bizjak  <ubizjak@gmail.com>
14072         * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
14073         (fixuns_trunc<mode>si2_avx512f): Ditto.
14074         (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
14075         (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
14076         Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
14078 2018-05-23  Alexander Monakov  <amonakov@ispras.ru>
14080         PR rtl-optimization/79985
14081         * df-scan.c (df_insn_refs_collect): Remove special case for
14082         global registers and asm statements.
14084 2018-05-23  Alexander Monakov  <amonakov@ispras.ru>
14086         * extend.texi (Global Register Variables): Rewrite the bullet list.
14087         Note that the register is available for allocation. Note that access
14088         via inline asm must use constraints. Add note about async-signal
14089         handlers. Remove paragraph about automagic register selection.
14091 2018-05-23  Richard Biener  <rguenther@suse.de>
14093         * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
14094         of fixed offset from memset VN.
14096 2018-05-23  Bill Schmidt  <wschmidt@linux.ibm.com>
14098         * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
14099         first_interp field.
14100         (alloc_cand_and_find_basis): Initialize first_interp field.
14101         (slsr_process_mul): Modify first_interp field.
14102         (slsr_process_add): Likewise.
14103         (slsr_process_cast): Modify first_interp field for each new
14104         interpretation.
14105         (slsr_process_copy): Likewise.
14106         (dump_candidate): Dump first_interp field.
14107         (replace_mult_candidate): Process all interpretations, not just
14108         subsequent ones.
14109         (replace_rhs_if_not_dup): Likewise.
14110         (replace_one_candidate): Likewise.
14112 2018-05-23  Wilco Dijkstra  <wdijkstr@arm.com>
14114         * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
14115         Add new boolean.
14116         (aarch64_needs_frame_chain): New function.
14117         (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
14119 2018-05-23  Sudakshina Das  <sudi.das@arm.com>
14121         PR target/84882
14122         * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
14123         Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
14124         * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
14125         * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
14126         as true for strict-align.
14127         (aarch64_can_inline_p): Perform checks even when callee has no
14128         attributes to check for strict alignment.
14129         * doc/extend.texi (AArch64 Function Attributes): Document
14130         no-strict-align.
14131         * doc/invoke.texi: (AArch64 Options): Likewise.
14133 2018-05-23  Richard Sandiford  <richard.sandiford@linaro.org>
14135         PR tree-optimization/85853
14136         * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
14137         the handling of the root of the node to...
14138         (vect_slp_analyze_node_operations_1): ...this new function,
14139         and run the whole thing with the child nodes' def types
14140         set according to their SLP node's def type.
14142 2018-05-23  Richard Biener  <rguenther@suse.de>
14144         PR middle-end/85874
14145         * tree-data-ref.c (create_runtime_alias_checks): Defer
14146         and ignore overflow warnings.
14148 2018-05-23  Yury Gribov  <tetra2005@gmail.com>
14150         PR tree-optimization/85822
14151         * tree-vrp.c (is_masked_range_test): Fix handling of negative
14152         constants.
14154 2018-05-23  Richard Biener  <rguenther@suse.de>
14156         * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
14157         memset constants via native_interpret_expr.
14159 2018-05-22  H.J. Lu  <hongjiu.lu@intel.com>
14161         PR target/85345
14162         * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
14163         attribute.
14164         (cgraph_node::create_alias): Likewise.
14165         (cgraph_node::get_availability): Check ifunc_resolver instead
14166         of looking up ifunc attribute.
14167         * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
14168         * varasm.c (do_assemble_alias): Likewise.
14169         (assemble_alias): Likewise.
14170         (default_binds_local_p_3): Likewise.
14171         * cgraph.h (cgraph_node): Add ifunc_resolver.
14172         (cgraph_node::only_called_directly_or_aliased_p): Return false
14173         for IFUNC resolver.
14174         * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
14175         attribute.
14176         * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
14177         is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
14178         (symtab_node::binds_to_current_def_p): Check ifunc_resolver
14179         instead of looking up ifunc attribute.
14181 2018-05-22  Luis Machado  <luis.machado@linaro.org>
14183         * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
14185 2018-05-22  Martin Sebor  <msebor@redhat.com>
14187         PR middle-end/85359
14188         * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
14189         only when expasion succeeds.
14190         (expand_builtin_strcmp): Same.
14191         (expand_builtin_strncmp): Same.
14193 2018-05-22  Martin Sebor  <msebor@redhat.com>
14195         * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
14197 2018-05-22  Jackson Woodruff  <jackson.woodruff@arm.com>
14198             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
14200         * config/aarch64/aarch64-ldpstp.md: Replace uses of
14201         aarch64_mem_pair_operand with memory_operand and delete operand swapping
14202         code.
14203         * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
14204         Add check for legitimate_address.
14205         (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
14206         (aarch64_swap_ldrstr_operands): New.
14207         * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
14208         Define prototype.
14210 2018-05-22  Jackson Woodruff  <jackson.woodruff@arm.com>
14211             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
14213         * config/aarch64/aarch64.md: New patterns to generate stp
14214         and ldp.
14215         (store_pair_sw, store_pair_dw): New patterns to generate stp for
14216         single words and double words.
14217         (load_pair_sw, load_pair_dw): Likewise.
14218         (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
14219         Delete.
14220         (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
14221         Delete.
14222         * config/aarch64/aarch64-ldpstp.md: Modify peephole
14223         for different mode ldpstp and add peephole for merged zero stores.
14224         Likewise for loads.
14225         * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
14226         Add size check.
14227         (aarch64_gen_store_pair): Rename calls to match new patterns.
14228         (aarch64_gen_load_pair): Rename calls to match new patterns.
14229         * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
14230         (load_pair<DREG:mode><DREG2:mode>): ... This.
14231         (store_pair<mode>): Rename to...
14232         (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
14233         * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
14234         New mode iterators.
14235         (V_INT_EQUIV): Handle SImode.
14236         * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
14237         New predicate.
14239 2018-05-22  Martin Sebor  <msebor@redhat.com>
14241         PR c/85623
14242         * calls.c (maybe_warn_nonstring_arg): Use string length to set
14243         or ajust the presumed bound on an operation to avoid unnecessary
14244         warnings.
14246 2018-05-22  Martin Sebor  <msebor@redhat.com>
14248         PR tree-optimization/85826
14249         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
14250         assuming that a DECL necesarily has a constant size.
14252 2018-05-22  Richard Sandiford  <richard.sandiford@linaro.org>
14254         PR middle-end/85862
14255         * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
14257 2018-05-22  Richard Biener  <rguenther@suse.de>
14259         PR tree-optimization/85834
14260         * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
14261         non-constant and non-zero memset arguments.
14263 2018-05-22  Martin Liska  <mliska@suse.cz>
14265         PR ipa/85607
14266         * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
14268 2018-05-22  Richard Biener  <rguenther@suse.de>
14270         PR tree-optimization/85863
14271         * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
14272         comparisons when vectype is specified.
14273         (vectorizable_condition): Do not specify vectype for
14274         vect_is_simple_cond when SLP vectorizing.
14276 2018-05-21  Michael Meissner  <meissner@linux.ibm.com>
14278         PR target/85657
14279         * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
14280         define __ibm128 as long double.
14281         * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
14282         as a distinct type when IEEE 128-bit support is enabled.
14283         (init_float128_ieee): Fix up conversions between IFmode and IEEE
14284         128-bit types to use the correct functions.
14285         (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
14286         convert between 128-bit floating point types that have different
14287         modes but the same representation, instead of using gen_lowpart to
14288         makean alias.
14289         * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
14290         KFmode.
14291         (IFKF_reg): New attributes to give the register constraints for
14292         IFmode and KFmode.
14293         (extend<mode>tf2_internal): New insns to mark an explicit
14294         conversion between 128-bit floating point types that have a
14295         different mode but share the same representation.
14297 2018-05-21  Richard Sandiford  <richard.sandiford@linaro.org>
14299         PR tree-optimization/85814
14300         * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
14301         a null return from get_strinfo when unsharing the next
14302         strinfo in the chain.
14304 2018-05-21  Vladimir Mezentsev  <vladimir.mezentsev@oracle.com>
14306         PR gcc/84923
14307         * varasm.c (weak_finish): Clean up weak_decls.
14309 2018-05-21  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
14311         * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
14312         UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
14313         UNSPEC_UADALP values.
14314         * config/aarch64/iterators.md (ABAL): New int iterator.
14315         (ABDL2): Likewise.
14316         (ADALP): Likewise.
14317         (sur): Add mappings for the above.
14318         * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
14319         New define_insn.
14320         (aarch64_<sur>abal<mode>_4): Likewise.
14321         (aarch64_<sur>adalp<mode>_3): Likewise.
14322         (<sur>sadv16qi): New define_expand.
14324 2018-05-21  Alexander Nesterovskiy  <alexander.nesterovskiy@intel.com>
14326         * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
14327         (*movdf_internal): Ditto.
14328         (*rcpsf2_sse): Ditto.
14329         (*rsqrtsf2_sse): Ditto.
14330         (*sqrt<mode>2_sse): Ditto.
14332 2018-05-21  Tamar Christina  <tamar.christina@arm.com>
14334         * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
14335         eor3q<mode>4.
14336         (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
14337         * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
14338         veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
14339         vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
14340         vbcaxq_s64): New.
14341         * config/aarch64/arm_neon.h: Likewise.
14342         * config/aarch64/iterators.md (VQ_I): New.
14344 2018-05-21  Alexey Brodkin <abrodkin@synopsys.com>
14346         * config.gcc: Add arc/t-multilib-linux to tmake_file for
14347         arc*-*-linux*.
14348         * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
14349         MULTILIB_DIRNAMES
14351 2018-05-20  Chung-Ju Wu  <jasonwucj@gmail.com>
14353         * config/nds32/constraints.md (S): New constraint.
14354         * config/nds32/nds32.md (call_internal): Use constraint S.
14355         (call_value_internal): Likewise.
14356         (sibcall_internal): Likewise.
14357         (sibcall_value_internal): Likewise.
14359 2018-05-20  Kito Cheng  <kito.cheng@gmail.com>
14360             Chung-Ju Wu  <jasonwucj@gmail.com>
14362         * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
14363         into consideration.
14365 2018-05-20  Kito Cheng  <kito.cheng@gmail.com>
14366             Chung-Ju Wu  <jasonwucj@gmail.com>
14368         * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
14369         (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
14370         (nds32_rtx_costs_impl): Simplify.
14371         (nds32_address_cost_impl): Simplify.
14372         (nds32_init_rtx_costs): New function.
14373         (nds32_rtx_costs_speed_prefer): Likewise.
14374         (nds32_rtx_costs_size_prefer): Likewise.
14375         (nds32_address_cost_speed_prefer): Likewise.
14376         (nds32_address_cost_speed_fwprop): Likewise.
14377         (nds32_address_cost_size_prefer): Likewise.
14378         * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
14379         * config/nds32/nds32.c (nds32_option_override): Use
14380         nds32_init_rtx_costs function.
14382 2018-05-20  Chung-Ju Wu  <jasonwucj@gmail.com>
14384         * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
14385         * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
14386         (TARGET_PIPELINE_N8): Likewise.
14387         (TARGET_PIPELINE_N10): Likewise.
14388         (TARGET_PIPELINE_N13): Likewise.
14389         (TARGET_PIPELINE_GRAYWOLF): Likewise.
14391 2018-05-19  Monk Chiang  <sh.chiang04@gmail.com>
14393         * config/nds32/nds32-fpu.md: Update copyright year.
14395 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
14397         * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
14399 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
14401         * config/nds32/nds32.c
14402         (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
14403         * config/nds32/nds32.opt (minline-asm-r15): New option.
14405 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
14407         * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
14408         MASK_HW_ABS.
14409         * config/nds32/nds32.md (abssi2): New pattern.
14411 2018-05-19  Uros Bizjak  <ubizjak@gmail.com>
14413         * config/i386/i386.md (rex64namesuffix): New mode attribute.
14414         * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
14415         Merge insn pattern from sse_cvtsi2ss<round_name> and
14416         sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
14417         (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
14418         from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
14419         using SWI48 mode iterator.
14420         (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
14421         sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
14422         (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
14423         pattern from sse_cvttss2si<round_saeonly_name>
14424         and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
14425         (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
14426         from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
14427         using SWI48 mode iterator.
14428         (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
14429         insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
14430         avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
14431         (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
14432         from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
14433         using SWI48 mode iterator.
14434         (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
14435         insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
14436         avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
14437         (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
14438         sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
14439         SWI48 mode iterator.
14440         (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
14441         sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
14442         (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
14443         pattern from sse_cvttsd2si<round_saeonly_name>
14444         and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
14446 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
14448         * config/nds32/nds32-md-auxiliary.c
14449         (nds32_valid_smw_lwm_base_p): Refine.
14450         (nds32_output_smw_single_word): Refine.
14451         (nds32_output_smw_double_word): New.
14452         * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
14454 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
14456         * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
14457         (nds32_output_stack_pop): Refine.
14458         (nds32_expand_unaligned_load): Refine.
14459         (nds32_expand_unaligned_store): Refine.
14461 2018-05-19  Kuan-Lin Chen <kuanlinchentw@gmail.com>
14462             Chung-Ju Wu  <jasonwucj@gmail.com>
14464         * config/nds32/constants.md: Add TP_REGNUM constant.
14465         (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
14466         UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
14467         UNSPEC_ADD32.
14468         * config/nds32/nds32-doubleword.md: Consider flag_pic.
14469         * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
14470         * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
14471         * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
14472         and PIC code generation.
14473         * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
14474         code generation.
14475         * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
14476         optimization.
14477         * config/nds32/nds32.md: Support TLS and PIC.
14478         * config/nds32/nds32.c: Support TLS and PIC.
14479         * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
14480         * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
14481         predicate.
14483 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
14485         * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
14486         mode with E_ prefix.
14488 2018-05-19  Kuan-Lin Chen <kuanlinchentw@gmail.com>
14489             Chung-Ju Wu  <jasonwucj@gmail.com>
14491         * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
14492         * config/nds32/nds32-md-auxiliary.c
14493         (symbolic_reference_mentioned_p): New.
14494         (nds32_legitimize_ict_address): New.
14495         (nds32_expand_ict_move): New.
14496         (nds32_indirect_call_referenced_p): New.
14497         (nds32_symbol_binds_local_p): Delete.
14498         (nds32_long_call_p): Modify.
14499         * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
14500         * config/nds32/nds32-protos.h
14501         (symbolic_reference_mentioned_p): Declare.
14502         (nds32_legitimize_ict_address): Declare.
14503         (nds32_expand_ict_move): Declare.
14504         (nds32_indirect_call_referenced_p): Declare.
14505         * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
14506         (nds32_relax_group): Use nds32_ict_const_p as condition.
14507         * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
14508         (nds32_asm_file_start): Output ict_model directive in asm code.
14509         (nds32_legitimate_address_p): Consider indirect call.
14510         (nds32_print_operand): Consider indirect call.
14511         (nds32_print_operand_address): Consider indirect call.
14512         (nds32_insert_attributes): Handle "indirect_call" attribute.
14513         (TARGET_LEGITIMATE_ADDRESS_P): Define.
14514         (TARGET_LEGITIMATE_CONSTANT_P): Define.
14515         (TARGET_CANNOT_FORCE_CONST_MEM): Define.
14516         (TARGET_DELEGITIMIZE_ADDRESS): Define.
14517         (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
14518         * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
14519         (TARGET_ICT_MODEL_SMALL): Define.
14520         (TARGET_ICT_MODEL_LARGE): Define.
14521         * config/nds32/nds32.md (movsi): Consider ict model.
14522         (call, call_value): Consider ict model.
14523         (sibcall, sibcall_value): Consider ict model.
14524         * config/nds32/nds32.opt (mict-model): New option.
14525         * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
14526         model.
14528 2018-05-18  Kito Cheng  <kito.cheng@gmail.com>
14529             Monk Chiang  <sh.chiang04@gmail.com>
14530             Jim Wilson <jimw@sifive.com>
14532         * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
14533         Add support to parse rv32e*.  Clear MASK_RVE for rv32i and rv64i.
14534         * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
14535         * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
14536         __riscv_32e when TARGET_RVE.  Handle ABI_ILP32E as soft-float ABI.
14537         * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
14538         * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
14539         compute save_libcall_adjustment properly.
14540         (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
14541         (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
14542         * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
14543         (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
14544         (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
14545         (ABI_SPEC): Handle mabi=ilp32e.
14546         * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
14547         (RVE): Add RVE mask.
14548         * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
14549         <-march>: Add rv32e as an example.
14551 2018-05-18  Marc Glisse  <marc.glisse@inria.fr>
14553         PR c++/82899
14554         * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
14555         (intra_create_variable_infos): Handle C++ constructors.
14557 2018-05-18  Martin Liska  <mliska@suse.cz>
14559         * passes.def: Remove a redundant pass.
14561 2018-05-18  Eric Botcazou  <ebotcazou@adacore.com>
14563         PR bootstrap/85838
14564         * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
14566 2018-05-18  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
14568         * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
14569         (ARMv4): Update.
14570         (ARMv2, ARMv3, ARMv3m): Delete fgroups.
14571         (ARMv6m): Update.
14572         (armv2, armv2a, armv3, armv3m): Delete architectures.
14573         (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
14574         arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
14575         arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
14576         Delete cpus.
14577         * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
14578         (*mulsidi3adddi): Likewise.
14579         (mulsidi3): Likewise.
14580         (*mulsidi3_nov6): Likewise.
14581         (umulsidi3): Likewise.
14582         (umulsidi3_nov6): Likewise.
14583         (umaddsidi4): Likewise.
14584         (*umulsidi3adddi): Likewise.
14585         (smulsi3_highpart): Likewise.
14586         (*smulsi3_highpart_nov6): Likewise.
14587         (umulsi3_highpart): Likewise.
14588         (*umulsi3_highpart_nov6): Likewise.
14589         * config/arm/arm.h (arm_arch3m): Delete.
14590         * config/arm/arm.c (arm_arch3m): Delete.
14591         (arm_option_override_internal): Update armv3-related comment.
14592         (arm_configure_build_target): Delete use of isa_bit_mode32.
14593         (arm_option_reconfigure_globals): Delete set of arm_ach3m.
14594         (arm_rtx_costs_internal): Delete check of arm_arch3m.
14595         * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
14596         (mulsa3): Likewise.
14597         (mulusa3): Likewise.
14598         * config/arm/arm-protos.h (arm_arch3m): Delete.
14599         * config/arm/arm-tables.opt: Regenerate.
14600         * config/arm/arm-tune.md: Likewise.
14601         * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
14602         deleted architectures.
14604 2018-05-18  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
14606         * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
14607         (armv5t, armv5te): New features.
14608         (ARMv5, ARMv5e): Delete fgroups.
14609         (ARMv5t, ARMv5te): Adjust for above changes.
14610         (ARMv6m): Likewise.
14611         (armv5, armv5e): Delete arches.
14612         * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
14613         arm_arch5.
14614         (*call_reg_arm): Likewise.
14615         (*call_value_reg_armv5): Likewise.
14616         (*call_value_reg_arm): Likewise.
14617         (*call_symbol): Likewise.
14618         (*call_value_symbol): Likewise.
14619         (*sibcall_insn): Likewise.
14620         (*sibcall_value_insn): Likewise.
14621         (clzsi2): Likewise.
14622         (prefetch): Likewise.
14623         (define_split and define_peephole2 dependent on arm_arch5):
14624         Likewise.
14625         * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
14626         arm_arch5e.
14627         (TARGET_ARM_QBIT): Likewise.
14628         (TARGET_DSP_MULTIPLY): Likewise.
14629         (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
14630         (arm_arch5, arm_arch5e): Delete.
14631         (arm_arch5t, arm_arch5te): Declare.
14632         * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
14633         (arm_arch5t): Declare.
14634         (arm_option_reconfigure_globals): Update for the above.
14635         (arm_options_perform_arch_sanity_checks): Update comment, replace
14636         use of arm_arch5 with arm_arch5t.
14637         (use_return_insn): Likewise.
14638         (arm_emit_call_insn): Likewise.
14639         (output_return_instruction): Likewise.
14640         (arm_final_prescan_insn): Likewise.
14641         (arm_coproc_builtin_available): Likewise.
14642         * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
14643         arm_arch5e with arm_arch5t and arm_arch5te.
14644         * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
14645         (arm_arch5t, arm_arch5te): Declare.
14646         * config/arm/arm-tables.opt: Regenerate.
14647         * config/arm/t-arm-elf: Remove references to armv5, armv5e.
14648         * config/arm/t-multilib: Likewise.
14649         * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
14650         instead of arm_arch5.
14651         (*call_reg_thumb1): Likewise.
14652         (*call_value_reg_thumb1_v5): Likewise.
14653         (*call_value_reg_thumb1): Likewise.
14654         * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
14655         unreachable path.
14656         * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
14658 2018-05-18  Martin Liska  <mliska@suse.cz>
14660         PR gcov-profile/84846
14661         * doc/gcov.texi: Document -t option of gcov tool.
14663 2018-05-18  Martin Liska  <mliska@suse.cz>
14665         PR gcov-profile/84846
14666         * gcov.c (print_usage): Add new -t option.
14667         (process_args): Handle the option.
14668         (generate_results): Use stdout as output when requested by
14669         the option.
14671 2018-05-18  Martin Liska  <mliska@suse.cz>
14673         PR gcov-profile/84846
14674         * coverage.c (coverage_init): Write PWD to .gcno file.
14675         * doc/gcov.texi: Document how working directory is printed.
14676         * gcov-dump.c (dump_gcov_file): Print PWD.
14677         * gcov.c (output_intermediate_file): Likewise.
14678         (read_graph_file): Read PWD string.
14679         (output_lines): Print PWD.
14681 2018-05-18  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
14683         PR middle-end/85817
14684         * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
14685         for retval and return false if all args to phi are zero.
14687 2018-05-18  Richard Biener  <rguenther@suse.de>
14689         * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
14690         method.
14691         (evrp_dom_walker::before_dom_children): Call it.
14693 2018-05-18  Richard Biener  <rguenther@suse.de>
14695         * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
14696         results when processing array refs with variable index.
14698 2018-05-18  Toon Moene  <toon@moene.org>
14700         * doc/invoke.texi: Move -floop-unroll-and-jam documentation
14701         directly after that of -floop-interchange. Indicate that both
14702         options are enabled by default when specifying -O3.
14704 2018-05-18  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
14706         * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
14707         iterator.  Delete separate integer-mode vec_set<mode> expander.
14708         (aarch64_simd_vec_setv2di): Delete.
14709         (vec_setv2di): Delete.
14710         (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
14711         Use VALL_F16 mode iterator.  Add LD1 alternative and use vwcore for
14712         the "w, r" alternative.
14714 2018-05-18  Martin Liska  <mliska@suse.cz>
14716         * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
14717         * tree-pass.h (make_pass_lower_switch_O0): New function.
14718         * tree-switch-conversion.c (node_has_low_bound): Remove.
14719         (node_has_high_bound): Likewise.
14720         (node_is_bounded): Likewise.
14721         (class pass_lower_switch): Make it a template type and create
14722         two instances.
14723         (pass_lower_switch::execute): Add template argument.
14724         (make_pass_lower_switch): New function.
14725         (make_pass_lower_switch_O0): New function.
14726         (do_jump_if_equal): Remove.
14727         (emit_case_nodes): Simplify to just handle all 3 cases and leave
14728         all the hard work to tree optimization passes.
14730 2018-05-18  Martin Liska  <mliska@suse.cz>
14732         * dbgcnt.c (limit_low): Renamed from limit.
14733         (limit_high): New variable.
14734         (dbg_cnt_is_enabled): Check for upper limit.
14735         (dbg_cnt): Adjust dumping.
14736         (dbg_cnt_set_limit_by_index): Add new argument for high
14737         value.
14738         (dbg_cnt_set_limit_by_name): Likewise.
14739         (dbg_cnt_process_single_pair): Parse new format.
14740         (dbg_cnt_process_opt): Use strtok.
14741         (dbg_cnt_list_all_counters): Remove 'value' and add
14742         'limit_high'.
14743         * doc/invoke.texi: Document changes.
14745 2018-05-18  Richard Sandiford  <richard.sandiford@linaro.org>
14747         * doc/sourcebuild.texi (scalar_all_fma): Document.
14748         * tree.def (FMA_EXPR): Delete.
14749         * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
14750         * internal-fn.c (ternary_direct): New macro.
14751         (expand_ternary_optab_fn): Likewise.
14752         (direct_ternary_optab_supported_p): Likewise.
14753         * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
14754         * builtins.c (fold_builtin_fma): Delete.
14755         (fold_builtin_3): Don't call it.
14756         * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
14757         * expr.c (expand_expr_real_2): Likewise.
14758         * fold-const.c (operand_equal_p): Likewise.
14759         (fold_ternary_loc): Likewise.
14760         * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
14761         * gimple.c (DEFTREECODE): Likewise.
14762         * gimplify.c (gimplify_expr): Likewise.
14763         * optabs-tree.c (optab_for_tree_code): Likewise.
14764         * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
14765         * tree-eh.c (operation_could_trap_p): Likewise.
14766         (stmt_could_throw_1_p): Likewise.
14767         * tree-inline.c (estimate_operator_cost): Likewise.
14768         * tree-pretty-print.c (dump_generic_node): Likewise.
14769         (op_code_prio): Likewise.
14770         * tree-ssa-loop-im.c (stmt_cost): Likewise.
14771         * tree-ssa-operands.c (get_expr_operands): Likewise.
14772         * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
14773         * fold-const-call.h (fold_fma): Delete.
14774         * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
14775         CFN_FNMA and CFN_FNMS.
14776         (fold_fma): Delete.
14777         * genmatch.c (combined_fn): New enum.
14778         (commutative_ternary_tree_code): Remove FMA_EXPR handling.
14779         (commutative_op): New function.
14780         (commutate): Use it.  Handle more than 2 operands.
14781         (dt_operand::gen_gimple_expr): Use commutative_op.
14782         (parser::parse_expr): Allow :c to be used with non-binary
14783         operators if the commutative operand is known.
14784         * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
14785         CFN_FMS, CFN_FNMA and CFN_FNMS.
14786         (backprop::process_assign_use): Remove FMA_EXPR handling.
14787         * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
14788         (gen_hsa_fma): New function.
14789         (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
14790         IFN_FNMA and IFN_FNMS.
14791         * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
14792         * gimple-fold.h (follow_all_ssa_edges): Declare.
14793         * gimple-fold.c (follow_all_ssa_edges): New function.
14794         * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
14795         gimple_build interface and use follow_all_ssa_edges to fold the result.
14796         (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
14797         instead of checking for optabs directly.
14798         * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
14799         rather than FMA_EXPRs.
14800         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
14801         call to IFN_FMA instead of an FMA_EXPR.
14803 2018-05-17  Jim Wilson  <jimw@sifive.com>
14805         * expr.c (do_tablejump): When converting index to Pmode, if we have a
14806         sign extended promoted subreg, and the range does not have the sign bit
14807         set, then do a sign extend.
14809         * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
14810         test, check for sign extended subreg and/or constant operands, and
14811         do a sign extend in that case.
14813 2018-05-17  Steve Ellcey  <sellcey@cavium.com>
14815         * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
14816         (thunderx2t99_multiple): Delete psuedo-units from used cpus.
14817         Add untyped.
14818         (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
14819         Change logics_shift_reg to logics_shift_imm.
14820         (thunderx2t99_fp_loadpair_basic): Delete.
14821         (thunderx2t99_fp_storepair_basic): Delete.
14822         (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
14823         (thunderx2t99_asimd_polynomial): Delete.
14824         (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
14825         and neon_fp_mul_d_scalar_q.
14826         (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
14827         (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
14828         (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
14829         (thunderx2t99_asimd_lut): Add missing tbl types.
14830         (thunderx2t99_asimd_ext): Delete.
14831         (thunderx2t99_asimd_load1_1_mult): Delete.
14832         (thunderx2t99_asimd_load1_2_mult): Delete.
14833         (thunderx2t99_asimd_load1_ldp): New.
14834         (thunderx2t99_asimd_load1): New.
14835         (thunderx2t99_asimd_load2): Add missing *load2* types.
14836         (thunderx2t99_asimd_load3): New.
14837         (thunderx2t99_asimd_load4): New.
14838         (thunderx2t99_asimd_store1_1_mult): Delete.
14839         (thunderx2t99_asimd_store1_2_mult): Delete.
14840         (thunderx2t99_asimd_store2_mult): Delete.
14841         (thunderx2t99_asimd_store2_onelane): Delete.
14842         (thunderx2t99_asimd_store_stp): New.
14843         (thunderx2t99_asimd_store1): New.
14844         (thunderx2t99_asimd_store2): New.
14845         (thunderx2t99_asimd_store3): New.
14846         (thunderx2t99_asimd_store4): New.
14848 2018-05-17  Jerome Lambourg  <lambourg@adacore.com>
14850         * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
14851         #include <stdint.h>.  Replace intptr_t with __INTPTR_TYPE__.
14853 2018-05-17  Pat Haugen  <pthaugen@us.ibm.com>
14854             Segher Boessenkool  <segher@kernel.crashing.org>
14856         PR target/85698
14857         * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
14858         operand.
14860 2018-05-17  Richard Biener  <rguenther@suse.de>
14862         * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
14863         for pruning loop and prune defs feeding only already visited PHIs.
14865 2018-05-17  Richard Biener  <rguenther@suse.de>
14867         * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
14869 2018-05-17  Bin Cheng  <bin.cheng@arm.com>
14870             Richard Biener  <rguenther@suse.de>
14872         PR tree-optimization/85793
14873         * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
14874         for VMAT_ELEMENTWISE.
14876 2018-05-17  Richard Sandiford  <richard.sandiford@linaro.org>
14878         * internal-fn.h (lookup_internal_fn): Declare
14879         * internal-fn.c (lookup_internal_fn): New function.
14880         * gimple.c (gimple_build_call_from_tree): Handle calls to
14881         internal functions.
14882         * gimple-pretty-print.c (dump_gimple_call): Print "." before
14883         internal function names.
14884         * tree-pretty-print.c (dump_generic_node): Likewise.
14885         * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
14887 2018-05-17  Richard Sandiford  <richard.sandiford@linaro.org>
14889         * gimple-fold.h (gimple_build): Make the function forms take
14890         combined_fn rather than built_in_function.
14891         (gimple_simplify): Likewise.
14892         * gimple-match-head.c (gimple_simplify): Likewise.
14893         * gimple-fold.c (gimple_build): Likewise.
14894         * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
14895         rather than gimple_build_call_internal.
14896         (get_initial_defs_for_reduction): Likewise.
14897         (vect_create_epilog_for_reduction): Likewise.
14898         (vectorizable_live_operation): Likewise.
14900 2018-05-17  Martin Liska  <mliska@suse.cz>
14902         * gimple-ssa-sprintf.c (format_directive): Do not use
14903         space in between 'G_' and '('.
14905 2018-05-17  Jakub Jelinek  <jakub@redhat.com>
14907         PR target/85323
14908         * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
14909         even if the mask is not all ones.
14911         PR target/85323
14912         * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
14913         vector.
14914         (ix86_gimple_fold_builtin): Likewise.
14916         PR target/85323
14917         * config/i386/i386.c: Include tree-vector-builder.h.
14918         (ix86_vector_shift_count): New function.
14919         (ix86_fold_builtin): Fold shift builtins by scalar count.
14920         (ix86_gimple_fold_builtin): Likewise.
14922         * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
14923         _mm512_setzero): New intrinsics.
14925 2018-05-17  James Greenhalgh  <james.greenhalgh@arm.com>
14926             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
14928         * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
14929         code generation for cases where splatting a value is not useful.
14930         * simplify-rtx.c (simplify_ternary_operation): Simplify
14931         vec_merge across a vec_duplicate and a paradoxical subreg forming
14932         a vector mode to a vec_concat.
14934 2018-05-17  Olga Makhotina  <olga.makhotina@intel.com>
14936         * config.gcc: Support "goldmont-plus".
14937         * config/i386/driver-i386.c (host_detect_local_cpu): Detect
14938         "goldmont-plus".
14939         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
14940         PROCESSOR_GOLDMONT_PLUS.
14941         * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
14942         (processor_target_table): Add "goldmont-plus".
14943         (PTA_GOLDMONT_PLUS): Define.
14944         (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
14945         (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
14946         (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
14947         (fold_builtin_cpu): Add "goldmont-plus".
14948         (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
14949         (ix86_option_override_internal): Add "goldmont-plus".
14950         * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
14951         (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
14952         * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
14953         * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
14955 2018-05-17  Richard Biener  <rguenther@suse.de>
14957         PR tree-optimization/85757
14958         * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
14959         remove defs that only feed that PHI from further processing.
14961 2018-05-16  Jim Wilson  <jimw@sifive.com>
14963         * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
14964         asterisk to name.
14965         (<optab>di3_mask, <optab>di3_mask_1): Likewise.
14967 2018-05-16  Mark Wielaard  <mark@klomp.org>
14969         * dwarf2out.c (count_index_strings): New function.
14970         (output_indirect_strings): Call count_index_strings and generate
14971         header for dwarf_version >= 5.
14973 2018-05-16  Mark Wielaard  <mark@klomp.org>
14975         * dwarf2out.c (dwarf_FORM): New function.
14976         (set_indirect_string): Use dwarf_FORM.
14977         (reset_indirect_string): Likewise.
14978         (size_of_die): Likewise.
14979         (value_format): Likewise.
14980         (output_die): Likewise.
14981         (add_skeleton_AT_string): Likewise.
14982         (output_macinfo_op): Likewise.
14983         (index_string): Likewise.
14984         (output_index_string_offset): Likewise.
14985         (output_index_string): Likewise.
14986         (count_index_strings): Likewise.
14988 2018-05-16  Carl Love  <cel@us.ibm.com>
14990         * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
14991         dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
14993 2018-05-16  Martin Jambor  <mjambor@suse.cz>
14995         * ipa-prop.c (ipa_free_all_edge_args): Remove.
14996         * ipa-prop.h (ipa_free_all_edge_args): Likewise.
14998 2018-05-16  Wilco Dijkstra  <wdijkstr@arm.com>
15000         * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
15001         (fnma<mode>4): Likewise.
15002         (fms<mode>4): Likewise.
15003         (fnms<mode>4): Likewise.
15004         (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
15005         (aarch64_fnma<mode>4): Likewise.
15006         (aarch64_fms<mode>4): Likewise.
15007         (aarch64_fnms<mode>4): Likewise.
15008         (aarch64_fnmadd<mode>4): Likewise.
15010 2018-05-16  Jason Merrill  <jason@redhat.com>
15012         * tree.c (warn_deprecated_use): Return bool.  Simplify logic.
15014 2018-05-16  Richard Biener  <rguenther@suse.de>
15016         * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
15017         (dump_stmt_cost): Declare.
15018         (add_stmt_cost): Dump cost we add.
15019         (add_stmt_costs): New function.
15020         (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
15021         No longer exported.
15022         (vect_analyze_stmt): Adjust prototype.
15023         (vectorizable_condition): Likewise.
15024         (vectorizable_live_operation): Likewise.
15025         (vectorizable_reduction): Likewise.
15026         (vectorizable_induction): Likewise.
15027         * tree-vect-loop.c (vect_analyze_loop_operations): Create local
15028         cost vector to pass to vectorizable_ and record afterwards.
15029         (vect_model_reduction_cost): Take cost vector argument and adjust.
15030         (vect_model_induction_cost): Likewise.
15031         (vectorizable_reduction): Likewise.
15032         (vectorizable_induction): Likewise.
15033         (vectorizable_live_operation): Likewise.
15034         * tree-vect-slp.c (vect_create_new_slp_node): Initialize
15035         SLP_TREE_NUMBER_OF_VEC_STMTS.
15036         (vect_analyze_slp_cost_1): Remove.
15037         (vect_analyze_slp_cost): Likewise.
15038         (vect_slp_analyze_node_operations): Take visited args and
15039         a target cost vector.  Avoid processing already visited stmt sets.
15040         (vect_slp_analyze_operations): Use a local cost vector to gather
15041         costs and register those of non-discarded instances.
15042         (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
15043         (vect_schedule_slp_instance): Remove copying of
15044         SLP_TREE_NUMBER_OF_VEC_STMTS.  Instead assert that it is not
15045         zero.
15046         * tree-vect-stmts.c (record_stmt_cost): Remove path directly
15047         adding cost.  Record cost entry location.
15048         (vect_prologue_cost_for_slp_op): Function to compute cost of
15049         a constant or invariant generated for SLP vect in the prologue,
15050         split out from vect_analyze_slp_cost_1.
15051         (vect_model_simple_cost): Make static.  Adjust for SLP costing.
15052         (vect_model_promotion_demotion_cost): Likewise.
15053         (vect_model_store_cost): Likewise, make static.
15054         (vect_model_load_cost): Likewise.
15055         (vectorizable_bswap): Add cost vector arg and adjust.
15056         (vectorizable_call): Likewise.
15057         (vectorizable_simd_clone_call): Likewise.
15058         (vectorizable_conversion): Likewise.
15059         (vectorizable_assignment): Likewise.
15060         (vectorizable_shift): Likewise.
15061         (vectorizable_operation): Likewise.
15062         (vectorizable_store): Likewise.
15063         (vectorizable_load): Likewise.
15064         (vectorizable_condition): Likewise.
15065         (vectorizable_comparison): Likewise.
15066         (can_vectorize_live_stmts): Likewise.
15067         (vect_analyze_stmt): Likewise.
15068         (vect_transform_stmt): Adjust calls to vectorizable_*.
15069         * tree-vectorizer.c: Include gimple-pretty-print.h.
15070         (dump_stmt_cost): New function.
15072 2018-05-16  Richard Biener  <rguenther@suse.de>
15074         * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
15075         * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
15076         * tree-ssa-dse.c: Include tree-ssa-loop.h.
15077         (check_name): New callback.
15078         (dse_classify_store): Track cycles via a visited bitmap of PHI
15079         defs and simplify handling of in-loop and across loop dead stores
15080         and properly fail for loop-variant refs.  Handle byte-tracking with
15081         multiple defs.  Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
15082         limiting the walk.
15084 2018-05-16  Richard Sandiford  <richard.sandiford@linaro.org>
15086         * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
15087         (vect_get_mask_type_for_stmt): Likewise.
15088         * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
15089         split out from...
15090         (vect_build_slp_tree_1): ...here.  Use vect_get_vector_types_for_stmt
15091         to determine the statement's vector type and the vector type that
15092         should be used for calculating nunits.  Deal with cases in which
15093         the type has to be deferred.
15094         (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
15095         and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
15096         * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
15097         (vect_determine_vf_for_stmt): New functions, split out from...
15098         (vect_determine_vectorization_factor): ...here.
15099         * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
15100         (vect_get_mask_type_for_stmt): New functions, split out from
15101         vect_determine_vectorization_factor.
15103 2018-05-16  Richard Biener  <rguenther@suse.de>
15105         * tree-cfg.c (verify_gimple_assign_ternary): Properly
15106         verify the [VEC_]COND_EXPR embedded comparison.
15108 2018-05-15  Martin Sebor  <msebor@redhat.com>
15110         PR tree-optimization/85753
15111         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
15112         RECORD_TYPE in addition to ARRAY_TYPE.
15114 2018-05-15  Martin Sebor  <msebor@redhat.com>
15116         PR middle-end/85643
15117         * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
15119 2018-05-15  Richard Biener  <rguenther@suse.de>
15121         * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
15122         add by_clobber_p one.  Change algorithm to collect all defs
15123         representing uses we need to walk and try reducing them to
15124         a single one before failing.
15125         (dse_dom_walker::dse_optimize_stmt): Adjust.
15127 2018-05-13  Mark Wielaard  <mark@klomp.org>
15129         * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
15130         (size_of_loc_descr): Likewise.
15131         (output_loc_operands): Likewise.
15132         (output_loc_operands_raw): Likewise.
15133         (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
15134         (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
15135         (hash_loc_operands): Likewise.
15136         (compare_loc_operands): Likewise.
15138 2018-05-14  Mark Wielaard  <mark@klomp.org>
15140         * dwarf2out.c (count_index_addrs): New function.
15141         (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
15143 2018-05-15  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
15145         PR tree-optimization/83648
15146         * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
15147         return value as malloc candidate.
15149 2018-05-15  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
15151         PR ipa/85734
15152         * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite
15153         param as true in call to suggest_attribute.
15155 2018-05-14  Segher Boessenkool  <segher@kernel.crashing.org>
15157         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
15158         -mreadonly-in-sdata.
15160 2018-05-14  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
15162         * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
15163         New pattern.
15164         (aarch64_crypto_aesd_fused): Likewise.
15166 2018-05-14  Wilco Dijkstra  <wdijkstr@arm.com>
15168         * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
15169         (movsi_aarch64): Likewise.
15170         (load_pairsi): Likewise.
15171         (load_pairdi): Likewise.
15172         (store_pairsi): Likewise.
15173         (store_pairdi): Likewise.
15174         (load_pairsf): Likewise.
15175         (load_pairdf): Likewise.
15176         (store_pairsf): Likewise.
15177         (store_pairdf): Likewise.
15178         (zero_extend): Likewise.
15179         (trunc): Swap alternatives.
15180         (fcvt_target): Add '?' to prefer w over r.
15182 2018-05-14  Jakub Jelinek  <jakub@redhat.com>
15184         PR target/85756
15185         * config/i386/i386.md: Disallow non-commutative arithmetics in
15186         last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
15187         optimization.  Use COMMUTATIVE_ARITH_P test rather than != MINUS
15188         in the peephole2 before it.
15190 2018-05-14  Sebastian Peryt  <sebastian.peryt@intel.com>
15192         * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
15193         OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
15194         (ix86_handle_option): Handle -mcldemote.
15195         * config.gcc: New header.
15196         * config/i386/cldemoteintrin.h: New file.
15197         * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
15198         * config/i386/driver-i386.c (host_detect_local_cpu): Detect
15199         -mcldemote.
15200         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
15201         OPTION_MASK_ISA_CLDEMOTE.
15202         * config/i386/i386.c (ix86_target_string): Add -mcldemote.
15203         (ix86_valid_target_attribute_inner_p): Ditto.
15204         (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
15205         (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
15206         (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
15207         * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
15208         * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
15209         (cldemote): New.
15210         * config/i386/i386.opt: Add -mcldemote.
15211         * config/i386/x86intrin.h: New header.
15212         * doc/invoke.texi: Add -mcldemote.
15214 2018-05-14  Richard Biener  <rguenther@suse.de>
15216         * doc/match-and-simplify.texi: Adjust :s documentation.
15218 2018-05-14  Alexander Monakov  <amonakov@ispras.ru>
15220         * sort.cc (REORDER_23): Pass the type for the temporaries instead of
15221         intended memcpy size.
15222         (REORDER_45): Likewise.
15224 2018-05-13  Alexander Monakov  <amonakov@ispras.ru>
15226         * sort.cc: New file.
15227         * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
15228         * vec.c (qsort_chk): Use gcc_qsort.
15229         * Makefile.in (OBJS-libcommon): Add sort.o.
15230         (build/sort.o): New target.  Use it...
15231         (BUILD_RTL): ... here, and...
15232         (build/gencfn-macros): ... here, and...
15233         (build/genmatch): ... here.
15235 2018-05-13  Kito Cheng  <kito.cheng@gmail.com>
15236             Chung-Ju Wu  <jasonwucj@gmail.com>
15238         * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
15239         * config/nds32/nds32-graywolf.md: New file.
15240         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
15241         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
15242         pipeline.
15243         * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
15244         * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
15245         * config/nds32/nds32.md (pipeline_model): Add graywolf.
15246         * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
15247         * config/nds32/pipelines.md: Include n15 settings.
15249 2018-05-13  Kito Cheng  <kito.cheng@gmail.com>
15250             Chung-Ju Wu  <jasonwucj@gmail.com>
15252         * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
15253         * config/nds32/nds32-n13.md: New file.
15254         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
15255         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
15256         pipeline.
15257         * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
15258         * config/nds32/nds32.md (pipeline_model): Add n13.
15259         * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
15260         * config/nds32/pipelines.md: Include n13 settings.
15262 2018-05-13  Kito Cheng  <kito.cheng@gmail.com>
15263             Chung-Ju Wu  <jasonwucj@gmail.com>
15265         * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
15266         * config/nds32/nds32-n10.md: New file.
15267         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
15268         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
15269         pipeline.
15270         * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
15271         * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
15272         * config/nds32/nds32.md (pipeline_model): Add n10.
15273         * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
15274         * config/nds32/pipelines.md: Include n10 settings.
15276 2018-05-13  Monk Chiang  <sh.chiang04@gmail.com>
15277             Kito Cheng  <kito.cheng@gmail.com>
15278             Chung-Ju Wu  <jasonwucj@gmail.com>
15280         * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
15281         * config/nds32/constants.md (unspec_element, unspec_volatile_element):
15282         Add enum values for DSP extension instructions.
15283         * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
15284         New constraints.
15285         * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
15286         sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
15287         New code iterators.
15288         (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
15289         * config/nds32/nds32-dspext.md: New file for DSP implementation.
15290         * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
15291         * config/nds32/nds32-intrinsic.md: Likewise.
15292         * config/nds32/nds32_intrinsic.h: Likewise.
15293         * config/nds32/nds32-md-auxiliary.c: Likewise.
15294         * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
15295         * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
15296         (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
15297         (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
15298         * config/nds32/nds32-protos.h: New declarations for DSP extension.
15299         * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
15300         TYPE_DMAC in switch statement.
15301         * config/nds32/nds32.c: New checking and implementation for DSP
15302         extension instructions.
15303         * config/nds32/nds32.h: Likewise.
15304         * config/nds32/nds32.md: Likewise.
15305         * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
15306         * config/nds32/predicates.md: Implement new predicates for DSP
15307         extension.
15309 2018-05-11  Michael Meissner  <meissner@linux.vnet.ibm.com>
15311         * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
15312         Reformat alternatives and attributes so it is easier to identify
15313         which constraints/attributes go with which instruction.
15314         (mov<mode>_hardfloat32, FMOVE64): Likewise.
15315         (mov<mode>_softfloat32, FMOVE64): Likewise.
15316         (mov<mode>_hardfloat64, FMOVE64): Likewise.
15317         (mov<mode>_softfloat64, FMOVE64): Likewise.
15319 2018-05-11  Kelvin Nilsen  <kelvin@gcc.gnu.org>
15321         * doc/extend.texi (PowerPC Built-in Functions): Rename this
15322         subsection.
15323         (Basic PowerPC Built-in Functions): The new name of the
15324         subsection previously known as "PowerPC Built-in Functions".
15325         (Basic PowerPC Built-in Functions Available on all Configurations):
15326         New subsubsection.
15327         (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
15328         (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
15329         (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
15330         (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
15332 2018-05-11  Martin Jambor  <mjambor@suse.cz>
15334         PR ipa/85655
15335         * ipa-cp.c (intersect_with_plats): Check that the lattice contains
15336         single const.
15338 2018-05-11  Richard Earnshaw  <rearnsha@arm.com>
15340         PR target/85733
15341         * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
15343 2018-05-11  Sebastian Peryt  <sebastian.peryt@intel.com>
15345         * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
15346         OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
15347         (ix86_handle_option): Handle -mwaitpkg.
15348         * config.gcc: New header.
15349         * config/i386/cpuid.h (bit_WAITPKG): New bit.
15350         * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
15351         * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
15352         function type.
15353         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
15354         OPTION_MASK_ISA_WAITPKG.
15355         * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
15356         (ix86_option_override_internal): Add PTA_WAITPKG.
15357         (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
15358         (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
15359         IX86_BUILTIN_TPAUSE.
15360         (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
15361         __builtin_ia32_umwait and __builtin_ia32_tpause.
15362         (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
15363         IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
15364         * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
15365         * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
15366         UNSPECV_TPAUSE): New.
15367         (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
15368         * config/i386/i386.opt: Add -mwaitpkg.
15369         * config/i386/waitpkgintrin.h: New file.
15370         * config/i386/x86intrin.h: New header.
15371         * doc/invoke.texi: Add -mwaitpkg.
15373 2018-05-11  Richard Earnshaw  <rearnsha@arm.com>
15375         PR target/85606
15376         * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
15377         equivalent.
15378         (cortex-m0): Use armv6s-m isa.
15379         (cortex-m0plus): Likewise.
15380         (cortex-m1): Likewise.
15381         (cortex-m0.small-multiply): Likewise.
15382         (cortex-m0plus.small-multiply): Likewise.
15383         (cortex-m1.small-multiply): Likewise.
15385 2018-05-11  Allan Sandfeld Jensen  <allan.jensen@qt.io>
15386             Jakub Jelinek  <jakub@redhat.com>
15388         PR tree-optimization/85692
15389         * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
15390         source permute as well.
15392 2018-05-11  Martin Liska  <mliska@suse.cz>
15394         PR sanitizer/85556
15395         * doc/extend.texi: Document LLVM style format for no_sanitize
15396         attribute.
15398 2018-05-10  Michael Meissner  <meissner@linux.vnet.ibm.com>
15400         * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
15401         mode_supports_vsx_dform_quad to mode_supports_dq_form.
15402         (mode_supports_vsx_dform_quad): Likewise.
15403         (mode_supports_vmx_dform): Move these functions to be next to the
15404         other mode_supports functions.
15405         (mode_supports_dq_form): Likewise.
15406         (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
15407         mode_supports_dq_form.
15408         (reg_offset_addressing_ok_p): Likewise.
15409         (offsettable_ok_by_alignment): Likewise.
15410         (rs6000_legitimate_offset_address_p): Likewise.
15411         (legitimate_lo_sum_address_p): Likewise.
15412         (rs6000_legitimize_address): Likewise.
15413         (rs6000_legitimize_reload_address): Likewise.
15414         (rs6000_secondary_reload_inner): Likewise.
15415         (rs6000_preferred_reload_class): Likewise.
15416         (rs6000_output_move_128bit): Likewise.
15418 2018-05-10  Uros Bizjak  <ubizjak@gmail.com>
15420         * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
15421         Generate SImode target register for null target.
15422         <case IX86_BUILTIN_XGETBV>: Ditto.
15423         <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
15424         * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
15426 2018-05-10  Carl Love  <cel@us.ibm.com>
15428         * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
15429         dcbtt and dcbtstt if operands[2] is 0.
15431 2018-05-10  Uros Bizjak  <ubizjak@gmail.com>
15433         PR target/85693
15434         * config/i386/sse.md (usadv64qi): New expander.
15436 2018-05-10  Segher Boessenkool  <segher@kernel.crashing.org>
15438         * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
15439         altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
15440         -maltivec=be support.
15441         (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
15442         vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
15443         vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
15444         vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
15445         vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
15446         vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
15447         altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
15448         altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
15449         altivec_vsumsws): Adjust.
15450         (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
15451         *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
15452         altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
15453         support.
15454         (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
15455         altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
15456         altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
15457         (altivec_lve<VI_char>x): Delete expand.
15458         (*altivec_lve<VI_char>x_internal): Rename to...
15459         (altivec_lve<VI_char>x): ... this.
15460         (altivec_lvxl_<mode>): Delete expand.
15461         (*altivec_lvxl_<mode>_internal): Rename to ...
15462         (altivec_lvxl_<mode>): ... this.
15463         (altivec_stvxl_<mode>): Delete expand.
15464         (*altivec_stvxl_<mode>_internal): Rename to ...
15465         (altivec_stvxl_<mode>): ... this.
15466         (altivec_stve<VI_char>x): Delete expand.
15467         (*altivec_stve<VI_char>x_internal): Rename to ...
15468         (altivec_stve<VI_char>x): ... this.
15469         (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
15470         doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
15471         reduc_plus_scal_<mode>): Adjust.
15472         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
15473         comment.
15474         (rs6000_cpu_cpp_builtins): Adjust.
15475         (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
15476         * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
15477         altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
15478         * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
15479         -maltivec=be support.
15480         (rs6000_split_vec_extract_var): Adjust.
15481         (rs6000_split_v4si_init): Adjust.
15482         (swap_selector_for_mode): Delete.
15483         (altivec_expand_lvx_be, altivec_expand_stvx_be,
15484         altivec_expand_stvex_be): Delete.
15485         (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
15486         -maltivec=be support.
15487         (rs6000_gimple_fold_builtin): Ditto.
15488         (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
15489         Adjust.
15490         * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
15491         (TARGET_DIRECT_MOVE_64BIT): Adjust.
15492         * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
15493         * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
15494         * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
15495         unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
15496         vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
15497         *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
15498         *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
15499         *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
15500         anonymous split): Adjust.
15501         (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
15502         (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
15504 2018-05-10  Eric Botcazou  <ebotcazou@adacore.com>
15506         * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
15507         when --with-gxx-include-dir is also specified.
15508         * configure: Regenerate.
15510 2018-05-09  Jim Wilson  <jimw@sifive.com>
15512         PR target/84797
15513         * config.gcc (riscv*-*-*): Handle --with-multilib-list.
15514         * config/riscv/t-withmultilib: New.
15515         * config/riscv/withmultilib.h: New.
15516         * doc/install.texi: Document RISC-V --with-multilib-list support.
15518 2018-05-09  Richard Biener  <rguenther@suse.de>
15520         * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
15521         vector.
15522         (vect_bb_vectorization_profitable_p): Adjust.  Compute
15523         actual scalar cost using the cost vector and the add_stmt_cost
15524         machinery.
15526 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
15528         PR rtl-optimization/85645
15529         * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
15530         in the REG_CFA_REGISTER note for LR, don't leave it empty.
15532 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
15534         PR rtl-optimization/85645
15535         * shrink-wrap.c (spread_components): Return a boolean saying if
15536         anything was changed.
15537         (try_shrink_wrapping_separate): Iterate spread_components until
15538         nothing changes anymore.
15540 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
15542         PR rtl-optimization/85645
15543         * regrename.c (build_def_use): Also kill the chains that include the
15544         destination of a REG_CFA_REGISTER note.
15546 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
15548         PR rtl-optimization/85645
15549         *  regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
15550         insn that has a REG_CFA_REGISTER note.
15552 2018-05-09  Richard Sandiford  <richard.sandiford@linaro.org>
15554         * cfgexpand.c (expand_clobber): New function.
15555         (expand_gimple_stmt_1): Use it.
15556         * tree-vect-stmts.c (vect_clobber_variable): New function,
15557         split out from...
15558         (vectorizable_simd_clone_call): ...here.
15559         (vectorizable_store): Emit a clobber either side of an
15560         IFN_STORE_LANES sequence.
15561         (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
15563 2018-05-09  Tom de Vries  <tom@codesourcery.com>
15565         PR target/85626
15566         * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
15567         (define_insn "trap_if_false"): Add exit after trap.
15569 2018-05-09  Eric Botcazou  <ebotcazou@adacore.com>
15571         PR rtl-optimization/85638
15572         * bb-reorder.c: Include common/common-target.h.
15573         (create_forwarder_block): New function extracted from...
15574         (fix_up_crossing_landing_pad): ...here.  Rename into...
15575         (dw2_fix_up_crossing_landing_pad): ...this.
15576         (sjlj_fix_up_crossing_landing_pad): New function.
15577         (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
15578         call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
15579         from both partitions and exit the loop after one iteration.
15581 2018-05-08  Kelvin Nilsen  <kelvin@gcc.gnu.org>
15583         Revert:
15584         * doc/extend.texi (PowerPC Built-in Functions): Rename this
15585         subsection.
15586         (Basic PowerPC Built-in Functions): The new name of the
15587         subsection previously known as "PowerPC Built-in Functions".
15588         (Basic PowerPC Built-in Functions Available on all Configurations):
15589         New subsubsection.
15590         (Basic PowerPC Built-in Functions Available on ISA 2.05): New
15591         subsubsection.
15592         (Basic PowerPC Built-in Functions Available on ISA 2.06): New
15593         subsubsection.
15594         (Basic PowerPC Built-in Functions Available on ISA 2.07): New
15595         subsubsection.
15596         (Basic PowerPC Built-in Functions Available on ISA 3.0): New
15597         subsubsection.
15599 2018-05-08  Jim Wilson  <jimw@sifive.com>
15601         * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
15602         (LD_EMUL_SUFFIX): New.
15603         (LINK_SPEC): Use it.
15605 2018-05-08  Kelvin Nilsen  <kelvin@gcc.gnu.org>
15607         * doc/extend.texi (PowerPC Built-in Functions): Rename this
15608         subsection.
15609         (Basic PowerPC Built-in Functions): The new name of the
15610         subsection previously known as "PowerPC Built-in Functions".
15611         (Basic PowerPC Built-in Functions Available on all Configurations):
15612         New subsubsection.
15613         (Basic PowerPC Built-in Functions Available on ISA 2.05): New
15614         subsubsection.
15615         (Basic PowerPC Built-in Functions Available on ISA 2.06): New
15616         subsubsection.
15617         (Basic PowerPC Built-in Functions Available on ISA 2.07): New
15618         subsubsection.
15619         (Basic PowerPC Built-in Functions Available on ISA 3.0): New
15620         subsubsection.
15622 2018-05-08  Jakub Jelinek  <jakub@redhat.com>
15624         PR target/85683
15625         * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
15626         after cmpelim optimization.
15628 2018-05-08  Olga Makhotina  <olga.makhotina@intel.com>
15630         * config.gcc: Support "goldmont".
15631         * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
15632         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
15633         PROCESSOR_GOLDMONT.
15634         * config/i386/i386.c (m_GOLDMONT): Define.
15635         (processor_target_table): Add "goldmont".
15636         (PTA_GOLDMONT): Define.
15637         (ix86_lea_outperforms): Add TARGET_GOLDMONT.
15638         (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
15639         (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
15640         (fold_builtin_cpu): Add "goldmont".
15641         (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
15642         (ix86_option_override_internal): Add "goldmont".
15643         * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
15644         (processor_type): Add PROCESSOR_GOLDMONT.
15645         * config/i386/i386.md: Add CPU "glm".
15646         * config/i386/glm.md: New file.
15647         * config/i386/x86-tune.def: Add m_GOLDMONT.
15648         * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
15650 2018-05-08  Jakub Jelinek  <jakub@redhat.com>
15652         PR target/85572
15653         * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
15654         E_V4DImode.
15655         * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
15656         VI1248_AVX512VL_AVX512BW.  Handle V2DImode and V4DImode if not
15657         TARGET_AVX512VL using ix86_expand_sse2_abs.  Formatting fixes.
15659         PR target/85317
15660         * config/i386/i386.c (ix86_fold_builtin): Handle
15661         IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
15663         PR target/85480
15664         * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
15665         (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
15667 2018-05-08  Richard Earnshaw  <rearnsha@arm.com>
15669         PR target/85658
15670         * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
15671         (check_arch): Likewise.
15672         (check_fpu): Return the result rather than printing it.
15673         (end arch): Fix operator precedence.
15674         (end cpu): Likewise.
15675         (END): Print the result from check_fpu.
15677 2018-05-08  Richard Sandiford  <richard.sandiford@linaro.org>
15678             Alan Hayward  <alan.hayward@arm.com>
15679             David Sherwood  <david.sherwood@arm.com>
15681         * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
15682         (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
15683         (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
15684         (*fcmuo<mode>_and): New patterns.
15686 2018-05-08  Richard Sandiford  <richard.sandiford@linaro.org>
15688         * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
15689         (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
15690         (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
15691         (cmp_op, sve_imm_con): New code attributes.
15692         (SVE_COND_INT_CMP, imm_con): Delete.
15693         (cmp_op): Remove above unspecs from int attribute.
15694         * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
15695         to...
15696         (*cmp<cmp_op><mode>): ...this.  Use UNSPEC_MERGE_PTRUE instead of
15697         comparison-specific unspecs.
15698         (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
15699         (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
15700         (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
15701         (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
15702         (*vec_fcm<cmp_op><mode>): Rename to...
15703         (*fcm<cmp_op><mode>): ...this and adjust likewise.
15704         (*vec_fcmuo<mode>): Rename to...
15705         (*fcmuo<mode>): ...this and adjust likewise.
15706         (*pred_fcm<cmp_op><mode>): New pattern.
15707         * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
15708         (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
15709         functions.
15710         (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
15711         and UNORDERED.
15712         (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
15713         (aarch64_emit_sve_predicated_cond): New function.
15714         (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
15715         (aarch64_emit_unspec_cond_or): Replace with...
15716         (aarch64_emit_sve_or_conds): ...this new function.  Use
15717         aarch64_emit_sve_ptrue_op for the individual comparisons and
15718         aarch64_emit_binop to OR them together.
15719         (aarch64_emit_inverted_unspec_cond): Replace with...
15720         (aarch64_emit_sve_inverted_cond): ...this new function.  Use
15721         aarch64_emit_sve_ptrue_op for the comparison and
15722         aarch64_emit_unop to invert the result.
15723         (aarch64_expand_sve_vec_cmp_float): Update after the above
15724         changes.  Use aarch64_emit_sve_ptrue_op for native comparisons.
15726 2018-05-07  Nathan Sidwell  <nathan@acm.org>
15728         * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
15729         * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
15730         (Backwards Compatibility): Likewise.
15732 2018-05-07  Luis Machado  <luis.machado@linaro.org>
15734         PR bootstrap/85681
15735         Revert:
15736         2018-05-07  Luis Machado  <luis.machado@linaro.org>
15738         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
15739         <prefetch_dynamic_strides>: New const bool field.
15740         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
15741         prefetch_dynamic_strides.
15742         (exynosm1_prefetch_tune): Likewise.
15743         (thunderxt88_prefetch_tune): Likewise.
15744         (thunderx_prefetch_tune): Likewise.
15745         (thunderx2t99_prefetch_tune): Likewise.
15746         (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
15747         to false.
15748         (aarch64_override_options_internal): Update to set
15749         PARAM_PREFETCH_DYNAMIC_STRIDES.
15750         * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
15751         * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
15752         * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
15753         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
15754         prefetch-dynamic-strides setting.
15756         2018-05-07  Luis Machado  <luis.machado@linaro.org>
15758         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
15759         <minimum_stride>: New const int field.
15760         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
15761         minimum_stride field.
15762         (exynosm1_prefetch_tune): Likewise.
15763         (thunderxt88_prefetch_tune): Likewise.
15764         (thunderx_prefetch_tune): Likewise.
15765         (thunderx2t99_prefetch_tune): Likewise.
15766         (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
15767         (aarch64_override_options_internal): Update to set
15768         PARAM_PREFETCH_MINIMUM_STRIDE.
15769         * doc/invoke.texi (prefetch-minimum-stride): Document new option.
15770         * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
15771         * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
15772         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
15773         stride is constant and is below the minimum stride threshold.
15775 2018-05-07  Luis Machado  <luis.machado@linaro.org>
15777         * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
15778         to 512.
15780 2018-05-07  Luis Machado  <luis.machado@linaro.org>
15782         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
15783         <prefetch_dynamic_strides>: New const bool field.
15784         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
15785         prefetch_dynamic_strides.
15786         (exynosm1_prefetch_tune): Likewise.
15787         (thunderxt88_prefetch_tune): Likewise.
15788         (thunderx_prefetch_tune): Likewise.
15789         (thunderx2t99_prefetch_tune): Likewise.
15790         (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
15791         to false.
15792         (aarch64_override_options_internal): Update to set
15793         PARAM_PREFETCH_DYNAMIC_STRIDES.
15794         * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
15795         * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
15796         * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
15797         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
15798         prefetch-dynamic-strides setting.
15800 2018-05-07  Luis Machado  <luis.machado@linaro.org>
15802         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
15803         <minimum_stride>: New const int field.
15804         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
15805         minimum_stride field.
15806         (exynosm1_prefetch_tune): Likewise.
15807         (thunderxt88_prefetch_tune): Likewise.
15808         (thunderx_prefetch_tune): Likewise.
15809         (thunderx2t99_prefetch_tune): Likewise.
15810         (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
15811         (aarch64_override_options_internal): Update to set
15812         PARAM_PREFETCH_MINIMUM_STRIDE.
15813         * doc/invoke.texi (prefetch-minimum-stride): Document new option.
15814         * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
15815         * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
15816         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
15817         stride is constant and is below the minimum stride threshold.
15819 2018-05-06  Jakub Jelinek  <jakub@redhat.com>
15821         PR c++/85659
15822         * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
15823         the type is addressable.  Don't force op into register if it has
15824         BLKmode.
15826 2018-05-05  Roland McGrath  <mcgrathr@google.com>
15828         PR other/77609
15829         * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
15830         any section for which we don't know a specific type it should have,
15831         regardless of name.  Previously this was done only for the exact
15832         names ".init_array", ".fini_array", and ".preinit_array".
15833         (default_elf_asm_named_section): Add comment about
15834         relationship with default_section_type_flags and SECTION_NOTYPE.
15835         (get_section): Don't consider it a type conflict if one side has
15836         SECTION_NOTYPE and the other doesn't, as long as neither has the
15837         SECTION_BSS et al used in the default_section_type_flags logic.
15839 2018-05-05  Tom de Vries  <tom@codesourcery.com>
15841         PR target/85653
15842         * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
15843         (workaround_barsyncs): New function.
15844         (nvptx_reorg): Use workaround_barsyncs.
15845         * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
15846         (define_expand "nvptx_membar_cta"): New define_expand.
15847         (define_insn "*nvptx_membar_cta"): New insn.
15849 2018-05-04  Pekka Jääskeläinen  <pekka.jaaskelainen@parmance.com>
15851         * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
15852         To improve optimization opportunities.
15853         * builtin-types.def: The new needed builtin types for the above.
15855 2018-05-04  Richard Biener  <rguenther@suse.de>
15857         * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
15858         * gimple-ssa-store-merging.c
15859         (imm_store_chain_info::output_merged_store): Remove redundant create,
15860         release split_store vector contents on failure.
15861         * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
15862         scalar stmt vector on cache hit.
15864 2018-05-04  Segher Boessenkool  <segher@kernel.crashing.org>
15866         * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
15867         Xilinx FP support.
15868         * config.gcc (powerpc-xilinx-eabi*): Remove.
15869         * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
15870         support.
15871         (fusion_addis_mem_combo_load): Ditto.
15872         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
15873         FP support.
15874         (rs6000_cpu_cpp_builtins): Ditto.
15875         * config/rs6000/rs6000-linux.c
15876         (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
15877         * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
15878         * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
15879         support.
15880         (rs6000_setup_reg_addr_masks): Ditto.
15881         (rs6000_init_hard_regno_mode_ok): Ditto.
15882         (rs6000_option_override_internal): Ditto.
15883         (legitimate_lo_sum_address_p): Ditto.
15884         (rs6000_legitimize_address): Ditto.
15885         (rs6000_legitimize_reload_address): Ditto.
15886         (rs6000_legitimate_address_p): Ditto.
15887         (abi_v4_pass_in_fpr): Ditto.
15888         (setup_incoming_varargs): Ditto.
15889         (rs6000_gimplify_va_arg): Ditto.
15890         (rs6000_split_multireg_move): Ditto.
15891         (rs6000_savres_strategy): Ditto.
15892         (rs6000_emit_prologue_components): Ditto.
15893         (rs6000_emit_epilogue_components): Ditto.
15894         (rs6000_emit_prologue): Ditto.
15895         (rs6000_emit_epilogue): Ditto.
15896         (rs6000_elf_file_end): Ditto.
15897         (rs6000_function_value): Ditto.
15898         (rs6000_libcall_value): Ditto.
15899         * config/rs6000/rs6000.h: Ditto.
15900         (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
15901         (TARGET_MINMAX): ... this.  New.
15902         (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
15903         * config/rs6000/rs6000.md: Remove Xilinx FP support.
15904         (*movsi_internal1_single): Delete.
15905         * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
15906         mfpu=, mxilinx-fpu): Delete.
15907         * config/rs6000/singlefp.h: Delete.
15908         * config/rs6000/sysv4.h: Remove Xilinx FP support.
15909         * config/rs6000/t-rs6000: Ditto.
15910         * config/rs6000/t-xilinx: Delete.
15911         * config/rs6000/titan.md: Adjust for fp_type removal.
15912         * config/rs6000/vsx.md: Remove Xilinx FP support.
15913         (VStype_simple): Delete.
15914         (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
15915         * config/rs6000/xfpu.h: Delete.
15916         * config/rs6000/xfpu.md: Delete.
15917         * config/rs6000/xilinx.h: Delete.
15918         * config/rs6000/xilinx.opt: Delete.
15919         * doc/invoke.texi (RS/6000 and PowerPC Options): Remove
15920         -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
15922 2018-05-04  Tom de Vries  <tom@codesourcery.com>
15924         PR libgomp/85639
15925         * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
15926         if ignore == 0.
15928 2018-05-04  Richard Biener  <rguenther@suse.de>
15930         PR middle-end/85627
15931         * tree-complex.c (update_complex_assignment): We are always in SSA form.
15932         (expand_complex_div_wide): Likewise.
15933         (expand_complex_operations_1): Likewise.
15934         (expand_complex_libcall): Preserve EH info of the original stmt.
15935         (tree_lower_complex): Handle removed blocks.
15936         * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
15937         on complex multiplication and division libcall builtins.
15939 2018-05-04  Richard Biener  <rguenther@suse.de>
15941         PR middle-end/85574
15942         * fold-const.c (negate_expr_p): Restrict negation of operand
15943         zero of a division to when we know that can happen without
15944         overflow.
15945         (fold_negate_expr_1): Likewise.
15947 2018-05-04  Jakub Jelinek  <jakub@redhat.com>
15949         PR libstdc++/85466
15950         * real.h (real_nextafter): Declare.
15951         * real.c (real_nextafter): New function.
15952         * fold-const-call.c (fold_const_nextafter): New function.
15953         (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
15954         CASE_CFN_NEXTTOWARD.
15955         (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
15956         even when arg1_mode is different from arg0_mode.
15958 2018-05-03  Nathan Sidwell  <nathan@acm.org>
15960         * doc/extend.texi (Deprecated Features): Remove
15961         -ffriend-injection.
15962         (Backwards Compatibility): Likewise.
15963         * doc/invoke.texi (C++ Language Options): Likewise.
15964         (C++ Dialect Options): Likewise.
15966 2018-05-03  Jakub Jelinek  <jakub@redhat.com>
15968         PR target/85530
15969         * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
15970         _mm512_mask_mullox_epi64): New intrinsics.
15972 2018-05-03  Tom de Vries  <tom@codesourcery.com>
15974         PR testsuite/85106
15975         * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
15976         dump files): Add offload-tree.
15978 2018-05-03  Richard Biener  <rguenther@suse.de>
15980         PR tree-optimization/85615
15981         * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
15982         to loops not nested in BBs loop father to avoid creating multi-entry
15983         loops.
15985 2018-05-03  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
15987         PR tree-optimization/70291
15988         * tree-complex.c (expand_complex_libcall): Add type, inplace_p
15989         arguments.  Change return type to tree.  Emit libcall as a new
15990         statement rather than replacing existing one when inplace_p is true.
15991         (expand_complex_multiplication_components): New function.
15992         (expand_complex_multiplication): Expand floating-point complex
15993         multiplication using the above.
15994         (expand_complex_division): Rename inner_type parameter to type.
15995         Update expand_complex_libcall call-site.
15996         (expand_complex_operations_1): Update expand_complex_multiplication
15997         and expand_complex_division call-sites.
15999 2018-05-02  Jakub Jelinek  <jakub@redhat.com>
16001         PR target/85582
16002         * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
16003         *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
16004         *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
16005         the highest significant bit of the shift count mask is clear.  In
16006         check whether and[sq]i3 is needed verify that all significant bits
16007         of the shift count other than the highest are set.
16009 2018-05-02  Tom de Vries  <tom@codesourcery.com>
16011         PR libgomp/82428
16012         * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
16013         * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
16014         (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
16015         * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
16016         (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
16017         * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
16018         __builtin_goacc_parlevel_size.
16020 2018-05-02  Richard Biener  <rguenther@suse.de>
16022         PR tree-optimization/85597
16023         * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
16024         do not use split vect_get_vec_defs call but call vect_get_slp_defs
16025         directly.
16027 2018-05-02  Tom de Vries  <tom@codesourcery.com>
16029         PR testsuite/85106
16030         * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
16031         dump files): Add ltrans-tree.
16033 2018-05-02  Tom de Vries  <tom@codesourcery.com>
16035         PR testsuite/85106
16036         * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
16037         dump files): Add wpa-ipa.
16039 2018-05-02  Segher Boessenkool  <segher@kernel.crashing.org>
16041         * config.gcc (powerpc*-*-*): Remove paired.h.  Unsupport the
16042         powerpc*-*-linux*paired* target.
16043         * config/rs6000/750cl.h: Delete.
16044         * config/rs6000/paired.h: Delete.
16045         * config/rs6000/paired.md: Delete.
16046         * config/rs6000/predicates.md (easy_vector_constant): Remove paired
16047         float support.
16048         * config/rs6000/rs6000-builtin.def: Remove paired float support.
16049         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
16050         comment.  Remove paired float support.
16051         * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
16052         * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
16053         VECTOR_PAIRED.
16054         * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
16055         paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
16056         declarations.
16057         * config/rs6000/rs6000.c: Remove paired float support.
16058         (paired_expand_vector_init, paired_expand_vector_move,
16059         paired_emit_vector_compare, paired_emit_vector_cond_expr,
16060         (paired_expand_lv_builtin, paired_expand_stv_builtin,
16061         paired_expand_builtin, paired_expand_predicate_builtin,
16062         paired_init_builtins): Delete.
16063         * config/rs6000/rs6000.h: Remove paired float support.
16064         * config/rs6000/rs6000.md: Remove paired float support.
16065         (move_from_CR_ov_bit): Delete.
16066         * config/rs6000/rs6000.opt (mpaired): Delete.
16067         * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
16068         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
16070 2018-05-02  Richard Biener  <rguenther@suse.de>
16072         PR middle-end/85567
16073         * gimplify.c (gimplify_save_expr): When in SSA form allow
16074         SAVE_EXPRs to compute to SSA vars.
16076 2018-05-02  Jakub Jelinek  <jakub@redhat.com>
16078         PR target/85582
16079         * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
16080         *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
16081         *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
16082         clobber operands[2], instead use a new pseudo.  Formatting fixes.
16084 2018-05-02  Richard Sandiford  <richard.sandiford@linaro.org>
16086         PR tree-optimization/85586
16087         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
16088         exit early for statements in the same group if the accesses are
16089         not strided.
16091 2018-05-02  Tom de Vries  <tom@codesourcery.com>
16093         PR lto/85451
16094         * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
16095         error message.
16097 2018-05-01  Marc Glisse  <marc.glisse@inria.fr>
16099         PR tree-optimization/85143
16100         * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
16102 2018-05-01  Tom de Vries  <tom@codesourcery.com>
16104         PR lto/85451
16105         * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
16106         not found" error message.
16108 2018-05-01  Tom de Vries  <tom@codesourcery.com>
16110         PR other/83786
16111         * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
16112         * vec.c (test_ordered_remove_if): New function.
16113         (vec_c_tests): Call test_ordered_remove_if.
16114         * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
16115         * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
16116         * tree-vect-patterns.c (vect_pattern_recog_1): Use
16117         VEC_ORDERED_REMOVE_IF.
16119 2018-05-01  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
16121         PR tree-optimization/82665
16122         * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
16123         pointer subtraction where arguments come from a memchr call.
16125 2018-05-01  Jakub Jelinek  <jakub@redhat.com>
16127         * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
16128         --push-state --as-needed and --pop-state instead of --as-needed and
16129         --no-as-needed if ld supports it.
16130         * configure: Regenerated.
16132         PR web/85578
16133         * doc/install.texi2html: Replace _002d with - and _002a with * in
16134         generated html files using sed.
16136 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
16138         PR c++/85523
16139         * gcc-rich-location.c (blank_line_before_p): New function.
16140         (use_new_line): New function.
16141         (gcc_rich_location::add_fixit_insert_formatted): New function.
16142         * gcc-rich-location.h
16143         (gcc_rich_location::add_fixit_insert_formatted): New function.
16145 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
16147         * selftest.c (assert_streq): Rename "expected" and "actual" to
16148         "val1" and "val2".  Extend NULL-handling to cover both inputs
16149         symmetrically, while still requiring both to be non-NULL for a pass.
16150         * selftest.h (assert_streq): Rename "expected" and "actual" to
16151         "val1" and "val2".
16152         (ASSERT_EQ): Likewise.
16153         (ASSERT_EQ_AT): Likewise.
16154         (ASSERT_KNOWN_EQ): Likewise.
16155         (ASSERT_KNOWN_EQ_AT): Likewise.
16156         (ASSERT_NE): Likewise.
16157         (ASSERT_MAYBE_NE): Likewise.
16158         (ASSERT_MAYBE_NE_AT): Likewise.
16159         (ASSERT_STREQ): Likewise.  Clarify that both must be non-NULL for
16160         the assertion to pass.
16161         (ASSERT_STREQ_AT): Likewise.
16163 2018-04-30  Jonathan Wakely  <jwakely@redhat.com>
16165         * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
16166         interaction with -pie.
16168 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
16170         * selftest.h: Fix alphabetization of per-source-file selftest
16171         declarations.
16173 2018-04-30  Jason Merrill  <jason@redhat.com>
16175         PR c++/61982 - dead stores to destroyed objects.
16176         * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
16177         of clobber.
16179 2018-04-30  Jason Merrill  <jason@redhat.com>
16181         * tree.c (build_clobber): New.
16182         * tree.h: Declare it.
16183         * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
16185 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
16187         * diagnostic-show-locus.c (layout::layout): Update for
16188         location_get_source_line returning a char_span.
16189         (struct char_span): Move to input.h.
16190         (struct correction): Update for fields in char_span becoming
16191         private.
16192         (struct source_line): Update for location_get_source_line
16193         returning a char_span.
16194         (layout::print_line): Likewise.
16195         * edit-context.c (edited_file::print_content): Likewise.
16196         (edited_file::print_diff_hunk): Likewise.
16197         (edited_file::print_run_of_changed_lines): Likewise.
16198         (edited_file::get_num_lines): Likewise.
16199         (edited_line::edited_line): Likewise.
16200         * final.c (asm_show_source): Likewise.
16201         * input.c (location_get_source_line): Convert return type
16202         from const char * to char_span, losing the final "line_len"
16203         param.
16204         (dump_location_info): Update for the above.
16205         (get_substring_ranges_for_loc): Likewise.  Use a char_span
16206         when handling the literal within the line.
16207         (test_reading_source_line): Update for location_get_source_line
16208         returning a char_span.
16209         * input.h (class char_span): Move here from
16210         diagnostic-show-locus.c, converting from a struct to a class.
16211         Make data members private.
16212         (char_span::operator bool): New.
16213         (char_span::length): New.
16214         (char_span::get_buffer): New.
16215         (char_span::operator[]): New.
16216         (char_span::subspan): Make const.
16217         (char_span::xstrdup): New.
16218         (location_get_source_line): Convert return type from const char *
16219         to char_span, losing the final "line_size" param.
16221 2018-04-30  Jan Hubicka  <jh@suse.cz>
16223         * lto-wrapper.c (ltrans_priorities): New static var.
16224         (cmp_priority): New.
16225         (run_gcc): Read priorities and if doing parallel build order
16226         the Makefile by them.
16228 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
16230         * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
16232 2018-04-30  Richard Biener  <rguenther@suse.de>
16234         * tree-cfg.c (verify_address): Remove base argument, add
16235         flag whether to check TREE_ADDRESSABLE and do that.
16236         (verify_expr): Remove.
16237         (verify_types_in_gimple_reference): Add pieces from verify_expr.
16238         (verify_gimple_assign_single): Likewise.
16239         (verify_gimple_switch): Likewise.
16240         (verify_expr_location_1): Dereference tp once.  Add (disabled)
16241         piece from verify_expr.
16242         (verify_gimple_in_cfg): Do not call verify_expr on all ops.
16244 2018-04-30  Claudiu Zissulescu  <claziss@synopsys.com>
16246         * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
16248 2018-04-30  Claudiu Zissulescu  <claziss@synopsys.com>
16250         * config/arc/arc-protos.h (prepare_extend_operands): Remove.
16251         (small_data_pattern): Likewise.
16252         (arc_rewrite_small_data): Likewise.
16253         * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
16254         (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
16255         (get_symbol_alignment): New function.
16256         (legitimate_small_data_address_p): Likewise.
16257         (legitimate_scaled_address): Update, call
16258         legitimate_small_data_address_p.
16259         (output_sdata): New static variable.
16260         (arc_print_operand): Update how we handle small data operands.
16261         (arc_print_operand_address): Likewise.
16262         (arc_legitimate_address_p): Update, use
16263         legitimate_small_data_address_p.
16264         (arc_rewrite_small_data_p): Remove.
16265         (arc_rewrite_small_data_1): Likewise.
16266         (arc_rewrite_small_data): Likewise.
16267         (small_data_pattern): Likewise.
16268         (compact_sda_memory_operand): Update to use
16269         legitimate_small_data_address_p and get_symbol_alignment.
16270         (prepare_move_operands): Don't rewite sdata pattern.
16271         (prepare_extend_operands): Remove.
16272         * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
16273         pattern.
16274         (zero_extendqisi2): Likewise.
16275         (zero_extendhisi2): Likewise.
16276         (extendqihi2): Likewise.
16277         (extendqisi2): Likewise.
16278         (extendhisi2): Likewise.
16279         (addsi3): Likewise.
16280         (subsi3): Likewise.
16281         (andsi3): Likewise.
16282         * config/arc/constraints.md (Usd): Change it to memory constraint.
16284 2018-04-30  Claudiu Zissulescu  <claziss@synopsys.com>
16286         * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
16287         as source of std instructions.
16288         * config/arc/arc.md (movsi_insn): Update pattern predicate to
16289         allow 6-bit constants as source for store instructions.
16290         (movdi_insn): Update instruction pattern to allow 6-bit constants
16291         as source for store instructions.
16293 2018-04-30  Jonathan Wakely  <jwakely@redhat.com>
16295         * doc/invoke.texi (-fdebug-types-section): Fix grammar.
16297 2018-04-30  Nathan Sidwell  <nathan@acm.org>
16298             Sandra Loosemore <sandra@codesourcery.com>
16300         * dumpfile.c (dump_open): Allow '-' for stdout.
16301         * doc/invoke.texi (Developer Options): Document dump filename
16302         determination early.  Document stdin/stdout selection.
16304 2018-04-30  Andrew Sadek  <andrew.sadek.se@gmail.com>
16306         Microblaze Target: PIC data text relative
16308         * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
16309         * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
16310         Add declaration.
16311         * config/microblaze/microblaze.h (microblaze_constant_address_p):
16312         CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
16313         * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
16314         New addressing mode for data-text relative position indepenedent code.
16315         (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
16316         'ADDRESS_SYMBOLIC_TXT_REL'.
16317         (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
16318         (microblaze_legitimate_pic_operand): Exclude function calls from
16319         pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
16320         (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
16321         addresses cases.
16322         (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
16323         (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
16324         (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
16325         for 'address + offset'.
16326         (microblaze_expand_prologue): Add new function prologue call for
16327         'r20' assignation.
16328         (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
16329         'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
16330         table in case of TARGET_PIC_DATA_TEXT_REL.
16331         (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
16332         * config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
16333         Add new macros 'UNSPEC_TEXT',
16334         'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
16335         + exclude function calls from 'UNSPEC_PLT' in case of data text
16336         relative mode.
16337         * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
16338         new target hook for generating address diff vector tables in case of
16339         flag_pic.
16340         * doc/tm.texi : Regenerate.
16341         * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
16342         'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
16343         of addr diff vector generation.
16344         * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
16345         target hook definition.
16346         * targhooks.h, targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
16347         Add default function for generate_pic_addr_diff_vec -> flag_pic.
16348         * doc/invoke.texi (Add new pic option): Add new microblaze pic
16349         option for data text relative.
16351 2018-04-30  Richard Biener  <rguenther@suse.de>
16353         * tree-chrec.h (evolution_function_is_constant_p): Remove
16354         redundant check.
16355         * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
16357 2018-04-30  Richard Biener  <rguenther@suse.de>
16359         PR bootstrap/85571
16360         * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
16362 2018-04-30  Richard Biener  <rguenther@suse.de>
16364         PR tree-optimization/28364
16365         PR tree-optimization/85275
16366         * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
16367         copying first exit test.
16369 2018-04-28  Mark Wielaard  <mark@klomp.org>
16371         * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
16372         dwarf_version >= 5.
16373         (dwarf_AT): Handle DW_AT_addr_base.
16374         (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
16376 2018-04-28  Uros Bizjak  <ubizjak@gmail.com>
16378         PR target/84431
16379         * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
16380         (*ashl<dwi>3_doubleword_mask_1): Ditto.
16381         (*<shift_insn><dwi>3_doubleword_mask): Ditto.
16382         (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
16384 2018-04-28  Richard Biener  <rguenther@suse.de>
16386         * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
16387         (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
16388         to reflect use.  Only add interesting stmts.
16390 2018-04-27  Martin Jambor  <mjambor@suse.cz>
16392         PR ipa/85549
16393         * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
16394         the jump function allows for passing through aggregate values.
16396 2018-04-27  David Malcolm  <dmalcolm@redhat.com>
16398         * input.h (in_system_header_at): Convert from macro to inline
16399         function.
16400         (from_macro_expansion_at): Likewise.
16401         (from_macro_definition_at): Likewise.
16403 2018-04-27  Jeff Law  <law@redhat.com>
16405         * config.gcc: Mark tile* targets as deprecated/obsolete.
16407 2018-04-27  Richard Biener  <rguenther@suse.de>
16409         * config/aarch64/aarch64.c: Simplify ap.__stack advance and
16410         fix for ILP32.
16412 2018-04-27  Richard Biener  <rguenther@suse.de>
16414         * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
16416 2018-04-27  Uros Bizjak  <ubizjak@gmail.com>
16418         * config/i386/i386.md (*movti_internal): Substitute Ye constraint
16419         with Yd constraint. Set "preferred_for_speed" attribute from
16420         TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
16421         with Yd constraint.
16422         (*movdi_internal): Ditto.
16423         (movti_interunit splitters): Remove
16424         TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
16425         (movdi_interunit splitters): Ditto.
16426         * config/i386/constraints.md (Ye): Remove.
16427         (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
16429 2018-04-27  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
16431         PR target/85512
16432         * config/aarch64/constraints.md (Usg): Limit to 31.
16433         (Usj): Limit to 63.
16435 2018-04-27  Jakub Jelinek  <jakub@redhat.com>
16437         PR tree-optimization/85529
16438         * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
16439         argument.  Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
16440         rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
16441         zero extension or masking of the MSB bit.
16442         (optimize_range_tests): Add FIRST_BB argument, pass it through
16443         to optimize_range_tests_var_bound.
16444         (maybe_optimize_range_tests, reassociate_bb): Adjust
16445         optimize_range_tests callers.
16447 2018-04-26  Richard Biener  <rguenther@suse.de>
16448             Jakub Jelinek  <jakub@redhat.com>
16450         * cgraph.h (symbol_table): Just declare debug method here.
16451         * symtab.c (symbol_table::debug): Define.
16453 2018-04-26  Eric Botcazou  <ebotcazou@adacore.com>
16455         * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
16457 2018-04-26  Uros Bizjak  <ubizjak@gmail.com>
16459         * config/i386/i386.md ("isa" attribute): Add x64_sse2.
16460         ("enabled" attribute): Handle x64_sse2 "isa" attribute.
16461         (*movdi_internal): Substitute Yi and Yj constraint with x
16462         and Ym and Yn constraint with y constraint.  Update "isa"
16463         attribute and set "preferred_for_speed" attribute from
16464         TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
16465         (*movsi_internal): Ditto.
16466         (*movdf_internal): Ditto.
16467         (*movsf_internal): Ditto.
16468         (*zero_extendsidi2): Ditto.
16469         * config/i386/sse.md (vec_set<mode>_0): Ditto.
16470         (sse2_loadld): Ditto.
16471         (*vec_extract<ssevecmodelower>_0): Ditto.
16472         (*vec_extractv4si_0_zext_sse4): Ditto.
16473         (vec_concatv2di): Ditto.
16474         (*vec_dup<mode>): Ditto.
16475         * config/i386/mmx.md (*mov<mode>_internal): Ditto.
16476         * config/i386/constraints.md (Yi): Remove.
16477         (Yj): Remove.
16478         (Ym): Remove.
16479         (Yn): Remove.
16481 2018-04-26  Nathan Sidwell  <nathan@acm.org>
16483         * dumpfile.c (dump_open): New.
16484         (dump_open_alternate_stream, dump_start, dump_begin): Call it.
16485         (dump_finish): Detect stdio/stderr by value not name.
16487 2018-04-26  Jonathan Wakely  <jwakely@redhat.com>
16489         * doc/invoke.texi (-Wreturn-type): Document default status for C++.
16491 2018-04-26  Tom de Vries  <tom@codesourcery.com>
16493         PR target/84952
16494         * config/nvptx/nvptx.c (verify_neutering_jumps)
16495         (verify_neutering_labels): New function
16496         (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
16498 2018-04-26  Tom de Vries  <tom@codesourcery.com>
16500         PR target/84025
16501         * config/nvptx/nvptx.c (needs_neutering_p): New function.
16502         (nvptx_single): Use needs_neutering_p to skip over insns that do not
16503         need neutering.
16505 2018-04-26  Richard Biener <rguenther@suse.de>
16506             Tom de Vries  <tom@codesourcery.com>
16508         PR lto/85422
16509         * lto-streamer-out.c (output_function): Fixup loops if required to match
16510         discovery done in the reader.
16512 2018-04-26  Richard Biener  <rguenther@suse.de>
16514         PR tree-optimization/85116
16515         * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
16516         have a loop exit from the single latch predecessor.  Remove
16517         case of header with just condition.
16518         (ch_base::copy_headers): Exclude infinite loops from any
16519         processing.
16520         (pass_ch::execute): Record exits.
16522 2018-04-26  Richard Biener  <rguenther@suse.de>
16524         * tree-vect-data-refs.c (vect_get_data_access_cost): Get
16525         prologue cost vector and pass it to vect_get_load_cost.
16526         (vect_get_peeling_costs_all_drs): Likewise.
16527         (vect_peeling_hash_get_lowest_cost): Likewise.
16528         (vect_enhance_data_refs_alignment): Likewise.
16530 2018-04-26  Richard Biener  <rguenther@suse.de>
16532         PR middle-end/85450
16533         * tree-cfg.c (verify_gimple_assign_unary): Restore proper
16534         checking of integer<->pointer conversions.
16535         * omp-expand.c (expand_omp_for_static_nochunk): Avoid
16536         sign-/zero-extending pointer types.
16537         (expand_omp_for_static_chunk): Likewise.
16539 2018-03-22  Hans-Peter Nilsson  <hp@axis.com>
16540             Jean Lee  <xiaoyur347@gmail.com>
16542         * config/mips/mips.c (mips_asan_shadow_offset): New function.
16543         (TARGET_ASAN_SHADOW_OFFSET): Define.
16544         * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
16545         true for -fsanitize=address.
16547 2018-04-25  Mark Wielaard  <mark@klomp.org>
16549         * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
16550         shorter ones.
16552 2018-04-25  Jakub Jelinek  <jakub@redhat.com>
16554         * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
16555         than "alu", remove explicit "memory" and "imm_disp" attributes.
16556         (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
16558         PR middle-end/85414
16559         * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
16560         case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
16561         gen_lowpart_no_emit.
16563 2018-04-25  Sebastian Peryt  <sebastian.peryt@intel.com>
16565         PR target/85473
16566         * config/i386/i386.c (ix86_expand_builtin): Change memory
16567         operand to XI, extend p0 to Pmode.
16568         * config/i386/i386.md: Change unspec volatile and operand
16569         1 mode to XI, change operand 0 mode to P.
16571 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
16573         * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
16574         GET_MODE_MASK before any checking.
16575         (nds32_can_use_bset_p): Likewise.
16576         (nds32_can_use_btgl_p): Likewise.
16578 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
16580         * config/nds32/nds32-doubleword.md: New define_split pattern for
16581         illegal register number.
16583 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
16585         * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
16587 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
16589         * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
16591 2018-04-25  Richard Biener  <rguenther@suse.de>
16593         * lto-streamer.h (LTO_major_version): Bump to 8.
16595 2018-04-25  Jakub Jelinek  <jakub@redhat.com>
16597         * BASE-VER: Set to 9.0.0.
16599 2018-04-24  Segher Boessenkool  <segher@kernel.crashing.org>
16601         * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
16602         in __abskf2 and __powikf2.
16604 2018-04-24  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
16606         PR target/85512
16607         * config/aarch64/constraints.md (Usg, Usj): New constraints.
16608         * config/aarch64/iterators.md (cmode_simd): New mode attribute.
16609         * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
16610         Use the above on operand 2.  Reindent.
16611         (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
16613 2018-04-24  H.J. Lu  <hongjiu.lu@intel.com>
16615         PR target/85485
16616         * common/config/i386/i386-common.c (ix86_handle_option): Don't
16617         handle OPT_mcet.
16618         * config/i386/i386.opt (mcet): Removed.
16619         * doc/install.texi: Remove -mcet documentation.
16620         * doc/invoke.texi: Likewise.
16622 2018-04-24  H.J. Lu  <hongjiu.lu@intel.com>
16624         PR target/85485
16625         * doc/install.texi: Remove -mcet from bootstrap-cet.
16627 2018-04-24  Jakub Jelinek  <jakub@redhat.com>
16629         PR target/85511
16630         * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
16631         __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
16632         if TARGET_64BIT.
16634         PR target/85503
16635         * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
16636         const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
16637         containing a CONST_VECTOR.
16639 2018-04-24  Cesar Philippidis  <cesar@codesourcery.com>
16641         * doc/install.texi: Update newlib dependency for nvptx.
16643 2018-04-24  Jakub Jelinek  <jakub@redhat.com>
16645         PR target/85508
16646         * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
16647         instead of INTVAL when shifting x left.
16649 2018-04-24  Andreas Krebbel  <krebbel@linux.ibm.com>
16651         PR tree-optimization/85478
16652         * tree-vect-loop.c (vect_analyze_loop_2): Do not call
16653         vect_grouped_store_supported for single element vectors.
16655 2018-04-24  Richard Biener  <rguenther@suse.de>
16657         PR target/85491
16658         * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
16659         load cost increase to the case of non-constant step.
16661 2018-04-24  Jakub Jelinek  <jakub@redhat.com>
16663         PR target/84828
16664         * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
16665         destination if any_malformed_asm.
16667 2018-04-23  Eric Botcazou  <ebotcazou@adacore.com>
16669         PR middle-end/85496
16670         * expr.c (store_field): In the bitfield case, if the value comes from
16671         a function call and is returned in registers by means of a PARALLEL,
16672         do not change the mode of the temporary unless BLKmode and VOIDmode.
16674 2018-04-23  Andrey Belevantsev  <abel@ispras.ru>
16676         PR rtl-optimization/85423
16677         * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
16678         dependencies to debug insns when the previous insn is non-debug.
16680 2018-04-23  Claudiu Zissulescu  <claziss@synopsys.com>
16682         * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
16683         enums into a single definition.
16684         (fls): Fix predicates and printing.
16685         (seti): Likewise.
16687 2018-04-23  Claudiu Zissulescu  <claziss@synopsys.com>
16689         * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
16690         * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
16691         and short u6 immediate.
16692         (check_if_valid_sleep_operand): Remove.
16693         * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
16695 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
16697         * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
16698         flag_always_save_lp condition.
16699         * config/nds32/nds32.opt (malways-save-lp): New option.
16701 2018-04-22  Shiva Chen  <shiva0217@gmail.com>
16703         * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
16704         * config/nds32/nds32.c (nds32_use_load_post_increment): New.
16705         * config/nds32/nds32.h
16706         (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
16707         (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
16709 2018-04-22  Shiva Chen  <shiva0217@gmail.com>
16711         * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
16712         * config/nds32/nds32.c (nds32_ls_333_p): Remove.
16714 2018-04-22  Shiva Chen  <shiva0217@gmail.com>
16715             Chung-Ju Wu  <jasonwucj@gmail.com>
16717         * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
16718         Declare.
16719         * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
16720         * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
16722 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
16724         * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
16726 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
16728         * config/nds32/nds32-protos.h (nds32_data_alignment,
16729         nds32_local_alignment): Declare.
16730         * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
16731         nds32_local_alignment): New functions.
16732         (TARGET_CONSTANT_ALIGNMENT): Define.
16733         * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
16735 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
16737         * config/nds32/nds32.c
16738         (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
16739         (TARGET_MODES_TIEABLE_P): Likewise.
16741 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
16743         * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
16744         level Ofast and Og.
16746 2018-04-22  Monk Chiang  <sh.chiang04@gmail.com>
16747             Chung-Ju Wu  <jasonwucj@gmail.com>
16749         * config/nds32/constants.md (unspec_volatile_element): Add enum values
16750         for unaligned access.
16751         * config/nds32/nds32-intrinsic.c: Implementation of expanding
16752         unaligned access.
16753         * config/nds32/nds32-intrinsic.md: Likewise.
16754         * config/nds32/nds32_intrinsic.h: Likewise.
16755         * config/nds32/nds32.h (nds32_builtins): Likewise.
16756         * config/nds32/nds32.opt (munaligned-access): New option.
16757         * config/nds32/nds32.c (nds32_asm_file_start): Display
16758         flag_unaligned_access status.
16760 2018-04-20  Kito Cheng  <kito.cheng@gmail.com>
16762         * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
16763         -mno-relax is present.
16764         * config/riscv/linux.h (LINK_SPEC): Ditto.
16766 2018-04-20  Martin Sebor  <msebor@redhat.com>
16768         PR c/85365
16769         * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
16770         for null pointers.
16771         (gimple_fold_builtin_stxcpy_chk): Same.
16772         * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
16774 2018-04-20  Michael Meissner  <meissner@linux.ibm.com>
16776         PR target/85456
16777         * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
16778         __powikf2 when long double is IEEE 128-bit.
16780 2018-04-20  Kito Cheng  <kito.cheng@gmail.com>
16782         * config/riscv/riscv.c (riscv_first_stack_step): Round up min
16783         step to make sure stack always aligned.
16785 2018-04-20  Carl Love  <cel@us.ibm.com>
16787         PR target/83402
16788         * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
16789         size check for arg0.
16791 2018-04-20  Nathan Sidwell  <nathan@codesourcery.com>
16792             Tom de Vries  <tom@codesourcery.com>
16794         PR target/85445
16795         * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
16796         Emit insns for calls too.
16797         (nvptx_find_par): Always look for worker-level predecessor insn.
16798         (nvptx_propagate): Add is_call parm, return bool.  Copy frame for
16799         calls.
16800         (nvptx_vpropagate, nvptx_wpropagate): Adjust.
16801         (nvptx_process_pars): Propagate frames for calls.
16803 2018-04-20  H.J. Lu  <hongjiu.lu@intel.com>
16805         PR target/85469
16806         * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
16807         Removed.
16808         (OPTION_MASK_ISA_IBT_UNSET): Likewise.
16809         (ix86_handle_option): Don't handle OPT_mibt.
16810         * config/i386/cet.h: Check __CET__ instead of __IBT__ and
16811         __SHSTK__.
16812         * config/i386/driver-i386.c (host_detect_local_cpu): Remove
16813         has_ibt and ibt.
16814         * config/i386/i386-c.c (ix86_target_macros_internal): Don't
16815         check OPTION_MASK_ISA_IBT nor flag_cf_protection.
16816         (ix86_target_macros): Define __CET__ with flag_cf_protection
16817         for -fcf-protection.
16818         * config/i386/i386.c (isa2_opts): Remove -mibt.
16819         * config/i386/i386.h (TARGET_IBT): Removed.
16820         (TARGET_IBT_P): Likewise.
16821         (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
16822         * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
16823         * config/i386/i386.opt (mcet): Update help message.
16824         (mshstk): Likewise.
16825         (mibt): Removed.
16826         * doc/invoke.texi: Remove -mibt.  Document __CET__.  Document
16827         -mcet as an alias for -mshstk.
16829 2018-04-20  Richard Biener <rguenther@suse.de>
16831         PR middle-end/85475
16832         * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
16833         complexity by forcing a single use of the multiply operand.
16835 2018-04-20  Martin Jambor  <mjambor@suse.cz>
16837         ipa/85449
16838         * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
16839         recursion dependency to only apply to non-clones.
16841 2018-04-20  Martin Jambor  <mjambor@suse.cz>
16843         ipa/85447
16844         * ipa-cp.c (create_specialized_node): Check that clones of
16845         self-recursive edges exist during IPA-CP.
16847 2018-04-19  Toon Moene  <toon@moene.org>
16849         * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
16850         by -O3.
16852 2018-04-19  Jakub Jelinek  <jakub@redhat.com>
16854         PR tree-optimization/85467
16855         * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
16856         VECTOR_TYPE_P macro.  If type is vector type, VIEW_CONVERT_EXPR the
16857         VECTOR_CST element to type.
16859 2018-04-19  H.J. Lu  <hongjiu.lu@intel.com>
16861         PR target/85397
16862         * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
16863         * config/i386/i386.md (builtin_setjmp_setup): Removed.
16864         (builtin_longjmp): Likewise.
16865         (save_stack_nonlocal): New pattern.
16866         (restore_stack_nonlocal): Likewise.
16868 2018-04-19  H.J. Lu  <hongjiu.lu@intel.com>
16870         PR target/85404
16871         * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
16872         Replace ASM_OUTPUT_LABEL with fprintf.
16874 2018-04-19  H.J. Lu  <hongjiu.lu@intel.com>
16876         PR target/85417
16877         * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
16878         Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
16879         * config/i386/i386-c.c (ix86_target_macros_internal): Also
16880         define __IBT__ and __SHSTK__ for -fcf-protection.
16881         * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
16882         TARGET_IBT.
16883         (ix86_trampoline_init): Likewise.
16884         (x86_output_mi_thunk): Likewise.
16885         (ix86_notrack_prefixed_insn_p): Likewise.
16886         (ix86_option_override_internal): Don't disallow -fcf-protection.
16887         * config/i386/i386.md (rdssp<mode>): Also enable for
16888         -fcf-protection.
16889         (incssp<mode>): Likewise.
16890         (nop_endbr): Likewise.
16891         * config/i386/i386.opt (mcet): Change help message to built-in
16892         functions only.
16893         (mibt): Likewise.
16894         (mshstk): Likewise.
16895         * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
16896         on -fcf-protection.  Change -mcet, -mibt and -mshstk to only
16897         enable CET built-in functions.
16899 2018-04-19  Sebastian Peryt  <sebastian.peryt@intel.com>
16901         * common/config/i386/i386-common.c
16902         (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
16903         OPTION_MASK_ISA_MOVDIRI_UNSET,
16904         OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
16905         (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
16906         * config.gcc (movdirintrin.h): New header.
16907         * config/i386/cpuid.h (bit_MOVDIRI,
16908         bit_MOVDIR64B): New bits.
16909         * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
16910         and -mmvodir64b.
16911         * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
16912         (VOID, PVOID, PCVOID)): New function types.
16913         * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
16914         __builtin_ia32_directstoreu_u64,
16915         __builtin_ia32_movdir64b): New builtins.
16916         * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
16917         * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
16918         and -mmovdiri.
16919         (ix86_valid_target_attribute_inner_p): Ditto.
16920         (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
16921         and VOID_FTYPE_PUNSIGNED_UNSIGNED.
16922         (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
16923         * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
16924         TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
16925         * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
16926         (movdiri<mode>, movdir64b_<mode>): New.
16927         * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
16928         * config/i386/immintrin.h: Include movdirintrin.h.
16929         * config/i386/movdirintrin.h: New file.
16930         * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
16932 2018-04-19  Richard Biener  <rguenther@suse.de>
16934         PR middle-end/85455
16935         * cfg.c (clear_bb_flags): When loop state says we have
16936         marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
16938 2018-04-19  Richard Biener  <rguenther@suse.de>
16940         PR tree-optimization/84737
16941         * tree-vect-data-refs.c (vect_copy_ref_info): New function
16942         copying restrict info.
16943         (vect_setup_realignment): Use it.
16944         * tree-vectorizer.h (vect_copy_ref_info): Declare.
16945         * tree-vect-stmts.c (vectorizable_store): Copy ref info from
16946         the first DR to all generated stores.
16947         (vectorizable_load): Likewise for loads.
16949 2018-04-19  Jakub Jelinek  <jakub@redhat.com>
16951         PR tree-optimization/85446
16952         * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
16953         the integral and pointer types to have the same precision.
16955         * doc/install.texi: Document --disable-cet being the default and
16956         --enable-cet=auto.
16958 2018-04-18  Martin Liska  <mliska@suse.cz>
16960         * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
16961         style.
16963 2018-04-18  Martin Liska  <mliska@suse.cz>
16965         Revert
16966         2018-03-02  Eric Botcazou  <ebotcazou@adacore.com>
16968         PR ipa/83983
16969         * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
16970         arguments if they are comparable.
16972 2018-04-18  Martin Liska  <mliska@suse.cz>
16974         Revert
16975         2018-03-13  Eric Botcazou  <ebotcazou@adacore.com>
16977         PR lto/84805
16978         * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
16979         incomplete types.
16981 2018-04-18  H.J. Lu  <hongjiu.lu@intel.com>
16983         PR target/85388
16984         * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
16985         ENDBR after calling __morestack.
16987 2018-04-18  David Malcolm  <dmalcolm@redhat.com>
16989         PR jit/85384
16990         * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
16991         by using gcc_base_ver to generate a gcc_driver_version, and use
16992         it when generating GCC_DRIVER_NAME.
16993         * configure: Regenerate.
16995 2018-04-18  Jakub Jelinek  <jakub@redhat.com>
16997         PR target/81084
16998         * config.gcc: Obsolete powerpc*-*-*spe*.
17000 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
17002         PR debug/84637
17003         * dbxout.c (dbxout_int): Perform negation in unsigned int type.
17004         (stabstr_D): Change type of unum from unsigned int to
17005         unsigned HOST_WIDE_INT.  Perform negation in unsigned HOST_WIDE_INT
17006         type.
17008 2018-04-17  Jim Wilson  <jimw@sifive.com>
17010         PR 84856
17011         * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
17012         RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
17013         Set arg_pointer_offset after using pretend_args_size.
17015 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
17017         PR rtl-optimization/85431
17018         * dse.c (record_store): Ignore zero width stores.
17020         PR sanitizer/85230
17021         * asan.c (handle_builtin_stack_restore): Adjust comment.  Emit
17022         __asan_allocas_unpoison call and last_alloca_addr = new_sp before
17023         __builtin_stack_restore rather than after it.
17024         * builtins.c (expand_asan_emit_allocas_unpoison): Pass
17025         arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
17026         argument instead of virtual_dynamic_stack_rtx.
17028 2018-04-17  Kelvin Nilsen  <kelvin@gcc.gnu.org>
17030         * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
17031         New prototype.
17032         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
17033         Add note to error message to explain internal mapping of overloaded
17034         built-in function name to non-overloaded built-in function name.
17035         * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
17036         function.
17038 2018-04-17  Michael Meissner  <meissner@linux.vnet.ibm.com>
17040         PR target/85424
17041         * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
17042         where the inputs overlap with the output.
17044 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
17046         * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
17047         (=v, v) alternative and explicit "memory" attribute.
17048         (vec_extract_lo_<mode><mask_name>): Likewise.  Also add
17049         "type", "prefix", "prefix_extra", "length_immediate" and "mode"
17050         attributes.
17051         (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
17052         "sselog1" type instead of "sselog".
17053         (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
17054         "sselog".  Remove explicit "memory" attribute.
17055         (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
17056         "type", "prefix", "prefix_extra", "length_immediate" and "mode"
17057         attributes.
17058         (vec_extract_hi_v32hi): Merge all alternatives into one, use
17059         "sselog1" type instead of "sselog".  Remove explicit "memory"
17060         attribute.
17061         (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
17062         use "sselog1" type instead of "sselog".  Remove explicit "memory"
17063         attribute.
17064         (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
17065         "type", "prefix", "prefix_extra", "length_immediate" and "mode"
17066         attributes.
17067         (vec_extract_hi_v64qi): Merge all alternatives into one, use
17068         "sselog1" type instead of "sselog".  Remove explicit "memory"
17069         attribute.
17070         (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
17071         use "sselog1" type instead of "sselog".  Remove explicit "memory"
17072         attribute.
17074         PR target/85430
17075         * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
17077         PR middle-end/85414
17078         * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
17079         on a SUBREG.
17081 2018-04-17  Martin Jambor  <mjambor@suse.cz>
17083         PR ipa/85421
17084         * ipa-cp.c (create_specialized_node): Call
17085         expand_all_artificial_thunks if necessary.
17087 2018-04-17  Martin Liska  <mliska@suse.cz>
17089         PR lto/85405
17090         * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
17091         in message, remote space in between '_G' and '('.
17093 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
17095         PR target/85281
17096         * config/i386/sse.md (reduces<mode><mask_scalar_name>,
17097         avx512f_vmcmp<mode>3<round_saeonly_name>,
17098         avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
17099         avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
17100         avx512f_rndscale<mode><round_saeonly_name>,
17101         avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
17102         avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
17103         Use %<iptr>2 instead of %2 for -masm=intel.
17104         (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
17105         avx512f_vcvttss2usi<round_saeonly_name>,
17106         avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
17107         -masm=intel.
17108         (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
17109         avx512f_vcvttsd2usi<round_saeonly_name>,
17110         avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
17111         Use %q1 instead of %1 for -masm=intel.
17112         (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
17113         avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
17114         of %3 for -masm=intel.
17115         (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
17116         -masm=intel.
17117         (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
17118         -masm=intel.
17119         (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
17120         -masm=intel.
17121         (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
17122         %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
17123         %g1.
17124         (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
17125         -masm=intel.
17126         (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
17127         %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
17128         %g1 and one with %0 and %1.
17129         (avx512er_vmrcp28<mode><round_saeonly_name>,
17130         avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
17131         %1 for -masm=intel.
17132         (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
17133         avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
17134         avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
17135         of %0 and %{%4%} for -masm=intel.
17136         (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
17137         avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
17138         avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
17139         order of %0 and %{%5%}%{z%} for -masm=intel.
17141 2018-04-17  Jan Hubicka  <jh@suse.cz>
17143         PR lto/85405
17144         * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
17146 2018-04-17  Martin Liska  <mliska@suse.cz>
17148         PR ipa/85329
17149         * multiple_target.c (create_dispatcher_calls): Set apostrophes
17150         for target_clone error message.  Make default implementation
17151         clone to be a local declaration.
17152         (separate_attrs): Add new argument and check for an empty
17153         string.
17154         (expand_target_clones): Handle it.
17155         (ipa_target_clone): Make redirection just for target_clones
17156         functions.
17158 2018-04-16  Cesar Philippidis  <cesar@codesourcery.com>
17159             Tom de Vries  <tom@codesourcery.com>
17161         PR middle-end/84955
17162         * omp-expand.c (expand_oacc_for): Add dummy false branch for
17163         tiled basic blocks without omp continue statements.
17165 2018-04-16  Aaron Sawdey  <acsawdey@linux.ibm.com>
17167         PR target/83660
17168         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
17169         vec_extract expression as having side effects to make sure it gets
17170         a cleanup point.
17172 2018-04-16  H.J. Lu  <hongjiu.lu@intel.com>
17174         PR target/85403
17175         * config/i386/i386.c (get_builtin_code_for_version): Check
17176         error_mark_node.
17178 2018-04-16  Olga Makhotina  <olga.makhotina@intel.com>
17180         PR target/84331
17181         * config.gcc: Support "skylake".
17182         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
17183         PROCESSOR_SKYLAKE.
17184         * config/i386/i386.c (m_SKYLAKE): Define.
17185         (processor_target_table): Add "skylake".
17186         (ix86_option_override_internal): Add "skylake".
17187         (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
17188         PROCESSOR_CANNONLAKE.
17189         (get_builtin_code_for_version): Fix priority for
17190         PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
17191         PROCESSOR_SKYLAKE-AVX512.
17192         * config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
17193         (processor_type): Add PROCESSOR_SKYLAKE.
17195 2018-04-16  Paolo Carlini  <paolo.carlini@oracle.com>
17196             Jason Merrill  <jason@redhat.com>
17198         PR c++/85112
17199         * convert.c (convert_to_integer_1): Use direct recursion for
17200         enumeral types and types with a precision less than the number
17201         of bits in their mode.
17203 2018-04-16  Julia Koval  <julia.koval@intel.com>
17205         PR target/84413
17206         * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
17207         X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
17209 2018-04-14  Segher Boessenkool  <segher@kernel.crashing.org>
17211         PR target/85293
17212         * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
17213         * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
17214         and -mno-direct-move.
17216 2018-04-13  Paul A. Clarke  <pc@us.ibm.com>
17218         PR target/83402
17219         * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
17220         Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
17221         Ensure negative shifts result in {0}.
17223 2018-04-13  Vladimir Makarov  <vmakarov@redhat.com>
17225         PR rtl-optimization/79916
17226         * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
17227         regs (if any) to define how to gnerate SD moves when LRA is in
17228         progress.
17230 2018-04-13  Jakub Jelinek  <jakub@redhat.com>
17232         PR rtl-optimization/85393
17233         * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
17234         * except.c (expand_dw2_landing_pad_for_region): Make static.
17235         * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
17236         a label and unconditional jump to old_bb, rather than
17237         expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
17238         basic block.
17240         PR rtl-optimization/85376
17241         * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
17242         zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
17243         instead of a specific value.
17245 2018-04-13  Jan Hubicka  <hubicka@ucw.cz>
17246             Bin Cheng  <bin.cheng@arm.com>
17248         PR tree-optimization/82965
17249         PR tree-optimization/83991
17250         * cfgloopanal.c (expected_loop_iterations_unbounded): Add
17251         by_profile_only parameter.
17252         * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
17253         information if the loop was predicted to iterate too many times.
17254         * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
17256 2018-04-13  Jan Hubicka  <hubicka@ucw.cz>
17258         PR lto/71991
17259         * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
17260         always inline.
17262 2018-04-13  Martin Liska  <mliska@suse.cz>
17263             Jakub Jelinek  <jakub@redhat.com>
17265         PR middle-end/81657
17266         * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
17267         * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
17268         * builtins.c (expand_builtin_memory_copy_args): Use
17269         BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
17270         handle dest_addr == pc_rtx.
17272 2018-04-12  Segher Boessenkool  <segher@kernel.crashing.org>
17274         PR target/85291
17275         * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
17276         asked to not generate direct moves.
17277         (fix_trunc<mode>si2_stfiwx): Similar.
17278         (fix_trunc<mode>si2_internal): Similar.
17280 2018-04-12  Jakub Jelinek  <jakub@redhat.com>
17282         PR debug/83157
17283         * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
17284         * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
17285         lookup if dest in some wider mode is known to be const0_rtx and
17286         if so, record permanent equivalence for it to be ZERO_EXTEND of
17287         the narrower mode destination.
17289 2018-04-12  Cesar Philippidis  <cesar@codesourcery.com>
17291         * lto-streamer-out.c (output_function): Revert 259346.
17292         * omp-expand.c (expand_oacc_for): Likewise.
17294 2018-04-12  Alexander Monakov  <amonakov@ispras.ru>
17296         PR rtl-optimization/85354
17297         * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
17298         * sel-sched.c (sel_global_init): ... here.
17300 2018-04-12  Eric Botcazou  <ebotcazou@adacore.com>
17302         PR target/85238
17303         * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
17304         * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
17305         mode for PE-COFF targets.
17306         * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
17307         (i386_pe_asm_lto_end): Likewise.
17308         * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
17309         (TARGET_ASM_LTO_END): Likewise.
17310         * config/i386/winnt.c (saved_debug_info_level): New static variable.
17311         (i386_pe_asm_lto_start): New function.
17312         (i386_pe_asm_lto_end): Likewise.
17314 2018-04-12  Cesar Philippidis  <cesar@codesourcery.com>
17315             Richard Biener  <rguenther@suse.de>
17317         PR middle-end/84955
17318         * lto-streamer-out.c (output_function): Fix CFG loop state before
17319         streaming out.
17320         * omp-expand.c (expand_oacc_for): Handle calls to internal
17321         functions like regular functions.
17323 2018-04-12  Richard Biener  <rguenther@suse.de>
17325         PR lto/85371
17326         * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
17327         for the early LTO debug to properly generate references to it
17328         during DIE emission.  Do not re-use that for the skeleton for
17329         split-dwarf.
17330         (dwarf2out_early_finish): Likewise.
17332 2018-04-12  Jakub Jelinek  <jakub@redhat.com>
17334         PR target/85328
17335         * config/i386/sse.md
17336         (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
17337         <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
17338         vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
17339         vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
17340         and output is a reg, avoid creating invalid lowpart subreg, but
17341         instead split into a 512-bit move.  Don't split if not AVX512VL,
17342         input is xmm16+ reg and output is a mem.
17343         (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
17344         vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
17345         xmm16+ reg and output is a mem.
17347 2018-04-12  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
17349         * config/s390/s390.c (s390_output_indirect_thunk_function): Check
17350         also for flag_dwarf2_cfi_asm.
17352 2018-04-12  Jakub Jelinek  <jakub@redhat.com>
17354         PR rtl-optimization/85342
17355         * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
17356         a bool scalar var inside of the loop instead.  Don't try to update
17357         recog_data.operand after failed apply_change_group.
17359 2018-04-12  Tom de Vries  <tom@codesourcery.com>
17361         PR target/85296
17362         * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
17363         (nvptx_assemble_decl_begin): Add undefined param.  Declare undefined
17364         array with flexible array member as array without given dimension.
17365         (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
17366         argument for undefined param to true.
17368 2018-04-11  Aaron Sawdey  <acsawdey@linux.ibm.com>
17370         PR target/85321
17371         * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
17372         -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
17373         from PowerPC section.
17374         * config/rs6000/sysv4.opt (mcall-): Improve help text.
17375         * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
17376         help text that is too long.
17377         * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
17378         help text that is too long.
17379         * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
17380         help text that is too long.
17382 2018-04-11  Uros Bizjak  <ubizjak@gmail.com>
17384         * config/alpha/alpha.md (stack_probe_internal): Rename
17385         from "probe_stack".  Update all callers.
17387 2018-04-11  Alexander Monakov  <amonakov@ispras.ru>
17389         PR rtl-optimization/84566
17390         * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
17391         sched_macro_fuse_insns.
17393 2018-04-11  Alexander Monakov  <amonakov@ispras.ru>
17395         PR target/84301
17396         * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
17397         (compute_block_dependences): ... from here.
17399 2018-04-11  Jakub Jelinek  <jakub@redhat.com>
17401         PR tree-optimization/85331
17402         * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
17403         from int to HOST_WIDE_INT.
17405 2018-04-11  Martin Jambor  <mjambor@suse.cz>
17407         PR ipa/84149
17408         * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
17409         (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
17410         not the same as the source val.
17411         (cgraph_edge_brings_value_p): New parameter.
17412         (gather_edges_for_value): Pass destination value to
17413         cgraph_edge_brings_value_p.
17414         (perhaps_add_new_callers): Likewise.
17415         (get_info_about_necessary_edges): Likewise and exclude values brought
17416         only by self-recursive edges.
17417         (create_specialized_node): Redirect only clones of self-calling edges.
17418         (+self_recursive_pass_through_p): New function.
17419         (find_more_scalar_values_for_callers_subset): Use it.
17420         (find_aggregate_values_for_callers_subset): Likewise.
17421         (known_aggs_to_agg_replacement_list): Removed.
17422         (decide_whether_version_node): Re-calculate known constants for all
17423         remaining context clones.
17425 2018-04-11  Richard Biener  <rguenther@suse.de>
17427         PR lto/85339
17428         * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
17429         from early DWARF output.
17430         (dwarf2out_early_finish): Output line info unconditionally into
17431         early DWARF and add reference to it.
17433 2018-04-11  Jakub Jelinek  <jakub@redhat.com>
17435         PR target/85281
17436         * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
17437         (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
17438         other than V2DFmode using iptr mode attribute.
17439         (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
17441 2018-04-11  Alexander Monakov  <amonakov@ispras.ru>
17443         PR rtl-optimization/84659
17444         * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
17446 2018-04-11  Jakub Jelinek  <jakub@redhat.com>
17448         PR debug/85302
17449         * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
17450         SIZEP is NULL.
17451         (output_loc_list): Pass address of a dummy size variable even in the
17452         locview handling loop.
17453         (index_location_lists): Add comment on why skip_loc_list_entry can't
17454         call size_of_locs.
17456 2018-04-11  Thomas Preud'homme  <thomas.preudhomme@arm.com>
17458         PR target/85261
17459         * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
17460         into register.
17462 2018-04-10  Aaron Sawdey  <acsawdey@linux.ibm.com>
17464         PR target/85321
17465         * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
17466         -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
17467         and -mstring-compare-inline-limit.
17469 2018-04-10  Segher Boessenkool  <segher@kernel.crashing.org>
17471         PR target/85287
17472         * config/rs6000/rs6000.md (allocate_stack): Put the residual size
17473         for stack clash protection in a register whenever we need it to be in
17474         a register.
17476 2018-04-10  Segher Boessenkool  <segher@kernel.crashing.org>
17478         * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
17479         Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
17481 2018-04-10  Segher Boessenkool  <segher@kernel.crashing.org>
17483         PR target/85321
17484         * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
17485         the help text.
17486         (mlong-double-): Ditto.
17487         * config/rs6000/sysv4.opt (msdata=): Ditto.
17488         (mtls-size=): Ditto.
17490 2018-04-10  Kelvin Nilsen  <kelvin@gcc.gnu.org>
17492         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
17493         erroneous entries for
17494         "vector int vec_ldl (int, long int *)", and
17495         "vector unsigned int vec_ldl (int, unsigned long int *)".
17496         Add comments and entries for
17497         "vector bool char vec_ldl (int, bool char *)",
17498         "vector bool short vec_ldl (int, bool short *)",
17499         "vector bool int vec_ldl (int, bool int *)",
17500         "vector bool long long vec_ldl (int, bool long long *)",
17501         "vector pixel vec_ldl (int, pixel *)",
17502         "vector long long vec_ldl (int, long long *)",
17503         "vector unsigned long long vec_ldl (int, unsigned long long *)".
17504         * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
17505         type tree bool_long_long_type_node and correct definition of
17506         bool_V2DI_type_node to make reference to this new type tree.
17507         (rs6000_mangle_type): Replace erroneous reference to
17508         bool_long_type_node with bool_long_long_type_node.
17509         * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
17510         comments to emphasize sign distinctions for char and int types and
17511         replace RS6000_BTI_bool_long constant with
17512         RS6000_BTI_bool_long_long constant.  Also add comment to restrict
17513         use of RS6000_BTI_pixel.
17514         (bool_long_type_node): Remove this macro definition.
17515         (bool_long_long_type_node): New macro definition
17517 2018-04-10  Jakub Jelinek  <jakub@redhat.com>
17519         PR rtl-optimization/85300
17520         * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
17521         into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
17522         simplify_unary_operation fails.
17524 2018-04-10  Martin Liska  <mliska@suse.cz>
17526         * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
17527         cgraph_edge and ipa_ref.
17529 2018-04-10  Jakub Jelinek  <jakub@redhat.com>
17531         PR target/85177
17532         PR target/85255
17533         * config/i386/sse.md
17534         (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
17535         computation of the VEC_MERGE selector from mask.
17536         (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
17537         Fix decoding of the VEC_MERGE selector into mask.
17539 2018-04-10  Richard Sandiford  <richard.sandiford@linaro.org>
17541         PR tree-optimization/85286
17542         * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
17544 2018-04-10  Richard Sandiford  <richard.sandiford@linaro.org>
17546         * final.c (final_1): Set insn_last_address as well as
17547         insn_current_address.
17549 2018-04-10  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
17551         PR target/85173
17552         * explow.c (emit_stack_probe): Call validize_mem on memory location
17553         before passing it to gen_probe_stack.  Create address operand and
17554         legitimize it for the probe_stack_address case.
17556 2018-04-09  Jan Hubicka  <jh@suse.cz>
17558         PR lto/85078
17559         * ipa-devirt.c (rebuild_type_inheritance-hash): New.
17560         * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
17561         * tree.c (free_lang_data_in_type): Fix handling of binfos;
17562         walk basetypes.
17563         (free_lang_data): Rebuild type inheritance graph.
17565 2018-04-09  Martin Sebor  <msebor@redhat.com>
17567         * invoke.texi (-finline-small-functions): Mention other optimization
17568         options.
17569         (-findirect-inlining, -fpartial-inlining): Same.
17570         (-finline-functions-called-once): Same.
17571         (-freorder-blocks-and-partition): Same.
17573 2018-04-09  Jan Hubicka  <jh@suse.cz>
17575         PR rtl/84058
17576         * cfgcleanup.c (try_forward_edges): Do not give up on crossing
17577         jumps; choose last target that matches the criteria (i.e.
17578         no partition changes for non-crossing jumps).
17579         * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
17580         support for redirecting crossing jumps to non-crossing.
17582 2018-04-09  Alexey Brodkin  <abrodkin@synopsys.com>
17584         * config/arc/arc.c (arc_expand_prologue): Set stack usage info
17585         also for naked functions.
17587 2018-04-09  Claudiu Zissulescu  <claziss@synopsys.com>
17589         * config/arc/arc.md (add_shift): New pattern.
17590         (add_shift2): Likewise.
17591         (sub_shift): Likewise.
17592         (sub_shift_cmp0_noout): Likewise.
17593         (compare_si_ashiftsi): Likewise.
17594         (xbfu_cmp0_noout): New combine pattern.
17595         (xbfu_cmp0"): Likewise.
17596         (movsi_set_cc_insn): Place the predicable variant first.
17597         (commutative_binary_cmp0_noout): Remove clobber.
17598         (commutative_binary_cmp0): New pattern.
17599         (noncommutative_binary_cmp0): Likewise.
17600         (noncommutative_binary_cmp0_noout): Likewise.
17601         (noncommutative_binary_comparison_result_used): Removed.
17602         (rsub_cmp0): New pattern.
17603         (rsub_cmp0_noout): Likewise.
17604         (extzvsi): Changed, keep only meaningful variants.
17605         (SQH, SEZ): New iterators.
17606         (SQH_postfix): New mode attribute.
17607         (SEZ_prefix): New code attribute.
17608         (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
17609         (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
17610         * config/arc/predicates.md (cc_set_register): Use CC_REG instead
17611         of numerical value.
17612         (noncommutative_operator): Check the availability of barrel
17613         shifter option.
17615 2018-04-09  Richard Biener  <rguenther@suse.de>
17617         PR tree-optimization/85284
17618         * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
17619         Only use the niter constraining form of simple_iv when the exit
17620         is always executed.
17622 2018-04-09  Tom de Vries  <tom@codesourcery.com>
17624         PR target/84041
17625         * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
17626         (define_expand "*memory_barrier"): New define_expand.
17627         (define_insn "memory_barrier"): New insn.
17629 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
17631         PR rtl-optimization/80463
17632         PR rtl-optimization/83972
17633         PR rtl-optimization/83480
17635         * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
17636         correct producer for the insn.
17637         (tidy_control_flow): Fixup seqnos in case of debug insns.
17639 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
17641         PR rtl-optimization/83913
17643         * sel-sched-ir.c (merge_expr_data): Choose the middle between two
17644         different sched-times when merging exprs.
17646 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
17648         PR rtl-optimization/83962
17650         * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
17651         tidy_fallthru_edge and tidy_control_flow.
17653 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
17655         PR rtl-optimization/83530
17657         * sel-sched.c (force_next_insn): New global variable.
17658         (remove_insn_for_debug): When force_next_insn is true, also leave only
17659         next insn in the ready list.
17660         (sel_sched_region): When the region wasn't scheduled, make another pass
17661         over it with force_next_insn set to 1.
17663 2018-04-08  Monk Chiang  <sh.chiang04@gmail.com>
17665         * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
17666         into tm_file.
17667         * config/nds32/constants.md (unspec_volatile_element): Add enum values
17668         for interrupt control.
17669         * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
17670         functions for interrupt control.
17671         * config/nds32/nds32-intrinsic.md: Likewise.
17672         * config/nds32/nds32_intrinsic.h: Likewise.
17673         * config/nds32/nds32.h (nds32_builtins): Likewise.
17675 2018-04-08  Chung-Ju Wu  <jasonwucj@gmail.com>
17677         * config/nds32/nds32.c (nds32_init_machine_status,
17678         nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
17679         strict_aligned_p field.
17680         (nds32_expand_to_rtl_hook): New function.
17681         (TARGET_EXPAND_TO_RTL_HOOK): Define.
17682         * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
17684 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
17685             Chung-Ju Wu  <jasonwucj@gmail.com>
17687         * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
17688         * config/nds32/nds32-n7.md: New file.
17689         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
17690         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
17691         pipeline.
17692         * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
17693         * config/nds32/nds32.md (pipeline_model): Add n7.
17694         * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
17695         * config/nds32/pipelines.md: Include n7 settings.
17697 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
17698             Chung-Ju Wu  <jasonwucj@gmail.com>
17700         * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
17701         * config/nds32/nds32-e8.md: New file.
17702         * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
17703         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
17704         pipeline.
17705         * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
17706         * config/nds32/nds32.md (pipeline_model): Add e8.
17707         * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
17708         * config/nds32/pipelines.md: Include e8 settings.
17710 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
17711             Chung-Ju Wu  <jasonwucj@gmail.com>
17713         * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
17714         * config/nds32/nds32-n8.md: New file.
17715         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
17716         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
17717         pipeline.
17718         * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
17719         * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
17720         * config/nds32/nds32.md (pipeline_model): Add n8.
17721         * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
17722         * config/nds32/pipelines.md: Include n8 settings.
17724 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
17725             Chung-Ju Wu  <jasonwucj@gmail.com>
17727         * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
17728         * config/nds32/nds32-n9-2r1w.md: New file.
17729         * config/nds32/nds32-n9-3r2w.md: New file.
17730         * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
17731         nds32_register_ports): New or modify for cpu n9.
17732         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
17733         pipeline.
17734         * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
17735         * config/nds32/nds32-utils.c: New file.
17736         * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
17737         TARGET_MUL_SLOW): Define.
17738         * config/nds32/nds32.md (pipeline_model): New attribute.
17739         * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
17740         New options that support cpu n9.
17741         * config/nds32/pipelines.md: Include n9 settings.
17742         * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
17744 2018-04-08  Chung-Ju Wu  <jasonwucj@gmail.com>
17746         * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
17747         information if necessary.
17748         (output_cond_branch_compare_zero): Likewise.
17749         * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
17750         (nds32_target_alignment): Refine for alignment.
17751         * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
17752         (FUNCTION_BOUNDARY): Modify.
17753         * config/nds32/nds32.md (call_internal, call_value_internal): Consider
17754         align case.
17755         * config/nds32/nds32.opt (malways-align, malign-functions): New.
17757 2018-04-08  Monk Chiang  <sh.chiang04@gmail.com>
17759         * config/nds32/constants.md (unspec_volatile_element): Add values for
17760         TLB operation and data prefetch.
17761         * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
17762         functions for TLB operation and data prefetch.
17763         * config/nds32/nds32-intrinsic.md: Likewise.
17764         * config/nds32/nds32_intrinsic.h: Likewise.
17765         * config/nds32/nds32.c (nds32_dpref_names): Likewise.
17766         (nds32_print_operand): Likewise.
17767         * config/nds32/nds32.h (nds32_builtins): Likewise.
17769 2018-04-07  Thomas Koenig  <tkoenig@gcc.gnu.org>
17770         Andrew Pinski <pinsika@gcc.gnu.org>
17772         PR middle-end/82976
17773         * match.pd: Use constant_boolean_node of correct type instead of
17774         boolean_true_node or boolean_false_node for simplifying
17775         pointer comparisons to zero.
17777 2018-04-07  Jakub Jelinek  <jakub@redhat.com>
17779         PR tree-optimization/80021
17780         * tree.c (verify_type_variant): Make error call in verify_variant_match
17781         translatable and remove final full stop.
17783 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
17785         * config/nds32/constants.md (unspec_volatile_element): Add
17786         UNSPEC_VOLATILE_EH_RETURN.
17787         * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
17788         nds32_output_stack_pop): Support dwarf exception handling process.
17789         * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
17790         * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
17791         exception handling process.
17792         (nds32_compute_stack_frame): Likewise.
17793         (nds32_return_addr_rtx): Likewise.
17794         (nds32_initial_elimination_offset): Likewise.
17795         (nds32_expand_prologue): Likewise.
17796         (nds32_expand_epilogue): Likewise.
17797         (nds32_dynamic_chain_address): New function.
17798         * config/nds32/nds32.h (machine_function): Add fields for dwarf
17799         exception handling.
17800         (DYNAMIC_CHAIN_ADDRESS): Define.
17801         (EH_RETURN_DATA_REGNO): Define.
17802         (EH_RETURN_STACKADJ_RTX): Define.
17803         * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
17804         patterns for dwarf exception handling.
17806 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
17808         * config/nds32/nds32.h: Clean up obsolete macros.
17810 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
17812         * config/nds32/constants.md (unspec_element, unspec_volatile_element):
17813         Add enum values for particular instructions.
17814         * config/nds32/nds32-intrinsic.c: Implementation of expanding
17815         particular intrinsic functions.
17816         * config/nds32/nds32-intrinsic.md: Likewise.
17817         * config/nds32/nds32_intrinsic.h: Likewise.
17818         * config/nds32/nds32.h (nds32_builtins): Likewise.
17819         * config/nds32/nds32.md (type): Add pbsad and pbsada.
17820         (btst, ave): New patterns for particular instructions.
17822 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
17824         * config/nds32/constants.md (unspec_element, unspec_volatile_element):
17825         Add enum values for atomic load/store and memory sync.
17826         * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
17827         and memory sync.
17828         * config/nds32/nds32-intrinsic.md: Likewise.
17829         * config/nds32/nds32_intrinsic.h: Likewise.
17830         * config/nds32/nds32.h (nds32_builtins): Likewise.
17832 2018-04-07  Jakub Jelinek  <jakub@redhat.com>
17834         PR tree-optimization/85257
17835         * fold-const.c (native_encode_vector): If not all elts could fit
17836         and off is -1, return 0 rather than offset.
17837         * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
17838         (offseti - offset2) / BITS_PER_UNIT as 4th argument to
17839         native_encode_expr.  Verify len * BITS_PER_UNIT >= maxsizei.  Don't
17840         adjust buffer in native_interpret_expr call.
17842 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
17844         * config/nds32/constants.md (unspec_volatile_element): Add cache
17845         control enum values.
17846         * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
17847         * config/nds32/nds32-intrinsic.md: Add cache control patterns.
17848         * config/nds32/nds32.c (nds32_cctl_names): New.
17849         (nds32_print_operand): Handle cache control register names.
17850         * config/nds32/nds32.h (nds32_builtins): New enum values.
17851         * config/nds32/nds32_intrinsic.h: Add cache control enum types and
17852         macros.
17853         * config/nds32/nds32.md (type): Add mmu.
17854         * config/nds32/pipelines.md (simple_insn): Add mmu.
17856 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
17858         * config/nds32/nds32.md (type): Remove call.
17859         * config/nds32/pipelines.md (simple_insn): Likewise.
17861 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
17863         * config/nds32/constants.md (unspec_volatile_element): Add
17864         UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
17865         UNSPEC_VOLATILE_FMFCFG.
17866         * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
17867         description for fmfcfg and fmfcsr.
17868         (bdesc_1arg): Add fmtcsr.
17869         (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
17870         (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
17871         * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
17872         unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
17873         unspec_fmfcfg): New patterns.
17874         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
17875         NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
17876         NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
17877         * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
17878         __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
17879         __nds32__fmfcfg): Define.
17881 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
17883         * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
17884         intrinsic register names.
17885         * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
17886         intrinsic register enum values and macros.
17888 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
17890         * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
17891         for load/store addressing form.
17892         (nds32_print_operand_address): Likewise.
17894 2018-04-06  Eric Botcazou  <ebotcazou@adacore.com>
17896         PR target/85196
17897         * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
17898         based on LABEL_REF.  Remove useless assertion.
17899         (pic_address_needs_scratch): Fix formatting.
17900         (sparc_legitimize_pic_address): Minor tweaks.
17901         (sparc_delegitimize_address): Adjust assertion accordingly.
17902         * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
17903         into symbolic_operand.
17904         (movsi_high_pic_label_ref): Likewise.
17905         (movsi_lo_sum_pic_label_ref): Likewise.
17906         (movdi_pic_label_ref): Likewise.
17907         (movdi_high_pic_label_ref): Likewise.
17908         (movdi_lo_sum_pic_label_ref): Likewise.
17910 2018-04-06  Amaan Cheval  <amaan.cheval@gmail.com>
17912         * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
17913         custom LIB_SPEC setup.
17915 2018-04-06  Ruslan Bukin  <br@bsdpad.com>
17916             Kito Cheng  <kito.cheng@gmail.com>
17918         * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
17919         * config/riscv/freebsd.h: New.
17921 2018-04-06  Chung-Ju Wu  <jasonwucj@gmail.com>
17923         * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
17924         * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
17925         file.
17927 2018-04-06  Chung-Ju Wu  <jasonwucj@gmail.com>
17928             Kito Cheng  <kito.cheng@gmail.com>
17930         * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
17931         nds32_output_call, nds32_symbol_binds_local_p): New functions.
17932         * config/nds32/nds32-protos.h (nds32_output_call,
17933         nds32_output_return): Declare.
17934         * config/nds32/nds32.md: Refine all the call and return patterns.
17936 2018-04-06  Jakub Jelinek  <jakub@redhat.com>
17938         PR debug/85252
17939         * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
17940         build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
17942         PR rtl-optimization/84872
17943         * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
17944         nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
17945         EDGE_CROSSING edge.
17947 2018-04-06  Tamar Christina  <tamar.christina@arm.com>
17949         * expr.c (copy_blkmode_to_reg): Revert 254862.
17950         * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
17952 2018-04-06  Richard Biener  <rguenther@suse.de>
17954         PR middle-end/85244
17955         * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
17956         after seeing a component reference with an adjacent field.  Treat
17957         refs to arrays at struct end of external decls similar to
17958         refs to unconstrained commons.
17960 2018-04-06  Jakub Jelinek  <jakub@redhat.com>
17962         PR sanitizer/85213
17963         * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
17964         look through SAVE_EXPRs with non-side-effects argument.  Adjust
17965         recursive calls.
17966         (fold_comparison): Adjust twoval_comparison_p caller, don't handle
17967         save_p here.
17969 2018-04-06  Richard Biener  <rguenther@suse.de>
17971         PR middle-end/85180
17972         * alias.c (find_base_term): New wrapper around find_base_term
17973         unwinding CSELIB_VAL_PTR changes.
17974         (find_base_term): Do not restore CSELIB_VAL_PTR during the
17975         recursion.
17977 2018-04-06  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
17979         * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
17980         instructions.
17981         * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
17982         constant definitions.
17983         ("nop"): lr 0,0 -> nopr r0
17984         ("nop_lr0", "nop_lr1"): New insn definitions.
17986 2018-04-06  Chung-Ju Wu  <jasonwucj@gmail.com>
17988         * config/nds32/nds32.md (*stack_push, *stack_pop): Use
17989         NDS32_V3PUSH_AVAILABLE_P macro.
17991 2018-04-06  Monk Chiang  <sh.chiang04@gmail.com>
17992             Chung-Ju Wu  <jasonwucj@gmail.com>
17994         * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
17995         (nds32*-*-*): Add float and fpu_config into supported_defaults.
17996         * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
17997         Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
17998         * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
17999         UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
18000         * config/nds32/constraints.md: New constraints and checking for hard
18001         float configuration.
18002         * config/nds32/iterators.md: New mode iterator and attribute for hard
18003         float configuration.
18004         * config/nds32/nds32-doubleword.md: Use hard float alternatives and
18005         patterns.
18006         * config/nds32/nds32-fpu.md: New file.
18007         * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
18008         deal with hard float code generation.
18009         * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
18010         ARCH_V3S.
18011         (abi_type, float_reg_number): New enum type.
18012         * config/nds32/nds32-predicates.c: New predicates for hard float.
18013         * config/nds32/nds32-protos.h: Declare functions for hard float.
18014         * config/nds32/nds32.c: Implementation for hard float configuration.
18015         * config/nds32/nds32.h: Definitions for hard float configuration.
18016         * config/nds32/nds32.md: Include hard float machine description and
18017         modify patterns for hard float configuration.
18018         * config/nds32/nds32.opt: New options for hard float configuration.
18019         * config/nds32/predicates.md: New predicates for hard float
18020         configuration.
18022 2018-04-06  Kuan-Lin Chen  <kuanlinchentw@gmail.com>
18024         * common/config/nds32/nds32-common.c
18025         (nds32_option_optimization_table): Enable -mreleax-hint by default.
18027 2018-04-05  Jakub Jelinek  <jakub@redhat.com>
18029         PR middle-end/85195
18030         * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
18031         CONSTRUCTOR_ELT (ctor, ...)->value.
18033 2018-04-05  Uros Bizjak  <ubizjak@gmail.com>
18035         PR target/85193
18036         * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
18038 2018-04-05  Tom de Vries  <tom@codesourcery.com>
18040         PR target/85204
18041         * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
18042         cond jump.
18044 2018-04-05  Shiva Chen  <shiva0217@gmail.com>
18045             Kito Cheng  <kito.cheng@gmail.com>
18047         * config/nds32/constraints.md (U33): Fine-tune checking condition.
18048         * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
18049         * config/nds32/nds32.h (nds32_16bit_address_type): Add
18050         ADDRESS_POST_MODIFY_LO_REG_IMM3U.
18052 2018-04-05  Shiva Chen  <shiva0217@gmail.com>
18053             Kito Cheng  <kito.cheng@gmail.com>
18055         * config/nds32/constraints.md (Ufe): New memory constraint.
18056         * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
18057         nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
18058         * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
18059         operands.
18060         * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
18061         * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
18063 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
18065         * config/nds32/nds32.md: Use optimize_size in the condition for
18066         alu-shift instructions.
18068 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
18070         * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
18072 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
18074         * config/nds32/nds32.md (negsi2): Refine pattern.
18076 2018-04-05  Kito Cheng  <kito.cheng@gmail.com>
18077             Chung-Ju Wu  <jasonwucj@gmail.com>
18079         * config/nds32/iterators.md (shift_rotate): New code iterator.
18080         (shift): New code attribute.
18081         * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
18082         * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
18083         * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
18084         * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
18085         bit-wise operations.
18086         (andsi3, *andsi3): Ditto.
18087         (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
18088         (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
18089         (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
18090         * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
18091         nds32_ior_operand, nds32_xor_operand): New predicates.
18093 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
18095         * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
18096         (addsi3, subsi3): ... this.
18098 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
18100         * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
18102 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
18104         * config/nds32/nds32.md: Adjust indention.
18106 2018-04-05  Kito Cheng  <kito.cheng@gmail.com>
18108         * config/nds32/nds32.md (feature): New attribute.
18110 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
18112         * config/nds32/nds32.md (subtype): New attribute.
18114 2018-04-04  Thomas Preud'homme  <thomas.preudhomme@arm.com>
18116         PR target/85203
18117         * config/arm/arm-builtins.c (arm_expand_builtin): Change
18118         expansion to perform a bitwise AND of the argument followed by a
18119         boolean negation of the result.
18121 2018-04-04  Peter Bergner  <bergner@vnet.ibm.com>
18123         PR rtl-optimization/84878
18124         * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
18125         the basic block.  Assert the use reference is not artificial and that
18126         it has an associated insn.
18128 2018-04-04  Michael Matz  <matz@suse.de>
18130         * builtins.c (compute_objsize): Pass correct operand
18131         to array_at_struct_end_p.
18133 2018-04-04  Richard Biener  <rguenther@suse.de>
18135         PR lto/85176
18136         * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
18137         from contexts for DINFO_LEVEL_TERSE and below.
18139 2018-04-04  Kito Cheng  <kito.cheng@gmail.com>
18141         * config/nds32/nds32-doubleword.md (move_<mode>): Require
18142         resiter_operand condition.
18143         * config/nds32/nds32.md (*move<mode>): Ditto.
18145 2018-04-04  Kito Cheng  <kito.cheng@gmail.com>
18146             Monk Chiang  <sh.chiang04@gmail.com>
18148         * config/nds32/nds32.md (movmisalign<mode>): New pattern.
18150 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
18152         * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
18154 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
18155             Kito Cheng  <kito.cheng@gmail.com>
18157         * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
18158         nds32_cond_code_str, output_cond_branch,
18159         output_cond_branch_compare_zero, nds32_expand_cbranch,
18160         nds32_expand_cstore, nds32_expand_movcc,
18161         nds32_output_cbranchsi4_equality_zero,
18162         nds32_output_cbranchsi4_equality_reg,
18163         nds32_output_cbranchsi4_equality_reg_or_const_int,
18164         nds32_output_cbranchsi4_greater_less_zero: New functions.
18165         * config/nds32/nds32-protos.h (nds32_expand_cbranch,
18166         nds32_expand_cstore, nds32_expand_movcc,
18167         nds32_output_cbranchsi4_equality_zero,
18168         nds32_output_cbranchsi4_equality_reg,
18169         nds32_output_cbranchsi4_equality_reg_or_const_int,
18170         nds32_output_cbranchsi4_greater_less_zero): Declare.
18171         * config/nds32/predicates.md (nds32_movecc_comparison_operator,
18172         nds32_rimm11s_operand): New predicates.
18173         * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
18174         * config/nds32/nds32.md: Rewrite all the branch and conditional move
18175         patterns.
18177 2018-04-04  Kito Cheng  <kito.cheng@gmail.com>
18179         * config/nds32/nds32-doubleword.md: Refine all the instruction type.
18180         * config/nds32/nds32.md: Ditto.
18181         * config/nds32/pipelines.md: Ditto.
18183 2018-04-04  Richard Biener  <rguenther@suse.de>
18185         PR tree-optimization/85168
18186         * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
18187         propagating abnormals.
18189 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
18191         * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
18193 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
18194             Kito Cheng  <kito.cheng@gmail.com>
18196         * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
18197         * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
18198         * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
18199         (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
18200         * config/nds32/nds32.md (sibcall_internal): New.
18201         (sibcall_register): Remove.
18202         (sibcall_immediate): Remove.
18203         (sibcall_value_internal): New.
18204         (sibcall_value_register): Remove.
18205         (sibcall_value_immediate): Remove.
18206         * config/nds32/predicates.md (nds32_general_register_operand): New.
18207         (nds32_call_address_operand): New.
18209 2018-04-03  Jakub Jelinek  <jakub@redhat.com>
18211         PR rtl-optimization/85167
18212         * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
18213         bb_defs if *split_p, instead preinitialize it to NULL.
18215         PR tree-optimization/85156
18216         * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
18217         evaluating the argument multiple times.
18219 2018-04-03  Bill Schmidt  <wschmidt@linux.ibm.com>
18221         * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
18222         than vector.
18223         (_mm_cvtpd_ps): Likewise.
18224         (_mm_cvttpd_epi32): Likewise.
18225         * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
18226         * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
18227         vector, pixel, and bool following altivec.h include.
18229 2018-04-03  Martin Sebor  <msebor@redhat.com>
18231         * doc/extend.texi (Common Function Attributes): Clarify.
18232         (const attribute): Likewise.
18233         (pure attribute): Likewise.
18235 2018-04-03  Jakub Jelinek  <jakub@redhat.com>
18237         PR target/85169
18238         * config/i386/i386.c (ix86_expand_vector_set): Use
18239         HOST_WIDE_INT_1U << elt instead of 1 << elt.  Formatting fix.
18241 2018-04-03  Uros Bizjak  <ubizjak@gmail.com>
18243         * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
18244         instructions when changing rounding bits to preserve precision bits
18245         in the x87 control word.
18247 2018-04-03  Martin Liska  <mliska@suse.cz>
18249         PR tree-optimization/82491
18250         * rtl.h (strip_offset_and_add): Replace += suboffset with
18251         poly_uint64 () + suboffset.
18253 2018-03-29  Martin Liska  <mliska@suse.cz>
18254             Martin Jambor  <mjambor@suse.cz>
18256         PR ipa/84947
18257         * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
18258         param_type is not an integral or pointer type.
18260 2018-04-03  Richard Biener  <rguenther@suse.de>
18262         * sese.h (recompute_all_dominators): Remove.
18264 2018-04-02  Martin Sebor  <msebor@redhat.com>
18266         * doc/invoke.texi (-Wrestrict): Fix typos.
18268 2018-04-02  Jim Wilson  <jimw@sifive.com>
18270         * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
18271         * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
18272         (<optab>di3, <optab>si3_extend): Likewise.
18273         (<optab>si3_mask, <optab>si3_mask_1): New.
18274         (<optab>di3_mask, <optab>di3_mask_1): New.
18275         (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
18276         (lshrsi3_zero_extend_1): Use VOIDmode shift count.
18277         * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
18279 2018-04-02  Gerald Pfeifer  <gerald@pfeifer.com>
18281         * doc/cpp.texi (Variadic Macros): Fix line continuation in an
18282         example.
18284 2018-04-02  Chung-Ju Wu  <jasonwucj@gmail.com>
18286         * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
18287         (nds32_canonicalize_comparison): New function.
18289 2018-04-01  Chung-Ju Wu  <jasonwucj@gmail.com>
18290             Kito Cheng  <kito.cheng@gmail.com>
18291             Kuan-Lin Chen  <kuanlinchentw@gmail.com>
18293         * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
18294         * config/nds32/constants.md (unspec_volatile_element): Add
18295         UNSPEC_VOLATILE_RELAX_GROUP.
18296         * config/nds32/nds32-relax-opt.c: New file.
18297         * config/nds32/nds32-predicates.c
18298         (nds32_symbol_load_store_p): New function.
18299         * config/nds32/nds32-protos.h
18300         (nds32_symbol_load_store_p): Declare function.
18301         (make_pass_nds32_relax_opt): Declare new rtl pass function.
18302         * config/nds32/nds32.c
18303         (nds32_register_pass): New function to register pass.
18304         (nds32_register_passes): New function to register passes.
18305         * config/nds32/nds32.md (relax_group): New pattern.
18306         * config/nds32/nds32.opt (mrelax-hint): New option.
18307         * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
18309 2018-04-01  Kito Cheng  <kito.cheng@gmail.com>
18311         * config/nds32/t-nds32: Modify files dependency.
18313 2018-04-01  Chung-Ju Wu  <jasonwucj@gmail.com>
18315         * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
18316         (PROFILE_HOOK): Define its implementation.
18318 2018-04-01  Chung-Ju Wu  <jasonwucj@gmail.com>
18320         * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
18321         type and 32-bit size.
18323 2018-04-01  Jakub Jelinek  <jakub@redhat.com>
18325         PR middle-end/85090
18326         * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
18327         (V_128_256): New mode iterator.
18328         (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
18329         (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
18330         (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
18331         of V.
18332         * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
18333         V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
18335 2018-03-31  Segher Boessenkool  <segher@kernel.crashing.org>
18337         PR target/83315
18338         * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
18339         NaN inputs correctly.
18341 2018-03-30  Peter Bergner  <bergner@vnet.ibm.com>
18343         PR target/80546
18344         * config/rs6000/vsx.md (??r): New mode attribute.
18345         (*vsx_mov<mode>_64bit): Use it.
18346         (*vsx_mov<mode>_32bit): Likewise.
18348 2018-03-30  Martin Sebor  <msebor@redhat.com>
18350         PR tree-optimization/84818
18351         * builtins.c (check_access): Use warning_n.
18353 2018-03-30  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
18355         PR target/83822
18356         * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
18357         condition.
18358         * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
18359         condition.
18361 2018-03-30  Julia Koval  <julia.koval@intel.com>
18363         PR target/84413
18364         * x86-tune.def (movx, partial_reg_dependency): Enable for
18365         m_SKYLAKE_AVX512.
18367 2018-03-29  Vladimir Makarov  <vmakarov@redhat.com>
18369         PR inline-asm/84985
18370         * lra-constraints.c (process_alt_operands): Move setting
18371         this_alternative_matches below.
18373 2018-03-29  Martin Liska  <mliska@suse.cz>
18375         PR lto/84995.
18376         * doc/invoke.texi: Document how LTO works with debug info.
18377         Describe auto-load support of binutils.  Mention 'x86-64'
18378         as valid option value of -march option.
18380 2018-03-29  Jakub Jelinek  <jakub@redhat.com>
18382         * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
18384         PR c/85094
18385         * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
18386         For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
18387         OEP_NO_HASH_CHECK for recursive call, to avoid exponential
18388         checking.
18390 2018-03-28  Peter Bergner  <bergner@vnet.ibm.com>
18392         PR target/84912
18393         * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
18394         (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
18395         * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
18396         for RS6000_BTM_POWERPC64.
18397         (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
18398         (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
18399         * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
18400         definition.
18401         (DIVDE): Use it.
18402         (DIVDEU): Likewise.
18404 2018-03-28  Carl Love  <cel@us.ibm.com>
18406         Revert
18407         2017-09-27  Carl Love  <cel@us.ibm.com>
18409         * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
18410         (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
18411         * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
18412         fctiw instruction.
18414 2018-03-28  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
18416         * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
18417         instead of __vector bool.
18418         (_mm_max_pu8): Likewise.
18419         (_mm_min_pi16): Likewise.
18421 2018-03-28  Peter Bergner  <bergner@vnet.ibm.com>
18423         PR target/84912
18424         * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
18425         (DIVWEUO): Likewise.
18426         (DIVDEO): Likewise.
18427         (DIVDEUO): Likewise.
18428         * config/rs6000/rs6000.c (builtin_function_type): Remove support for
18429         DIVWEUO and DIVDEUO.
18430         * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
18431         (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
18432         (div_extend): Likewise.
18433         * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
18434         builtin function.
18435         (__builtin_divweuo): Likewise.
18436         (__builtin_divdeo): Likewise.
18437         (__builtin_divdeuo): Likewise.
18439 2018-03-28  Jakub Jelinek  <jakub@redhat.com>
18441         PR target/85095
18442         * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
18443         *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
18445         PR tree-optimization/82004
18446         * gimple-match-head.c (optimize_pow_to_exp): New function.
18447         * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
18448         Don't fold to exp if optimize_pow_to_exp is false.
18450 2018-03-28  Martin Liska  <mliska@suse.cz>
18452         PR other/84819
18453         * calls.c (initialize_argument_information): Fix trailing space.
18454         * common.opt: Fix typo and provide better explanation for
18455         -fsanitize-coverage option.
18456         * config/i386/i386.opt: Fix typo.
18458 2018-03-28  Jakub Jelinek  <jakub@redhat.com>
18459             Martin Liska  <mliska@suse.cz>
18461         PR sanitizer/85081
18462         * gimplify.c (asan_poison_variable): Don't do the check for
18463         gimplify_omp_ctxp here.
18464         (gimplify_decl_expr): Do it here.
18465         (gimplify_target_expr): Likewise.
18467 2018-03-28  Martin Liska  <mliska@suse.cz>
18469         PR target/84988
18470         * config/i386/i386.c (ix86_function_arg_advance): Do not call
18471         chkp_type_bounds_count if MPX is not enabled.
18473 2018-03-27  Chung-Ju Wu  <jasonwucj@gmail.com>
18475         * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
18477 2018-03-27  Michael Meissner  <meissner@linux.vnet.ibm.com>
18479         PR target/84914
18480         * config/rs6000/rs6000.c (create_complex_muldiv): New helper
18481         function to create the function decl for complex long double
18482         multiply and divide for -mabi=ieeelongdouble.
18483         (init_float128_ieee): Call it.
18485 2018-03-27  H.J. Lu  <hongjiu.lu@intel.com>
18487         PR target/85044
18488         * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
18489         -fcf-protection=branch -mibt.
18490         * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
18492 2018-03-27  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
18494         PR target/81863
18495         * config/arm/arm.c (arm_valid_symbolic_address): Handle
18496         arm_word_relocations.
18498 2018-03-27  Cesar Philippidis  <cesar@codesourcery.com>
18500         PR target/85056
18501         * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
18502         extern array declarations.
18504 2018-03-27  Richard Biener  <rguenther@suse.de>
18506         PR middle-end/84067
18507         * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
18508         explicit single_use checks.
18510 2018-03-27  Richard Biener  <rguenther@suse.de>
18512         PR tree-optimization/85082
18513         * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
18514         Valueize the VUSE.
18516 2018-03-27  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
18518         * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
18519         * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
18520         Turn on fasynchronous-unwind-tables and funwind-tables.
18522 2018-03-26  Uros Bizjak  <ubizjak@gmail.com>
18524         PR target/85073
18525         * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
18526         (*bmi_blsr_<mode>_ccz): Ditto.
18528 2018-03-26  Tom de Vries  <tom@codesourcery.com>
18530         PR tree-optimization/85063
18531         * omp-general.c (offloading_function_p): New function.  Factor out
18532         of ...
18533         * omp-offload.c (pass_omp_target_link::gate): ... here.
18534         * omp-general.h (offloading_function_p): Declare.
18535         * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
18536         with attribute omp declare target for offloading functions.
18538 2018-03-24  Richard Sandiford  <richard.sandiford@linaro.org>
18540         PR tree-optimization/84005
18541         * tree-data-ref.h (get_base_for_alignment): Declare.
18542         * tree-data-ref.c (get_base_for_alignment_1): New function.
18543         (get_base_for_alignment): Likewise.
18544         * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
18545         get_base_for_alignment to find a suitable base object, instead
18546         of always using drb->base_address.
18548 2018-03-23  Jakub Jelinek  <jakub@redhat.com>
18550         PR inline-asm/85022
18551         * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
18552         known size by default.
18554 2018-03-23  Vladimir Makarov  <vmakarov@redhat.com>
18556         PR inline-asm/85030
18557         * lra-constraints.c (process_alt_operands): Don't match BLKmode
18558         and non BLKmode operands.
18560 2018-03-23  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
18562         PR target/85026
18563         * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
18564         Clean up attributes.
18566 2018-03-23  Richard Biener  <rguenther@suse.de>
18568         PR debug/85020
18569         * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
18570         we are going to emit early debug for LTO.
18572 2018-03-23  Jakub Jelinek  <jakub@redhat.com>
18574         PR inline-asm/85034
18575         * function.c (match_asm_constraints_1): Don't optimize if input
18576         doesn't satisfy general_operand predicate for output's mode.
18578         PR inline-asm/85022
18579         * alias.c (write_dependence_p): Don't require for x_canonicalized
18580         non-VOIDmode if x has VOIDmode.
18582         PR sanitizer/85029
18583         * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
18584         just don't try to optimize it rather than assert it never happens.
18586 2018-03-22  Kelvin Nilsen  <kelvin@gcc.gnu.org>
18588         * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
18589         macro expansions for definition of ST_INTERNAL_<mode> and
18590         LD_INTERNAL_<mode> builtins.
18591         * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
18592         Remove prototype.
18593         * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
18594         function.
18595         (altivec_expand_st_builtin): Likewise.
18596         (altivec_expand_builtin): Remove calls to deleted functions.
18597         (rs6000_address_for_altivec): Delete this function.
18598         * config/rs6000/vector.md: Remove expands for
18599         vector_altivec_load_<mode> and vector_altivec_store_<mode>.
18601 2018-03-22  Sudakshina Das  <sudi.das@arm.com>
18603         PR target/84826
18604         * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
18605         * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
18606         re-computing once computed.
18607         (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
18608         (arm_init_machine_status): Initialize
18609         machine->static_chain_stack_bytes.
18611 2018-03-22  Kelvin Nilsen  <kelvin@gcc.gnu.org>
18613         PR target/84760
18614         * doc/extend.texi: Add four new prototypes for vec_ld.
18615         * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
18616         definitions for more logical presentation.
18617         * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
18618         entries for V1TI variants of __builtin_altivec_ld builtin.
18619         * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
18620         handling of V1TI variant of LVX icode pattern.
18621         (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
18622         (rs6000_gimple_fold_builtin): Likewise.
18623         (altivec_init_builtins): Add code to define
18624         __builtin_altivec_lvx_v1ti function.
18626 2018-03-22  Jakub Jelinek  <jakub@redhat.com>
18628         PR inline-asm/84941
18629         * function.c (match_asm_constraints_1): Don't do the optimization
18630         if input isn't a REG, SUBREG, MEM or constant.
18632 2018-03-22  Tom de Vries  <tom@codesourcery.com>
18634         PR tree-optimization/84956
18635         * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
18636         bb_has_abnormal_pred.
18638 2018-03-22  Jakub Jelinek  <jakub@redhat.com>
18640         PR sanitizer/85018
18641         * dwarf2asm.c (dw2_output_indirect_constant_1): Set
18642         DECL_INITIAL (decl) to decl at the end.
18643         * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
18644         adjust the comment.
18646 2018-03-21  Joseph Myers  <joseph@codesourcery.com>
18648         * doc/extend.texi (__builtin_tgmath): Document when complex
18649         integer types are treated as _Complex _Float64.
18651 2018-03-21  Tom de Vries  <tom@codesourcery.com>
18653         * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
18655 2018-03-21  Jakub Jelinek  <jakub@redhat.com>
18657         PR tree-optimization/84960
18658         * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
18659         if it is ENTRY block, move them into single succ of ENTRY in that case.
18661 2018-03-21  Richard Sandiford  <richard.sandiford@linaro.org>
18663         PR tree-optimization/84811
18664         * poly-int.h (poly_span_traits): Remove the T3 parameter and
18665         promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
18666         (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
18667         (known_subrange_p): Update accordingly.  Cast each value involved
18668         in the size comparison, rather than casting the result of the
18669         subtraction.
18671 2018-03-21  Jakub Jelinek  <jakub@redhat.com>
18673         PR tree-optimization/84982
18674         * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
18675         by flipping the least significant bit rather than all bits from
18676         bitpos to bitpos + bitsize - 1.
18678 2018-03-21  Nathan Sidwell  <nathan@acm.org>
18680         * doc/extend.texi (Deprecated Features): Remove mention of
18681         long-deleted deprecations.
18683 2018-03-21  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
18685         PR jit/84288
18686         * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
18687         * configure: Regenerate.
18689 2018-03-21  Tom de Vries  <tom@codesourcery.com>
18691         PR tree-optimization/83126
18692         * tree-parloops.c (num_phis): New function.
18693         (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
18695 2018-03-21  Nathan Sidwell  <nathan@acm.org>
18697         * doc/extend.texi (Deprecated Features): Update deprecated flags,
18698         mention anon-struct/union members and trailing attributes.
18700 2018-03-21  Bin Cheng  <bin.cheng@arm.com>
18702         PR tree-optimization/84969
18703         * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
18704         builtin memset partitions if they set different rhs values.
18706 2018-03-21  Jakub Jelinek  <jakub@redhat.com>
18708         PR rtl-optimization/84989
18709         * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
18710         VEC_DUPLICATE with scalar result mode.
18712 2018-03-21  Martin Liska  <mliska@suse.cz>
18714         PR ipa/84963
18715         * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
18716         not intended return statement.
18718 2018-03-21  Martin Liska  <mliska@suse.cz>
18720         PR target/84988
18721         * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
18722         (chkp_find_bound_slots_1): Limit number of iterations.
18724 2018-03-20  David H. Gutteridge  <dhgutteridge@sympatico.ca>
18726         PR target/84838
18727         * Minor grammar fixes for x86 options.
18729 2018-03-20  Jakub Jelinek  <jakub@redhat.com>
18731         PR debug/84875
18732         * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
18733         holding REG_CFA_RESTORE notes, instead turn them into a USE.
18735 2018-03-20  Peter Bergner  <bergner@vnet.ibm.com>
18737         PR target/83789
18738         * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
18739         (altivec_lvx_<mode>_1op): Likewise.
18740         (altivec_stvx_<mode>_2op): Likewise.
18741         (altivec_stvx_<mode>_1op): Likewise.
18742         (altivec_lvx_<VM2:mode>): New define_expand.
18743         (altivec_stvx_<VM2:mode>): Likewise.
18744         (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
18745         (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
18746         (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
18747         (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
18748         * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
18749         (rs6000_gen_lvx): Likewise.
18750         * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
18751         (altivec_expand_stv_builtin): Likewise.
18752         (altivec_expand_builtin): Likewise.
18753         * config/rs6000/vector.md: Likewise.
18755 2018-03-20  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
18757         PR target/82518
18758         * config/arm/arm.c (arm_array_mode_supported_p): Return false for
18759         BYTES_BIG_ENDIAN.
18761 2018-03-20  Richard Biener  <rguenther@suse.de>
18763         PR target/84986
18764         * config/i386/i386.c (ix86_add_stmt_cost): Only cost
18765         sign-conversions as zero, fall back to standard scalar_stmt
18766         cost for the rest.
18768 2018-03-20  Martin Liska  <mliska@suse.cz>
18770         PR ipa/84825
18771         * predict.c (rebuild_frequencies): Handle case when we have
18772         PROFILE_ABSENT, but flag_guess_branch_prob is false.
18774 2018-03-20  Jakub Jelinek  <jakub@redhat.com>
18776         PR target/84990
18777         * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
18778         flag_section_anchors.
18779         * varasm.c (use_blocks_for_decl_p): Remove hack for
18780         dw2_force_const_mem.
18782         PR target/84845
18783         * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
18784         to ...
18785         (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this.  If pseudos can't
18786         be created, use lowpart_subreg of operands[0] rather than operands[0]
18787         itself.
18788         (*aarch64_reg_<mode>3_minus_mask): Rename to ...
18789         (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
18790         (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
18791         and n constraint instead of aarch64_shift_imm_di and Usd.
18792         (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
18793         (*aarch64_<optab>_reg_minus<mode>3): ... this.
18795 2018-03-20  Sudakshina Das  <sudi.das@arm.com>
18797         PR target/82989
18798         * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
18799         to favor GPR over NEON registers.
18800         (<shift>di3_neon): Likewise.
18802 2018-03-20  Tom de Vries  <tom@codesourcery.com>
18804         PR target/84952
18805         * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
18806         (nvptx_process_pars): Emit bar.sync asap and alap.
18808 2018-03-20  Tom de Vries  <tom@codesourcery.com>
18810         PR target/84954
18811         * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
18812         seen_label if seen_label is already set.
18814 2018-03-20  Jakub Jelinek  <jakub@redhat.com>
18816         PR target/84945
18817         * config/i386/i386.c (fold_builtin_cpu): For features above 31
18818         use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
18819         Use 1U instead of 1.  Formatting fixes.
18821         PR c/84953
18822         * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
18823         instead of TREE_TYPE (s1) for the return value.
18825 2018-03-19  Jakub Jelinek  <jakub@redhat.com>
18827         PR tree-optimization/84946
18828         * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
18829         bitsize + bitsize in poly_uint64 rather than poly_int64.
18831         PR sanitizer/78651
18832         * dwarf2asm.c: Include fold-const.c.
18833         (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
18834         of decl rather than decl itself.
18836         PR rtl-optimization/84643
18837         * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
18839 2018-03-19  Maxim Ostapenko  <m.ostapenko@samsung.com>
18841         PR sanitizer/78651
18842         * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
18843         calling assemble_variable.
18845 2018-03-19  Sudakshina Das  <sudi.das@arm.com>
18847         PR target/81647
18848         * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
18849         instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
18851 2018-03-19  Jim Wilson  <jimw@sifive.com>
18853         PR bootstrap/84856
18854         * config/riscv/riscv.c (riscv_function_arg_boundary): Use
18855         PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
18856         (riscv_first_stack_step): Likewise.
18857         (riscv_option_override): Use STACK_BOUNDARY instead of
18858         MIN_STACK_BOUNDARY.
18859         * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
18860         MIN_STACK_BOUNDARY.
18861         (BIGGEST_ALIGNMENT): Set to 128.
18862         (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
18863         (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
18864         STACK_BOUNDARY.
18866 2018-03-19  Richard Biener  <rguenther@suse.de>
18868         PR tree-optimization/84933
18869         * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
18870         values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
18872 2018-03-19  Richard Biener  <rguenther@suse.de>
18874         PR tree-optimization/84859
18875         * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
18876         (cond_if_else_store_replacement): Perform sinking operation on
18877         single-store BBs regardless of MAX_STORES_TO_SINK setting.
18878         Generalize what a BB with a single eligible store is.
18880 2018-03-19  Richard Biener  <rguenther@suse.de>
18882         PR tree-optimization/84929
18883         * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
18884         chrec_is_positive against non-chrec arg.
18886 2018-03-19  Tamar Christina  <tamar.christina@arm.com>
18888         PR target/84711
18889         * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
18891 2018-03-18  Martin Liska  <mliska@suse.cz>
18893         PR rtl-optimization/84635
18894         * regrename.c (build_def_use): Use matches_mode only when
18895         matches >= 0.
18897 2018-03-18  Richard Sandiford  <richard.sandiford@linaro.org>
18899         PR tree-optimization/84913
18900         * tree-vect-loop.c (vectorizable_reduction): Don't try to
18901         vectorize chains of COND_EXPRs.
18903 2018-03-18  Chung-Ju Wu  <jasonwucj@gmail.com>
18905         * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
18907 2018-03-18  Chung-Ju Wu  <jasonwucj@gmail.com>
18909         * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
18911 2018-03-18  Chung-Ju Wu  <jasonwucj@gmail.com>
18913         * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
18915 2018-03-17  Chung-Ju Wu  <jasonwucj@gmail.com>
18916             Kito Cheng  <kito.cheng@gmail.com>
18918         * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
18919         * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
18920         (nds32_adjust_reg_alloc_order): New function.
18921         * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
18923 2018-03-17  Kito Cheng  <kito.cheng@gmail.com>
18925         * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
18926         nds32_print_operand, nds32_print_operand_address): Use
18927         HOST_WIDE_INT_PRINT_DEC instead.
18929 2018-03-17  Chung-Ju Wu  <jasonwucj@gmail.com>
18931         * config/nds32/nds32.c (nds32_register_priority): Modify cost.
18933 2018-03-17  Jakub Jelinek  <jakub@redhat.com>
18935         PR target/84902
18936         * config/i386/i386.c (initial_ix86_tune_features,
18937         initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
18938         unsigned long long.
18939         (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
18940         to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
18941         rather than 1u << ix86_tune.  Formatting fix.
18942         (ix86_option_override_internal): Change ix86_arch_mask from
18943         unsigned int to unsigned HOST_WIDE_INT, initialize to
18944         HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
18945         (ix86_function_specific_restore): Likewise.
18947 2018-03-16  Jakub Jelinek  <jakub@redhat.com>
18949         PR target/84899
18950         * postreload.c (reload_combine_recognize_pattern): Perform
18951         INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
18952         truncate_int_for_mode the result for the destination's mode.
18954         PR c/84909
18955         * hsa-gen.c (mem_type_for_type): Fix comment typo.
18956         * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
18957         Likewise.
18958         * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
18959         Likewise.
18961 2018-03-16  Vladimir Makarov  <vmakarov@redhat.com>
18963         PR target/84876
18964         * lra-assigns.c (lra_split_hard_reg_for): Don't use
18965         regno_allocno_class_array and sorted_pseudos.
18966         * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
18967         insns where regno is used.
18969 2018-03-16  Martin Liska  <mliska@suse.cz>
18971         PR ipa/84833
18972         * multiple_target.c (create_dispatcher_calls): Redirect
18973         reference in the symbol table.
18975 2018-03-16  Martin Liska  <mliska@suse.cz>
18977         PR ipa/84722
18978         * multiple_target.c (create_dispatcher_calls): Redirect also
18979         an alias.
18981 2018-03-16  Jakub Jelinek  <jakub@redhat.com>
18983         PR c++/79937
18984         PR c++/82410
18985         * tree.h (TARGET_EXPR_NO_ELIDE): Define.
18986         * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
18987         TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
18989 2018-03-16  Julia Koval  <julia.koval@intel.com>
18991         * doc/invoke.texi (Skylake Server): Add CLWB.
18992         Cannonlake): Remove CLWB.
18994 2018-03-16  Jakub Jelinek  <jakub@redhat.com>
18996         PR tree-optimization/84841
18997         * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
18998         1 << 3.
18999         (FLOAT_ONE_CONST_TYPE): Define.
19000         (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
19001         (sort_by_operand_rank): Put entries with higher constant_type last
19002         rather than first to match comments.
19004 2018-03-15  Sandra Loosemore  <sandra@codesourcery.com>
19006         * config/nios2/nios2.md (movsi_internal): Fix thinko in
19007         split predicate.
19009 2018-03-15  Jakub Jelinek  <jakub@redhat.com>
19011         PR c++/79085
19012         * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
19013         check and use address of target always.
19015 2018-03-15  H.J. Lu  <hongjiu.lu@intel.com>
19017         PR target/84574
19018         * config/i386/i386.c (indirect_thunk_needed): Update comments.
19019         (indirect_thunk_bnd_needed): Likewise.
19020         (indirect_thunks_used): Likewise.
19021         (indirect_thunks_bnd_used): Likewise.
19022         (indirect_return_needed): New.
19023         (indirect_return_bnd_needed): Likewise.
19024         (output_indirect_thunk_function): Add a bool argument for
19025         function return.
19026         (output_indirect_thunk_function): Don't generate alias for
19027         function return thunk.
19028         (ix86_code_end): Call output_indirect_thunk_function to generate
19029         function return thunks.
19030         (ix86_output_function_return): Set indirect_return_bnd_needed
19031         and indirect_return_needed instead of indirect_thunk_bnd_needed
19032         and indirect_thunk_needed.
19034 2018-03-15  Olga Makhotina  <olga.makhotina@intel.com>
19036         * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
19037         (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
19038         (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
19040 2018-03-15  David Malcolm  <dmalcolm@redhat.com>
19041             Paul Hua <paul.hua.gm@gmail.com>
19043         PR c/84852
19044         * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
19046 2018-03-15  Segher Boessenkool  <segher@kernel.crashing.org>
19048         * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
19049         TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
19050         resp. SFmode cases.
19052 2018-03-15  Tamar Christina  <tamar.christina@arm.com>
19054         PR target/84711
19055         * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
19056         instead of GET_MODE_SIZE when comparing Units.
19058 2018-03-15  Vladimir Mezentsev  <vladimir.mezentsev@oracle.com>
19060         PR target/68256
19061         * varasm.c (hash_section): Return an unchangeble hash value
19062         * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
19063         Return !aarch64_can_use_per_function_literal_pools_p ().
19065 2018-03-15  Jakub Jelinek  <jakub@redhat.com>
19067         PR target/84860
19068         * optabs.c (emit_conditional_move): Pass address of cmode's copy
19069         rather than address of cmode as last argument to prepare_cmp_insn.
19071 2018-03-15  Julia Koval  <julia.koval@intel.com>
19073         * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
19074         F_AVX512VNNI, F_AVX512BITALG): New.
19076 2018-03-14  John David Anglin  <danglin@gcc.gnu.org>
19078         PR target/83451
19079         * config/pa/pa.c (pa_emit_move_sequence):  Always emit secondary reload
19080         insn for floating-point loads and stores.
19082 2018-03-14  Carl Love  <cel@us.ibm.com>
19084         * config/rs6000/rs6000-c.c: Add macro definitions for
19085         ALTIVEC_BUILTIN_VEC_PERMXOR.
19086         * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
19087         * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
19088         * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
19089         * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
19090         UNSPEC_VPERMXOR.
19091         * config/doc/extend.texi: Add prototypes for vec_permxor.
19093 2018-03-14  David Malcolm  <dmalcolm@redhat.com>
19095         PR c/84852
19096         * diagnostic-show-locus.c (class layout_point): Convert m_line
19097         from int to linenum_type.
19098         (line_span::comparator): Use linenum "compare" function when
19099         comparing line numbers.
19100         (test_line_span): New function.
19101         (layout_range::contains_point): Convert param "row" from int to
19102         linenum_type.
19103         (layout_range::intersects_line_p): Likewise.
19104         (layout::will_show_line_p): Likewise.
19105         (layout::print_source_line): Likewise.
19106         (layout::should_print_annotation_line_p): Likewise.
19107         (layout::print_annotation_line): Likewise.
19108         (layout::print_leading_fixits): Likewise.
19109         (layout::annotation_line_showed_range_p): Likewise.
19110         (struct line_corrections): Likewise for field m_row.
19111         (line_corrections::line_corrections): Likewise for param "row".
19112         (layout::print_trailing_fixits): Likewise.
19113         (layout::get_state_at_point): Likewise.
19114         (layout::get_x_bound_for_row): Likewise.
19115         (layout::print_line): Likewise.
19116         (diagnostic_show_locus): Likewise for locals "last_line" and "row".
19117         (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
19118         * input.c (selftest::test_linenum_comparisons): New function.
19119         (selftest::input_c_tests): Call it.
19120         * selftest.c (selftest::test_assertions): Test ASSERT_GT,
19121         ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
19122         * selftest.h (ASSERT_GT): New macro.
19123         (ASSERT_GT_AT): New macro.
19124         (ASSERT_LT): New macro.
19125         (ASSERT_LT_AT): New macro.
19127 2018-03-14  Segher Boessenkool  <segher@kernel.crashing.org>
19129         PR rtl-optimization/84780
19130         * combine.c (distribute_links): Don't make a link based on pc_rtx.
19132 2018-03-14  Martin Liska  <mliska@suse.cz>
19134         * tree.c (record_node_allocation_statistics): Use
19135         get_stats_node_kind.
19136         (get_stats_node_kind): New function extracted from
19137         record_node_allocation_statistics.
19138         (free_node): Use get_stats_node_kind.
19140 2018-03-14  Richard Biener  <rguenther@suse.de>
19142         * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
19143         that the value-set of ANTIC_IN doesn't grow.
19145         Revert
19146         * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
19147         member.
19148         (BB_VISITED_WITH_VISITED_SUCCS): New define.
19149         (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
19151 2018-03-14  Julia Koval  <julia.koval@intel.com>
19153         * config.gcc (icelake-client, icelake-server): New.
19154         (icelake): Remove.
19155         * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
19156         (initial_ix86_arch_features): Ditto.
19157         (PTA_SKYLAKE): Add SGX.
19158         (PTA_ICELAKE): Remove.
19159         (PTA_ICELAKE_CLIENT): New.
19160         (PTA_ICELAKE_SERVER): New.
19161         (ix86_option_override_internal): Split up icelake on icelake client and
19162         icelake server.
19163         (get_builtin_code_for_version): Ditto.
19164         (fold_builtin_cpu): Ditto.
19165         * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
19166         * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
19167         * config/i386/i386.h (processor_type): Ditto.
19168         * doc/invoke.texi: Ditto.
19170 2018-03-14  Jakub Jelinek  <jakub@redhat.com>
19172         PR sanitizer/83392
19173         * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
19174         INTEGER_CST offset, add it together with bitpos / 8 and
19175         sign extend based on POINTER_SIZE.
19177         PR target/84844
19178         Revert
19179         2017-04-20  Uros Bizjak  <ubizjak@gmail.com>
19181         PR target/78090
19182         * config/i386/constraints.md (Yc): New register constraint.
19183         * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
19184         Use Yc constraint for alternative 2 of operand 0.  Remove
19185         preferred_for_speed attribute.
19187 2018-03-14  Richard Biener  <rguenther@suse.de>
19189         PR tree-optimization/84830
19190         * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
19191         with the old one to avoid oscillations.
19193 2018-03-13  Vladimir Makarov  <vmakarov@redhat.com>
19195         PR target/83712
19196         * lra-assigns.c (find_all_spills_for): Ignore uninteresting
19197         pseudos.
19198         (assign_by_spills): Return a flag of reload assignment failure.
19199         Do not process the reload assignment failures.  Do not spill other
19200         reload pseudos if they has the same reg class.  Update n if
19201         necessary.
19202         (lra_assign): Add a return arg.  Set up from the result of
19203         assign_by_spills call.
19204         (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
19205         * lra-constraints.c (split_reg): Add a new arg.  Use it instead of
19206         usage_insns if it is not NULL.
19207         (spill_hard_reg_in_range): New function.
19208         (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
19209         * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
19210         function prototypes.
19211         (lra_assign): Change prototype.
19212         * lra.c (lra): Add code to deal with fails by splitting hard reg
19213         live ranges.
19215 2018-03-01  Palmer Dabbelt  <palmer@sifive.com>
19217         * config/riscv/riscv.opt (mrelax): New option.
19218         * config/riscv/riscv.c (riscv_file_start): Emit ".option
19219         "norelax" when riscv_mrelax is disabled.
19220         * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
19222 2018-03-13  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
19224         PR target/84743
19225         * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
19226         reassociation for int modes.
19228 2018-03-13  Richard Sandiford  <richard.sandiford@linaro.org>
19230         * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
19231         Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
19232         for big-endian.
19233         * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
19234         * config/aarch64/aarch64-sve.md
19235         (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
19236         (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
19237         (*extend<mode><Vwide>2): Rename to...
19238         (aarch64_sve_extend<mode><Vwide>2): ...this.
19239         (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
19240         renaming the old pattern to...
19241         (aarch64_sve_punpk<perm_hilo>_<mode>): ...this.  Only define
19242         unsigned packs.
19243         (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
19244         define_expand, renaming the old pattern to...
19245         (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
19246         (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
19247         (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
19248         account when deciding which SVE instruction the optab should use.
19249         (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
19251 2018-03-13  Richard Sandiford  <richard.sandiford@linaro.org>
19253         * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
19254         (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
19255         (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
19256         (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
19257         (tlsdesc_small_<mode>): Turn a define_expand and use
19258         tlsdesc_small_sve_<mode> for SVE.  Rename original define_insn to...
19259         (tlsdesc_small_advsimd_<mode>): ...this.
19260         (tlsdesc_small_sve_<mode>): New pattern.
19262 2018-03-13  Richard Sandiford  <richard.sandiford@linaro.org>
19264         * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
19265         (UNSPEC_UMUL_HIGHPART): New constants.
19266         (MUL_HIGHPART): New int iteraor.
19267         (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
19268         * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
19269         define_expand.
19270         (*<su>mul<mode>3_highpart): New define_insn.
19272 2018-03-13  Eric Botcazou  <ebotcazou@adacore.com>
19274         PR lto/84805
19275         * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
19276         incomplete types.
19278 2018-03-13  Martin Liska  <mliska@suse.cz>
19280         PR ipa/84658.
19281         * (sem_item_optimizer::sem_item_optimizer): Initialize new
19282         vector.
19283         (sem_item_optimizer::~sem_item_optimizer): Release it.
19284         (sem_item_optimizer::merge_classes): Register variable aliases.
19285         (sem_item_optimizer::fixup_pt_set): New function.
19286         (sem_item_optimizer::fixup_points_to_sets): Likewise.
19287         * ipa-icf.h: Declare new variables and functions.
19289 2018-03-13  Jakub Jelinek  <jakub@redhat.com>
19291         PR middle-end/84834
19292         * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
19293         integer_pow2p@2 and test integer_pow2p in condition.
19294         (A < 0 ? C : 0): Similarly for @1.
19296         PR middle-end/84831
19297         * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
19298         characters starting at p contain '\0' character, don't look beyond
19299         that.
19301         PR target/84827
19302         * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
19303         pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
19305         PR target/84828
19306         * reg-stack.c (change_stack): Change update_end var from int to
19307         rtx_insn *, if non-NULL don't update just BB_END (current_block), but
19308         also call set_block_for_insn on the newly added insns and rescan.
19310         PR target/84786
19311         * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
19312         on the last operand.
19314         PR c++/84704
19315         * tree.c (stabilize_reference_1): Return save_expr (e) for
19316         STATEMENT_LIST even if it doesn't have side-effects.
19318 2018-03-12  Jonathan Wakely  <jwakely@redhat.com>
19320         * doc/invoke.texi (-mclflushopt): Fix spelling of option.
19322 2018-03-12  Renlin Li  <renlin.li@arm.com>
19324         * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
19325         aarch64_output_scalar_simd_mov_immediate.
19327 2018-03-12  Martin Sebor  <msebor@redhat.com>
19329         PR tree-optimization/83456
19330         * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
19331         for perfectly overlapping calls to memcpy.
19332         (gimple_fold_builtin_memory_chk): Same.
19333         (gimple_fold_builtin_strcpy): Handle no-warning.
19334         (gimple_fold_builtin_stxcpy_chk): Same.
19335         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
19337 2018-03-12  Segher Boessenkool  <segher@kernel.crashing.org>
19339         * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
19340         parameter.  Use it for SFmode.
19341         (rs6000_function_arg_advance_1): Adjust.
19342         (rs6000_function_arg): Adjust.
19343         (rs6000_gimplify_va_arg): Pass false for that new parameter.
19345 2018-03-12  Segher Boessenkool  <segher@kernel.crashing.org>
19347         PR rtl-optimization/84169
19348         PR rtl-optimization/84780
19349         * combine.c (can_combine_p): Check for a 2-insn combination whether
19350         the destination register is used between the two insns, too.
19352 2018-03-12  Richard Biener  <rguenther@suse.de>
19354         PR tree-optimization/84803
19355         * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
19356         for refs DR analysis didn't process.
19358 2018-03-12  Richard Biener  <rguenther@suse.de>
19360         PR tree-optimization/84777
19361         * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
19362         force-vectorize loops ignore whether we are optimizing for size.
19364 2018-03-12  Chung-Ju Wu  <jasonwucj@gmail.com>
19366         * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
19367         (TARGET_MD_ASM_ADJUST): Define.
19369 2018-03-12  Monk Chiang  <sh.chiang04@gmail.com>
19370             Kito Cheng  <kito.cheng@gmail.com>
19371             Chung-Ju Wu  <jasonwucj@gmail.com>
19373         * config/nds32/nds32.c (nds32_compute_stack_frame,
19374         nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
19375         nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
19376         nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
19377         nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
19378         * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
19379         NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
19380         * config/nds32/nds32.md (prologue, epilogue): Use macro
19381         NDS32_V3PUSH_AVAILABLE_P to do checking.
19383 2018-03-11  Jakub Jelinek  <jakub@redhat.com>
19385         PR debug/58150
19386         * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
19387         DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
19388         but on TYPE_SIZE.  Don't do anything for ENUM_IS_OPAQUE if not creating
19389         a new die.  Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE.  Guard
19390         addition of most attributes on !orig_type_die or the attribute not
19391         being present already.  Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
19393 2018-03-11  Kito Cheng  <kito.cheng@gmail.com>
19394             Chung-Ju Wu  <jasonwucj@gmail.com>
19396         * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
19397         __NDS32_VH__ macro.
19398         * config/nds32/nds32.opt (mvh): New option.
19400 2018-03-11  Kito Cheng  <kito.cheng@gmail.com>
19401             Chung-Ju Wu  <jasonwucj@gmail.com>
19403         * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
19404         function.
19405         * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
19406         * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
19407         definition.
19409 2018-03-11  Kito Cheng  <kito.cheng@gmail.com>
19410             Chung-Ju Wu  <jasonwucj@gmail.com>
19412         * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
19413         function.
19414         * config/nds32/nds32-multiple.md (strlensi): New pattern.
19415         * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
19417 2018-03-11  Monk Chiang  <sh.chiang04@gmail.com>
19418             Kito Cheng  <kito.cheng@gmail.com>
19419             Chung-Ju Wu  <jasonwucj@gmail.com>
19421         * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
19422         UNSPEC_FFMISM and UNSPEC_FLMISM.
19423         * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
19424         for ffb, ffmism and flmism.
19425         * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
19426         (unspec_ffmism): Ditto.
19427         (unspec_flmism): Ditto.
19428         (nds32_expand_builtin_impl): Check if string extension is available.
19429         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
19430         NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
19432 2018-03-10  Vladimir Makarov  <vmakarov@redhat.com>
19434         Reverting patch:
19435         2018-03-09  Vladimir Makarov  <vmakarov@redhat.com>
19437         PR target/83712
19438         * lra-assigns.c (assign_by_spills): Return a flag of reload
19439         assignment failure.  Do not process the reload assignment
19440         failures.  Do not spill other reload pseudos if they has the same
19441         reg class.
19442         (lra_assign): Add a return arg.  Set up from the result of
19443         assign_by_spills call.
19444         (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
19445         * lra-constraints.c (split_reg): Add a new arg.  Use it instead of
19446         usage_insns if it is not NULL.
19447         (spill_hard_reg_in_range): New function.
19448         (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
19449         * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
19450         function prototypes.
19451         (lra_assign): Change prototype.
19452         * lra.c (lra): Add code to deal with fails by splitting hard reg
19453         live ranges.
19455 2018-03-10  H.J. Lu  <hongjiu.lu@intel.com>
19457         PR target/84807
19458         * config/i386/i386.opt: Replace Enforcment with Enforcement.
19460 2018-03-10  Alexandre Oliva  <aoliva@redhat.com>
19462         PR debug/84620
19463         * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
19464         (dw_val_node): Add val_symbolic_view.
19465         * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
19466         (symview_upper_bound): New.
19467         (new_line_info_table): Initialize symviews_since_reset.
19468         (dwarf2out_source_line): Count symviews_since_reset and set
19469         symview_upper_bound.
19470         (dw_val_equal_p): Handle symview.
19471         (add_AT_symview): New.
19472         (print_dw_val): Handle symview.
19473         (attr_checksum, attr_checksum_ordered): Likewise.
19474         (same_dw_val_p, size_of_die): Likewise.
19475         (value_format, output_die): Likewise.
19476         (add_high_low_attributes): Use add_AT_symview for entry_view.
19477         (dwarf2out_finish): Reset symview_upper_bound, clear
19478         zero_view_p.
19480 2018-03-09  Peter Bergner  <bergner@vnet.ibm.com>
19482         PR target/83969
19483         * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
19484         Add strict argument and use it.
19485         (rs6000_split_multireg_move): Update for new strict argument.
19486         (mem_operand_gpr): Disallow all non-offsettable addresses.
19487         * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
19489 2018-03-09  Jakub Jelinek  <jakub@redhat.com>
19491         PR target/84772
19492         * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
19493         temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
19494         * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
19496         PR c++/84767
19497         * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
19498         decl, use remap_type if we want to use the type.
19500 2018-03-09  Martin Sebor  <msebor@redhat.com>
19502         PR tree-optimization/84526
19503         * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
19504         Remove dead code.
19505         (builtin_access::generic_overlap): Be prepared to handle non-array
19506         base objects.
19508 2018-03-09  Alexandre Oliva  <aoliva@redhat.com>
19510         PR rtl-optimization/84682
19511         * lra-constraints.c (process_address_1): Check is_address flag
19512         for address constraints.
19513         (process_alt_operands): Likewise.
19514         * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
19515         preprocess_constraints.
19516         * recog.h (preprocess_constraints): Add oploc parameter.
19517         Adjust callers.
19518         * recog.c (preprocess_constraints): Test address_operand for
19519         CT_ADDRESS constraints.
19521 2018-03-09  Vladimir Makarov  <vmakarov@redhat.com>
19523         PR target/83712
19524         * lra-assigns.c (assign_by_spills): Return a flag of reload
19525         assignment failure.  Do not process the reload assignment
19526         failures.  Do not spill other reload pseudos if they has the same
19527         reg class.
19528         (lra_assign): Add a return arg.  Set up from the result of
19529         assign_by_spills call.
19530         (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
19531         * lra-constraints.c (split_reg): Add a new arg.  Use it instead of
19532         usage_insns if it is not NULL.
19533         (spill_hard_reg_in_range): New function.
19534         (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
19535         * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
19536         function prototypes.
19537         (lra_assign): Change prototype.
19538         * lra.c (lra): Add code to deal with fails by splitting hard reg
19539         live ranges.
19541 2018-03-09  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
19543         PR target/83193
19544         * common/config/arm/arm-common.c (arm_parse_arch_option_name):
19545         Accept complain bool parameter.  Only emit errors if it is true.
19546         (arm_parse_cpu_option_name): Likewise.
19547         (arm_target_thumb_only): Adjust callers of the above.
19548         * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
19549         prototype to take a default true bool parameter.
19550         (arm_parse_arch_option_name): Likewise.
19552 2018-03-09  David Malcolm  <dmalcolm@redhat.com>
19553             Francois-Xavier Coudert  <fxcoudert@gcc.gnu.org>
19555         PR jit/64089
19556         PR jit/84288
19557         * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
19558         * configure: Regenerate.
19559         * configure.ac ("linker --version-script option"): New.
19560         ("linker soname option"): New.
19562 2018-03-09  Richard Biener  <rguenther@suse.de>
19564         PR tree-optimization/84775
19565         * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
19566         immediate uses of predicate stmts and mark them modified.
19568         Revert
19569         PR tree-optimization/84178
19570         * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
19571         to caller.
19572         (version_loop_for_if_conversion): Delay update_ssa call.
19573         (tree_if_conversion): Delay update_ssa until after predicate
19574         insertion.
19576 2018-03-09  Eric Botcazou  <ebotcazou@adacore.com>
19578         PR target/84763
19579         * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
19580         when the function accesses prior frames.
19582 2018-03-08  Jakub Jelinek  <jakub@redhat.com>
19584         PR debug/84456
19585         * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
19586         gen_llsym, otherwise call maybe_gen_llsym.
19588         PR inline-asm/84742
19589         * recog.c (asm_operand_ok): Return 0 if multi-character constraint
19590         has ',' character inside of it.
19592 2018-03-08  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
19594         PR target/84748
19595         * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
19596         as clobbering CC_REGNUM.
19598 2018-03-08  Richard Biener  <rguenther@suse.de>
19600         PR middle-end/84552
19601         * tree-scalar-evolution.c: Include tree-into-ssa.h.
19602         (follow_copies_to_constant): Do not follow SSA names registered
19603         for update.
19605 2018-03-08  Richard Biener  <rguenther@suse.de>
19607         PR tree-optimization/84178
19608         * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
19609         to caller.
19610         (version_loop_for_if_conversion): Delay update_ssa call.
19611         (tree_if_conversion): Delay update_ssa until after predicate
19612         insertion.
19614 2018-03-08  David Malcolm  <dmalcolm@redhat.com>
19616         PR tree-optimization/84178
19617         * tree-if-conv.c (release_bb_predicate): Remove the
19618         the assertion that the stmts have NULL use_ops.
19619         Discard the statements, asserting that they haven't
19620         yet been added to a BB.
19622 2018-03-08  Richard Biener  <rguenther@suse.de>
19624         PR tree-optimization/84746
19625         * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
19626         (phi_translate): Pass in destination ANTIC_OUT set.
19627         (phi_translate_1): Likewise.  For a simplified result lookup
19628         a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
19629         (phi_translate_set): Adjust.
19630         (do_pre_regular_insertion): Likewise.
19631         (do_pre_partial_partial_insertion): Likewise.
19633 2018-03-08  Martin Liska  <mliska@suse.cz>
19635         PR gcov-profile/84735
19636         * doc/gcov.texi: Document usage of profile files.
19637         * gcov-io.h: Document changes in the format.
19639 2018-03-08  Alexandre Oliva  <aoliva@redhat.com>
19641         PR debug/84404
19642         PR debug/84408
19643         * dwarf2out.c (struct dw_line_info_table): Update comments for
19644         view == -1.
19645         (FORCE_RESET_NEXT_VIEW): New.
19646         (FORCE_RESETTING_VIEW_P): New.
19647         (RESETTING_VIEW_P): Check for -1 too.
19648         (ZERO_VIEW_P): Likewise.
19649         (new_line_info_table): Force-reset next view.
19650         (dwarf2out_begin_function): Likewise.
19651         (dwarf2out_source_line): Simplify zero_view_p initialization.
19652         Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
19653         view directly.  Omit view when omitting .loc at line 0.
19655 2018-03-08  Jakub Jelinek  <jakub@redhat.com>
19657         PR tree-optimization/84740
19658         * tree-switch-conversion.c (process_switch): Call build_constructors
19659         only if info.phi_count is non-zero.
19661         PR tree-optimization/84739
19662         * tree-tailcall.c (find_tail_calls): Check call arguments against
19663         DECL_ARGUMENTS (current_function_decl) rather than
19664         DECL_ARGUMENTS (func) when checking for tail recursion.
19666 2018-03-07  Jakub Jelinek  <jakub@redhat.com>
19668         * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
19669         Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
19670         Volker Reichelt's entry and add entries for people that perform
19671         GCC fuzzy testing and report numerous bugs.
19673 2018-03-07  Segher Boessenkool  <segher@kernel.crashing.org>
19675         PR target/82411
19676         * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
19677         readonly data in sdata, if that is disabled.
19678         * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
19679         * doc/invoke.texi (RS/6000 and PowerPC Options): Document
19680         -mreadonly-in-sdata option.
19682 2018-03-07  Martin Sebor  <msebor@redhat.com>
19684         PR tree-optimization/84468
19685         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
19686         basic block when looking for nul assignment.
19688 2018-03-07  Eric Botcazou  <ebotcazou@adacore.com>
19690         PR target/84277
19691         * except.h (output_function_exception_table): Adjust prototype.
19692         * except.c (output_function_exception_table): Remove FNNAME parameter
19693         and add SECTION parameter.  Ouput one part of the table at a time.
19694         * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
19695         the first part of the exception table and emit unwind directives.
19696         * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
19697         (i386_pe_seh_cold_init): Likewise.
19698         * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
19699         (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
19700         * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
19701         (ix86_output_call_insn): Emit a nop in one more case for SEH.
19702         * config/i386/winnt.c: Include except.h.
19703         (struct seh_frame_state): Add reg_offset, after_prologue and
19704         in_cold_section fields.
19705         (i386_pe_seh_end_prologue): Set seh->after_prologue.
19706         (i386_pe_seh_cold_init): New function.
19707         (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
19708         to seh->in_cold_section.
19709         (seh_emit_push): Record the offset of the push.
19710         (seh_emit_save): Record the offet of the save.
19711         (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
19712         Test seh->after_prologue to disregard the epilogue.
19713         (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
19714         (i386_pe_end_cold_function): New function.
19716 2018-03-07  Jakub Jelinek  <jakub@redhat.com>
19718         PR fortran/84565
19719         * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
19720         aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
19722         PR c++/84704
19723         * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
19724         on tmp_var.
19725         * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
19726         don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
19728         PR middle-end/84723
19729         * multiple_target.c: Include tree-inline.h and intl.h.
19730         (expand_target_clones): Diagnose and fail if node->definition and
19731         !tree_versionable_function_p (node->decl).
19733 2018-03-06  John David Anglin  <danglin@gcc.gnu.org>
19735         * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
19736         sprint_ul.
19737         (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
19738         (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
19739         * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
19741 2018-03-06  Jakub Jelinek  <jakub@redhat.com>
19743         PR target/84710
19744         * combine.c (try_combine): Use reg_or_subregno instead of handling
19745         just paradoxical SUBREGs and REGs.
19747 2018-03-06  Claudiu Zissulescu  <claziss@synopsys.com>
19749         * config/arc/arc.c (arc_finalize_pic): Remove function.
19750         (arc_must_save_register): We use single base PIC register, remove
19751         checks to save/restore the PIC register.
19752         (arc_expand_prologue): Likewise.
19753         * config/arc/arc-protos.h (arc_set_default_type_attributes):
19754         Remove.
19755         (arc_verify_short): Likewise.
19756         (arc_attr_type): Likewise.
19757         * config/arc/arc.c (arc_set_default_type_attributes): Remove.
19758         (walk_stores): Likewise.
19759         (arc_address_cost): Make it static.
19760         (arc_verify_short): Likewise.
19761         (branch_dest): Likewise.
19762         (arc_attr_type): Likewise.
19763         * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
19764         (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
19765         (arc_final_prescan_insn): Remove inserting the nops due to
19766         hardware hazards.  It is done in reorg step.
19767         (insn_length_variant_t): Remove.
19768         (insn_length_parameters_t): Likewise.
19769         (arc_insn_length_parameters): Likewise.
19770         (arc_get_insn_variants): Likewise.
19771         * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
19773 2018-03-06  Jakub Jelinek  <jakub@redhat.com>
19775         PR inline-asm/84683
19776         * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
19777         assertion failure.
19779         PR tree-optimization/84687
19780         * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
19781         on new_node->decl.
19782         * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
19784 2018-03-05  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
19786         * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
19787         Rename to ppc_speculation_barrier.
19788         * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
19789         __builtin_ppc_speculation_barrier.
19791 2018-03-05  Jakub Jelinek  <jakub@redhat.com>
19793         PR target/84700
19794         * combine.c (combine_simplify_rtx): Don't try to simplify if
19795         if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
19796         are equal to x.
19798 2018-03-05  Segher Boessenkool  <segher@kernel.crashing.org>
19800         * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
19801         to 32 bytes when compiling for POWER9.
19803 2018-03-05  Jakub Jelinek  <jakub@redhat.com>
19805         PR target/84564
19806         * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
19807         regparm >= 3 with no arg reg available also for calls with
19808         flag_force_indirect_call.  Pass decl to ix86_function_regparm.
19810         PR target/84524
19811         * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
19812         orig,vex.
19813         (*<plusminus_insn><mode>3): Likewise.  Remove <mask_operand3> uses.
19815 2018-03-05  Peter Bergner  <bergner@vnet.ibm.com>
19817         PR target/84264
19818         * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
19820 2018-03-05  Richard Biener  <rguenther@suse.de>
19822         PR tree-optimization/84486
19823         * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
19824         When inserting a __builtin_assume_aligned call set the LHS
19825         SSA name alignment info accordingly.
19827 2018-03-05  Wilco Dijkstra  <wdijkstr@arm.com>
19829         PR tree-optimization/84114
19830         * config/aarch64/aarch64.c (aarch64_reassociation_width)
19831         Avoid reassociation of FLOAT_MODE addition.
19833 2018-03-05  Olga Makhotina  <olga.makhotina@intel.com>
19835         * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
19836         OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
19837         OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
19838         (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
19839         * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
19840         * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
19841         * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
19842         and -mwbnoinvd.
19843         * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
19844         __builtin_ia32_wbinvd): New builtins.
19845         (SPECIAL_ARGS2): New.
19846         * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
19847         (SPECIAL_ARGS2): New.
19848         * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
19849         (ix86_valid_target_attribute_inner_p): Ditto.
19850         (ix86_init_mmx_sse_builtins): Add special_args2.
19851         * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
19852         TARGET_WBNOINVD_P): New.
19853         * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
19854         (define_insn "wbinvd", define_insn "wbnoinvd"): New.
19855         * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
19856         * config/i386/immintrin.h (_wbinvd): New intrinsic.
19857         * config/i386/pconfigintrin.h: New file.
19858         * config/i386/wbnoinvdintrin.h: Ditto.
19859         * config/i386/x86intrin.h: Add headers pconfigintrin.h and
19860         wbnoinvdintrin.h.
19861         * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
19863 2018-03-05  Richard Biener  <rguenther@suse.de>
19865         PR tree-optimization/84670
19866         * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
19867         member.
19868         (BB_VISITED_WITH_VISITED_SUCCS): New define.
19869         (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
19870         (compute_antic_aux): Only assert the number of values in ANTIC_IN
19871         doesn't grow if all successors (recursively) were visited at least
19872         once.
19874 2018-03-05  Richard Biener  <rguenther@suse.de>
19876         PR tree-optimization/84650
19877         * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
19878         if executed in the loop pipeline.
19880 2018-03-05  Sandra Loosemore  <sandra@codesourcery.com>
19882         * doc/configfiles.texi (Configuration Files): Move info about
19883         conditionalizing $target-protos.h to...
19884         * doc/sourcebuild.texi (Back End): Here.  Explain how $target.h
19885         differs from $target-protos.h.
19887 2018-03-05  Kito Cheng  <kito.cheng@gmail.com>
19888             Chung-Ju Wu  <jasonwucj@gmail.com>
19890         * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
19891         * config/nds32/nds32-multiple.md (setmemsi): Define.
19892         * config/nds32/nds32-memory-manipulation.c
19893         (nds32_gen_dup_4_byte_to_word_value): New.
19894         (emit_setmem_word_loop): New.
19895         (emit_setmem_byte_loop): New.
19896         (nds32_expand_setmem_loop): New.
19897         (nds32_expand_setmem_loop_v3m): New.
19898         (nds32_expand_setmem_unroll): New.
19899         (nds32_expand_setmem): New.
19901 2018-03-04  Kito Cheng  <kito.cheng@gmail.com>
19902             Chung-Ju Wu  <jasonwucj@gmail.com>
19904         * config/nds32/nds32-memory-manipulation.c
19905         (nds32_emit_load_store): New.
19906         (nds32_emit_post_inc_load_store): New.
19907         (nds32_emit_mem_move): New.
19908         (nds32_emit_mem_move_block): New.
19909         (nds32_expand_movmemsi_loop_unknown_size): New.
19910         (nds32_expand_movmemsi_loop_known_size): New.
19911         (nds32_expand_movmemsi_loop): New.
19912         (nds32_expand_movmemsi_unroll): New.
19913         (nds32_expand_movmemqi): Rename ...
19914         (nds32_expand_movmemsi): ... to this.
19915         * config/nds32/nds32-multiple.md (movmemqi): Rename ...
19916         (movmemsi): ... to this.
19917         * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
19918         (nds32_expand_movmemsi): ... to this.
19920 2018-03-04  Kito Cheng  <kito.cheng@gmail.com>
19921             Monk Chiang  <sh.chiang04@gmail.com>
19922             Chung-Ju Wu  <jasonwucj@gmail.com>
19924         * config/nds32/nds32-protos.h
19925         (nds32_expand_load_multiple): New arguments.
19926         (nds32_expand_store_multiple): Ditto.
19927         (nds32_valid_multiple_load_store): Rename ...
19928         (nds32_valid_multiple_load_store_p): ... to this.
19929         * config/nds32/nds32-memory-manipulation.c
19930         (nds32_expand_load_multiple): Refine implementation.
19931         (nds32_expand_store_multiple): Ditto.
19932         * config/nds32/nds32-multiple.md
19933         (load_multiple): Update nds32_expand_load_multiple interface.
19934         (store_multiple): Update nds32_expand_store_multiple interface.
19935         * config/nds32/nds32-predicates.c
19936         (nds32_valid_multiple_load_store): Rename ...
19937         (nds32_valid_multiple_load_store_p): ... to this and refine
19938         implementation.
19939         * config/nds32/predicates.md
19940         (nds32_load_multiple_and_update_address_operation): New predicate.
19941         (nds32_store_multiple_and_update_address_operation): New predicate.
19943 2018-03-04  Kito Cheng  <kito.cheng@gmail.com>
19944             Chung-Ju Wu  <jasonwucj@gmail.com>
19946         * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
19947         (combo): New attribute.
19948         * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
19950 2018-03-03  Chung-Ju Wu  <jasonwucj@gmail.com>
19952         * config/nds32/nds32.opt: Change -mcmodel= default value.
19954 2018-03-03  Kito Cheng  <kito.cheng@gmail.com>
19955             Monk Chiang  <sh.chiang04@gmail.com>
19956             Chung-Ju Wu  <jasonwucj@gmail.com>
19958         * config/nds32/constants.md (unspec_element): New enum.
19959         * config/nds32/constraints.md (Umw): New constraint.
19960         * config/nds32/nds32-intrinsic.c: Add more builtin functions.
19961         * config/nds32/nds32-intrinsic.md: Likewise.
19962         * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
19963         (nds32_valid_smw_lwm_base_p): New.
19964         (nds32_output_smw_single_word): New.
19965         (nds32_output_lmw_single_word): New.
19966         (nds32_expand_unaligned_load): New.
19967         (nds32_expand_unaligned_store): New.
19968         * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
19969         (nds32_output_smw_single_word): Declare.
19970         (nds32_output_lmw_single_word): Declare.
19971         (nds32_expand_unaligned_load): Declare.
19972         (nds32_expand_unaligned_store): Declare.
19973         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
19974         NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
19975         NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
19976         NDS32_BUILTIN_UASTORE_DW.
19977         * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
19978         predicate.
19980 2018-03-03  Monk Chiang  <sh.chiang04@gmail.com>
19981             Kito Cheng  <kito.cheng@gmail.com>
19982             Chung-Ju Wu  <jasonwucj@gmail.com>
19984         * config/nds32/nds32-intrinsic.c
19985         (nds32_expand_builtin_null_ftype_reg): Delete.
19986         (nds32_expand_builtin_reg_ftype_imm): Ditto.
19987         (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
19988         (nds32_read_argument): New.
19989         (nds32_legitimize_target): Ditto.
19990         (nds32_legitimize_argument): Ditto.
19991         (nds32_check_constant_argument): Ditto.
19992         (nds32_expand_unop_builtin): Ditto.
19993         (nds32_expand_unopimm_builtin): Ditto.
19994         (nds32_expand_binop_builtin): Ditto.
19995         (nds32_builtin_decl_impl): Ditto.
19996         (builtin_description): Ditto.
19997         (nds32_expand_builtin_impl): Rewrite with new infrastructure.
19998         (nds32_init_builtins_impl): Ditto.
19999         * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
20000         (nds32_builtin_decl): New.
20001         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
20002         * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
20004 2018-03-02  Jeff Law  <law@redhat.com>
20006         * reorg.c (stop_search_p): Handle DEBUG_INSN.
20007         (redundant_insn, fill_simple_delay_slots): Likewise.
20008         (fill_slots_from_thread): Likewise.
20009         * resource.c (mark_referenced_resources): Likewise.
20010         (mark_set_resources, find_dead_or_set_registers): Likewise.
20012 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
20014         * substring-locations.h (format_warning_va): Formatting fix for
20015         ATTRIBUTE_GCC_DIAG.
20016         (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
20017         argument.
20018         (format_warning_n_va, format_warning_at_substring_n): New prototypes.
20019         * substring-locations.c: Include intl.h.
20020         (format_warning_va): Turned into small wrapper around
20021         format_warning_n_va, renamed to ...
20022         (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
20023         rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
20024         use ngettext.
20025         (format_warning_at_substring_n): New function.
20026         * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
20027         (fmtwarn): Add ATTRIBUTE_GCC_DIAG.  Turn into a copy of
20028         format_warning_at_substring with just a shorter name instead of
20029         const function pointer.
20030         (fmtwarn_n): New function.
20031         (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
20032         appropriate, get rid of all the fmtstr temporaries, move conditionals
20033         with G_() wrapped string literals directly into fmtwarn arguments,
20034         cast dir.len to (int), formatting fixes.
20036 2018-03-02  Thomas Schwinge  <thomas@codesourcery.com>
20038         * doc/invoke.texi: Remove "Cilk Plus" references.
20040 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
20041             Richard Biener  <rguenther@suse.de>
20043         PR ipa/84628
20044         * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
20045         for error or warning attributes if CALL_FROM_THUNK_P is set.
20046         Formatting fixes.
20048 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
20050         PR target/56540
20051         * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
20052         __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
20054         PR target/56540
20055         * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
20056         __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
20058         * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
20059         instead of -1U in last predictors element's probability member.
20061 2018-03-02  Eric Botcazou  <ebotcazou@adacore.com>
20063         PR ipa/83983
20064         * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
20065         arguments if they are comparable.
20067 2018-03-02  Richard Sandiford  <richard.sandiford@linaro.org>
20069         PR tree-optimization/84634
20070         * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
20071         masks and masked_loop_p with a single loop_masks, making sure it's
20072         null for bb vectorization.
20074 2018-03-02  Richard Sandiford  <richard.sandiford@linaro.org>
20076         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
20077         (vect_analyze_data_ref_access): Use loop->safe_len rather than
20078         loop->force_vectorize to check whether there is no alias.
20080 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
20082         PR target/84614
20083         * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
20084         prototypes.
20085         * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
20086         comments.
20087         (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
20088         * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
20089         instead of a loop around prev_real_insn.
20090         * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
20091         prev_real_insn.
20093         PR inline-asm/84625
20094         * config/i386/i386.c (ix86_print_operand): Use conditional
20095         output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
20096         zero vector.
20098 2018-03-02  Richard Biener  <rguenther@suse.de>
20100         PR tree-optimization/84427
20101         * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
20102         (bitmap_set_subtract_values): Rewrite to handle multiple
20103         exprs per value.
20104         (clean): Likewise.
20105         (prune_clobbered_mems): Likewise.
20106         (phi_translate): Take edge instead of pred/phiblock.
20107         (phi_translate_1): Likewise.
20108         (phi_translate_set): Likewise.  Insert all translated
20109         exprs for a value into the set, keeping possibly multiple
20110         expressions per value.
20111         (compute_antic_aux): Adjust for phi_translate changes.
20112         When intersecting union the expressions and prune those
20113         not in the final value set, keeping possibly multiple
20114         expressions per value.  Do not use value-insertion
20115         for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
20116         all expressions.  Add verification that the value-sets
20117         only shrink during iteration.
20118         (compute_partial_antic_aux): Adjust for the phi_translate changes.
20119         (do_pre_regular_insertion): Likewise.
20120         (do_pre_partial_partial_insertion): Likewise.
20122 2018-03-02  Richard Biener  <rguenther@suse.de>
20124         PR target/82005
20125         * config/darwin.c (saved_debug_info_level): New static global.
20126         (darwin_asm_lto_start): Disable debug info generation for LTO out.
20127         (darwin_asm_lto_end): Restore debug info generation settings.
20129 2018-03-01  Martin Liska  <mliska@suse.cz>
20131         PR sanitizer/82484
20132         * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
20133         volatile arguments.
20135 2018-03-01  Richard Biener  <rguenther@suse.de>
20137         PR debug/84645
20138         * dwarf2out.c (gen_variable_die): Properly handle late VLA
20139         type annotation with LTO when debug was disabled at compile-time.
20141 2018-03-01  Matthew Fortune  <mfortune@gmail.com>
20143         * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
20144         XINT with INTVAL.
20145         (mips_final_postscan_insn): Likewise.
20147 2018-03-01  Richard Sandiford  <richard.sandiford@linaro.org>
20149         PR rtl-optimization/84528
20150         * alias.c (init_alias_target): Add commentary.
20151         (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
20152         a unique base value if the frame pointer is not eliminated
20153         to the stack pointer.
20155 2018-03-01  Tom de Vries  <tom@codesourcery.com>
20157         PR rtl-optimization/83327
20158         * lra-int.h (hard_regs_spilled_into): Declare.
20159         * lra.c (hard_regs_spilled_into): Define.
20160         (init_reg_info): Init hard_regs_spilled_into.
20161         * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
20162         * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
20163         (process_bb_lives): Handle hard_regs_spilled_into.
20164         (lra_create_live_ranges_1): Before doing liveness propagation, clear
20165         regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
20167 2018-02-28  David Edelsohn  <dje.gcc@gmail.com>
20169         * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
20170         (powerpc-ibm-aix[789]*): Default to AIX 7.2.
20171         * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
20172         * config/rs6000/aix72.h: New file.
20174 2018-02-28  Jakub Jelinek  <jakub@redhat.com>
20176         * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
20177         instead of warning_at with conditional singular and plural messages
20178         where possible.
20180         PR target/52991
20181         * stor-layout.c (update_alignment_for_field): For
20182         targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
20183         && !DECL_PACKED (field), do the alignment update, just use
20184         only desired_align instead of MAX (type_align, desired_align)
20185         as the alignment.
20186         (place_field): Don't do known_align < desired_align handling
20187         early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
20188         is non-NULL, instead do it after rli->prev_field handling and
20189         only if not within a bitfield word.  For DECL_PACKED (field)
20190         use type_align of BITS_PER_UNIT.
20192 2018-02-28  Eric Botcazou  <ebotcazou@adacore.com>
20194         * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
20195         superfluous parentheses and trailing spaces.
20197 2018-02-28  Richard Biener  <rguenther@suse.de>
20199         PR tree-optimization/84584
20200         * graphite-scop-detection.c (scop_detection::add_scop): Discard
20201         SCoPs with fake exit edge.
20203 2018-02-28  Martin Liska  <mliska@suse.cz>
20205         PR testsuite/84597
20206         * timevar.c (timer::print): Fix format to properly print 100%
20207         values.
20209 2018-02-28  Richard Biener  <rguenther@suse.de>
20211         PR middle-end/84607
20212         * genmatch.c (capture_info::walk_match): Do not mark
20213         captured expressions without operands as expr_p given
20214         they act more like predicates and should be subject to
20215         "lost tail" side-effect preserving.
20217 2018-02-28  Alexandre Oliva  <aoliva@redhat.com>
20219         PR rtl-optimization/81611
20220         * auto-inc-dec.c (attempt_change): Move dead note from
20221         mem_insn if it's the next use of regno
20222         (find_address): Take address use of reg holding
20223         non-incremented value.  Add parm to limit search to the named
20224         reg only.
20225         (merge_in_block): Attempt to use a mem insn that is the next
20226         use of the original regno.
20228 2018-02-27  Martin Sebor  <msebor@redhat.com>
20230         PR c++/83871
20231         * doc/invoke.texi (-Wmissing-attributes): New option.
20232         * print-tree.c (print_node): Handle DECL_UNINLINABLE.
20234 2018-02-27  Martin Sebor  <msebor@redhat.com>
20236         PR translation/84207
20237         * diagnostic-core.h (warning_n, error_n, inform_n): Change
20238         n argument to unsigned HOST_WIDE_INT.
20239         * diagnostic.c (warning_n, error_n, inform_n): Ditto.
20240         (diagnostic_n_impl): Ditto.  Handle arguments in excess of LONG_MAX.
20241         * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
20242         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
20244 2018-02-27  Richard Biener  <rguenther@suse.de>
20246         PR tree-optimization/84512
20247         * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
20248         Do not use the estimate returned from record_stmt_cost for
20249         the scalar iteration cost but sum properly using add_stmt_cost.
20251 2018-02-27  Richard Biener  <rguenther@suse.de>
20253         PR tree-optimization/84466
20254         * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
20255         Adjust last change to less strictly validate use operands.
20257 2018-02-27  Martin Liska  <mliska@suse.cz>
20259         PR gcov-profile/84548
20260         * gcov.c (process_file): Allow partial overlap and consider it
20261         also as group functions.
20262         (output_lines): Properly calculate range of lines for a group.
20264 2018-02-27  Martin Liska  <mliska@suse.cz>
20266         * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
20267         'ggc' suffixes.  Change first column width.
20268         (timer::print): Fix formatting of the column.
20270 2018-02-27  Alexandre Oliva  <aoliva@redhat.com>
20272         * tree-ssa-live.c (remove_unused_scope_block_p): Do not
20273         preserve inline entry blocks for the sake of debug inline
20274         entry point markers alone.
20275         (remove_unused_locals): Suggest in comments a better place to
20276         force the preservation of inline entry blocks that are
20277         otherwise unused, but do not preserve them.
20279 2018-02-26  H.J. Lu  <hongjiu.lu@intel.com>
20281         * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
20283 2018-02-26  H.J. Lu  <hongjiu.lu@intel.com>
20285         PR target/84039
20286         * config/i386/constraints.md (Bs): Replace
20287         ix86_indirect_branch_register with
20288         TARGET_INDIRECT_BRANCH_REGISTER.
20289         (Bw): Likewise.
20290         * config/i386/i386.md (indirect_jump): Likewise.
20291         (tablejump): Likewise.
20292         (*sibcall_memory): Likewise.
20293         (*sibcall_value_memory): Likewise.
20294         Peepholes of indirect call and jump via memory: Likewise.
20295         (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
20296         (*sibcall_value_GOT_32): Likewise.
20297         * config/i386/predicates.md (indirect_branch_operand): Likewise.
20298         (GOT_memory_operand): Likewise.
20299         (call_insn_operand): Likewise.
20300         (sibcall_insn_operand): Likewise.
20301         (GOT32_symbol_operand): Likewise.
20302         * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
20304 2018-02-26  Eric Botcazou  <ebotcazou@adacore.com>
20306         PR rtl-optimization/83496
20307         * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
20308         booleans to RTXes.  Call fix_reg_dead_note on every non-null element.
20309         (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
20310         redundant insn, if any.
20311         (relax_delay_slots): Likewise.
20312         (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
20314 2018-02-26  Richard Sandiford  <richard.sandiford@linaro.org>
20316         PR tree-optimization/83965
20317         * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
20318         that grouped statements are part of a reduction chain.  Return
20319         true if the statement is not marked as a reduction itself but
20320         is part of a group.
20321         (vect_recog_dot_prod_pattern): Don't check whether the statement
20322         is part of a group here.
20323         (vect_recog_sad_pattern): Likewise.
20324         (vect_recog_widen_sum_pattern): Likewise.
20326 2018-02-26  Eric Botcazou  <ebotcazou@adacore.com>
20328         PR debug/84545
20329         * final.c (rest_of_clean_state): Also look for calls inside sequences.
20331 2018-02-26  H.J. Lu  <hongjiu.lu@intel.com>
20333         PR target/84530
20334         * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
20335         the bool argument.
20336         (ix86_output_indirect_function_return): New prototype.
20337         (ix86_split_simple_return_pop_internal): Likewise.
20338         * config/i386/i386.c (indirect_return_via_cx): New.
20339         (indirect_return_via_cx_bnd): Likewise.
20340         (indirect_thunk_name): Handle return va CX_REG.
20341         (output_indirect_thunk_function): Create alias for
20342         __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
20343         (ix86_output_indirect_jmp): Remove the bool argument.
20344         (ix86_output_indirect_function_return): New function.
20345         (ix86_split_simple_return_pop_internal): Likewise.
20346         * config/i386/i386.md (*indirect_jump): Don't pass false
20347         to ix86_output_indirect_jmp.
20348         (*tablejump_1): Likewise.
20349         (simple_return_pop_internal): Change it to define_insn_and_split.
20350         Call ix86_split_simple_return_pop_internal to split it for
20351         -mfunction-return=.
20352         (simple_return_indirect_internal): Call
20353         ix86_output_indirect_function_return instead of
20354         ix86_output_indirect_jmp.
20356 2018-02-26  Jakub Jelinek  <jakub@redhat.com>
20358         PR bootstrap/84405
20359         * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
20360         memset and value initialization afterwards.
20362 2018-02-26  Christophe Lyon  <christophe.lyon@linaro.org>
20364         * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
20366 2018-02-26  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
20368         PR target/84521
20369         * common/config/aarch64/aarch64-common.c
20370         (aarch_option_optimization_table[]): Switch
20371         off fomit-frame-pointer
20373 2018-02-26  Kito Cheng  <kito.cheng@gmail.com>
20374             Chung-Ju Wu  <jasonwucj@gmail.com>
20376         * config/nds32/nds32-multiple.md (load_multiple): Disallow
20377         volatile memory.
20378         (store_multiple): Ditto.
20380 2018-02-26  Kito Cheng  <kito.cheng@gmail.com>
20382         * config.gcc: Add --with-cpu support for nds32 target.
20383         * config/nds32/nds32-opts.h (nds32_cpu_type): New.
20384         * config/nds32/nds32.opt: Add -mcpu= option.
20386 2018-02-25  Segher Boessenkool  <segher@kernel.crashing.org>
20388         * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
20389         isel=yes): Warn for these deprecated options.
20391 2018-02-23  David Edelsohn  <dje.gcc@gmail.com>
20393         * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
20394         ISA_2_5_MASKS_EMBEDDED.
20396 2018-02-23  Jakub Jelinek  <jakub@redhat.com>
20398         * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
20399         p->max as pointers rather than using iterative_hash_expr.
20401 2018-02-23  Carl Love  <cel@us.ibm.com>
20403         * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
20404         macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
20405         BU_P8V_OVERLOAD_2.
20406         * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
20407         P8V_BUILTIN_VEC_VSIGNED2.  Change VSX_BUILTIN_VEC_VUNSIGNED2 to
20408         P8V_BUILTIN_VEC_VUNSIGNED2.
20410 2018-02-22  Vladimir Makarov  <vmakarov@redhat.com>
20412         PR target/81572
20413         * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
20414         * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
20415         LRA_UNKNOWN_ALT.
20416         * lra-constraints.c (curr_insn_transform): Set up
20417         LRA_NON_CLOBBERED_ALT for moves processed on the fast path.  Use
20418         LRA_UNKNOWN_ALT.
20419         (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
20420         * lra-eliminations.c (spill_pseudos): Ditto.
20421         (process_insn_for_elimination): Ditto.
20422         * lra-lives.c (reg_early_clobber_p): Use the new macros.
20423         * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
20424         LRA_NON_CLOBBERED_ALT.
20426 2018-02-22  Martin Sebor  <msebor@redhat.com>
20428         PR tree-optimization/84480
20429         * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
20430         to maybe_diag_stxncpy_trunc.  Call it.
20431         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
20432         from gimple_fold_builtin_strcpy.  Print inlining stack.
20433         (handle_builtin_stxncpy): Print inlining stack.
20434         * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
20436 2018-02-22  H.J. Lu  <hongjiu.lu@intel.com>
20438         PR target/84176
20439         * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
20440         error when -mindirect-branch=thunk-extern, -fcf-protection=branch
20441         and -fcheck-pointer-bounds are used together.
20442         (indirect_thunk_prefix): New enum.
20443         (indirect_thunk_need_prefix): New function.
20444         (indirect_thunk_name): Replace need_bnd_p with need_prefix.  Use
20445         "_nt" instead of "_bnd" for NOTRACK prefix.
20446         (output_indirect_thunk): Replace need_bnd_p with need_prefix.
20447         (output_indirect_thunk_function): Likewise.
20448         (): Likewise.
20449         (ix86_code_end): Update output_indirect_thunk_function calls.
20450         (ix86_output_indirect_branch_via_reg): Replace
20451         ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
20452         (ix86_output_indirect_branch_via_push): Likewise.
20453         (ix86_output_function_return): Likewise.
20454         * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
20455         incompatible with -fcf-protection=branch and
20456         -fcheck-pointer-bounds.
20458 2018-02-22  Steve Ellcey  <sellcey@cavium.com>
20460         PR target/83335
20461         * config/aarch64/aarch64.c (aarch64_print_address_internal):
20462         Change gcc_assert call to output_operand_lossage.
20464 2018-02-22  Steve Ellcey  <sellcey@cavium.com>
20466         * doc/extend.texi (__builtin_extend_pointer): Document builtin.
20468 2018-02-22  DJ Delorie  <dj@redhat.com>
20469             Sebastian Perta  <sebastian.perta@renesas.com>
20470             Oleg Endo  <olegendo@gcc.gnu.org>
20472         * config/rx/rx.c (rx_rtx_costs): New function.
20473         (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
20475 2018-02-22  Thomas Preud'homme  <thomas.preudhomme@arm.com>
20477         * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
20479 2018-02-22  Martin Liska  <mliska@suse.cz>
20481         PR driver/83193
20482         * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
20483         Add "native" as a possible value.
20485 2018-02-22  Martin Liska  <mliska@suse.cz>
20487         PR driver/83193
20488         * config/i386/i386.c (ix86_option_override_internal):
20489         Add "native" as a possible value for -march and -mtune.
20491 2018-02-22  Jakub Jelinek  <jakub@redhat.com>
20493         PR target/84502
20494         * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
20495         to all type variants.
20497         PR tree-optimization/84503
20498         * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
20499         width as info->bitpos + info->bitsize - start.
20500         (merged_store_group::merge_overlapping): Simplify width computation.
20501         (check_no_overlap): New function.
20502         (imm_store_chain_info::try_coalesce_bswap): Compute expected
20503         start + width and last_order of the group, fail if check_no_overlap
20504         fails.
20505         (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
20506         to group if check_no_overlap fails.
20508 2018-02-21  Segher Boessenkool  <segher@kernel.crashing.org>
20510         * config/rs6000/altivec.md: Delete contraint arguments to
20511         define_expand, define_split, and define_peephole2, and in
20512         define_insn_and_split if always unused.
20513         * config/rs6000/darwin.md: Ditto.
20514         * config/rs6000/dfp.md: Ditto.
20515         * config/rs6000/rs6000.md: Ditto.
20516         * config/rs6000/sync.md: Ditto.
20517         * config/rs6000/vector.md: Ditto.
20518         * config/rs6000/vsx.md: Ditto.
20520 2018-02-21  Segher Boessenkool  <segher@kernel.crashing.org>
20522         * config/rs6000/altivec.md: Write output control strings as braced
20523         blocks instead of double-quoted strings.
20524         * config/rs6000/darwin.md: Ditto.
20525         * config/rs6000/rs6000.md: Ditto.
20526         * config/rs6000/vector.md: Ditto.
20527         * config/rs6000/vsx.md: Ditto.
20529 2018-02-21  Jason Merrill  <jason@redhat.com>
20531         PR c++/84314 - ICE with templates and fastcall attribute.
20532         * attribs.c (build_type_attribute_qual_variant): Remove assert.
20534 2018-02-21  Jan Hubicka  <hubicka@ucw.cz>
20536         * ipa-cp.c (determine_versionability): Fix comment typos.
20538 2018-02-21  Jan Hubicka  <hubicka@ucw.cz>
20540         PR c/84229
20541         * ipa-cp.c (determine_versionability): Do not version functions caling
20542         va_arg_pack.
20544 2018-02-21  Martin Liska  <mliska@suse.cz>
20546         PR driver/83193
20547         * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
20548         Add "native" as a possible value.
20549         * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT):  Define
20550         the macro when native cpu detection is available.
20552 2018-02-21  Martin Liska  <mliska@suse.cz>
20554         PR driver/83193
20555         * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
20556         Add "native" as a possible value.
20557         * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
20558         when native cpu detection is available.
20560 2018-02-21  Jakub Jelinek  <jakub@redhat.com>
20561             Martin Sebor  <msebor@redhat.com>
20563         PR tree-optimization/84478
20564         * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
20565         false.
20566         * gimple-fold.c (get_range_strlen): Make minlen const and assume it
20567         can't be NULL.  Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
20568         support which is conservatively correct, for 2 only stay conservative
20569         for maxlen.  Formatting and comment capitalization fixes.  Add STRICT
20570         argument to the 2 argument get_range_strlen, adjust 6 arg
20571         get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
20572         false.
20573         (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
20574         (gimple_fold_builtin_strlen): Pass true as last argument to
20575         get_range_strlen.
20577 2018-02-20  Martin Sebor  <msebor@redhat.com>
20579         PR middle-end/84095
20580         * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
20581         (builtin_memref::set_base_and_offset): Same.  Handle inner references.
20582         (builtin_memref::builtin_memref): Factor out parts into
20583         set_base_and_offset and call it.
20585 2018-02-20  Richard Sandiford  <richard.sandiford@linaro.org>
20587         PR middle-end/84406
20588         * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
20589         is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
20590         greater precision.  If to_mode is a MODE_PARTIAL_INT, stop the
20591         search at the associated MODE_INT.
20593 2018-02-20  Jeff Law  <law@redhat.com>
20595         PR middle-end/82123
20596         PR tree-optimization/81592
20597         PR middle-end/79257
20598         * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
20599         for range data rather than using global data.
20600         * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
20601         range data rather than using global data.
20602         * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
20603         pass it to children as needed.
20604         (struct directive::fmtresult): Similarly.
20605         (struct directive::set_width): Similarly.
20606         (struct directive::set_precision): Similarly.
20607         (format_integer, format_directive, parse_directive): Similarly.
20608         (format_none): Accept unnamed vr_values parameter.
20609         (format_percent, format_floating, format_character): Similarly.
20610         (format_string, format_plain): Similarly.
20611         * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
20612         the EVRP range analyzer for range data rather than using global data.
20613         * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
20614         gimple-ssa-evrp-analyze.h
20615         (class sprintf_dom_walker): Add after_dom_children member function.
20616         Add evrp_range_analyzer member.
20617         (sprintf_dom_walker::before_dom_children): Call into the EVRP
20618         range analyzer as needed.
20619         (sprintf_dom_walker::after_dom_children): New member function.
20620         * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
20621         if not optimizing.
20622         (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
20623         (evrp_range_analyzer::pop_to_marker): Likewise.
20625 2018-02-20  Richard Sandiford  <richard.sandiford@linaro.org>
20627         PR tree-optimization/84419
20628         * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
20629         with the required type if its current type is compatible but
20630         different.
20632 2018-02-20  Jakub Jelinek  <jakub@redhat.com>
20634         PR middle-end/82004
20635         * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
20636         after vectorization.
20638 2018-02-20  Martin Liska  <mliska@suse.cz>
20640         PR driver/83193
20641         * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
20642         possible values if we don't have a hint.
20644 2018-02-20  Martin Liska  <mliska@suse.cz>
20646         PR c/84310
20647         PR target/79747
20648         * final.c (shorten_branches): Build align_tab array with one
20649         more element.
20650         * opts.c (finish_options): Add alignment option limit check.
20651         (MAX_CODE_ALIGN): Likewise.
20652         (MAX_CODE_ALIGN_VALUE): Likewise.
20653         * doc/invoke.texi: Document maximum allowed option value for
20654         all -falign-* options.
20656 2018-02-19  Jakub Jelinek  <jakub@redhat.com>
20658         PR target/84146
20659         * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
20660         * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
20661         * var-tracking.c (emit_note_insn_var_location): Remove all references
20662         to NOTE_INSN_CALL_ARG_LOCATION.
20663         (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
20664         the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
20665         Use copy_rtx_if_shared.
20666         * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
20667         NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
20668         (dwarf2out_var_location): Remove handling of
20669         NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
20670         on call_insn.
20671         * final.c (final_scan_insn): Remove all references to
20672         NOTE_INSN_CALL_ARG_LOCATION.
20673         (rest_of_clean_state): Likewise.  Remove REG_CALL_ARG_LOCATION notes
20674         before dumping final insns.
20675         * except.c (emit_note_eh_region_end): Remove all references to
20676         NOTE_INSN_CALL_ARG_LOCATION.
20677         * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
20678         * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
20679         * config/arc/arc.c (hwloop_optimize): Likewise.
20680         * config/arm/arm.c (create_fix_barrier): Likewise.
20681         * config/s390/s390.c (s390_chunkify_start): Likewise.
20682         * config/sh/sh.c (find_barrier): Likewise.
20683         * config/i386/i386.c (rest_of_insert_endbranch,
20684         ix86_seh_fixup_eh_fallthru): Likewise.
20685         * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
20686         * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
20687         * config/frv/frv.c (frv_function_prologue): Likewise.
20688         * emit-rtl.c (try_split): Likewise.  Copy over REG_CALL_ARG_LOCATION
20689         reg note.
20690         (note_outside_basic_block_p): Remove all references to
20691         NOTE_INSN_CALL_ARG_LOCATION.
20692         * gengtype.c (adjust_field_rtx_def): Likewise.
20693         * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
20694         Likewise.
20695         * jump.c (cleanup_barriers, delete_related_insns): Likewise.
20696         * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
20698         PR c++/84444
20699         * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
20700         is ADDR_EXPR.
20702         PR tree-optimization/84452
20703         * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
20704         expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
20705         is NULL.
20707 2018-02-19  Martin Liska  <mliska@suse.cz>
20709         PR sanitizer/82183
20710         * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
20712 2018-02-19  Martin Liska  <mliska@suse.cz>
20713             Richard Sandiford  <richard.sandiford@linaro.org>
20715         PR tree-optimization/82491
20716         * gimple-fold.c (get_base_constructor): Make earlier bail out
20717         to prevent ubsan.
20719 2018-02-19  Carl Love  <cel@us.ibm.com>
20721         * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
20722         BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
20723         BU_P8V_OVERLOAD_1.
20724         * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
20725         P8V_BUILTIN_VEC_NEG.
20727 2018-02-19  Sebastian Perta  <sebastian.perta@renesas.com>
20729         * config/rl78/rl78.md (movdf): New define expand.
20731 2018-02-19  Martin Liska  <mliska@suse.cz>
20733         PR other/80589
20734         * doc/invoke.texi: Fix typo.
20735         * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
20737 2018-02-18  Segher Boessenkool  <segher@kernel.crashing.org>
20739         * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
20740         handle rs6000_single_float and rs6000_double_float specially for
20741         e500 family CPUs.
20743 2018-02-16  Jeff Law  <law@redhat.com>
20745         * config/rx/rx.c (add_pop_cfi_notes): New function.;
20746         (pop_regs): Use it.
20748 2018-02-16  Jakub Jelinek  <jakub@redhat.com>
20750         PR ipa/84425
20751         * ipa-inline.c (inline_small_functions): Fix a typo.
20753 2018-02-16  Nathan Sidwell  <nathan@acm.org>
20755         * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
20757 2018-02-16  Carl Love  <cel@us.ibm.com>
20759         * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
20760         Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
20761         from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
20762         * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
20763         expansion to P8V_BUILTIN_VEC_FLOAT2.
20765 2018-02-16  Vladimir Makarov  <vmakarov@redhat.com>
20767         PR rtl-optimization/70023
20768         * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
20769         src_regno into account.
20771 2018-02-16  Carl Love  <cel@us.ibm.com>
20773         * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
20774         * config/rs6000/rs6000-builtin.def: Remove macro expansion for
20775         VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
20776         * config/rs6000/rs6000.c: Remove case statements for
20777         P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
20778         P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
20779         and P9V_BUILTIN_VEC_VINSERT4B.
20780         * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
20781         P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
20782         * config/rs6000/vsx.md:
20783         * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
20784         vec_insert4b.
20786 2018-02-16  Carl Love  <cel@us.ibm.com>
20788         * config/rs6000/altivec.h: Add builtin names vec_extract4b
20789         vec_insert4b.
20790         * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
20791         definitions.
20792         * config/rs6000/rs6000-c.c: Add the definitions for
20793         P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
20794         * config/rs6000/rs6000.c (altivec_expand_builtin): Add
20795         P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
20796         * config/rs6000/vsx.md: Add define_insn extract4b.  Add define_expand
20797         definition for insert4b and define insn *insert3b_internal.
20798         * doc/extend.texi: Add documentation for vec_extract4b.
20800 2018-02-16  Nathan Sidwell  <nathan@acm.org>
20802         * doc/extend.texi (Backwards Compatibility): Mention friend
20803         injection.  Note for-scope is deprecated.
20804         * doc/invoke.texi (-ffriend-injection): Deprecate.
20806 2018-02-16  Segher Boessenkool  <segher@kernel.crashing.org>
20808         * combine.c (try_combine): When adjusting LOG_LINKS for the destination
20809         that moved to I2, also allow destinations that are a paradoxical
20810         subreg (instead of a normal reg).
20812 2018-02-16  Oleg Endo  <olegendo@gcc.gnu.org>
20814         PR target/83831
20815         * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
20816         to QImode.
20818 2018-02-16  Richard Biener  <rguenther@suse.de>
20820         PR tree-optimization/84037
20821         PR tree-optimization/84016
20822         PR target/82862
20823         * config/i386/i386.c (ix86_builtin_vectorization_cost):
20824         Adjust vec_construct for the fact we need additional higher latency
20825         128bit inserts for AVX256 and AVX512 vector builds.
20826         (ix86_add_stmt_cost): Scale vector construction cost for
20827         elementwise loads.
20829 2018-02-16  Richard Biener  <rguenther@suse.de>
20831         PR tree-optimization/84417
20832         * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
20833         the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
20834         (non_rewritable_lvalue_p): Likewise, use poly-ints.
20836 2018-02-16  Martin Liska  <mliska@suse.cz>
20838         PR sanitizer/84307
20839         * internal-fn.def (ASAN_CHECK): Set proper flags.
20840         (ASAN_MARK): Likewise.
20842 2018-02-16  Julia Koval  <julia.koval@intel.com>
20844         * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
20845         from PTA_CANNONLAKE.
20847 2018-02-16  Jakub Jelinek  <jakub@redhat.com>
20849         PR target/84272
20850         * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
20851         Use ++iter rather than iter++ for std::list iterators.
20852         (func_fma_steering::dfs): Likewise.  Don't delete nodes right away,
20853         defer deleting them until all nodes in the forest are processed.  Do
20854         free even leaf nodes.  Change to_process into auto_vec.
20856         PR bootstrap/84405
20857         * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
20858         * vec.h (vec_default_construct): Use memset instead of placement new
20859         if BROKEN_VALUE_INITIALIZATION is defined.
20860         * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
20861         memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
20862         is defined.
20864         PR rtl-optimization/83723
20865         * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
20866         * lra.c (lra_substitute_pseudo): Likewise.  If true, use
20867         gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG.  Pass DEBUG_P to
20868         recursive calls.
20869         (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
20870         callers.
20871         * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
20873 2018-02-16  Eric Botcazou  <ebotcazou@adacore.com>
20875         PR rtl-optimization/81443
20876         * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
20877         from inner REGs to paradoxical SUBREGs.
20879 2018-02-16  Richard Biener  <rguenther@suse.de>
20881         PR tree-optimization/84399
20882         * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
20883         For operands we can analyze at their definition make sure we can
20884         analyze them at each use as well.
20886 2018-02-16  Richard Biener  <rguenther@suse.de>
20888         PR tree-optimization/84190
20889         * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
20890         volatile accesses if the decl isn't volatile.
20892 2018-02-15  Jason Merrill  <jason@redhat.com>
20894         PR c++/84314 - ICE with templates and fastcall attribute.
20895         * attribs.c (build_type_attribute_qual_variant): Don't clobber
20896         TYPE_CANONICAL on an existing type.
20898 2018-02-15  Jakub Jelinek  <jakub@redhat.com>
20900         PR tree-optimization/84383
20901         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
20902         dstoff nor call operand_equal_p if dstbase is NULL.
20904         PR tree-optimization/84334
20905         * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
20906         also a CONSTANT_CLASS_P, punt.
20908 2018-02-14  Jim Wilson  <jimw@sifive.com>
20910         * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
20911         first SMALL_OPERAND check.  New local min_second_step.  Move assert
20912         to where locals are set.  Add TARGET_RVC support.
20913         * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
20915 2018-02-14  Indu Bhagat  <indu.bhagat@oracle.com>
20917         * doc/invoke.texi: Correct -Wformat-overflow code sample.
20919 2018-02-14  Martin Sebor  <msebor@redhat.com>
20921         PR tree-optimization/83698
20922         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
20923         arrays constrain the offset range to their bounds.
20924         (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
20925         (builtin_access::overlap): Avoid setting the size of overlap if it's
20926         already been set.
20927         (maybe_diag_overlap): Also consider arrays when deciding what values
20928         of offsets to include in diagnostics.
20930 2018-02-14  Martin Sebor  <msebor@redhat.com>
20932         PR c/84108
20933         * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
20934         that correspond to the kind of a declaration.
20936 2018-02-14  John David Anglin  <danglin@gcc.gnu.org>
20938         PR target/83984
20939         * config/pa/pa.md: Load address of PIC label using the linkage table
20940         if the label is nonlocal.
20942 2018-02-14  Kelvin Nilsen  <kelvin@gcc.gnu.org>
20944         * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
20945         warning message if user requests -maltivec=be.
20946         * doc/invoke.texi: Document deprecation of -maltivec=be.
20948 2018-02-14  Will Schmidt  <will_schmidt@vnet.ibm.com>
20950         PR target/84220
20951         * config/rs6000/rs6000-c.c: Update definitions for
20952         ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
20953         VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
20955 2018-02-14  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
20957         PR target/84239
20958         * config/i386/cetintrin.h: Remove _rdssp[d|q] and
20959         add _get_ssp intrinsics. Remove argument from
20960         __builtin_ia32_rdssp[d|q].
20961         * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
20962         * config/i386/i386-builtin.def: Remove argument from
20963         __builtin_ia32_rdssp[d|q].
20964         * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
20965         ix86_expand_special_args_builtin for _rdssp[d|q].
20966         * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
20967         Clear register before usage.
20968         * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
20969         Add documentation for new _get_ssp and _inc_ssp intrinsics.
20971 2018-02-14  Richard Sandiford  <richard.sandiford@linaro.org>
20973         PR tree-optimization/84357
20974         * tree-data-ref.c (object_address_invariant_in_loop_p): Check
20975         operand 1 of an ARRAY_REF too.
20977 2018-02-14  Oleg Endo  <olegendo@gcc.gnu.org>
20979         PR target/83831
20980         * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
20981         rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
20982         declarations.
20983         (set_of_reg): New struct.
20984         (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
20985         * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
20986         rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
20987         functions.
20988         * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
20989         Split into bitclr, bitset, bitinvert patterns if appropriate.
20990         (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
20991         use rx_fuse_in_memory_bitop.
20992         (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
20993         to named insn, correct maximum insn length.
20995 2018-02-14  Jozef Lawrynowicz  <jozefl.gcc@gmail.com>
20997         PR target/79242
20998         * machmode.def: Define a complex mode for PARTIAL_INT.
20999         * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
21000         MODE_PARTIAL_INT.
21001         * doc/rtl.texi: Document CSPImode.
21002         * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
21003         handling.
21004         (msp430_hard_regno_nregs_with_padding): Likewise.
21006 2018-02-13  Peter Bergner  <bergner@vnet.ibm.com>
21008         PR target/84279
21009         * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
21011 2018-02-13  Segher Boessenkool  <segher@kernel.crashing.org>
21013         PR rtl-optimization/84169
21014         * combine.c (try_combine): New variable split_i2i3.  Set it to true if
21015         we generated a parallel as new i3 and we split that to new i2 and i3
21016         instructions.  Handle split_i2i3 similar to swap_i2i3: scan the
21017         LOG_LINKs of i3 to see which of those need to link to i2 now.  Link
21018         those to i2, not i1.  Partially rewrite this scan code.
21020 2018-02-13  Jakub Jelinek  <jakub@redhat.com>
21022         PR c/82210
21023         * stor-layout.c (place_field): For variable length fields, adjust
21024         offset_align afterwards not just based on the field's alignment,
21025         but also on the size.
21027         PR middle-end/84309
21028         * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
21029         of exps and logs in the use_exp2 case.
21031 2018-02-13  Jeff Law  <law@redhat.com>
21033         * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
21034         entry for "vector".
21036         * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
21037         ARGS as unused.
21039 2018-02-13  Alexandre Oliva  <aoliva@redhat.com>
21041         PR debug/84342
21042         PR debug/84319
21043         * common.opt (gas-loc-support, gas-locview-support): New.
21044         (ginline-points, ginternal-reset-location-views): New.
21045         * doc/invoke.texi: Document them.  Use @itemx where intended.
21046         (gvariable-location-views): Adjust.
21047         * target.def (reset_location_view): New.
21048         * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
21049         (TARGET_RESET_LOCATION_VIEW): New.
21050         * doc/tm.texi: Rebuilt.
21051         * dwarf2out.c (dwarf2out_default_as_loc_support): New.
21052         (dwarf2out_default_as_locview_support): New.
21053         (output_asm_line_debug_info): Use option variables.
21054         (dwarf2out_maybe_output_loclist_view_pair): Likewise.
21055         (output_loc_list): Likewise.
21056         (add_high_low_attributes): Check option variables.
21057         Don't output entry view attribute in strict mode.
21058         (gen_inlined_subroutine_die): Check option variables.
21059         (dwarf2out_inline_entry): Likewise.
21060         (init_sections_and_labels): Likewise.
21061         (dwarf2out_early_finish): Likewise.
21062         (maybe_reset_location_view): New, from...
21063         (dwarf2out_var_location): ... here.  Call it.
21064         * debug.h (dwarf2out_default_as_loc_support): Declare.
21065         (dwarf2out_default_as_locview_support): Declare.
21066         * hooks.c (hook_int_rtx_insn_0): New.
21067         * hooks.h (hook_int_rtx_insn_0): Declare.
21068         * toplev.c (process_options): Take -gas-loc-support and
21069         -gas-locview-support from dwarf2out.  Enable
21070         -gvariable-location-views by default only with locview
21071         assembler support.  Enable -ginternal-reset-location-views by
21072         default only if the target defines the corresponding hook.
21073         Enable -ginline-points by default if location views are
21074         enabled; force it disabled if statement frontiers are
21075         disabled.
21076         * tree-inline.c (expand_call_inline): Check option variables.
21077         * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
21079 2018-02-13  Richard Sandiford  <richard.sandiford@linaro.org>
21081         PR tree-optimization/84321
21082         * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
21083         handling.  Also check whether the anti-range contains any values
21084         that satisfy the mask; switch to a VR_RANGE if not.
21086 2018-02-13  Paolo Bonzini  <bonzini@gnu.org>
21088         PR sanitizer/84340
21089         * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
21091 2018-02-13  Martin Jambor  <mjambor@suse.cz>
21093         PR c++/83990
21094         * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
21095         of call statements, also set location of a load to a temporary.
21097 2018-02-13  Sebastian Perta  <sebastian.perta@renesas.com>
21099         * config/rl78/rl78.c (add_vector_labels): New function.
21100         * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
21101         * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
21102         * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
21103         which checks that no arguments are passed.
21104         * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
21105         * doc/extend.texi: Documentation for the new attribute.
21107 2018-02-13  Andreas Schwab  <schwab@suse.de>
21109         * config/riscv/linux.h (CPP_SPEC): Define.
21111 2018-02-13  Jakub Jelinek  <jakub@redhat.com>
21113         PR target/84335
21114         * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
21115         OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
21116         OPTION_MASK_ISA_AES as first argument to def_builtin_const
21117         for AES builtins.  Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
21118         instead of OPTION_MASK_ISA_PCLMUL as first argument to
21119         def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
21120         * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
21121         temporarily for AES and PCLMUL builtins.
21123         PR tree-optimization/84339
21124         * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
21125         ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
21126         Formatting fixes.
21128         PR middle-end/84309
21129         * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
21130         exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
21131         * generic-match-head.c (canonicalize_math_after_vectorization_p): New
21132         inline function.
21133         * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
21134         inline function.
21135         * omp-simd-clone.h: New file.
21136         * omp-simd-clone.c: Include omp-simd-clone.h.
21137         (expand_simd_clones): No longer static.
21138         * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
21139         cgraph.h and omp-simd-clone.h.
21140         (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
21141         (vect_recog_widen_shift_pattern): Formatting fix.
21142         (vect_pattern_recog_1): Don't check optab for calls.
21144         PR target/84336
21145         * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
21146         operands[2] into a REG before using gen_lowpart on it.
21148 2018-02-12  Jeff Law  <law@redhat.com>
21150         PR target/83760
21151         * config/sh/sh.c (find_barrier): Consider a sibling call
21152         a barrier as well.
21154         * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
21155         successfully back substituting a reg.
21157 2018-02-12  Richard Biener  <rguenther@suse.de>
21159         PR tree-optimization/84037
21160         * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
21161         parameter, move visited init to caller.
21162         (vect_slp_analyze_operations): Separate cost from validity
21163         check, initialize visited once for all instances.
21164         (vect_schedule_slp): Analyze map to CSE vectorized nodes once
21165         for all instances.
21166         * tree-vect-stmts.c (vect_model_simple_cost): Make early
21167         out an assert.
21168         (vect_model_promotion_demotion_cost): Likewise.
21169         (vectorizable_bswap): Guard cost modeling with !slp_node
21170         instead of !PURE_SLP_STMT to avoid double-counting on hybrid
21171         SLP stmts.
21172         (vectorizable_call): Likewise.
21173         (vectorizable_conversion): Likewise.
21174         (vectorizable_assignment): Likewise.
21175         (vectorizable_shift): Likewise.
21176         (vectorizable_operation): Likewise.
21177         (vectorizable_store): Likewise.
21178         (vectorizable_load): Likewise.
21179         (vectorizable_condition): Likewise.
21180         (vectorizable_comparison): Likewise.
21182 2018-02-12  Paolo Bonzini  <bonzini@gnu.org>
21184         PR sanitizer/84307
21185         * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
21186         (ASAN_MARK): Fix fnspec to account for return value, change pointer
21187         argument from 'R' to 'W' so that the pointed-to datum is clobbered.
21189 2018-02-08  Jan Hubicka  <hubicka@ucw.cz>
21191         PR middle-end/83665
21192         * params.def (inline-min-speedup): Increase from 8 to 15.
21193         (max-inline-insns-auto): Decrease from 40 to 30.
21194         * ipa-split.c (consider_split): Add some buffer for function to
21195         be considered inlining candidate.
21196         * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
21197         default values.
21199 2018-02-12  Richard Biener  <rguenther@suse.de>
21201         PR tree-optimization/84037
21202         * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
21203         matched stmts if we cannot swap the non-matched ones.
21205 2018-02-12  Olga Makhotina  <olga.makhotina@intel.com>
21207         * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
21208         _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
21209         _mm_maskz_scalef_round_ss): New intrinsics.
21210         (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
21211         * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
21212         __builtin_ia32_scalefss_round): Remove.
21213         (__builtin_ia32_scalefsd_mask_round,
21214         __builtin_ia32_scalefss_mask_round): New intrinsics.
21215         * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
21216         (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
21217         ((match_operand:VF_128 2 "<round_nimm_predicate>"
21218         "<round_constraint>")): Changed to ...
21219         ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
21220         "<round_scalar_constraint>")): ... this.
21221         ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
21222         %0, %1, %2<round_op3>}"): Changed to ...
21223         ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
21224         %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
21225         %2<round_scalar_mask_op3>}"): ... this.
21226         * config/i386/subst.md (round_scalar_nimm_predicate): New.
21228 2018-02-12  Olga Makhotina  <olga.makhotina@intel.com>
21230         * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
21231         (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
21232         (_mm_maskz_sqrt_round_ss): New intrinsics.
21233         (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
21234         (__builtin_ia32_sqrtsd_mask_round)
21235         (__builtin_ia32_sqrtss_mask_round): New builtins.
21236         * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
21237         (__builtin_ia32_sqrtss_round): Remove.
21238         (__builtin_ia32_sqrtsd_mask_round)
21239         (__builtin_ia32_sqrtss_mask_round): New builtins.
21240         * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
21241         (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
21242         ((match_operand:VF_128 1 "vector_operand"
21243         "xBm,<round_constraint>")): Changed to ...
21244         ((match_operand:VF_128 1 "vector_operand"
21245         "xBm,<round_scalar_constraint>")): ... this.
21246         (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
21247         %0, %2, %<iptr>1<round_op3>}): Changed to ...
21248         (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
21249         %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
21250         %<iptr>1<round_scalar_mask_op3>}): ... this.
21251         ((set_attr "prefix" "<round_prefix>")): Changed to ...
21252         ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
21254 2018-02-11  Steven Munroe  <munroesj@gcc.gnu.org>
21256         PR target/84266
21257         * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
21258         Cast vec_cmpeq result to correct type.
21259         * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
21260         Cast vec_cmpgt result to correct type.
21262 2018-02-11  Alexandre Oliva  <aoliva@redhat.com>
21264         * final.c (final_scan_insn_1): Renamed from...
21265         (final_scan_insn): ... this.  New wrapper, to recover
21266         seen from the outermost call in recursive ones.
21267         * config/sparc/sparc.c (output_return): Drop seen from call.
21268         (output_sibcall): Likewise.
21269         * config/visium/visium.c (output_branch): Likewise.
21271 2018-02-10  John David Anglin  <danglin@gcc.gnu.org>
21273         * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
21274         function label.
21276 2018-02-10  Alan Modra  <amodra@gmail.com>
21278         PR target/84300
21279         * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
21280         Specify LR as an input.
21282 2018-02-10  Jakub Jelinek  <jakub@redhat.com>
21284         PR sanitizer/83987
21285         * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
21286         remove_member_access_dummy_vars): New functions.
21287         (lower_omp_for, lower_omp_taskreg, lower_omp_target,
21288         lower_omp_1, execute_lower_omp): Use them.
21290         PR rtl-optimization/84308
21291         * shrink-wrap.c (spread_components): Release todo vector.
21293 2018-02-09  Vladimir Makarov  <vmakarov@redhat.com>
21295         PR rtl-optimization/57193
21296         * ira-color.c (struct allocno_color_data): Add member
21297         conflict_allocno_hard_prefs.
21298         (update_conflict_allocno_hard_prefs): New.
21299         (bucket_allocno_compare_func): Add a preference based on
21300         conflict_allocno_hard_prefs.
21301         (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
21302         (color_allocnos): Remove a dead code.  Initiate
21303         conflict_allocno_hard_prefs.  Call update_costs_from_prefs.
21305 2018-02-09  Jakub Jelinek  <jakub@redhat.com>
21307         PR target/84226
21308         * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
21309         constraint from =wa to wa.  Avoid a subreg on the output operand,
21310         instead use a pseudo and subreg it in a move.
21311         (p9_xxbrd_<mode>): Changed to ...
21312         (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
21313         (p9_xxbrd_v2df): New expander.
21314         (p9_xxbrw_<mode>): Changed to ...
21315         (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
21316         (p9_xxbrw_v4sf): New expander.
21318 2018-02-09  Sebastian Perta  <sebastian.perta@renesas.com>
21320         * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
21322 2018-02-09  Peter Bergner  <bergner@vnet.ibm.com>
21324         PR target/83926
21325         * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
21326         multiply in 32-bit mode.
21327         (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
21328         (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
21329         mode.
21331 2018-02-09  Sebastian Perta  <sebastian.perta@renesas.com>
21333         * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
21334         to allow or block "symbol_ref" depending on the value of TARGET_JSR.
21335         * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
21336         * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
21338 2018-02-09  Pierre-Marie de Rodat  <derodat@adacore.com>
21340         PR lto/84213
21341         * dwarf2out.c (is_trivial_indirect_ref): New function.
21342         (dwarf2out_late_global_decl): Do not generate a location
21343         attribute for variables that have a non-trivial DECL_VALUE_EXPR
21344         and that are not defined in the current unit.
21346 2018-02-09  Eric Botcazou  <ebotcazou@adacore.com>
21348         * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
21349         instead of a libcall for UNORDERED.
21351 2018-02-09  Tamar Christina  <tamar.christina@arm.com>
21353         PR target/82641
21354         * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
21355         __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
21357 2018-02-09  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
21359         PR target/PR84295
21360         * config/s390/s390.c (s390_set_current_function): Invoke
21361         s390_indirect_branch_settings also if fndecl didn't change.
21363 2018-02-09  Alexandre Oliva  <aoliva@redhat.com>
21365         * config/rs6000/rs6000.md (blockage): Set length to zero.
21367 2018-02-09  Eric Botcazou  <ebotcazou@adacore.com>
21369         * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
21371 2018-02-09  Jakub Jelinek  <jakub@redhat.com>
21373         PR sanitizer/84285
21374         * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
21375         STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
21376         -static-lib*san.
21378         PR debug/84252
21379         * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
21380         PARALLEL incoming that failed vt_get_decl_and_offset check.
21382         PR middle-end/84237
21383         * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
21384         * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
21385         TREE_READONLY bit.
21386         (get_variable_section): For decls in named .bss* sections pass true as
21387         second argument to bss_initializer_p.
21389 2018-02-09  Marek Polacek  <polacek@redhat.com>
21390             Jakub Jelinek  <jakub@redhat.com>
21392         PR c++/83659
21393         * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
21394         Formatting fixes.  Verify first that tree_fits_poly_int64_p (op01).
21395         Sync some changes from cxx_fold_indirect_ref.
21397 2018-02-09  Alexandre Oliva  <aoliva@redhat.com>
21399         * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
21400         markers.
21401         * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
21402         (BLOCK_INLINE_ENTRY_LABEL): New.
21403         (dwarf2out_var_location): Disregard inline entry markers.
21404         (inline_entry_data): New struct.
21405         (inline_entry_data_hasher): New hashtable type.
21406         (inline_entry_data_hasher::hash): New.
21407         (inline_entry_data_hasher::equal): New.
21408         (inline_entry_data_table): New variable.
21409         (add_high_low_attributes): Add DW_AT_entry_pc and
21410         DW_AT_GNU_entry_view attributes if a pending entry is found
21411         in inline_entry_data_table.  Add old entry_pc attribute only
21412         if debug nonbinding markers are disabled.
21413         (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
21414         markers are enabled.
21415         (block_within_block_p, dwarf2out_inline_entry): New.
21416         (dwarf2out_finish): Check that no entries remained in
21417         inline_entry_data_table.
21418         * final.c (reemit_insn_block_notes): Handle inline entry notes.
21419         (final_scan_insn, notice_source_line): Likewise.
21420         (rest_of_clean_state): Skip inline entry markers.
21421         * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
21422         markers.
21423         * gimple.c (gimple_build_debug_inline_entry): New.
21424         * gimple.h (enum gimple_debug_subcode): Add
21425         GIMPLE_DEBUG_INLINE_ENTRY.
21426         (gimple_build_debug_inline_entry): Declare.
21427         (gimple_debug_inline_entry_p): New.
21428         (gimple_debug_nonbind_marker_p): Adjust.
21429         * insn-notes.def (INLINE_ENTRY): New.
21430         * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
21431         inline entry marker notes.
21432         (print_insn): Likewise.
21433         * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
21434         (INSN_DEBUG_MARKER_KIND): Likewise.
21435         (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
21436         * tree-inline.c (expand_call_inline): Build and insert
21437         debug_inline_entry stmt.
21438         * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
21439         inline entry blocks early, if nonbind markers are enabled.
21440         (dump_scope_block): Dump fragment info.
21441         * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
21442         * doc/gimple.texi (gimple_debug_inline_entry_p): New.
21443         (gimple_build_debug_inline_entry): New.
21444         * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
21445         Enable/disable inline entry points too.
21446         * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
21447         (DEBUG_INSN): Describe inline entry markers.
21449         * common.opt (gvariable-location-views): New.
21450         (gvariable-location-views=incompat5): New.
21451         * config.in: Rebuilt.
21452         * configure: Rebuilt.
21453         * configure.ac: Test assembler for view support.
21454         * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
21455         * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
21456         * dwarf2out.c (var_loc_view): New typedef.
21457         (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
21458         (dwarf2out_locviews_in_attribute): New.
21459         (dwarf2out_locviews_in_loclist): New.
21460         (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
21461         (enum dw_line_info_opcode): Add LI_adv_address.
21462         (struct dw_line_info_table): Add view.
21463         (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
21464         (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
21465         (zero_view_p): New variable.
21466         (ZERO_VIEW_P): New macro.
21467         (output_asm_line_debug_info): New.
21468         (struct var_loc_node): Add view.
21469         (add_AT_view_list, AT_loc_list): New.
21470         (add_var_loc_to_decl): Add view param.  Test it against last.
21471         (new_loc_list): Add view params.  Record them.
21472         (AT_loc_list_ptr): Handle loc and view lists.
21473         (view_list_to_loc_list_val_node): New.
21474         (print_dw_val): Handle dw_val_class_view_list.
21475         (size_of_die): Likewise.
21476         (value_format): Likewise.
21477         (loc_list_has_views): New.
21478         (gen_llsym): Set vl_symbol too.
21479         (maybe_gen_llsym, skip_loc_list_entry): New.
21480         (dwarf2out_maybe_output_loclist_view_pair): New.
21481         (output_loc_list): Output view list or entries too.
21482         (output_view_list_offset): New.
21483         (output_die): Handle dw_val_class_view_list.
21484         (output_dwarf_version): New.
21485         (output_compilation_unit_header): Use it.
21486         (output_skeleton_debug_sections): Likewise.
21487         (output_rnglists, output_line_info): Likewise.
21488         (output_pubnames, output_aranges): Update version comments.
21489         (output_one_line_info_table): Output view numbers in asm comments.
21490         (dw_loc_list): Determine current endview, pass it to new_loc_list.
21491         Call maybe_gen_llsym.
21492         (loc_list_from_tree_1): Adjust.
21493         (add_AT_location_description): Create view list attribute if
21494         needed, check it's absent otherwise.
21495         (convert_cfa_to_fb_loc_list): Adjust.
21496         (maybe_emit_file): Call output_asm_line_debug_info for test.
21497         (dwarf2out_var_location): Reset views as needed.  Precompute
21498         add_var_loc_to_decl args.  Call get_attr_min_length only if we have the
21499         attribute.  Set view.
21500         (new_line_info_table): Reset next view.
21501         (set_cur_line_info_table): Call output_asm_line_debug_info for test.
21502         (dwarf2out_source_line): Likewise.  Output view resets and labels to
21503         the assembler, or select appropriate line info opcodes.
21504         (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
21505         (optimize_string_length): Catch it.  Adjust.
21506         (resolve_addr): Copy vl_symbol along with ll_symbol.  Handle
21507         dw_val_class_view_list, and remove it if no longer needed.
21508         (hash_loc_list): Hash view numbers.
21509         (loc_list_hasher::equal): Compare them.
21510         (optimize_location_lists): Check whether a view list symbol is
21511         needed, and whether the locview attribute is present, and
21512         whether they match.  Remove the locview attribute if no longer
21513         needed.
21514         (index_location_lists): Call skip_loc_list_entry for test.
21515         (dwarf2out_finish): Call output_asm_line_debug_info for test.
21516         Use output_dwarf_version.
21517         * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
21518         (struct dw_val_node): Add val_view_list.
21519         * final.c (SEEN_NEXT_VIEW): New.
21520         (set_next_view_needed): New.
21521         (clear_next_view_needed): New.
21522         (maybe_output_next_view): New.
21523         (final_start_function): Rename to...
21524         (final_start_function_1): ... this.  Take pointer to FIRST,
21525         add SEEN parameter.  Emit param bindings in the initial view.
21526         (final_start_function): Reintroduce SEEN-less interface.
21527         (final): Rename to...
21528         (final_1): ... this.  Take SEEN parameter.  Output final pending
21529         next view at the end.
21530         (final): Reintroduce seen-less interface.
21531         (final_scan_insn): Output pending next view before switching
21532         sections or ending a block.  Mark the next view as needed when
21533         outputting variable locations.  Notify debug backend of section
21534         changes, and of location view changes.
21535         (rest_of_handle_final): Adjust.
21536         * toplev.c (process_options): Autodetect value for debug variable
21537         location views option.  Warn on incompat5 without -gdwarf-5.
21538         * doc/invoke.texi (gvariable-location-views): New.
21539         (gvariable-location-views=incompat5): New.
21540         (gno-variable-location-views): New.
21542 2018-02-08  David Malcolm  <dmalcolm@redhat.com>
21544         PR tree-optimization/84136
21545         * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
21546         that the result of find_edge is non-NULL.
21548 2018-02-08  Sergey Shalnov  <sergey.shalnov@intel.com>
21550         PR target/83008
21551         * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
21552         storing integer register in SImode.  Fix cost of 256 and 512
21553         byte aligned SSE register store.
21555 2018-02-08  Sergey Shalnov  <sergey.shalnov@intel.com>
21557         * config/i386/i386.c (ix86_multiplication_cost): Fix
21558         multiplication cost for TARGET_AVX512DQ.
21560 2018-02-08  Marek Polacek  <polacek@redhat.com>
21562         PR tree-optimization/84238
21563         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
21564         get_range_strlen.
21566 2018-02-08  Richard Sandiford  <richard.sandiford@linaro.org>
21568         PR tree-optimization/84265
21569         * tree-vect-stmts.c (vectorizable_store): Don't treat
21570         VMAT_CONTIGUOUS accesses as grouped.
21571         (vectorizable_load): Likewise.
21573 2018-02-08  Richard Sandiford  <richard.sandiford@linaro.org>
21575         PR tree-optimization/81635
21576         * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
21577         * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
21578         (test_round_for_mask): New functions.
21579         (wide_int_cc_tests): Call test_round_for_mask.
21580         * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
21581         * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
21582         * tree-data-ref.c (split_constant_offset_1): Use it to refine the
21583         range returned by get_range_info.
21585 2018-02-08  Jan Hubicka  <hubicka@ucw.cz>
21587         PR ipa/81360
21588         * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
21589         * symtab.c: Include builtins.h
21590         (symtab_node::output_to_lto_symbol_table_p): Move here
21591         from lto-streamer-out.c:output_symbol_p.
21592         * lto-streamer-out.c (write_symbol): Turn early exit to assert.
21593         (output_symbol_p): Move all logic to symtab.c
21594         (produce_symtab): Update.
21596 2018-02-08  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
21598         * config/s390/s390-opts.h (enum indirect_branch): Define.
21599         * config/s390/s390-protos.h (s390_return_addr_from_memory)
21600         (s390_indirect_branch_via_thunk)
21601         (s390_indirect_branch_via_inline_thunk): Add function prototypes.
21602         (enum s390_indirect_branch_type): Define.
21603         * config/s390/s390.c (struct s390_frame_layout, struct
21604         machine_function): Remove.
21605         (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
21606         (indirect_branch_table_label_no, indirect_branch_table_name):
21607         Define variables.
21608         (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
21609         (enum s390_indirect_branch_option): Define.
21610         (s390_return_addr_from_memory): New function.
21611         (s390_handle_string_attribute): New function.
21612         (s390_attribute_table): Add new attribute handler.
21613         (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
21614         (s390_indirect_branch_via_thunk): New function.
21615         (s390_indirect_branch_via_inline_thunk): New function.
21616         (s390_function_ok_for_sibcall): When jumping via thunk disallow
21617         sibling call optimization for non z10 compiles.
21618         (s390_emit_call): Force indirect branch target to be a single
21619         register.  Add r1 clobber for non-z10 compiles.
21620         (s390_emit_epilogue): Emit return jump via return_use expander.
21621         (s390_reorg): Handle JUMP_INSNs as execute targets.
21622         (s390_option_override_internal): Perform validity checks for the
21623         new command line options.
21624         (s390_indirect_branch_attrvalue): New function.
21625         (s390_indirect_branch_settings): New function.
21626         (s390_set_current_function): Invoke s390_indirect_branch_settings.
21627         (s390_output_indirect_thunk_function):  New function.
21628         (s390_code_end): Implement target hook.
21629         (s390_case_values_threshold): Implement target hook.
21630         (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
21631         macros.
21632         * config/s390/s390.h (struct s390_frame_layout)
21633         (struct machine_function): Move here from s390.c.
21634         (TARGET_INDIRECT_BRANCH_NOBP_RET)
21635         (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
21636         (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
21637         (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
21638         (TARGET_INDIRECT_BRANCH_NOBP_CALL)
21639         (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
21640         (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
21641         (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
21642         (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
21643         * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
21644         (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
21645         (mnemonic attribute): Add values which aren't recognized
21646         automatically.
21647         ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
21648         pattern for branch conversion.  Fix mnemonic attribute.
21649         ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
21650         indirect branch via thunk if requested.
21651         ("indirect_jump", "<code>"): Expand patterns for branch conversion.
21652         ("*indirect_jump"): Disable for branch conversion using out of
21653         line thunks.
21654         ("indirect_jump_via_thunk<mode>_z10")
21655         ("indirect_jump_via_thunk<mode>")
21656         ("indirect_jump_via_inlinethunk<mode>_z10")
21657         ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
21658         ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
21659         ("casesi_jump_via_inlinethunk<mode>_z10")
21660         ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
21661         ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
21662         ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
21663         ("*indirect2_jump"): Disable for branch conversion.
21664         ("casesi_jump"): Turn into expander and expand patterns for branch
21665         conversion.
21666         ("return_use"): New expander.
21667         ("*return"): Emit return via thunk and rename it to ...
21668         ("*return<mode>"): ... this one.
21669         * config/s390/s390.opt: Add new options and and enum for the
21670         option values.
21672 2018-02-08  Richard Sandiford  <richard.sandiford@linaro.org>
21674         * lra-constraints.c (match_reload): Unconditionally use
21675         gen_lowpart_SUBREG, rather than selecting between that
21676         and equivalent gen_rtx_SUBREG code.
21678 2018-02-08  Richard Biener  <rguenther@suse.de>
21680         PR tree-optimization/84233
21681         * tree-ssa-phiprop.c (propagate_with_phi): Use separate
21682         changed flag instead of boguously re-using phi_inserted.
21684 2018-02-08  Martin Jambor  <mjambor@suse.cz>
21686         * hsa-gen.c (get_symbol_for_decl): Set program allocation for
21687         static local variables.
21689 2018-02-08  Richard Biener  <rguenther@suse.de>
21691         PR tree-optimization/84278
21692         * tree-vect-stmts.c (vectorizable_store): When looking for
21693         smaller vector types to perform grouped strided loads/stores
21694         make sure the mode is supported by the target.
21695         (vectorizable_load): Likewise.
21697 2018-02-08  Wilco Dijkstra  <wdijkstr@arm.com>
21699         * config/aarch64/aarch64.c (aarch64_components_for_bb):
21700         Increase LDP/STP opportunities by adding adjacent callee-saves.
21702 2018-02-08  Wilco Dijkstra  <wdijkstr@arm.com>
21704         PR rtl-optimization/84068
21705         PR rtl-optimization/83459
21706         * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
21708 2018-02-08  Aldy Hernandez  <aldyh@redhat.com>
21710         PR tree-optimization/84224
21711         * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
21712         * calls.c (gimple_alloca_call_p): Only return TRUE when we have
21713         non-zero arguments.
21715 2018-02-07  Iain Sandoe  <iain@codesourcery.com>
21717         PR target/84113
21718         * config/rs6000/altivec.md (*restore_world): Remove LR use.
21719         * config/rs6000/predicates.md (restore_world_operation): Adjust op
21720         count, remove one USE.
21722 2018-02-07  Michael Meissner  <meissner@linux.vnet.ibm.com>
21724         * doc/install.texi (Configuration): Document the
21725         --with-long-double-format={ibm,ieee} PowerPC configuration
21726         options.
21728         PR target/84154
21729         * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
21730         Convert from define_expand to be define_insn_and_split.  Rework
21731         float/double/_Float128 conversions to QI/HI/SImode to work with
21732         both ISA 2.07 (power8) or ISA 3.0 (power9).  Fix regression where
21733         conversions to QI/HImode types did a store and then a load to
21734         truncate the value.  For conversions to VSX registers, don't split
21735         the insn, instead emit the code directly.  Use the code iterator
21736         any_fix to combine signed and unsigned conversions.
21737         (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
21738         (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
21739         (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
21740         (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
21741         (fix_<mode>di2_hw): Likewise.
21742         (fixuns_<mode>di2_hw): Likewise.
21743         (fix_<mode>si2_hw): Likewise.
21744         (fixuns_<mode>si2_hw): Likewise.
21745         (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
21746         (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
21747         (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
21748         fix<uns>_trunc<SFDF:mode>si2_p8.
21749         (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer used.
21750         (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
21751         (fix<uns>_<mode>_mem): Likewise.
21752         (fctiw<u>z_<mode>_mem): Likewise.
21753         (fix<uns>_<mode>_mem): Likewise.
21754         (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
21755         the register allocator from doing a direct move to the GPRs to do
21756         a store, and instead use the ISA 3.0 store byte/half-word from
21757         vector register instruction.  For IEEE 128-bit floating point,
21758         also optimize stores of 32-bit ints.
21759         (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
21761 2018-02-07  Alan Hayward  <alan.hayward@arm.com>
21763         * genextract.c (push_pathstr_operand): New function to support
21764         [a-zA-Z].
21765         (walk_rtx): Call push_pathstr_operand.
21766         (print_path): Support [a-zA-Z].
21768 2018-02-07  Richard Biener  <rguenther@suse.de>
21770         PR tree-optimization/84037
21771         * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
21772         (cse_and_gimplify_to_preheader): Declare.
21773         (vect_get_place_in_interleaving_chain): Likewise.
21774         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
21775         ivexpr_map.
21776         (_loop_vec_info::~_loop_vec_info): Delete it.
21777         (cse_and_gimplify_to_preheader): New function.
21778         * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
21779         * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
21780         (vectorizable_load): Likewise.  For grouped stores always base
21781         the IV on the first element.
21782         * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
21783         condition before gimplifying.
21785 2018-02-07  Jakub Jelinek  <jakub@redhat.com>
21787         * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
21788         *DIV_EXPR and *MOD_EXPR.
21790 2018-02-07  H.J. Lu  <hongjiu.lu@intel.com>
21792         PR target/84248
21793         * config/i386/i386.c (ix86_option_override_internal): Mask out
21794         the CF_SET bit when checking -fcf-protection.
21796 2018-02-07  Tom de Vries  <tom@codesourcery.com>
21798         PR libgomp/84217
21799         * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
21800         enough.
21802 2018-02-07  Richard Biener  <rguenther@suse.de>
21804         PR tree-optimization/84204
21805         * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
21806         this place.
21808         PR tree-optimization/84205
21809         * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
21810         special-case isl_ast_op_zdiv_r.
21812         PR tree-optimization/84223
21813         * graphite-scop-detection.c (gather_bbs::before_dom_children):
21814         Only add conditions from within the region.
21815         (gather_bbs::after_dom_children): Adjust.
21817 2018-02-07  Georg-Johann Lay  <avr@gjlay.de>
21819         PR target/84209
21820         * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
21821         * config/avr/avr.md: Only post-reload split REG-REG moves if
21822         either register is GENERAL_REG_P.
21824 2018-02-07  Jakub Jelinek  <jakub@redhat.com>
21826         PR tree-optimization/84235
21827         * tree-ssa-scopedtables.c
21828         (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
21829         if the subtraction is performed in floating point type where NaNs are
21830         honored.  For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
21831         build 1.  Formatting fix.
21833 2018-02-06  Jakub Jelinek  <jakub@redhat.com>
21835         PR target/84146
21836         * config/i386/i386.c (rest_of_insert_endbranch): Only skip
21837         NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
21838         and skip it regardless of bb boundaries.  Use CALL_P macro,
21839         don't test INSN_P (insn) together with CALL_P or JUMP_P check
21840         unnecessarily, formatting fix.
21842 2018-02-06  Michael Collison  <michael.collison@arm.com>
21844         * config/arm/thumb2.md:
21845         (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
21846         (*thumb_mov_notscc): Ditto.
21848 2018-02-06  Michael Meissner  <meissner@linux.vnet.ibm.com>
21850         PR target/84154
21851         * config/rs6000/rs6000.md (su code attribute): Use "u" for
21852         unsigned_fix, not "s".
21854 2018-02-06  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
21856         * configure.ac (gcc_fn_eh_frame_ro): New function.
21857         (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
21858         correct .eh_frame permissions.
21859         * configure: Regenerate.
21861 2018-02-06  Andrew Jenner  <andrew@codeourcery.com>
21863         * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
21864         irrelevant options.
21866 2018-02-06  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
21868         * config/rs6000/rs6000.c (rs6000_option_override_internal):
21869         Display warning message for -mno-speculate-indirect-jumps.
21871 2018-02-06  Andrew Jenner  <andrew@codesourcery.com>
21873         * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
21874         Undocumented.
21875         * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
21877 2018-02-06  Aldy Hernandez  <aldyh@redhat.com>
21879         PR tree-optimization/84225
21880         * tree-eh.c (find_trapping_overflow): Only call
21881         operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
21883 2018-02-06  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
21885         PR target/84145
21886         * config/i386/i386.c: Reimplement the check of possible options
21887         -mibt/-mshstk conbination. Change error messages.
21888         * doc/invoke.texi: Fix a typo: remove extra '='.
21890 2018-02-06  Marek Polacek  <polacek@redhat.com>
21892         PR tree-optimization/84228
21893         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
21895 2018-02-06  Tamar Christina  <tamar.christina@arm.com>
21897         PR target/82641
21898         * config/arm/arm.c (arm_print_asm_arch_directives): Record already
21899         emitted arch directives.
21900         * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
21901         __ARM_FEATURE_COPROC before changing architectures.
21903 2018-02-06  Richard Biener  <rguenther@suse.de>
21905         * config/i386/i386.c (print_reg): Fix typo.
21906         (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
21908 2018-02-06  Eric Botcazou  <ebotcazou@adacore.com>
21910         * configure: Regenerate.
21912 2018-02-05  Martin Sebor  <msebor@redhat.com>
21914         PR tree-optimization/83369
21915         * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
21916         inlining context.
21918 2018-02-05  Martin Liska  <mliska@suse.cz>
21920         * doc/invoke.texi: Cherry-pick upstream r323995.
21922 2018-02-05  Richard Sandiford  <richard.sandiford@linaro.org>
21924         * ira.c (ira_init_register_move_cost): Adjust comment.
21926 2018-02-05  Martin Liska  <mliska@suse.cz>
21928         PR gcov-profile/84137
21929         * doc/gcov.texi: Fix typo in documentation.
21931 2018-02-05  Martin Liska  <mliska@suse.cz>
21933         PR gcov-profile/83879
21934         * doc/gcov.texi: Document necessity of --dynamic-list-data when
21935         using dlopen functionality.
21937 2018-02-05  Olga Makhotina  <olga.makhotina@intel.com>
21939         * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
21940         _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
21941         _mm_maskz_range_ss, _mm_mask_range_round_ss,
21942         _mm_maskz_range_round_ss): New intrinsics.
21943         (__builtin_ia32_rangesd128_round)
21944         (__builtin_ia32_rangess128_round): Remove.
21945         (__builtin_ia32_rangesd128_mask_round,
21946         __builtin_ia32_rangess128_mask_round): New builtins.
21947         * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
21948         __builtin_ia32_rangess128_round): Remove.
21949         (__builtin_ia32_rangesd128_mask_round,
21950         __builtin_ia32_rangess128_mask_round): New builtins.
21951         * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
21952         (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
21953         ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
21954         "<round_saeonly_constraint>")): Changed to ...
21955         ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
21956         "<round_saeonly_scalar_constraint>")): ... this.
21957         ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
21958         %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
21959         ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
21960         %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
21961         %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
21963 2018-02-02  Andrew Jenner  <andrew@codesourcery.com>
21965         * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
21966         options.
21967         * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
21968         Remove all values except native, 8540 and 8548.
21970 2018-02-02  H.J. Lu  <hongjiu.lu@intel.com>
21972         * config/i386/i386.c (ix86_output_function_return): Pass
21973         INVALID_REGNUM, instead of -1, as invalid register number to
21974         indirect_thunk_name and output_indirect_thunk.
21976 2018-02-02  Julia Koval  <julia.koval@intel.com>
21978         * config.gcc: Add -march=icelake.
21979         * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
21980         * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
21981         * config/i386/i386.c (processor_costs): Add m_ICELAKE.
21982         (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
21983         PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
21984         (processor_target_table): Add icelake.
21985         (ix86_option_override_internal): Handle new PTAs.
21986         (get_builtin_code_for_version): Handle icelake.
21987         (M_INTEL_COREI7_ICELAKE): New.
21988         (fold_builtin_cpu): Handle icelake.
21989         * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
21990         * doc/invoke.texi: Add -march=icelake.
21992 2018-02-02  Julia Koval  <julia.koval@intel.com>
21994         * config/i386/i386.c (ix86_option_override_internal): Change flags type
21995         to wide_int_bitmask.
21996         * wide-int-bitmask.h: New.
21998 2018-02-02  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
22000         PR target/84066
22001         * config/i386/i386.md: Replace Pmode with word_mode in
22002         builtin_setjmp_setup and builtin_longjmp to support x32.
22004 2018-02-01  Peter Bergner  <bergner@vnet.ibm.com>
22006         PR target/56010
22007         PR target/83743
22008         * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
22009         #include "opts.h".
22010         (rs6000_supported_cpu_names): New static variable.
22011         (linux_cpu_translation_table): Likewise.
22012         (elf_platform) <cpu>: Define new static variable and use it.
22013         Translate kernel AT_PLATFORM name to canonical name if needed.
22014         Error if platform name is unknown.
22016 2018-02-01  Aldy Hernandez  <aldyh@redhat.com>
22018         PR target/84089
22019         * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
22021 2018-02-01  Jeff Law  <law@redhat.com>
22023         PR target/84128
22024         * config/i386/i386.c (release_scratch_register_on_entry): Add new
22025         OFFSET and RELEASE_VIA_POP arguments.  Use SP+OFFSET to restore
22026         the scratch if RELEASE_VIA_POP is false.
22027         (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
22028         If we have to save a temporary register, decrement SIZE appropriately.
22029         Pass new arguments to release_scratch_register_on_entry.
22030         (ix86_adjust_stack_and_probe): Likewise.
22031         (ix86_emit_probe_stack_range): Pass new arguments to
22032         release_scratch_register_on_entry.
22034 2018-02-01  Uros Bizjak  <ubizjak@gmail.com>
22036         PR rtl-optimization/84157
22037         * combine.c (change_zero_ext): Use REG_P predicate in
22038         front of HARD_REGISTER_P predicate.
22040 2018-02-01  Georg-Johann Lay  <avr@gjlay.de>
22042         * config/avr/avr.c (avr_option_override): Move disabling of
22043         -fdelete-null-pointer-checks to...
22044         * common/config/avr/avr-common.c (avr_option_optimization_table):
22045         ...here.
22047 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
22049         PR tree-optimization/81635
22050         * tree-data-ref.c (split_constant_offset_1): For types that
22051         wrap on overflow, try to use range info to prove that wrapping
22052         cannot occur.
22054 2018-02-01  Renlin Li  <renlin.li@arm.com>
22056         PR target/83370
22057         * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
22058         TAILCALL_ADDR_REGS.
22059         (aarch64_register_move_cost): Likewise.
22060         * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
22061         TAILCALL_ADDR_REGS.
22062         (REG_CLASS_NAMES): Likewise.
22063         (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
22064         TAILCALL_ADDR_REGS. Remove IP registers.
22065         * config/aarch64/aarch64.md (Ucs): Update register constraint.
22067 2018-02-01  Richard Biener  <rguenther@suse.de>
22069         * domwalk.h (dom_walker::dom_walker): Add additional constructor
22070         for specifying RPO order and allow NULL for that.
22071         * domwalk.c (dom_walker::dom_walker): Likewise.
22072         (dom_walker::walk): Handle NULL RPO order.
22073         * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
22074         in RPO order.
22075         (rewrite_update_dom_walker): Likewise.
22076         (mark_def_dom_walker): Likewise.
22078 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
22080         * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
22081         (aarch64_maybe_expand_sve_subreg_move): Declare.
22082         * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
22083         * config/aarch64/predicates.md (aarch64_any_register_operand): New
22084         predicate.
22085         * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
22086         that are semantically a reverse operation.
22087         (*aarch64_sve_mov<mode>_subreg_be): New pattern.
22088         * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
22089         (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
22090         functions.
22091         (aarch64_can_change_mode_class): For big-endian, forbid changes
22092         between two SVE modes if they have different element sizes.
22094 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
22096         * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
22097         the TImode handling for big-endian targets.
22099 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
22101         * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
22102         (*sve_ld1rq<Vesize>): ... this new pattern.  Handle all element sizes,
22103         not just bytes.
22104         * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
22105         Remove BSWAP handing for big-endian targets and use the form of
22106         LD1RQ appropariate for the mode.
22108 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
22110         * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
22111         all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
22112         duplicated element.
22114 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
22116         PR tearget/83845
22117         * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
22118         check for operands that need to go through aarch64_sve_reload_be.
22120 2018-02-01  Jakub Jelinek  <jakub@redhat.com>
22122         PR tree-optimization/81661
22123         PR tree-optimization/84117
22124         * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
22125         * tree-eh.c: Include gimplify.h.
22126         (find_trapping_overflow, replace_trapping_overflow,
22127         rewrite_to_non_trapping_overflow): New functions.
22128         * tree-vect-loop.c: Include tree-eh.h.
22129         (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
22130         * tree-data-ref.c: Include tree-eh.h.
22131         (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
22133 2018-01-31  Uros Bizjak  <ubizjak@gmail.com>
22135         PR rtl-optimization/84123
22136         * combine.c (change_zero_ext): Check if hard register satisfies
22137         can_change_dest_mode before calling gen_lowpart_SUBREG.
22139 2018-01-31  Vladimir Makarov  <vmakarov@redhat.com>
22141         PR target/82444
22142         * ira.c (ira_init_register_move_cost): Remove assert.
22144 2018-01-31  Eric Botcazou  <ebotcazou@adacore.com>
22146         PR rtl-optimization/84071
22147         * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
22148         * doc/tm.texi: Regenerate.
22150 2018-01-31  Richard Biener  <rguenther@suse.de>
22152         PR tree-optimization/84132
22153         * tree-data-ref.c (analyze_miv_subscript): Properly
22154         check whether evolution_function_is_affine_multivariate_p
22155         before calling gcd_of_steps_may_divide_p.
22157 2018-01-31  Julia Koval  <julia.koval@intel.com>
22159         PR target/83618
22160         * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
22161         * config/i386/i386.md (rdpid_rex64) New.
22162         (rdpid): Make 32bit only.
22164 2018-01-29  Aldy Hernandez  <aldyh@redhat.com>
22166         PR lto/84105
22167         * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
22168         an IDENTIFIER_NODE for FUNCTION_TYPE's.
22170 2018-01-31  Eric Botcazou  <ebotcazou@adacore.com>
22172         Revert
22173         2018-01-12  Eric Botcazou  <ebotcazou@adacore.com>
22175         * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
22177 2018-01-31  Eric Botcazou  <ebotcazou@adacore.com>
22179         PR rtl-optimization/84071
22180         * combine.c (record_dead_and_set_regs_1): Record the source unmodified
22181         for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
22183 2018-01-31  Claudiu Zissulescu  <claziss@synopsys.com>
22185         * config/arc/arc.c (arc_handle_aux_attribute): New function.
22186         (arc_attribute_table): Add 'aux' attribute.
22187         (arc_in_small_data_p): Consider aux like variables.
22188         (arc_is_aux_reg_p): New function.
22189         (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
22190         (arc_get_aux_arg): New function.
22191         (prepare_move_operands): Handle aux-register access.
22192         (arc_handle_aux_attribute): New function.
22193         * doc/extend.texi (ARC Variable attributes): Add subsection.
22195 2018-01-31  Claudiu Zissulescu  <claziss@synopsys.com>
22197         * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
22198         * config/arc/arc.c (arc_handle_uncached_attribute): New function.
22199         (arc_attribute_table): Add 'uncached' attribute.
22200         (arc_print_operand): Print '.di' flag for uncached memory
22201         accesses.
22202         (arc_in_small_data_p): Do not consider for small data the uncached
22203         types.
22204         (arc_is_uncached_mem_p): New function.
22205         * config/arc/predicates.md (compact_store_memory_operand): Check
22206         for uncached memory accesses.
22207         (nonvol_nonimm_operand): Likewise.
22208         * doc/extend.texi (ARC Type Attribute): New subsection.
22210 2018-01-31  Jakub Jelinek  <jakub@redhat.com>
22212         PR c/84100
22213         * common.opt (falign-functions=, falign-jumps=, falign-labels=,
22214         falign-loops=): Add Optimization flag.
22216 2018-01-30  Jeff Law  <law@redhat.com>
22218         PR target/84064
22219         * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
22220         INT_REGISTERS_SAVED.  Check it prior to calling
22221         get_scratch_register_on_entry.
22222         (ix86_adjust_stack_and_probe): Similarly.
22223         (ix86_emit_probe_stack_range): Similarly.
22224         (ix86_expand_prologue): Corresponding changes.
22226 2018-01-30  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
22228         PR target/40411
22229         * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
22230         -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
22232 2018-01-30  Vladimir Makarov  <vmakarov@redhat.com>
22234         PR target/84112
22235         * lra-constraints.c (curr_insn_transform): Process AND in the
22236         address.
22238 2018-01-30  Jakub Jelinek  <jakub@redhat.com>
22240         PR rtl-optimization/83986
22241         * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
22242         dependence against last_pending_memory_flush in addition to
22243         pending_jump_insns.
22245 2018-01-30  Alexandre Oliva  <aoliva@redhat.com>
22247         PR tree-optimization/81611
22248         * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
22249         copies.
22251 2018-01-30  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
22253         PR target/83758
22254         * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
22255         a reg rtx.
22257 2018-01-30  Richard Biener  <rguenther@suse.de>
22258             Jakub Jelinek  <jakub@redhat.com>
22260         PR tree-optimization/84111
22261         * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
22262         inner loops added during recursion, as they don't have up-to-date
22263         SSA form.
22265 2018-01-30  Jan Hubicka  <hubicka@ucw.cz>
22267         PR ipa/81360
22268         * ipa-inline.c (can_inline_edge_p): Break out late tests to...
22269         (can_inline_edge_by_limits_p): ... here.
22270         (can_early_inline_edge_p, check_callers,
22271         update_caller_keys, update_callee_keys, recursive_inlining,
22272         add_new_edges_to_heap, speculation_useful_p,
22273         inline_small_functions,
22274         inline_small_functions, flatten_function,
22275         inline_to_all_callers_1): Update.
22277 2018-01-30  Jan Hubicka  <hubicka@ucw.cz>
22279         * profile-count.c (profile_count::combine_with_ipa_count): Handle
22280         zeros correctly.
22282 2018-01-30  Richard Biener  <rguenther@suse.de>
22284         PR tree-optimization/83008
22285         * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
22286         invariant and constant vector uses in stmts when they need
22287         more than one stmt.
22289 2018-01-30  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
22291         PR bootstrap/84017
22292         * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
22293         * configure: Regenerate.
22295 2018-01-30  Richard Sandiford  <richard.sandiford@linaro.org>
22297         * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
22298         pattern.
22299         (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
22300         Use gen_rtx_REG rather than gen_lowpart.
22302 2018-01-30  Richard Sandiford  <richard.sandiford@linaro.org>
22304         * lra-constraints.c (match_reload): Use subreg_lowpart_offset
22305         rather than 0 when creating partial subregs.
22307 2018-01-30  Richard Sandiford  <richard.sandiford@linaro.org>
22309         * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
22310         of usage.
22312 2018-01-29  Michael Meissner  <meissner@linux.vnet.ibm.com>
22314         PR target/81550
22315         * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
22316         and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
22317         -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
22318         flags.  This restores the settings used before the 2017-07-24.
22319         Turning off pre increment/decrement/modify allows IVOPTS to
22320         optimize DF/SF loops where the index is an int.
22322 2018-01-29  Richard Biener  <rguenther@suse.de>
22323             Kelvin Nilsen  <kelvin@gcc.gnu.org>
22325         PR bootstrap/80867
22326         * tree-vect-stmts.c (vectorizable_call): Don't call
22327         targetm.vectorize_builtin_md_vectorized_function if callee is
22328         NULL.
22330 2018-01-22  Carl Love  <cel@us.ibm.com>
22332         * doc/extend.tex: Fix typo in second arg in
22333         __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
22335 2018-01-29  Richard Biener  <rguenther@suse.de>
22337         PR tree-optimization/84086
22338         * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
22339         (flush_ssaname_freelist): When SSA names were released reset
22340         the SCEV hash table.
22342 2018-01-29  Richard Biener  <rguenther@suse.de>
22344         PR tree-optimization/84057
22345         * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
22346         removed paths when removing edges.
22348 2018-01-27  H.J. Lu  <hongjiu.lu@intel.com>
22350         * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
22351         -mfunction-return=@var{choice}.
22353 2018-01-27  Bernd Edlinger  <bernd.edlinger@hotmail.de>
22355         PR diagnostic/84034
22356         * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
22357         Handle CR like TAB.
22358         (layout::print_source_line): Likewise.
22359         (test_get_line_width_without_trailing_whitespace): Add test cases.
22361 2018-01-27  Jakub Jelinek  <jakub@redhat.com>
22363         PR middle-end/84040
22364         * sched-deps.c (sched_macro_fuse_insns): Return immediately for
22365         debug insns.
22367 2018-01-26  Jim Wilson  <jimw@sifive.com>
22369         * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
22371         * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
22372         specified.
22374 2018-01-26  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
22376         * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
22377         and CMP + SUB-immediate -> SUBS.
22379 2018-01-26  Martin Sebor  <msebor@redhat.com>
22381         PR tree-optimization/83896
22382         * tree-ssa-strlen.c (get_string_len): Rename...
22383         (get_string_cst_length): ...to this.  Return HOST_WIDE_INT.
22384         Avoid assuming length is constant.
22385         (handle_char_store): Use HOST_WIDE_INT for string length.
22387 2018-01-26  Uros Bizjak  <ubizjak@gmail.com>
22389         PR target/81763
22390         * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
22391         to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
22393 2018-01-26  Richard Biener  <rguenther@suse.de>
22395         PR rtl-optimization/84003
22396         * dse.c (record_store): Only record redundant stores when
22397         the earlier store aliases at least all accesses the later one does.
22399 2018-01-26  Jakub Jelinek  <jakub@redhat.com>
22401         PR rtl-optimization/83985
22402         * dce.c (deletable_insn_p): Return false for separate shrink wrapping
22403         REG_CFA_RESTORE insns.
22404         (delete_unmarked_insns): Don't ignore separate shrink wrapping
22405         REG_CFA_RESTORE insns here.
22407         PR c/83989
22408         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
22409         use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
22411 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
22413         * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
22414         * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
22415         (arc_init): Likewise.
22416         (arc_override_options): Likewise.
22417         (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
22418         value.
22419         (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
22420         support.
22421         * config/arc/arc.h (TARGET_DBNZ): Define.
22422         * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
22423         properly set the tune attribute.
22424         (dbnz): Use TARGET_DBNZ guard.
22425         * config/arc/arc.opt (mtune): Add core3 option.
22427 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
22429         * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
22430         recognize new pic like addresses.
22431         (arc_delegitimize_address): Clean up.
22433 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
22435         * config/arc/arc-arches.def: Option mrf16 valid for all
22436         architectures.
22437         * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
22438         * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
22439         * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
22440         * config/arc/arc-tables.opt: Regenerate.
22441         * config/arc/arc.c (arc_conditional_register_usage): Handle
22442         reduced register file case.
22443         (arc_file_start): Set must have build attributes.
22444         * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
22445         mrf16 option value.
22446         * config/arc/arc.opt (mrf16): Add new option.
22447         * config/arc/elf.h (ATTRIBUTE_PCS): Define.
22448         * config/arc/genmultilib.awk: Handle new mrf16 option.
22449         * config/arc/linux.h (ATTRIBUTE_PCS): Define.
22450         * config/arc/t-multilib: Regenerate.
22451         * doc/invoke.texi (ARC Options): Document mrf16 option.
22453 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
22455         * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
22456         * config/arc/arc.c (arc_handle_secure_attribute): New function.
22457         (arc_attribute_table): Add 'secure_call' attribute.
22458         (arc_print_operand): Print secure call operand.
22459         (arc_function_ok_for_sibcall): Don't optimize tail calls when
22460         secure.
22461         (arc_is_secure_call_p): New function.  * config/arc/arc.md
22462         (call_i): Add support for sjli instruction.
22463         (call_value_i): Likewise.
22464         * config/arc/constraints.md (Csc): New constraint.
22466 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
22467             John Eric Martin  <John.Martin@emmicro-us.com>
22469         * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
22470         * config/arc/arc.c (_arc_jli_section): New struct.
22471         (arc_jli_section): New type.
22472         (rc_jli_sections): New static variable.
22473         (arc_handle_jli_attribute): New function.
22474         (arc_attribute_table): Add jli_always and jli_fixed attribute.
22475         (arc_file_end): New function.
22476         (TARGET_ASM_FILE_END): Define.
22477         (arc_print_operand): Reuse 'S' letter for JLI output instruction.
22478         (arc_add_jli_section): New function.
22479         (jli_call_scan): Likewise.
22480         (arc_reorg): Call jli_call_scan.
22481         (arc_output_addsi): Remove 'S' from printing asm operand.
22482         (arc_is_jli_call_p): New function.
22483         * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
22484         operand.
22485         (movhi_insn): Likewise.
22486         (movsi_insn): Likewise.
22487         (movsi_set_cc_insn): Likewise.
22488         (loadqi_update): Likewise.
22489         (load_zeroextendqisi_update): Likewise.
22490         (load_signextendqisi_update): Likewise.
22491         (loadhi_update): Likewise.
22492         (load_zeroextendhisi_update): Likewise.
22493         (load_signextendhisi_update): Likewise.
22494         (loadsi_update): Likewise.
22495         (loadsf_update): Likewise.
22496         (movsicc_insn): Likewise.
22497         (bset_insn): Likewise.
22498         (bxor_insn): Likewise.
22499         (bclr_insn): Likewise.
22500         (bmsk_insn): Likewise.
22501         (bicsi3_insn): Likewise.
22502         (cmpsi_cc_c_insn): Likewise.
22503         (movsi_ne): Likewise.
22504         (movsi_cond_exec): Likewise.
22505         (clrsbsi2): Likewise.
22506         (norm_f): Likewise.
22507         (normw): Likewise.
22508         (swap): Likewise.
22509         (divaw): Likewise.
22510         (flag): Likewise.
22511         (sr): Likewise.
22512         (kflag): Likewise.
22513         (ffs): Likewise.
22514         (ffs_f): Likewise.
22515         (fls): Likewise.
22516         (call_i): Remove 'S' asm letter, add jli instruction.
22517         (call_value_i): Likewise.
22518         * config/arc/arc.op (mjli-always): New option.
22519         * config/arc/constraints.md (Cji): New constraint.
22520         * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
22521         operand.
22522         (subsf3_fpx): Likewise.
22523         (mulsf3_fpx): Likewise.
22524         * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
22525         asm operand.
22526         * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
22527         function attrbutes.
22528         * doc/invoke.texi (ARC): Document mjli-always option.
22530 2018-01-26  Sebastian Perta  <sebastian.perta@renesas.com>
22532         * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
22533         avoid addition with 0 and use incw and decw where possible.
22535 2018-01-26  Richard Biener  <rguenther@suse.de>
22537         PR tree-optimization/81082
22538         * fold-const.c (fold_plusminus_mult_expr): Do not perform the
22539         association if it requires casting to unsigned.
22540         * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
22541         from fold_plusminus_mult_expr to catch important cases late when
22542         range info is available.
22544 2018-01-26  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
22546         * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
22547         * configure.ac (hidden_linkonce): New test.
22548         * configure: Regenerate.
22549         * config.in: Regenerate.
22551 2018-01-26  Julia Koval  <julia.koval@intel.com>
22553         * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
22554         _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
22555         _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
22556         _mm_mask_bitshuffle_epi64_mask): Fix type.
22557         * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
22558         USI_FTYPE_V4DI_V4DI_USI): Remove.
22559         * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
22560         __builtin_ia32_vpshufbitqmb256_mask,
22561         __builtin_ia32_vpshufbitqmb128_mask): Fix types.
22562         * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
22563         * config/i386/sse.md (VI1_AVX512VLBW): Change types.
22565 2018-01-26  Alan Modra  <amodra@gmail.com>
22567         PR target/84033
22568         * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
22569         UNSPEC_VBPERMQ.  Sort other unspecs.
22571 2018-01-25  David Edelsohn  <dje.gcc@gmail.com>
22573         * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
22575 2018-01-25  Jan Hubicka  <hubicka@ucw.cz>
22577         PR middle-end/83055
22578         * predict.c (drop_profile): Do not push/pop cfun; update also
22579         node->count.
22580         (handle_missing_profiles): Fix logic looking for zero profiles.
22582 2018-01-25  Jakub Jelinek  <jakub@redhat.com>
22584         PR middle-end/83977
22585         * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
22586         on functions with #pragma omp declare simd or functions with simd
22587         attribute.
22588         * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
22589         * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
22590         Remove trailing \n from warning_at calls.
22592 2018-01-25  Tom de Vries  <tom@codesourcery.com>
22594         PR target/84028
22595         * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
22596         for neutered workers.
22598 2018-01-24  Joseph Myers  <joseph@codesourcery.com>
22600         PR target/68467
22601         * config/m68k/m68k.c (m68k_promote_function_mode): New function.
22602         (TARGET_PROMOTE_FUNCTION_MODE): New macro.
22604 2018-01-24  Jeff Law  <law@redhat.com>
22606         PR target/83994
22607         * i386.c (get_probe_interval): Move to earlier point.
22608         (ix86_compute_frame_layout): If -fstack-clash-protection and
22609         the frame is larger than the probe interval, then use pushes
22610         to save registers rather than reg->mem moves.
22611         (ix86_expand_prologue): Remove conditional for int_registers_saved
22612         assertion.
22614 2018-01-24  Vladimir Makarov  <vmakarov@redhat.com>
22616         PR target/84014
22617         * ira-build.c (setup_min_max_allocno_live_range_point): Set up
22618         min/max for never referenced object.
22620 2018-01-24  Jakub Jelinek  <jakub@redhat.com>
22622         PR middle-end/83977
22623         * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
22624         here.
22625         * omp-low.c (create_omp_child_function): Remove "omp declare simd"
22626         attributes from DECL_ATTRIBUTES (decl) without affecting
22627         DECL_ATTRIBUTES (current_function_decl).
22628         * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
22629         functions with non-NULL DECL_ABSTRACT_ORIGIN.
22631 2018-01-24  Richard Sandiford  <richard.sandiford@linaro.org>
22633         PR tree-optimization/83979
22634         * fold-const.c (fold_comparison): Use constant_boolean_node
22635         instead of boolean_{true,false}_node.
22637 2018-01-24  Jan Hubicka  <hubicka@ucw.cz>
22639         * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
22640         with zero counts.
22642 2018-01-24  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
22644         * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
22645         Simplify the clause that sets the length attribute.
22646         (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
22647         (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
22648         clause that sets the length attribute.
22649         (*sibcall_value_nonlocal_sysv<mode>): Likewise.
22651 2018-01-24  Tom de Vries  <tom@codesourcery.com>
22653         PR target/83589
22654         * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
22655         (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
22656         Add strict parameter.
22657         (prevent_branch_around_nothing): Insert dummy insn between branch to
22658         label and label with no ptx insn inbetween.
22659         * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
22661 2018-01-24  Tom de Vries  <tom@codesourcery.com>
22663         PR target/81352
22664         * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
22665         for neutered threads in warp.
22666         * config/nvptx/nvptx.md (define_insn "exit"): New insn.
22668 2018-01-24  Richard Biener  <rguenther@suse.de>
22670         PR tree-optimization/83176
22671         * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
22672         operands.
22674 2018-01-24  Richard Biener  <rguenther@suse.de>
22676         PR tree-optimization/82819
22677         * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
22678         code generating pluses that are no-ops in the target precision.
22680 2018-01-24  Richard Biener  <rguenther@suse.de>
22682         PR middle-end/84000
22683         * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
22685 2018-01-23  Jan Hubicka  <hubicka@ucw.cz>
22687         * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
22688         to merge probabilities.
22689         * predict.c (probably_never_executed): Also mark as cold functions
22690         with global 0 profile and guessed local profile.
22691         * profile-count.c (profile_probability::combine_with_count): New
22692         member function.
22693         * profile-count.h (profile_probability::operator*,
22694         profile_probability::operator*=, profile_probability::operator/,
22695         profile_probability::operator/=): Reduce precision to adjusted
22696         and set value to guessed on contradictory divisions.
22697         (profile_probability::combine_with_freq): Remove.
22698         (profile_probability::combine_wiht_count): Declare.
22699         (profile_count::force_nonzero):: Set to adjusted.
22700         (profile_count::probability_in):: Set quality to adjusted.
22701         * tree-ssa-tail-merge.c (replace_block_by): Use
22702         combine_with_count.
22704 2018-01-23  Andrew Waterman  <andrew@sifive.com>
22705             Jim Wilson  <jimw@sifive.com>
22707         * config/riscv/riscv.c (riscv_stack_boundary): New.
22708         (riscv_option_override): Set riscv_stack_boundary.  Handle
22709         riscv_preferred_stack_boundary_arg.
22710         * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
22711         (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
22712         (STACK_BOUNDARY): Set to riscv_stack_boundary.
22713         (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
22714         * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
22715         * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
22717 2018-01-23  H.J. Lu  <hongjiu.lu@intel.com>
22719         PR target/83905
22720         * config/i386/i386.c (ix86_expand_prologue): Use cost reference
22721         of struct ix86_frame.
22722         (ix86_expand_epilogue): Likewise.  Add a local variable for
22723         the reg_save_offset field in struct ix86_frame.
22725 2018-01-23  Bin Cheng  <bin.cheng@arm.com>
22727         PR tree-optimization/82604
22728         * tree-loop-distribution.c (enum partition_kind): New enum item
22729         PKIND_PARTIAL_MEMSET.
22730         (partition_builtin_p): Support above new enum item.
22731         (generate_code_for_partition): Ditto.
22732         (compute_access_range): Differentiate cases that equality can be
22733         proven at all loops, the innermost loops or no loops.
22734         (classify_builtin_st, classify_builtin_ldst): Adjust call to above
22735         function.  Set PKIND_PARTIAL_MEMSET for partition appropriately.
22736         (finalize_partitions, distribute_loop): Don't fuse partition of
22737         PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
22738         (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
22739         parloop is enabled.
22741 2018-01-23  Martin Liska  <mliska@suse.cz>
22743         * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
22744         order to ignore the predictor.
22745         (PRED_POLYMORPHIC_CALL): Likewise.
22746         (PRED_RECURSIVE_CALL): Likewise.
22748 2018-01-23  Martin Liska  <mliska@suse.cz>
22750         * tree-profile.c (tree_profiling): Print function header to
22751         aware reader which function we are working on.
22752         * value-prof.c (gimple_find_values_to_profile): Do not print
22753         not interesting value histograms.
22755 2018-01-23  Martin Liska  <mliska@suse.cz>
22757         * profile-count.h (enum profile_quality): Add
22758         profile_uninitialized as the first value. Do not number values
22759         as they are zero based.
22760         (profile_count::verify): Update sanity check.
22761         (profile_probability::verify): Likewise.
22763 2018-01-23  Nathan Sidwell  <nathan@acm.org>
22765         * doc/invoke.texi (ffor-scope): Deprecate.
22767 2018-01-23  David Malcolm  <dmalcolm@redhat.com>
22769         PR tree-optimization/83510
22770         * domwalk.c (set_all_edges_as_executable): New function.
22771         (dom_walker::dom_walker): Convert bool param
22772         "skip_unreachable_blocks" to enum reachability.  Move setup of
22773         edge flags to set_all_edges_as_executable and only do it when
22774         reachability is REACHABLE_BLOCKS.
22775         * domwalk.h (enum dom_walker::reachability): New enum.
22776         (dom_walker::dom_walker): Convert bool param
22777         "skip_unreachable_blocks" to enum reachability.
22778         (set_all_edges_as_executable): New decl.
22779         * graphite-scop-detection.c  (gather_bbs::gather_bbs): Convert
22780         from false for "skip_unreachable_blocks" to ALL_BLOCKS for
22781         "reachability".
22782         * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
22783         but converting true to REACHABLE_BLOCKS.
22784         * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
22785         * tree-vrp.c
22786         (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
22787         Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
22788         (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
22789         REACHABLE_BLOCKS.
22790         (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
22791         if check_all_array_refs will be called.
22793 2018-01-23  David Malcolm  <dmalcolm@redhat.com>
22795         * tree.c (selftest::test_location_wrappers): Add more test
22796         coverage.
22798 2018-01-23  David Malcolm  <dmalcolm@redhat.com>
22800         * sbitmap.c (selftest::test_set_range): Fix memory leaks.
22801         (selftest::test_bit_in_range): Likewise.
22803 2018-01-23  Richard Sandiford  <richard.sandiford@linaro.org>
22805         PR testsuite/83888
22806         * doc/sourcebuild.texi (vect_float): Say that the selector
22807         only describes the situation when -funsafe-math-optimizations is on.
22808         (vect_float_strict): Document.
22810 2018-01-23  Richard Sandiford  <richard.sandiford@linaro.org>
22812         PR tree-optimization/83965
22813         * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
22814         (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
22815         instead of checking only for a reduction.
22816         (vect_recog_widen_sum_pattern): Likewise.
22818 2018-01-23  Jan Hubicka  <hubicka@ucw.cz>
22820         * predict.c (probably_never_executed): Only use precise profile info.
22821         (compute_function_frequency): Skip after inlining hack since we now
22822         have quality checking.
22824 2018-01-23  Jan Hubicka  <hubicka@ucw.cz>
22826         * profile-count.h (profile_probability::very_unlikely,
22827         profile_probability::unlikely, profile_probability::even): Set
22828         precision to guessed.
22830 2018-01-23  Richard Biener  <rguenther@suse.de>
22832         PR tree-optimization/83963
22833         * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
22834         Properly terminate dominator walk when crossing the exit edge not
22835         when visiting its source block.
22837 2018-01-23  Jakub Jelinek  <jakub@redhat.com>
22839         PR c++/83918
22840         * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
22841         VIEW_CONVERT_EXPR to wrap CONST_DECLs.
22843 2018-01-22  Jakub Jelinek  <jakub@redhat.com>
22845         PR tree-optimization/83957
22846         * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs.  Remove
22847         semicolon after for body surrounded by braces.
22849         PR tree-optimization/83081
22850         * profile-count.h (profile_probability::split): New method.
22851         * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
22852         Use profile_probability::split.
22853         (do_compare_rtx_and_jump): Fix adjustment of probabilities
22854         when splitting a single conditional jump into 2.
22856 2018-01-22  David Malcolm  <dmalcolm@redhat.com>
22858         PR tree-optimization/69452
22859         * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
22860         decl.
22862 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
22864         * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
22865         * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
22866         * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
22868 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
22870         * config/rl78/rl78-protos.h (rl78_split_movdi): New function
22871         declaration.
22872         * config/rl78/rl78.md (movdi): New define_expand.
22873         * config/rl78/rl78.c (rl78_split_movdi): New function.
22875 2018-01-22  Michael Meissner  <meissner@linux.vnet.ibm.com>
22877         PR target/83862
22878         * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
22879         no longer used.
22880         * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
22881         * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
22882         128-bit to produce an UNSPEC move to get the double word with the
22883         signbit and then a shift directly to do signbit.
22884         (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
22885         implementation with a new version that just does either a direct
22886         move or a regular move.  Move memory interface to separate insns.
22887         Move insns so they are next to the expander.
22888         (signbit<mode>2_dm_mem_be): New combiner insns to combine load
22889         with signbit move.  Split big and little endian case.
22890         (signbit<mode>2_dm_mem_le): Likewise.
22891         (signbit<mode>2_dm_<su>ext): Delete, no longer used.
22892         (signbit<mode>2_dm2): Likewise.
22894 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
22896         * config/rl78/rl78.md (anddi3): New define_expand.
22898 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
22900         * config/rl78/rl78.md (umindi3): New define_expand.
22902 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
22904         * config/rl78/rl78.md (smindi3): New define_expand.
22906 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
22908         * config/rl78/rl78.md (smaxdi3): New define_expand.
22910 2018-01-22  Carl Love  <cel@us.ibm.com>
22912         * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
22913         LVX_V1TI): Add macro expansion.
22914         * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
22915         definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
22916         VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
22917         * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
22918         Change check to determine if the instruction is a byte reversing
22919         entry.  Fix typo in comment.
22920         * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
22921         for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
22922         Add def_builtin calls for new builtins.
22923         * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
22924         Add define_insn expansion.
22926 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
22928         * config/rl78/rl78.md (umaxdi3): New define_expand.
22930 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
22932         * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
22933         for non-QImode registers.
22935 2018-01-22  Richard Biener  <rguenther@suse.de>
22937         PR tree-optimization/83963
22938         * graphite-scop-detection.c (scop_detection::get_sese): Delay
22939         including the loop exit block.
22940         (scop_detection::merge_sese): Likewise.
22941         (scop_detection::add_scop): Do it here instead.
22943 2018-01-22  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
22945         * doc/sourcebuild.texi (arm_softfloat): Document.
22947 2018-01-21  John David Anglin  <danglin@gcc.gnu.org>
22949         PR gcc/77734
22950         * config/pa/pa.c (pa_function_ok_for_sibcall): Use
22951         targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
22952         Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
22954 2018-01-21  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
22955             David Edelsohn  <dje.gcc@gmail.com>
22957         PR target/83946
22958         * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
22959         Change "crset eq" to "crset 2".
22960         (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
22961         (*call_indirect_aix<mode>_nospec): Likewise.
22962         (*call_value_indirect_aix<mode>_nospec): Likewise.
22963         (*call_indirect_elfv2<mode>_nospec): Likewise.
22964         (*call_value_indirect_elfv2<mode>_nospec): Likewise.
22965         (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
22966         change assembly output from . to $.
22967         (*sibcall_value_nonlocal_sysv<mode>): Likewise.
22968         (indirect_jump<mode>_nospec): Change assembly output from . to $.
22969         (*tablejump<mode>_internal1_nospec): Likewise.
22971 2018-01-21  Oleg Endo  <olegendo@gcc.gnu.org>
22973         PR target/80870
22974         * config/sh/sh_optimize_sett_clrt.cc:
22975         Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
22977 2018-01-20  Richard Sandiford  <richard.sandiford@linaro.org>
22979         PR tree-optimization/83940
22980         * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
22981         offset_dt to vect_constant_def rather than vect_unknown_def_type.
22982         (vect_check_load_store_mask): Add a mask_dt_out parameter and
22983         use it to pass back the definition type.
22984         (vect_check_store_rhs): Likewise rhs_dt_out.
22985         (vect_build_gather_load_calls): Add a mask_dt argument and use
22986         it instead of a call to vect_is_simple_use.
22987         (vectorizable_store): Update calls to vect_check_load_store_mask
22988         and vect_check_store_rhs.  Use the dt returned by the latter instead
22989         of scatter_src_dt.  Use the cached mask_dt and gs_info.offset_dt
22990         instead of calls to vect_is_simple_use.  Pass the scalar rather
22991         than the vector operand to vect_is_simple_use when handling
22992         second and subsequent copies of an rhs value.
22993         (vectorizable_load): Update calls to vect_check_load_store_mask
22994         and vect_build_gather_load_calls.  Use the cached mask_dt and
22995         gs_info.offset_dt instead of calls to vect_is_simple_use.
22997 2018-01-20  Jakub Jelinek  <jakub@redhat.com>
22999         PR middle-end/83945
23000         * tree-emutls.c: Include gimplify.h.
23001         (lower_emutls_2): New function.
23002         (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
23003         with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
23004         it before further processing.
23006         PR target/83930
23007         * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
23008         UINTVAL (trueop1) instead of INTVAL (op1).
23010 2018-01-19  Jakub Jelinek  <jakub@redhat.com>
23012         PR debug/81570
23013         PR debug/83728
23014         * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
23015         INCOMING_FRAME_SP_OFFSET if not defined.
23016         (scan_trace): Add ENTRY argument.  If true and
23017         DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
23018         emit a note to adjust the CFA offset.
23019         (create_cfi_notes): Adjust scan_trace callers.
23020         (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
23021         INCOMING_FRAME_SP_OFFSET in the CIE.
23022         * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
23023         * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
23024         Likewise.
23025         * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
23026         * doc/tm.texi: Regenerated.
23028 2018-01-19  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
23030         PR rtl-optimization/83147
23031         * lra-constraints.c (remove_inheritance_pseudos): Use
23032         lra_substitute_pseudo_within_insn.
23034 2018-01-19  Tom de Vries  <tom@codesourcery.com>
23035             Cesar Philippidis  <cesar@codesourcery.com>
23037         PR target/83920
23038         * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
23040 2018-01-19  Cesar Philippidis  <cesar@codesourcery.com>
23042         PR target/83790
23043         * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
23044         spaces for function labels.
23046 2018-01-19  Martin Liska  <mliska@suse.cz>
23048         * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
23049         (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
23050         (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
23051         (PRED_OPCODE_POSITIVE): Change from 64 to 59.
23052         (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
23053         (PRED_CONST_RETURN): Change from 69 to 65.
23054         (PRED_NULL_RETURN): Change from 91 to 71.
23055         (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
23056         (PRED_LOOP_GUARD): Change from 66 to 73.
23058 2018-01-19  Martin Liska  <mliska@suse.cz>
23060         * predict.c (predict_insn_def): Add new assert.
23061         (struct branch_predictor): Change type to signed integer.
23062         (test_prediction_value_range): Amend test to cover
23063         PROB_UNINITIALIZED.
23064         * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
23065         (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
23066         (PRED_LOOP_ITERATIONS_MAX): Likewise.
23067         (PRED_LOOP_IV_COMPARE): Likewise.
23068         * predict.h (PROB_UNINITIALIZED): Define new constant.
23070 2018-01-19  Martin Liska  <mliska@suse.cz>
23072         * predict.c (dump_prediction): Add new format for
23073         analyze_brprob.py script which is enabled with -details
23074         suboption.
23075         * profile-count.h (precise_p): New function.
23077 2018-01-19  Richard Sandiford  <richard.sandiford@linaro.org>
23079         PR tree-optimization/83922
23080         * tree-vect-loop.c (vect_verify_full_masking): Return false if
23081         there are no statements that need masking.
23082         (vect_active_double_reduction_p): New function.
23083         (vect_analyze_loop_operations): Use it when handling phis that
23084         are not in the loop header.
23086 2018-01-19  Richard Sandiford  <richard.sandiford@linaro.org>
23088         PR tree-optimization/83914
23089         * tree-vect-loop.c (vectorizable_induction): Don't convert
23090         init_expr or apply the peeling adjustment for inductions
23091         that are nested within the vectorized loop.
23093 2018-01-19  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
23095         * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
23096         instead of NEG.
23098 2018-01-18  Jakub Jelinek  <jakub@redhat.com>
23100         PR sanitizer/81715
23101         PR testsuite/83882
23102         * function.h (gimplify_parameters): Add gimple_seq * argument.
23103         * function.c: Include gimple.h and options.h.
23104         (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
23105         for the added local temporaries if needed.
23106         * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
23107         if there are any parameter cleanups, wrap whole body into a
23108         try/finally with the cleanups.
23110 2018-01-18  Wilco Dijkstra  <wdijkstr@arm.com>
23112         PR target/82964
23113         * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
23114         Use GET_MODE_CLASS for scalar floating point.
23116 2018-01-18  Jan Hubicka  <hubicka@ucw.cz>
23118         PR ipa/82256
23119         patch by PaX Team
23120         * cgraphclones.c (cgraph_node::create_version_clone_with_body):
23121         Fix call of call_cgraph_insertion_hooks.
23123 2018-01-18  Martin Sebor  <msebor@redhat.com>
23125         * doc/invoke.texi (-Wclass-memaccess): Tweak text.
23127 2018-01-18  Jan Hubicka  <hubicka@ucw.cz>
23129         PR ipa/83619
23130         * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
23131         frequencies.
23133 2018-01-18  Boris Kolpackov  <boris@codesynthesis.com>
23135         PR other/70268
23136         * common.opt: (-ffile-prefix-map): New option.
23137         * opts.c (common_handle_option): Defer it.
23138         * opts-global.c (handle_common_deferred_options): Handle it.
23139         * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
23140         * file-prefix-map.h: New file.
23141         (remap_debug_filename, add_debug_prefix_map): ...here.
23142         (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
23143         * final.c (debug_prefix_map, add_debug_prefix_map
23144         remap_debug_filename): Move to...
23145         * file-prefix-map.c: New file.
23146         (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
23147         generalize, get rid of alloca(), use strrchr() instead of strchr().
23148         (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
23149         Implement in terms of add_prefix_map().
23150         (remap_macro_filename, remap_debug_filename): Implement in term of
23151         remap_filename().
23152         * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
23153         * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
23154         * dbxout.c: Include file-prefix-map.h.
23155         * varasm.c: Likewise.
23156         * vmsdbgout.c: Likewise.
23157         * xcoffout.c: Likewise.
23158         * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
23159         * doc/cppopts.texi (-fmacro-prefix-map): Document.
23160         * doc/invoke.texi (-ffile-prefix-map): Document.
23161         (-fdebug-prefix-map): Update description.
23163 2018-01-18  Martin Liska  <mliska@suse.cz>
23165         * config/i386/i386.c (indirect_thunk_name): Document that also
23166         lfence is emitted.
23167         (output_indirect_thunk): Document why both instructions
23168         (pause and lfence) are generated.
23170 2018-01-18  Richard Biener  <rguenther@suse.de>
23172         PR tree-optimization/83887
23173         * graphite-scop-detection.c
23174         (scop_detection::get_nearest_dom_with_single_entry): Remove.
23175         (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
23176         (scop_detection::merge_sese): Re-implement with a flood-fill
23177         algorithm that properly finds a SESE region if it exists.
23179 2018-01-18  Jakub Jelinek  <jakub@redhat.com>
23181         PR c/61240
23182         * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
23183         pointer_diff optimizations use view_convert instead of convert.
23185 2018-01-17  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
23187         * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
23188         Generate different code for -mno-speculate-indirect-jumps.
23189         (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
23190         (*call_indirect_aix<mode>): Disable for
23191         -mno-speculate-indirect-jumps.
23192         (*call_indirect_aix<mode>_nospec): New define_insn.
23193         (*call_value_indirect_aix<mode>): Disable for
23194         -mno-speculate-indirect-jumps.
23195         (*call_value_indirect_aix<mode>_nospec): New define_insn.
23196         (*sibcall_nonlocal_sysv<mode>): Generate different code for
23197         -mno-speculate-indirect-jumps.
23198         (*sibcall_value_nonlocal_sysv<mode>): Likewise.
23200 2018-01-17  Michael Meissner  <meissner@linux.vnet.ibm.com>
23202         * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
23203         long double type, set the flags for noting the default long double
23204         type, even if we don't pass or return a long double type.
23206 2018-01-17  Jan Hubicka  <hubicka@ucw.cz>
23208         PR ipa/83051
23209         * ipa-inline.c (flatten_function): Do not overwrite final inlining
23210         failure.
23212 2018-01-17  Will Schmidt  <will_schmidt@vnet.ibm.com>
23214         * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
23215         support for merge[hl].
23216         (fold_mergehl_helper): New helper function.
23217         (tree-vector-builder.h): New #include for tree_vector_builder usage.
23218         * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
23219         (altivec_vmrglw_direct): Add xxmrglw insn.
23221 2018-01-17  Andrew Waterman  <andrew@sifive.com>
23223         * config/riscv/riscv.c (riscv_conditional_register_usage): If
23224         UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
23226 2018-01-17  David Malcolm  <dmalcolm@redhat.com>
23228         PR lto/83121
23229         * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
23230         call the lto_location_cache before reading the
23231         DECL_SOURCE_LOCATION of the types.
23233 2018-01-17  Wilco Dijkstra  <wdijkstr@arm.com>
23234             Richard Sandiford  <richard.sandiford@linaro.org>
23236         * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
23237         * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
23238         (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
23239         SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
23240         * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
23241         Add declaration.
23242         * config/aarch64/constraints.md (aarch64_movti_operand):
23243         Limit immediates.
23244         * config/aarch64/predicates.md (Uti): Add new constraint.
23246 2018-01-17  Carl Love  <cel@us.ibm.com>
23248         * config/rs6000/vsx.md (define_expand xl_len_r,
23249         define_expand stxvl, define_expand *stxvl): Add match_dup argument.
23250         (define_insn): Add, match_dup 1 argument to define_insn stxvll and
23251         lxvll.
23252         (define_expand, define_insn): Move the shift left from  the
23253         define_insn to the define_expand for lxvl and stxvl instructions.
23254         * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
23255         and XL_LEN_R definitions to PURE.
23257 2018-01-17  Uros Bizjak  <ubizjak@gmail.com>
23259         * config/i386/i386.c (indirect_thunk_name): Declare regno
23260         as unsigned int.  Compare regno with INVALID_REGNUM.
23261         (output_indirect_thunk): Ditto.
23262         (output_indirect_thunk_function): Ditto.
23263         (ix86_code_end): Declare regno as unsigned int.  Use INVALID_REGNUM
23264         in the call to output_indirect_thunk_function.
23266 2018-01-17  Richard Sandiford  <richard.sandiford@linaro.org>
23268         PR middle-end/83884
23269         * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
23270         rather than the size of inner_type to determine the stack slot size
23271         when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
23273 2018-01-16  Sebastian Peryt  <sebastian.peryt@intel.com>
23275         PR target/83546
23276         * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
23277         to PTA_SILVERMONT.
23279 2018-01-16  Michael Meissner  <meissner@linux.vnet.ibm.com>
23281         * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
23282         endian Linux systems to optionally enable multilibs for selecting
23283         the long double type if the user configured an explicit type.
23284         * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
23285         have no long double multilibs if not defined.
23286         * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
23287         warn if the user used -mabi={ieee,ibm}longdouble and we built
23288         multilibs for long double.
23289         * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
23290         appropriate multilib option.
23291         (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
23292         multilib options.
23293         * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
23294         for building long double multilibs.
23295         * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
23297 2018-01-16  John David Anglin  <danglin@gcc.gnu.org>
23299         * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
23300         copies.
23302         * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
23303         64 bits.
23304         * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
23305         128 bits.
23307         * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
23308         variables.
23310         * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
23311         return value.
23313 2018-01-16  Eric Botcazou  <ebotcazou@adacore.com>
23315         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
23316         ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
23318 2018-01-16  Kelvin Nilsen  <kelvin@gcc.gnu.org>
23320         * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
23321         different rtl trees depending on TARGET_64BIT.
23322         (rs6000_gen_lvx): Likewise.
23324 2018-01-16  Eric Botcazou  <ebotcazou@adacore.com>
23326         * config/visium/visium.md (nop): Tweak comment.
23327         (hazard_nop): Likewise.
23329 2018-01-16  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
23331         * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
23332         -mspeculate-indirect-jumps.
23333         * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
23334         for -mno-speculate-indirect-jumps.
23335         (*call_indirect_elfv2<mode>_nospec): New define_insn.
23336         (*call_value_indirect_elfv2<mode>): Disable for
23337         -mno-speculate-indirect-jumps.
23338         (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
23339         (indirect_jump): Emit different RTL for
23340         -mno-speculate-indirect-jumps.
23341         (*indirect_jump<mode>): Disable for
23342         -mno-speculate-indirect-jumps.
23343         (*indirect_jump<mode>_nospec): New define_insn.
23344         (tablejump): Emit different RTL for
23345         -mno-speculate-indirect-jumps.
23346         (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
23347         (tablejumpsi_nospec): New define_expand.
23348         (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
23349         (tablejumpdi_nospec): New define_expand.
23350         (*tablejump<mode>_internal1): Disable for
23351         -mno-speculate-indirect-jumps.
23352         (*tablejump<mode>_internal1_nospec): New define_insn.
23353         * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
23354         option.
23356 2018-01-16  Artyom Skrobov tyomitch@gmail.com
23358         * caller-save.c (insert_save): Drop unnecessary parameter.  All
23359         callers updated.
23361 2018-01-16  Jakub Jelinek  <jakub@redhat.com>
23362             Richard Biener  <rguenth@suse.de>
23364         PR libgomp/83590
23365         * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
23366         return early, inline manually is_gimple_sizepos.  Make sure if we
23367         call gimplify_expr we don't end up with a gimple constant.
23368         * tree.c (variably_modified_type_p): Don't return true for
23369         is_gimple_constant (_t).  Inline manually is_gimple_sizepos.
23370         * gimplify.h (is_gimple_sizepos): Remove.
23372 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
23374         PR tree-optimization/83857
23375         * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
23376         vectorizable_live_operation for pure SLP statements.
23377         (vectorizable_live_operation): Handle PHIs.
23379 2018-01-16  Richard Biener  <rguenther@suse.de>
23381         PR tree-optimization/83867
23382         * tree-vect-stmts.c (vect_transform_stmt): Precompute
23383         nested_in_vect_loop_p since the scalar stmt may get invalidated.
23385 2018-01-16  Jakub Jelinek  <jakub@redhat.com>
23387         PR c/83844
23388         * stor-layout.c (handle_warn_if_not_align): Use byte_position and
23389         multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
23390         If off is not INTEGER_CST, issue a may not be aligned warning
23391         rather than isn't aligned.  Use isn%'t rather than isn't.
23392         * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
23393         into MULT_EXPR.
23394         <case MULT_EXPR>: Improve the case when bottom and one of the
23395         MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
23396         operand, in that case check if the other operand is multiple of
23397         bottom divided by the INTEGER_CST operand.
23399 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
23401         PR target/83858
23402         * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
23403         * config/pa/pa-protos.h (pa_function_arg_size): Declare.
23404         * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
23405         pa_function_arg_size instead of FUNCTION_ARG_SIZE.
23406         * config/pa/pa.c (pa_function_arg_advance): Likewise.
23407         (pa_function_arg, pa_arg_partial_bytes): Likewise.
23408         (pa_function_arg_size): New function.
23410 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
23412         * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
23413         in a separate statement.
23415 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
23417         PR tree-optimization/83847
23418         * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
23419         group gathers and scatters.
23421 2018-01-16  Jakub Jelinek  <jakub@redhat.com>
23423         PR rtl-optimization/86620
23424         * params.def (max-sched-ready-insns): Bump minimum value to 1.
23426         PR rtl-optimization/83213
23427         * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
23428         to last if both are JUMP_INSNs.
23430         PR tree-optimization/83843
23431         * gimple-ssa-store-merging.c
23432         (imm_store_chain_info::output_merged_store): Handle bit_not_p on
23433         store_immediate_info for bswap/nop orig_stores.
23435 2018-01-15  Andrew Waterman  <andrew@sifive.com>
23437         * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
23438         !TARGET_MUL.
23439         <UDIV>: Increase cost if !TARGET_DIV.
23441 2018-01-15  Segher Boessenkool  <segher@kernel.crashing.org>
23443         * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
23444         (define_attr "cr_logical_3op"): New.
23445         (cceq_ior_compare): Adjust.
23446         (cceq_ior_compare_complement): Adjust.
23447         (*cceq_rev_compare): Adjust.
23448         * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
23449         (is_cracked_insn): Adjust.
23450         (insn_must_be_first_in_group): Adjust.
23451         * config/rs6000/40x.md: Adjust.
23452         * config/rs6000/440.md: Adjust.
23453         * config/rs6000/476.md: Adjust.
23454         * config/rs6000/601.md: Adjust.
23455         * config/rs6000/603.md: Adjust.
23456         * config/rs6000/6xx.md: Adjust.
23457         * config/rs6000/7450.md: Adjust.
23458         * config/rs6000/7xx.md: Adjust.
23459         * config/rs6000/8540.md: Adjust.
23460         * config/rs6000/cell.md: Adjust.
23461         * config/rs6000/e300c2c3.md: Adjust.
23462         * config/rs6000/e500mc.md: Adjust.
23463         * config/rs6000/e500mc64.md: Adjust.
23464         * config/rs6000/e5500.md: Adjust.
23465         * config/rs6000/e6500.md: Adjust.
23466         * config/rs6000/mpc.md: Adjust.
23467         * config/rs6000/power4.md: Adjust.
23468         * config/rs6000/power5.md: Adjust.
23469         * config/rs6000/power6.md: Adjust.
23470         * config/rs6000/power7.md: Adjust.
23471         * config/rs6000/power8.md: Adjust.
23472         * config/rs6000/power9.md: Adjust.
23473         * config/rs6000/rs64.md: Adjust.
23474         * config/rs6000/titan.md: Adjust.
23476 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
23478         * config/i386/predicates.md (indirect_branch_operand): Rewrite
23479         ix86_indirect_branch_register logic.
23481 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
23483         * config/i386/constraints.md (Bs): Update
23484         ix86_indirect_branch_register check.  Don't check
23485         ix86_indirect_branch_register with GOT_memory_operand.
23486         (Bw): Likewise.
23487         * config/i386/predicates.md (GOT_memory_operand): Don't check
23488         ix86_indirect_branch_register here.
23489         (GOT32_symbol_operand): Likewise.
23491 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
23493         * config/i386/predicates.md (constant_call_address_operand):
23494         Rewrite ix86_indirect_branch_register logic.
23495         (sibcall_insn_operand): Likewise.
23497 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
23499         * config/i386/constraints.md (Bs): Replace
23500         ix86_indirect_branch_thunk_register with
23501         ix86_indirect_branch_register.
23502         (Bw): Likewise.
23503         * config/i386/i386.md (indirect_jump): Likewise.
23504         (tablejump): Likewise.
23505         (*sibcall_memory): Likewise.
23506         (*sibcall_value_memory): Likewise.
23507         Peepholes of indirect call and jump via memory: Likewise.
23508         * config/i386/i386.opt: Likewise.
23509         * config/i386/predicates.md (indirect_branch_operand): Likewise.
23510         (GOT_memory_operand): Likewise.
23511         (call_insn_operand): Likewise.
23512         (sibcall_insn_operand): Likewise.
23513         (GOT32_symbol_operand): Likewise.
23515 2018-01-15  Jakub Jelinek  <jakub@redhat.com>
23517         PR middle-end/83837
23518         * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
23519         type rather than type addr's type points to.
23520         (expand_omp_atomic_mutex): Likewise.
23521         (expand_omp_atomic): Likewise.
23523 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
23525         PR target/83839
23526         * config/i386/i386.c (output_indirect_thunk_function): Use
23527         ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
23528         for  __x86_return_thunk.
23530 2018-01-15  Richard Biener  <rguenther@suse.de>
23532         PR middle-end/83850
23533         * expmed.c (extract_bit_field_1): Fix typo.
23535 2018-01-15  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
23537         PR target/83687
23538         * config/arm/iterators.md (VF): New mode iterator.
23539         * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
23540         Remove integer-related logic from pattern.
23541         (neon_vabd<mode>_3): Likewise.
23543 2018-01-15  Jakub Jelinek  <jakub@redhat.com>
23545         PR middle-end/82694
23546         * common.opt (fstrict-overflow): No longer an alias.
23547         (fwrapv-pointer): New option.
23548         * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
23549         also for pointer types based on flag_wrapv_pointer.
23550         * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
23551         opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
23552         opts->x_flag_wrapv got set.
23553         * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
23554         changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
23555         POINTER_TYPE_OVERFLOW_UNDEFINED.
23556         * match.pd: Likewise in address comparison pattern.
23557         * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
23559 2018-01-15  Richard Biener  <rguenther@suse.de>
23561         PR lto/83804
23562         * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
23563         from TYPE_FIELDS.  Free TYPE_BINFO if not used by devirtualization.
23564         Reset type names to their identifier if their TYPE_DECL doesn't
23565         have linkage (and thus is used for ODR and devirt).
23566         (save_debug_info_for_decl): Remove.
23567         (save_debug_info_for_type): Likewise.
23568         (add_tree_to_fld_list): Adjust.
23569         * tree-pretty-print.c (dump_generic_node): Make dumping of
23570         type names more robust.
23572 2018-01-15  Richard Biener  <rguenther@suse.de>
23574         * BASE-VER: Bump to 8.0.1.
23576 2018-01-14  Martin Sebor  <msebor@redhat.com>
23578         PR other/83508
23579         * builtins.c (check_access): Avoid warning when the no-warning bit
23580         is set.
23582 2018-01-14  Cory Fields  <cory-nospam-@coryfields.com>
23584         * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
23585         * ira-color (allocno_hard_regs_compare): Likewise.
23587 2018-01-14  Nathan Rossi  <nathan@nathanrossi.com>
23589         PR target/83013
23590         * config/microblaze/microblaze.c (microblaze_asm_output_ident):
23591         Use .pushsection/.popsection.
23593 2018-01-14  Martin Sebor  <msebor@redhat.com>
23595         PR c++/81327
23596         * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
23598 2018-01-14  Jakub Jelinek  <jakub@redhat.com>
23600         * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
23601         entry from extra_headers.
23602         (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
23603         extra_headers, make the list bitwise identical to the i?86-*-* one.
23605 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
23607         * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
23608         -mcmodel=large with -mindirect-branch=thunk,
23609         -mindirect-branch=thunk-extern, -mfunction-return=thunk and
23610         -mfunction-return=thunk-extern.
23611         * doc/invoke.texi: Document -mcmodel=large is incompatible with
23612         -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
23613         -mfunction-return=thunk and -mfunction-return=thunk-extern.
23615 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
23617         * config/i386/i386.c (print_reg): Print the name of the full
23618         integer register without '%'.
23619         (ix86_print_operand): Handle 'V'.
23620         * doc/extend.texi: Document 'V' modifier.
23622 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
23624         * config/i386/constraints.md (Bs): Disallow memory operand for
23625         -mindirect-branch-register.
23626         (Bw): Likewise.
23627         * config/i386/predicates.md (indirect_branch_operand): Likewise.
23628         (GOT_memory_operand): Likewise.
23629         (call_insn_operand): Likewise.
23630         (sibcall_insn_operand): Likewise.
23631         (GOT32_symbol_operand): Likewise.
23632         * config/i386/i386.md (indirect_jump): Call convert_memory_address
23633         for -mindirect-branch-register.
23634         (tablejump): Likewise.
23635         (*sibcall_memory): Likewise.
23636         (*sibcall_value_memory): Likewise.
23637         Disallow peepholes of indirect call and jump via memory for
23638         -mindirect-branch-register.
23639         (*call_pop): Replace m with Bw.
23640         (*call_value_pop): Likewise.
23641         (*sibcall_pop_memory): Replace m with Bs.
23642         * config/i386/i386.opt (mindirect-branch-register): New option.
23643         * doc/invoke.texi: Document -mindirect-branch-register option.
23645 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
23647         * config/i386/i386-protos.h (ix86_output_function_return): New.
23648         * config/i386/i386.c (ix86_set_indirect_branch_type): Also
23649         set function_return_type.
23650         (indirect_thunk_name): Add ret_p to indicate thunk for function
23651         return.
23652         (output_indirect_thunk_function): Pass false to
23653         indirect_thunk_name.
23654         (ix86_output_indirect_branch_via_reg): Likewise.
23655         (ix86_output_indirect_branch_via_push): Likewise.
23656         (output_indirect_thunk_function): Create alias for function
23657         return thunk if regno < 0.
23658         (ix86_output_function_return): New function.
23659         (ix86_handle_fndecl_attribute): Handle function_return.
23660         (ix86_attribute_table): Add function_return.
23661         * config/i386/i386.h (machine_function): Add
23662         function_return_type.
23663         * config/i386/i386.md (simple_return_internal): Use
23664         ix86_output_function_return.
23665         (simple_return_internal_long): Likewise.
23666         * config/i386/i386.opt (mfunction-return=): New option.
23667         (indirect_branch): Mention -mfunction-return=.
23668         * doc/extend.texi: Document function_return function attribute.
23669         * doc/invoke.texi: Document -mfunction-return= option.
23671 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
23673         * config/i386/i386-opts.h (indirect_branch): New.
23674         * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
23675         * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
23676         with local indirect jump when converting indirect call and jump.
23677         (ix86_set_indirect_branch_type): New.
23678         (ix86_set_current_function): Call ix86_set_indirect_branch_type.
23679         (indirectlabelno): New.
23680         (indirect_thunk_needed): Likewise.
23681         (indirect_thunk_bnd_needed): Likewise.
23682         (indirect_thunks_used): Likewise.
23683         (indirect_thunks_bnd_used): Likewise.
23684         (INDIRECT_LABEL): Likewise.
23685         (indirect_thunk_name): Likewise.
23686         (output_indirect_thunk): Likewise.
23687         (output_indirect_thunk_function): Likewise.
23688         (ix86_output_indirect_branch_via_reg): Likewise.
23689         (ix86_output_indirect_branch_via_push): Likewise.
23690         (ix86_output_indirect_branch): Likewise.
23691         (ix86_output_indirect_jmp): Likewise.
23692         (ix86_code_end): Call output_indirect_thunk_function if needed.
23693         (ix86_output_call_insn): Call ix86_output_indirect_branch if
23694         needed.
23695         (ix86_handle_fndecl_attribute): Handle indirect_branch.
23696         (ix86_attribute_table): Add indirect_branch.
23697         * config/i386/i386.h (machine_function): Add indirect_branch_type
23698         and has_local_indirect_jump.
23699         * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
23700         to true.
23701         (tablejump): Likewise.
23702         (*indirect_jump): Use ix86_output_indirect_jmp.
23703         (*tablejump_1): Likewise.
23704         (simple_return_indirect_internal): Likewise.
23705         * config/i386/i386.opt (mindirect-branch=): New option.
23706         (indirect_branch): New.
23707         (keep): Likewise.
23708         (thunk): Likewise.
23709         (thunk-inline): Likewise.
23710         (thunk-extern): Likewise.
23711         * doc/extend.texi: Document indirect_branch function attribute.
23712         * doc/invoke.texi: Document -mindirect-branch= option.
23714 2018-01-14  Jan Hubicka  <hubicka@ucw.cz>
23716         PR ipa/83051
23717         * ipa-inline.c (edge_badness): Tolerate roundoff errors.
23719 2018-01-14  Richard Sandiford  <richard.sandiford@linaro.org>
23721         * ipa-inline.c (want_inline_small_function_p): Return false if
23722         inlining has already failed with CIF_FINAL_ERROR.
23723         (update_caller_keys): Call want_inline_small_function_p before
23724         can_inline_edge_p.
23725         (update_callee_keys): Likewise.
23727 2018-01-10  Kelvin Nilsen  <kelvin@gcc.gnu.org>
23729         * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
23730         New function.
23731         (rs6000_quadword_masked_address_p): Likewise.
23732         (quad_aligned_load_p): Likewise.
23733         (quad_aligned_store_p): Likewise.
23734         (const_load_sequence_p): Add comment to describe the outer-most loop.
23735         (mimic_memory_attributes_and_flags): New function.
23736         (rs6000_gen_stvx): Likewise.
23737         (replace_swapped_aligned_store): Likewise.
23738         (rs6000_gen_lvx): Likewise.
23739         (replace_swapped_aligned_load): Likewise.
23740         (replace_swapped_load_constant): Capitalize argument name in
23741         comment describing this function.
23742         (rs6000_analyze_swaps): Add a third pass to search for vector loads
23743         and stores that access quad-word aligned addresses and replace
23744         with stvx or lvx instructions when appropriate.
23745         * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
23746         New function prototype.
23747         (rs6000_quadword_masked_address_p): Likewise.
23748         (rs6000_gen_lvx): Likewise.
23749         (rs6000_gen_stvx): Likewise.
23750         * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
23751         VSX_D (V2DF, V2DI), modify this split to select lvx instruction
23752         when memory address is aligned.
23753         (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
23754         this split to select lvx instruction when memory address is aligned.
23755         (*vsx_le_perm_load_v8hi): Modify this split to select lvx
23756         instruction when memory address is aligned.
23757         (*vsx_le_perm_load_v16qi): Likewise.
23758         (four unnamed splitters): Modify to select the stvx instruction
23759         when memory is aligned.
23761 2018-01-13  Jan Hubicka  <hubicka@ucw.cz>
23763         * predict.c (determine_unlikely_bbs): Handle correctly BBs
23764         which appears in the queue multiple times.
23766 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
23767             Alan Hayward  <alan.hayward@arm.com>
23768             David Sherwood  <david.sherwood@arm.com>
23770         * tree-vectorizer.h (vec_lower_bound): New structure.
23771         (_loop_vec_info): Add check_nonzero and lower_bounds.
23772         (LOOP_VINFO_CHECK_NONZERO): New macro.
23773         (LOOP_VINFO_LOWER_BOUNDS): Likewise.
23774         (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
23775         * tree-data-ref.h (dr_with_seg_len): Add access_size and align
23776         fields.  Make seg_len the distance travelled, not including the
23777         access size.
23778         (dr_direction_indicator): Declare.
23779         (dr_zero_step_indicator): Likewise.
23780         (dr_known_forward_stride_p): Likewise.
23781         * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
23782         tree-ssanames.h.
23783         (runtime_alias_check_p): Allow runtime alias checks with
23784         variable strides.
23785         (operator ==): Compare access_size and align.
23786         (prune_runtime_alias_test_list): Rework for new distinction between
23787         the access_size and seg_len.
23788         (create_intersect_range_checks_index): Likewise.  Cope with polynomial
23789         segment lengths.
23790         (get_segment_min_max): New function.
23791         (create_intersect_range_checks): Use it.
23792         (dr_step_indicator): New function.
23793         (dr_direction_indicator): Likewise.
23794         (dr_zero_step_indicator): Likewise.
23795         (dr_known_forward_stride_p): Likewise.
23796         * tree-loop-distribution.c (data_ref_segment_size): Return
23797         DR_STEP * (niters - 1).
23798         (compute_alias_check_pairs): Update call to the dr_with_seg_len
23799         constructor.
23800         * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
23801         (vect_preserves_scalar_order_p): New function, split out from...
23802         (vect_analyze_data_ref_dependence): ...here.  Check for zero steps.
23803         (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
23804         (vect_vfa_access_size): New function.
23805         (vect_vfa_align): Likewise.
23806         (vect_compile_time_alias): Take access_size_a and access_b arguments.
23807         (dump_lower_bound): New function.
23808         (vect_check_lower_bound): Likewise.
23809         (vect_small_gap_p): Likewise.
23810         (vectorizable_with_step_bound_p): Likewise.
23811         (vect_prune_runtime_alias_test_list): Ignore cross-iteration
23812         depencies if the vectorization factor is 1.  Convert the checks
23813         for nonzero steps into checks on the bounds of DR_STEP.  Try using
23814         a bunds check for variable steps if the minimum required step is
23815         relatively small. Update calls to the dr_with_seg_len
23816         constructor and to vect_compile_time_alias.
23817         * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
23818         function.
23819         (vect_loop_versioning): Call it.
23820         * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
23821         when retrying.
23822         (vect_estimate_min_profitable_iters): Account for any bounds checks.
23824 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
23825             Alan Hayward  <alan.hayward@arm.com>
23826             David Sherwood  <david.sherwood@arm.com>
23828         * doc/sourcebuild.texi (vect_scatter_store): Document.
23829         * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
23830         optabs.
23831         * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
23832         Document.
23833         * genopinit.c (main): Add supports_vec_scatter_store and
23834         supports_vec_scatter_store_cached to target_optabs.
23835         * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
23836         IFN_MASK_SCATTER_STORE.
23837         * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
23838         functions.
23839         * internal-fn.h (internal_store_fn_p): Declare.
23840         (internal_fn_stored_value_index): Likewise.
23841         * internal-fn.c (scatter_store_direct): New macro.
23842         (expand_scatter_store_optab_fn): New function.
23843         (direct_scatter_store_optab_supported_p): New macro.
23844         (internal_store_fn_p): New function.
23845         (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
23846         IFN_MASK_SCATTER_STORE.
23847         (internal_fn_mask_index): Likewise.
23848         (internal_fn_stored_value_index): New function.
23849         (internal_gather_scatter_fn_supported_p): Adjust operand numbers
23850         for scatter stores.
23851         * optabs-query.h (supports_vec_scatter_store_p): Declare.
23852         * optabs-query.c (supports_vec_scatter_store_p): New function.
23853         * tree-vectorizer.h (vect_get_store_rhs): Declare.
23854         * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
23855         true for scatter stores.
23856         (vect_gather_scatter_fn_p): Handle scatter stores too.
23857         (vect_check_gather_scatter): Consider using scatter stores if
23858         supports_vec_scatter_store_p.
23859         * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
23860         scatter stores too.
23861         * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
23862         internal_fn_stored_value_index.
23863         (check_load_store_masking): Handle scatter stores too.
23864         (vect_get_store_rhs): Make public.
23865         (vectorizable_call): Use internal_store_fn_p.
23866         (vectorizable_store): Handle scatter store internal functions.
23867         (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
23868         when deciding whether the end of the group has been reached.
23869         * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
23870         * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
23871         (mask_scatter_store<mode>): New insns.
23873 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
23874             Alan Hayward  <alan.hayward@arm.com>
23875             David Sherwood  <david.sherwood@arm.com>
23877         * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
23878         * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
23879         * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
23880         function.
23881         (vect_use_strided_gather_scatters_p): Take a masked_p argument.
23882         Use vect_truncate_gather_scatter_offset if we can't treat the
23883         operation as a normal gather load or scatter store.
23884         (get_group_load_store_type): Take the gather_scatter_info
23885         as argument.  Try using a gather load or scatter store for
23886         single-element groups.
23887         (get_load_store_type): Update calls to get_group_load_store_type
23888         and vect_use_strided_gather_scatters_p.
23890 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
23891             Alan Hayward  <alan.hayward@arm.com>
23892             David Sherwood  <david.sherwood@arm.com>
23894         * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
23895         optional tree argument.
23896         * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
23897         null target hooks.
23898         (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
23899         but continue to use the current value as a fallback.
23900         (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
23901         to compare the updates.
23902         * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
23903         (get_load_store_type): Use it when handling a strided access.
23904         (vect_get_strided_load_store_ops): New function.
23905         (vect_get_data_ptr_increment): Likewise.
23906         (vectorizable_load): Handle strided gather loads.  Always pass
23907         a step to vect_create_data_ref_ptr and bump_vector_ptr.
23909 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
23910             Alan Hayward  <alan.hayward@arm.com>
23911             David Sherwood  <david.sherwood@arm.com>
23913         * doc/md.texi (gather_load@var{m}): Document.
23914         (mask_gather_load@var{m}): Likewise.
23915         * genopinit.c (main): Add supports_vec_gather_load and
23916         supports_vec_gather_load_cached to target_optabs.
23917         * optabs-tree.c (init_tree_optimization_optabs): Use
23918         ggc_cleared_alloc to allocate target_optabs.
23919         * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
23920         * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
23921         functions.
23922         * internal-fn.h (internal_load_fn_p): Declare.
23923         (internal_gather_scatter_fn_p): Likewise.
23924         (internal_fn_mask_index): Likewise.
23925         (internal_gather_scatter_fn_supported_p): Likewise.
23926         * internal-fn.c (gather_load_direct): New macro.
23927         (expand_gather_load_optab_fn): New function.
23928         (direct_gather_load_optab_supported_p): New macro.
23929         (direct_internal_fn_optab): New function.
23930         (internal_load_fn_p): Likewise.
23931         (internal_gather_scatter_fn_p): Likewise.
23932         (internal_fn_mask_index): Likewise.
23933         (internal_gather_scatter_fn_supported_p): Likewise.
23934         * optabs-query.c (supports_at_least_one_mode_p): New function.
23935         (supports_vec_gather_load_p): Likewise.
23936         * optabs-query.h (supports_vec_gather_load_p): Declare.
23937         * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
23938         and memory_type field.
23939         (NUM_PATTERNS): Bump to 15.
23940         * tree-vect-data-refs.c: Include internal-fn.h.
23941         (vect_gather_scatter_fn_p): New function.
23942         (vect_describe_gather_scatter_call): Likewise.
23943         (vect_check_gather_scatter): Try using internal functions for
23944         gather loads.  Recognize existing calls to a gather load function.
23945         (vect_analyze_data_refs): Consider using gather loads if
23946         supports_vec_gather_load_p.
23947         * tree-vect-patterns.c (vect_get_load_store_mask): New function.
23948         (vect_get_gather_scatter_offset_type): Likewise.
23949         (vect_convert_mask_for_vectype): Likewise.
23950         (vect_add_conversion_to_patterm): Likewise.
23951         (vect_try_gather_scatter_pattern): Likewise.
23952         (vect_recog_gather_scatter_pattern): New pattern recognizer.
23953         (vect_vect_recog_func_ptrs): Add it.
23954         * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
23955         internal_fn_mask_index and internal_gather_scatter_fn_p.
23956         (check_load_store_masking): Take the gather_scatter_info as an
23957         argument and handle gather loads.
23958         (vect_get_gather_scatter_ops): New function.
23959         (vectorizable_call): Check internal_load_fn_p.
23960         (vectorizable_load): Likewise.  Handle gather load internal
23961         functions.
23962         (vectorizable_store): Update call to check_load_store_masking.
23963         * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
23964         * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
23965         * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
23966         (aarch64_gather_scale_operand_d): New predicates.
23967         * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
23968         (mask_gather_load<mode>): New insns.
23970 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
23971             Alan Hayward  <alan.hayward@arm.com>
23972             David Sherwood  <david.sherwood@arm.com>
23974         * optabs.def (fold_left_plus_optab): New optab.
23975         * doc/md.texi (fold_left_plus_@var{m}): Document.
23976         * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
23977         * internal-fn.c (fold_left_direct): Define.
23978         (expand_fold_left_optab_fn): Likewise.
23979         (direct_fold_left_optab_supported_p): Likewise.
23980         * fold-const-call.c (fold_const_fold_left): New function.
23981         (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
23982         * tree-parloops.c (valid_reduction_p): New function.
23983         (gather_scalar_reductions): Use it.
23984         * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
23985         (vect_finish_replace_stmt): Declare.
23986         * tree-vect-loop.c (fold_left_reduction_fn): New function.
23987         (needs_fold_left_reduction_p): New function, split out from...
23988         (vect_is_simple_reduction): ...here.  Accept reductions that
23989         forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
23990         (vect_force_simple_reduction): Also store the reduction type in
23991         the assignment's STMT_VINFO_REDUC_TYPE.
23992         (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
23993         (merge_with_identity): New function.
23994         (vect_expand_fold_left): Likewise.
23995         (vectorize_fold_left_reduction): Likewise.
23996         (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION.  Leave the
23997         scalar phi in place for it.  Check for target support and reject
23998         cases that would reassociate the operation.  Defer the transform
23999         phase to vectorize_fold_left_reduction.
24000         * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
24001         * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
24002         (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
24004 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
24006         * tree-if-conv.c (predicate_mem_writes): Remove redundant
24007         call to ifc_temp_var.
24009 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
24010             Alan Hayward  <alan.hayward@arm.com>
24011             David Sherwood  <david.sherwood@arm.com>
24013         * target.def (legitimize_address_displacement): Take the original
24014         offset as a poly_int.
24015         * targhooks.h (default_legitimize_address_displacement): Update
24016         accordingly.
24017         * targhooks.c (default_legitimize_address_displacement): Likewise.
24018         * doc/tm.texi: Regenerate.
24019         * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
24020         as an argument, moving assert of ad->disp == ad->disp_term to...
24021         (process_address_1): ...here.  Update calls to base_plus_disp_to_reg.
24022         Try calling targetm.legitimize_address_displacement before expanding
24023         the address rather than afterwards, and adjust for the new interface.
24024         * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
24025         Match the new hook interface.  Handle SVE addresses.
24026         * config/sh/sh.c (sh_legitimize_address_displacement): Make the
24027         new hook interface.
24029 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
24031         * Makefile.in (OBJS): Add early-remat.o.
24032         * target.def (select_early_remat_modes): New hook.
24033         * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
24034         * doc/tm.texi: Regenerate.
24035         * targhooks.h (default_select_early_remat_modes): Declare.
24036         * targhooks.c (default_select_early_remat_modes): New function.
24037         * timevar.def (TV_EARLY_REMAT): New timevar.
24038         * passes.def (pass_early_remat): New pass.
24039         * tree-pass.h (make_pass_early_remat): Declare.
24040         * early-remat.c: New file.
24041         * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
24042         function.
24043         (TARGET_SELECT_EARLY_REMAT_MODES): Define.
24045 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
24046             Alan Hayward  <alan.hayward@arm.com>
24047             David Sherwood  <david.sherwood@arm.com>
24049         * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
24050         vfm1 with a bound_epilog parameter.
24051         (vect_do_peeling): Update calls accordingly, and move the prologue
24052         call earlier in the function.  Treat the base bound_epilog as 0 for
24053         fully-masked loops and retain vf - 1 for other loops.  Add 1 to
24054         this base when peeling for gaps.
24055         * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
24056         with fully-masked loops.
24057         (vect_estimate_min_profitable_iters): Handle the single peeled
24058         iteration in that case.
24060 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
24061             Alan Hayward  <alan.hayward@arm.com>
24062             David Sherwood  <david.sherwood@arm.com>
24064         * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
24065         single-element interleaving even if the size is not a power of 2.
24066         * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
24067         accesses for single-element interleaving if the group size is
24068         not a power of 2.
24070 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
24071             Alan Hayward  <alan.hayward@arm.com>
24072             David Sherwood  <david.sherwood@arm.com>
24074         * doc/md.texi (fold_extract_last_@var{m}): Document.
24075         * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
24076         * optabs.def (fold_extract_last_optab): New optab.
24077         * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
24078         * internal-fn.c (fold_extract_direct): New macro.
24079         (expand_fold_extract_optab_fn): Likewise.
24080         (direct_fold_extract_optab_supported_p): Likewise.
24081         * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
24082         * tree-vect-loop.c (vect_model_reduction_cost): Handle
24083         EXTRACT_LAST_REDUCTION.
24084         (get_initial_def_for_reduction): Do not create an initial vector
24085         for EXTRACT_LAST_REDUCTION reductions.
24086         (vectorizable_reduction): Leave the scalar phi in place for
24087         EXTRACT_LAST_REDUCTIONs.  Try using EXTRACT_LAST_REDUCTION
24088         ahead of INTEGER_INDUC_COND_REDUCTION.  Do not check for an
24089         epilogue code for EXTRACT_LAST_REDUCTION and defer the
24090         transform phase to vectorizable_condition.
24091         * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
24092         split out from...
24093         (vect_finish_stmt_generation): ...here.
24094         (vect_finish_replace_stmt): New function.
24095         (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
24096         * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
24097         pattern.
24098         * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
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         * doc/md.texi (extract_last_@var{m}): Document.
24105         * optabs.def (extract_last_optab): New optab.
24106         * internal-fn.def (EXTRACT_LAST): New internal function.
24107         * internal-fn.c (cond_unary_direct): New macro.
24108         (expand_cond_unary_optab_fn): Likewise.
24109         (direct_cond_unary_optab_supported_p): Likewise.
24110         * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
24111         loops using EXTRACT_LAST.
24112         * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
24113         (extract_last_<mode>): ...this optab.
24114         (vec_extract<mode><Vel>): Update accordingly.
24116 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
24117             Alan Hayward  <alan.hayward@arm.com>
24118             David Sherwood  <david.sherwood@arm.com>
24120         * target.def (empty_mask_is_expensive): New hook.
24121         * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
24122         * doc/tm.texi: Regenerate.
24123         * targhooks.h (default_empty_mask_is_expensive): Declare.
24124         * targhooks.c (default_empty_mask_is_expensive): New function.
24125         * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
24126         if the target says that empty masks are expensive.
24127         * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
24128         New function.
24129         (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
24131 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
24132             Alan Hayward  <alan.hayward@arm.com>
24133             David Sherwood  <david.sherwood@arm.com>
24135         * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
24136         (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
24137         (vect_use_loop_mask_for_alignment_p): New function.
24138         (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
24139         * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
24140         niters_skip argument.  Make sure that the first niters_skip elements
24141         of the first iteration are inactive.
24142         (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
24143         Update call to vect_set_loop_masks_directly.
24144         (get_misalign_in_elems): New function, split out from...
24145         (vect_gen_prolog_loop_niters): ...here.
24146         (vect_update_init_of_dr): Take a code argument that specifies whether
24147         the adjustment should be added or subtracted.
24148         (vect_update_init_of_drs): Likewise.
24149         (vect_prepare_for_masked_peels): New function.
24150         (vect_do_peeling): Skip prologue peeling if we're using a mask
24151         instead.  Update call to vect_update_inits_of_drs.
24152         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
24153         mask_skip_niters.
24154         (vect_analyze_loop_2): Allow fully-masked loops with peeling for
24155         alignment.  Do not include the number of peeled iterations in
24156         the minimum threshold in that case.
24157         (vectorizable_induction): Adjust the start value down by
24158         LOOP_VINFO_MASK_SKIP_NITERS iterations.
24159         (vect_transform_loop): Call vect_prepare_for_masked_peels.
24160         Take the number of skipped iterations into account when calculating
24161         the loop bounds.
24162         * tree-vect-stmts.c (vect_gen_while_not): New function.
24164 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
24165             Alan Hayward  <alan.hayward@arm.com>
24166             David Sherwood  <david.sherwood@arm.com>
24168         * doc/sourcebuild.texi (vect_fully_masked): Document.
24169         * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
24170         default value to 0.
24171         * tree-vect-loop.c (vect_analyze_loop_costing): New function,
24172         split out from...
24173         (vect_analyze_loop_2): ...here. Don't check the vectorization
24174         factor against the number of loop iterations if the loop is
24175         fully-masked.
24177 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
24178             Alan Hayward  <alan.hayward@arm.com>
24179             David Sherwood  <david.sherwood@arm.com>
24181         * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
24182         (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
24183         (dump_groups): Update accordingly.
24184         (iv_use::mem_type): New member variable.
24185         (address_p): New function.
24186         (record_use): Add a mem_type argument and initialize the new
24187         mem_type field.
24188         (record_group_use): Add a mem_type argument.  Use address_p.
24189         Remove obsolete null checks of base_object.  Update call to record_use.
24190         (find_interesting_uses_op): Update call to record_group_use.
24191         (find_interesting_uses_cond): Likewise.
24192         (find_interesting_uses_address): Likewise.
24193         (get_mem_type_for_internal_fn): New function.
24194         (find_address_like_use): Likewise.
24195         (find_interesting_uses_stmt): Try find_address_like_use before
24196         calling find_interesting_uses_op.
24197         (addr_offset_valid_p): Use the iv mem_type field as the type
24198         of the addressed memory.
24199         (add_autoinc_candidates): Likewise.
24200         (get_address_cost): Likewise.
24201         (split_small_address_groups_p): Use address_p.
24202         (split_address_groups): Likewise.
24203         (add_iv_candidate_for_use): Likewise.
24204         (autoinc_possible_for_pair): Likewise.
24205         (rewrite_groups): Likewise.
24206         (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
24207         (determine_group_iv_cost): Update after split of USE_ADDRESS.
24208         (get_alias_ptr_type_for_ptr_address): New function.
24209         (rewrite_use_address): Rewrite address uses in calls that were
24210         identified by find_address_like_use.
24212 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
24213             Alan Hayward  <alan.hayward@arm.com>
24214             David Sherwood  <david.sherwood@arm.com>
24216         * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
24217         TARGET_MEM_REFs.
24218         * gimple-expr.h (is_gimple_addressable: Likewise.
24219         * gimple-expr.c (is_gimple_address): Likewise.
24220         * internal-fn.c (expand_call_mem_ref): New function.
24221         (expand_mask_load_optab_fn): Use it.
24222         (expand_mask_store_optab_fn): Likewise.
24224 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
24225             Alan Hayward  <alan.hayward@arm.com>
24226             David Sherwood  <david.sherwood@arm.com>
24228         * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
24229         (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
24230         (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
24231         (cond_umax@var{mode}): Document.
24232         * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
24233         (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
24234         (cond_umin_optab, cond_umax_optab): New optabs.
24235         * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
24236         (COND_IOR, COND_XOR): New internal functions.
24237         * internal-fn.h (get_conditional_internal_fn): Declare.
24238         * internal-fn.c (cond_binary_direct): New macro.
24239         (expand_cond_binary_optab_fn): Likewise.
24240         (direct_cond_binary_optab_supported_p): Likewise.
24241         (get_conditional_internal_fn): New function.
24242         * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
24243         Cope with reduction statements that are vectorized as calls rather
24244         than assignments.
24245         * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
24246         * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
24247         (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
24248         (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
24249         (UNSPEC_COND_EOR): New unspecs.
24250         (optab): Add mappings for them.
24251         (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
24252         (sve_int_op, sve_fp_op): New int attributes.
24254 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
24255             Alan Hayward  <alan.hayward@arm.com>
24256             David Sherwood  <david.sherwood@arm.com>
24258         * optabs.def (while_ult_optab): New optab.
24259         * doc/md.texi (while_ult@var{m}@var{n}): Document.
24260         * internal-fn.def (WHILE_ULT): New internal function.
24261         * internal-fn.h (direct_internal_fn_supported_p): New override
24262         that takes two types as argument.
24263         * internal-fn.c (while_direct): New macro.
24264         (expand_while_optab_fn): New function.
24265         (convert_optab_supported_p): Likewise.
24266         (direct_while_optab_supported_p): New macro.
24267         * wide-int.h (wi::udiv_ceil): New function.
24268         * tree-vectorizer.h (rgroup_masks): New structure.
24269         (vec_loop_masks): New typedef.
24270         (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
24271         and fully_masked_p.
24272         (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
24273         (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
24274         (vect_max_vf): New function.
24275         (slpeel_make_loop_iterate_ntimes): Delete.
24276         (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
24277         (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
24278         (vect_record_loop_mask, vect_get_loop_mask): Likewise.
24279         * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
24280         internal-fn.h, stor-layout.h and optabs-query.h.
24281         (vect_set_loop_mask): New function.
24282         (add_preheader_seq): Likewise.
24283         (add_header_seq): Likewise.
24284         (interleave_supported_p): Likewise.
24285         (vect_maybe_permute_loop_masks): Likewise.
24286         (vect_set_loop_masks_directly): Likewise.
24287         (vect_set_loop_condition_masked): Likewise.
24288         (vect_set_loop_condition_unmasked): New function, split out from
24289         slpeel_make_loop_iterate_ntimes.
24290         (slpeel_make_loop_iterate_ntimes): Rename to..
24291         (vect_set_loop_condition): ...this.  Use vect_set_loop_condition_masked
24292         for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
24293         (vect_do_peeling): Update call accordingly.
24294         (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
24295         loops.
24296         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
24297         mask_compare_type, can_fully_mask_p and fully_masked_p.
24298         (release_vec_loop_masks): New function.
24299         (_loop_vec_info): Use it to free the loop masks.
24300         (can_produce_all_loop_masks_p): New function.
24301         (vect_get_max_nscalars_per_iter): Likewise.
24302         (vect_verify_full_masking): Likewise.
24303         (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
24304         retries, and free the mask rgroups before retrying.  Check loop-wide
24305         reasons for disallowing fully-masked loops.  Make the final decision
24306         about whether use a fully-masked loop or not.
24307         (vect_estimate_min_profitable_iters): Do not assume that peeling
24308         for the number of iterations will be needed for fully-masked loops.
24309         (vectorizable_reduction): Disable fully-masked loops.
24310         (vectorizable_live_operation): Likewise.
24311         (vect_halve_mask_nunits): New function.
24312         (vect_double_mask_nunits): Likewise.
24313         (vect_record_loop_mask): Likewise.
24314         (vect_get_loop_mask): Likewise.
24315         (vect_transform_loop): Handle the case in which the final loop
24316         iteration might handle a partial vector.  Call vect_set_loop_condition
24317         instead of slpeel_make_loop_iterate_ntimes.
24318         * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
24319         (check_load_store_masking): New function.
24320         (prepare_load_store_mask): Likewise.
24321         (vectorizable_store): Handle fully-masked loops.
24322         (vectorizable_load): Likewise.
24323         (supportable_widening_operation): Use vect_halve_mask_nunits for
24324         booleans.
24325         (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
24326         (vect_gen_while): New function.
24327         * config/aarch64/aarch64.md (umax<mode>3): New expander.
24328         (aarch64_uqdec<mode>): New insn.
24330 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
24331             Alan Hayward  <alan.hayward@arm.com>
24332             David Sherwood  <david.sherwood@arm.com>
24334         * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
24335         (reduc_xor_scal_optab): New optabs.
24336         * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
24337         (reduc_xor_scal_@var{m}): Document.
24338         * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
24339         * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
24340         internal functions.
24341         * fold-const-call.c (fold_const_call): Handle them.
24342         * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
24343         internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
24344         * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
24345         (*reduc_<bit_reduc>_scal_<mode>): New patterns.
24346         * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
24347         (UNSPEC_XORV): New unspecs.
24348         (optab): Add entries for them.
24349         (BITWISEV): New int iterator.
24350         (bit_reduc_op): New int attributes.
24352 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
24353             Alan Hayward  <alan.hayward@arm.com>
24354             David Sherwood  <david.sherwood@arm.com>
24356         * doc/md.texi (vec_shl_insert_@var{m}): New optab.
24357         * internal-fn.def (VEC_SHL_INSERT): New internal function.
24358         * optabs.def (vec_shl_insert_optab): New optab.
24359         * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
24360         (duplicate_and_interleave): Likewise.
24361         * tree-vect-loop.c: Include internal-fn.h.
24362         (neutral_op_for_slp_reduction): New function, split out from
24363         get_initial_defs_for_reduction.
24364         (get_initial_def_for_reduction): Handle option 2 for variable-length
24365         vectors by loading the neutral value into a vector and then shifting
24366         the initial value into element 0.
24367         (get_initial_defs_for_reduction): Replace the code argument with
24368         the neutral value calculated by neutral_op_for_slp_reduction.
24369         Use gimple_build_vector for constant-length vectors.
24370         Use IFN_VEC_SHL_INSERT for variable-length vectors if all
24371         but the first group_size elements have a neutral value.
24372         Use duplicate_and_interleave otherwise.
24373         (vect_create_epilog_for_reduction): Take a neutral_op parameter.
24374         Update call to get_initial_defs_for_reduction.  Handle SLP
24375         reductions for variable-length vectors by creating one vector
24376         result for each scalar result, with the elements associated
24377         with other scalar results stubbed out with the neutral value.
24378         (vectorizable_reduction): Call neutral_op_for_slp_reduction.
24379         Require IFN_VEC_SHL_INSERT for double reductions on
24380         variable-length vectors, or SLP reductions that have
24381         a neutral value.  Require can_duplicate_and_interleave_p
24382         support for variable-length unchained SLP reductions if there
24383         is no neutral value, such as for MIN/MAX reductions.  Also require
24384         the number of vector elements to be a multiple of the number of
24385         SLP statements when doing variable-length unchained SLP reductions.
24386         Update call to vect_create_epilog_for_reduction.
24387         * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
24388         and remove initial values.
24389         (duplicate_and_interleave): Make public.
24390         * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
24391         * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
24393 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
24394             Alan Hayward  <alan.hayward@arm.com>
24395             David Sherwood  <david.sherwood@arm.com>
24397         * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
24398         (can_duplicate_and_interleave_p): New function.
24399         (vect_get_and_check_slp_defs): Take the vector of statements
24400         rather than just the current one.  Remove excess parentheses.
24401         Restriction rejectinon of vect_constant_def and vect_external_def
24402         for variable-length vectors to boolean types, or types for which
24403         can_duplicate_and_interleave_p is false.
24404         (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
24405         (duplicate_and_interleave): New function.
24406         (vect_get_constant_vectors): Use gimple_build_vector for
24407         constant-length vectors and suitable variable-length constant
24408         vectors.  Use duplicate_and_interleave for other variable-length
24409         vectors.  Don't defer the update when inserting new statements.
24411 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
24412             Alan Hayward  <alan.hayward@arm.com>
24413             David Sherwood  <david.sherwood@arm.com>
24415         * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
24416         min_profitable_iters doesn't go negative.
24418 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
24419             Alan Hayward  <alan.hayward@arm.com>
24420             David Sherwood  <david.sherwood@arm.com>
24422         * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
24423         (vec_mask_store_lanes@var{m}@var{n}): Likewise.
24424         * optabs.def (vec_mask_load_lanes_optab): New optab.
24425         (vec_mask_store_lanes_optab): Likewise.
24426         * internal-fn.def (MASK_LOAD_LANES): New internal function.
24427         (MASK_STORE_LANES): Likewise.
24428         * internal-fn.c (mask_load_lanes_direct): New macro.
24429         (mask_store_lanes_direct): Likewise.
24430         (expand_mask_load_optab_fn): Handle masked operations.
24431         (expand_mask_load_lanes_optab_fn): New macro.
24432         (expand_mask_store_optab_fn): Handle masked operations.
24433         (expand_mask_store_lanes_optab_fn): New macro.
24434         (direct_mask_load_lanes_optab_supported_p): Likewise.
24435         (direct_mask_store_lanes_optab_supported_p): Likewise.
24436         * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
24437         parameter.
24438         (vect_load_lanes_supported): Likewise.
24439         * tree-vect-data-refs.c (strip_conversion): New function.
24440         (can_group_stmts_p): Likewise.
24441         (vect_analyze_data_ref_accesses): Use it instead of checking
24442         for a pair of assignments.
24443         (vect_store_lanes_supported): Take a masked_p parameter.
24444         (vect_load_lanes_supported): Likewise.
24445         * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
24446         vect_store_lanes_supported and vect_load_lanes_supported.
24447         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
24448         * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
24449         parameter.  Don't allow gaps for masked accesses.
24450         Use vect_get_store_rhs.  Update calls to vect_store_lanes_supported
24451         and vect_load_lanes_supported.
24452         (get_load_store_type): Take a masked_p parameter and update
24453         call to get_group_load_store_type.
24454         (vectorizable_store): Update call to get_load_store_type.
24455         Handle IFN_MASK_STORE_LANES.
24456         (vectorizable_load): Update call to get_load_store_type.
24457         Handle IFN_MASK_LOAD_LANES.
24459 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
24460             Alan Hayward  <alan.hayward@arm.com>
24461             David Sherwood  <david.sherwood@arm.com>
24463         * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
24464         modes for SVE.
24465         * config/aarch64/aarch64-protos.h
24466         (aarch64_sve_struct_memory_operand_p): Declare.
24467         * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
24468         (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
24469         (VPRED, vpred): Handle SVE structure modes.
24470         * config/aarch64/constraints.md (Utx): New constraint.
24471         * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
24472         (aarch64_sve_struct_nonimmediate_operand): New predicates.
24473         * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
24474         * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
24475         (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
24476         structure modes.  Split into pieces after RA.
24477         (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
24478         (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
24479         New patterns.
24480         * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
24481         SVE structure modes.
24482         (aarch64_classify_address): Likewise.
24483         (sizetochar): Move earlier in file.
24484         (aarch64_print_operand): Handle SVE register lists.
24485         (aarch64_array_mode): New function.
24486         (aarch64_sve_struct_memory_operand_p): Likewise.
24487         (TARGET_ARRAY_MODE): Redefine.
24489 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
24490             Alan Hayward  <alan.hayward@arm.com>
24491             David Sherwood  <david.sherwood@arm.com>
24493         * target.def (array_mode): New target hook.
24494         * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
24495         * doc/tm.texi: Regenerate.
24496         * hooks.h (hook_optmode_mode_uhwi_none): Declare.
24497         * hooks.c (hook_optmode_mode_uhwi_none): New function.
24498         * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
24499         targetm.array_mode.
24500         * stor-layout.c (mode_for_array): Likewise.  Support polynomial
24501         type sizes.
24503 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
24504             Alan Hayward  <alan.hayward@arm.com>
24505             David Sherwood  <david.sherwood@arm.com>
24507         * fold-const.c (fold_binary_loc): Check the argument types
24508         rather than the result type when testing for a vector operation.
24510 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
24512         * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
24513         * doc/tm.texi: Regenerate.
24515 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
24516             Alan Hayward  <alan.hayward@arm.com>
24517             David Sherwood  <david.sherwood@arm.com>
24519         * doc/invoke.texi (-msve-vector-bits=): Document new option.
24520         (sve): Document new AArch64 extension.
24521         * doc/md.texi (w): Extend the description of the AArch64
24522         constraint to include SVE vectors.
24523         (Upl, Upa): Document new AArch64 predicate constraints.
24524         * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
24525         enum.
24526         * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
24527         (msve-vector-bits=): New option.
24528         * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
24529         SVE when these are disabled.
24530         (sve): New extension.
24531         * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
24532         modes.  Adjust their number of units based on aarch64_sve_vg.
24533         (MAX_BITSIZE_MODE_ANY_MODE): Define.
24534         * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
24535         aarch64_addr_query_type.
24536         (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
24537         (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
24538         (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
24539         (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
24540         (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
24541         (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
24542         (aarch64_simd_imm_zero_p): Delete.
24543         (aarch64_check_zero_based_sve_index_immediate): Declare.
24544         (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
24545         (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
24546         (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
24547         (aarch64_sve_float_mul_immediate_p): Likewise.
24548         (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
24549         rather than an rtx.
24550         (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
24551         (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
24552         (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
24553         (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
24554         (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
24555         (aarch64_regmode_natural_size): Likewise.
24556         * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
24557         (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
24558         left one place.
24559         (AARCH64_ISA_SVE, TARGET_SVE): New macros.
24560         (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
24561         for VG and the SVE predicate registers.
24562         (V_ALIASES): Add a "z"-prefixed alias.
24563         (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
24564         (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
24565         (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
24566         (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
24567         (REG_CLASS_NAMES): Add entries for them.
24568         (REG_CLASS_CONTENTS): Likewise.  Update ALL_REGS to include VG
24569         and the predicate registers.
24570         (aarch64_sve_vg): Declare.
24571         (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
24572         (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
24573         (REGMODE_NATURAL_SIZE): Define.
24574         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
24575         SVE macros.
24576         * config/aarch64/aarch64.c: Include cfgrtl.h.
24577         (simd_immediate_info): Add a constructor for series vectors,
24578         and an associated step field.
24579         (aarch64_sve_vg): New variable.
24580         (aarch64_dbx_register_number): Handle VG and the predicate registers.
24581         (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
24582         (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
24583         (VEC_ANY_DATA, VEC_STRUCT): New constants.
24584         (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
24585         (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
24586         (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
24587         (aarch64_get_mask_mode): New functions.
24588         (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
24589         and FP_LO_REGS.  Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
24590         (aarch64_hard_regno_mode_ok): Handle VG.  Also handle the SVE
24591         predicate modes and predicate registers.  Explicitly restrict
24592         GPRs to modes of 16 bytes or smaller.  Only allow FP registers
24593         to store a vector mode if it is recognized by
24594         aarch64_classify_vector_mode.
24595         (aarch64_regmode_natural_size): New function.
24596         (aarch64_hard_regno_caller_save_mode): Return the original mode
24597         for predicates.
24598         (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
24599         (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
24600         (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
24601         (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
24602         functions.
24603         (aarch64_add_offset): Add a temp2 parameter.  Assert that temp1
24604         does not overlap dest if the function is frame-related.  Handle
24605         SVE constants.
24606         (aarch64_split_add_offset): New function.
24607         (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
24608         them aarch64_add_offset.
24609         (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
24610         and update call to aarch64_sub_sp.
24611         (aarch64_add_cfa_expression): New function.
24612         (aarch64_expand_prologue): Pass extra temporary registers to the
24613         functions above.  Handle the case in which we need to emit new
24614         DW_CFA_expressions for registers that were originally saved
24615         relative to the stack pointer, but now have to be expressed
24616         relative to the frame pointer.
24617         (aarch64_output_mi_thunk): Pass extra temporary registers to the
24618         functions above.
24619         (aarch64_expand_epilogue): Likewise.  Prevent inheritance of
24620         IP0 and IP1 values for SVE frames.
24621         (aarch64_expand_vec_series): New function.
24622         (aarch64_expand_sve_widened_duplicate): Likewise.
24623         (aarch64_expand_sve_const_vector): Likewise.
24624         (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
24625         Handle SVE constants.  Use emit_move_insn to move a force_const_mem
24626         into the register, rather than emitting a SET directly.
24627         (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
24628         (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
24629         (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
24630         (offset_9bit_signed_scaled_p): New functions.
24631         (aarch64_replicate_bitmask_imm): New function.
24632         (aarch64_bitmask_imm): Use it.
24633         (aarch64_cannot_force_const_mem): Reject expressions involving
24634         a CONST_POLY_INT.  Update call to aarch64_classify_symbol.
24635         (aarch64_classify_index): Handle SVE indices, by requiring
24636         a plain register index with a scale that matches the element size.
24637         (aarch64_classify_address): Handle SVE addresses.  Assert that
24638         the mode of the address is VOIDmode or an integer mode.
24639         Update call to aarch64_classify_symbol.
24640         (aarch64_classify_symbolic_expression): Update call to
24641         aarch64_classify_symbol.
24642         (aarch64_const_vec_all_in_range_p): New function.
24643         (aarch64_print_vector_float_operand): Likewise.
24644         (aarch64_print_operand): Handle 'N' and 'C'.  Use "zN" rather than
24645         "vN" for FP registers with SVE modes.  Handle (const ...) vectors
24646         and the FP immediates 1.0 and 0.5.
24647         (aarch64_print_address_internal): Handle SVE addresses.
24648         (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
24649         (aarch64_regno_regclass): Handle predicate registers.
24650         (aarch64_secondary_reload): Handle big-endian reloads of SVE
24651         data modes.
24652         (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
24653         (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
24654         (aarch64_convert_sve_vector_bits): New function.
24655         (aarch64_override_options): Use it to handle -msve-vector-bits=.
24656         (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
24657         rather than an rtx.
24658         (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
24659         Handle SVE vector and predicate modes.  Accept VL-based constants
24660         that need only one temporary register, and VL offsets that require
24661         no temporary registers.
24662         (aarch64_conditional_register_usage): Mark the predicate registers
24663         as fixed if SVE isn't available.
24664         (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
24665         Return true for SVE vector and predicate modes.
24666         (aarch64_simd_container_mode): Take the number of bits as a poly_int64
24667         rather than an unsigned int.  Handle SVE modes.
24668         (aarch64_preferred_simd_mode): Update call accordingly.  Handle
24669         SVE modes.
24670         (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
24671         if SVE is enabled.
24672         (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
24673         (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
24674         (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
24675         (aarch64_sve_float_mul_immediate_p): New functions.
24676         (aarch64_sve_valid_immediate): New function.
24677         (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
24678         Explicitly reject structure modes.  Check for INDEX constants.
24679         Handle PTRUE and PFALSE constants.
24680         (aarch64_check_zero_based_sve_index_immediate): New function.
24681         (aarch64_simd_imm_zero_p): Delete.
24682         (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
24683         vector modes.  Accept constants in the range of CNT[BHWD].
24684         (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
24685         ask for an Advanced SIMD mode.
24686         (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
24687         (aarch64_simd_vector_alignment): Handle SVE predicates.
24688         (aarch64_vectorize_preferred_vector_alignment): New function.
24689         (aarch64_simd_vector_alignment_reachable): Use it instead of
24690         the vector size.
24691         (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
24692         (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
24693         functions.
24694         (MAX_VECT_LEN): Delete.
24695         (expand_vec_perm_d): Add a vec_flags field.
24696         (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
24697         (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
24698         (aarch64_evpc_ext): Don't apply a big-endian lane correction
24699         for SVE modes.
24700         (aarch64_evpc_rev): Rename to...
24701         (aarch64_evpc_rev_local): ...this.  Use a predicated operation for SVE.
24702         (aarch64_evpc_rev_global): New function.
24703         (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
24704         (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
24705         MAX_VECT_LEN.
24706         (aarch64_evpc_sve_tbl): New function.
24707         (aarch64_expand_vec_perm_const_1): Update after rename of
24708         aarch64_evpc_rev.  Handle SVE permutes too, trying
24709         aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
24710         than aarch64_evpc_tbl.
24711         (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
24712         (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
24713         (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
24714         (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
24715         (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
24716         (aarch64_expand_sve_vcond): New functions.
24717         (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
24718         of aarch64_vector_mode_p.
24719         (aarch64_dwarf_poly_indeterminate_value): New function.
24720         (aarch64_compute_pressure_classes): Likewise.
24721         (aarch64_can_change_mode_class): Likewise.
24722         (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
24723         (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
24724         (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
24725         (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
24726         (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
24727         (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
24728         * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
24729         (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
24730         constraints.
24731         (Dn, Dl, Dr): Accept const as well as const_vector.
24732         (Dz): Likewise.  Compare against CONST0_RTX.
24733         * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
24734         of "vector" where appropriate.
24735         (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
24736         (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
24737         (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
24738         (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
24739         (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
24740         (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
24741         (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
24742         (v_int_equiv): Extend to SVE modes.
24743         (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
24744         mode attributes.
24745         (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
24746         (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
24747         (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
24748         (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
24749         (SVE_COND_FP_CMP): New int iterators.
24750         (perm_hilo): Handle the new unpack unspecs.
24751         (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
24752         attributes.
24753         * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
24754         (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
24755         (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
24756         (aarch64_equality_operator, aarch64_constant_vector_operand)
24757         (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
24758         (aarch64_sve_nonimmediate_operand): Likewise.
24759         (aarch64_sve_general_operand): Likewise.
24760         (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
24761         (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
24762         (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
24763         (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
24764         (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
24765         (aarch64_sve_float_arith_immediate): Likewise.
24766         (aarch64_sve_float_arith_with_sub_immediate): Likewise.
24767         (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
24768         (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
24769         (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
24770         (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
24771         (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
24772         (aarch64_sve_float_arith_operand): Likewise.
24773         (aarch64_sve_float_arith_with_sub_operand): Likewise.
24774         (aarch64_sve_float_mul_operand): Likewise.
24775         (aarch64_sve_vec_perm_operand): Likewise.
24776         (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
24777         (aarch64_mov_operand): Accept const_poly_int and const_vector.
24778         (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
24779         as well as const_vector.
24780         (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
24781         in file.  Use CONST0_RTX and CONSTM1_RTX.
24782         (aarch64_simd_or_scalar_imm_zero): Likewise.  Add match_codes.
24783         (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
24784         Use aarch64_simd_imm_zero.
24785         * config/aarch64/aarch64-sve.md: New file.
24786         * config/aarch64/aarch64.md: Include it.
24787         (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
24788         (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
24789         (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
24790         (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
24791         (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
24792         (sve): New attribute.
24793         (enabled): Disable instructions with the sve attribute unless
24794         TARGET_SVE.
24795         (movqi, movhi): Pass CONST_POLY_INT operaneds through
24796         aarch64_expand_mov_immediate.
24797         (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
24798         CNT[BHSD] immediates.
24799         (movti): Split CONST_POLY_INT moves into two halves.
24800         (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
24801         Split additions that need a temporary here if the destination
24802         is the stack pointer.
24803         (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
24804         (*add<mode>3_poly_1): New instruction.
24805         (set_clobber_cc): New expander.
24807 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
24809         * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
24810         parameter and use it instead of GET_MODE_SIZE (innermode).  Use
24811         inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
24812         Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
24813         GET_MODE_NUNITS (innermode).  Also add a first_elem parameter.
24814         Change innermode from fixed_mode_size to machine_mode.
24815         (simplify_subreg): Update call accordingly.  Handle a constant-sized
24816         subreg of a variable-length CONST_VECTOR.
24818 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
24819             Alan Hayward  <alan.hayward@arm.com>
24820             David Sherwood  <david.sherwood@arm.com>
24822         * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
24823         (add_offset_to_base): New function, split out from...
24824         (create_mem_ref): ...here.  When handling a scale other than 1,
24825         check first whether the address is valid without the offset.
24826         Add it into the base if so, leaving the index and scale as-is.
24828 2018-01-12  Jakub Jelinek  <jakub@redhat.com>
24830         PR c++/83778
24831         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
24832         fold_for_warn before checking if arg2 is INTEGER_CST.
24834 2018-01-12  Segher Boessenkool  <segher@kernel.crashing.org>
24836         * config/rs6000/predicates.md (load_multiple_operation): Delete.
24837         (store_multiple_operation): Delete.
24838         * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
24839         * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
24840         * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
24841         guarded by TARGET_STRING.
24842         (rs6000_output_load_multiple): Delete.
24843         * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
24844         OPTION_MASK_STRING / TARGET_STRING handling.
24845         (print_operand) <'N', 'O'>: Add comment that these are unused now.
24846         (const rs6000_opt_masks) <"string">: Change mask to 0.
24847         * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
24848         (MASK_STRING): Delete.
24849         * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
24850         parts.  Simplify.
24851         (load_multiple): Delete.
24852         (*ldmsi8): Delete.
24853         (*ldmsi7): Delete.
24854         (*ldmsi6): Delete.
24855         (*ldmsi5): Delete.
24856         (*ldmsi4): Delete.
24857         (*ldmsi3): Delete.
24858         (store_multiple): Delete.
24859         (*stmsi8): Delete.
24860         (*stmsi7): Delete.
24861         (*stmsi6): Delete.
24862         (*stmsi5): Delete.
24863         (*stmsi4): Delete.
24864         (*stmsi3): Delete.
24865         (movmemsi_8reg): Delete.
24866         (corresponding unnamed define_insn): Delete.
24867         (movmemsi_6reg): Delete.
24868         (corresponding unnamed define_insn): Delete.
24869         (movmemsi_4reg): Delete.
24870         (corresponding unnamed define_insn): Delete.
24871         (movmemsi_2reg): Delete.
24872         (corresponding unnamed define_insn): Delete.
24873         (movmemsi_1reg): Delete.
24874         (corresponding unnamed define_insn): Delete.
24875         * config/rs6000/rs6000.opt (mno-string): New.
24876         (mstring): Replace by deprecation warning stub.
24877         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
24879 2018-01-12  Jakub Jelinek  <jakub@redhat.com>
24881         * regrename.c (regrename_do_replace): If replacing the same
24882         reg multiple times, try to reuse last created gen_raw_REG.
24884         PR debug/81155
24885         * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
24886         main to workaround a bug in GDB.
24888 2018-01-12  Tom de Vries  <tom@codesourcery.com>
24890         PR target/83737
24891         * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
24893 2018-01-12  Vladimir Makarov  <vmakarov@redhat.com>
24895         PR rtl-optimization/80481
24896         * ira-color.c (get_cap_member): New function.
24897         (allocnos_conflict_by_live_ranges_p): Use it.
24898         (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
24899         (setup_slot_coalesced_allocno_live_ranges): Ditto.
24901 2018-01-12  Uros Bizjak  <ubizjak@gmail.com>
24903         PR target/83628
24904         * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
24905         (*saddl_se_1): Ditto.
24906         (*ssubsi_1): Ditto.
24907         (*ssubl_se_1): Ditto.
24909 2018-01-12  Richard Sandiford  <richard.sandiford@linaro.org>
24911         * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
24912         rather than wi::to_widest for DR_INITs.
24913         * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
24914         wi::to_poly_offset rather than wi::to_offset for DR_INIT.
24915         (vect_analyze_data_ref_accesses): Require both DR_INITs to be
24916         INTEGER_CSTs.
24917         (vect_analyze_group_access_1): Note that here.
24919 2018-01-12  Richard Sandiford  <richard.sandiford@linaro.org>
24921         * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
24922         polynomial type sizes.
24924 2018-01-12  Richard Sandiford  <richard.sandiford@linaro.org>
24926         * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
24927         poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
24928         (gimple_add_tmp_var): Likewise.
24930 2018-01-12  Martin Liska  <mliska@suse.cz>
24932         * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
24933         (gimple_alloc_sizes): Likewise.
24934         (dump_gimple_statistics): Use PRIu64 in printf format.
24935         * gimple.h: Change uint64_t to int.
24937 2018-01-12  Martin Liska  <mliska@suse.cz>
24939         * tree-core.h: Use uint64_t instead of int.
24940         * tree.c (tree_node_counts): Likewise.
24941         (tree_node_sizes): Likewise.
24942         (dump_tree_statistics): Use PRIu64 in printf format.
24944 2018-01-12  Martin Liska  <mliska@suse.cz>
24946         * Makefile.in: As qsort_chk is implemented in vec.c, add
24947         vec.o to linkage of gencfn-macros.
24948         * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
24949         passing the info to record_node_allocation_statistics.
24950         (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
24951         and pass the info.
24952         * ggc-common.c (struct ggc_usage): Add operator== and use
24953         it in operator< and compare function.
24954         * mem-stats.h (struct mem_usage): Likewise.
24955         * vec.c (struct vec_usage): Remove operator< and compare
24956         function. Can be simply inherited.
24958 2018-01-12  Martin Jambor  <mjambor@suse.cz>
24960         PR target/81616
24961         * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
24962         * tree-ssa-math-opts.c: Include domwalk.h.
24963         (convert_mult_to_fma_1): New function.
24964         (fma_transformation_info): New type.
24965         (fma_deferring_state): Likewise.
24966         (cancel_fma_deferring): New function.
24967         (result_of_phi): Likewise.
24968         (last_fma_candidate_feeds_initial_phi): Likewise.
24969         (convert_mult_to_fma): Added deferring logic, split actual
24970         transformation to convert_mult_to_fma_1.
24971         (math_opts_dom_walker): New type.
24972         (math_opts_dom_walker::after_dom_children): New method, body moved
24973         here from pass_optimize_widening_mul::execute, added deferring logic
24974         bits.
24975         (pass_optimize_widening_mul::execute): Moved most of code to
24976         math_opts_dom_walker::after_dom_children.
24977         * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
24978         * config/i386/i386.c (ix86_option_override_internal): Added
24979         maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
24981 2018-01-12  Richard Biener  <rguenther@suse.de>
24983         PR debug/83157
24984         * dwarf2out.c (gen_variable_die): Do not reset old_die for
24985         inline instance vars.
24987 2018-01-12  Oleg Endo  <olegendo@gcc.gnu.org>
24989         PR target/81819
24990         * config/rx/rx.c (rx_is_restricted_memory_address):
24991         Handle SUBREG case.
24993 2018-01-12  Richard Biener  <rguenther@suse.de>
24995         PR tree-optimization/80846
24996         * target.def (split_reduction): New target hook.
24997         * targhooks.c (default_split_reduction): New function.
24998         * targhooks.h (default_split_reduction): Declare.
24999         * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
25000         target requests first reduce vectors by combining low and high
25001         parts.
25002         * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
25003         (get_vectype_for_scalar_type_and_size): Export.
25004         * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
25005         * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
25006         * doc/tm.texi: Regenerate.
25007         * config/i386/i386.c (ix86_split_reduction): Implement
25008         TARGET_VECTORIZE_SPLIT_REDUCTION.
25010 2018-01-12  Eric Botcazou  <ebotcazou@adacore.com>
25012         PR target/83368
25013         * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
25014         in PIC mode except for TARGET_VXWORKS_RTP.
25015         * config/sparc/sparc.c: Include cfgrtl.h.
25016         (TARGET_INIT_PIC_REG): Define.
25017         (TARGET_USE_PSEUDO_PIC_REG): Likewise.
25018         (sparc_pic_register_p): New predicate.
25019         (sparc_legitimate_address_p): Use it.
25020         (sparc_legitimize_pic_address): Likewise.
25021         (sparc_delegitimize_address): Likewise.
25022         (sparc_mode_dependent_address_p): Likewise.
25023         (gen_load_pcrel_sym): Remove 4th parameter.
25024         (load_got_register): Adjust call to above.  Remove obsolete stuff.
25025         (sparc_expand_prologue): Do not call load_got_register here.
25026         (sparc_flat_expand_prologue): Likewise.
25027         (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
25028         (sparc_use_pseudo_pic_reg): New function.
25029         (sparc_init_pic_reg): Likewise.
25030         * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
25031         (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
25033 2018-01-12  Christophe Lyon  <christophe.lyon@linaro.org>
25035         * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
25036         Add item for branch_cost.
25038 2018-01-12  Eric Botcazou  <ebotcazou@adacore.com>
25040         PR rtl-optimization/83565
25041         * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
25042         not extend the result to a larger mode for rotate operations.
25043         (num_sign_bit_copies1): Likewise.
25045 2018-01-12  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
25047         PR target/40411
25048         * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
25049         -symbolic.
25050         Use values-Xc.o for -pedantic.
25051         Link with values-xpg4.o for C90, values-xpg6.o otherwise.
25053 2018-01-12  Martin Liska  <mliska@suse.cz>
25055         PR ipa/83054
25056         * ipa-devirt.c (final_warning_record::grow_type_warnings):
25057         New function.
25058         (possible_polymorphic_call_targets): Use it.
25059         (ipa_devirt): Likewise.
25061 2018-01-12  Martin Liska  <mliska@suse.cz>
25063         * profile-count.h (enum profile_quality): Use 0 as invalid
25064         enum value of profile_quality.
25066 2018-01-12  Chung-Ju Wu  <jasonwucj@gmail.com>
25068         * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
25069         -mext-string options.
25071 2018-01-12  Richard Biener  <rguenther@suse.de>
25073         * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
25074         DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
25075         * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
25076         Likewise.
25077         * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
25079 2018-01-11  Michael Meissner  <meissner@linux.vnet.ibm.com>
25081         * configure.ac (--with-long-double-format): Add support for the
25082         configuration option to change the default long double format on
25083         PowerPC systems.
25084         * config.gcc (powerpc*-linux*-*): Likewise.
25085         * configure: Regenerate.
25086         * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
25087         double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
25088         used without modification.
25090 2018-01-11  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
25092         * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
25093         (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
25094         * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
25095         MISC_BUILTIN_SPEC_BARRIER.
25096         (rs6000_init_builtins): Likewise.
25097         * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
25098         enum value.
25099         (speculation_barrier): New define_insn.
25100         * doc/extend.texi: Document __builtin_speculation_barrier.
25102 2018-01-11  Jakub Jelinek  <jakub@redhat.com>
25104         PR target/83203
25105         * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
25106         is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
25107         * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
25108         iterators.
25109         (ssescalarmodesuffix): Add 512-bit vectors.  Use "d" or "q" for
25110         integral modes instead of "ss" and "sd".
25111         (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
25112         vectors with 32-bit and 64-bit elements.
25113         (vecdupssescalarmodesuffix): New mode attribute.
25114         (vec_dup<mode>): Use it.
25116 2018-01-11  H.J. Lu  <hongjiu.lu@intel.com>
25118         PR target/83330
25119         * config/i386/i386.c (ix86_compute_frame_layout): Align stack
25120         frame if argument is passed on stack.
25122 2018-01-11  Jakub Jelinek  <jakub@redhat.com>
25124         PR target/82682
25125         * ree.c (combine_reaching_defs): Optimize also
25126         reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
25127         reg2=any_extend(exp); reg1=reg2;, formatting fix.
25129 2018-01-11  Jan Hubicka  <hubicka@ucw.cz>
25131         PR middle-end/83189
25132         * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
25134 2018-01-11  Jan Hubicka  <hubicka@ucw.cz>
25136         PR middle-end/83718
25137         * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
25138         after they are computed.
25140 2018-01-11  Bin Cheng  <bin.cheng@arm.com>
25142         PR tree-optimization/83695
25143         * gimple-loop-linterchange.cc
25144         (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
25145         reset cached scev information after interchange.
25146         (pass_linterchange::execute): Remove call to scev_reset_htab.
25148 2018-01-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
25150         * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
25151         vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
25152         vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
25153         vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
25154         vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
25155         vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
25156         * config/arm/arm_neon_builtins.def (vfmal_lane_low,
25157         vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
25158         vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
25159         vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
25160         vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
25161         * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
25162         (V_lane_reg): Likewise.
25163         * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
25164         New define_expand.
25165         (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
25166         (vfmal_lane_low<mode>_intrinsic,
25167         vfmal_lane_low<vfmlsel2><mode>_intrinsic,
25168         vfmal_lane_high<vfmlsel2><mode>_intrinsic,
25169         vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
25170         vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
25171         vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
25172         vfmsl_lane_high<mode>_intrinsic): New define_insns.
25174 2018-01-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
25176         * config/arm/arm-cpus.in (fp16fml): New feature.
25177         (ALL_SIMD): Add fp16fml.
25178         (armv8.2-a): Add fp16fml as an option.
25179         (armv8.3-a): Likewise.
25180         (armv8.4-a): Add fp16fml as part of fp16.
25181         * config/arm/arm.h (TARGET_FP16FML): Define.
25182         * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
25183         when appropriate.
25184         * config/arm/arm-modes.def (V2HF): Define.
25185         * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
25186         vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
25187         vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
25188         * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
25189         vfmsl_low, vfmsl_high): New set of builtins.
25190         * config/arm/iterators.md (PLUSMINUS): New code iterator.
25191         (vfml_op): New code attribute.
25192         (VFMLHALVES): New int iterator.
25193         (VFML, VFMLSEL): New mode attributes.
25194         (V_reg): Define mapping for V2HF.
25195         (V_hi, V_lo): New mode attributes.
25196         (VF_constraint): Likewise.
25197         (vfml_half, vfml_half_selector): New int attributes.
25198         * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
25199         define_expand.
25200         (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
25201         vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
25202         New define_insn.
25203         * config/arm/t-arm-elf (v8_fps): Add fp16fml.
25204         * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
25205         * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
25206         * doc/invoke.texi (ARM Options): Document fp16fml.  Update armv8.4-a
25207         documentation.
25208         * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
25209         Document new effective target and option set.
25211 2018-01-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
25213         * config/arm/arm-cpus.in (armv8_4): New feature.
25214         (ARMv8_4a): New fgroup.
25215         (armv8.4-a): New arch.
25216         * config/arm/arm-tables.opt: Regenerate.
25217         * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
25218         * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
25219         * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
25220         Add matching rules for -march=armv8.4-a and extensions.
25221         * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
25223 2018-01-11  Oleg Endo  <olegendo@gcc.gnu.org>
25225         PR target/81821
25226         * config/rx/rx.md (BW): New mode attribute.
25227         (sync_lock_test_and_setsi): Add mode suffix to insn output.
25229 2018-01-11  Richard Biener  <rguenther@suse.de>
25231         PR tree-optimization/83435
25232         * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
25233         * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
25234         * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
25236 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
25237             Alan Hayward  <alan.hayward@arm.com>
25238             David Sherwood  <david.sherwood@arm.com>
25240         * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
25241         field.
25242         (aarch64_classify_address): Initialize it.  Track polynomial offsets.
25243         (aarch64_print_address_internal): Use it to check for a zero offset.
25245 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
25246             Alan Hayward  <alan.hayward@arm.com>
25247             David Sherwood  <david.sherwood@arm.com>
25249         * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
25250         * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
25251         Return a poly_int64 rather than a HOST_WIDE_INT.
25252         (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
25253         rather than a HOST_WIDE_INT.
25254         * config/aarch64/aarch64.h (aarch64_frame): Protect with
25255         HAVE_POLY_INT_H rather than HOST_WIDE_INT.  Change locals_offset,
25256         hard_fp_offset, frame_size, initial_adjust, callee_offset and
25257         final_offset from HOST_WIDE_INT to poly_int64.
25258         * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
25259         to_constant when getting the number of units in an Advanced SIMD
25260         mode.
25261         (aarch64_builtin_vectorized_function): Check for a constant number
25262         of units.
25263         * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
25264         GET_MODE_SIZE.
25265         (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
25266         attribute instead of GET_MODE_NUNITS.
25267         * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
25268         (aarch64_class_max_nregs): Use the constant_lowest_bound of the
25269         GET_MODE_SIZE for fixed-size registers.
25270         (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
25271         (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
25272         (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
25273         (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
25274         (aarch64_print_operand, aarch64_print_address_internal)
25275         (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
25276         (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
25277         (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
25278         Handle polynomial GET_MODE_SIZE.
25279         (aarch64_hard_regno_caller_save_mode): Likewise.  Return modes
25280         wider than SImode without modification.
25281         (tls_symbolic_operand_type): Use strip_offset instead of split_const.
25282         (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
25283         (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
25284         passing and returning SVE modes.
25285         (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
25286         rather than GEN_INT.
25287         (aarch64_emit_probe_stack_range): Take the size as a poly_int64
25288         rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
25289         (aarch64_allocate_and_probe_stack_space): Likewise.
25290         (aarch64_layout_frame): Cope with polynomial offsets.
25291         (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
25292         start_offset as a poly_int64 rather than a HOST_WIDE_INT.  Track
25293         polynomial offsets.
25294         (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
25295         (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
25296         poly_int64 rather than a HOST_WIDE_INT.
25297         (aarch64_get_separate_components, aarch64_process_components)
25298         (aarch64_expand_prologue, aarch64_expand_epilogue)
25299         (aarch64_use_return_insn_p): Handle polynomial frame offsets.
25300         (aarch64_anchor_offset): New function, split out from...
25301         (aarch64_legitimize_address): ...here.
25302         (aarch64_builtin_vectorization_cost): Handle polynomial
25303         TYPE_VECTOR_SUBPARTS.
25304         (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
25305         GET_MODE_NUNITS.
25306         (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
25307         number of elements from the PARALLEL rather than the mode.
25308         (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
25309         rather than GET_MODE_BITSIZE.
25310         (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
25311         (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
25312         (aarch64_expand_vec_perm_const_1): Handle polynomial
25313         d->perm.length () and d->perm elements.
25314         (aarch64_evpc_tbl): Likewise.  Use nelt rather than GET_MODE_NUNITS.
25315         Apply to_constant to d->perm elements.
25316         (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
25317         polynomial CONST_VECTOR_NUNITS.
25318         (aarch64_move_pointer): Take amount as a poly_int64 rather
25319         than an int.
25320         (aarch64_progress_pointer): Avoid temporary variable.
25321         * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
25322         the mode attribute instead of GET_MODE.
25324 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
25325             Alan Hayward  <alan.hayward@arm.com>
25326             David Sherwood  <david.sherwood@arm.com>
25328         * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
25329         x exists before using it.
25330         (aarch64_add_constant_internal): Rename to...
25331         (aarch64_add_offset_1): ...this.  Replace regnum with separate
25332         src and dest rtxes.  Handle the case in which they're different,
25333         including when the offset is zero.  Replace scratchreg with an rtx.
25334         Use 2 additions if there is no spare register into which we can
25335         move a 16-bit constant.
25336         (aarch64_add_constant): Delete.
25337         (aarch64_add_offset): Replace reg with separate src and dest
25338         rtxes.  Take a poly_int64 offset instead of a HOST_WIDE_INT.
25339         Use aarch64_add_offset_1.
25340         (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
25341         an rtx rather than an int.  Take the delta as a poly_int64
25342         rather than a HOST_WIDE_INT.  Use aarch64_add_offset.
25343         (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
25344         (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
25345         aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
25346         (aarch64_expand_epilogue): Update calls to aarch64_add_offset
25347         and aarch64_add_sp.
25348         (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
25349         aarch64_add_constant.
25351 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
25353         * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
25354         Use scalar_float_mode.
25356 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
25358         * config/aarch64/aarch64-simd.md
25359         (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
25360         (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
25361         (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
25362         (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
25363         (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
25364         (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
25365         (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
25366         (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
25367         (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
25368         (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
25370 2018-01-11  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
25372         PR target/83514
25373         * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
25374         targ_options->x_arm_arch_string is non NULL.
25376 2018-01-11  Tamar Christina  <tamar.christina@arm.com>
25378         * config/aarch64/aarch64.h
25379         (AARCH64_FL_FOR_ARCH8_4): Add  AARCH64_FL_DOTPROD.
25381 2018-01-11  Sudakshina Das  <sudi.das@arm.com>
25383         PR target/82096
25384         * expmed.c (emit_store_flag_force): Swap if const op0
25385         and change VOIDmode to mode of op0.
25387 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
25389         PR rtl-optimization/83761
25390         * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
25391         than bytes to mode_for_size.
25393 2018-01-10  Jan Hubicka  <hubicka@ucw.cz>
25395         PR middle-end/83189
25396         * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
25397         * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
25398         profile.
25400 2018-01-10  Jan Hubicka  <hubicka@ucw.cz>
25402         PR middle-end/83575
25403         * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
25404         when in layout mode.
25405         (cfg_layout_finalize): Do not verify cfg before we are out of layout.
25406         * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
25407         partition fixup.
25409 2018-01-10  Michael Collison  <michael.collison@arm.com>
25411         * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
25412         * config/aarch64/aarch64-option-extension.def: Add
25413         AARCH64_OPT_EXTENSION of 'fp16fml'.
25414         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
25415         (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
25416         * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
25417         * config/aarch64/constraints.md (Ui7): New constraint.
25418         * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
25419         (VFMLA_SEL_W): Ditto.
25420         (f16quad): Ditto.
25421         (f16mac1): Ditto.
25422         (VFMLA16_LOW): New int iterator.
25423         (VFMLA16_HIGH): Ditto.
25424         (UNSPEC_FMLAL): New unspec.
25425         (UNSPEC_FMLSL): Ditto.
25426         (UNSPEC_FMLAL2): Ditto.
25427         (UNSPEC_FMLSL2): Ditto.
25428         (f16mac): New code attribute.
25429         * config/aarch64/aarch64-simd-builtins.def
25430         (aarch64_fmlal_lowv2sf): Ditto.
25431         (aarch64_fmlsl_lowv2sf): Ditto.
25432         (aarch64_fmlalq_lowv4sf): Ditto.
25433         (aarch64_fmlslq_lowv4sf): Ditto.
25434         (aarch64_fmlal_highv2sf): Ditto.
25435         (aarch64_fmlsl_highv2sf): Ditto.
25436         (aarch64_fmlalq_highv4sf): Ditto.
25437         (aarch64_fmlslq_highv4sf): Ditto.
25438         (aarch64_fmlal_lane_lowv2sf): Ditto.
25439         (aarch64_fmlsl_lane_lowv2sf): Ditto.
25440         (aarch64_fmlal_laneq_lowv2sf): Ditto.
25441         (aarch64_fmlsl_laneq_lowv2sf): Ditto.
25442         (aarch64_fmlalq_lane_lowv4sf): Ditto.
25443         (aarch64_fmlsl_lane_lowv4sf): Ditto.
25444         (aarch64_fmlalq_laneq_lowv4sf): Ditto.
25445         (aarch64_fmlsl_laneq_lowv4sf): Ditto.
25446         (aarch64_fmlal_lane_highv2sf): Ditto.
25447         (aarch64_fmlsl_lane_highv2sf): Ditto.
25448         (aarch64_fmlal_laneq_highv2sf): Ditto.
25449         (aarch64_fmlsl_laneq_highv2sf): Ditto.
25450         (aarch64_fmlalq_lane_highv4sf): Ditto.
25451         (aarch64_fmlsl_lane_highv4sf): Ditto.
25452         (aarch64_fmlalq_laneq_highv4sf): Ditto.
25453         (aarch64_fmlsl_laneq_highv4sf): Ditto.
25454         * config/aarch64/aarch64-simd.md:
25455         (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
25456         (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
25457         (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
25458         (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
25459         (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
25460         (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
25461         (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
25462         (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
25463         (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
25464         (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
25465         (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
25466         (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
25467         (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
25468         (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
25469         (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
25470         (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
25471         (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
25472         (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
25473         (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
25474         (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
25475         * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
25476         (vfmlsl_low_u32): Ditto.
25477         (vfmlalq_low_u32): Ditto.
25478         (vfmlslq_low_u32): Ditto.
25479         (vfmlal_high_u32): Ditto.
25480         (vfmlsl_high_u32): Ditto.
25481         (vfmlalq_high_u32): Ditto.
25482         (vfmlslq_high_u32): Ditto.
25483         (vfmlal_lane_low_u32): Ditto.
25484         (vfmlsl_lane_low_u32): Ditto.
25485         (vfmlal_laneq_low_u32): Ditto.
25486         (vfmlsl_laneq_low_u32): Ditto.
25487         (vfmlalq_lane_low_u32): Ditto.
25488         (vfmlslq_lane_low_u32): Ditto.
25489         (vfmlalq_laneq_low_u32): Ditto.
25490         (vfmlslq_laneq_low_u32): Ditto.
25491         (vfmlal_lane_high_u32): Ditto.
25492         (vfmlsl_lane_high_u32): Ditto.
25493         (vfmlal_laneq_high_u32): Ditto.
25494         (vfmlsl_laneq_high_u32): Ditto.
25495         (vfmlalq_lane_high_u32): Ditto.
25496         (vfmlslq_lane_high_u32): Ditto.
25497         (vfmlalq_laneq_high_u32): Ditto.
25498         (vfmlslq_laneq_high_u32): Ditto.
25499         * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
25500         (AARCH64_FL_FOR_ARCH8_4): New.
25501         (AARCH64_ISA_F16FML): New ISA flag.
25502         (TARGET_F16FML): New feature flag for fp16fml.
25503         (doc/invoke.texi): Document new fp16fml option.
25505 2018-01-10  Michael Collison  <michael.collison@arm.com>
25507         * config/aarch64/aarch64-builtins.c:
25508         (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
25509         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
25510         (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
25511         * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
25512         (AARCH64_ISA_SHA3): New ISA flag.
25513         (TARGET_SHA3): New feature flag for sha3.
25514         * config/aarch64/iterators.md (sha512_op): New int attribute.
25515         (CRYPTO_SHA512): New int iterator.
25516         (UNSPEC_SHA512H): New unspec.
25517         (UNSPEC_SHA512H2): Ditto.
25518         (UNSPEC_SHA512SU0): Ditto.
25519         (UNSPEC_SHA512SU1): Ditto.
25520         * config/aarch64/aarch64-simd-builtins.def
25521         (aarch64_crypto_sha512hqv2di): New builtin.
25522         (aarch64_crypto_sha512h2qv2di): Ditto.
25523         (aarch64_crypto_sha512su0qv2di): Ditto.
25524         (aarch64_crypto_sha512su1qv2di): Ditto.
25525         (aarch64_eor3qv8hi): Ditto.
25526         (aarch64_rax1qv2di): Ditto.
25527         (aarch64_xarqv2di): Ditto.
25528         (aarch64_bcaxqv8hi): Ditto.
25529         * config/aarch64/aarch64-simd.md:
25530         (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
25531         (aarch64_crypto_sha512su0qv2di): Ditto.
25532         (aarch64_crypto_sha512su1qv2di): Ditto.
25533         (aarch64_eor3qv8hi): Ditto.
25534         (aarch64_rax1qv2di): Ditto.
25535         (aarch64_xarqv2di): Ditto.
25536         (aarch64_bcaxqv8hi): Ditto.
25537         * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
25538         (vsha512h2q_u64): Ditto.
25539         (vsha512su0q_u64): Ditto.
25540         (vsha512su1q_u64): Ditto.
25541         (veor3q_u16): Ditto.
25542         (vrax1q_u64): Ditto.
25543         (vxarq_u64): Ditto.
25544         (vbcaxq_u16): Ditto.
25545         * config/arm/types.md (crypto_sha512): New type attribute.
25546         (crypto_sha3): Ditto.
25547         (doc/invoke.texi): Document new sha3 option.
25549 2018-01-10  Michael Collison  <michael.collison@arm.com>
25551         * config/aarch64/aarch64-builtins.c:
25552         (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
25553         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
25554         (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
25555         (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
25556         * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
25557         (AARCH64_ISA_SM4): New ISA flag.
25558         (TARGET_SM4): New feature flag for sm4.
25559         * config/aarch64/aarch64-simd-builtins.def
25560         (aarch64_sm3ss1qv4si): Ditto.
25561         (aarch64_sm3tt1aq4si): Ditto.
25562         (aarch64_sm3tt1bq4si): Ditto.
25563         (aarch64_sm3tt2aq4si): Ditto.
25564         (aarch64_sm3tt2bq4si): Ditto.
25565         (aarch64_sm3partw1qv4si): Ditto.
25566         (aarch64_sm3partw2qv4si): Ditto.
25567         (aarch64_sm4eqv4si): Ditto.
25568         (aarch64_sm4ekeyqv4si): Ditto.
25569         * config/aarch64/aarch64-simd.md:
25570         (aarch64_sm3ss1qv4si): Ditto.
25571         (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
25572         (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
25573         (aarch64_sm4eqv4si): Ditto.
25574         (aarch64_sm4ekeyqv4si): Ditto.
25575         * config/aarch64/iterators.md (sm3tt_op): New int iterator.
25576         (sm3part_op): Ditto.
25577         (CRYPTO_SM3TT): Ditto.
25578         (CRYPTO_SM3PART): Ditto.
25579         (UNSPEC_SM3SS1): New unspec.
25580         (UNSPEC_SM3TT1A): Ditto.
25581         (UNSPEC_SM3TT1B): Ditto.
25582         (UNSPEC_SM3TT2A): Ditto.
25583         (UNSPEC_SM3TT2B): Ditto.
25584         (UNSPEC_SM3PARTW1): Ditto.
25585         (UNSPEC_SM3PARTW2): Ditto.
25586         (UNSPEC_SM4E): Ditto.
25587         (UNSPEC_SM4EKEY): Ditto.
25588         * config/aarch64/constraints.md (Ui2): New constraint.
25589         * config/aarch64/predicates.md (aarch64_imm2): New predicate.
25590         * config/arm/types.md (crypto_sm3): New type attribute.
25591         (crypto_sm4): Ditto.
25592         * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
25593         (vsm3tt1aq_u32): Ditto.
25594         (vsm3tt1bq_u32): Ditto.
25595         (vsm3tt2aq_u32): Ditto.
25596         (vsm3tt2bq_u32): Ditto.
25597         (vsm3partw1q_u32): Ditto.
25598         (vsm3partw2q_u32): Ditto.
25599         (vsm4eq_u32): Ditto.
25600         (vsm4ekeyq_u32): Ditto.
25601         (doc/invoke.texi): Document new sm4 option.
25603 2018-01-10  Michael Collison  <michael.collison@arm.com>
25605         * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
25606         * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
25607         (AARCH64_FL_FOR_ARCH8_4): New.
25608         (AARCH64_FL_V8_4): New flag.
25609         (doc/invoke.texi): Document new armv8.4-a option.
25611 2018-01-10  Michael Collison  <michael.collison@arm.com>
25613         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
25614         (__ARM_FEATURE_AES): Define if TARGET_AES is true.
25615         (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
25616         * config/aarch64/aarch64-option-extension.def: Add
25617         AARCH64_OPT_EXTENSION of 'sha2'.
25618         (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
25619         (crypto): Disable sha2 and aes if crypto disabled.
25620         (crypto): Enable aes and sha2 if enabled.
25621         (simd): Disable sha2 and aes if simd disabled.
25622         * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
25623         New flags.
25624         (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
25625         (TARGET_SHA2): New feature flag for sha2.
25626         (TARGET_AES): New feature flag for aes.
25627         * config/aarch64/aarch64-simd.md:
25628         (aarch64_crypto_aes<aes_op>v16qi): Make pattern
25629         conditional on TARGET_AES.
25630         (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
25631         (aarch64_crypto_sha1hsi): Make pattern conditional
25632         on TARGET_SHA2.
25633         (aarch64_crypto_sha1hv4si): Ditto.
25634         (aarch64_be_crypto_sha1hv4si): Ditto.
25635         (aarch64_crypto_sha1su1v4si): Ditto.
25636         (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
25637         (aarch64_crypto_sha1su0v4si): Ditto.
25638         (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
25639         (aarch64_crypto_sha256su0v4si): Ditto.
25640         (aarch64_crypto_sha256su1v4si): Ditto.
25641         (doc/invoke.texi): Document new aes and sha2 options.
25643 2018-01-10  Martin Sebor  <msebor@redhat.com>
25645         PR tree-optimization/83781
25646         * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
25647         as string arrays.
25649 2018-01-11  Martin Sebor  <msebor@gmail.com>
25650             Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
25652         PR tree-optimization/83501
25653         PR tree-optimization/81703
25655         * tree-ssa-strlen.c (get_string_cst): Rename...
25656         (get_string_len): ...to this.  Handle global constants.
25657         (handle_char_store): Adjust.
25659 2018-01-10  Kito Cheng  <kito.cheng@gmail.com>
25660             Jim Wilson  <jimw@sifive.com>
25662         * config/riscv/riscv-protos.h (riscv_output_return): New.
25663         * config/riscv/riscv.c (struct machine_function): New naked_p field.
25664         (riscv_attribute_table, riscv_output_return),
25665         (riscv_handle_fndecl_attribute, riscv_naked_function_p),
25666         (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
25667         (riscv_compute_frame_info): Only compute frame->mask if not a naked
25668         function.
25669         (riscv_expand_prologue): Add early return for naked function.
25670         (riscv_expand_epilogue): Likewise.
25671         (riscv_function_ok_for_sibcall): Return false for naked function.
25672         (riscv_set_current_function): New.
25673         (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
25674         (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
25675         * config/riscv/riscv.md (simple_return): Call riscv_output_return.
25676         * doc/extend.texi (RISC-V Function Attributes): New.
25678 2018-01-10  Michael Meissner  <meissner@linux.vnet.ibm.com>
25680         * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
25681         check for 128-bit long double before checking TCmode.
25682         * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
25683         128-bit long doubles before checking TFmode or TCmode.
25684         (FLOAT128_IBM_P): Likewise.
25686 2018-01-10  Martin Sebor  <msebor@redhat.com>
25688         PR tree-optimization/83671
25689         * builtins.c (c_strlen): Unconditionally return zero for the empty
25690         string.
25691         Use -Warray-bounds for warnings.
25692         * gimple-fold.c (get_range_strlen): Handle non-constant lengths
25693         for non-constant array indices with COMPONENT_REF, arrays of
25694         arrays, and pointers to arrays.
25695         (gimple_fold_builtin_strlen): Determine and set length range for
25696         non-constant character arrays.
25698 2018-01-10  Aldy Hernandez  <aldyh@redhat.com>
25700         PR middle-end/81897
25701         * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
25702         empty blocks.
25704 2018-01-10  Eric Botcazou  <ebotcazou@adacore.com>
25706         * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
25708 2018-01-10  Peter Bergner  <bergner@vnet.ibm.com>
25710         PR target/83399
25711         * config/rs6000/rs6000.c (print_operand) <'y'>: Use
25712         VECTOR_MEM_ALTIVEC_OR_VSX_P.
25713         * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
25714         indexed_or_indirect_operand predicate.
25715         (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
25716         (*vsx_le_perm_load_v8hi): Likewise.
25717         (*vsx_le_perm_load_v16qi): Likewise.
25718         (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
25719         (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
25720         (*vsx_le_perm_store_v8hi): Likewise.
25721         (*vsx_le_perm_store_v16qi): Likewise.
25722         (eight unnamed splitters): Likewise.
25724 2018-01-10  Peter Bergner  <bergner@vnet.ibm.com>
25726         * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
25727         * config/rs6000/emmintrin.h: Likewise.
25728         * config/rs6000/mmintrin.h: Likewise.
25729         * config/rs6000/xmmintrin.h: Likewise.
25731 2018-01-10  David Malcolm  <dmalcolm@redhat.com>
25733         PR c++/43486
25734         * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
25735         "public_flag".
25736         * tree.c (tree_nop_conversion): Return true for location wrapper
25737         nodes.
25738         (maybe_wrap_with_location): New function.
25739         (selftest::check_strip_nops): New function.
25740         (selftest::test_location_wrappers): New function.
25741         (selftest::tree_c_tests): Call it.
25742         * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
25743         (maybe_wrap_with_location): New decl.
25744         (EXPR_LOCATION_WRAPPER_P): New macro.
25745         (location_wrapper_p): New inline function.
25746         (tree_strip_any_location_wrapper): New inline function.
25748 2018-01-10  H.J. Lu  <hongjiu.lu@intel.com>
25750         PR target/83735
25751         * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
25752         stack_realign_offset for the largest alignment of stack slot
25753         actually used.
25754         (ix86_find_max_used_stack_alignment): New function.
25755         (ix86_finalize_stack_frame_flags): Use it.  Set
25756         max_used_stack_alignment if we don't realign stack.
25757         * config/i386/i386.h (machine_function): Add
25758         max_used_stack_alignment.
25760 2018-01-10  Christophe Lyon  <christophe.lyon@linaro.org>
25762         * config/arm/arm.opt (-mbranch-cost): New option.
25763         * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
25764         account.
25766 2018-01-10  Segher Boessenkool  <segher@kernel.crashing.org>
25768         PR target/83629
25769         * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
25770         load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
25772 2018-01-10  Richard Biener  <rguenther@suse.de>
25774         PR debug/83765
25775         * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
25776         early out so it also covers the case where we have a non-NULL
25777         origin.
25779 2018-01-10  Richard Sandiford  <richard.sandiford@linaro.org>
25781         PR tree-optimization/83753
25782         * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
25783         for non-strided grouped accesses if the number of elements is 1.
25785 2018-01-10  Jan Hubicka  <hubicka@ucw.cz>
25787         PR target/81616
25788         * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
25789         * i386.h (TARGET_USE_GATHER): Define.
25790         * x86-tune.def (X86_TUNE_USE_GATHER): New.
25792 2018-01-10  Martin Liska  <mliska@suse.cz>
25794         PR bootstrap/82831
25795         * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
25796         * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
25797         partitioning.
25798         * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
25799         CLEANUP_NO_PARTITIONING is not set.
25801 2018-01-10  Richard Sandiford  <richard.sandiford@linaro.org>
25803         * doc/rtl.texi: Remove documentation of (const ...) wrappers
25804         for vectors, as a partial revert of r254296.
25805         * rtl.h (const_vec_p): Delete.
25806         (const_vec_duplicate_p): Don't test for vector CONSTs.
25807         (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
25808         * expmed.c (make_tree): Likewise.
25810         Revert:
25811         * common.md (E, F): Use CONSTANT_P instead of checking for
25812         CONST_VECTOR.
25813         * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
25814         checking for CONST_VECTOR.
25816 2018-01-09  Jan Hubicka  <hubicka@ucw.cz>
25818         PR middle-end/83575
25819         * predict.c (force_edge_cold): Handle in more sane way edges
25820         with no prediction.
25822 2018-01-09  Carl Love  <cel@us.ibm.com>
25824         * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
25825         V4SI, V4SF types.
25826         (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
25827         * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
25828         VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
25829         VMRGOW_V2DI, VMRGOW_V2DF.  Remove definition for VMRGOW.
25830         * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
25831         P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW):  Add definitions.
25832         * config/rs6000/rs6000-protos.h: Add extern defition for
25833         rs6000_generate_float2_double_code.
25834         * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
25835         function.
25836         * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
25837         (float2_v2df): Add define_expand.
25839 2018-01-09  Uros Bizjak  <ubizjak@gmail.com>
25841         PR target/83628
25842         * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
25843         op_mode in the force_to_mode call.
25845 2018-01-09  Richard Sandiford  <richard.sandiford@linaro.org>
25847         * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
25848         instead of checking each element individually.
25849         (aarch64_evpc_uzp): Likewise.
25850         (aarch64_evpc_zip): Likewise.
25851         (aarch64_evpc_ext): Likewise.
25852         (aarch64_evpc_rev): Likewise.
25853         (aarch64_evpc_dup): Test the encoding for a single duplicated element,
25854         instead of checking each element individually.  Return true without
25855         generating rtl if
25856         (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
25857         whether all selected elements come from the same input, instead of
25858         checking each element individually.  Remove calls to gen_rtx_REG,
25859         start_sequence and end_sequence and instead assert that no rtl is
25860         generated.
25862 2018-01-09  Richard Sandiford  <richard.sandiford@linaro.org>
25864         * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
25865         order of HIGH and CONST checks.
25867 2018-01-09  Richard Sandiford  <richard.sandiford@linaro.org>
25869         * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
25870         if the destination isn't an SSA_NAME.
25872 2018-01-09  Richard Biener  <rguenther@suse.de>
25874         PR tree-optimization/83668
25875         * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
25876         move prologue...
25877         (canonicalize_loop_form): ... here, renamed from ...
25878         (canonicalize_loop_closed_ssa_form): ... this and amended to
25879         swap successor edges for loop exit blocks to make us use
25880         the RPO order we need for initial schedule generation.
25882 2018-01-09  Joseph Myers  <joseph@codesourcery.com>
25884         PR tree-optimization/64811
25885         * match.pd: When optimizing comparisons with Inf, avoid
25886         introducing or losing exceptions from comparisons with NaN.
25888 2018-01-09  Martin Liska  <mliska@suse.cz>
25890         PR sanitizer/82517
25891         * asan.c (shadow_mem_size): Add gcc_assert.
25893 2018-01-09  Georg-Johann Lay  <avr@gjlay.de>
25895         Don't save registers in main().
25897         PR target/83738
25898         * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
25899         * config/avr/avr.opt (-mmain-is-OS_task): New target option.
25900         * config/avr/avr.c (avr_set_current_function): Don't error if
25901         naked, OS_task or OS_main are specified at the same time.
25902         (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
25903         OS_main.
25904         (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
25905         attribute.
25906         * common/config/avr/avr-common.c (avr_option_optimization_table):
25907         Switch on -mmain-is-OS_task for optimizing compilations.
25909 2018-01-09  Richard Biener  <rguenther@suse.de>
25911         PR tree-optimization/83572
25912         * graphite.c: Include cfganal.h.
25913         (graphite_transform_loops): Connect infinite loops to exit
25914         and remove fake edges at the end.
25916 2018-01-09  Jan Hubicka  <hubicka@ucw.cz>
25918         * ipa-inline.c (edge_badness): Revert accidental checkin.
25920 2018-01-09  Jan Hubicka  <hubicka@ucw.cz>
25922         PR ipa/80763
25923         * ipa-comdats.c (set_comdat_group): Only set comdat group of real
25924         symbols; not inline clones.
25926 2018-01-09  Jakub Jelinek  <jakub@redhat.com>
25928         PR target/83507
25929         * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
25930         hard registers.  Formatting fixes.
25932         PR preprocessor/83722
25933         * gcc.c (try_generate_repro): Pass
25934         &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
25935         &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
25936         do_report_bug.
25938 2018-01-08  Monk Chiang  <sh.chiang04@gmail.com>
25939             Kito Cheng  <kito.cheng@gmail.com>
25941         * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
25942         (riscv_leaf_function_p): Delete.
25943         (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
25945 2018-01-08  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
25947         * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
25948         function.
25949         (do_ifelse): New function.
25950         (do_isel): New function.
25951         (do_sub3): New function.
25952         (do_add3): New function.
25953         (do_load_mask_compare): New function.
25954         (do_overlap_load_compare): New function.
25955         (expand_compare_loop): New function.
25956         (expand_block_compare): Call expand_compare_loop() when appropriate.
25957         * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
25958         option description.
25959         (-mblock-compare-inline-loop-limit): New option.
25961 2018-01-08  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
25963         PR target/83677
25964         * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
25965         Reverse order of second and third operands in first alternative.
25966         * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
25967         of first and second elements in UNSPEC_VPERMR vector.
25968         (altivec_expand_vec_perm_le): Likewise.
25970 2018-01-08  Jeff Law  <law@redhat.com>
25972         PR rtl-optimizatin/81308
25973         * tree-switch-conversion.c (cfg_altered): New file scoped static.
25974         (process_switch): If group_case_labels makes a change, then set
25975         cfg_altered.
25976         (pass_convert_switch::execute): If a switch is converted, then
25977         set cfg_altered.  Return TODO_cfg_cleanup if cfg_altered is true.
25979         PR rtl-optimization/81308
25980         * recog.c (split_all_insns): Conditionally cleanup the CFG after
25981         splitting insns.
25983 2018-01-08  Vidya Praveen  <vidyapraveen@arm.com>
25985         PR target/83663 - Revert r255946
25986         * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
25987         generation for cases where splatting a value is not useful.
25988         * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
25989         across a vec_duplicate and a paradoxical subreg forming a vector
25990         mode to a vec_concat.
25992 2018-01-08  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
25994         * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
25995         -march=armv8.3-a variants.
25996         * config/arm/t-multilib: Likewise.
25997         * config/arm/t-arm-elf: Likewise.  Handle dotprod extension.
25999 2018-01-08  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
26001         * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
26002         to generate rtl.
26003         (cceq_ior_compare_complement): Give it a name so I can use it, and
26004         change boolean_or_operator predicate to boolean_operator so it can
26005         be used to generate a crand.
26006         (eqne): New code iterator.
26007         (bd/bd_neg): New code_attrs.
26008         (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
26009         a single define_insn.
26010         (<bd>tf_<mode>): A new insn pattern for the conditional form branch
26011         decrement (bdnzt/bdnzf/bdzt/bdzf).
26012         * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
26013         with the new names of the branch decrement patterns, and added the
26014         names of the branch decrement conditional patterns.
26016 2018-01-08  Richard Biener  <rguenther@suse.de>
26018         PR tree-optimization/83563
26019         * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
26020         cache.
26022 2018-01-08  Richard Biener  <rguenther@suse.de>
26024         PR middle-end/83713
26025         * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
26027 2018-01-08  Richard Biener  <rguenther@suse.de>
26029         PR tree-optimization/83685
26030         * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
26031         references to abnormals.
26033 2018-01-08  Richard Biener  <rguenther@suse.de>
26035         PR lto/83719
26036         * dwarf2out.c (output_indirect_strings): Handle empty
26037         skeleton_debug_str_hash.
26038         (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
26040 2018-01-08  Claudiu Zissulescu  <claziss@synopsys.com>
26042         * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
26043         (emit_store_direct): Likewise.
26044         (arc_trampoline_adjust_address): Likewise.
26045         (arc_asm_trampoline_template): New function.
26046         (arc_initialize_trampoline): Use asm_trampoline_template.
26047         (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
26048         * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
26049         * config/arc/arc.md (flush_icache): Delete pattern.
26051 2018-01-08  Claudiu Zissulescu  <claziss@synopsys.com>
26053         * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
26054         * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
26055         munaligned-access.
26057 2018-01-08  Sebastian Huber  <sebastian.huber@embedded-brains.de>
26059         PR target/83681
26060         * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
26061         by not USED_FOR_TARGET.
26062         (make_pass_resolve_sw_modes): Likewise.
26064 2018-01-08  Sebastian Huber  <sebastian.huber@embedded-brains.de>
26066         * config/nios2/nios2.h (nios2_section_threshold): Guard by not
26067         USED_FOR_TARGET.
26069 2018-01-08  Richard Biener  <rguenther@suse.de>
26071         PR middle-end/83580
26072         * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
26074 2018-01-08  Richard Biener  <rguenther@suse.de>
26076         PR middle-end/83517
26077         * match.pd ((t * 2) / 2) -> t): Add missing :c.
26079 2018-01-06  Aldy Hernandez  <aldyh@redhat.com>
26081         PR middle-end/81897
26082         * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
26083         basic blocks with a small number of successors.
26084         (convert_control_dep_chain_into_preds): Improve handling of
26085         forwarder blocks.
26086         (dump_predicates): Split apart into...
26087         (dump_pred_chain): ...here...
26088         (dump_pred_info): ...and here.
26089         (can_one_predicate_be_invalidated_p): Add debugging printfs.
26090         (can_chain_union_be_invalidated_p): Improve check for invalidation
26091         of paths.
26092         (uninit_uses_cannot_happen): Avoid unnecessary if
26093         convert_control_dep_chain_into_preds yielded nothing.
26095 2018-01-06  Martin Sebor  <msebor@redhat.com>
26097         PR tree-optimization/83640
26098         * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
26099         subtracting negative offset from size.
26100         (builtin_access::overlap): Adjust offset bounds of the access to fall
26101         within the size of the object if possible.
26103 2018-01-06  Richard Sandiford  <richard.sandiford@linaro.org>
26105         PR rtl-optimization/83699
26106         * expmed.c (extract_bit_field_1): Restrict the vector usage of
26107         extract_bit_field_as_subreg to cases in which the extracted
26108         value is also a vector.
26110         * lra-constraints.c (process_alt_operands): Test for the equivalence
26111         substitutions when detecting a possible reload cycle.
26113 2018-01-06  Jakub Jelinek  <jakub@redhat.com>
26115         PR debug/83480
26116         * toplev.c (process_options): Don't enable debug_nonbind_markers_p
26117         by default if flag_selective_schedling{,2}.  Formatting fixes.
26119         PR rtl-optimization/83682
26120         * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
26121         if it has non-VECTOR_MODE element mode.
26122         (vec_duplicate_p): Likewise.
26124         PR middle-end/83694
26125         * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
26126         and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
26128 2018-01-05  Jakub Jelinek  <jakub@redhat.com>
26130         PR target/83604
26131         * config/i386/i386-builtin.def
26132         (__builtin_ia32_vgf2p8affineinvqb_v64qi,
26133         __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
26134         Require also OPTION_MASK_ISA_AVX512F in addition to
26135         OPTION_MASK_ISA_GFNI.
26136         (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
26137         __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
26138         OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
26139         to OPTION_MASK_ISA_GFNI.
26140         (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
26141         OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
26142         OPTION_MASK_ISA_AVX512BW.
26143         (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
26144         OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
26145         addition to OPTION_MASK_ISA_GFNI.
26146         (__builtin_ia32_vgf2p8affineinvqb_v16qi,
26147         __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
26148         Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
26149         to OPTION_MASK_ISA_GFNI.
26150         * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
26151         a requirement for all ISAs rather than any of them with a few
26152         exceptions.
26153         (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
26154         processing.
26155         (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
26156         bitmasks to be enabled with 3 exceptions, instead of requiring any
26157         enabled ISA with lots of exceptions.
26158         * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
26159         vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
26160         Change avx512bw in isa attribute to avx512f.
26161         * config/i386/sgxintrin.h: Add license boilerplate.
26162         * config/i386/vaesintrin.h: Likewise.  Fix macro spelling __AVX512F
26163         to __AVX512F__ and __AVX512VL to __AVX512VL__.
26164         (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
26165         _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
26166         defined.
26167         * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
26168         _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
26169         temporarily sse2 rather than sse if not enabled already.
26171         PR target/83604
26172         * config/i386/sse.md (VI248_VLBW): Rename to ...
26173         (VI248_AVX512VL): ... this.  Don't guard V32HI with TARGET_AVX512BW.
26174         (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
26175         vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
26176         vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
26177         vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
26178         mode iterator instead of VI248_VLBW.
26180 2018-01-05  Jan Hubicka  <hubicka@ucw.cz>
26182         * ipa-fnsummary.c (record_modified_bb_info): Add OP.
26183         (record_modified): Skip clobbers; add debug output.
26184         (param_change_prob): Use sreal frequencies.
26186 2018-01-05  Richard Sandiford  <richard.sandiford@linaro.org>
26188         * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
26189         punt for user-aligned variables.
26191 2018-01-05  Richard Sandiford  <richard.sandiford@linaro.org>
26193         * tree-chrec.c (chrec_contains_symbols): Return true for
26194         POLY_INT_CST.
26196 2018-01-05  Sudakshina Das  <sudi.das@arm.com>
26198         PR target/82439
26199         * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
26200         of (x|y) == x for BICS pattern.
26202 2018-01-05  Jakub Jelinek  <jakub@redhat.com>
26204         PR tree-optimization/83605
26205         * gimple-ssa-strength-reduction.c: Include tree-eh.h.
26206         (find_candidates_dom_walker::before_dom_children): Ignore stmts that
26207         can throw.
26209 2018-01-05  Sebastian Huber  <sebastian.huber@embedded-brains.de>
26211         * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
26212         * config/epiphany/rtems.h: New file.
26214 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
26215             Uros Bizjak  <ubizjak@gmail.com>
26217         PR target/83554
26218         * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
26219         QIreg_operand instead of register_operand predicate.
26220         * config/i386/i386.c (ix86_rop_should_change_byte_p,
26221         set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
26222         comments instead of -fmitigate[-_]rop.
26224 2018-01-04  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
26226         PR bootstrap/81926
26227         * cgraphunit.c (symbol_table::compile): Switch to text_section
26228         before calling assembly_start debug hook.
26229         * run-rtl-passes.c (run_rtl_passes): Likewise.
26230         Include output.h.
26232 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
26234         * tree-vrp.c (extract_range_from_binary_expr_1): Check
26235         range_int_cst_p rather than !symbolic_range_p before calling
26236         extract_range_from_multiplicative_op_1.
26238 2018-01-04  Jeff Law  <law@redhat.com>
26240         * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
26241         redundant test in assertion.
26243 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
26245         * doc/rtl.texi: Document machine_mode wrapper classes.
26247 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
26249         * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
26250         using tree_to_uhwi.
26252 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
26254         * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
26255         the VEC_PERM_EXPR fold to fail.
26257 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
26259         PR debug/83585
26260         * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
26261         to switched_sections.
26263 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
26265         PR target/83680
26266         * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
26267         test for d.testing.
26269 2018-01-04  Peter Bergner  <bergner@vnet.ibm.com>
26271         PR target/83387
26272         * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
26273         allow arguments in FP registers if TARGET_HARD_FLOAT is false.
26275 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
26277         PR debug/83666
26278         * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
26279         is BLKmode and bitpos not zero or mode change is needed.
26281 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
26283         PR target/83675
26284         * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
26285         TARGET_VIS2.
26287 2018-01-04  Uros Bizjak  <ubizjak@gmail.com>
26289         PR target/83628
26290         * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
26291         instead of MULT rtx.  Update all corresponding splitters.
26292         (*saddl_se): Ditto.
26293         (*ssub<modesuffix>): Ditto.
26294         (*ssubl_se): Ditto.
26295         (*cmp_sadd_di): Update split patterns.
26296         (*cmp_sadd_si): Ditto.
26297         (*cmp_sadd_sidi): Ditto.
26298         (*cmp_ssub_di): Ditto.
26299         (*cmp_ssub_si): Ditto.
26300         (*cmp_ssub_sidi): Ditto.
26301         * config/alpha/predicates.md (const23_operand): New predicate.
26302         * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
26303         Look for ASHIFT, not MULT inner operand.
26304         (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
26306 2018-01-04  Martin Liska  <mliska@suse.cz>
26308         PR gcov-profile/83669
26309         * gcov.c (output_intermediate_file): Add version to intermediate
26310         gcov file.
26311         * doc/gcov.texi: Document new field 'version' in intermediate
26312         file format. Fix location of '-k' option of gcov command.
26314 2018-01-04  Martin Liska  <mliska@suse.cz>
26316         PR ipa/82352
26317         * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
26319 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
26321         * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
26323 2018-01-03  Martin Sebor  <msebor@redhat.com>
26325         PR tree-optimization/83655
26326         * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
26327         checking calls with invalid arguments.
26329 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
26331         * tree-vect-stmts.c (vect_get_store_rhs): New function.
26332         (vectorizable_mask_load_store): Delete.
26333         (vectorizable_call): Return false for masked loads and stores.
26334         (vectorizable_store): Handle IFN_MASK_STORE.  Use vect_get_store_rhs
26335         instead of gimple_assign_rhs1.
26336         (vectorizable_load): Handle IFN_MASK_LOAD.
26337         (vect_transform_stmt): Don't set is_store for call_vec_info_type.
26339 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
26341         * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
26342         split out from..,
26343         (vectorizable_mask_load_store): ...here.
26344         (vectorizable_load): ...and here.
26346 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
26348         * tree-vect-stmts.c (vect_build_all_ones_mask)
26349         (vect_build_zero_merge_argument): New functions, split out from...
26350         (vectorizable_load): ...here.
26352 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
26354         * tree-vect-stmts.c (vect_check_store_rhs): New function,
26355         split out from...
26356         (vectorizable_mask_load_store): ...here.
26357         (vectorizable_store): ...and here.
26359 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
26361         * tree-vect-stmts.c (vect_check_load_store_mask): New function,
26362         split out from...
26363         (vectorizable_mask_load_store): ...here.
26365 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
26367         * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
26368         (vect_model_store_cost): Take a vec_load_store_type instead of a
26369         vect_def_type.
26370         * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
26371         (vect_model_store_cost): Take a vec_load_store_type instead of a
26372         vect_def_type.
26373         (vectorizable_mask_load_store): Update accordingly.
26374         (vectorizable_store): Likewise.
26375         * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
26377 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
26379         * tree-vect-loop.c (vect_transform_loop): Stub out scalar
26380         IFN_MASK_LOAD calls here rather than...
26381         * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
26383 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
26384             Alan Hayward  <alan.hayward@arm.com>
26385             David Sherwood  <david.sherwood@arm.com>
26387         * expmed.c (extract_bit_field_1): For vector extracts,
26388         fall back to extract_bit_field_as_subreg if vec_extract
26389         isn't available.
26391 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
26392             Alan Hayward  <alan.hayward@arm.com>
26393             David Sherwood  <david.sherwood@arm.com>
26395         * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
26396         they are variable or constant sized.
26397         (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
26398         slots for constant-sized data.
26400 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
26401             Alan Hayward  <alan.hayward@arm.com>
26402             David Sherwood  <david.sherwood@arm.com>
26404         * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
26405         handling COND_EXPRs with boolean comparisons, try to find a better
26406         basis for the mask type than the boolean itself.
26408 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
26410         * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
26411         is calculated and how it can be overridden.
26412         * genmodes.c (max_bitsize_mode_any_mode): New variable.
26413         (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
26414         if defined.
26415         (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
26416         if nonzero.
26418 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
26419             Alan Hayward  <alan.hayward@arm.com>
26420             David Sherwood  <david.sherwood@arm.com>
26422         * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
26423         Remove the mode argument.
26424         (aarch64_simd_valid_immediate): Remove the mode and inverse
26425         arguments.
26426         * config/aarch64/iterators.md (bitsize): New iterator.
26427         * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
26428         (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
26429         * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
26430         aarch64_simd_valid_immediate.
26431         * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
26432         (aarch64_reg_or_bic_imm): Likewise.
26433         * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
26434         with an insn_type enum and msl with a modifier_type enum.
26435         Replace element_width with a scalar_mode.  Change the shift
26436         to unsigned int.  Add constructors for scalar_float_mode and
26437         scalar_int_mode elements.
26438         (aarch64_vect_float_const_representable_p): Delete.
26439         (aarch64_can_const_movi_rtx_p)
26440         (aarch64_simd_scalar_immediate_valid_for_move)
26441         (aarch64_simd_make_constant): Update call to
26442         aarch64_simd_valid_immediate.
26443         (aarch64_advsimd_valid_immediate_hs): New function.
26444         (aarch64_advsimd_valid_immediate): Likewise.
26445         (aarch64_simd_valid_immediate): Remove mode and inverse
26446         arguments.  Rewrite to use the above.  Use const_vec_duplicate_p
26447         to detect duplicated constants and use aarch64_float_const_zero_rtx_p
26448         and aarch64_float_const_representable_p on the result.
26449         (aarch64_output_simd_mov_immediate): Remove mode argument.
26450         Update call to aarch64_simd_valid_immediate and use of
26451         simd_immediate_info.
26452         (aarch64_output_scalar_simd_mov_immediate): Update call
26453         accordingly.
26455 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
26456             Alan Hayward  <alan.hayward@arm.com>
26457             David Sherwood  <david.sherwood@arm.com>
26459         * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
26460         (mode_nunits): Likewise CONST_MODE_NUNITS.
26461         * machmode.def (ADJUST_NUNITS): Document.
26462         * genmodes.c (mode_data::need_nunits_adj): New field.
26463         (blank_mode): Update accordingly.
26464         (adj_nunits): New variable.
26465         (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
26466         parameter.
26467         (emit_mode_size_inline): Set need_bytesize_adj for all modes
26468         listed in adj_nunits.
26469         (emit_mode_nunits_inline): Set need_nunits_adj for all modes
26470         listed in adj_nunits.  Don't emit case statements for such modes.
26471         (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
26472         and CONST_MODE_PRECISION.  Make CONST_MODE_SIZE expand to
26473         nothing if adj_nunits is nonnull.
26474         (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
26475         (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
26476         (emit_mode_fbit): Update use of print_maybe_const_decl.
26477         (emit_move_size): Likewise.  Treat the array as non-const
26478         if adj_nunits.
26479         (emit_mode_adjustments): Handle adj_nunits.
26481 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
26483         * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
26484         * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
26485         (VECTOR_MODES): Use it.
26486         (make_vector_modes): Take the prefix as an argument.
26488 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
26489             Alan Hayward  <alan.hayward@arm.com>
26490             David Sherwood  <david.sherwood@arm.com>
26492         * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
26493         * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
26494         for MODE_VECTOR_BOOL.
26495         * machmode.def (VECTOR_BOOL_MODE): Document.
26496         * genmodes.c (VECTOR_BOOL_MODE): New macro.
26497         (make_vector_bool_mode): New function.
26498         (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
26499         MODE_VECTOR_BOOL.
26500         * lto-streamer-in.c (lto_input_mode_table): Likewise.
26501         * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
26502         Likewise.
26503         * stor-layout.c (int_mode_for_mode): Likewise.
26504         * tree.c (build_vector_type_for_mode): Likewise.
26505         * varasm.c (output_constant_pool_2): Likewise.
26506         * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
26507         CONSTM1_RTX (BImode) are the same thing.  Initialize const_tiny_rtx
26508         for MODE_VECTOR_BOOL.
26509         * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
26510         of mode class checks.
26511         * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
26512         instead of a list of mode class checks.
26513         (expand_vector_scalar_condition): Likewise.
26514         (type_for_widest_vector_mode): Handle BImode as an inner mode.
26516 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
26517             Alan Hayward  <alan.hayward@arm.com>
26518             David Sherwood  <david.sherwood@arm.com>
26520         * machmode.h (mode_size): Change from unsigned short to
26521         poly_uint16_pod.
26522         (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
26523         (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
26524         or if measurement_type is not polynomial.
26525         (fixed_size_mode::includes_p): Check for constant-sized modes.
26526         * genmodes.c (emit_mode_size_inline): Make mode_size_inline
26527         return a poly_uint16 rather than an unsigned short.
26528         (emit_mode_size): Change the type of mode_size from unsigned short
26529         to poly_uint16_pod.  Use ZERO_COEFFS for the initializer.
26530         (emit_mode_adjustments): Cope with polynomial vector sizes.
26531         * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
26532         for GET_MODE_SIZE.
26533         * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
26534         for GET_MODE_SIZE.
26535         * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
26536         * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
26537         * caller-save.c (setup_save_areas): Likewise.
26538         (replace_reg_with_saved_mem): Likewise.
26539         * calls.c (emit_library_call_value_1): Likewise.
26540         * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
26541         * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
26542         (gen_lowpart_for_combine): Likewise.
26543         * convert.c (convert_to_integer_1): Likewise.
26544         * cse.c (equiv_constant, cse_insn): Likewise.
26545         * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
26546         (cselib_subst_to_values): Likewise.
26547         * dce.c (word_dce_process_block): Likewise.
26548         * df-problems.c (df_word_lr_mark_ref): Likewise.
26549         * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
26550         * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
26551         (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
26552         (rtl_for_decl_location): Likewise.
26553         * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
26554         * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
26555         * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
26556         (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
26557         (expand_expr_real_1): Likewise.
26558         * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
26559         (pad_below): Likewise.
26560         * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
26561         * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
26562         * ira.c (get_subreg_tracking_sizes): Likewise.
26563         * ira-build.c (ira_create_allocno_objects): Likewise.
26564         * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
26565         (ira_sort_regnos_for_alter_reg): Likewise.
26566         * ira-costs.c (record_operand_costs): Likewise.
26567         * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
26568         (resolve_simple_move): Likewise.
26569         * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
26570         (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
26571         (lra_constraints): Likewise.
26572         (CONST_POOL_OK_P): Reject variable-sized modes.
26573         * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
26574         (add_pseudo_to_slot, lra_spill): Likewise.
26575         * omp-low.c (omp_clause_aligned_alignment): Likewise.
26576         * optabs-query.c (get_best_extraction_insn): Likewise.
26577         * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
26578         * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
26579         (expand_mult_highpart, valid_multiword_target_p): Likewise.
26580         * recog.c (offsettable_address_addr_space_p): Likewise.
26581         * regcprop.c (maybe_mode_change): Likewise.
26582         * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
26583         * regrename.c (build_def_use): Likewise.
26584         * regstat.c (dump_reg_info): Likewise.
26585         * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
26586         (find_reloads, find_reloads_subreg_address): Likewise.
26587         * reload1.c (eliminate_regs_1): Likewise.
26588         * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
26589         * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
26590         (simplify_binary_operation_1, simplify_subreg): Likewise.
26591         * targhooks.c (default_function_arg_padding): Likewise.
26592         (default_hard_regno_nregs, default_class_max_nregs): Likewise.
26593         * tree-cfg.c (verify_gimple_assign_binary): Likewise.
26594         (verify_gimple_assign_ternary): Likewise.
26595         * tree-inline.c (estimate_move_cost): Likewise.
26596         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
26597         * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
26598         (get_address_cost_ainc): Likewise.
26599         * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
26600         (vect_supportable_dr_alignment): Likewise.
26601         * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
26602         (vectorizable_reduction): Likewise.
26603         * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
26604         (vectorizable_operation, vectorizable_load): Likewise.
26605         * tree.c (build_same_sized_truth_vector_type): Likewise.
26606         * valtrack.c (cleanup_auto_inc_dec): Likewise.
26607         * var-tracking.c (emit_note_insn_var_location): Likewise.
26608         * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
26609         (ADDR_VEC_ALIGN): Likewise.
26611 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
26612             Alan Hayward  <alan.hayward@arm.com>
26613             David Sherwood  <david.sherwood@arm.com>
26615         * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
26616         unsigned short.
26617         (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
26618         or if measurement_type is polynomial.
26619         * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
26620         * combine.c (make_extraction): Likewise.
26621         * dse.c (find_shift_sequence): Likewise.
26622         * dwarf2out.c (mem_loc_descriptor): Likewise.
26623         * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
26624         (extract_bit_field, extract_low_bits): Likewise.
26625         * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
26626         (optimize_bitfield_assignment_op, expand_assignment): Likewise.
26627         (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
26628         * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
26629         * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
26630         * reload.c (find_reloads): Likewise.
26631         * reload1.c (alter_reg): Likewise.
26632         * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
26633         * targhooks.c (default_secondary_memory_needed_mode): Likewise.
26634         * tree-if-conv.c (predicate_mem_writes): Likewise.
26635         * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
26636         * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
26637         * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
26638         * valtrack.c (dead_debug_insert_temp): Likewise.
26639         * varasm.c (mergeable_constant_section): Likewise.
26640         * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
26642 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
26643             Alan Hayward  <alan.hayward@arm.com>
26644             David Sherwood  <david.sherwood@arm.com>
26646         * expr.c (expand_assignment): Cope with polynomial mode sizes
26647         when assigning to a CONCAT.
26649 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
26650             Alan Hayward  <alan.hayward@arm.com>
26651             David Sherwood  <david.sherwood@arm.com>
26653         * machmode.h (mode_precision): Change from unsigned short to
26654         poly_uint16_pod.
26655         (mode_to_precision): Return a poly_uint16 rather than an unsigned
26656         short.
26657         (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
26658         or if measurement_type is not polynomial.
26659         (HWI_COMPUTABLE_MODE_P): Turn into a function.  Optimize the case
26660         in which the mode is already known to be a scalar_int_mode.
26661         * genmodes.c (emit_mode_precision): Change the type of mode_precision
26662         from unsigned short to poly_uint16_pod.  Use ZERO_COEFFS for the
26663         initializer.
26664         * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
26665         for GET_MODE_PRECISION.
26666         * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
26667         for GET_MODE_PRECISION.
26668         * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
26669         as polynomial.
26670         (try_combine, find_split_point, combine_simplify_rtx): Likewise.
26671         (expand_field_assignment, make_extraction): Likewise.
26672         (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
26673         (get_last_value): Likewise.
26674         * convert.c (convert_to_integer_1): Likewise.
26675         * cse.c (cse_insn): Likewise.
26676         * expr.c (expand_expr_real_1): Likewise.
26677         * lra-constraints.c (simplify_operand_subreg): Likewise.
26678         * optabs-query.c (can_atomic_load_p): Likewise.
26679         * optabs.c (expand_atomic_load): Likewise.
26680         (expand_atomic_store): Likewise.
26681         * ree.c (combine_reaching_defs): Likewise.
26682         * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
26683         * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
26684         * tree.h (type_has_mode_precision_p): Likewise.
26685         * ubsan.c (instrument_si_overflow): Likewise.
26687 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
26688             Alan Hayward  <alan.hayward@arm.com>
26689             David Sherwood  <david.sherwood@arm.com>
26691         * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
26692         polynomial numbers of units.
26693         (SET_TYPE_VECTOR_SUBPARTS): Likewise.
26694         (valid_vector_subparts_p): New function.
26695         (build_vector_type): Remove temporary shim and take the number
26696         of units as a poly_uint64 rather than an int.
26697         (build_opaque_vector_type): Take the number of units as a
26698         poly_uint64 rather than an int.
26699         * tree.c (build_vector_from_ctor): Handle polynomial
26700         TYPE_VECTOR_SUBPARTS.
26701         (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
26702         (uniform_vector_p, vector_type_mode, build_vector): Likewise.
26703         (build_vector_from_val): If the number of units is variable,
26704         use build_vec_duplicate_cst for constant operands and
26705         VEC_DUPLICATE_EXPR otherwise.
26706         (make_vector_type): Remove temporary is_constant ().
26707         (build_vector_type, build_opaque_vector_type): Take the number of
26708         units as a poly_uint64 rather than an int.
26709         (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
26710         VECTOR_CST_NELTS.
26711         * cfgexpand.c (expand_debug_expr): Likewise.
26712         * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
26713         (store_constructor, expand_expr_real_1): Likewise.
26714         (const_scalar_mask_from_tree): Likewise.
26715         * fold-const-call.c (fold_const_reduction): Likewise.
26716         * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
26717         (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
26718         (native_encode_vector, vec_cst_ctor_to_array): Likewise.
26719         (fold_relational_const): Likewise.
26720         (native_interpret_vector): Likewise.  Change the size from an
26721         int to an unsigned int.
26722         * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
26723         TYPE_VECTOR_SUBPARTS.
26724         (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
26725         (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
26726         duplicating a non-constant operand into a variable-length vector.
26727         * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
26728         TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
26729         * ipa-icf.c (sem_variable::equals): Likewise.
26730         * match.pd: Likewise.
26731         * omp-simd-clone.c (simd_clone_subparts): Likewise.
26732         * print-tree.c (print_node): Likewise.
26733         * stor-layout.c (layout_type): Likewise.
26734         * targhooks.c (default_builtin_vectorization_cost): Likewise.
26735         * tree-cfg.c (verify_gimple_comparison): Likewise.
26736         (verify_gimple_assign_binary): Likewise.
26737         (verify_gimple_assign_ternary): Likewise.
26738         (verify_gimple_assign_single): Likewise.
26739         * tree-pretty-print.c (dump_generic_node): Likewise.
26740         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
26741         (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
26742         * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
26743         (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
26744         (vect_shift_permute_load_chain): Likewise.
26745         * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
26746         (expand_vector_condition, optimize_vector_constructor): Likewise.
26747         (lower_vec_perm, get_compute_type): Likewise.
26748         * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
26749         (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
26750         * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
26751         (vect_recog_mask_conversion_pattern): Likewise.
26752         * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
26753         (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
26754         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
26755         (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
26756         (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
26757         (vectorizable_shift, vectorizable_operation, vectorizable_store)
26758         (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
26759         (supportable_widening_operation): Likewise.
26760         (supportable_narrowing_operation): Likewise.
26761         * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
26762         Likewise.
26763         * varasm.c (output_constant): Likewise.
26765 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
26766             Alan Hayward  <alan.hayward@arm.com>
26767             David Sherwood  <david.sherwood@arm.com>
26769         * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
26770         so that both the length == 3 and length != 3 cases set up their
26771         own permute vectors.  Add comments explaining why we know the
26772         number of elements is constant.
26773         (vect_permute_load_chain): Likewise.
26775 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
26776             Alan Hayward  <alan.hayward@arm.com>
26777             David Sherwood  <david.sherwood@arm.com>
26779         * machmode.h (mode_nunits): Change from unsigned char to
26780         poly_uint16_pod.
26781         (ONLY_FIXED_SIZE_MODES): New macro.
26782         (pod_mode::measurement_type, scalar_int_mode::measurement_type)
26783         (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
26784         (complex_mode::measurement_type, fixed_size_mode::measurement_type):
26785         New typedefs.
26786         (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
26787         (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
26788         or if measurement_type is not polynomial.
26789         * genmodes.c (ZERO_COEFFS): New macro.
26790         (emit_mode_nunits_inline): Make mode_nunits_inline return a
26791         poly_uint16.
26792         (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
26793         Use ZERO_COEFFS when emitting initializers.
26794         * data-streamer.h (bp_pack_poly_value): New function.
26795         (bp_unpack_poly_value): Likewise.
26796         * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
26797         for GET_MODE_NUNITS.
26798         * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
26799         for GET_MODE_NUNITS.
26800         * tree.c (make_vector_type): Remove temporary shim and make
26801         the real function take the number of units as a poly_uint64
26802         rather than an int.
26803         (build_vector_type_for_mode): Handle polynomial nunits.
26804         * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
26805         * emit-rtl.c (const_vec_series_p_1): Likewise.
26806         (gen_rtx_CONST_VECTOR): Likewise.
26807         * fold-const.c (test_vec_duplicate_folding): Likewise.
26808         * genrecog.c (validate_pattern): Likewise.
26809         * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
26810         * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
26811         * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
26812         (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
26813         (expand_vec_cond_expr, expand_mult_highpart): Likewise.
26814         * rtlanal.c (subreg_get_info): Likewise.
26815         * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
26816         (vect_grouped_load_supported): Likewise.
26817         * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
26818         * tree-vect-loop.c (have_whole_vector_shift): Likewise.
26819         * simplify-rtx.c (simplify_unary_operation_1): Likewise.
26820         (simplify_const_unary_operation, simplify_binary_operation_1)
26821         (simplify_const_binary_operation, simplify_ternary_operation)
26822         (test_vector_ops_duplicate, test_vector_ops): Likewise.
26823         (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
26824         instead of CONST_VECTOR_NUNITS.
26825         * varasm.c (output_constant_pool_2): Likewise.
26826         * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
26827         explicit-encoded elements in the XVEC for variable-length vectors.
26829 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
26831         * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
26833 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
26834             Alan Hayward  <alan.hayward@arm.com>
26835             David Sherwood  <david.sherwood@arm.com>
26837         * coretypes.h (fixed_size_mode): Declare.
26838         (fixed_size_mode_pod): New typedef.
26839         * builtins.h (target_builtins::x_apply_args_mode)
26840         (target_builtins::x_apply_result_mode): Change type to
26841         fixed_size_mode_pod.
26842         * builtins.c (apply_args_size, apply_result_size, result_vector)
26843         (expand_builtin_apply_args_1, expand_builtin_apply)
26844         (expand_builtin_return): Update accordingly.
26846 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
26848         * cse.c (hash_rtx_cb): Hash only the encoded elements.
26849         * cselib.c (cselib_hash_rtx): Likewise.
26850         * expmed.c (make_tree): Build VECTOR_CSTs directly from the
26851         CONST_VECTOR encoding.
26853 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
26854             Jeff Law  <law@redhat.com>
26856         PR target/83641
26857         * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
26858         noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
26859         only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
26860         and add REG_CFA_ADJUST_CFA notes in that case to both insns.
26862         PR target/83641
26863         * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
26864         explicitly probe *sp in a noreturn function if there were any callee
26865         register saves or frame pointer is needed.
26867 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
26869         PR debug/83621
26870         * cfgexpand.c (expand_debug_expr): Return NULL if mode is
26871         BLKmode for ternary, binary or unary expressions.
26873         PR debug/83645
26874         * var-tracking.c (delete_vta_debug_insn): New inline function.
26875         (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
26876         insns from get_insns () to NULL instead of each bb separately.
26877         Use delete_vta_debug_insn.  No longer static.
26878         (vt_debug_insns_local, variable_tracking_main_1): Adjust
26879         delete_vta_debug_insns callers.
26880         * rtl.h (delete_vta_debug_insns): Declare.
26881         * final.c (rest_of_handle_final): Call delete_vta_debug_insns
26882         instead of variable_tracking_main.
26884 2018-01-03  Martin Sebor  <msebor@redhat.com>
26886         PR tree-optimization/83603
26887         * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
26888         arguments past the endof the argument list in functions declared
26889         without a prototype.
26890         * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
26891         Avoid checking when arguments are null.
26893 2018-01-03  Martin Sebor  <msebor@redhat.com>
26895         PR c/83559
26896         * doc/extend.texi (attribute const): Fix a typo.
26897         * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
26898         issuing -Wsuggest-attribute for void functions.
26900 2018-01-03  Martin Sebor  <msebor@redhat.com>
26902         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
26903         offset_int::from instead of wide_int::to_shwi.
26904         (maybe_diag_overlap): Remove assertion.
26905         Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
26906         * gimple-ssa-sprintf.c (format_directive): Same.
26907         (parse_directive): Same.
26908         (sprintf_dom_walker::compute_format_length): Same.
26909         (try_substitute_return_value): Same.
26911 2018-01-03  Jeff Law  <law@redhat.com>
26913         PR middle-end/83654
26914         * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
26915         non-constant residual for zero at runtime and avoid probing in
26916         that case.  Reorganize code for trailing problem to mirror handling
26917         of the residual.
26919 2018-01-03  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
26921         PR tree-optimization/83501
26922         * tree-ssa-strlen.c (get_string_cst): New.
26923         (handle_char_store): Call get_string_cst.
26925 2018-01-03  Martin Liska  <mliska@suse.cz>
26927         PR tree-optimization/83593
26928         * tree-ssa-strlen.c: Include tree-cfg.h.
26929         (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
26930         (strlen_dom_walker): Add new member variable m_cleanup_cfg.
26931         (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
26932         to false.
26933         (strlen_dom_walker::before_dom_children): Call
26934         gimple_purge_dead_eh_edges. Dump tranformation with details
26935         dump flags.
26936         (strlen_dom_walker::before_dom_children): Update call by adding
26937         new argument cleanup_eh.
26938         (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
26940 2018-01-03  Martin Liska  <mliska@suse.cz>
26942         PR ipa/83549
26943         * cif-code.def (VARIADIC_THUNK): New enum value.
26944         * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
26945         thunks.
26947 2018-01-03  Jan Beulich  <jbeulich@suse.com>
26949         * sse.md (mov<mode>_internal): Tighten condition for when to use
26950         vmovdqu<ssescalarsize> for TI and OI modes.
26952 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
26954         Update copyright years.
26956 2018-01-03  Martin Liska  <mliska@suse.cz>
26958         PR ipa/83594
26959         * ipa-visibility.c (function_and_variable_visibility): Skip
26960         functions with noipa attribure.
26962 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
26964         * gcc.c (process_command): Update copyright notice dates.
26965         * gcov-dump.c (print_version): Ditto.
26966         * gcov.c (print_version): Ditto.
26967         * gcov-tool.c (print_version): Ditto.
26968         * gengtype.c (create_file): Ditto.
26969         * doc/cpp.texi: Bump @copying's copyright year.
26970         * doc/cppinternals.texi: Ditto.
26971         * doc/gcc.texi: Ditto.
26972         * doc/gccint.texi: Ditto.
26973         * doc/gcov.texi: Ditto.
26974         * doc/install.texi: Ditto.
26975         * doc/invoke.texi: Ditto.
26977 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
26979         * vector-builder.h (vector_builder::m_full_nelts): Change from
26980         unsigned int to poly_uint64.
26981         (vector_builder::full_nelts): Update prototype accordingly.
26982         (vector_builder::new_vector): Likewise.
26983         (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
26984         (vector_builder::operator ==): Likewise.
26985         (vector_builder::finalize): Likewise.
26986         * int-vector-builder.h (int_vector_builder::int_vector_builder):
26987         Take the number of elements as a poly_uint64 rather than an
26988         unsigned int.
26989         * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
26990         from unsigned int to poly_uint64.
26991         (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
26992         (vec_perm_indices::new_vector): Likewise.
26993         (vec_perm_indices::length): Likewise.
26994         (vec_perm_indices::nelts_per_input): Likewise.
26995         (vec_perm_indices::input_nelts): Likewise.
26996         * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
26997         number of elements per input as a poly_uint64 rather than an
26998         unsigned int.  Use the original encoding for variable-length
26999         vectors, rather than clamping each individual element.
27000         For the second and subsequent elements in each pattern,
27001         clamp the step and base before clamping their sum.
27002         (vec_perm_indices::series_p): Handle polynomial element counts.
27003         (vec_perm_indices::all_in_range_p): Likewise.
27004         (vec_perm_indices_to_tree): Likewise.
27005         (vec_perm_indices_to_rtx): Likewise.
27006         * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
27007         * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
27008         (tree_vector_builder::new_binary_operation): Handle polynomial
27009         element counts.  Return false if we need to know the number
27010         of elements at compile time.
27011         * fold-const.c (fold_vec_perm): Punt if the number of elements
27012         isn't known at compile time.
27014 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27016         * vec-perm-indices.h (vec_perm_builder): Change element type
27017         from HOST_WIDE_INT to poly_int64.
27018         (vec_perm_indices::element_type): Update accordingly.
27019         (vec_perm_indices::clamp): Handle polynomial element_types.
27020         * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
27021         (vec_perm_indices::all_in_range_p): Likewise.
27022         (tree_to_vec_perm_builder): Check for poly_int64 trees rather
27023         than shwi trees.
27024         * vector-builder.h (vector_builder::stepped_sequence_p): Handle
27025         polynomial vec_perm_indices element types.
27026         * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
27027         * fold-const.c (fold_vec_perm): Likewise.
27028         * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
27029         * tree-vect-generic.c (lower_vec_perm): Likewise.
27030         * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
27031         * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
27032         element type to HOST_WIDE_INT.
27034 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27035             Alan Hayward  <alan.hayward@arm.com>
27036             David Sherwood  <david.sherwood@arm.com>
27038         * alias.c (addr_side_effect_eval): Take the size as a poly_int64
27039         rather than an int.  Use plus_constant.
27040         (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
27041         Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
27043 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27044             Alan Hayward  <alan.hayward@arm.com>
27045             David Sherwood  <david.sherwood@arm.com>
27047         * calls.c (emit_call_1, expand_call): Change struct_value_size from
27048         a HOST_WIDE_INT to a poly_int64.
27050 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27051             Alan Hayward  <alan.hayward@arm.com>
27052             David Sherwood  <david.sherwood@arm.com>
27054         * calls.c (load_register_parameters): Cope with polynomial
27055         mode sizes.  Require a constant size for BLKmode parameters
27056         that aren't described by a PARALLEL.  If BLOCK_REG_PADDING
27057         forces a parameter to be padded at the lsb end in order to
27058         fill a complete number of words, require the parameter size
27059         to be ordered wrt UNITS_PER_WORD.
27061 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27062             Alan Hayward  <alan.hayward@arm.com>
27063             David Sherwood  <david.sherwood@arm.com>
27065         * reload1.c (spill_stack_slot_width): Change element type
27066         from unsigned int to poly_uint64_pod.
27067         (alter_reg): Treat mode sizes as polynomial.
27069 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27070             Alan Hayward  <alan.hayward@arm.com>
27071             David Sherwood  <david.sherwood@arm.com>
27073         * reload.c (complex_word_subreg_p): New function.
27074         (reload_inner_reg_of_subreg, push_reload): Use it.
27076 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27077             Alan Hayward  <alan.hayward@arm.com>
27078             David Sherwood  <david.sherwood@arm.com>
27080         * lra-constraints.c (process_alt_operands): Reject matched
27081         operands whose sizes aren't ordered.
27082         (match_reload): Refer to this check here.
27084 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27085             Alan Hayward  <alan.hayward@arm.com>
27086             David Sherwood  <david.sherwood@arm.com>
27088         * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
27089         that the mode size is in the set {1, 2, 4, 8, 16}.
27091 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27092             Alan Hayward  <alan.hayward@arm.com>
27093             David Sherwood  <david.sherwood@arm.com>
27095         * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
27096         Use plus_constant instead of gen_rtx_PLUS.
27098 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27099             Alan Hayward  <alan.hayward@arm.com>
27100             David Sherwood  <david.sherwood@arm.com>
27102         * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
27103         * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
27104         * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
27105         * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
27106         * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
27107         * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
27108         * config/i386/i386-protos.h (ix86_push_rounding): Declare.
27109         * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
27110         * config/i386/i386.c (ix86_push_rounding): ...this new function.
27111         * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
27112         a poly_int64.
27113         * config/m32c/m32c.c (m32c_push_rounding): Likewise.
27114         * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
27115         * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
27116         * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
27117         * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
27118         * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
27119         * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
27120         * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
27121         * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
27122         * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
27123         function.
27124         * expr.c (emit_move_resolve_push): Treat the input and result
27125         of PUSH_ROUNDING as a poly_int64.
27126         (emit_move_complex_push, emit_single_push_insn_1): Likewise.
27127         (emit_push_insn): Likewise.
27128         * lra-eliminations.c (mark_not_eliminable): Likewise.
27129         * recog.c (push_operand): Likewise.
27130         * reload1.c (elimination_effects): Likewise.
27131         * rtlanal.c (nonzero_bits1): Likewise.
27132         * calls.c (store_one_arg): Likewise.  Require the padding to be
27133         known at compile time.
27135 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27136             Alan Hayward  <alan.hayward@arm.com>
27137             David Sherwood  <david.sherwood@arm.com>
27139         * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
27140         Use plus_constant instead of gen_rtx_PLUS.
27142 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27143             Alan Hayward  <alan.hayward@arm.com>
27144             David Sherwood  <david.sherwood@arm.com>
27146         * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
27147         rather than an int.
27149 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27150             Alan Hayward  <alan.hayward@arm.com>
27151             David Sherwood  <david.sherwood@arm.com>
27153         * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
27154         instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
27155         via stack temporaries.  Treat the mode size as polynomial too.
27157 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27158             Alan Hayward  <alan.hayward@arm.com>
27159             David Sherwood  <david.sherwood@arm.com>
27161         * expr.c (expand_expr_real_2): When handling conversions involving
27162         unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
27163         multiplying int_size_in_bytes by BITS_PER_UNIT.  Treat GET_MODE_BISIZE
27164         as a poly_uint64 too.
27166 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27167             Alan Hayward  <alan.hayward@arm.com>
27168             David Sherwood  <david.sherwood@arm.com>
27170         * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
27172 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27173             Alan Hayward  <alan.hayward@arm.com>
27174             David Sherwood  <david.sherwood@arm.com>
27176         * combine.c (can_change_dest_mode): Handle polynomial
27177         REGMODE_NATURAL_SIZE.
27178         * expmed.c (store_bit_field_1): Likewise.
27179         * expr.c (store_constructor): Likewise.
27180         * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
27181         and polynomial REGMODE_NATURAL_SIZE.
27182         (gen_lowpart_common): Likewise.
27183         * reginfo.c (record_subregs_of_mode): Likewise.
27184         * rtlanal.c (read_modify_subreg_p): Likewise.
27186 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27187             Alan Hayward  <alan.hayward@arm.com>
27188             David Sherwood  <david.sherwood@arm.com>
27190         * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
27191         numbers of elements.
27193 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27194             Alan Hayward  <alan.hayward@arm.com>
27195             David Sherwood  <david.sherwood@arm.com>
27197         * match.pd: Cope with polynomial numbers of vector elements.
27199 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27200             Alan Hayward  <alan.hayward@arm.com>
27201             David Sherwood  <david.sherwood@arm.com>
27203         * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
27204         in a POINTER_PLUS_EXPR.
27206 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27207             Alan Hayward  <alan.hayward@arm.com>
27208             David Sherwood  <david.sherwood@arm.com>
27210         * omp-simd-clone.c (simd_clone_subparts): New function.
27211         (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
27212         (ipa_simd_modify_function_body): Likewise.
27214 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27215             Alan Hayward  <alan.hayward@arm.com>
27216             David Sherwood  <david.sherwood@arm.com>
27218         * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
27219         (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
27220         (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
27221         (expand_vector_condition, vector_element): Likewise.
27222         (subparts_gt): New function.
27223         (get_compute_type): Use subparts_gt.
27224         (count_type_subparts): Delete.
27225         (expand_vector_operations_1): Use subparts_gt instead of
27226         count_type_subparts.
27228 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27229             Alan Hayward  <alan.hayward@arm.com>
27230             David Sherwood  <david.sherwood@arm.com>
27232         * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
27233         (vect_compile_time_alias): ...this new function.  Do the calculation
27234         on poly_ints rather than trees.
27235         (vect_prune_runtime_alias_test_list): Update call accordingly.
27237 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27238             Alan Hayward  <alan.hayward@arm.com>
27239             David Sherwood  <david.sherwood@arm.com>
27241         * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
27242         numbers of units.
27243         (vect_schedule_slp_instance): Likewise.
27245 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27246             Alan Hayward  <alan.hayward@arm.com>
27247             David Sherwood  <david.sherwood@arm.com>
27249         * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
27250         constant and extern definitions for variable-length vectors.
27251         (vect_get_constant_vectors): Note that the number of units
27252         is known to be constant.
27254 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27255             Alan Hayward  <alan.hayward@arm.com>
27256             David Sherwood  <david.sherwood@arm.com>
27258         * tree-vect-stmts.c (vectorizable_conversion): Treat the number
27259         of units as polynomial.  Choose between WIDE and NARROW based
27260         on multiple_p.
27262 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27263             Alan Hayward  <alan.hayward@arm.com>
27264             David Sherwood  <david.sherwood@arm.com>
27266         * tree-vect-stmts.c (simd_clone_subparts): New function.
27267         (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
27269 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27270             Alan Hayward  <alan.hayward@arm.com>
27271             David Sherwood  <david.sherwood@arm.com>
27273         * tree-vect-stmts.c (vectorizable_call): Treat the number of
27274         vectors as polynomial.  Use build_index_vector for
27275         IFN_GOMP_SIMD_LANE.
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         * tree-vect-stmts.c (get_load_store_type): Treat the number of
27282         units as polynomial.  Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
27283         for variable-length vectors.
27284         (vectorizable_mask_load_store): Treat the number of units as
27285         polynomial, asserting that it is constant if the condition has
27286         already been enforced.
27287         (vectorizable_store, vectorizable_load): Likewise.
27289 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27290             Alan Hayward  <alan.hayward@arm.com>
27291             David Sherwood  <david.sherwood@arm.com>
27293         * tree-vect-loop.c (vectorizable_live_operation): Treat the number
27294         of units as polynomial.  Punt if we can't tell at compile time
27295         which vector contains the final result.
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         * tree-vect-loop.c (vectorizable_induction): Treat the number
27302         of units as polynomial.  Punt on SLP inductions.  Use an integer
27303         VEC_SERIES_EXPR for variable-length integer reductions.  Use a
27304         cast of such a series for variable-length floating-point
27305         reductions.
27307 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27308             Alan Hayward  <alan.hayward@arm.com>
27309             David Sherwood  <david.sherwood@arm.com>
27311         * tree.h (build_index_vector): Declare.
27312         * tree.c (build_index_vector): New function.
27313         * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
27314         of units as polynomial, forcibly converting it to a constant if
27315         vectorizable_reduction has already enforced the condition.
27316         (vect_create_epilog_for_reduction): Likewise.  Use build_index_vector
27317         to create a {1,2,3,...} vector.
27318         (vectorizable_reduction): Treat the number of units as polynomial.
27319         Choose vectype_in based on the largest scalar element size rather
27320         than the smallest number of units.  Enforce the restrictions
27321         relied on above.
27323 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27324             Alan Hayward  <alan.hayward@arm.com>
27325             David Sherwood  <david.sherwood@arm.com>
27327         * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
27328         number of units as polynomial.
27330 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27331             Alan Hayward  <alan.hayward@arm.com>
27332             David Sherwood  <david.sherwood@arm.com>
27334         * target.h (vector_sizes, auto_vector_sizes): New typedefs.
27335         * target.def (autovectorize_vector_sizes): Return the vector sizes
27336         by pointer, using vector_sizes rather than a bitmask.
27337         * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
27338         * targhooks.c (default_autovectorize_vector_sizes): Likewise.
27339         * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
27340         Likewise.
27341         * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
27342         * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
27343         * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
27344         * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
27345         * omp-general.c (omp_max_vf): Likewise.
27346         * omp-low.c (omp_clause_aligned_alignment): Likewise.
27347         * optabs-query.c (can_vec_mask_load_store_p): Likewise.
27348         * tree-vect-loop.c (vect_analyze_loop): Likewise.
27349         * tree-vect-slp.c (vect_slp_bb): Likewise.
27350         * doc/tm.texi: Regenerate.
27351         * tree-vectorizer.h (current_vector_size): Change from an unsigned int
27352         to a poly_uint64.
27353         * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
27354         the vector size as a poly_uint64 rather than an unsigned int.
27355         (current_vector_size): Change from an unsigned int to a poly_uint64.
27356         (get_vectype_for_scalar_type): Update accordingly.
27357         * tree.h (build_truth_vector_type): Take the size and number of
27358         units as a poly_uint64 rather than an unsigned int.
27359         (build_vector_type): Add a temporary overload that takes
27360         the number of units as a poly_uint64 rather than an unsigned int.
27361         * tree.c (make_vector_type): Likewise.
27362         (build_truth_vector_type): Take the number of units as a poly_uint64
27363         rather than an unsigned int.
27365 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27366             Alan Hayward  <alan.hayward@arm.com>
27367             David Sherwood  <david.sherwood@arm.com>
27369         * target.def (get_mask_mode): Take the number of units and length
27370         as poly_uint64s rather than unsigned ints.
27371         * targhooks.h (default_get_mask_mode): Update accordingly.
27372         * targhooks.c (default_get_mask_mode): Likewise.
27373         * config/i386/i386.c (ix86_get_mask_mode): Likewise.
27374         * doc/tm.texi: Regenerate.
27376 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27377             Alan Hayward  <alan.hayward@arm.com>
27378             David Sherwood  <david.sherwood@arm.com>
27380         * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
27381         * omp-general.c (omp_max_vf): Likewise.
27382         * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
27383         (expand_omp_simd): Handle polynomial safelen.
27384         * omp-low.c (omplow_simd_context): Add a default constructor.
27385         (omplow_simd_context::max_vf): Change from int to poly_uint64.
27386         (lower_rec_simd_input_clauses): Update accordingly.
27387         (lower_rec_input_clauses): Likewise.
27389 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27390             Alan Hayward  <alan.hayward@arm.com>
27391             David Sherwood  <david.sherwood@arm.com>
27393         * tree-vectorizer.h (vect_nunits_for_cost): New function.
27394         * tree-vect-loop.c (vect_model_reduction_cost): Use it.
27395         * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
27396         (vect_analyze_slp_cost): Likewise.
27397         * tree-vect-stmts.c (vect_model_store_cost): Likewise.
27398         (vect_model_load_cost): Likewise.
27400 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27401             Alan Hayward  <alan.hayward@arm.com>
27402             David Sherwood  <david.sherwood@arm.com>
27404         * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
27405         (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
27406         from an unsigned int * to a poly_uint64_pod *.
27407         (calculate_unrolling_factor): New function.
27408         (vect_analyze_slp_instance): Use it.  Track polynomial max_nunits.
27410 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27411             Alan Hayward  <alan.hayward@arm.com>
27412             David Sherwood  <david.sherwood@arm.com>
27414         * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
27415         from an unsigned int to a poly_uint64.
27416         (_loop_vec_info::slp_unrolling_factor): Likewise.
27417         (_loop_vec_info::vectorization_factor): Change from an int
27418         to a poly_uint64.
27419         (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
27420         (vect_get_num_vectors): New function.
27421         (vect_update_max_nunits, vect_vf_for_cost): Likewise.
27422         (vect_get_num_copies): Use vect_get_num_vectors.
27423         (vect_analyze_data_ref_dependences): Change max_vf from an int *
27424         to an unsigned int *.
27425         (vect_analyze_data_refs): Change min_vf from an int * to a
27426         poly_uint64 *.
27427         (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
27428         than an unsigned HOST_WIDE_INT.
27429         * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
27430         (vect_analyze_data_ref_dependence): Change max_vf from an int *
27431         to an unsigned int *.
27432         (vect_analyze_data_ref_dependences): Likewise.
27433         (vect_compute_data_ref_alignment): Handle polynomial vf.
27434         (vect_enhance_data_refs_alignment): Likewise.
27435         (vect_prune_runtime_alias_test_list): Likewise.
27436         (vect_shift_permute_load_chain): Likewise.
27437         (vect_supportable_dr_alignment): Likewise.
27438         (dependence_distance_ge_vf): Take the vectorization factor as a
27439         poly_uint64 rather than an unsigned HOST_WIDE_INT.
27440         (vect_analyze_data_refs): Change min_vf from an int * to a
27441         poly_uint64 *.
27442         * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
27443         vfm1 as a poly_uint64 rather than an int.  Make the same change
27444         for the returned bound_scalar.
27445         (vect_gen_vector_loop_niters): Handle polynomial vf.
27446         (vect_do_peeling): Likewise.  Update call to
27447         vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
27448         (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
27449         be constant.
27450         * tree-vect-loop.c (vect_determine_vectorization_factor)
27451         (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
27452         (vect_get_known_peeling_cost): Likewise.
27453         (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
27454         (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
27455         (vect_transform_loop): Likewise.  Use the lowest possible VF when
27456         updating the upper bounds of the loop.
27457         (vect_min_worthwhile_factor): Make static.  Return an unsigned int
27458         rather than an int.
27459         * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
27460         polynomial unroll factors.
27461         (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
27462         (vect_make_slp_decision): Likewise.
27463         (vect_supported_load_permutation_p): Likewise, and polynomial
27464         vf too.
27465         (vect_analyze_slp_cost): Handle polynomial vf.
27466         (vect_slp_analyze_node_operations): Likewise.
27467         (vect_slp_analyze_bb_1): Likewise.
27468         (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
27469         than an unsigned HOST_WIDE_INT.
27470         * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
27471         (vectorizable_load): Handle polynomial vf.
27472         * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
27473         a poly_uint64.
27474         (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
27476 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27477             Alan Hayward  <alan.hayward@arm.com>
27478             David Sherwood  <david.sherwood@arm.com>
27480         * match.pd: Handle bit operations involving three constants
27481         and try to fold one pair.
27483 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27485         * tree-vect-loop-manip.c: Include gimple-fold.h.
27486         (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
27487         niters_maybe_zero parameters.  Handle other cases besides a step of 1.
27488         (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
27489         Add a path that uses a step of VF instead of 1, but disable it
27490         for now.
27491         (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
27492         and niters_no_overflow parameters.  Update calls to
27493         slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
27494         Create a new SSA name if the latter choses to use a ste other
27495         than zero, and return it via niters_vector_mult_vf_var.
27496         * tree-vect-loop.c (vect_transform_loop): Update calls to
27497         vect_do_peeling, vect_gen_vector_loop_niters and
27498         slpeel_make_loop_iterate_ntimes.
27499         * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
27500         (vect_gen_vector_loop_niters): Update declarations after above changes.
27502 2018-01-02  Michael Meissner  <meissner@linux.vnet.ibm.com>
27504         * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
27505         128-bit round to integer instructions.
27506         (ceil<mode>2): Likewise.
27507         (btrunc<mode>2): Likewise.
27508         (round<mode>2): Likewise.
27510 2018-01-02  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
27512         * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
27513         unaligned VSX load/store on P8/P9.
27514         (expand_block_clear): Allow the use of unaligned VSX
27515         load/store on P8/P9.
27517 2018-01-02  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
27519         * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
27520         New function.
27521         (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
27522         swap associated with both a load and a store.
27524 2018-01-02  Andrew Waterman  <andrew@sifive.com>
27526         * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
27527         * config/riscv/riscv.md (clear_cache): Use it.
27529 2018-01-02  Artyom Skrobov  <tyomitch@gmail.com>
27531         * web.c: Remove out-of-date comment.
27533 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
27535         * expr.c (fixup_args_size_notes): Check that any existing
27536         REG_ARGS_SIZE notes are correct, and don't try to re-add them.
27537         (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
27538         (emit_single_push_insn): ...here.
27540 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
27542         * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
27543         (const_vector_encoded_nelts): New function.
27544         (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
27545         (const_vector_int_elt, const_vector_elt): Declare.
27546         * emit-rtl.c (const_vector_int_elt_1): New function.
27547         (const_vector_elt): Likewise.
27548         * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
27549         of CONST_VECTOR_ELT.
27551 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
27553         * expr.c: Include rtx-vector-builder.h.
27554         (const_vector_mask_from_tree): Use rtx_vector_builder and operate
27555         directly on the tree encoding.
27556         (const_vector_from_tree): Likewise.
27557         * optabs.c: Include rtx-vector-builder.h.
27558         (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
27559         sequence of "u" values.
27560         * vec-perm-indices.c: Include rtx-vector-builder.h.
27561         (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
27562         directly on the vec_perm_indices encoding.
27564 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
27566         * doc/rtl.texi (const_vector): Describe new encoding scheme.
27567         * Makefile.in (OBJS): Add rtx-vector-builder.o.
27568         * rtx-vector-builder.h: New file.
27569         * rtx-vector-builder.c: Likewise.
27570         * rtl.h (rtx_def::u2): Add a const_vector field.
27571         (CONST_VECTOR_NPATTERNS): New macro.
27572         (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
27573         (CONST_VECTOR_DUPLICATE_P): Likewise.
27574         (CONST_VECTOR_STEPPED_P): Likewise.
27575         (CONST_VECTOR_ENCODED_ELT): Likewise.
27576         (const_vec_duplicate_p): Check for a duplicated vector encoding.
27577         (unwrap_const_vec_duplicate): Likewise.
27578         (const_vec_series_p): Check for a non-duplicated vector encoding.
27579         Say that the function only returns true for integer vectors.
27580         * emit-rtl.c: Include rtx-vector-builder.h.
27581         (gen_const_vec_duplicate_1): Delete.
27582         (gen_const_vector): Call gen_const_vec_duplicate instead of
27583         gen_const_vec_duplicate_1.
27584         (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
27585         (gen_const_vec_duplicate): Use rtx_vector_builder.
27586         (gen_const_vec_series): Likewise.
27587         (gen_rtx_CONST_VECTOR): Likewise.
27588         * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
27589         (swap_const_vector_halves): Take an rtx pointer rather than rtx.
27590         Build a new vector rather than modifying a CONST_VECTOR in-place.
27591         (handle_special_swappables): Update call accordingly.
27592         * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
27593         (swap_const_vector_halves): Take an rtx pointer rather than rtx.
27594         Build a new vector rather than modifying a CONST_VECTOR in-place.
27595         (handle_special_swappables): Update call accordingly.
27597 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
27599         * simplify-rtx.c (simplify_const_binary_operation): Use
27600         CONST_VECTOR_ELT instead of XVECEXP.
27602 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
27604         * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
27605         the selector elements to be different from the data elements
27606         if the selector is a VECTOR_CST.
27607         * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
27608         ssizetype for the selector.
27610 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
27612         * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
27613         before testing each element individually.
27614         * tree-vect-generic.c (lower_vec_perm): Likewise.
27616 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
27618         * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
27619         * selftest-run-tests.c (selftest::run_tests): Call it.
27620         * vector-builder.h (vector_builder::operator ==): New function.
27621         (vector_builder::operator !=): Likewise.
27622         * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
27623         (vec_perm_indices::all_from_input_p): New function.
27624         * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
27625         (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
27626         * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
27627         instead of reading the VECTOR_CST directly.  Detect whether both
27628         vector inputs are the same before constructing the vec_perm_indices,
27629         and update the number of inputs argument accordingly.  Use the
27630         utility functions added above.  Only construct sel2 if we need to.
27632 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
27634         * optabs.c (expand_vec_perm_var): Use an explicit encoding for
27635         the broadcast of the low byte.
27636         (expand_mult_highpart): Use an explicit encoding for the permutes.
27637         * optabs-query.c (can_mult_highpart_p): Likewise.
27638         * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
27639         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
27640         (vectorizable_bswap): Likewise.
27641         * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
27642         explicit encoding for the power-of-2 permutes.
27643         (vect_permute_store_chain): Likewise.
27644         (vect_grouped_load_supported): Likewise.
27645         (vect_permute_load_chain): Likewise.
27647 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
27649         * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
27650         * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
27651         * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
27652         * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
27653         * tree-vect-stmts.c (vectorizable_bswap): Likewise.
27654         (vect_gen_perm_mask_any): Likewise.
27656 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
27658         * int-vector-builder.h: New file.
27659         * vec-perm-indices.h: Include int-vector-builder.h.
27660         (vec_perm_indices): Redefine as an int_vector_builder.
27661         (auto_vec_perm_indices): Delete.
27662         (vec_perm_builder): Redefine as a stand-alone class.
27663         (vec_perm_indices::vec_perm_indices): New function.
27664         (vec_perm_indices::clamp): Likewise.
27665         * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
27666         (vec_perm_indices::new_vector): New function.
27667         (vec_perm_indices::new_expanded_vector): Update for new
27668         vec_perm_indices class.
27669         (vec_perm_indices::rotate_inputs): New function.
27670         (vec_perm_indices::all_in_range_p): Operate directly on the
27671         encoded form, without computing elided elements.
27672         (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
27673         encoding.  Update for new vec_perm_indices class.
27674         * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
27675         the given vec_perm_builder.
27676         (expand_vec_perm_var): Update vec_perm_builder constructor.
27677         (expand_mult_highpart): Use vec_perm_builder instead of
27678         auto_vec_perm_indices.
27679         * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
27680         vec_perm_indices instead of auto_vec_perm_indices.  Use a single
27681         or double series encoding as appropriate.
27682         * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
27683         vec_perm_indices instead of auto_vec_perm_indices.
27684         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
27685         * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
27686         (vect_permute_store_chain): Likewise.
27687         (vect_grouped_load_supported): Likewise.
27688         (vect_permute_load_chain): Likewise.
27689         (vect_shift_permute_load_chain): Likewise.
27690         * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
27691         (vect_transform_slp_perm_load): Likewise.
27692         (vect_schedule_slp_instance): Likewise.
27693         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
27694         (vectorizable_mask_load_store): Likewise.
27695         (vectorizable_bswap): Likewise.
27696         (vectorizable_store): Likewise.
27697         (vectorizable_load): Likewise.
27698         * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
27699         vec_perm_indices instead of auto_vec_perm_indices.  Use
27700         tree_to_vec_perm_builder to read the vector from a tree.
27701         * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
27702         vec_perm_builder instead of a vec_perm_indices.
27703         (have_whole_vector_shift): Use vec_perm_builder and
27704         vec_perm_indices instead of auto_vec_perm_indices.  Leave the
27705         truncation to calc_vec_perm_mask_for_shift.
27706         (vect_create_epilog_for_reduction): Likewise.
27707         * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
27708         from auto_vec_perm_indices to vec_perm_indices.
27709         (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
27710         instead of changing individual elements.
27711         (aarch64_vectorize_vec_perm_const): Use new_vector to install
27712         the vector in d.perm.
27713         * config/arm/arm.c (expand_vec_perm_d::perm): Change
27714         from auto_vec_perm_indices to vec_perm_indices.
27715         (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
27716         instead of changing individual elements.
27717         (arm_vectorize_vec_perm_const): Use new_vector to install
27718         the vector in d.perm.
27719         * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
27720         Update vec_perm_builder constructor.
27721         (rs6000_expand_interleave): Likewise.
27722         * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
27723         (rs6000_expand_interleave): Likewise.
27725 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
27727         * optabs-query.c (can_vec_perm_var_p): Check whether lowering
27728         to qimode could truncate the indices.
27729         * optabs.c (expand_vec_perm_var): Likewise.
27731 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
27733         * Makefile.in (OBJS): Add vec-perm-indices.o.
27734         * vec-perm-indices.h: New file.
27735         * vec-perm-indices.c: Likewise.
27736         * target.h (vec_perm_indices): Replace with a forward class
27737         declaration.
27738         (auto_vec_perm_indices): Move to vec-perm-indices.h.
27739         * optabs.h: Include vec-perm-indices.h.
27740         (expand_vec_perm): Delete.
27741         (selector_fits_mode_p, expand_vec_perm_var): Declare.
27742         (expand_vec_perm_const): Declare.
27743         * target.def (vec_perm_const_ok): Replace with...
27744         (vec_perm_const): ...this new hook.
27745         * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
27746         (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
27747         * doc/tm.texi: Regenerate.
27748         * optabs.def (vec_perm_const): Delete.
27749         * doc/md.texi (vec_perm_const): Likewise.
27750         (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
27751         * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
27752         expand_vec_perm for constant permutation vectors.  Assert that
27753         the mode of variable permutation vectors is the integer equivalent
27754         of the mode that is being permuted.
27755         * optabs-query.h (selector_fits_mode_p): Declare.
27756         * optabs-query.c: Include vec-perm-indices.h.
27757         (selector_fits_mode_p): New function.
27758         (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
27759         is defined, instead of checking whether the vec_perm_const_optab
27760         exists.  Use targetm.vectorize.vec_perm_const instead of
27761         targetm.vectorize.vec_perm_const_ok.  Check whether the indices
27762         fit in the vector mode before using a variable permute.
27763         * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
27764         vec_perm_indices instead of an rtx.
27765         (expand_vec_perm): Replace with...
27766         (expand_vec_perm_const): ...this new function.  Take the selector
27767         as a vec_perm_indices rather than an rtx.  Also take the mode of
27768         the selector.  Update call to shift_amt_for_vec_perm_mask.
27769         Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
27770         Use vec_perm_indices::new_expanded_vector to expand the original
27771         selector into bytes.  Check whether the indices fit in the vector
27772         mode before using a variable permute.
27773         (expand_vec_perm_var): Make global.
27774         (expand_mult_highpart): Use expand_vec_perm_const.
27775         * fold-const.c: Includes vec-perm-indices.h.
27776         * tree-ssa-forwprop.c: Likewise.
27777         * tree-vect-data-refs.c: Likewise.
27778         * tree-vect-generic.c: Likewise.
27779         * tree-vect-loop.c: Likewise.
27780         * tree-vect-slp.c: Likewise.
27781         * tree-vect-stmts.c: Likewise.
27782         * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
27783         Delete.
27784         * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
27785         * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
27786         (aarch64_vectorize_vec_perm_const_ok): Fuse into...
27787         (aarch64_vectorize_vec_perm_const): ...this new function.
27788         (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
27789         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
27790         * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
27791         * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
27792         * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
27793         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
27794         (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
27795         into...
27796         (arm_vectorize_vec_perm_const): ...this new function.  Explicitly
27797         check for NEON modes.
27798         * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
27799         * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
27800         * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
27801         (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
27802         into...
27803         (ix86_vectorize_vec_perm_const): ...this new function.  Incorporate
27804         the old VEC_PERM_CONST conditions.
27805         * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
27806         * config/ia64/vect.md (vec_perm_const<mode>): Delete.
27807         * config/ia64/ia64.c (ia64_expand_vec_perm_const)
27808         (ia64_vectorize_vec_perm_const_ok): Merge into...
27809         (ia64_vectorize_vec_perm_const): ...this new function.
27810         * config/mips/loongson.md (vec_perm_const<mode>): Delete.
27811         * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
27812         * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
27813         * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
27814         * config/mips/mips.c (mips_expand_vec_perm_const)
27815         (mips_vectorize_vec_perm_const_ok): Merge into...
27816         (mips_vectorize_vec_perm_const): ...this new function.
27817         * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
27818         * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
27819         * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
27820         * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
27821         * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
27822         (rs6000_expand_vec_perm_const): Delete.
27823         * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
27824         Delete.
27825         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
27826         (altivec_expand_vec_perm_const_le): Take each operand individually.
27827         Operate on constant selectors rather than rtxes.
27828         (altivec_expand_vec_perm_const): Likewise.  Update call to
27829         altivec_expand_vec_perm_const_le.
27830         (rs6000_expand_vec_perm_const): Delete.
27831         (rs6000_vectorize_vec_perm_const_ok): Delete.
27832         (rs6000_vectorize_vec_perm_const): New function.
27833         (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
27834         an element count and rtx array.
27835         (rs6000_expand_extract_even): Update call accordingly.
27836         (rs6000_expand_interleave): Likewise.
27837         * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
27838         * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
27839         * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
27840         * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
27841         (rs6000_expand_vec_perm_const): Delete.
27842         * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
27843         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
27844         (altivec_expand_vec_perm_const_le): Take each operand individually.
27845         Operate on constant selectors rather than rtxes.
27846         (altivec_expand_vec_perm_const): Likewise.  Update call to
27847         altivec_expand_vec_perm_const_le.
27848         (rs6000_expand_vec_perm_const): Delete.
27849         (rs6000_vectorize_vec_perm_const_ok): Delete.
27850         (rs6000_vectorize_vec_perm_const): New function.  Remove stray
27851         reference to the SPE evmerge intructions.
27852         (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
27853         an element count and rtx array.
27854         (rs6000_expand_extract_even): Update call accordingly.
27855         (rs6000_expand_interleave): Likewise.
27856         * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
27857         * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
27858         new function.
27859         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
27861 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
27863         * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
27864         vector mode and that that mode matches the mode of the data
27865         being permuted.
27866         (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
27867         out into expand_vec_perm_var.  Do all CONST_VECTOR handling here,
27868         directly using expand_vec_perm_1 when forcing selectors into
27869         registers.
27870         (expand_vec_perm_var): New function, split out from expand_vec_perm.
27872 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
27874         * optabs-query.h (can_vec_perm_p): Delete.
27875         (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
27876         * optabs-query.c (can_vec_perm_p): Split into...
27877         (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
27878         (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
27879         particular selector is valid.
27880         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
27881         * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
27882         (vect_grouped_load_supported): Likewise.
27883         (vect_shift_permute_load_chain): Likewise.
27884         * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
27885         (vect_transform_slp_perm_load): Likewise.
27886         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
27887         (vectorizable_bswap): Likewise.
27888         (vect_gen_perm_mask_checked): Likewise.
27889         * fold-const.c (fold_ternary_loc): Likewise.  Don't take
27890         implementations of variable permutation vectors into account
27891         when deciding which selector to use.
27892         * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
27893         vec_perm_const_optab is supported; instead use can_vec_perm_const_p
27894         with a false third argument.
27895         * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
27896         to test whether the constant selector is valid and can_vec_perm_var_p
27897         to test whether a variable selector is valid.
27899 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
27901         * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
27902         * optabs-query.c (can_vec_perm_p): Likewise.
27903         * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
27904         instead of vec_perm_indices.
27905         * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
27906         (vect_gen_perm_mask_checked): Likewise,
27907         * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
27908         (vect_gen_perm_mask_checked): Likewise,
27910 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
27912         * optabs-query.h (qimode_for_vec_perm): Declare.
27913         * optabs-query.c (can_vec_perm_p): Split out qimode search to...
27914         (qimode_for_vec_perm): ...this new function.
27915         * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
27917 2018-01-02  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
27919         * rtlanal.c (canonicalize_condition): Return 0 if final rtx
27920         does not have a conditional at the top.
27922 2018-01-02  Richard Biener  <rguenther@suse.de>
27924         * ipa-inline.c (big_speedup_p): Fix expression.
27926 2018-01-02  Jan Hubicka  <hubicka@ucw.cz>
27928         PR target/81616
27929         * config/i386/x86-tune-costs.h: Increase cost of integer load costs
27930         for generic 4->6.
27932 2018-01-02  Jan Hubicka  <hubicka@ucw.cz>
27934         PR target/81616
27935         Generic tuning.
27936         * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
27937         cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
27938         and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
27939         cond_taken_branch_cost 3->4.
27941 2018-01-01  Jakub Jelinek  <jakub@redhat.com>
27943         PR tree-optimization/83581
27944         * tree-loop-distribution.c (pass_loop_distribution::execute): Return
27945         TODO_cleanup_cfg if any changes have been made.
27947         PR middle-end/83608
27948         * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
27949         convert_modes if target mode has the right side, but different mode
27950         class.
27952         PR middle-end/83609
27953         * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
27954         last argument when extracting from CONCAT.  If either from_real or
27955         from_imag is NULL, use expansion through memory.  If result is not
27956         a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
27957         the parts directly to inner mode, if even that fails, use expansion
27958         through memory.
27960         PR middle-end/83623
27961         * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
27962         check for bswap in mode rather than HImode and use that in expand_unop
27963         too.
27965 Copyright (C) 2018 Free Software Foundation, Inc.
27967 Copying and distribution of this file, with or without modification,
27968 are permitted in any medium without royalty provided the copyright
27969 notice and this notice are preserved.