re PR fortran/83522 (ICE on allocatable string reference, string(:)(:))
[official-gcc.git] / gcc / ChangeLog
blobb4dc11fe5feb3b94a4066e2bd228081ce0dfa9b4
1 2018-10-09  Eric Botcazou  <ebotcazou@adacore.com>
3         PR tree-optimization/86659
4         * gimple-match.h (gimple_match_op constructors): Initialize reverse.
6 2018-10-09  Richard Biener  <rguenther@suse.de>
8         PR tree-optimization/63155
9         * tree-ssa-structalias.c: Include tree-ssa.h.
10         (get_constraint_for_ssa_var): For undefs return nothing_id.
11         (find_func_aliases): Cleanup PHI handling.
13 2018-10-09  Richard Biener  <rguenther@suse.de>
15         * tree-vectorizer.c (dump_stmt_cost): Fix cut&paste missing
16         replacements.
18 2018-10-09  Martin Liska  <mliska@suse.cz>
20         * asan.c (asan_emit_stack_protection): If a stack variable
21         is located in a same file as current function, then emit
22         line info into variable definition string.
24 2018-10-08  Eric Botcazou  <ebotcazou@adacore.com>
26         * print-rtl.c (rtx_writer::print_rtx_operand_code_i): Print column
27         information.
29 2018-10-08  Eric Botcazou  <ebotcazou@adacore.com>
31         * cgraphunit.c (cgraph_node::expand_thunk): Do not force DECL_IGNORED_P
32         on the thunk.
34 2018-10-08  Eric Botcazou  <ebotcazou@adacore.com>
36         PR tree-optimization/86659
37         * gimple-match.h (struct gimple_match_op): Add reverse field.
38         (gimple_match_op::set_op): New overloaded method.
39         * gimple-match-head.c (maybe_build_generic_op) <BIT_FIELD_REF>: Set
40         the REF_REVERSE_STORAGE_ORDER flag on the value.
41         (gimple_simplify) <GIMPLE_ASSIGN>: For BIT_FIELD_REF, propagate the
42         REF_REVERSE_STORAGE_ORDER flag and avoid simplifying if it is set.
43         
44 2018-10-08  Richard Sandiford  <richard.sandiford@arm.com>
46         PR middle-end/63155
47         * gimple-ssa-backprop.c (backprop::intersect_uses): Use
48         FOR_EACH_IMM_USE_FAST instead of FOR_EACH_IMM_USE_STMT.
50 2018-10-08  H.J. Lu  <hongjiu.lu@intel.com>
52         PR target/87517
53         * config/i386/avx512fintrin.h (_mm512_mask_fmaddsub_round_pd):
54         Defined with __builtin_ia32_vfmaddsubpd512_mask.
56 2018-10-08  Richard Biener  <rguenther@suse.de>
58         * config/i386/x86-tune-costs.h (znver1_cost): Make AVX256 vector loads
59         cost the same as AVX128 ones.
61 2018-10-08  Paul Koning  <ni1d@arrl.net>
63         * config/pdp11/pdp11-protos.h (output_block_move): Remove.
64         (expand_block_move): New function.
65         * config/pdp11/pdp11.c (output_block_move): Remove.
66         (expand_block_move): New function.
67         * config/pdp11/pdp11.h (MOVE_RATIO): New definition.
68         * config/pdp11/pdp11.md (movmemhi): Use expand_block_move.
69         (*movmemhi1): Remove.
71 2018-10-08  Robin Dapp  <rdapp@linux.ibm.com>
73         * config/s390/2827.md: Increase latencies for some FP instructions.
75 2018-10-08  Richard Biener  <rguenther@suse.de>
77         * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
78         Open a dump scope.
79         * tree-vectorizer.c (dump_stmt_cost): Add cost param and dump it.
80         * tree-vectorizer.h (dump_stmt_cost): Adjust.
81         (add_stmt_cost): Dump return value of the hook.
83 2018-10-08  Richard Biener  <rguenther@suse.de>
85         PR tree-optimization/63155
86         * tree-ssa-propagate.c (add_ssa_edge): Do cheap check first.
87         (ssa_propagation_engine::ssa_propagate): Remove redundant
88         bitmap bit clearing.
90 2018-10-05  Peter Bergner  <bergner@linux.ibm.com>
92         PR rtl-optimization/86939
93         PR rtl-optimization/87479
94         * ira.h (non_conflicting_reg_copy_p): New prototype.
95         * ira-lives.c (ignore_reg_for_conflicts): New static variable.
96         (make_hard_regno_dead): Don't add conflicts for register
97         ignore_reg_for_conflicts.
98         (make_object_dead): Likewise.
99         (non_conflicting_reg_copy_p): New function.
100         (process_bb_node_lives): Set ignore_reg_for_conflicts for copies.
101         Remove special conflict handling of REAL_PIC_OFFSET_TABLE_REGNUM.
102         * lra-lives.c (ignore_reg_for_conflicts): New static variable.
103         (make_hard_regno_dead): Don't add conflicts for register
104         ignore_reg_for_conflicts.  Remove special conflict handling of
105         REAL_PIC_OFFSET_TABLE_REGNUM.  Remove now unused argument
106         check_pic_pseudo_p and update callers.
107         (mark_pseudo_dead): Don't add conflicts for register
108         ignore_reg_for_conflicts.
109         (process_bb_lives): Set ignore_reg_for_conflicts for copies.
111 2018-10-05  Andrew Waterman  <andrew@sifive.com>
112             Jim Wilson  <jimw@sifive.com>
114         * config/riscv/riscv.md (f<quiet_pattern>_quiet<ANYF:mode><X:mode>4):
115         Add define_expand.  Add ! HONOR_SNANS check to current pattern.  Add
116         new pattern using HONOR_SNANS that emits one extra instruction.
118 2018-10-05  Segher Boessenkool  <segher@kernel.crashing.org>
120         * config/rs6000/rs6000.md (unnamed mfcr scc_comparison_operator
121         patterns): Merge SI and DI patterns to a GPR pattern.
122         (unnamed define_insn and define_split for record form of that): Merge
123         to a single define_insn_and_split pattern.
125 2018-10-05  David Malcolm  <dmalcolm@redhat.com>
127         PR c++/56856
128         * input.c (expand_location_to_spelling_point): Add param "aspect"
129         and use rather than hardcoding LOCATION_ASPECT_CARET.
130         (get_substring_ranges_for_loc): Handle the case of a single token
131         within a macro expansion.
132         * input.h (expand_location_to_spelling_point): Add "aspect" param,
133         defaulting to LOCATION_ASPECT_CARET.
135 2018-10-05  Paul Koning  <ni1d@arrl.net>
137         * config/pdp11/pdp11.c (TARGET_CXX_GUARD_TYPE): Define.
138         (TARGET_CXX_CLASS_DATA_ALWAYS_COMDAT): Define.
139         (TARGET_CXX_LIBRARY_RTTI_COMDAT): Define.
140         (pdp11_guard_type): New function.
142 2018-10-05  Paul Koning  <ni1d@arrl.net>
144         * config/pdp11/pdp11.h (FLOAT_TYPE_SIZE): Always 32.
145         * config/pdp11/pdp11.opt (mfloat32): Remove.
146         (mfloat64): Remove.
147         * doc/invoke.texi (pdp11 -mfloat32): Remove:
148         (pdp11 -mfloat64): Remove.
150 2018-10-05  Uros Bizjak  <ubizjak@gmail.com>
152         * config/i386/i386.md (*cmpxf_cc_i387): Remove pattern.
153         (*cmp<mode>_cc_i387): Ditto.
154         (*cmpu<mode>_cc_i387): Ditto.
155         (*cmp<X87MODEF:mode>_<SWI24:mode>_cc_i387): Ditto.
156         * config/i386/i386.c (ix86_expand_fp_compare): Remove
157         "scratch" argument.
158         <case IX86_FPCMP_SAHF>: Do not generate pattern with HImode clobber.
159         Emit x86_sahf_1 pattern.
160         (ix86_expand_compare): Update call to ix86_expand_fp_compare.
161         (ix86_expand_carry_flag_compare): Ditto.
163 2018-10-05  Uros Bizjak  <ubizjak@gmail.com>
165         * config/i386/i386.md (*cmpxf_i387): Change operand 2 predicate
166         to reg_or_0_operand.  Add "C" constraint.
167         (*cmpxf_cc_i387): Ditto.
168         (*cmp<mode>_i387): Change operand 2 predicate
169         to nonimm_or_0_operand.  Add "C" constraint.
170         (*cmp<mode>_cc_i387): Ditto.
171         (*cmp<mode>_0_i387): Remove insn pattern.
172         (*cmp<mode>_0_cc_i387): Ditto.
174 2018-10-05  Uros Bizjak  <ubizjak@gmail.com>
176         * config/i386/constraints.md ("C"): Do not depend on TARGET_SSE.
177         * config/i386/predicates.md (nonimm_or_0_operand): Rename
178         from vector_move_operand.  Update all uses.
180 2018-10-05  Martin Sebor  <msebor@redhat.com>
182         PR tree-optimization/87490
183         * builtins.c (expand_builtin_strnlen): Handle a null data.decl
184         consistently.
186 2018-10-05  Richard Biener  <rguenther@suse.de>
188         PR tree-optimization/63155
189         * tree-ssa-ccp.c (ccp_propagate::visit_phi): Avoid excess
190         vertical space in dumpfiles.
191         * tree-ssa-propagate.h
192         (ssa_propagation_engine::process_ssa_edge_worklist): Remove.
193         * tree-ssa-propagate.c (cfg_blocks_back): New global.
194         (ssa_edge_worklist_back): Likewise.
195         (curr_order): Likewise.
196         (cfg_blocks_get): Remove abstraction.
197         (cfg_blocks_add): Likewise.
198         (cfg_blocks_empty_p): Likewise.
199         (add_ssa_edge): Add to current or next worklist based on
200         RPO index.
201         (add_control_edge): Likewise.
202         (ssa_propagation_engine::process_ssa_edge_worklist): Fold
203         into ...
204         (ssa_propagation_engine::ssa_propagate): ... here.  Unify
205         iteration from CFG and SSA edge worklist so we process
206         everything in RPO order, prioritizing forward progress
207         over iteration.
208         (ssa_prop_init): Allocate new worklists, do not dump
209         immediate uses.
210         (ssa_prop_fini): Free new worklists.
212 2018-10-05  Richard Biener  <rguenther@suse.de>
214         * tree-core.h (tree_block::abstract_flag): Remove.
215         (tree_block::block_num): Make full 32bits.
216         * tree.def (BLOCK): Remove docs about BLOCK_ABSTRACT.
217         * tree.h (BLOCK_ABSTRACT): Remove.
218         * dwarf2out.c (gen_lexical_block_die): Remove dead code
219         resulting from BLOCK_ABSTRACT being always false.
220         (gen_inlined_subroutine_die): Likewise.
221         (gen_block_die): Likewise.
222         * tree.c (block_ultimate_origin): Likewise.
223         * tree-pretty-print.c (dump_block_node): Remove code dealing
224         with BLOCK_ABSTRACT.
225         * tree-ssa-live.c (dump_scope_block): Likewise.
226         * tree-streamer-in.c (unpack_ts_block_value_fields): Likewise.
227         * tree-streamer-out.c (pack_ts_block_value_fields): Likewise.
229 2018-10-05   Richard Biener  <rguenther@suse.de>
231         * config/i386/i386.c (ix86_add_stmt_cost): When scalar cost
232         is asked for initialize mode to the component mode of the
233         vector type.
235 2018-10-05  H.J. Lu  <hongjiu.lu@intel.com>
237         PR target/87522
238         * config/i386/gnu-user.h (ASM_SPEC): Don't pass -msse2avx to
239         assembler for -mavx.
240         * config/i386/gnu-user64.h (ASM_SPEC): Likewise.
242 2018-10-05  Segher Boessenkool  <segher@kernel.crashing.org>
244         PR target/87509
245         * config/rs6000/rs6000-builtin.def (RS6000_BUILTIN_SET_FPSCR_DRN): Use
246         RS6000_BTM_DFP.
247         * config/rs6000/rs6000.md (rs6000_set_fpscr_rn): Require the operand
248         to be DImode.  When using mffscrn, force the operand to a register.
250 2018-10-04  Uros Bizjak  <ubizjak@gmail.com>
252         * config/i386/i386.md (*fop_<X87MODEF:mode>_2_i387): Macroize insn
253         from *fop_<MODEF:mode>_2_i387 and *fop_xf_2_i387 using
254         X87MODEF mode iterator.
255         (*fop_<X87MODEF:mode>_3_i387): Macroize insn from
256         *fop_<MODEF:mode>_3_i387 and *fop_xf_3_i387 using
257         X87MODEF mode iterator.
259 2018-10-04  Vinay Kumar  <vinay.kumar@blackfigtech.com>
261         * doc/invoke.texi (-Wno-prio-ctor-dtor): Document new warning
262         -Wno-prio-ctor-dtor.
264 2018-10-04  David Malcolm  <dmalcolm@redhat.com>
266         * Makefile.in (OBJS): Add opt-problem.o.
267         * dump-context.h: Include "selftest.h.
268         (selftest::temp_dump_context): New forward decl.
269         (class dump_context): Make friend of class
270         selftest::temp_dump_context.
271         (dump_context::dump_loc_immediate): New decl.
272         (class dump_pretty_printer): Move here from dumpfile.c.
273         (class temp_dump_context): Move to namespace selftest.
274         (temp_dump_context::temp_dump_context): Add param
275         "forcibly_enable_dumping".
276         (selftest::verify_dumped_text):
277         (ASSERT_DUMPED_TEXT_EQ): Move here from dumpfile.c.
278         (selftest::verify_item):
279         (ASSERT_IS_TEXT): Move here from dumpfile.c.
280         (ASSERT_IS_TREE): Likewise.
281         (ASSERT_IS_GIMPLE): Likewise.
282         * dumpfile.c (dump_context::dump_loc): Move immediate dumping
283         to...
284         (dump_context::dump_loc_immediate): ...this new function.
285         (class dump_pretty_printer): Move to dump-context.h.
286         (dump_switch_p_1): Don't enable MSG_PRIORITY_REEMITTED.
287         (opt_info_switch_p_1): Enable MSG_PRIORITY_REEMITTED.
288         (temp_dump_context::temp_dump_context): Move to "selftest"
289         namespace.  Add param "forcibly_enable_dumping", and use it to
290         conditionalize the use of m_pp;
291         (selftest::verify_dumped_text): Make non-static.
292         (ASSERT_DUMPED_TEXT_EQ): Move to dump-context.h.
293         (selftest::verify_item): Make non-static.
294         (ASSERT_IS_TEXT): Move to dump-context.h.
295         (ASSERT_IS_TREE): Likewise.
296         (ASSERT_IS_GIMPLE): Likewise.
297         (selftest::test_capture_of_dump_calls): Pass "true" for new
298         param of temp_dump_context.
299         * dumpfile.h (enum dump_flag): Add MSG_PRIORITY_REEMITTED, adding
300         it to MSG_ALL_PRIORITIES.  Update values of TDF_COMPARE_DEBUG and
301         TDF_COMPARE_DEBUG.
302         * opt-problem.cc: New file.
303         * opt-problem.h: New file.
304         * optinfo-emit-json.cc
305         (selftest::test_building_json_from_dump_calls): Pass "true" for
306         new param of temp_dump_context.
307         * optinfo.cc (optinfo_kind_to_dump_flag): New function.
308         (optinfo::emit_for_opt_problem): New function.
309         (optinfo::emit): Clarity which emit_item is used.
310         * optinfo.h (optinfo::get_dump_location): New accessor.
311         (optinfo::emit_for_opt_problem): New decl.
312         (optinfo::emit): Make const.
313         * selftest-run-tests.c (selftest::run_tests): Call
314         selftest::opt_problem_cc_tests.
315         * selftest.h (selftest::opt_problem_cc_tests): New decl.
316         * tree-data-ref.c (dr_analyze_innermost): Convert return type from
317         bool to opt_result, converting fprintf messages to
318         opt_result::failure_at calls.  Add "stmt" param for use by the
319         failure_at calls.
320         (create_data_ref): Pass "stmt" to the dr_analyze_innermost call.
321         (runtime_alias_check_p): Convert return type from bool to
322         opt_result, converting dump_printf calls to
323         opt_result::failure_at, using the statement DDR_A for their
324         location.
325         (find_data_references_in_stmt): Convert return type from bool to
326         opt_result, converting "return false" to opt_result::failure_at
327         with a new message.
328         * tree-data-ref.h: Include "opt-problem.h".
329         (dr_analyze_innermost): Convert return type from bool to opt_result,
330         and add a const gimple * param.
331         (find_data_references_in_stmt): Convert return type from bool to
332         opt_result.
333         (runtime_alias_check_p): Likewise.
334         * tree-predcom.c (find_looparound_phi): Pass "init_stmt" to
335         dr_analyze_innermost.
336         * tree-vect-data-refs.c (vect_mark_for_runtime_alias_test):
337         Convert return type from bool to opt_result, adding a message for
338         the PARAM_VECT_MAX_VERSION_FOR_ALIAS_CHECKS zero case.
339         (vect_analyze_data_ref_dependence): Convert return type from bool
340         to opt_result.  Change sense of return type from "false"
341         effectively meaning "no problems" to "false" meaning a problem,
342         so that "return false" becomes "return opt_result::success".
343         Convert "return true" calls to opt_result::failure_at, using
344         the location of statement A rather than vect_location.
345         (vect_analyze_data_ref_dependences): Convert return type from bool
346         to opt_result.
347         (verify_data_ref_alignment): Likewise, converting dump_printf_loc
348         calls to opt_result::failure_at, using the stmt location rather
349         than vect_location.
350         (vect_verify_datarefs_alignment): Convert return type from bool
351         to opt_result.
352         (vect_enhance_data_refs_alignment): Likewise.  Split local "stat"
353         into multiple more-tightly-scoped copies.
354         (vect_analyze_data_refs_alignment): Convert return type from bool
355         to opt_result.
356         (vect_analyze_data_ref_accesses): Likewise, converting a
357         "return false" to a "return opt_result::failure_at", adding a
358         new message.
359         (vect_prune_runtime_alias_test_list): Convert return type from
360         bool to opt_result, converting dump_printf_loc to
361         opt_result::failure_at.  Add a %G to show the pertinent statement,
362         and use the stmt's location rather than vect_location.
363         (vect_find_stmt_data_reference): Convert return type from
364         bool to opt_result, converting dump_printf_loc to
365         opt_result::failure_at, using stmt's location.
366         (vect_analyze_data_refs):  Convert return type from bool to
367         opt_result.  Convert "return false" to "return
368         opt_result::failure_at", adding messages as needed.
369         * tree-vect-loop.c (vect_determine_vf_for_stmt_1): Convert return
370         type from bool to opt_result.
371         (vect_determine_vf_for_stmt): Likewise.
372         (vect_determine_vectorization_factor): Likewise, converting
373         dump_printf_loc to opt_result::failure_at, using location of phi
374         rather than vect_location.
375         (vect_analyze_loop_form_1): Convert return type from bool to
376         opt_result, converting dump_printf_loc calls, retaining the use of
377         vect_location.
378         (vect_analyze_loop_form): Convert return type from loop_vec_info
379         to opt_loop_vec_info.
380         (vect_analyze_loop_operations): Convert return type from bool to
381         opt_result, converting dump_printf_loc calls, using the location
382         of phi/stmt rather than vect_location where available.  Convert
383         various "return false" to "return opt_result::failure_at" with
384         "unsupported phi" messages.
385         (vect_get_datarefs_in_loop): Convert return type from bool to
386         opt_result.  Add a message for the
387         PARAM_LOOP_MAX_DATAREFS_FOR_DATADEPS failure.
388         (vect_analyze_loop_2): Convert return type from bool to
389         opt_result.  Ensure "ok" is set to a opt_result::failure_at before
390         each "goto again;", adding new messages where needed.
391         Add "unsupported grouped {store|load}" messages.
392         (vect_analyze_loop): Convert return type from loop_vec_info to
393         opt_loop_vec_info.
394         * tree-vect-slp.c (vect_analyze_slp): Convert return type from
395         bool to opt_result.
396         * tree-vect-stmts.c (process_use): Likewise, converting
397         dump_printf_loc call and using stmt location, rather than
398         vect_location.
399         (vect_mark_stmts_to_be_vectorized): Likeise.
400         (vect_analyze_stmt): Likewise, adding a %G.
401         (vect_get_vector_types_for_stmt): Convert return type from bool to
402         opt_result, converting dump_printf_loc calls and using stmt
403         location, rather than vect_location.
404         (vect_get_mask_type_for_stmt): Convert return type from tree to
405         opt_tree, converting dump_printf_loc calls and using stmt location.
406         * tree-vectorizer.c: Include "opt-problem.h.
407         (try_vectorize_loop_1): Flag "Analyzing loop at" dump message as
408         MSG_PRIORITY_INTERNALS.  Convert local "loop_vinfo" from
409         loop_vec_info to opt_loop_vec_info.  If if fails, and dumping is
410         enabled, use it to report at the top level "couldn't vectorize
411         loop" followed by the problem.
412         * tree-vectorizer.h (opt_loop_vec_info): New typedef.
413         (vect_mark_stmts_to_be_vectorized): Convert return type from bool
414         to opt_result.
415         (vect_analyze_stmt): Likewise.
416         (vect_get_vector_types_for_stmt): Likewise.
417         (tree vect_get_mask_type_for_stmt): Likewise.
418         (vect_analyze_data_ref_dependences): Likewise.
419         (vect_enhance_data_refs_alignment): Likewise.
420         (vect_analyze_data_refs_alignment): Likewise.
421         (vect_verify_datarefs_alignment): Likewise.
422         (vect_analyze_data_ref_accesses): Likewise.
423         (vect_prune_runtime_alias_test_list): Likewise.
424         (vect_find_stmt_data_reference): Likewise.
425         (vect_analyze_data_refs): Likewise.
426         (vect_analyze_loop): Convert return type from loop_vec_info to
427         opt_loop_vec_info.
428         (vect_analyze_loop_form): Likewise.
429         (vect_analyze_slp): Convert return type from bool to opt_result.
431 2018-10-04  David Malcolm  <dmalcolm@redhat.com>
433         * doc/invoke.texi (-fopt-info): Document new "internals"
434         sub-option.
435         * dump-context.h (dump_context::apply_dump_filter_p): New decl.
436         * dumpfile.c (dump_options): Update for renaming of MSG_ALL to
437         MSG_ALL_KINDS.
438         (optinfo_verbosity_options): Add "internals".
439         (kind_as_string): Update for renaming of MSG_ALL to MSG_ALL_KINDS.
440         (dump_context::apply_dump_filter_p): New member function.
441         (dump_context::dump_loc): Use apply_dump_filter_p rather than
442         explicitly masking the dump_kind.
443         (dump_context::begin_scope): Increment the scope depth first.  Use
444         apply_dump_filter_p rather than explicitly masking the dump_kind.
445         (dump_context::emit_item): Use apply_dump_filter_p rather than
446         explicitly masking the dump_kind.
447         (dump_dec): Likewise.
448         (dump_hex): Likewise.
449         (dump_switch_p_1): Default to MSG_ALL_PRIORITIES.
450         (opt_info_switch_p_1): Default to MSG_PRIORITY_USER_FACING.
451         (opt_info_switch_p): Update handling of default
452         MSG_OPTIMIZED_LOCATIONS to cope with default of
453         MSG_PRIORITY_USER_FACING.
454         (dump_basic_block): Use apply_dump_filter_p rather than explicitly
455         masking the dump_kind.
456         (selftest::test_capture_of_dump_calls): Update test_dump_context
457         instances to use MSG_ALL_KINDS | MSG_PRIORITY_USER_FACING rather
458         than MSG_ALL.  Generalize scope test to be run at all four
459         combinations of with/without MSG_PRIORITY_USER_FACING and
460         MSG_PRIORITY_INTERNALS, adding examples of explicit priority
461         for each of the two values.
462         * dumpfile.h (enum dump_flag): Add comment about the MSG_* flags.
463         Rename MSG_ALL to MSG_ALL_KINDS.  Add MSG_PRIORITY_USER_FACING,
464         MSG_PRIORITY_INTERNALS, and MSG_ALL_PRIORITIES, updating the
465         values for TDF_COMPARE_DEBUG and TDF_ALL_VALUES.
466         (AUTO_DUMP_SCOPE): Add a note to the comment about the interaction
467         with MSG_PRIORITY_*.
468         * tree-vect-loop-manip.c (vect_loop_versioning): Mark versioning
469         dump messages as MSG_PRIORITY_USER_FACING.
470         * tree-vectorizer.h (DUMP_VECT_SCOPE): Add a note to the comment
471         about the interaction with MSG_PRIORITY_*.
473 2018-10-04  Bernd Edlinger  <bernd.edlinger@hotmail.de>
475         * varasm.c (output_constant): Add new parameter merge_strings.
476         Make strings properly zero terminated in merge string sections.
477         (mergeable_string_section): Don't fail if the last char is non-zero.
478         (assemble_variable_contents): Handle merge string sections.
479         (assemble_variable): Likewise.
480         (assemble_constant_contents): Likewise.
481         (output_constant_def_contents): Likewise.
482         (output_constructor_array_range,
483         output_constructor_regular_field): Adjust call to output_constant.
484         (output_object_block): Adjust call to assemble_constant_contents
485         and assemble_variable_contents.
487 2018-10-04  Martin Liska  <mliska@suse.cz>
489         PR c/87483
490         * cgraphunit.c (process_function_and_variable_attributes):
491         Warn about a function with alias attribute and a body.
493 2018-10-04  Martin Liska  <mliska@suse.cz>
495         PR ipa/82625
496         * multiple_target.c (redirect_to_specific_clone): New function.
497         (ipa_target_clone): Use it.
498         * tree-inline.c: Fix comment.
500 2018-10-04  David Malcolm  <dmalcolm@redhat.com>
502         * dumpfile.c (gcc::dump_manager::dump_manager): Initialize new
503         fields.
504         (gcc::dump_manager::~dump_manager): Free m_optinfo_filename.
505         (gcc::dump_manager::register_pass): New member function, adapted
506         from loop body in gcc::pass_manager::register_pass, adding a
507         call to update_dfi_for_opt_info.
508         (gcc::dump_manager::opt_info_enable_passes): Store the
509         -fopt-info options into the new fields.  Move the loop
510         bodies into...
511         (gcc::dump_manager::update_dfi_for_opt_info): ...this new member
512         function.
513         * dumpfile.h (struct opt_pass): New forward decl.
514         (gcc::dump_manager::register_pass): New decl.
515         (gcc::dump_manager::update_dfi_for_opt_info): New decl.
516         (class gcc::dump_manager): Add fields "m_optgroup_flags",
517         "m_optinfo_flags", and "m_optinfo_filename".
518         * passes.c (gcc::pass_manager::register_pass): Move all of the
519         dump-handling code to gcc::dump_manager::register_pass.
521 2018-10-04  Peter Bergner  <bergner@linux.ibm.com>
523         PR rtl-optimization/87466
524         * target.def (setjmp_preserves_nonvolatile_regs_p): New target hook.
525         * doc/tm.texi.in (TARGET_SETJMP_PRESERVES_NONVOLATILE_REGS_P): New hook.
526         * doc/tm.texi: Regenerate.
527         * ira-lives.c (process_bb_node_lives): Use the new target hook.
528         * lra-lives.c (process_bb_lives): Likewise.
529         * config/rs6000/rs6000.c (TARGET_SETJMP_PRESERVES_NONVOLATILE_REGS_P):
530         Define.
532 2018-10-04  Tamar Christina  <tamar.christina@arm.com>
534         * params.c (add_params): Fix initialization.
536 2018-10-04  Martin Liska  <mliska@suse.cz>
538         PR gcov-profile/84107
539         * tree-profile.c (init_ic_make_global_vars):
540         Remove ic_void_ptr_var and ic_gcov_type_ptr_var.
541         Come up with new ic_tuple* variables.  Emit
542         __gcov_indirect_call{,_topn} variables.
543         (gimple_gen_ic_profiler): Access the variable
544         and emit gimple.
545         (gimple_gen_ic_func_profiler): Access
546         __gcov_indirect_call.callee field.
547         (gimple_init_gcov_profiler): Use ptr_type_node.
548         * value-prof.c (gimple_ic): Use ptr_type_node.
550 2018-10-04  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
552         PR tree-optimization/85787
553         * ipa-pure-const.c (malloc_candidate_p_1): Move most of malloc_candidate_p
554         into this function and add support for detecting multiple phis.
555         (DUMP_AND_RETURN): Move from malloc_candidate_p into top-level macro.
557 2018-10-04  Martin Liska  <mliska@suse.cz>
559         PR ipa/87491
560         * ipa-inline.c (inline_to_all_callers_1):
561         Call ultimate_alias_target for node being inlined.
563 2018-10-03  Jeff Law  <law@redhat.com>
565         * gimple-ssa-sprintf.c (format_string): Do not hardcode size of
566         target's wchar_t.
567         * tree.c (get_typenode_from_name): Moved from fortran/trans-types.c.
568         * tree.h (get_typenode_from_name): Prototype.
570 2018-10-03  Uros Bizjak  <ubizjak@gmail.com>
572         * config/i386/i386.md (*cmp<X87MODEF:mode>_<SWI24:mode>_i387):
573         Change operand 2 predicate to nonimmediate_operand.
574         (*cmp<X87MODEF:mode>_<SWI24:mode>_cc_i387): Ditto.
576 2018-10-03  Martin Sebor  <msebor@redhat.com>
577             Jeff Law  <law@redhat.com>
579         * gimple-ssa-sprintf.c (struct fmtresult): Add new member and
580         initialize it.
581         (get_string_length): Detect unterminated arrays.
582         (format_string): Same.
583         (format_directive): Warn about unterminated arrays.
584         (handle_gimple_call): Mark statements with no_warning as needed.
586 2018-10-03  Jim Wilson  <jimw@sifive.com>
588         * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): For ABI_ILP32E,
589         also define __riscv_abi_rve.  Delete trailing white space.
591 2018-10-03  Paul Koning  <ni1d@arrl.net>
593         Enable LRA register allocator for PDP11.
594         * config/pdp11/constraints.md (Q): Use define_memory_constraint.
595         (R): Likewise.
596         (D): Likewise.
597         * config/pdp11/pdp11.c (pdp11_lra_p): New function.
598         * config/pdp11/pdp11.opt (-mlra): New option.
599         * doc/invoke.texi (PDP-11 Options): Document -mlra.
601 2018-10-03  Uros Bizjak  <ubizjak@gmail.com>
603         * config/i386/i386.md (*<absneg:code>extendsfdf2): Remove.
604         (*<absneg:code>extend<mode>xf2): Ditto.
606 2018-10-03  Aldy Hernandez  <aldyh@redhat.com>
608         PR tree-optimization/87415
609         * tree-vrp.c (set_value_range_with_overflow): Special case one bit
610         precision fields.
612 2018-10-02  Jeff Law  <law@redhat.com>
614         * gimple-fold.c (get_range_strlen): Only set *nonstr when
615         an unterminated string is discovered.  Bubble up range
616         even for unterminated strings.
617         (gimple_fold_builtin_strlen): Do not fold if get_range_strlen
618         indicates the string was not terminated via NONSTR.
620 2018-10-03  Aldy Hernandez  <aldyh@redhat.com>
622         * tree-vrp.c (extract_range_from_unary_expr): Special case all
623         pointer conversions.
624         Do not do anything special for anti-ranges.
626 2018-10-03  Jérôme Lambourg  <lambourg@adacore.com>
628         * config/arm/vxworks.h (ARM_TARGET2_DWARF_FORMAT): Adjust to
629         DW_EH_PE_pcrel | DW_EH_PE_indirect for RTPs.
631 2018-10-03  Martin Liska  <mliska@suse.cz>
633         PR gcov-profile/86109
634         * coverage.c (coverage_begin_function): Do not
635         mark lambdas as artificial.
636         * tree-core.h (struct GTY): Remove tm_clone_flag
637         and introduce new lambda_function.
638         * tree.h (DECL_LAMBDA_FUNCTION): New macro.
640 2018-10-02  Aaron Sawdey  <acsawdey@linux.ibm.com>
642         PR target/87474
643         * config/rs6000/rs6000-string.c (expand_strn_compare): Check that both
644         P8_VECTOR and VSX are enabled.
646 2018-10-02  Andreas Krebbel  <krebbel@linux.ibm.com>
648         * config/s390/driver-native.c (s390_host_detect_local_cpu): Add
649         0x3907 as CPU model number.
651 2018-10-02  Andreas Krebbel  <krebbel@linux.ibm.com>
653         * common/config/s390/s390-common.c: Rename PF_ARCH12 to PF_Z14.
654         * config/s390/s390.h (enum processor_flags): Rename PF_ARCH12 to
655         PF_Z14.  Rename TARGET_CPU_ARCH12 to TARGET_CPU_Z14,
656         TARGET_CPU_ARCH12_P to TARGET_CPU_Z14_P, TARGET_ARCH12 to
657         TARGET_Z14, and TARGET_ARCH12_P to TARGET_Z14_P.
658         * config/s390/s390.md: Likewise. Rename also the cpu attribute
659         value from arch12 to z14.
661 2018-10-02  Uros Bizjak  <ubizjak@gmail.com>
663         * config/i386/i386.md (fxam<mode>2_i387_with_temp): Remove.
664         (isinfxf2): Ditto.
665         (isinf<mode>2): Ditto.
667 2018-10-02  Uros Bizjak  <ubizjak@gmail.com>
669         * config/i386/i386.c (ix86_emit_i387_round): Extend op1 to XFmode
670         before emitting fxam.  Perform calculations in XFmode.
672 2018-10-02  Marc Glisse  <marc.glisse@inria.fr>
674         * match.pd (((X /[ex] A) +- B) * A): New transformation.
676 2018-10-02  Marc Glisse  <marc.glisse@inria.fr>
678         PR middle-end/87319
679         * fold-const.c (fold_plusminus_mult_expr): Handle complex and vectors.
680         * tree.c (signed_or_unsigned_type_for): Handle complex.
682 2018-10-02  Jeff Law  <law@redhat.com>
684         * gimple-fold.c (get_range_strlen): Remove dead code.
686 2018-10-02  Martin Sebor  <msebor@redhat.com>
687             Jeff Law  <law@redhat.com>
689         * builtins.c (unterminated_array): Add new arguments.
690         If argument is not terminated, bubble up size and exact
691         state to callers.
692         (expand_builtin_strnlen): Detect, avoid expanding
693         and diagnose unterminated arrays.
694         (c_strlen): Fill in offset of start of unterminated strings.
695         * builtins.h (unterminated_array): Update prototype.
697 2018-10-02  Richard Biener  <rguenther@suse.de>
699         * config/i386/sse.md (reduc_plus_scal_v4df): Avoid the use
700         of haddv4df, first reduce to SSE width and exploit the fact
701         that we only need element zero with the reduction result.
702         (reduc_plus_scal_v2df): Likewise.
704 2018-10-02  Eric Botcazou  <ebotcazou@adacore.com>
706         * dojump.h (do_jump): Delete.
707         (do_jump_1): Likewise.
708         (split_comparison): Move around.
709         * dojump.c (do_jump): Make static.
710         (do_jump_1): Likewise.
711         (jumpifnot): Move around.
712         (jumpifnot_1): Likewise.
713         (jumpif): Likewise.
714         (jumpif_1): Likewise.
715         * expr.c (expand_expr_real_1): Call jumpif[not] instead of do_jump.
717 2018-10-02  Eric Botcazou  <ebotcazou@adacore.com>
719         * reorg.c (make_return_insns): Use emit_copy_of_insn_after for the
720         insns in the delay slot and add_insn_after for the jump insn.
722 2018-10-02  Richard Biener  <rguenther@suse.de>
724         * tree-inline.c (expand_call_inline): Use the location of
725         the callee declaration for the inline-entry marker.
726         * final.c (notice_source_line): Remove special-casing of
727         NOTE_INSN_INLINE_ENTRY.
729 2018-10-01  Carl Love  <cel@us.ibm.com>
731         PR 69431
732         * config/rs6000/rs6000-builtin.def (__builtin_mffsl): New.
733         (__builtin_mtfsb0): New.
734         (__builtin_mtfsb1): New.
735         ( __builtin_set_fpscr_rn): New.
736         (__builtin_set_fpscr_drn): New.
737         * config/rs6000/rs6000.c (rs6000_expand_mtfsb_builtin): Add.
738         (rs6000_expand_set_fpscr_rn_builtin): Add.
739         (rs6000_expand_set_fpscr_drn_builtin): Add.
740         (rs6000_expand_builtin): Add case statement entries for
741         RS6000_BUILTIN_MTFSB0, RS6000_BUILTIN_MTFSB1,
742         RS6000_BUILTIN_SET_FPSCR_RN, RS6000_BUILTIN_SET_FPSCR_DRN,
743         RS6000_BUILTIN_MFFSL.
744         (rs6000_init_builtins): Add ftype initialization and def_builtin
745         calls for __builtin_mffsl, __builtin_mtfsb0, __builtin_mtfsb1,
746         __builtin_set_fpscr_rn, __builtin_set_fpscr_drn.
747         * config/rs6000.md (rs6000_mtfsb0, rs6000_mtfsb1, rs6000_mffscrn,
748         rs6000_mffscdrn): Add define_insn.
749         (rs6000_set_fpscr_rn, rs6000_set_fpscr_drn): Add define_expand.
750         * doc/extend.texi: Add documentation for the builtins.
752 2018-10-01  Richard Biener  <rguenther@suse.de>
754         PR tree-optimization/87465
755         * tree-ssa-loop-ivcanon.c (tree_estimate_loop_size): Fix typo
756         causing branch miscounts.
758 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
760         * common/config/aarch64/aarch64-common.c (TARGET_OPTION_DEFAULT_PARAM,
761         aarch64_option_default_param):  New.
762         (params.h): Include.
763         (TARGET_OPTION_VALIDATE_PARAM, aarch64_option_validate_param): New.
764         * config/aarch64/aarch64.c (aarch64_override_options_internal): Simplify
765         stack-clash protection validation code.
767 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
769         * params.c (validate_param): New.
770         (add_params): Use it.
771         (set_param_value): Refactor param validation into validate_param.
772         (diagnostic.h): Include.
773         * diagnostic.h (diagnostic_ready_p): New.
775 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
777         * params.c (set_param_value):
778         Add index of parameter being validated.
779         * common/common-target.def (option_validate_param): New.
780         * common/common-targhooks.h (default_option_validate_param): New.
781         * common/common-targhooks.c (default_option_validate_param): New.
782         * doc/tm.texi.in (TARGET_OPTION_VALIDATE_PARAM): New.
783         * doc/tm.texi: Regenerate.
785 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
787         PR target/86486
788         * config/aarch64/aarch64.c (aarch64_override_options_internal):
789         Add validation for stack-clash parameters and set defaults.
791 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
793         PR target/86486
794         * configure.ac: Add stack-clash-protection-guard-size.
795         * doc/install.texi: Document it.
796         * config.in (DEFAULT_STK_CLASH_GUARD_SIZE): New.
797         * params.def: Update comment for guard-size.
798         (PARAM_STACK_CLASH_PROTECTION_GUARD_SIZE,
799         PARAM_STACK_CLASH_PROTECTION_PROBE_INTERVAL): Update description.
800         * configure: Regenerate.
802 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
804         PR target/86486
805         * config/aarch64/aarch64.h (STACK_CLASH_MIN_BYTES_OUTGOING_ARGS,
806         STACK_DYNAMIC_OFFSET): New.
807         * config/aarch64/aarch64.c (aarch64_layout_frame):
808         Update outgoing args size.
809         (aarch64_stack_clash_protection_alloca_probe_range,
810         TARGET_STACK_CLASH_PROTECTION_ALLOCA_PROBE_RANGE): New.
812 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
814         PR target/86486
815         * explow.c (anti_adjust_stack_and_probe_stack_clash): Support custom
816         probe ranges.
817         * target.def (stack_clash_protection_alloca_probe_range): New.
818         (stack_clash_protection_final_dynamic_probe): Remove.
819         * targhooks.h (default_stack_clash_protection_alloca_probe_range) New.
820         (default_stack_clash_protection_final_dynamic_probe): Remove.
821         * targhooks.c: Likewise.
822         * doc/tm.texi.in (TARGET_STACK_CLASH_PROTECTION_ALLOCA_PROBE_RANGE): New.
823         (TARGET_STACK_CLASH_PROTECTION_FINAL_DYNAMIC_PROBE): Remove.
824         * doc/tm.texi: Regenerate.
826 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
828         PR target/86486
829         * config/aarch64/aarch64-protos.h (aarch64_output_probe_sve_stack_clash): New.
830         * config/aarch64/aarch64.c (aarch64_output_probe_sve_stack_clash,
831         aarch64_clamp_to_uimm12_shift): New.
832         (aarch64_allocate_and_probe_stack_space): Add SVE specific section.
833         * config/aarch64/aarch64.md (probe_sve_stack_clash): New.
835 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
837         PR target/86486
838         * config/aarch64/aarch64.c (aarch64_layout_frame): Add assert.
840 2018-10-01  Jeff Law  <law@redhat.com>
841             Richard Sandiford <richard.sandiford@linaro.org>
842             Tamar Christina  <tamar.christina@arm.com>
844         PR target/86486
845         * config/aarch64/aarch64.md
846         (probe_stack_range): Add k (SP) constraint.
847         * config/aarch64/aarch64.h (STACK_CLASH_CALLER_GUARD,
848         STACK_CLASH_MAX_UNROLL_PAGES): New.
849         * config/aarch64/aarch64.c (aarch64_output_probe_stack_range): Emit
850         stack probes for stack clash.
851         (aarch64_allocate_and_probe_stack_space): New.
852         (aarch64_expand_prologue): Use it.
853         (aarch64_expand_epilogue): Likewise and update IP regs re-use criteria.
854         (aarch64_sub_sp): Add emit_move_imm optional param.
856 2018-10-01  MCC CS <deswurstes@users.noreply.github.com>
858         PR tree-optimization/87261
859         * match.pd: Remove trailing whitespace.
860         Add (x & y) | ~(x | y) -> ~(x ^ y),
861         (~x | y) ^ (x ^ y) -> x | ~y and (x ^ y) | ~(x | y) -> ~(x & y)
863 2018-10-01  Claudiu Zissulescu  <claziss@synopsys.com>
865         * config/arc/arc.md (*add_n): Clean up pattern, update instruction
866         constraints.
867         (ashlsi3_insn): Update instruction constraints.
868         (ashrsi3_insn): Likewise.
869         (rotrsi3): Likewise.
870         (add_shift): Likewise.
871         * config/arc/constraints.md (Csz): New 32 bit constraint. It
872         avoids placing in the limm field small constants which, otherwise,
873         could end into a small instruction.
875 2018-10-01  Claudiu Zissulescu  <claziss@synopsys.com>
877         * config/arc/arc.md (maddsidi4_split): Don't use dmac if the
878         destination register is not odd-even.
879         (umaddsidi4_split): Likewise.
881 2018-10-01  Richard Biener  <rguenther@suse.de>
883         * tree-inline.c (expand_call_inline): Store origin of fn
884         in BLOCK_ABSTRACT_ORIGIN for the inline BLOCK.
885         * tree.c (block_ultimate_origin): Simplify and do some
886         checking.
888 2018-09-30  Uros Bizjak  <ubizjak@gmail.com>
890         * config/i386/mmx.md (EMMS): New int iterator.
891         (emms): New int attribute.
892         (mmx_<emms>): Macroize insn from *mmx_emms and *mmx_femms using
893         EMMS int iterator.  Explicitly declare clobbers.
894         (mmx_emms): Remove expander.
895         (mmx_femms): Ditto.
896         * config/i386/predicates.md (emms_operation): Remove predicate.
897         (vzeroall_pattern): New predicate.
898         (vzeroupper_pattern): Rename from vzeroupper_operation.
899         * config/i386/i386.c (ix86_avx_u128_mode_after): Use
900         vzeroupper_pattern and vzeroall_pattern predicates.
902 2018-09-30  Peter Bergner  <bergner@linux.ibm.com>
904         PR rtl-optimization/86939
905         * ira-lives.c (make_hard_regno_born): Rename from this...
906         (make_hard_regno_live): ... to this.  Remove update to conflict
907         information.  Update function comment.
908         (make_hard_regno_dead): Add conflict information update.  Update
909         function comment.
910         (make_object_born): Rename from this...
911         (make_object_live): ... to this.  Remove update to conflict information.
912         Update function comment.
913         (make_object_dead):  Add conflict information update.  Update function
914         comment.
915         (mark_pseudo_regno_live): Call make_object_live.
916         (mark_pseudo_regno_subword_live): Likewise.
917         (mark_hard_reg_dead): Update function comment.
918         (mark_hard_reg_live): Call make_hard_regno_live.
919         (process_bb_node_lives): Likewise.
920         * lra-lives.c (make_hard_regno_born): Rename from this...
921         (make_hard_regno_live): ... to this.  Remove update to conflict
922         information.  Remove now uneeded check_pic_pseudo_p argument.
923         Update function comment.
924         (make_hard_regno_dead): Add check_pic_pseudo_p argument and add update
925         to conflict information.  Update function comment.
926         (mark_pseudo_live): Remove update to conflict information.  Update
927         function comment.
928         (mark_pseudo_dead): Add conflict information update.
929         (mark_regno_live): Call make_hard_regno_live.
930         (mark_regno_dead): Call make_hard_regno_dead with new arguement.
931         (process_bb_lives): Call make_hard_regno_live and make_hard_regno_dead.
933 2018-09-29  H.J. Lu  <hongjiu.lu@intel.com>
935         PR target/87370
936         * config/i386/i386.c (construct_container): Use TImode for
937         BLKmode values in 2 integer registers.
939 2018-09-29  Jeff Law  <law@redhat.com>
941         * builtins.c (unterminated_array): Pass in c_strlen_data * to
942         c_strlen rather than just a tree *.
943         (c_strlen): Change NONSTR argument to a c_strlen_data pointer.
944         Update recursive calls appropriately.  If caller did not provide a
945         suitable data pointer, create a local one.  When a non-terminated
946         string is discovered, bubble up information about the string via the
947         c_strlen_data object.
948         * builtins.h (c_strlen): Update prototype.
949         (c_strlen_data): New structure.
950         * gimple-fold.c (get_range_strlen): Update calls to c_strlen.
951         For a type 2 call, if c_strlen indicates a non-terminated string
952         use the length of the non-terminated string.
953         (gimple_fold_builtin_stpcpy): Update calls to c_strlen.
955 2018-09-29  Jakub Jelinek  <jakub@redhat.com>
957         PR target/87467
958         * config/i386/avx512fintrin.h (_mm512_abs_pd, _mm512_mask_abs_pd): Use
959         __m512d type for __A argument rather than __m512.
961 2018-09-28  John David Anglin  <danglin@gcc.gnu.org>
963         * match.pd (simple_comparison): Don't optimize if either operand is
964         a function pointer when target needs function pointer canonicalization.
966 2018-09-28  Segher Boessenkool  <segher@kernel.crashing.org>
968         * config/rs6000/driver-rs6000.c (asm_names): Adjust the entries for
969         power5 .. power9 to remove indirection.
970         * config/rs6000/rs6000.h (ASM_CPU_POWER5_SPEC, ASM_CPU_POWER6_SPEC,
971         ASM_CPU_POWER7_SPEC, ASM_CPU_POWER8_SPEC, ASM_CPU_POWER9_SPEC,
972         ASM_CPU_476_SPEC): Delete.
973         (ASM_CPU_SPEC): Adjust.
974         (EXTRA_SPECS): Delete asm_cpu_power5, asm_cpu_power6, asm_cpu_power7,
975         asm_cpu_power8, asm_cpu_power9, asm_cpu_476.
977 2018-09-28  Segher Boessenkool  <segher@kernel.crashing.org>
979         * config.in: Delete HAVE_AS_DCI.
980         * config/powerpcspe/powerpcspe.h: Treat HAVE_AS_DCI as always true.
981         * config/rs6000/rs6000.h: Ditto.
982         * configure.ac: Delete HAVE_AS_DCI.
983         * configure: Regenerate.
985 2018-09-28  Segher Boessenkool  <segher@kernel.crashing.org>
987         * config.in (HAVE_AS_LWSYNC): Delete.
988         * config/powerpcspe/powerpcspe.h (TARGET_LWSYNC_INSTRUCTION): Delete.
989         * config/powerpcspe/sync.md (*lwsync): Always generate lwsync, never
990         do it as a .long .
991         * config/rs6000/rs6000.h (TARGET_LWSYNC_INSTRUCTION): Delete.
992         * config/rs6000/sync.md (*lwsync): Always generate lwsync, never do it
993         as a .long .
994         * configure.ac: Delete HAVE_AS_LWSYNC.
995         * configure: Regenerate.
997 2018-09-28  Eric Botcazou  <ebotcazou@adacore.com>
998             Pierre-Marie de Rodat  <derodat@adacore.com>
1000         * calls.c (expand_call): Try to do a tail call for thunks at -O0 too.
1001         * cgraph.h (struct cgraph_thunk_info): Add indirect_offset.
1002         (cgraph_node::create_thunk): Add indirect_offset parameter.
1003         (thunk_adjust): Likewise.
1004         * cgraph.c (cgraph_node::create_thunk): Add indirect_offset parameter
1005         and initialize the corresponding field with it.
1006         (cgraph_node::dump): Dump indirect_offset field.
1007         * cgraphclones.c (duplicate_thunk_for_node): Deal with indirect_offset.
1008         * cgraphunit.c (cgraph_node::analyze): Be prepared for external thunks.
1009         (thunk_adjust): Add indirect_offset parameter and deal with it.
1010         (cgraph_node::expand_thunk): Deal with the indirect_offset field and
1011         pass it to thunk_adjust.  Do not call the target hook if it's non-zero
1012         or if the thunk is external or local.  Fix formatting.  Do not chain
1013         the RESULT_DECL to BLOCK_VARS.  Pass the static chain to the target,
1014         if any, in the GIMPLE representation.
1015         * ipa-icf.c (sem_function::equals_wpa): Deal with indirect_offset.
1016         * lto-cgraph.c (lto_output_node): Write indirect_offset field.
1017         (input_node): Read indirect_offset field.
1018         * tree-inline.c (expand_call_inline): Pass indirect_offset field in the
1019         call to thunk_adjust.
1020         * tree-nested.c (struct nesting_info): Add thunk_p field.
1021         (create_nesting_tree): Set it.
1022         (convert_all_function_calls): Copy static chain from targets to thunks.
1023         (finalize_nesting_tree_1): Return early for thunks.
1024         (unnest_nesting_tree_1): Do not finalize thunks.
1025         (gimplify_all_functions): Do not gimplify thunks.
1027 2018-09-28  David Malcolm  <dmalcolm@redhat.com>
1029         * opt-suggestions.c (option_proposer::build_option_suggestions):
1030         Release "option_values".
1032 2018-09-28  David Malcolm  <dmalcolm@redhat.com>
1034         * coverage.c (get_coverage_counts): Convert problem-reporting dump
1035         messages from MSG_OPTIMIZED_LOCATIONS to MSG_MISSED_OPTIMIZATION.
1036         * dumpfile.c (kind_as_string): New function.
1037         (dump_loc): Rather than a hardcoded prefix of "note: ", use
1038         kind_as_string to vary the prefix based on dump_kind.
1039         (selftest::test_capture_of_dump_calls): Update for above.
1041 2018-09-28  Uros Bizjak  <ubizjak@gmail.com>
1043         * config/i386/i386.h (SSE_REGNO): Fix check for FIRST_REX_SSE_REG.
1044         (GET_SSE_REGNO): Rename from SSE_REGNO.  Update all uses for rename.
1046 2018-09-28  Uros Bizjak  <ubizjak@gmail.com>
1048         * config/i386/i386.h (CC_REGNO): Remove FPSR_REG.
1049         * config/i386/i386.c (ix86_fixed_condition_code_regs): Use
1050         INVALID_REGNUM instead of FPSR_REG.
1051         (ix86_md_asm_adjust): Do not clobber FPSR_REG.
1052         * config/i386/i386.md: Update comment of FP compares.
1053         (fldenv): Do not clobber FPSR_REG.
1055 2018-09-28  Richard Biener  <rguenther@suse.de>
1057         * tree.h (BLOCK_ORIGIN): New.
1058         * omp-expand.c (grid_expand_target_grid_body): Assign
1059         BLOCK_ORIGIN to BLOCK_ABSTRACT_ORIGIN.
1060         * tree-inline.c (remap_block): Likewise.
1061         * auto-profile.c (get_function_decl_from_block): Simplify
1062         by eliding the BLOCK_ABSTRACT_ORIGIN chasing.
1063         * langhooks.c (lhd_print_error_function): Likewise.
1064         * optinfo-emit-json.cc (optrecord_json_writer::inlining_chain_to):
1065         Likewise.
1066         * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
1067         * tree.c (block_nonartificial_location): Likewise.
1068         (block_ultimate_origin): Likewise.
1069         * tree-pretty-print.c (percent_K_format): Likewise.  Remove
1070         no longer needed LTO case.
1072 2018-09-28  Andrew Stubbs  <ams@codesourcery.com>
1073             Jan Hubicka  <jh@suse.cz>
1074             Martin Jambor  <mjambor@suse.cz>
1076         * simplify-rtx.c (simplify_merge_mask): New function.
1077         (simplify_ternary_operation): Use it, also see if VEC_MERGEs with the
1078         same masks are used in op1 or op2.
1079         (test_vec_merge): New function.
1080         (test_vector_ops): Call test_vec_merge.
1082 2018-09-28  Eric Botcazou  <ebotcazou@adacore.com>
1084         * config/sparc/sparc-protos.h (sparc_branch_cost): Declare.
1085         * config/sparc/sparc.h (BRANCH_COST): Call sparc_branch_cost.
1086         * config/sparc/sparc.c (struct processor_costs): Add branch_cost field.
1087         (cypress_costs): Set it.
1088         (supersparc_costs): Likewise.
1089         (hypersparc_costs): Likewise.
1090         (leon_cost): Likewise.
1091         (leon3_costs): Likewise.
1092         (sparclet_costs): Likewise.
1093         (ultrasparc_costs): Likewise.
1094         (ultrasparc_costs): Likewise.
1095         (niagara_costs): Likewise.
1096         (niagara2_costs): Likewise.
1097         (niagara3_costs): Likewise.
1098         (niagara4_costs): Likewise.
1099         (niagara7_costs): Likewise.
1100         (m8_costs): Likewise.
1101         (TARGET_CAN_FOLLOW_JUMP): Define.
1102         (pass_work_around_errata::gate): Minor tweak.
1103         (sparc_option_override): Remove MASK_FSMULD mask for V7 processors.
1104         Do not set both MASK_VIS4 and MASK_VIS4B for M8 processor.
1105         Automaitcally clear MASK_FSMULD mask for V7 processors.
1106         (sparc_can_follow_jump): New static function.
1107         (output_ubranch): Deal with CROSSING_JUMP_P.
1108         (sparc_use_sched_lookahead): Rewrite using switch statement.
1109         (sparc_issue_rate): Reorder.
1110         (sparc_branch_cost): New function.
1112 2018-09-27  Martin Sebor  <msebor@redhat.com>
1114         * tree.h (tree_to_shwi): Add attribute nonnull and pure.
1115         (tree_to_poly_int64, tree_to_uhwi, tree_to_poly_uint64): Same.
1116         (int_fits_type_p): Same.
1118 2018-09-27  Uros Bizjak  <ubizjak@gmail.com>
1120         * config/i386/i386.md (FPCR_REG): Remove.
1121         (UNSPEC_FLDCW): Remove.
1122         (x86_fnstcw_1): Use (const_int 0) instead of FPCR_REG.
1123         (x86_fldcw_1): Remove insn pattern.
1124         (fnstenv): Do not clobber FPCR_REG.
1125         (fldenv): Ditto.
1126         * config/i386/i386.h (FIXED_REGISTERS) Remove fpsr register.
1127         (CALL_USED_REGISTERS): Ditto.
1128         (REG_ALLOC_ORDER): Ditto.
1129         (REG_CLASS_CONTENTS): Ditto.
1130         (HI_REGISTER_NAMES): Ditto.
1131         (ADDITIONAL_REGISTER_NAMES): Use defines instead
1132         of numerical constants.
1133         * config/i386/i386.c (regclass_map): Remove fpsr register.
1134         (dbx_register_map): Ditto.
1135         (dbx64_register_map): Ditto.
1136         (svr4_dbx_register_map): Ditto.
1137         (print_reg): Do not handle FPCR_REG.
1139 2018-09-27  Segher Boessenkool  <segher@kernel.crashing.org>
1141         PR target/87149
1142         * config.in (HAVE_AS_CMPB, HAVE_AS_DFP, HAVE_AS_FPRND, HAVE_AS_MFPGPR,
1143         HAVE_AS_POPCNTB, HAVE_AS_POPCNTD, HAVE_AS_POWER8, HAVE_AS_POWER9):
1144         Delete, always treat as true.
1145         * config/powerpcspe/powerpcspe.c (rs6000_option_override_internal):
1146         Ditto.  Simplify remaining code.
1147         * config/powerpcspe/powerpcspe.h: Ditto.
1148         * config/rs6000/rs6000.c (rs6000_option_override_internal): Ditto.
1149         Simplify remaining code.
1150         (rs6000_expand_builtin): Ditto.
1151         * config/rs6000/rs6000.h: Ditto.
1152         * configure.ac: Ditto.
1153         * configure: Regenerate.
1155 2018-09-27  Martin Liska  <mliska@suse.cz>
1157         * coverage.c (get_coverage_counts): Revert the formatting
1158         of missing profile opt info.
1160 2018-09-27  Richard Biener  <rguenther@suse.de>
1162         PR debug/37801
1163         PR debug/87440
1164         * dwarf2out.c (set_block_origin_self): Do not mark outermost
1165         block as we do not output that.
1166         (gen_inlined_subroutine_die): Elide the originally outermost
1167         block, matching what we do for concrete instances.
1168         (decls_for_scope): Add parameter specifying whether to recurse
1169         to subblocks.
1171 2018-09-27  Andrew Stubbs  <ams@codesourcery.com>
1172             Tom de Vries  <tom@codesourcery.com>
1174         PR 82089
1176         * expmed.c (emit_cstore): Fix handling of result_mode == BImode and
1177         STORE_FLAG_VALUE == 1.
1179 2018-09-27  Andreas Krebbel  <krebbel@linux.ibm.com>
1181         * config/s390/s390.md (PPA_TX_ABORT, PPA_OOO_BARRIER): New
1182         constant definitions.
1183         ("tx_assist"): Replace magic number with PPA_TX_ABORT.
1184         ("*ppa"): Enable pattern also for -march=zEC12 -mno-htm.
1185         ("speculation_barrier"): New expander definition.
1187 2018-09-26  Indu Bhagat  <indu.bhagat@oracle.com>
1189         PR gcov-profile/86957
1190         * common.opt: New warning option -Wmissing-profile.
1191         * coverage.c (get_coverage_counts): Add warning for missing .gcda file.
1192         * doc/invoke.texi: Document -Wmissing-profile.
1194 2018-09-26  Jim Wilson  <jimw@sifive.com>
1196         * config/riscv/riscv.md (subsi3_extended2): Add J constraint.
1197         (negdi2, negsi2, negsi2_extended, negsi2_extended2): New.
1199 2018-09-26  Martin Sebor  <msebor@redhat.com>
1201         * tree.c (zerop): Change return type to bool.
1202         (integer_zerop, integer_onep, integer_each_onep): Same.
1203         (integer_all_onesp, integer_minus_onep, integer_pow2p): Same.
1204         (integer_nonzerop, integer_truep, tree_ctz, real_zerop): Same.
1205         (real_onep, real_minus_onep, chain_index): Same.
1206         (print_type_hash_statistics, type_list_equal): Same.
1207         * tree.h (zerop): Same.
1208         (zerop, integer_zerop, integer_onep, integer_each_onep): Same.
1209         (integer_all_onesp, integer_minus_onep, integer_pow2p): Same.
1210         (integer_nonzerop, integer_truep, tree_ctz, real_zerop): Same.
1211         (real_onep, real_minus_onep, chain_index): Same.
1212         (print_type_hash_statistics, type_list_equal): Same.
1214 2018-09-26  Jim Wilson  <jimw@sifive.com>
1216         * config/riscv/riscv.h (FUNCTION_ARG_REGNO_P): Fix comment.
1218 2018-09-26  Jakub Jelinek  <jakub@redhat.com>
1220         PR target/87414
1221         * config/i386/i386.c: Include debug.h and dwarf2out.h.
1222         (output_indirect_thunk): Emit DW_CFA_def_cfa_offset after the
1223         call.
1225 2018-09-25  Andrew Stubbs  <ams@codesourcery.com>
1227         * builtins.c (get_builtin_sync_mem): Force address mode conversion.
1229 2018-09-26  Uros Bizjak  <ubizjak@gmail.com>
1231         * config/i386/i386.h (enum reg_class): Remove FP_TOP_SSE_REGS
1232         and FP_SECOND_SSE_REGS.
1233         (REG_CLASS_NAMES): Ditto.
1234         (REG_CLASS_CONTENTS): Ditto.
1235         * config/i386/i386.c (ix86_preferred_reload_class) Do not handle
1236         FP_TOP_SSE_REGS and FP_SECOND_SSE_REGS classes.
1237         (ix86_preferred_output_reload_class): Ditto.
1238         * config/i386/i386.md (fix_trunc<mode>_i387_fisttp): Change "=&1f"
1239         clobber constraint to "=&f".
1240         (fix_truncdi_i387): Ditto.
1241         (lrintxfdi2): Ditto.
1242         (fistdi2_<rounding>): Ditto.
1243         (fpremxf4_i387): Change "=u" constraint to "=f".
1244         (fprem1xf4_i387): Ditto.
1245         (sincosxf3): Ditto.
1246         (fptanxf4_i387): Ditto.
1247         (fxtractxf3_i387): Ditto.
1248         (fscalexf4_i387): Ditto.
1249         (atan2xf3): Change "u" constraint to "f".
1250         (fyl2xxf3_i387): Ditto.
1251         (fyl2xp1xf3_i387): Ditto.
1253 2018-09-26  Uros Bizjak  <ubizjak@gmail.com>
1255         PR target/87439
1256         * config/i386/i386.h (NUM_MODES_FOR_MODE_SWITCHING): Update
1257         for removed I387_MASK_PM entity.
1260 2018-09-26  Jeff Law  <law@redhat.com>
1261         Revert
1262         2018-09-26  Alexey Neyman  <stilor@att.net>
1264         * graphite.h: Include <isl/id.h> and <isl/space.h>; these
1265         headers are no longer pulled in by <isl/val.h>.
1267 2018-09-26  Richard Biener  <rguenther@suse.de>
1269         PR debug/87443
1270         * dwarf2out.c (gen_lexical_block_die): Do not equate inline
1271         or concrete instance DIE to the tree.  Create abstract origin
1272         attributes also for concrete instances.
1274 2018-09-26  Alexey Neyman  <stilor@att.net>
1276         * graphite.h: Include <isl/id.h> and <isl/space.h>; these
1277         headers are no longer pulled in by <isl/val.h>.
1279 2018-09-26  Matthew Malcomson  <matthew.malcomson@arm.com>
1281         * config/arm/arm.c (arm_split_compare_and_swap, arm_split_atomic_op):
1282         Use new helper functions.
1283         * config/arm/sync.md (atomic_load<mode>, atomic_store<mode>):
1284         Use new helper functions.
1285         * config/arm/aarch-common-protos.h (aarch_mm_needs_acquire,
1286         aarch_mm_needs_release): New declarations.
1287         * config/arm/aarch-common.c (aarch_mm_needs_acquire,
1288         aarch_mm_needs_release): New.
1290 2018-09-26  Eric Botcazou  <ebotcazou@adacore.com>
1292         * config/arm/arm.c (arm_reorg): Skip Thumb reorg pass for thunks.
1293         (arm32_output_mi_thunk): Deal with long calls.
1295 2018-09-26  Richard Biener  <rguenther@suse.de>
1297         PR debug/87428
1298         PR debug/87362
1299         * tree-inline.c (expand_call_inline): When the location
1300         of the call is UNKNOWN_LOCATION use DECL_SOURCE_LOCATION
1301         or BUILTINS_LOCATION for the BLOCK_SOURCE_LOCATION of
1302         the inserted BLOCK to make inlined_function_outer_scope_p
1303         recognize it.
1304         * dwarf2out.c (add_call_src_coords_attributes): Do not add
1305         coords for reserved locations.
1307 2018-09-25  Segher Boessenkool  <segher@kernel.crashing.org>
1309         * config/rs6000/rs6000.md (*movcc_internal1): Use set_attr_alternative.
1310         (*call_indirect_nonlocal_sysv<mode>): Ditto.
1311         (*call_value_indirect_nonlocal_sysv<mode>): Ditto.
1312         (*sibcall_nonlocal_sysv<mode>): Ditto.
1313         (*sibcall_value_nonlocal_sysv<mode>): Ditto.
1314         (<bd>_<mode>): Ditto.
1315         (<bd>tf_<mode>): Ditto.
1317 2018-09-25  Segher Boessenkool  <segher@kernel.crashing.org>
1319         * config/rs6000/altivec.md (*altivec_mov<mode>): Write the output
1320         control string as a list of templates instead of as C code.
1321         (*altivec_movti): Ditto.
1322         * config/rs6000/darwin.md (movdf_low_di): Ditto.
1324 2018-09-25  Jim Wilson  <jimw@sifive.com>
1326         * config/riscv/riscv.c (riscv_split_symbol): Mark auipc label as weak
1327         when target symbol is weak.
1329 2018-09-25  Bernd Edlinger  <bernd.edlinger@hotmail.de>
1331         PR c/87387
1332         * builtins.c (unterminated_array): Simplify.
1333         * expr.c (string_constant): Handle SSA_NAME.  Add more exceptions
1334         where pointer arithmetic is safe.
1336 2018-09-25  Segher Boessenkool  <segher@kernel.crashing.org>
1338         PR target/86987
1339         * config/rs6000/altivec.md (altivec_vspltb): Use
1340         const_0_to_15_operand instead of u5bit_cint_operand.
1341         (*altivec_vspltb_internal): Ditto.
1342         (altivec_vspltb_direct): Ditto.
1343         (altivec_vsplth): Use const_0_to_7_operand instead of
1344         u5bit_cint_operand.
1345         (*altivec_vsplth_internal): Ditto.
1346         (altivec_vsplth_direct): Ditto.
1347         (altivec_vspltw): Use const_0_to_3_operand instead of
1348         u5bit_cint_operand.
1349         (*altivec_vspltw_internal): Ditto.
1350         (altivec_vspltw_direct): Ditto.
1351         (altivec_vspltsf): Ditto.
1352         (*altivec_vspltsf_internal): Ditto.
1353         * config/rs6000/rs6000.c (rs6000_expand_binop_builtin): Handle the
1354         various splats with the proper size immediate.  Reorder the various
1355         cases by ascending size of immediate, and put all such together.
1357 2018-09-25  Richard Biener  <rguenther@suse.de>
1359         PR debug/83941
1360         * dwarf2out.c (add_AT_external_die_ref): Remove now redundant
1361         GC-ification.
1362         (maybe_create_die_with_external_ref): Do not create
1363         DW_TAG_imported_unit here.
1364         (add_abstract_origin_attribute): Handle external BLOCK refs.
1365         (dwarf2out_abstract_function): Simplify LTO case.
1366         (dwarf2out_early_finish): Create DW_TAG_imported_unit explicitely
1367         rather than using maybe_create_die_with_external_ref.
1369 2018-09-25  Uros Bizjak  <ubizjak@gmail.com>
1371         PR target/71278
1372         * config/i386/i386.md (frndintxf2_mask_pm): Remove.
1373         (frndintxf2_mask_pm_i387): Ditto.
1374         (nearbyintxf2): Rewrite expander pattern to match rintxf2.
1375         Enable for !flag_trapping_math.
1376         (nearbyint<mode>2): Enable x87 modes for !flag_trapping_math.
1377         Enable SSE modes for TARGET_SSE4_1 and expand them with round insn.
1378         Change operand 1 predicate to nonimmediate_operand.
1379         (attr "i387_cw"): Remove mask_pm.
1380         * config/i386/i386.h (enum ix86_stack_slot): Remove SLOT_CW_MASK_PM.
1381         (enum ix86_entity): Remove I387_MASK_PM.
1382         * config/i386/i386.c (ix86_i387_mode_needed): Do not
1383         handle I387_MASK_PM.
1384         (ix86_mode_needed): Ditto.
1385         (ix86_mode_after): Ditto.
1386         (ix86_mode_entry): Ditto.
1387         (ix86_mode_exit): Ditto.
1388         (emit_i387_cw_initialization): Do not handle I387_CW_MASK_PM.
1390 2018-09-25  Jakub Jelinek  <jakub@redhat.com>
1392         * vr-values.c (vr_values::vr_values): Initialize to_remove_edges and
1393         to_update_switch_stmts to vNULL instead of calling create on them
1394         immediately.
1396 2018-09-25  Richard Biener  <rguenther@suse.de>
1398         PR tree-optimization/87402
1399         * tree-ssa-sccvn.c (SSA_VISITED): Remove unused function.
1400         (visit_phi): Re-instantiate handling of supposed to be VARYING
1401         but non-VARYING backedge value.
1403 2018-09-25  Richard Biener  <rguenther@suse.de>
1405         PR debug/83941
1406         * dwarf2out.c (struct sym_off_pair): New.
1407         (external_die_map): New global.
1408         (lookup_decl_die): When in LTO create DIEs lazily from the
1409         external_die_map.
1410         (lookup_block_die): New function, create DIEs lazily in LTO.
1411         (equate_block_to_die): New function.
1412         (dwarf2out_die_ref_for_decl): During WPA get the association
1413         from the external DIE map.
1414         (dwarf2out_register_external_die): Record mapping into the
1415         external DIE map.
1416         (maybe_create_die_with_external_ref): New function split out from
1417         DIE generation part of old dwarf2out_register_external_die.
1418         (add_abstract_origin_attribute): Do not return the DIE.  When
1419         in LTO reference externals directly.
1420         (dwarf2out_abstract_function): When in LTO ignore calls for
1421         decls with external DIEs (already present abstract instances).
1422         (gen_call_site_die): Adjust.
1423         (add_high_low_attributes): Likewise.
1424         (gen_lexical_block_die): Likewise.
1425         (gen_inlined_subroutine_die): Likewie.
1426         (gen_block_die): Likewise.
1427         (dwarf2out_inline_entry): Likewise.
1428         (dwarf2out_early_finish): In LTRANS phase create DW_TAG_imported_unit
1429         DIEs.
1431 2018-09-25  Martin Liska  <mliska@suse.cz>
1433         * ipa-fnsummary.c (estimate_node_size_and_time): Scale by two
1434         integers and not by a float value.
1436 2018-09-25  Martin Liska  <mliska@suse.cz>
1438         PR fortran/87394
1439         * dbgcnt.c (dbg_cnt_process_single_pair): Return false
1440         instead of NULL.
1441         * dumpfile.c (dump_enable_all): Remove extra parenthesis.
1442         * gcov-tool.c: Declare the function with ATTRIBUTE_NORETURN.
1443         * godump.c (go_format_type): Remove extra parenthesis.
1445 2018-09-25  Martin Liska  <mliska@suse.cz>
1447         * alias.c (set_dest_equal_p): Remove unused function.
1448         * config/i386/i386.c (def_builtin_pure2): Likewise.
1449         * diagnostic-show-locus.c (class layout): Remove
1450         unused field.
1451         (layout::layout): Likewise here.
1452         * dump-context.h (class temp_dump_context): Likewise.
1453         * dwarf2out.c (add_AT_fde_ref): Remove unused function.
1454         (add_AT_loclistsptr): Likewise.
1455         (add_AT_offset): Likewise.
1456         (get_AT_hi_pc): Likewise.
1457         (is_comdat_die): Likewise.
1458         (type_is_enum): Likewise.
1459         (ceiling): Likewise.
1460         (add_AT_vms_delta): Likewise.
1461         (is_class_die): Likewise.
1462         * edit-context.c (class line_event): Remove unused field.
1463         * graphite-sese-to-poly.c (tree_int_to_gmp): Remove
1464         unused function.
1465         * ipa-cp.c (ipa_get_vr_lat): Likewise.
1466         * lra-constraints.c (ok_for_index_p_nonstrict): Likewise.
1467         (ok_for_base_p_nonstrict): Likewise.
1468         * tree-chrec.c (is_not_constant_evolution): Likewise.
1469         (chrec_fold_poly_cst): Likewise.
1470         * tree-if-conv.c (has_pred_critical_p): Likewise.
1471         * tree-ssa-coalesce.c (print_exprs): Likewise.
1472         * tree-ssa-pre.c (bitmap_set_contains_expr): Likewise.
1473         * tree-ssa-uninit.c (is_and_or_or_p): Likewise.
1474         * tree-vrp.c (value_ranges_intersect_p): Likewise.
1475         (value_range_nonnegative_p): Likewise.
1477 2018-09-25  Martin Liska  <mliska@suse.cz>
1479         * config/powerpcspe/powerpcspe.c (rs6000_output_function_epilogue):
1480         Do not handle "GNU Pascal".
1481         * config/rs6000/rs6000.c (rs6000_output_function_epilogue):
1482         Likewise.
1483         * config/sparc/sparc.c (sparc_pass_by_reference): Remove Pascal
1484         from documentation. Likewise.
1485         * dbxout.c (dbxout_range_type): Likewise.
1486         * doc/cpp.texi: Likewise.
1487         * doc/extend.texi: Likewise.
1488         * doc/frontends.texi: Likewise.
1489         * doc/invoke.texi: Remove Pascal entry.
1490         * tree.def (CLEANUP_POINT_EXPR): Likewise.
1491         * doc/rtl.texi (MODE_FUNCTION): Remove not used entry.
1493 2018-09-25  Martin Liska  <mliska@suse.cz>
1495         PR middle-end/86078
1496         * doc/invoke.texi: Document all parameters and remove default
1497         of the parameters.
1499 2018-09-25  Ilya Leoshkevich  <iii@linux.ibm.com>
1501         PR bootstrap/87417
1502         * rtl.c (rtx_code_size): Take into account that EQ_ATTR_ALT
1503         contains HOST_WIDE_INTs when computing its size.
1505 2018-09-24  Jim Wilson  <jimw@sifive.com>
1507         PR target/87391
1508         * config/riscv/riscv.h (STACK_BOUNDARY): Test riscv_abi == ABI_ILP32E
1509         not TARGET_RVE.
1510         (ABI_STACK_BOUNDARY, MAX_ARGS_IN_REGISTERS): Likewise.
1512 2018-09-24  Andrew Pinski  <apinski@marvell.com>
1514         *  config/aarch64/aarch64.c (aarch_macro_fusion_pair_p): Don't
1515         access prev before checking it for NULLness in the
1516         AARCH64_FUSE_CMP_BRANCH case.
1518 2018-09-24  H.J. Lu  <hongjiu.lu@intel.com>
1520         PR target/82699
1521         * config/i386/i386.c (rest_of_insert_endbranch): Set
1522         endbr_queued_at_entrance to true and don't insert ENDBR if
1523         x86_function_profiler will be called.
1524         (x86_function_profiler): Insert ENDBR if endbr_queued_at_entrance
1525         is true.
1526         * config/i386/i386.h (machine_function): Add
1527         endbr_queued_at_entrance.
1529 2018-09-24  Ilya Leoshkevich  <iii@linux.ibm.com>
1531         * genattrtab.c (mk_attr_alt): Use alternative_mask.
1532         (attr_rtx_1): Adjust caching to match the new EQ_ATTR_ALT field
1533         types.
1534         (check_attr_test): Use alternative_mask.
1535         (get_attr_value): Likewise.
1536         (compute_alternative_mask): Use alternative_mask and XWINT.
1537         (make_alternative_compare): Use alternative_mask.
1538         (attr_alt_subset_p): Use XWINT.
1539         (attr_alt_subset_of_compl_p): Likewise.
1540         (attr_alt_intersection): Use alternative_mask and XWINT.
1541         (attr_alt_union): Likewise.
1542         (attr_alt_complement): Use HOST_WIDE_INT and XWINT.
1543         (mk_attr_alt): Use alternative_mask and HOST_WIDE_INT.
1544         (simplify_test_exp): Use alternative_mask and XWINT.
1545         (write_test_expr): Use alternative_mask and XWINT, adjust bit
1546         number calculation to support 64 bits.  Generate code that
1547         checks 64-bit masks.
1548         (main): Use alternative_mask.
1549         * rtl.def (EQ_ATTR_ALT): Change field types from ii to ww.
1551 2018-09-24  Ilya Leoshkevich  <iii@linux.ibm.com>
1553         PR target/80080
1554         * config/s390/s390.c (s390_emit_epilogue): Do not use PARALLEL
1555         RETURN+USE when returning via %r14.
1557 2018-09-24  Martin Liska  <mliska@suse.cz>
1559         * gcov.c (output_lines): Print colorization legend
1560         for both flag_use_colors and flag_use_hotness_colors.
1561         Reword the help.
1563 2018-09-24  Martin Liska  <mliska@suse.cz>
1565         * coverage.c (get_coverage_counts): Use warning_at
1566         with current_function_decl location. Use %qD in warning
1567         message.
1569 2018-09-24  Martin Liska  <mliska@suse.cz>
1571         * memory-block.h (memory_block_pool::release): Annotate with
1572         valgrind that the memory is not accessible.
1574 2018-09-24  Martin Liska  <mliska@suse.cz>
1576         PR sanitizer/85774
1577         * asan.c: Make asan_handled_variables extern.
1578         * asan.h: Likewise.
1579         * cfgexpand.c (expand_stack_vars): Make sure
1580         a representative is unpoison if another
1581         variable in the partition is handled by
1582         use-after-scope sanitization.
1584 2018-09-24  Richard Biener  <rguenther@suse.de>
1586         PR tree-optimization/63155
1587         * tree-ssa-propagate.c (add_ssa_edge): Avoid adding PHIs to
1588         the worklist when the edge of the respective argument isn't
1589         executable.
1591 2018-09-23   Uros Bizjak  <ubizjak@gmail.com>
1593         * config/i386/i386.h (enum reg_class): Rename MASK_REGS to
1594         ALL_MASK_REGS and MASK_EVEX_REGS to MASK_REGS.
1595         (MASK_CLASS_P): Update for rename.
1596         (MAYBE_MASK_CLASS_P): Ditto.
1597         (REG_CLASS_NAMES): Update.
1598         (REG_CLASS_CONTENT): Update.
1599         * config/i386/i386.c (regclass_map): Update for MASK_REG
1600         and ALL_MASK_REGS rename.
1601         * config/i386/constraints.md (Yk): Update for rename.
1602         (k): Ditto.
1604 2018-09-23   Uros Bizjak  <ubizjak@gmail.com>
1606         * config/i386/i386.h (enum reg_class): Remove
1607         EVEX_SSE_REGS and MOD4_SSE_REGS.
1608         (REG_CLASS_NAMES): Update.
1609         (REG_CLASS_CONTENT): Update.
1610         * config/i386/i386.c (regclass_map): Declare AVX-512 SSE
1611         registers as ALL_SSE_REGS.
1612         (ix86_additional_allocno_class_p): Remove.
1613         (TARGET_ADDITIONAL_ALLOCNO_CLASS_P): Remove.
1614         (ix86_register_priority): Lower priority of EVEX SSE registers.
1615         Use IN_RANGE macro where appropriate.
1616         (ix86_hard_regno_mode_ok): Merge AVX-5124FMAPS and
1617         AVX-5124VNNIW checks.
1618         (ix86_modes_tieable_p): Tie 512-bit SSE modes.
1619         * config/i386/sse.md (avx5124fmaddps_4fmaddps)
1620         (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddps_maskz)
1621         (avx5124fmaddps_4fmaddss, avx5124fmaddps_4fmaddss_mask)
1622         (avx5124fmaddps_4fmaddss_maskz, avx5124fmaddps_4fnmaddps)
1623         (avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddps_maskz)
1624         (avx5124fmaddps_4fnmaddss, avx5124fmaddps_4fnmaddss_mask)
1625         (avx5124fmaddps_4fnmaddss_maskz, avx5124vnniw_vp4dpwssd)
1626         (avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssd_maskz)
1627         (avx5124vnniw_vp4dpwssds, avx5124vnniw_vp4dpwssds_mask)
1628         (avx5124vnniw_vp4dpwssds_maskz): Use "v" instead of "Yh" constraint.
1629         * config/i386/constraints.md (Yh): Remove.
1631 2018-09-23   Uros Bizjak  <ubizjak@gmail.com>
1633         * config/i386/i386.c (regclass_map): Declare integer REX registers
1634         as GENERAL_REGS.
1636 2018-09-23  Gerald Pfeifer  <gerald@pfeifer.com>
1638         * doc/service.texi (Service): Switch the fsf.org link to https.
1640 2018-09-22  Chung-Ju Wu  <jasonwucj@gmail.com>
1642         PR target/86798
1643         * config/nds32/nds32.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
1644         Define to speculation_safe_value_not_needed.
1646 2018-09-21  Florian Weimer  <fweimer@redhat.com>
1648         PR middle-end/81035
1649         * doc/extend.texi (Common Function Attributes): Mention that
1650         noreturn suppresses tail call optimization.
1652 2018-09-21  Jeff Law  <law@redhat.com>
1654         * gimple-ssa-evrp.c (evrp_dom_walker::cleanup): Call
1655         vr_values::cleanup_edges_and_switches.
1656         * tree-vrp.c (to_remove_edges, to_update_switch_stmts): Moved into
1657         vr_values class.
1658         (identify_jump_threads): Remove EDGE_IGNORE handling.
1659         (execute_vrp): Move handling of to_remove_edges and
1660         to_update_switch_stmts into vr_values class member functions.
1661         * tree-vrp.h (switch_update, to_remove_edges): Remove declarations.
1662         (to_update_switch_stmts): Likewise.
1663         * vr-values.c: Include cfghooks.h.
1664         (vr_values::vr_values): Initialize to_remove_edges and
1665         to_update_switch_stmts.
1666         (vr_values::~vr_values): Verify to_remove_edges and
1667         to_update_switch_stmts are empty.
1668         (vr_values::simplify_switch_using_ranges): Set EDGE_IGNORE as needed.
1669         (vr_values::cleanup_edges_and_switches): New member function.
1670         * vr-values.h (vr_values): Add cleanup_edges_and_switches member
1671         function.  Add new data members.
1673 2018-09-21  David Malcolm  <dmalcolm@redhat.com>
1675         PR tree-optimization/87309
1676         * dumpfile.c (dump_context::begin_scope): Filter the dump_loc
1677         calls with pflags and alt_flags.
1678         (selftest::test_capture_of_dump_calls): Add test of interaction of
1679         MSG_OPTIMIZED_LOCATIONS with AUTO_DUMP_SCOPE.
1681 2018-09-21  Olivier Hainque  <hainque@adacore.com>
1683         * config.gcc: Factorize and comment inclusion of vxworks-dummy.h.
1685 2018-09-21  Olivier Hainque  <hainque@adacore.com>
1687         * config/vxworks.h (CLEAR_INSN_CACHE): #define to 1.
1689 2018-09-21  Olivier Hainque  <hainque@adacore.com>
1691         * config/vxworks.h (VXWORKS_LIBGCC_SPEC): Remove -lc_internal.
1692         Merge block comment with the one ahead of VXWORKS_LIBS_RTP. Then:
1693         (VXWORKS_LIBS_RTP): Minor reordering.
1695 2018-09-21  Olivier Hainque  <hainque@adacore.com>
1697         * config/vxworks.h (STARTFILE_PREFIX_SPEC): Define.
1698         (VXWORKS_LIBS_DIR_RTP): Remove definition and use.
1700 2018-09-21  Olivier Hainque  <hainque@adacore.com>
1702         * config/vxworks.h (SIZE_TYPE): Account for TARGET_VXWORKS64.
1703         (PTRDIFF_TYPE): Likewise.
1705 2018-09-21  Olivier Hainque  <hainque@adacore.com>
1707         * config.gcc: Enforce def of TARGET_VXWORKS64 to 1 from
1708         triplet, similar to support for VxWorks7.
1709         * config/vxworks-dummy.h: Provide a default definition
1710         of TARGET_VXWORKS64 to 0.
1712 2018-09-21  Olivier Hainque  <hainque@adacore.com>
1714         * config/vxworks.h (TARGET_VXWORKS7): Move default definition ...
1715         * config/vxworks-dummy.h: here.
1717 2018-09-21  Olivier Hainque  <hainque@adacore.com>
1719         * config.gcc: Prepend vxworks-dummy.h to tm_file for powerpc*
1721 2018-09-21  Shaokun Zhang  <zhangshaokun@hisilicon.com>
1722             Bo Zhou  <zbo.zhou@hisilicon.com>
1724         * config/aarch64/aarch64-cores.def (tsv110): New CPU.
1725         * config/aarch64/aarch64-tune.md: Regenerated.
1726         * doc/invoke.texi (AArch64 Options/-mtune): Add "tsv110".
1727         * config/aarch64/aarch64.c (tsv110_tunings): New tuning table.
1728         * config/aarch64/aarch64-cost-tables.h: Add "tsv110" extra costs.
1730 2018-09-21  Andrew Stubbs  <ams@codesourcery.com>
1731             Julian Brown  <julian@codesourcery.com>
1733         * builtins.c (get_builtin_sync_mem): Handle address spaces.
1735 2018-09-21  Eric Botcazou  <ebotcazou@adacore.com>
1737         * config/rs6000/rs6000.c (rs6000_function_ok_for_sibcall): Return false
1738         if the call takes a static chain.
1740 2018-09-21  Martin Liska  <mliska@suse.cz>
1742         * auto-profile.c (autofdo_source_profile::read): Do not
1743         set sum_all.
1744         (read_profile): Do not add working sets.
1745         (read_autofdo_file): Remove sum_all.
1746         (afdo_callsite_hot_enough_for_early_inline): Remove const
1747         qualifier.
1748         * coverage.c (struct counts_entry): Remove gcov_summary.
1749         (read_counts_file): Read new GCOV_TAG_OBJECT_SUMMARY,
1750         do not support GCOV_TAG_PROGRAM_SUMMARY.
1751         (get_coverage_counts): Remove summary and expected
1752         arguments.
1753         * coverage.h (get_coverage_counts): Likewise.
1754         * doc/gcov-dump.texi: Remove -w option.
1755         * gcov-dump.c (dump_working_sets): Remove.
1756         (main): Do not support '-w' option.
1757         (print_usage): Likewise.
1758         (tag_summary): Likewise.
1759         * gcov-io.c (gcov_write_summary): Do not dump
1760         histogram.
1761         (gcov_read_summary): Likewise.
1762         (gcov_histo_index): Remove.
1763         (gcov_histogram_merge): Likewise.
1764         (compute_working_sets): Likewise.
1765         * gcov-io.h (GCOV_TAG_OBJECT_SUMMARY): Mark
1766         it not obsolete.
1767         (GCOV_TAG_PROGRAM_SUMMARY): Mark it obsolete.
1768         (GCOV_TAG_SUMMARY_LENGTH): Adjust.
1769         (GCOV_HISTOGRAM_SIZE): Remove.
1770         (GCOV_HISTOGRAM_BITVECTOR_SIZE): Likewise.
1771         (struct gcov_summary): Simplify rapidly just
1772         to runs and sum_max fields.
1773         (gcov_histo_index): Remove.
1774         (NUM_GCOV_WORKING_SETS): Likewise.
1775         (compute_working_sets): Likewise.
1776         * gcov-tool.c (print_overlap_usage_message): Remove
1777         trailing empty line.
1778         * gcov.c (read_count_file): Read GCOV_TAG_OBJECT_SUMMARY.
1779         (output_lines): Remove program related line.
1780         * ipa-profile.c (ipa_profile): Do not consider GCOV histogram.
1781         * lto-cgraph.c (output_profile_summary): Do not stream GCOV
1782         histogram.
1783         (input_profile_summary): Do not read it.
1784         (merge_profile_summaries): And do not merge it.
1785         (input_symtab): Do not call removed function.
1786         * modulo-sched.c (sms_schedule): Do not print sum_max.
1787         * params.def (HOT_BB_COUNT_FRACTION): Reincarnate param that was
1788         removed when histogram method was invented.
1789         (HOT_BB_COUNT_WS_PERMILLE): Mention that it's used only in LTO
1790         mode.
1791         * postreload-gcse.c (eliminate_partially_redundant_load): Fix
1792         GCOV coding style.
1793         * predict.c (get_hot_bb_threshold): Use HOT_BB_COUNT_FRACTION
1794         and dump selected value.
1795         * profile.c (add_working_set): Remove.
1796         (get_working_sets): Likewise.
1797         (find_working_set): Likewise.
1798         (get_exec_counts): Do not work with working sets.
1799         (read_profile_edge_counts): Do not inform as sum_max is removed.
1800         (compute_branch_probabilities): Likewise.
1801         (compute_value_histograms): Remove argument for call of
1802         get_coverage_counts.
1803         * profile.h: Do not make gcov_summary const.
1805 2018-09-21  Monk Chiang  <sh.chiang04@gmail.com>
1807         * config.gcc (nds32*-*-*): Set TARGET_DEFAULT_TLSDESC_TRAMPOLINE=0.
1809 2018-09-21  Eric Botcazou  <ebotcazou@adacore.com>
1811         PR tree-optimization/86990
1812         * gimple-ssa-store-merging.c (imm_store_chain_info:coalesce_immediate):
1813         Check that the entire merged store group is made of constants only for
1814         overlapping stores.
1816 2018-09-20  Allan Sandfeld Jensen  <allan.jensen@qt.io>
1818         * gcc.c (LINK_COMMAND_SPEC): Handle -r like -nostdlib.
1819         (VTABLE_VERIFICATION_SPEC): Likewise.
1820         (SANITIZER_EARLY_SPEC): Likewise.
1821         (SANITIZER_SPEC): Likewise.
1822         * config/darwin.h (LINK_COMMAND_SPEC): Likewise.
1823         * doc/invoke.texi (Link Options): Document -r.
1825 2018-09-20  Richard Biener <rguenther@suse.de>
1827         PR middle-end/87054
1828         * gimplify.c (gimplify_expr): Retain alignment of
1829         addressable lvalue in dereference.
1831 2018-09-20  Alexandre Oliva <aoliva@redhat.com>
1833         PR bootstrap/87013
1834         * configure.ac: Check for .loc is_stmt support.
1835         * configure, config.in: Rebuilt.
1836         * dwarf2out.c (dwarf2out_source_line): Skip is_stmt
1837         if not supported.
1839 2018-09-20  Segher Boessenkool  <segher@kernel.crashing.org>
1841         * config/rs6000/rs6000.opt (misel=no, misel=yes): Delete.
1842         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -misel=yes and
1843         -misel=no.
1845 2018-09-20  Segher Boessenkool  <segher@kernel.crashing.org>
1847         * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
1848         VECTOR_OTHER.
1849         * config/rs6000/rs6000.c (rs6000_debug_vector_unit): Delete
1850         case VECTOR_OTHER.
1852 2018-09-20  Marek Polacek  <polacek@redhat.com>
1854         * doc/invoke.texi: Add -Wno-init-list-lifetime to C++ Language Options.
1856 2018-09-20  Richard Sandiford  <richard.sandiford@arm.com>
1858         PR tree-optimization/87288
1859         * tree-vect-loop.c (vect_analyze_loop_2): Take PEELING_FOR_GAPS
1860         into account when determining PEELING_FOR_NITERS.
1862 2018-09-20  Richard Sandiford  <richard.sandiford@arm.com>
1864         PR tree-optimization/86877
1865         * tree-vect-loop.c (vect_analyze_loop_2): Call
1866         vect_verify_datarefs_alignment.
1868 2018-09-19  Marek Polacek  <polacek@redhat.com>
1870         * doc/invoke.texi: Document -Wclass-conversion.
1872 2018-09-19  John David Anglin  <danglin@gcc.gnu.org>
1874         * config/pa/pa.c (pa_adjust_priority): Delete.
1875         (TARGET_SCHED_ADJUST_PRIORITY): Delete define.
1877         * config/pa/pa.md (atomic_storeqi): Restore deleted expander.
1878         (atomic_storehi): Likewise.
1879         (atomic_storesi): Likewise.
1880         (atomic_loaddi): Restore compare and swap exchange loop code.
1882 2018-09-19  Segher Boessenkool  <segher@kernel.crashing.org>
1884         PR rtl-optimization/86902
1885         * combine.c (try_combine): When changing the CC mode used, don't change
1886         an unrelated mode in other_insn to that new CC mode.
1888 2018-09-19  David Malcolm  <dmalcolm@redhat.com>
1890         * tree-data-ref.c (runtime_alias_check_p): Use formatted printing
1891         with %T in place of calls to dump_generic_expr.
1892         (prune_runtime_alias_test_list): Likewise.
1893         (create_runtime_alias_checks): Likewise.
1894         * tree-vect-data-refs.c (vect_check_nonzero_value): Likewise.
1895         (vect_analyze_data_ref_dependence): Likewise.
1896         (vect_slp_analyze_data_ref_dependence): Likewise.
1897         (vect_record_base_alignment): Likewise.  Use %G in place of call
1898         to dump_gimple_stmt.
1899         (vect_compute_data_ref_alignment): Likewise.
1900         (verify_data_ref_alignment): Likewise.
1901         (vect_find_same_alignment_drs): Likewise.
1902         (vect_analyze_group_access_1): Likewise.
1903         (vect_analyze_data_ref_accesses): Likewise.
1904         (dependence_distance_ge_vf): Likewise.
1905         (dump_lower_bound): Likewise.
1906         (vect_prune_runtime_alias_test_list): Likewise.
1907         (vect_find_stmt_data_reference): Likewise.
1908         (vect_analyze_data_refs): Likewise.
1909         (vect_create_addr_base_for_vector_ref): Likewise.
1910         (vect_create_data_ref_ptr): Likewise.
1911         * tree-vect-loop-manip.c (vect_set_loop_condition): Likewise.
1912         (vect_can_advance_ivs_p): Likewise.
1913         (vect_update_ivs_after_vectorizer): Likewise.
1914         (vect_gen_prolog_loop_niters): Likewise.
1915         (vect_prepare_for_masked_peels): Likewise.
1916         * tree-vect-loop.c (vect_determine_vf_for_stmt): Likewise.
1917         (vect_determine_vectorization_factor): Likewise.
1918         (vect_is_simple_iv_evolution): Likewise.
1919         (vect_analyze_scalar_cycles_1): Likewise.
1920         (vect_analyze_loop_operations): Likewise.
1921         (report_vect_op): Likewise.
1922         (vect_is_slp_reduction): Likewise.
1923         (check_reduction_path): Likewise.
1924         (vect_is_simple_reduction): Likewise.
1925         (vect_create_epilog_for_reduction): Likewise.
1926         (vect_finalize_reduction:): Likewise.
1927         (vectorizable_induction): Likewise.
1928         (vect_transform_loop_stmt): Likewise.
1929         (vect_transform_loop): Likewise.
1930         (optimize_mask_stores): Likewise.
1931         * tree-vect-patterns.c (vect_pattern_detected): Likewise.
1932         (vect_split_statement): Likewise.
1933         (vect_recog_over_widening_pattern): Likewise.
1934         (vect_recog_average_pattern): Likewise.
1935         (vect_determine_min_output_precision_1): Likewise.
1936         (vect_determine_precisions_from_range): Likewise.
1937         (vect_determine_precisions_from_users): Likewise.
1938         (vect_mark_pattern_stmts): Likewise.
1939         (vect_pattern_recog_1): Likewise.
1940         * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
1941         (vect_record_max_nunits): Likewise.
1942         (vect_build_slp_tree_1): Likewise.
1943         (vect_build_slp_tree_2): Likewise.
1944         (vect_print_slp_tree): Likewise.
1945         (vect_analyze_slp_instance): Likewise.
1946         (vect_detect_hybrid_slp_stmts): Likewise.
1947         (vect_detect_hybrid_slp_1): Likewise.
1948         (vect_slp_analyze_operations): Likewise.
1949         (vect_slp_analyze_bb_1): Likewise.
1950         (vect_transform_slp_perm_load): Likewise.
1951         (vect_schedule_slp_instance): Likewise.
1952         * tree-vect-stmts.c (vect_mark_relevant): Likewise.
1953         (vect_mark_stmts_to_be_vectorized): Likewise.
1954         (vect_init_vector_1): Likewise.
1955         (vect_get_vec_def_for_operand): Likewise.
1956         (vect_finish_stmt_generation_1): Likewise.
1957         (vect_check_load_store_mask): Likewise.
1958         (vectorizable_call): Likewise.
1959         (vectorizable_conversion): Likewise.
1960         (vectorizable_operation): Likewise.
1961         (vectorizable_load): Likewise.
1962         (vect_analyze_stmt): Likewise.
1963         (vect_is_simple_use): Likewise.
1964         (vect_get_vector_types_for_stmt): Likewise.
1965         (vect_get_mask_type_for_stmt): Likewise.
1966         * tree-vectorizer.c (increase_alignment): Likewise.
1968 2018-09-19  Andrew Stubbs  <ams@codesourcery.com>
1970         * doc/rtl.texi: Adjust vec_select description.
1971         * simplify-rtx.c (simplify_binary_operation_1): Allow VEC_SELECT to use
1972         non-constant selectors.
1974 2018-09-19  Matthew Malcomson  <matthew.malcomson@arm.com>
1976         * config/aarch64/aarch64-protos.h
1977         (aarch64_offset_9bit_signed_unscaled_p): New declaration.
1978         * config/aarch64/aarch64.md (arches): New "rcpc8_4" attribute value.
1979         (arch_enabled): Add check for "rcpc8_4" attribute value of "arch".
1980         * config/aarch64/aarch64.h (AARCH64_FL_RCPC8_4): New bitfield.
1981         (AARCH64_FL_FOR_ARCH8_4): Include AARCH64_FL_RCPC8_4.
1982         (AARCH64_FL_PROFILE): Move index so flags are ordered.
1983         (AARCH64_ISA_RCPC8_4): New flag.
1984         * config/aarch64/aarch64.c (offset_9bit_signed_unscaled_p): Renamed
1985         to aarch64_offset_9bit_signed_unscaled_p.
1986         * config/aarch64/atomics.md (atomic_store<mode>): Allow offset
1987         and use stlur.
1988         * config/aarch64/constraints.md (Ust): New constraint.
1989         * config/aarch64/predicates.md.
1990         (aarch64_9bit_offset_memory_operand): New predicate.
1991         (aarch64_rcpc_memory_operand): New predicate.
1993 2018-09-19  Eric Botcazou  <ebotcazou@adacore.com>
1995         PR rtl-optimization/87361
1996         * rtlanal.c (nonzero_bits1): Revert accidental change.
1998 2018-09-19  Richard Biener  <rguenther@suse.de>
2000         PR tree-optimization/87349
2001         PR tree-optimization/87342
2002         * tree-ssa-sccvn.c (do_rpo_vn): Iterate max_rpo computation.
2004 2018-09-18  Marek Polacek  <polacek@redhat.com>
2006         P1064R0 - Allowing Virtual Function Calls in Constant Expressions
2007         * gimple-fold.c (gimple_get_virt_method_for_vtable): Adjust assert.
2009 2018-09-18  Segher Boessenkool  <segher@kernel.crashing.org>
2011         * config/rs6000/rs6000.md: Remove old "Cygnus sibcall" comment.
2013 2018-09-18  Segher Boessenkool  <segher@kernel.crashing.org>
2015         PR rtl-optimization/86882
2016         * rtlanal.c (reg_overlap_mentioned_p): Handle CLOBBER.
2018 2018-09-18  Uros Bizjak  <ubizjak@gmail.com>
2020         * config/i386/i386.md (*<code>extend<mode>xf2): Macroize insn from
2021         *<code>extendsfxf2 and *<code>extenddfxf2 using MODEF mode iterator.
2023 2018-09-18  Jonathan Wakely  <jwakely@redhat.com>
2025         PR other/87353
2026         * doc/invoke.texi (Link Options): Fix formatting and grammar.
2028 2018-09-18  Richard Biener  <rguenther@suse.de>
2030         PR middle-end/63155
2031         * tree-ssa-coalesce.c (tree_int_map_hasher): Remove.
2032         (compute_samebase_partition_bases): Likewise.
2033         (coalesce_ssa_name): Always use compute_optimized_partition_bases.
2034         (gimple_can_coalesce_p): Simplify.
2036 2018-09-18  Hans-Peter Nilsson  <hp@bitrange.com>
2038         Handle a library implementation of ffs calling __builtin_ffs.
2039         * config/mmix/mmix.c (TARGET_INIT_LIBFUNCS): Override with...
2040         (mmix_init_libfuncs): New function: make __builtin_ffs expand
2041         to __ffsdi2.
2043 2018-09-17  David Malcolm  <dmalcolm@redhat.com>
2045         * diagnostic-show-locus.c (class layout_range): Add field
2046         "m_original_idx".
2047         (layout_range::layout_range): Add "original_idx" param and use it
2048         to initialize new field.
2049         (make_range): Use 0 for original_idx.
2050         (layout::layout): Pass in index to calls to
2051         maybe_add_location_range.
2052         (layout::maybe_add_location_range): Add param "original_idx" and
2053         pass it on to layout_range.
2054         (layout::print_any_labels): Pass on range->m_original_idx to
2055         get_text call.
2056         (gcc_rich_location::add_location_if_nearby): Use 0 for
2057         original_idx.
2058         * gcc-rich-location.h (text_range_label::get_text): Update for new
2059         param.
2060         (range_label_for_type_mismatch::get_text): Likewise.
2062 2018-09-17  Uros Bizjak  <ubizjak@gmail.com>
2064         * config/i386/i386.c (ix86_emit_i387_log1p): Emit fldln2 earlier.
2066 2018-09-17  David Malcolm  <dmalcolm@redhat.com>
2068         * gimple-ssa-sprintf.c (fmtwarn): Update for introduction of
2069         format_string_diagnostic_t.
2070         (fmtwarn_n): Likewise.
2071         * substring-locations.c
2072         (format_string_diagnostic_t::format_string_diagnostic_t) New ctor.
2073         (format_warning_n_va): Convert to...
2074         (format_string_diagnostic_t::emit_warning_n_va): ...this.
2075         (format_warning_va): Convert to...
2076         (format_string_diagnostic_t::emit_warning_va): ...this.
2077         (format_warning_at_substring): Convert to...
2078         (format_string_diagnostic_t::emit_warning): ...this.
2079         (format_warning_at_substring_n): Convert to...
2080         (format_string_diagnostic_t::emit_warning_n): ...this.
2081         * substring-locations.h (class format_string_diagnostic_t): New
2082         class.
2083         (format_warning_va): Convert to
2084         format_string_diagnostic_t::emit_warning_va.
2085         (format_warning_n_va): Convert to
2086         format_string_diagnostic_t::emit_warning_n_va.
2087         (format_warning_at_substring): Convert to
2088         format_string_diagnostic_t::emit_warning.
2089         (format_warning_at_substring_n): Convert to
2090         format_string_diagnostic_t::emit_warning_n.
2092 2018-09-17  Cesar Philippidis  <cesar@codesourcery.com>
2093             Bernd Schmidt <bernds_cb1@t-online.de>
2095         * config/nvptx/nvptx.md (atomic_fetch_<logic><mode>): Enable with
2096         SImode args.
2098 2018-09-17  Uros Bizjak  <ubizjak@gmail.com>
2100         * config/i386/i386.md (truncxf<mode>2_i387_noop_unspec): Change
2101         operand 0 predicate to nonimmediate operand.
2102         (rint<mode>2_frndint): Remove insn pattern.
2103         (rint<mode>2): Change operand 1 predicate to general_operand.
2104         Extend operand 1 to XFmode and generate rintxf2 insn.
2105         (frndintxf2_<rounding>): Rename from frndint<mode>2_<rounding>.
2106         Do not use X87MODEF mode macro.
2107         (frndintxf2_<rounding>_i387): Rename from
2108         frndint<mode>2_<rounding>_i387.  Do not use X87MODEF mode macro.
2109         (<rounding_insn><mode>2): For non-SSE modes, extend operand 1
2110         to XFmode and generate significandxf3 insn.
2112 2018-09-17  Richard Biener  <rguenther@suse.de>
2114         PR tree-optimization/87328
2115         * tree-ssa-sccvn.c (process_bb): Remove assertion about not
2116         visiting unexecutable backedges when not iterating.
2117         (do_rpo_vn): Mark all edges not executable even when not
2118         iterating.
2120 2018-09-17  Martin Jambor  <mjambor@suse.cz>
2122         PR c/63886
2123         * doc/invoke.texi (Warning Options): Likewise.
2125 2018-09-17  Richard Biener  <rguenther@suse.de>
2127         PR tree-optimization/87301
2128         * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_cleanup): Properly
2129         clean EH info from leftover copy assignments.
2131 2018-09-17  Martin Liska  <mliska@suse.cz>
2133         PR gcov-profile/85871
2134         * gcov.c (output_intermediate_file): Fix out of bounds
2135         access.
2137 2018-09-17  Vineet Gupta  <vgupta@synopsys.com>
2139         * config/arc/arc.c: Object attributes for core4 not reflected
2140         correctly.
2141         * config/arc/arc.h: Don't restrict DBNZ to core3 (core4 includes
2142         core3).
2144 2018-09-17  Alexey Brodkin  <abrodkin@synopsys.com>
2146         * config/arc/linux.h (LINK_EH_SPEC): Add missing space.
2148 2018-09-17  Martin Liska  <mliska@suse.cz>
2150         * doc/gcov.texi: Document new option --use-hotness-colors.
2151         * gcov.c (struct source_info): Declare new field.
2152         (source_info::source_info): Set default for maximum_count.
2153         (print_usage): Add new -q option.
2154         (process_args): Process it.
2155         (accumulate_line_info): Save src->maximum_count.
2156         (output_line_beginning): Make color line number if
2157         flag_use_hotness_colors is set.
2158         (output_line_details): Pass default argument value.
2159         (output_lines): Pass src->maximum_count.
2161 2018-09-17  Martin Liska  <mliska@suse.cz>
2163         * common/config/i386/i386-common.c (ix86_get_valid_option_values):
2164         Use processor_names table.
2165         * config/i386/i386.c (ix86_default_align): Use
2166         processor_cost_table for alignment values.
2167         (ix86_option_override_internal): Use processor_names.
2168         (ix86_function_specific_print): Likewise.
2169         * config/i386/i386.h (struct processor_costs):
2170         Add alignment values.
2171         (struct ptt): Remove and replace with const char *.
2172         * config/i386/x86-tune-costs.h (struct processor_costs):
2173         Declare default alignments for all costs.
2175 2018-09-17  Aldy Hernandez  <aldyh@redhat.com>
2177         * tree-vrp.c (extract_range_from_unary_expr): Do not special case
2178         symbolics or VR_VARYING ranges for ABS_EXPR.
2179         * wide-int-range.cc (wide_int_range_abs): Return positive numbers
2180         when range will wrap.
2182 2018-09-15  Eric Botcazou  <ebotcazou@adacore.com>
2184         PR middle-end/86864
2185         * cfgexpand.c (expand_gimple_basic_block): Be prepared for a BARRIER
2186         before and after a JUMP_TABLE_DATA.
2188 2018-09-14  John David Anglin  <danglin@gcc.gnu.org>
2190         PR middle-end/87188
2191         * dojump.c (do_compare_and_jump): Canonicalize function pointers
2192         when one operand is a function pointer.  Use POINTER_TYPE_P and
2193         FUNC_OR_METHOD_TYPE_P.
2194         * expr.c (do_store_flag): Use POINTER_TYPE_P and FUNC_OR_METHOD_TYPE_P.
2195         * fold-const.c (build_range_check): Likewise.
2196         * match.pd (simple_comparison): Likewise.
2198 2018-09-14  David Malcolm  <dmalcolm@redhat.com>
2200         PR c/82967
2201         * spellcheck.c (get_edit_distance_cutoff): New function.
2202         (selftest::test_edit_distance_unit_test_oneway): Rename to...
2203         (selftest::test_get_edit_distance_one_way): ...this.
2204         (selftest::test_get_edit_distance_unit): Rename to...
2205         (selftest::test_get_edit_distance_both_ways): ...this.
2206         (selftest::test_edit_distances): Move tests to this new function,
2207         and test some more pairs of strings.  Update for above renaming.
2208         (selftest::get_old_cutoff): New function.
2209         (selftest::test_get_edit_distance_cutoff): New function.
2210         (selftest::assert_suggested_for): New function.
2211         (ASSERT_SUGGESTED_FOR): New macro.
2212         (selftest::assert_not_suggested_for): New function.
2213         (ASSERT_NOT_SUGGESTED_FOR): New macro.
2214         (selftest::test_suggestions): New function.
2215         (selftest::spellcheck_c_tests): Move test_get_edit_distance_unit
2216         tests to selftest::test_edit_distances and call it.  Add calls to
2217         selftest::test_get_edit_distance_cutoff and
2218         selftest::test_suggestions.
2219         * spellcheck.h (get_edit_distance_cutoff): New function declaration.
2220         (best_match::consider): Replace hard-coded cutoff calculation with
2221         a call to...
2222         (best_match::get_cutoff): New declaration.
2223         (best_match::get_best_meaningful_candidate): Likewise.
2225 2018-09-14  Bernd Edlinger  <bernd.edlinger@hotmail.de>
2227         * builtins.c (fold_builtin_strlen): Remove TODO comment.
2229 2018-09-14  Bernd Edlinger  <bernd.edlinger@hotmail.de>
2231         revert:
2232         2018-07-30  Bernd Edlinger  <bernd.edlinger@hotmail.de>
2234         * tree-ssa-forwprop.c (simplify_builtin_call): Don't create a not NUL
2235         terminated string literal.
2237 2018-09-14  Martin Sebor  <msebor@redhat.com>
2239         * builtins.c (unterminated_array): Handle ARRAY_REF.
2240         (expand_builtin_stpcpy_1): Detect unterminated char arrays.
2241         * builtins.h (unterminated_array): Declare extern.
2242         * gimple-fold.c (gimple_fold_builtin_stpcpy): Detect unterminated
2243         arrays.
2244         (gimple_fold_builtin_sprintf): Propagate NO_WARNING to transformed
2245         calls.
2247 2018-09-14  Martin Sebor  <msebor@redhat.com>
2248             Jeff Law  <law@redhat.com>
2250         * builtins.c (unterminated_array): New.
2251         (expand_builtin_strcpy): Adjust.
2252         (expand_builtin_strcpy_args): Detect unterminated arrays.
2253         * gimple-fold.c (get_maxval_strlen): Add argument.  Detect
2254         unterminated arrays.
2255         * gimple-fold.h (get_maxval_strlen): Add argument.
2256         (gimple_fold_builtin_strcpy): Detec unterminated arrays.
2258         * gimple-fold.c (get_range_strlen): Add argument.
2259         (get_maxval_strlen): Adjust.
2260         * gimple-fold.h (get_range_strlen): Add argument.
2262 2018-09-14  Wei Xiao  <wei3.xiao@intel.com>
2264         * config/i386/movdirintrin.h: Fix copyright year.
2266 2018-09-14  Uros Bizjak  <ubizjak@gmail.com>
2268         * reg-stack.c: Include regs.h.
2269         (replace_reg): Assert that mode is MODE_FLOAT or MODE_COMPLEX_FLOAT.
2270         (emit_pop_insn): Default pop insn mode to the reg_raw_mode of
2271         FIRST_STACK_REG, not DFmode.
2272         (emit_swap_insn): Default swap insn mode to the reg_raw_mode of
2273         FIRST_STACK_REG, not XFmode.  Explicitly construct swap RTX.
2274         (change stack): Default register mode to the reg_raw_mode of
2275         FIRST_STACK_REG, not DFmode.
2276         * config/i386/i386.md (*swap<mode>): Remove insn pattern.
2277         (*swapxf): Rename from swapxf.
2279 2018-09-14  Carl Love  <cel@us.ibm.com>
2281         * config/rs6000/emmintrin.h: Add _MM_SHUFFLE2.
2282         * config/rs6000/xmmintrin.h: Add _MM_SHUFFLE.
2284 2018-09-14  Segher Boessenkool  <segher@kernel.crashing.org>
2286         PR target/87224
2287         * config/rs6000/rs6000.md (*mov<mode>_hardfloat64): Add Z to the Y
2288         alternatives.
2290 2018-09-14  Sam Tebbs  <sam.tebbs@arm.com>
2292         PR target/85628
2293         * config/aarch64/aarch64.md (*aarch64_bfxilsi_uxtw): Define.
2295 2018-09-14  Jason Merrill  <jason@redhat.com>
2297         Fix --enable-gather-detailed-mem-stats.
2298         * hash-table.c (hash_table_usage): Change from variable to function.
2299         * hash-table.h: Adjust.
2300         * Makefile.in: Add missing dependencies on hash-table.h.
2302 2018-09-14  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
2304         PR tree-optimization/87259
2305         PR lto/87283
2306         (pass_cse_reciprocals::execute): Run optimize_recip_sqrt after
2307         execute_cse_reciprocals_1 has tried transforming.
2309 2018-09-14  Aldy Hernandez  <aldyh@redhat.com>
2311         * tree-vrp.c (extract_range_from_binary_expr_1): Normalize
2312         VR_VARYING for PLUS/MINUS_EXPR.
2314 2018-09-14  Ilya Leoshkevich  <iii@linux.ibm.com>
2316         * config/s390/s390-passes.def (INSERT_PASS_BEFORE): Improve
2317         formatting.
2319 2018-09-14  Richard Biener  <rguenther@suse.de>
2321         PR middle-end/63155
2322         * tree-ssa-coalesce.c (ssa_conflicts_merge): Remove conflict
2323         bits for the merged partition.
2325 2018-09-13  Martin Sebor  <msebor@redhat.com>
2326             Bernd Edlinger  <bernd.edlinger@hotmail.de>
2328         * builtins.h (c_srlen): Add argument.
2329         * builtins.c (warn_string_no_nul): New function.
2330         (c_strlen): Add argument and use it.  Update recursive calls.
2331         Pass DECL argument to string_constant to get info on non
2332         terminated strings.  Update *NONSTR as needed.
2333         (fold_builtin_strlen): Add argument to calls to c_strlen.
2334         Warn for unterminated arrays.
2335         (warn_string_no_null): Add prototype.
2336         * expr.c (string_constant): Update arguments.  Update recursive
2337         calls appropriately.  Detect missing NUL terminator and outermost
2338         declaration its missing in.
2339         Improve checks for arrays with nonzero lower bound or elements
2340         that are not a single byte.  Simplify offset computation.
2341         Simplify checks for non-NUL terminated strings.
2342         * gimple-fold.c (get_range_strlen): Add argument to c_strlen call.
2343         * gimple-ssa-sprintf.c (get_string_length): Remove unnecessary code.
2345 2018-09-13  Bernd Edlinger  <bernd.edlinger@hotmail.de>
2347         * builtins.c (c_strlen): Handle not zero terminated STRING_CSTs
2348         correctly.
2349         * fold-const.c (c_getstr): Fix function comment.  Remove unused third
2350         argument.  Fix range checks.
2351         * fold-const.h (c_getstr): Adjust protoype.
2352         * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid folding when
2353         string is constant but contains no NUL byte.
2355         * expr.c (string_constant): Adjust function comment.
2356         Remove bogus check for zero termination.
2358         * fold-const.c (c_getstr): Clamp STRING_LENGTH to STRING_SIZE.
2360         * varasm.c (compare_constant): Compare type size of STRING_CSTs.
2361         (get_constant_size): Don't make STRING_CSTs larger than they are.
2362         (check_string_literal): New check function for STRING_CSTs.
2363         (output_constant): Use it.
2365 2018-09-13  Eric Botcazou  <ebotcazou@adacore.com>
2367         PR target/86812
2368         * config/visium/visium.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define.
2370 2018-09-13  Richard Biener  <rguenther@suse.de>
2372         PR tree-optimization/87263
2373         * tree-ssa-sccvn.c (visit_phi): Revert some earlier changes.
2374         (struct unwind_state): Add max_rpo field.
2375         (do_rpo_vn): Allow up-to-date loop state to be used when not iterating.
2376         Compute max_rpo, the max RPO number a block can be backwards reached
2377         from.  Re-write non-iterating mode to a RPO ordered worklist approach,
2378         separating it from the iterating mode.
2380 2018-09-13  Vlad Lazar  <vlad.lazar@arm.com>
2382         * haifa-sched.c (rank_for_schedule): Schedule by INSN_COST.
2383         (rfs_decision): New scheduling decision.
2385 2018-09-13  Richard Biener  <rguenther@suse.de>
2387         PR bootstrap/87134
2388         * tree-ssa-sccvn.c (vn_nary_op_insert_into): Fix assert.
2389         (vn_nary_op_insert_pieces_predicated): Do not write useless
2390         valid_dominated_by_p entry outside of the allocated storage.
2392 2018-09-13  Omar Sandoval  <osandov@osandov.com>
2393             Tom de Vries  <tdevries@suse.de>
2395         PR debug/86985
2396         * dwarf2out.c (is_c): New function.
2397         (add_subscript_info): Add DW_AT_count of 0 for C zero-length arrays.
2399 2018-09-13  Sam Tebbs  <sam.tebbs@arm.com>
2401         PR target/85628
2402         * config/aarch64/aarch64.md (*aarch64_bfxil):
2403         Define.
2404         * config/aarch64/constraints.md (Ulc): Define.
2405         * config/aarch64/aarch64-protos.h (aarch64_high_bits_all_ones_p):
2406         Define.
2407         * config/aarch64/aarch64.c (aarch64_high_bits_all_ones_p):
2408         New function.
2410 2018-09-13  Vlad Lazar  <vlad.lazar@arm.com>
2412         * config/aarch64/aarch64.h (TARGET_COMPUTE_FRAME_LAYOUT): Define.
2413         * config/aarch64/aarch64.c (aarch64_expand_prologue): Remove
2414         aarch64_layout_frame call.
2415         (aarch64_expand_epilogue): Likewise.
2416         (aarch64_initial_elimination_offset): Likewise.
2417         (aarch64_get_separate_components): Likewise.
2418         (aarch64_use_return_insn_p): Likewise.
2419         (aarch64_layout_frame): Remove unneeded check.
2421 2018-09-13  Jakub Jelinek  <jakub@redhat.com>
2423         * configure.ac: Only append
2424         " : (reconfigured) $TOPLEVEL_CONFIGURE_ARGUMENTS" to
2425         gcc_config_arguments if it was never reconfigured or last reconfigure
2426         was with different arguments.
2427         * configure: Regenerated.
2429 2018-09-13  Jakub Jelinek  <jakub@redhat.com>
2430             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
2432         PR middle-end/87290
2433         * expr.c (maybe_optimize_pow2p_mod_cmp): New function.
2434         (maybe_optimize_mod_cmp): Use it if integer_pow2p treeop1.
2436 2018-09-13  Jakub Jelinek  <jakub@redhat.com>
2438         PR tree-optimization/87287
2439         * fold-const.c (fold_binary_loc) <case EQ_EXPR>: Move signed modulo
2440         X % C == 0 to X % (unsigned) C == 0 optimization to ...
2441         * match.pd (X % C == 0): ... here.  New optimization.
2443 2018-09-12  Jakub Jelinek  <jakub@redhat.com>
2445         PR middle-end/82853
2446         * expr.h (maybe_optimize_mod_cmp): Declare.
2447         * expr.c (mod_inv): New function.
2448         (maybe_optimize_mod_cmp): New function.
2449         (do_store_flag): Use it.
2450         * cfgexpand.c (expand_gimple_cond): Likewise.
2452 2018-09-09  Cesar Philippidis  <cesar@codesourcery.com>
2453             Julian Brown  <julian@codesourcery.com>
2455         PR middle-end/86336
2456         * gimplify.c (gimplify_scan_omp_clauses): Set
2457         target_firstprivatize_array_bases in OpenACC parallel and kernels
2458         region contexts.  Remove GOMP_MAP_FIRSTPRIVATE_REFERENCE clauses from
2459         OpenACC data regions.
2461 2018-09-12  Uros Bizjak  <ubizjak@gmail.com>
2463         * config/i386/i386.md (sqrt_extend<mode>xf3_i387): Remove.
2464         (sqrt<mode>2): Extend operand 1 to XFmode and generate sqrtxf3 insn.
2466 2018-09-12  Richard Biener  <rguenther@suse.de>
2468         PR tree-optimization/87280
2469         * tree-ssa-sccvn.c (process_bb): Handle the case of executable
2470         edge but unreachable target.
2471         (do_rpo_vn): For conservatively handling a PHI only mark
2472         the backedge executable but not the block reachable.
2474 2018-09-12  Richard Biener  <rguenther@suse.de>
2476         PR tree-optimization/87266
2477         * tree-ssa-sccvn.c (do_rpo_vn): Always iterate to not yet
2478         visited blocks.
2480 2018-09-12  Andreas Krebbel  <krebbel@linux.ibm.com>
2482         * config/s390/s390.md (PFPO_RND_MODE_DFP, PFPO_RND_MODE_BFP): New
2483         constants.
2484         ("trunc<BFP:mode><DFP_ALL:mode>2")
2485         ("trunc<DFP_ALL:mode><BFP:mode>2")
2486         ("extend<BFP:mode><DFP_ALL:mode>2")
2487         ("extend<DFP_ALL:mode><BFP:mode>2"): Set proper rounding mode
2488         according to the target operand type.
2490 2018-09-12  Jakub Jelinek  <jakub@redhat.com>
2491             Andreas Krebbel  <krebbel@linux.ibm.com>
2493         PR tree-optimization/86844
2494         * gimple-ssa-store-merging.c
2495         (imm_store_chain_info::coalesce_immediate): For overlapping stores, if
2496         there are any overlapping stores in between them, make sure they are
2497         also coalesced or we give up completely.
2499 2018-09-12  Jakub Jelinek  <jakub@redhat.com>
2501         PR middle-end/87248
2502         * fold-const.c (fold_ternary_loc) <case COND_EXPR>: Verify also that
2503         BIT_AND_EXPR's second operand is a power of two.  Formatting fix.
2505 2018-09-12  Tom de Vries  <tdevries@suse.de>
2507         * common.opt (gdescribe-dies): Add option.
2508         * dwarf2out.c (add_name_and_src_coords_attributes): Add description
2509         attribute for artifical and nameless decls.
2510         (dwarf2out_register_external_die): Add description attribute to
2511         external reference die.
2512         (add_desc_attribute): New functions.
2513         (gen_subprogram_die): Add description attribute to
2514         DW_TAG_call_site_parameter.
2515         * tree-pretty-print.c (print_generic_expr_to_str): New function.
2516         * tree-pretty-print.h (print_generic_expr_to_str): Declare.
2517         * doc/invoke.texi (@item Debugging Options): Add -gdescribe-dies and
2518         -gno-describe-dies.
2519         (@item -gdescribe-dies): Add.
2521 2018-09-12  Aldy Hernandez  <aldyh@redhat.com>
2523         * tree-vrp.c (vrp_shift_undefined_p): Remove.
2524         (extract_range_from_binary_expr_1: Call
2525         wide_int_range_shift_undefined_p instead of vrp_shift_undefined_p.
2526         * wide-int-range.h (wide_int_range_shift_undefined_p): Do not
2527         depend on sign.
2529 2018-09-12  Aldy Hernandez  <aldyh@redhat.com>
2531         * gimple-ssa-warn-alloca.c
2532         (alloca_type_and_limit::alloca_type_and_limit): Initialize limit
2533         field for ALLOCA_BOUND_*_LARGE.
2535 2018-09-11  Nathan Sidwell  <nathan@acm.org>
2537         * gcc.c (load_specs, execute, run_attempt): Use %qs not '%s'.
2539 2018-09-11  Uros Bizjak  <ubizjak@gmail.com>
2541         * reg-stack.c (subst_asm_stack_regs): Call replace_reg also
2542         for clobbers.  Remove obsolete comment.
2544 2018-09-11  Uros Bizjak  <ubizjak@gmail.com>
2546         * config/i386/i386.md (define_attr "type"): Remove mpxmov, mpxmk,
2547         mpxchk, mpxld and mpxst types.
2548         (define_attr length_immediate): Remove all processing of mpx types.
2549         (define_attr prefix_0f): Ditto.
2550         (define_attr memory): Ditto.
2552 2018-09-11  Uros Bizjak  <ubizjak@gmail.com>
2554         * config/i386/i386.md (fyl2x_extend<mode>xf3_i387): Remove.
2555         (log<mode>2): Change operand 1 predicate to general_operand.
2556         Extend operand 1 to XFmode and generate logxf3 insn.
2557         (log10<mode>2): Change operand 1 predicate to general_operand.
2558         Extend operand 1 to XFmode and generate log10xf3 insn.
2559         (log2<mode>2): Change operand 1 predicate to general_operand.
2560         Extend operand 1 to XFmode and generate log2xf3 insn.
2561         (fyl2xp1_extend<mode>xf3_i387): Remove.
2562         (log1p<mode>2): Change operand 1 predicate to general_operand.
2563         Extend operand 1 to XFmode and generate log1pxf3 insn.
2564         (fxtract_extend<mode>xf3_i387): Remove.
2565         (logb<mode>2): Change operand 1 predicate to general_operand.
2566         Extend operand 1 to XFmode and generate logbxf3 insn.
2567         (ilogb<mode>2): Change operand 1 predicate to general_operand.
2568         Extend operand 1 to XFmode and generate fxtractxf3_i387 insn.
2569         (significand<mode>2): Change operand 1 predicate to general_operand.
2570         Extend operand 1 to XFmode and generate significandxf3 insn.
2572 2018-09-11  Nathan Sidwell  <nathan@acm.org>
2574         * gcc.c (perror_with_name, pfatal_with_name): Delete.
2575         (load_specs): Use fatal_error.
2576         (DELETE_IF_ORDINARY, process_command): Use error.
2577         (execute, run_attempt): Use fatal_error.
2579 2018-09-11  Andrew Stubbs  <ams@codesourcery.com>
2581         * diagnostic-core.h (sorry_at): New prototype.
2582         * diagnostic.c (sorry_at): New function.
2584 2018-09-11  Aldy Hernandez  <aldyh@redhat.com>
2586         * tree-vrp.c (extract_range_from_binary_expr_1): Treat all divisions
2587         by zero as VR_UNDEFINED.
2589 2018-09-10  Uros Bizjak  <ubizjak@gmail.com>
2591         * config/i386/i386.md (<sincos>xf2): Rename from *<sincos>xf2_i387.
2592         (*<sincos>_extend<mode>xf2_i387): Remove insn pattern.
2593         (<sincos>mode2): New expander.
2594         (sincos_extend<mode>xf3_i387): Remove insn pattern.
2595         (sincos -> sin, cos splitters): Remove splitter patterns.
2596         (sincos<mode>3): Change operand 2 predicate to general_operand.
2597         Extend operand 2 to XFmode and generate sincosxf3 insn.
2598         (fptanxf4_i387): Change mode of operands 0 and 3 to SFmode.
2599         Change operand 3 predicate to const1_operand.
2600         (fptan_extend<mode>xf4_i387): Remove insn pattern.
2601         (tanxf2): Update operands in the call to fptanxf4_i387.
2602         (tan<mode>2): Change operand 1 predicate to general_operand.
2603         Extend operand 1 to XFmode and generate tanxf3 insn.
2604         (atan2xf3): Rename from *fpatanxf3_i387.
2605         (fpatan_extend<mode>xf3_i387): Remove insn pattern.
2606         (atan2xf3): Remove expander.
2607         (atan2<mode<3):  Change operand 1 and 2 predicates to general_operand.
2608         Extend operands 1 and 2 to XFmode and generate atan2xf3 insn.
2609         (atan<mode>2): Change operand 1 predicate to general_operand.
2610         Extend operand 1 to XFmode and generate atanxf3 insn.
2612 2018-09-10  Uros Bizjak  <ubizjak@gmail.com>
2614         * config/i386/i386.md (x87/SSE constant load splitter): Use
2615         memory_operand instead of nonimmediate_operand for input operand
2616         predicate.
2618 2018-09-09  Uros Bizjak  <ubizjak@gmail.com>
2620         * config/i386/i386.md (float partial SSE register stall splitter): Move
2621         splitter near its instruction pattern.
2622         (float_extend partial SSE register stall splitter): Ditto.
2623         (float_truncate partial SSE register stall splitter): Ditto.
2625 2018-09-09  Hans-Peter Nilsson  <hp@bitrange.com>
2627         PR target/86794
2628         * config/mmix/mmix.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
2629         to speculation_safe_value_not_needed.
2631         PR target/85666
2632         * config/mmix/mmix.c (mmix_assemble_integer): Handle byte-size
2633         non-CONST_INT rtx:es using assemble_integer_with_op ".byte".
2634         (MMIX_CFUN_NEEDS_SAVED_EH_RETURN_ADDRESS): Don't call
2635         leaf_function_p, instead use has_hard_reg_initial_val.
2637 2018-09-09  Nathan Sidwell  <nathan@acm.org>
2639         * gcc.h (pfatal_with_name): Don't declare here.
2640         * gcc.c (pfatal_with_name): Make static.
2642 2018-09-09  Xianmiao Qu  <xianmiao_qu@c-sky.com>
2644         * config/csky/csky.md (*cskyv2_adddi3, *cskyv2_subdi3): Add
2645         earlyclobber.
2647 2018-09-08  John David Anglin  <danglin@gcc.gnu.org>
2649         PR rtl-optimization/85458
2650         * sel-sched.c (sel_target_adjust_priority): Allow backend adjust
2651         priority hook to reduce the priority of EXPR.
2653 2018-09-07  Uros Bizjak  <ubizjak@gmail.com>
2655         * config/i386/i386.md (float<SWI48x:mode><MODEF:mode>2) Enable
2656         DImode for x87 on 32bit targets.  Conditionally disable x87 modes
2657         with X87_ENABLE_FLOAT.  Remove preparation code.
2658         (*float<SWI48:mode><MODEF:mode>2): Rename from
2659         *float<SWI48:mode><MODEF:mode>2_mixed.  Handle x87, SSE and mixed
2660         math using "enabled" attribute.
2661         (*floatdi<MODEF:mode>2_i387): Rename from
2662         *float<SWI48x:mode><MODEF:mode>2_i387.  Handle only DImode and
2663         enable for 32bit targets only.
2664         (floatdi<X87MODEF:mode>2_i387_with_xmm pre-reload splitter): New
2665         splitter.
2666         (floatdi<X87MODEF:mode>2_i387_with_xmm): Use register_operand
2667         as operand 1 predicate.  Rewrite as define_insn_and_split.
2668         (floatdi<X87MODEF:mode>2_i387_with_xmm memory input splitter): Remove.
2670 2018-09-06  Uros Bizjak  <ubizjak@gmail.com>
2672         * reg-stack.c (get_true_reg) <case FLOAT_TRUNCATE>: Reorder
2673         to fallthru to FLOAT case.
2675 2018-09-06  Will Schmidt  <will_schmidt@vnet.ibm.com>
2677         PR target/86731
2678         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Update logic
2679         around folding of vec_sl to handle out of range shift values.
2681 2018-09-06  Uros Bizjak  <ubizjak@gmail.com>
2683         * config/i386/i386.md (fix_trunc<mode>_fisttp_i387_1): Remove.
2684         Update callers to gen_fix_trunc<mode>_i387_fisttp
2685         (fix_trunc<mode>_i387_fisttp): Change operand 0 predicate to
2686         nonimmediate_operand.
2687         (fix_trunc<mode>_i387_fisttp_with_temp): Remove insn pattern
2688         and corresponding splitters.
2689         (*fix_trunc<mode>_i387_1): Always expand with fix_trunc<mode>_i387.
2690         (fix_truncdi_i387): Change operand 0 predicate to nonimmediate_operand.
2691         (fix_truncdi_i387_with_temp): Remove insn pattern
2692         and corresponding splitters.
2693         (fix_trunc<mode>_i387): Change operand 0 predicate to
2694         nonimmediate_operand.
2695         (fix_trunc<mode>_i387_with_temp): Remove insn pattern
2696         and corresponding splitters.
2697         (*fistdi2_1): Remove.
2698         (fistdi2): Ditto.
2699         (fistdi2_with_temp): Remove insn pattern and corresponding splitters.
2700         (lrintxfdi2): Remove expander.  Reimplement as define_insn.
2701         (*fist<mode>2_1): Remove.
2702         (fist<mode>2): Ditto.
2703         (fist<mode>2_with_temp): Remove insn pattern and corresponding
2704         splitters.
2705         (lrintxf<mode>2): Remove expander.  Reimplement as define_insn.
2706         (*fist<mode>2_<rounding>_1): Always expand with fist<mode>2_<rounding>.
2707         (fistdi2_<rounding>): Change operand 0 predicate to
2708         nonimmediate_operand.
2709         (fistdi2_<rounding>_with_temp): Remove insn pattern
2710         and corresponding splitters.
2711         (fist<mode>2_<rounding>): Change operand 0 predicate to
2712         nonimmediate_operand.
2713         (fist<mode>2_<rounding>_with_temp): Remove insn pattern
2714         and corresponding splitters.
2716         (*fixuns_trunc<mode>si2_avx512f_zext): Depend on TARGET_SSE_MATH.
2718 2018-09-06  Bernd Edlinger  <bernd.edlinger@hotmail.de>
2720         * varasm.c (output_constructor_regular_field): Check TYPE_SIZE_UNIT of
2721         the init value.
2723 2018-09-06  Will Schmidt  <will_schmidt@vnet.ibm.com>
2725         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support for
2726         early gimple folding of vec_splat().
2727         * tree-vect-generic.c: Remove static from tree_vec_extract() definition.
2728         * gimple-fold.h: Add an extern define for tree_vec_extract().
2730 2018-09-06  Will Schmidt  <will_schmidt@vnet.ibm.com>
2732         * config/rs6000/rs6000.c (fold_mergehl_helper): Add types_compatible_p
2733         wrappers around TREE_TYPE comparisons.
2735 2018-09-06  Ilya Leoshkevich  <iii@linux.ibm.com>
2737         PR target/80080
2738         * config/s390/predicates.md: Add nonsym_memory_operand.
2739         * config/s390/s390.c (s390_legitimize_cs_operand): If operand
2740         contains a SYMBOL_REF, load it into an intermediate pseudo.
2741         (s390_emit_compare_and_swap): Legitimize operand.
2742         * config/s390/s390.md: Use the new nonsym_memory_operand
2743         with UNSPECV_CAS patterns.
2745 2018-09-06  Ilya Leoshkevich  <iii@linux.ibm.com>
2747         PR target/80080
2748         * config/s390/s390-passes.def: New file.
2749         * config/s390/s390-protos.h (class rtl_opt_pass): Add forward
2750         declaration.
2751         (make_pass_s390_early_mach): Add declaration.
2752         * config/s390/s390.c (make_pass_s390_early_mach):
2753         (s390_option_override): Remove dynamic registration.
2754         * config/s390/t-s390: Add s390-passes.def.
2756 2018-09-06  Ilya Leoshkevich  <iii@linux.ibm.com>
2758         * config/s390/s390.c (s390_decompose_constant_pool_ref):
2759         Remove UNSPEC_LTREL_BASE check.
2760         (annotate_constant_pool_refs): Likewise.
2761         (find_constant_pool_ref): Likewise.
2762         (find_ltrel_base): Removed.
2763         (replace_ltrel_base): Removed.
2764         (s390_mainpool_finish): Remove replace_ltrel_base call.
2765         (s390_chunkify_start): Remove pending LTREL_BASE logic.
2766         (s390_chunkify_finish): Remove replace_ltrel_base call.
2767         * config/s390/s390.md: Remove UNSPEC_LTREL_BASE.
2769 2018-09-06  Hans-Peter Nilsson  <hp@axis.com>
2771         PR target/86779
2772         * config/cris/cris.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
2773         to speculation_safe_value_not_needed.
2775 2018-09-05  Cesar Philippidis  <cesar@codesourcery.com>
2776             Bernd Schmidt  <bernds_cb1@t-online.de>
2778         * config/nvptx/nvptx-opts.h: New file.
2779         * config/nvptx/nvptx.c (nvptx_file_start): Print the correct .target.
2780         * config/nvptx/nvptx.h: Include "nvptx-opts.h".
2781         (ASM_SPEC): Define.
2782         (TARGET_SM35): New macro.
2783         * config/nvptx/nvptx.md (atomic_fetch_<logic><mode>): Enable with the
2784         correct predicate.
2785         * config/nvptx/nvptx.opt (ptx_isa, sm_30, sm_35): New enum and its
2786         values.
2787         (misa=): New option.
2788         * doc/invoke.texi (Nvidia PTX Options): Document -misa.
2790 2018-09-05  Uros Bizjak  <ubizjak@gmail.com>
2792         * config/i386/i386.md (truncdfsf2): Remove expander.
2793         (truncdfsf2_with_temp): Ditto.
2794         (truncxf<mode>2): Ditto.
2795         (*truncdfsf_fast_mixed): Remove insn pattern.
2796         (*truncdfsf_fast_i387): Ditto.
2797         (*truncdfsf_mixed): Ditto.
2798         (*truncdfsf_i387): Ditto.
2799         (*truncdfsf2_i387_1): Ditto.
2800         (*truncxfsf2_mixed): Ditto.
2801         (*truncxfdf2_mixed): Ditto.
2802         (*truncxf<mode>2_i387_noop): Ditto. Update callers
2803         to call gen_truncxf<mode>2 instead.
2804         (*truncxf<mode>2_i387): Remove.
2805         (reg->reg splitters): Remove splitter pattern.
2806         (reg->mem splitters): Ditto.
2808         (truncdfsf2): New insn pattern.
2809         (truncxf<mode>2): Ditto.
2811 2018-09-05  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
2813         * tree-ssa-math-opts.c (is_mult_by): New function.
2814         (is_square_of): Use the above.
2815         (optimize_recip_sqrt): New function.
2816         (pass_cse_reciprocals::execute): Use the above.
2818 2018-09-05  Richard Biener  <rguenther@suse.de>
2820         PR bootstrap/87134
2821         * tree-ssa-sccvn.c (rpo_elim::eliminate_push_avail): Make sure
2822         to zero-init the emplaced vec.
2824 2018-09-05  Martin Liska  <mliska@suse.cz>
2826         PR tree-optimization/87205
2827         * tree-switch-conversion.c (pass_lower_switch::execute):
2828         Group cases for switch statements.
2830 2018-09-05  Richard Biener  <rguenther@suse.de>
2832         PR tree-optimization/87217
2833         * tree-ssa-sccvn.c (vuse_valueize): New.
2834         (vn_reference_lookup_pieces): Use it.
2835         (vn_reference_lookup): Likewise.
2837 2018-09-05  Nathan Sidwell  <nathan@acm.org>
2839         PR c++/87137
2840         * stor-layout.c (place_field): Scan forwards to check last
2841         bitfield when ms_bitfield_placement is in effect.
2843 2018-09-05  Richard Biener  <rguenther@suse.de>
2845         PR bootstrap/87225
2846         * tree-vect-stmts.c (vectorizable_simd_clone_call): Fix bogus
2847         return.
2849 2018-09-05  Siddhesh Poyarekar  <siddhesh@sourceware.org>
2850             Kugan Vivekanandarajah  <kugan.vivekanandarajah@linaro.org>
2852         * config/aarch64/falkor-tag-collision-avoidance.c: New file.
2853         * config.gcc (extra_objs): Build it.
2854         * config/aarch64/t-aarch64 (falkor-tag-collision-avoidance.o):
2855         Likewise.
2856         * config/aarch64/aarch64-passes.def
2857         (pass_tag_collision_avoidance): New pass.
2858         * config/aarch64/aarch64.c (qdf24xx_tunings): Add
2859         AARCH64_EXTRA_TUNE_RENAME_LOAD_REGS to tuning_flags.
2860         (aarch64_classify_address): Remove static qualifier.
2861         (aarch64_address_info, aarch64_address_type): Move to...
2862         * config/aarch64/aarch64-protos.h: ... here.
2863         (make_pass_tag_collision_avoidance): New function.
2864         * config/aarch64/aarch64-tuning-flags.def (rename_load_regs):
2865         New tuning flag.
2867 2018-09-05  Martin Liska  <mliska@suse.cz>
2869         * doc/gcov.texi: Update documentation of humar
2870         readable mode.
2871         * gcov.c (format_count): Print one decimal place, it provides
2872         more fine number of situations like '1G' vs. '1.4G'.
2874 2018-09-05  Martin Liska  <mliska@suse.cz>
2876         PR target/87164
2877         * config/rs6000/rs6000.opt: Mark the option as Deprecated.
2878         * optc-gen.awk: Allow 'Var' for Deprecated options in order
2879         to generate a MASK value.
2881 2018-09-04  H.J. Lu  <hongjiu.lu@intel.com>
2883         PR debug/86593
2884         * dwarf2out.c (based_loc_descr): Allow hard frame pointer even
2885         if frame pointer isn't used.
2886         (compute_frame_pointer_to_fb_displacement): Likewise.
2888 2018-09-04  Jakub Jelinek  <jakub@redhat.com>
2890         PR target/87198
2891         * common/config/i386/i386-common.c (OPTION_MASK_ISA_XSAVEOPT_SET,
2892         OPTION_MASK_ISA_XSAVES_SET, OPTION_MASK_ISA_XSAVEC_SET): Use
2893         OPTION_MASK_ISA_XSAVE_SET instead of OPTION_MASK_ISA_XSAVE.
2894         (OPTION_MASK_ISA_XSAVE_UNSET): Add OPTION_MASK_ISA_XSAVES_UNSET
2895         and OPTION_MASK_ISA_XSAVEC_UNSET.
2897 2018-09-04  Max Filippov  <jcmvbkbc@gmail.com>
2899         * config/xtensa/xtensa.c (xtensa_expand_atomic): Reorder AND and
2900         XOR operations in NAND case.
2902 2018-09-04  Aldy Hernandez  <aldyh@redhat.com>
2904         * wide-int-range.cc (wide_int_range_convert): New.
2905         * wide-int-range.h (wide_int_range_convert): New.
2906         * tree-vrp.c (extract_range_from_unary_expr): Abstract wide int
2907         code into wide_int_range_convert.
2908         (extract_range_into_wide_ints): Do not munge anti range constants
2909         into the entire domain.  Just return the range back.
2911 2018-09-04  Martin Liska  <mliska@suse.cz>
2913         * genmatch.c (output_line_directive): Add new argument
2914         fnargs.
2915         (dt_simplify::gen_1): Encapsulate dump within __builtin_expect.
2917 2018-09-04  Jonathan Wakely  <jwakely@redhat.com>
2919         * doc/invoke.texi (Option Summary): Add whitespace.
2921         * doc/invoke.texi (Option Summary): Add -Waligned-new.
2923 2018-09-04  Richard Biener  <rguenther@suse.de>
2925         PR tree-optimization/87211
2926         * tree-ssa-sccvn.c (visit_phi): When value-numbering to a
2927         backedge value we're supposed to treat as VARYING also number
2928         the PHI to VARYING in case it got a different value-number already.
2930 2018-09-04  Aldy Hernandez  <aldyh@redhat.com>
2932         * tree-vrp.c (vrp_can_optimize_bit_op): Remove.
2933         (extract_range_from_binary_expr_1): Do not call
2934         vrp_can_optimize_bit_op.
2935         * wide-int-range.cc (wide_int_range_can_optimize_bit_op): Make
2936         static.
2937         (wide_int_range_get_mask_and_bounds): New.
2938         (wide_int_range_optimize_bit_op): New.
2939         (wide_int_range_bit_ior): Call wide_int_range_optimize_bit_op.
2940         (wide_int_range_bit_and): Same.
2941         * wide-int-range.h (wide_int_range_can_optimize_bit_op): Remove.
2942         (wide_int_range_optimize_bit_op): New.
2943         (wide_int_range_get_mask_and_bounds): New.
2945 2018-09-04  Richard Biener  <rguenther@suse.de>
2947         PR tree-optimization/87176
2948         * tree-ssa-sccvn.c (visit_phi): Remove redundant allsame
2949         variable.  When value-numbering a virtual PHI node make sure
2950         to not value-number to the backedge value.
2952 2018-09-04  Jonathan Wakely  <jwakely@redhat.com>
2954         * doc/extend.texi (Long Long, Hex Floats): Document support for
2955         long long and hex floats in more recent versions of ISO C++.
2957 2018-09-03  Richard Biener  <rguenther@suse.de>
2959         PR tree-optimization/87177
2960         * tree-ssa-sccvn.c (vuse_ssa_val): Revert previous change, keep
2961         cleanup.
2963 2018-09-03  Alexander Monakov  <amonakov@ispras.ru>
2965         * bb-reorder.c (edge_order): Convert to C-qsort-style
2966         tri-state comparator.
2967         (reorder_basic_blocks_simple): Change std::stable_sort to
2968         gcc_stablesort.
2970 2018-09-03  Alexander Monakov  <amonakov@ispras.ru>
2972         * tree-loop-distribution.c (offset_cmp): Convert to C-qsort-style
2973         tri-state comparator.
2974         (fuse_memset_builtins): Change std::stable_sort to gcc_stablesort.
2976 2018-09-03  Alexander Monakov  <amonakov@ispras.ru>
2978         * sort.cc (struct sort_ctx): New field 'nlim'.  Use it...
2979         (mergesort): ... here as maximum count for using netsort.
2980         (gcc_qsort): Set nlim to 3 if stable sort is requested.
2981         (gcc_stablesort): New.
2982         * system.h (gcc_stablesort): Declare.
2984 2018-09-03  Alexander Monakov  <amonakov@ispras.ru>
2986         * sort.cc (gcc_qsort) [CHECKING_P]: Call qsort_chk.
2987         * system.h (qsort): Always redirect to gcc_qsort.  Update comment.
2988         * vec.c (qsort_chk): Do not call gcc_qsort.  Update comment.
2990 2018-09-03  Segher Boessenkool  <segher@kernel.crashing.org>
2992         * config/rs6000/rs6000.md (*mov<mode>_hardfloat32): Remove %U from the
2993         lxsdx and stxsdx alternatives.
2994         (*mov<mode>_hardfloat64): Ditto.
2995         * config/rs6000/vsx.md (*vsx_extract_<mode>_store): Ditto.
2997 2018-09-03  Richard Biener  <rguenther@suse.de>
2999         PR tree-optimization/87200
3000         * tree-ssa-sccvn.c (vn_nary_build_or_lookup_1): Valueize a
3001         simplify result.
3003 2018-09-03  Martin Liska  <mliska@suse.cz>
3005         PR tree-optimization/87201
3006         * tree-switch-conversion.c (switch_decision_tree::balance_case_nodes):
3007         Fix parenthesis in an expression.
3009 2018-09-03  Richard Biener  <rguenther@suse.de>
3011         PR tree-optimization/87197
3012         * tree-ssa-sccvn.c (vn_nary_build_or_lookup_1): Mark the new def
3013         visited.  CSE the VN_INFO hashtable lookup.
3015         PR tree-optimization/87169
3016         * tree-ssa-sccvn.c (do_rpo_vn): When marking loops for not
3017         iterating make sure there's no extra backedges from irreducible
3018         regions feeding the header.  Mark the destination block
3019         executable.
3021 2018-09-03  Martin Liska  <mliska@suse.cz>
3023         PR driver/83193
3024         * common/common-target.def: Add TARGET_GET_VALID_OPTION_VALUES.
3025         * common/common-targhooks.c (default_get_valid_option_values):
3026         New function.
3027         * common/common-targhooks.h (default_get_valid_option_values):
3028         Likewise.
3029         * common/config/i386/i386-common.c: Move processor_target_table
3030         from i386.c.
3031         (ix86_get_valid_option_values): New function.
3032         (TARGET_GET_VALID_OPTION_VALUES): New macro.
3033         * config/i386/i386.c (struct ptt): Move to i386-common.c.
3034         (PTA_*): Move all defined masks into i386-common.c.
3035         (ix86_function_specific_restore): Use new processor_cost_table.
3036         * config/i386/i386.h (struct ptt): Moved from i386.c.
3037         (struct pta): Likewise.
3038         * doc/tm.texi: Document new TARGET_GET_VALID_OPTION_VALUES.
3039         * doc/tm.texi.in: Likewise.
3040         * opt-suggestions.c (option_proposer::suggest_option):
3041         Pass prefix to build_option_suggestions.
3042         (option_proposer::get_completions): Likewise.
3043         (option_proposer::build_option_suggestions): Use the new target
3044         hook.
3045         * opts.c (struct option_help_tuple): New struct.
3046         (print_filtered_help): Use the new target hook.
3048 2018-09-03  Martin Liska  <mliska@suse.cz>
3050         PR middle-end/59521
3051         * predict.c (set_even_probabilities): Add likely_edges
3052         argument and handle cases where we have precisely one
3053         likely edge.
3054         (combine_predictions_for_bb): Catch also likely_edges.
3055         (tree_predict_by_opcode): Handle gswitch statements.
3056         * tree-cfg.h (find_case_label_for_value): New declaration.
3057         (find_taken_edge_switch_expr): Likewise.
3058         * tree-switch-conversion.c (switch_decision_tree::balance_case_nodes):
3059         Find pivot in decision tree based on probabily, not by number of
3060         nodes.
3062 2018-09-02  Gerald Pfeifer  <gerald@pfeifer.com>
3064         * doc/standards.texi (Standards): Update Objective-C reference.
3066 2018-09-01  Gerald Pfeifer  <gerald@pfeifer.com>
3068         * doc/install.texi (Prerequisites): Update link for MPC.
3070 2018-09-01  Michael Matz  <matz@suse.de>
3072         PR tree-optimization/87074
3073         * gimple-loop-jam.c (unroll_jam_possible_p): Check loop exit
3074         PHIs for outer-loop uses.
3076 2018-09-01  Gerald Pfeifer  <gerald@pfeifer.com>
3078         * doc/generic.texi (OpenMP): Adjust link to openmp.org.
3079         * doc/invoke.texi (C Dialect Options): Ditto.
3081 2018-09-01  Gerald Pfeifer  <gerald@pfeifer.com>
3083         * doc/install.texi (Prerequisites): Adjust link mpfr.org.
3085 2018-08-31  Richard Biener  <rguenther@suse.de>
3087         PR tree-optimization/87168
3088         * tree-ssa-sccvn.c (SSA_VAL): Add visited output parameter.
3089         (rpo_elim::eliminate_avail): When OP was not visited it must
3090         be available.
3092 2018-08-31  David Malcolm  <dmalcolm@redhat.com>
3094         * tree-vrp.c (copy_value_range): Convert param "from" from
3095         "value_range *" to "const value_range *".
3096         (range_is_null): Likewise for param "vr".
3097         (range_int_cst_p): Likewise.
3098         (range_int_cst_singleton_p): Likewise.
3099         (symbolic_range_p): Likewise.
3100         (value_ranges_intersect_p): Likewise for both params.
3101         (value_range_nonnegative_p): Likewise for param "vr".
3102         (value_range_constant_singleton): Likewise.
3103         (vrp_set_zero_nonzero_bits): Likewise for param "ar".
3104         (extract_range_into_wide_ints): Likewise for param "vr".
3105         (extract_range_from_multiplicative_op): Likewise for params "vr0"
3106         and "vr1".
3107         (vrp_can_optimize_bit_op): Likewise.
3108         (extract_range_from_binary_expr_1): Likewise for params "vr0_" and
3109         "vr1_".
3110         (extract_range_from_unary_expr): Likewise.
3111         (debug_value_range): Likewise for param "vr".
3112         (value_range::dump): Add "const" qualifier.
3113         (vrp_prop::check_array_ref): Convert local "vr" from
3114         "value_range *" to "const value_range *".
3115         (vrp_prop::check_mem_ref): Likewise.
3116         (vrp_prop::visit_stmt): Likewise for local "old_vr".
3117         (vrp_intersect_ranges_1): Likewise for param "vr_1".
3118         (vrp_intersect_ranges): Likewise.
3119         (simplify_stmt_for_jump_threading): Likewise for local "vr".
3120         (vrp_prop::vrp_finalize): Likewise.
3121         * tree-vrp.h (value_range::dump): Add "const" qualifier.
3122         (vrp_intersect_ranges): Add "const" qualifier to params as above.
3123         (extract_range_from_unary_expr): Likewise.
3124         (value_range_constant_singleton): Likewise.
3125         (symbolic_range_p): Likewise.
3126         (copy_value_range): Likewise.
3127         (extract_range_from_binary_expr_1): Likewise.
3128         (range_int_cst_p): Likewise.
3129         (vrp_set_zero_nonzero_bits): Likewise.
3130         (range_int_cst_singleton_p): Likewise.
3132 2018-08-31  Vlad Lazar  <vlad.lazar@arm.com>
3134         * config/aarch64/arm_neon.h (vabsd_s64): New.
3135         (vnegd_s64): Likewise.
3137 2018-08-31  Martin Jambor  <mjambor@suse.cz>
3139         * ipa-cp.c (estimate_local_effects): Replace wrong MAX with MIN.
3141 2018-08-31  Martin Liska  <mliska@suse.cz>
3143         * ipa-icf.c (sem_item::add_type): Use
3144         sem_item::m_type_hash_cache.
3145         * ipa-icf.h: Move the cache from sem_item_optimizer
3146         to sem_item.
3148 2018-08-31  Nathan Sidwell  <nathan@acm.org>
3150         * doc/extend.texi (Backwards Compatibility): Remove implicit
3151         extern C leeway of () being (...).
3153 2018-08-31  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
3155         * ipa-inline.c (can_inline_edge_by_limits_p): Fix typos in comment.
3157 2018-08-31  Segher Boessenkool  <segher@kernel.crashing.org>
3159         PR target/86684
3160         PR target/87149
3161         * config/rs6000/rs6000.md (lround<mode>di2): Gate on TARGET_FPRND.
3163 2018-08-31  Jakub Jelinek  <jakub@redhat.com>
3165         PR middle-end/87138
3166         * expmed.c (expand_mult_const): Use immed_wide_int_const instead of
3167         gen_int_mode.  Formatting fixes.
3169 2018-08-30  Sandra Loosemore  <sandra@codesourcery.com>
3171         * target.def (custom_function_descriptors): Improve documentation.
3172         * doc/tm.texi.in (Trampolines): Expand discussion of function
3173         descriptors and move TARGET_CUSTOM_FUNCTION_DESCRIPTORS to the
3174         beginning of the section.
3175         * doc/tm.texi: Regenerated.
3177 2018-08-30  Jose E. Marchesi  <jose.marchesi@oracle.com>
3179         * cfg.h (class auto_edge_flag): Spell out the template-id of the
3180         base class in the initializer list.  This is a workaround for
3181         building with older GCC.
3182         (class auto_bb_flag): Likewise.
3184 2018-08-30  Aaron Sawdey  <acsawdey@linux.ibm.com>
3186         * config/rs6000/altivec.md (altivec_eq<mode>): Remove star.
3187         (altivec_vcmpequ<VI_char>_p): Remove star.
3188         * config/rs6000/rs6000-string.c (do_load_for_compare): Support
3189         vector load modes.
3190         (expand_strncmp_vec_sequence): New function.
3191         (emit_final_str_compare_vec): New function.
3192         (expand_strn_compare): Add support for vector strncmp.
3193         * config/rs6000/rs6000.opt (-mstring-compare-inline-limit): Change
3194         length specification to bytes.
3195         * config/rs6000/vsx.md (vsx_ld_elemrev_v16qi_internal): Remove star.
3196         (vcmpnezb_p): New pattern.
3197         * doc/invoke.texi (RS/6000 and PowerPC Options): Update documentation
3198         for option -mstring-compare-inline-limit.
3200 2018-08-30  Thiago Macieira  <thiago.macieira@intel.com>
3202         * config/i386/i386.c (PTA_WESTMERE): Remove PTA_AES.
3203         (PTA_SKYLAKE): Add PTA_AES.
3204         (PTA_GOLDMONT): Likewise.
3206 2018-08-29  Jan Hubicka  <jh@suse.cz>
3208         PR lto/86517
3209         * lto-opts.c (lto_write_options): Always stream PIC/PIE mode.
3210         * lto-wrapper.c (merge_and_complain): Fix merging of PIC/PIE.
3212 2018-08-29  Jan Hubicka  <jh@suse.cz>
3214         * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not follow
3215         TYPE_STUB_DECL.
3216         (hash_tree): Do not visit TYPE_STUB_DECL.
3217         * tree-streamer-out.c (write_ts_type_common_tree_pointers): Do not
3218         stream TYPE_STUB_DECL.
3219         * tree-streamer-in.c (lto_input_ts_type_common_tree_pointers): Likewise.
3220         * ipa-utils.h (type_with_linkage_p): Do not rely on TYPE_STUB_DECL
3221         after free_lang_data.
3222         (type_in_anonymous_namespace_p): Likewise.
3224 2018-08-29  Jan Hubicka  <jh@suse.cz>
3226         * sreal.h (SREAL_PART_BITS): Change to 31; remove seemingly unnecessary
3227         comment that it has to be even number.
3228         (class sreal): Change m_sig type to int32_t.
3229         * sreal.c (sreal::dump, sreal::to_int, opreator+, operator-): Use
3230         int64_t for temporary calculations.
3231         (sreal_verify_basics): Drop one bit from minimum and maximum.
3233 2018-08-30  Richard Biener  <rguenther@suse.de>
3235         PR tree-optimization/87147
3236         * tree-ssa-sccvn.c (SSA_VISITED): New function.
3237         (visit_phi): When the degenerate result is from the backedge and
3238         we didn't visit its definition yet drop to VARYING.
3239         (do_rpo_vn): Properly mark blocks with incoming backedges as executable.
3241 2018-08-29  Jan Hubicka  <jh@suse.cz>
3243         * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not walk
3244         DECL_VINDEX.
3245         (hash_tree): Likewise.
3247 2018-08-29  Jan Hubicka  <jh@suse.cz>
3249         * tree.c (find_decls_types_r): Walk also TYPE_NEXT_PTR_TO
3250         and TYPE_NEXT_REF_TO.
3252 2018-08-29  Jan Hubicka  <jh@suse.cz>
3254         * sreal.h (SREAL_PART_BITS): Change to 31; remove seemingly unnecessary
3255         comment that it has to be even number.
3256         (class sreal): Change m_sig type to int32_t.
3257         * sreal.c (sreal::dump, sreal::to_int, opreator+, operator-): Use
3258         int64_t for temporary calculations.
3259         (sreal_verify_basics): Drop one bit from minimum and maximum.
3261 2018-08-30  Tamar Christina  <tamar.christina@arm.com>
3263         * config/aarch64/aarch64.c (aarch64_expand_movmem): Set TImode max.
3265 2018-08-30  Vlad Lazar  <vlad.lazar@arm.com>
3267         PR middle-end/86995
3268         * expmed.c (canonicalize_comparison): Use wi::sub instead of wi::add
3269         if to_add is negative.
3271 2018-08-29  Bernd Edlinger  <bernd.edlinger@hotmail.de>
3273         PR middle-end/87053
3274         * builtins.c (c_strlen): Improve range checks.
3276 2018-08-29  Martin Sebor  <msebor@redhat.com>
3277             Jeff Law  <law@redhat.com>
3279         PR tree-optimization/86714
3280         PR tree-optimization/86711
3281         * builtins.c (c_strlen): Add arguments to call to string_constant.
3282         * expr.c (string_constant): Add argument.  Detect missing nul
3283         terminator and outermost declaration it's missing in.
3284         * expr.h (string_constant): Add argument.
3285         * fold-const.c (read_from_constant_string): Add arguments to call to
3286         string_constant.
3287         (c_getstr): Likewise.
3288         * tree-ssa-forwprop.c (simplify_builtin_call): Likewise.
3289         to string_constant.
3290         * tree-ssa-strlen.c (get_stridx): Likewise.
3292 2018-08-29  Jan Hubicka  <jh@suse.cz>
3294         * tree-streamer-in.c (lto_input_ts_function_decl_tree_pointers):
3295         Do not stream DECL_VINDEX.
3296         * tree-streamer-out.c (write_ts_function_decl_tree_pointers): Likewise.
3297         * tree.c (free_lang_data_in_decl): Clear DECL_VINDEX.
3298         (decl_function_context): Use DECL_VIRTUAL_P rather than DECL_VINDEX.
3300 2018-08-29  Richard Biener  <rguenther@suse.de>
3302         * tree-ssa-sccvn.c (vuse_ssa_val): Return NULL for unvisited
3303         virtual operands that are not default defs to honor region
3304         boundaries.
3305         (rpo_vn_valueize): Remove ineffective code here.
3307 2018-08-29  Richard Biener  <rguenther@suse.de>
3309         PR tree-optimization/87132
3310         * tree-ssa-alias.c (get_continuation_for_phi): Do not translate
3311         when skipping defs reachable over backedges.
3313 2018-08-29  Richard Biener  <rguenther@suse.de>
3315         * tree-core.h: Document use of deprecated_flag in SSA_NAME.
3316         * tree.h (SSA_NAME_POINTS_TO_READONLY_MEMORY): Define.
3317         * tree-into-ssa.c (pass_build_ssa::execute): Initialize
3318         function parameters SSA_NAME_POINTS_TO_READONLY_MEMORY from fnspec.
3319         * tree-ssa-sccvn.c (const_parms, init_const_parms): Remove.
3320         (vn_reference_lookup_3): Remove use of const_parms.
3321         (free_rpo_vn): Do not free const_parms.
3322         (do_rpo_vn): Do not call init_const_parms.
3323         * tree-ssa-alias.c (refs_may_alias_p_1): Honor
3324         SSA_NAME_POINTS_TO_READONLY_MEMORY.
3325         (call_may_clobber_ref_p_1): Likewise.
3327 2018-08-29  Alexander Monakov  <amonakov@ispras.ru>
3329         PR other/86726
3330         * invoke.texi (Optimization Options): List -ftree-scev-cprop.
3331         (-O): Ditto.
3332         (-ftree-scev-cprop): Document.
3334 2018-08-29  Jan Hubicka  <jh@suse.cz>
3336         * sreal.h (normalize, normalize_up, normalize_down): Add new_sig/new_exp
3337         parameters.
3338         (sreal constructor): Update.
3339         * sreal.c (sreal:operator+, sreal:operator-, sreal:operator*,
3340         sreal:operator/): Update.
3342 2018-08-29  Martin Liska  <mliska@suse.cz>
3344         * tree-switch-conversion.c (switch_conversion::expand):
3345         Strenghten assumption about gswitch statements.
3347 2018-08-29  Richard Biener  <rguenther@suse.de>
3349         PR tree-optimization/87117
3350         * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_stmt): Only
3351         re-value-number released SSA VDEFs.
3353 2018-08-29  Richard Biener  <rguenther@suse.de>
3355         PR tree-optimization/87126
3356         * tree-ssa-sccvn.c (vn_reference_insert): Remove assert.
3358 2018-08-28  Jim Wilson  <jimw@sifive.com>
3360         * config/riscv/pic.md: Rewrite.
3361         * config/riscv/riscv.c (riscv_address_insns): Return cost of 3 for
3362         invalid address.
3363         * config/riscv/riscv.md (ZERO_EXTEND_LOAD): Delete.
3364         (SOFTF, default_load, softload, softstore): New.
3366 2018-08-28  Jeff Law  <law@redhat.com>
3368         * fold-const.c (fold_binary_loc): Remove recently added assert.
3370 2018-08-28  Joern Rennecke  <joern.rennecke@riscy-ip.com>
3372         * genpreds.c (write_predicate_subfunction): Also add ATTRIBUTE_UNUSED
3373         to OP parmeter of generated function.
3375 2018-08-28  MCC CS  <deswurstes@users.noreply.github.com>
3377         PR tree-optimization/87009
3378         * match.pd: Add boolean optimizations.
3380 2018-08-28  Martin Sebor  <msebor@redhat.com>
3382         PR middle-end/86631
3383         * calls.c (alloc_max_size): Treat HOST_WIDE_INT special.
3384         * gimple-ssa-warn-alloca.c (adjusted_warn_limit): New function.
3385         (pass_walloca::gate): Use it.
3386         (alloca_call_type): Same.
3387         (pass_walloca::execute): Same.
3388         * stor-layout.c (layout_decl): Treat HOST_WIDE_INT special.
3390 2018-08-28  David Malcolm  <dmalcolm@redhat.com>
3392         * dumpfile.h (ATTRIBUTE_GCC_DUMP_PRINTF): Change version check on
3393         GCC_VERSION for usage of "__gcc_dump_printf__" format from
3394         >= 3005 to >= 9000.
3396 2018-08-28  Richard Biener  <rguenther@suse.de>
3398         PR tree-optimization/87124
3399         * tree-ssa-sccvn.c (vn_lookup_simplify_result): Guard against
3400         constants before looking up avail.
3402 2018-08-28  Jakub Jelinek  <jakub@redhat.com>
3404         PR middle-end/87099
3405         * calls.c (maybe_warn_nonstring_arg): Punt early if
3406         warn_stringop_overflow is zero.  Don't call get_range_strlen
3407         on 3rd argument, keep iterating until lenrng[1] is INTEGER_CST.
3408         Swap comparison operands to have constants on rhs.  Only use
3409         lenrng[1] if non-NULL and INTEGER_CST.  Don't uselessly
3410         increment lenrng[0].
3412 2018-08-28  Richard Sandiford  <richard.sandiford@arm.com>
3414         * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Fix unguarded
3415         use of tree_to_shwi.  Remove duplicated test for the size being
3416         a whole number of bytes.
3418 2018-08-28  Richard Biener  <rguenther@suse.de>
3420         PR tree-optimization/87117
3421         * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_cleanup):
3422         Handle removed stmt without LHS (GIMPLE_NOP).
3424 2018-08-28  Richard Biener  <rguenther@suse.de>
3426         PR tree-optimization/87117
3427         * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Exclude
3428         void which is is_gimple_reg_type by checking for COMPLETE_TYPE_P.
3430 2018-08-28  Richard Biener  <rguenther@suse.de>
3432         PR tree-optimization/87117
3433         * tree-ssa-pre.c (compute_avail): Do not make expressions
3434         with predicated values available.
3435         (get_expr_value_id): Assert we do not run into predicated value
3436         expressions.
3438 2018-08-28  Richard Biener  <rguenther@suse.de>
3440         PR tree-optimization/87117
3441         * tree-ssa-operands.c (add_stmt_operand): STRING_CST may
3442         get virtual operands.
3443         (get_expr_operands): Handle STRING_CST like other decls.
3445 2018-08-28  Martin Liska  <mliska@suse.cz>
3447         * tree.h: Update documentation of fndecl_built_in_p
3448         functions.
3451 2018-08-27  Jeff Law  <law@redhat.com>
3452         PR tree-optimization/87110
3453         * tree-ssa-dse.c (compute_trims): Handle non-constant
3454         TYPE_SIZE_UNIT.
3456 2018-08-27  Martin Sebor  <msebor@redhat.com>
3458         PR tree-optimization/86914
3459         * tree-ssa-strlen.c (maybe_set_strlen_range): Avoid MEM_REF.
3461 2018-08-27  Martin Sebor  <msebor@redhat.com>
3463         PR tree-optimization/87112
3464         * builtins.c (expand_builtin_strnlen): Convert c_strlen result to
3465         the type of the bound argument.
3467 2018-08-27  Jeff Law  <law@redhat.com>
3469         * tree-ssa-dse.c (compute_trims): Handle case where the reference's
3470         type does not have a TYPE_SIZE_UNIT.
3472 2018-08-27  Steve Ellcey  <sellcey@cavium.com>
3474         * config/aarch64/aarch64-speculation.cc: Replace include of cfg.h
3475         with include of backend.h.
3477 2018-08-27  Richard Biener  <rguenther@suse.de>
3479         PR tree-optimization/86927
3480         * tree-vect-loop.c (vect_create_epilog_for_reduction): Properly
3481         use const cond reduction code.
3483 2018-08-27  Alexander Monakov  <amonakov@ispras.ru>
3485         PR tree-optimization/85758
3486         * match.pd ((X & Y) ^ Y): Add :s qualifier to inner expression.
3488 2018-08-27  David Malcolm  <dmalcolm@redhat.com>
3490         PR c++/87091
3491         * diagnostic-show-locus.c (class layout_range): Update for
3492         conversion of show_caret_p to a tri-state.
3493         (layout_range::layout_range): Likewise.
3494         (make_range): Likewise.
3495         (layout::maybe_add_location_range): Likewise.
3496         (layout::should_print_annotation_line_p): Don't show annotation
3497         lines for ranges that are SHOW_LINES_WITHOUT_RANGE.
3498         (layout::get_state_at_point): Update for conversion of
3499         show_caret_p to a tri-state.  Bail out early for
3500         SHOW_LINES_WITHOUT_RANGE, so that such ranges don't affect
3501         underlining or source colorization.
3502         (gcc_rich_location::add_location_if_nearby): Update for conversion
3503         of show_caret_p to a tri-state.
3504         (selftest::test_one_liner_multiple_carets_and_ranges): Likewise.
3505         (selftest::test_one_liner_fixit_replace_equal_secondary_range):
3506         Likewise.
3507         (selftest::test_one_liner_labels): Likewise.
3508         * gcc-rich-location.c (gcc_rich_location::add_expr): Update for
3509         conversion of show_caret_p to a tri-state.
3510         * pretty-print.c (text_info::set_location): Likewise.
3511         * pretty-print.h (text_info::set_location): Likewise.
3512         * substring-locations.c (format_warning_n_va): Likewise.
3513         * tree-diagnostic.c (default_tree_printer): Likewise.
3514         * tree-pretty-print.c (newline_and_indent): Likewise.
3516 2018-08-27  David Malcolm  <dmalcolm@redhat.com>
3518         PR c++/87091
3519         * diagnostic-show-locus.c (get_line_span_for_fixit_hint): Show the
3520         line above for line-insertion fix-it hints.
3521         (selftest::test_fixit_insert_containing_newline): Update the
3522         expected results, and add a test with line-numbering enabled.
3524 2018-08-27  Martin Liska  <mliska@suse.cz>
3526         PR sanitizer/86962
3527         * sanopt.c (sanitize_rewrite_addressable_params): Ignore
3528         params with DECL_HAS_VALUE_EXPR_P.
3530 2018-08-27  Martin Liska  <mliska@suse.cz>
3532         * config/i386/i386.c (ix86_expand_set_or_movmem): Dump
3533         selected expansion strategy.
3535 2018-08-27  Martin Liska  <mliska@suse.cz>
3537         * builtins.h (is_builtin_fn): Remove and fndecl_built_in_p.
3538         * builtins.c (is_builtin_fn): Likewise.
3539         * attribs.c (diag_attr_exclusions): Use new function
3540         fndecl_built_in_p and remove check for FUNCTION_DECL if
3541         possible.
3542         (builtin_mathfn_code): Likewise.
3543         (fold_builtin_expect): Likewise.
3544         (fold_call_expr): Likewise.
3545         (fold_builtin_call_array): Likewise.
3546         (fold_call_stmt): Likewise.
3547         (set_builtin_user_assembler_name): Likewise.
3548         (is_simple_builtin): Likewise.
3549         * calls.c (gimple_alloca_call_p): Likewise.
3550         (maybe_warn_nonstring_arg): Likewise.
3551         * cfgexpand.c (expand_call_stmt): Likewise.
3552         * cgraph.c (cgraph_update_edges_for_call_stmt_node): Likewise.
3553         (cgraph_edge::verify_corresponds_to_fndecl): Likewise.
3554         (cgraph_node::verify_node): Likewise.
3555         * cgraphclones.c (build_function_decl_skip_args): Likewise.
3556         (cgraph_node::create_clone): Likewise.
3557         * config/arm/arm.c (arm_insert_attributes): Likewise.
3558         * config/i386/i386.c (ix86_gimple_fold_builtin): Likewise.
3559         * dse.c (scan_insn): Likewise.
3560         * expr.c (expand_expr_real_1): Likewise.
3561         * fold-const.c (operand_equal_p): Likewise.
3562         (fold_binary_loc): Likewise.
3563         * gimple-fold.c (gimple_fold_stmt_to_constant_1): Likewise.
3564         * gimple-low.c (lower_stmt): Likewise.
3565         * gimple-pretty-print.c (dump_gimple_call): Likewise.
3566         * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
3567         Likewise.
3568         * gimple.c (gimple_build_call_from_tree): Likewise.
3569         (gimple_call_builtin_p): Likewise.
3570         (gimple_call_combined_fn): Likewise.
3571         * gimplify.c (gimplify_call_expr): Likewise.
3572         (gimple_boolify): Likewise.
3573         (gimplify_modify_expr): Likewise.
3574         (gimplify_addr_expr): Likewise.
3575         * hsa-gen.c (gen_hsa_insns_for_call): Likewise.
3576         * ipa-cp.c (determine_versionability): Likewise.
3577         * ipa-fnsummary.c (compute_fn_summary): Likewise.
3578         * ipa-param-manipulation.c (ipa_modify_formal_parameters): Likewise.
3579         * ipa-split.c (visit_bb): Likewise.
3580         (split_function): Likewise.
3581         * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
3582         * lto-cgraph.c (input_node): Likewise.
3583         * lto-streamer-out.c (write_symbol): Likewise.
3584         * omp-low.c (setjmp_or_longjmp_p): Likewise.
3585         (lower_omp_1): Likewise.
3586         * predict.c (strip_predict_hints): Likewise.
3587         * print-tree.c (print_node): Likewise.
3588         * symtab.c (symtab_node::output_to_lto_symbol_table_p): Likewise.
3589         * trans-mem.c (is_tm_irrevocable): Likewise.
3590         (is_tm_load): Likewise.
3591         (is_tm_simple_load): Likewise.
3592         (is_tm_store): Likewise.
3593         (is_tm_simple_store): Likewise.
3594         (is_tm_abort): Likewise.
3595         (tm_region_init_1): Likewise.
3596         * tree-call-cdce.c (gen_shrink_wrap_conditions): Likewise.
3597         * tree-cfg.c (verify_gimple_call): Likewise.
3598         (move_stmt_r): Likewise.
3599         (stmt_can_terminate_bb_p): Likewise.
3600         * tree-eh.c (lower_eh_constructs_2): Likewise.
3601         * tree-if-conv.c (if_convertible_stmt_p): Likewise.
3602         * tree-inline.c (remap_gimple_stmt): Likewise.
3603         (copy_bb): Likewise.
3604         (estimate_num_insns): Likewise.
3605         (fold_marked_statements): Likewise.
3606         * tree-sra.c (scan_function): Likewise.
3607         * tree-ssa-ccp.c (surely_varying_stmt_p): Likewise.
3608         (optimize_stack_restore): Likewise.
3609         (pass_fold_builtins::execute): Likewise.
3610         * tree-ssa-dce.c (mark_stmt_if_obviously_necessary): Likewise.
3611         (mark_all_reaching_defs_necessary_1): Likewise.
3612         * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Likewise.
3613         * tree-ssa-forwprop.c (simplify_builtin_call): Likewise.
3614         (pass_forwprop::execute): Likewise.
3615         * tree-ssa-loop-im.c (stmt_cost): Likewise.
3616         * tree-ssa-math-opts.c (pass_cse_reciprocals::execute): Likewise.
3617         * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Likewise.
3618         * tree-ssa-strlen.c (get_string_length): Likewise.
3619         * tree-ssa-structalias.c (handle_lhs_call): Likewise.
3620         (find_func_aliases_for_call): Likewise.
3621         * tree-ssa-ter.c (find_replaceable_in_bb): Likewise.
3622         * tree-stdarg.c (optimize_va_list_gpr_fpr_size): Likewise.
3623         * tree-tailcall.c (find_tail_calls): Likewise.
3624         * tree.c (need_assembler_name_p): Likewise.
3625         (free_lang_data_in_decl): Likewise.
3626         (get_call_combined_fn): Likewise.
3627         * ubsan.c (is_ubsan_builtin_p): Likewise.
3628         * varasm.c (incorporeal_function_p): Likewise.
3629         * tree.h (DECL_BUILT_IN): Remove and replace with
3630         fndecl_built_in_p.
3631         (DECL_BUILT_IN_P): Transfort to fndecl_built_in_p.
3632         (fndecl_built_in_p): New.
3634 2018-08-27  Martin Liska  <mliska@suse.cz>
3636         PR tree-optimization/86847
3637         * tree-switch-conversion.c (switch_decision_tree::dump_case_nodes):
3638         Dump also subtree probability.
3639         (switch_decision_tree::do_jump_if_equal): New function.
3640         (switch_decision_tree::emit_case_nodes): Handle special
3641         situations in balanced tree that can be emitted much simpler.
3642         Fix calculation of probabilities that happen in tree expansion.
3643         * tree-switch-conversion.h (struct cluster): Add
3644         is_single_value_p.
3645         (struct simple_cluster): Likewise.
3646         (struct case_tree_node): Add new function has_child.
3647         (do_jump_if_equal): New.
3649 2018-08-27  Martin Liska  <mliska@suse.cz>
3651         * tree-switch-conversion.c (bit_test_cluster::find_bit_tests):
3652         Add new argument to bit_test_cluster constructor.
3653         (bit_test_cluster::emit): Set bits really number of values
3654         handlel by a test.
3655         (bit_test_cluster::hoist_edge_and_branch_if_true): Add
3656         probability argument.
3657         * tree-switch-conversion.h (struct bit_test_cluster):
3658         Add m_handles_entire_switch.
3660 2018-08-27  Martin Liska  <mliska@suse.cz>
3662         PR tree-optimization/86702
3663         * tree-switch-conversion.c (jump_table_cluster::emit):
3664         Make probabilities even for values in jump table
3665         according to number of cases handled.
3666         (switch_decision_tree::compute_cases_per_edge): Pass
3667         argument to reset_out_edges_aux function.
3668         (switch_decision_tree::analyze_switch_statement): Likewise.
3669         * tree-switch-conversion.h (switch_decision_tree::reset_out_edges_aux):
3670         Make it static.
3672 2018-08-27  Martin Liska  <mliska@suse.cz>
3674         * cfgexpand.c (expand_asm_stmt): Use label_to_block and pass
3675         cfun argument explicitly.
3676         * gimple-pretty-print.c (dump_gimple_switch): Likewise.
3677         * hsa-gen.c (gen_hsa_insns_for_switch_stmt): Use new
3678         function gimple_switch_default_bb.
3679         (convert_switch_statements):
3680         (expand_builtins):
3681         * ipa-fnsummary.c (set_switch_stmt_execution_predicate):
3682         * stmt.c (label_to_block_fn): Use label_to_block and pass
3683         cfun argument explicitly and use gimple_switch_label_bb.
3684         (expand_case): Likewise.
3685         * tree-cfg.c (lower_phi_internal_fn): Use label_to_block and pass
3686         cfun argument explicitly. Likewise.
3687         (make_edges_bb): Likewise.
3688         (make_cond_expr_edges): Likewise.
3689         (get_cases_for_edge): Likewise.
3690         (make_gimple_switch_edges): Likewise.
3691         (label_to_block_fn): Likewise.
3692         (label_to_block): Likewise.
3693         (make_goto_expr_edges): Likewise.
3694         (make_gimple_asm_edges): Likewise.
3695         (main_block_label): Likewise.
3696         (group_case_labels_stmt): Likewise.
3697         (find_taken_edge_computed_goto): Likewise.
3698         (find_taken_edge_switch_expr): Likewise.
3699         (gimple_verify_flow_info): Likewise.
3700         (gimple_redirect_edge_and_branch): Likewise.
3701         (gimple_switch_label_bb): New function.
3702         (gimple_switch_default_bb): Likewise.
3703         (gimple_switch_edge): Likewise.
3704         (gimple_switch_default_edge): Likewise.
3705         * tree-cfg.h (label_to_block_fn): Remove and replace ...
3706         (label_to_block): ... with this.
3707         (gimple_switch_label_bb): New.
3708         (gimple_switch_default_bb): Likewise.
3709         (gimple_switch_edge): Likewise.
3710         (gimple_switch_default_edge): Likewise.
3711         * tree-cfgcleanup.c (convert_single_case_switch): Use
3712         new gimple functions and pass new argument to label_to_block.
3713         (cleanup_control_flow_bb):
3714         * tree-eh.c (make_eh_dispatch_edges): Use label_to_block and pass
3715         cfun argument explicitly.
3716         (make_eh_edges): Likewise.
3717         (redirect_eh_dispatch_edge): Likewise.
3718         (lower_resx): Likewise.
3719         (lower_eh_dispatch): Likewise.
3720         (maybe_remove_unreachable_handlers): Likewise.
3721         (unsplit_eh): Likewise.
3722         (cleanup_empty_eh): Likewise.
3723         (verify_eh_edges): Likewise.
3724         (verify_eh_dispatch_edge): Likewise.
3725         * tree-ssa-dom.c (record_edge_info): Likewise.
3726         * tree-ssa-forwprop.c (simplify_gimple_switch_label_vec): Likewise.
3727         * tree-ssa-threadedge.c (thread_around_empty_blocks): Likewise.
3728         (thread_through_normal_block): Likewise.
3729         * tree-ssa-uncprop.c (associate_equivalences_with_edges): Likewise.
3730         * tree-ssa-uninit.c (convert_control_dep_chain_into_preds):
3731         * tree-switch-conversion.c (switch_conversion::collect): Use new
3732         gimple functions.
3733         (switch_conversion::check_final_bb): Likewise.
3734         (switch_conversion::gather_default_values): Pass new argument
3735         to label_to_block.
3736         (switch_conversion::build_constructors): Likewise.
3737         (switch_decision_tree::compute_cases_per_edge): Use new
3738         gimple_switch_edge function.
3739         (switch_decision_tree::analyze_switch_statement): Pass new argument
3740         to label_to_block.
3741         (switch_decision_tree::try_switch_expansion): Use
3742         gimple_switch_default_edge.
3743         * tree-vrp.c (find_switch_asserts): Pass new argument
3744         to label_to_block.
3745         * vr-values.c (vr_values::vrp_visit_switch_stmt): Likewise.
3746         (vr_values::simplify_switch_using_ranges): Likewise.
3748 2018-08-27  Richard Biener  <rguenther@suse.de>
3750         * cfganal.h (rev_post_order_and_mark_dfs_back_seme): Declare.
3751         * cfganal.c (rev_post_order_and_mark_dfs_back_seme): New function.
3753         * tree-ssa-sccvn.h (struct vn_pval): New structure.
3754         (struct vn_nary_op_s): Add unwind_to member.  Add
3755         predicated_values flag and put result into a union together
3756         with a linked list of vn_pval.
3757         (struct vn_ssa_aux): Add name member to make maintaining
3758         a map of SSA name to vn_ssa_aux possible.  Remove no longer
3759         needed info, dfsnum, low, visited, on_sccstack, use_processed
3760         and range_info_anti_range_p members.
3761         (run_scc_vn, vn_eliminate, free_scc_vn, vn_valueize): Remove.
3762         (do_rpo_vn, run_rpo_vn, eliminate_with_rpo_vn, free_rpo_vn):
3763         New functions.
3764         (vn_valueize): New global.
3765         (vn_context_bb): Likewise.
3766         (VN_INFO_RANGE_INFO, VN_INFO_ANTI_RANGE_P, VN_INFO_RANGE_TYPE,
3767         VN_INFO_PTR_INFO): Remove.
3768         * tree-ssa-sccvn.c: ... (rewrite)
3769         (pass_fre::execute): For -O2+ initialize loops and run
3770         RPO VN in optimistic mode (iterating).  For -O1 and -Og
3771         run RPO VN in non-optimistic mode.
3772         * params.def (PARAM_SCCVN_MAX_SCC_SIZE): Remove.
3773         (PARAM_RPO_VN_MAX_LOOP_DEPTH): Add.
3774         * doc/invoke.texi (sccvn-max-scc-size): Remove.
3775         (rpo-vn-max-loop-depth): Document.
3776         * tree-ssa-alias.c (walk_non_aliased_vuses): Stop walking
3777         when valuezing the VUSE signals we walked out of the region.
3778         * tree-ssa-pre.c (phi_translate_1): Ignore predicated values.
3779         (phi_translate): Set VN context block to use for availability
3780         lookup.
3781         (compute_avail): Likewise.
3782         (pre_valueize): New function.
3783         (pass_pre::execute): Adjust to the RPO VN API.
3785         * tree-ssa-loop-ivcanon.c: Include tree-ssa-sccvn.h.
3786         (propagate_constants_for_unrolling): Remove.
3787         (tree_unroll_loops_completely): Perform value-numbering
3788         on the unrolled bodies loop parent.
3790 2018-08-27  Richard Biener  <rguenther@suse.de>
3792         * tree-ssa-pre.c (compute_antic): Re-use inverted postorder
3793         for partial antic compute.
3795 2018-08-27  Jakub Jelinek  <jakub@redhat.com>
3797         PR rtl-optimization/87065
3798         * combine.c (simplify_if_then_else): Formatting fix.
3799         (if_then_else_cond): Guard MULT optimization with SCALAR_INT_MODE_P
3800         check.
3801         (known_cond): Don't return const_true_rtx for vector modes.  Use
3802         CONST0_RTX instead of const0_rtx.  Formatting fixes.
3804 2018-08-27  Martin Liska  <mliska@suse.cz>
3806         PR gcov-profile/87069
3807         * gcov.c (process_file): Record files already processed
3808         and warn about a file being processed multiple times.
3810 2018-08-27  Martin Liska  <mliska@suse.cz>
3812         PR driver/83193
3813         * config/aarch64/aarch64.c (aarch64_override_options_internal):
3814         Set default values for x_aarch64_*_string strings.
3815         * config/aarch64/aarch64.opt: Remove --{march,mcpu,mtune}==
3816         prefix.  For -mabi do not print '=ABI' in help and use
3817         <option_value> format for -msve-vector-bits and -moverride
3818         options.
3820 2018-08-26  Jeff Law  <law@redhat.com>
3822         * config/mips/frame-header-opt.c: Include "backend.h" rather than
3823         "cfg.h"
3825 2018-08-26  Marek Polacek  <polacek@redhat.com>
3827         PR c++/87029, Implement -Wredundant-move.
3828         * doc/invoke.texi: Document -Wredundant-move.
3830 2018-08-25  Martin Sebor  <msebor@redhat.com>
3832         PR tree-optimization/87059
3833         * builtins.c (expand_builtin_strncmp): Convert MIN_EXPR operand
3834         to the same type as the other.
3835         * fold-const.c (fold_binary_loc): Assert expectation.
3837 2018-08-25  Iain Sandoe  <iain@sandoe.co.uk>
3839         * config/darwin.c (machopic_legitimize_pic_address): Clean up
3840         extraneous parentheses, dead code section and formatting.
3842 2018-08-24  David Malcolm  <dmalcolm@redhat.com>
3844         PR c++/87091
3845         * diagnostic-show-locus.c (layout::layout): Ensure the margin is
3846         wide enough for jumps in the line-numbering to be visible.
3847         (layout::print_gap_in_line_numbering): New member function.
3848         (layout::calculate_line_spans): When using line numbering, merge
3849         line spans that are only 1 line apart.
3850         (diagnostic_show_locus): When printing line numbers, show gaps in
3851         line numbering directly, rather than printing headers.
3852         (selftest::test_diagnostic_show_locus_fixit_lines): Add test of
3853         line-numbering with multiple line spans.
3854         (selftest::test_fixit_insert_containing_newline_2): Add test of
3855         line-numbering, in which the spans are close enough to be merged.
3857 2018-08-24  Aldy Hernandez  <aldyh@redhat.com>
3859         * gimple-ssa-evrp-analyze.c (set_ssa_range_info): Pass value_range
3860         to range_includes_zero_p.  Do not special case VR_ANTI_RANGE.
3861         * tree-vrp.c (range_is_nonnull): Remove.
3862         (range_includes_zero_p): Accept value_range instead of min/max.
3863         (extract_range_from_binary_expr_1): Do not early bail on
3864         POINTER_PLUS_EXPR.
3865         Use range_includes_zero_p instead of range_is_nonnull.
3866         (extract_range_from_unary_expr): Use range_includes_zero_p instead
3867         of range_is_nonnull.
3868         (vrp_meet_1): Pass value_range to range_includes_zero_p.  Do not
3869         special case VR_ANTI_RANGE.
3870         (vrp_finalize): Same.
3871         * tree-vrp.h (range_includes_zero_p): Pass value_range as argument
3872         instead of min/max.
3873         (range_is_nonnull): Remove.
3874         * vr-values.c (vrp_stmt_computes_nonzero): Use
3875         range_includes_zero_p instead of range_is_nonnull.
3876         (extract_range_basic): Pass value_range to range_includes_zero_p
3877         instead of range_is_nonnull.
3879 2018-08-24  Uros Bizjak  <ubizjak@gmail.com>
3881         * emit-rtl.c (init_emit_once): Do not emit MODE_POINTER_BOUNDS RTXes.
3882         * emit-rtl.h (rtl_data): Remove return_bnd.
3883         * explow.c (trunc_int_for_mode): Do not handle POINTER_BOUNDS_MODE_P.
3884         * function.c (diddle_return_value): Do not handle crtl->return_bnd.
3885         * genmodes.c (complete_mode): Do not handle MODE_POINTER_BOUNDS.
3886         (POINTER_BOUNDS_MODE): Remove definition.
3887         (make_pointer_bounds_mode): Remove.
3888         (get_mode_class): Do not handle MODE_POINTER_BOUNDS.
3889         * machmode.h (POINTER_BOUNDS_MODE_P): Remove definition.
3890         (scalare_mode::includes_p): Do not handle MODE_POINTER_BOUNDS.
3891         * mode-classes.def: Do not define MODE_POINTER_BOUNDS.
3892         * stor-layout.c (int_mode_for_mode): Do not handle MODE_POINTER_BOUNDS.
3893         * tree-core.h (enum tree_index): Remove TI_POINTER_BOUNDS_TYPE.
3894         * varasm.c (output_constant_pool_2): Do not handle MODE_POINTER_BOUNDS.
3896         * config/i386/i386-modes.def (BND32, BND64): Remove.
3897         * config/i386/i386.c (dbx_register_map): Remove bound registers.
3898         (dbx64_register_map): Ditto.
3899         (svr4_dbx_register_map): Ditto.
3900         (indirect_thunk_bnd_needed): Remove.
3901         (indirect_thunks_bnd_used): Ditto.
3902         (indirect_return_bnd_needed): Ditto.
3903         (indirect_return_via_cx_bnd): Ditto.
3904         (enum indirect_thunk_prefix): Remove indirect_thunk_prefix_bnd.
3905         (indirect_thunk_name): Remove handling of indirect_thunk_prefix_bnd.
3906         (output_indirect_thunk): Ditto.  Remove need_prefix argument.
3907         (output_indirect_thunk_function): Remove handling of
3908         indirect_return_bnd_needed, indirect_return_via_cx_bnd,
3909         indirect_thunk_bnd_needed and indirect_thunks_bnd_used variables.
3910         (ix86_save_reg): Remove handling of crtl->return_bnd.
3911         (ix86_legitimate_constant_p): Remove handling of POINTER_BOUNDS_MODE_P.
3912         (ix86_print_operand_address_as): Remove handling of UNSPEC_BNDMK_ADDR
3913         and UNSPEC_BNDLX_ADDR.
3914         (ix86_output_indirect_branch_via_reg): Remove handling of
3915         indirect_thunk_prefix_bnd.
3916         (ix86_output_indirect_branch_via_push): Ditto.
3917         (ix86_output_function_return): Ditto.
3918         (ix86_output_indirect_function_return): Ditto.
3919         (avoid_func_arg_motion): Do not handle UNSPEC_BNDSTX.
3920         * config/i386/i386.h (FIXED_REGISTERS): Remove bound registers.
3921         (CALL_USED_REGISTERS): Ditto.
3922         (REG_ALLOC_ORDER): Update for removal of bound registers.
3923         (HI_REGISTER_NAMES): Ditto.
3924         * config/i386/i386.md (UNSPEC_BNDMK, UNSPEC_BNDMK_ADDR, UNSPEC_BNDSTX)
3925         (UNSPEC_BNDLDX, UNSPEC_BNDLDX_ADDR, UNSPEC_BNDCL, UNSPEC_BNDCU)
3926         (UNSPEC_BNDCN, UNSPEC_MPX_FENCE): Remove.
3927         (BND0_REG, BND1_REG, BND2_REG, BND3_REG): Remove
3928         (FIRST_PSEUDO_REG): Update.
3929         (BND): Remove mode iterator.
3930         * config/i386/predicates.md (bnd_mem_operator): Remove.
3932 2018-08-24  Richard Sandiford  <richard.sandiford@arm.com>
3934         * tree-vect-stmts.c (vectorizable_bswap): Handle variable-length
3935         vectors.
3937 2018-08-24  Richard Sandiford  <richard.sandiford@arm.com>
3939         * tree-vect-slp.c (vect_transform_slp_perm_load): Separate out
3940         the case in which the permute needs only a single element and
3941         repeats for every vector of the result.  Extend that case to
3942         handle variable-length vectors.
3943         * tree-vect-stmts.c (vectorizable_load): Update accordingly.
3945 2018-08-24  H.J. Lu  <hongjiu.lu@intel.com>
3947         PR debug/79342
3948         * dwarf2out.c (save_macinfo_strings): Call set_indirect_string
3949         on DW_MACINFO_start_file for -gsplit-dwarf -g3.
3951 2018-08-24  Richard Biener  <rguenther@suse.de>
3953         * cfg.h (struct control_flow_graph): Add edge_flags_allocated and
3954         bb_flags_allocated members.
3955         (auto_flag): New RAII class for allocating flags.
3956         (auto_edge_flag): New RAII class for allocating edge flags.
3957         (auto_bb_flag): New RAII class for allocating bb flags.
3958         * cfgloop.c (verify_loop_structure): Allocate temporary edge
3959         flag dynamically.
3960         * cfganal.c (dfs_enumerate_from): Remove use of visited sbitmap
3961         in favor of temporarily allocated BB flag.
3962         * hsa-brig.c: Re-order includes.
3963         * hsa-dump.c: Likewise.
3964         * hsa-regalloc.c: Likewise.
3965         * print-rtl.c: Likewise.
3966         * profile-count.c: Likewise.
3968 2018-08-24  Segher Boessenkool  <segher@kernel.crashing.org>
3970         PR target/86989
3971         * config/rs6000/rs6000.c (toc_relative_expr_p): Check that the base is
3972         the TOC register.
3974 2018-08-24  Aldy Hernandez  <aldyh@redhat.com>
3976         PR 87073/bootstrap
3977         * wide-int-range.cc (wide_int_range_div): Do not ignore result
3978         from wide_int_range_multiplicative_op.
3980 2018-08-23  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
3982         * tree-vect-data-refs.c (vect_grouped_store_supported): Fix typo
3983         "permutaion".
3985 2018-08-23  Giuliano Belinassi  <giuliano.belinassi@usp.br>
3987         * genmatch.c (parser::parse_operation): Fix typo 'exapnded'
3988         to 'expanded'.
3990 2018-08-23  Alexander Monakov  <amonakov@ispras.ru>
3992         * tree-scalar-evolution.c (final_value_replacement_loop): Dump
3993         full GENERIC expression used for replacement.
3995 2018-08-23  Aldy Hernandez  <aldyh@redhat.com>
3997         * tree-vrp.c (abs_extent_range): Remove.
3998         (extract_range_into_wide_ints): Pass wide ints by reference.
3999         (extract_range_from_binary_expr_1): Rewrite the *DIV_EXPR code.
4000         Pass wide ints by reference in all calls to
4001         extract_range_into_wide_ints.
4002         * wide-int-range.cc (wide_int_range_div): New.
4003         * wide-int-range.h (wide_int_range_div): New.
4004         (wide_int_range_includes_zero_p): New.
4005         (wide_int_range_zero_p): New.
4007 2018-08-23  Matthew Malcomson  <matthew.malcomson@arm.com>
4009         * config/aarch64/aarch64.md (arches): New enum.
4010         (arch): New enum attr.
4011         (arch_enabled): New attr.
4012         (enabled): Now uses arch_enabled only.
4013         (simd, sve, fp16): Removed attribute.
4014         (fp): Attr now defined in terms of 'arch'.
4015         (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64, *movti_aarch64,
4016         *movhf_aarch64, <optab><fcvt_target><GPF:mode>2,
4017         <FCVT_F2FIXED:fcvt_fixed_insn><GPF:mode>3,
4018         <FCVT_FIXED2F:fcvt_fixed_insn><GPI:mode>3): Merge 'fp' and 'simd'
4019         attributes into 'arch'.
4020         (*movsf_aarch64, *movdf_aarch64, *movtf_aarch64, *add<mode>3_aarch64,
4021         subdi3, neg<mode>2, <optab><mode>3, one_cmpl<mode>2,
4022         *<NLOGICAL:optab>_one_cmpl<mode>3, *xor_one_cmpl<mode>3,
4023         *aarch64_ashl_sisd_or_int_<mode>3, *aarch64_lshr_sisd_or_int_<mode>3,
4024         *aarch64_ashr_sisd_or_int_<mode>3, *aarch64_sisd_ushl): Convert use of
4025         'simd' attribute into 'arch'.
4026         (load_pair_sw_<SX:mode><SX2:mode>, load_pair_dw_<DX:mode><DX2:mode>,
4027         store_pair_sw_<SX:mode><SX2:mode>, store_pair_dw_<DX:mode><DX2:mode>):
4028         Convert use of 'fp' attribute to 'arch'.
4029         * config/aarch64/aarch64-simd.md (move_lo_quad_internal_<mode>,
4030         move_lo_quad_internal_<mode>): (different modes) Merge 'fp' and 'simd'
4031         into 'arch'.
4032         (move_lo_quad_internal_be_<mode>, move_lo_quad_internal_be_<mode>):
4033         (different modes) Merge 'fp' and 'simd' into 'arch'.
4034         (*aarch64_combinez<mode>, *aarch64_combinez_be<mode>): Merge 'fp' and
4035         'simd' into 'arch'.
4037 2018-08-23  Segher Boessenkool  <segher@kernel.crashing.org>
4039         PR rtl-optimization/87026
4040         * expmed.c (canonicalize_comparison): If we can no longer create
4041         pseudoregisters, don't.
4043 2018-08-23  Richard Earnshaw  <rearnsha@arm.com>
4045         PR target/86951
4046         * config/arm/arm-protos.h (arm_emit_speculation_barrier): New
4047         prototype.
4048         * config/arm/arm.c (speculation_barrier_libfunc): New static
4049         variable.
4050         (arm_init_libfuncs): Initialize it.
4051         (arm_emit_speculation_barrier): New function.
4052         * config/arm/arm.md (speculation_barrier): Call
4053         arm_emit_speculation_barrier for architectures that do not have
4054         DSB or ISB.
4055         (speculation_barrier_insn): Only match on Armv7 or later.
4057 2018-08-23  Richard Biener  <rguenther@suse.de>
4059         PR middle-end/87024
4060         * tree-inline.c (copy_bb): Drop unused __builtin_va_arg_pack_len
4061         calls.
4063 2018-08-23  Richard Sandiford  <richard.sandiford@arm.com>
4065         * config/aarch64/aarch64.c (aarch64_evpc_sve_tbl): Fix handling
4066         of single-vector TBLs.
4067         (aarch64_vectorize_vec_perm_const): Set one_vector_p when only
4068         one input is given.
4070 2018-08-23  Richard Sandiford  <richard.sandiford@arm.com>
4072         PR target/85910
4073         * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const_1): Fix
4074         aarch64_evpc_tbl guard.
4076 2018-08-22  Bernd Edlinger  <bernd.edlinger@hotmail.de>
4078         * tree-ssa-dse.c (compute_trims): Avoid folding away undefined
4079         behaviour.
4081 2018-08-22  Martin Sebor  <msebor@redhat.com>
4083         PR middle-end/87052
4084         * tree-pretty-print.c (pretty_print_string): Add argument.
4085         (dump_generic_node): Call to pretty_print_string with string size.
4087 2018-08-22  Segher Boessenkool  <segher@kernel.crashing.org>
4089         PR rtl-optimization/86771
4090         * combine.c (try_combine): Do not allow splitting a resulting PARALLEL
4091         of two SETs into those two SETs, one to be placed at i2, if that SETs
4092         destination is modified between i2 and i3.
4094 2018-08-22  Richard Sandiford  <richard.sandiford@arm.com>
4096         PR tree-optimization/86725
4097         * tree-vect-loop.c (vect_inner_phi_in_double_reduction_p): New
4098         function.
4099         (vect_analyze_scalar_cycles_1): Check it.
4101 2018-08-22  Richard Sandiford  <richard.sandiford@arm.com>
4103         PR tree-optimization/86725
4104         * tree-vect-loop.c (vect_is_simple_reduction): When treating
4105         an outer loop phi as a double reduction, make sure that the
4106         single user of the phi result is an inner loop phi.
4108 2018-08-22  Richard Sandiford  <richard.sandiford@arm.com>
4110         * tree-vect-data-refs.c (vect_analyze_group_access_1): Convert
4111         grouped stores with gaps to a strided group.
4113 2018-08-22  Richard Sandiford  <richard.sandiford@arm.com>
4115         * tree-vect-stmts.c (get_group_load_store_type)
4116         (get_load_store_type): Only test STMT_VINFO_STRIDED_P for the
4117         first statement in a group.
4119 2018-08-22  Iain Sandoe  <iain@sandoe.co.uk>
4121         * config/darwin.h (LINK_COMMAND_SPEC_A): Sync LTO options with
4122         the sequence used in gcc/gcc.c.
4124 2018-08-22  Iain Sandoe  <iain@sandoe.co.uk>
4126         PR other/704
4127         * gcc-ar.c (main): Don’t try to invoke the plug-in if we’re not
4128         building it.
4130 2018-08-22  Iain Sandoe  <iain@sandoe.co.uk>
4132         * config/darwin10.h (LINK_GCC_C_SEQUENCE_SPEC): Adjust to use the
4133         Darwin10-specific unwinder-shim.
4134         * config/darwin12.h (LINK_GCC_C_SEQUENCE_SPEC): Remove.
4135         * config/rs6000/darwin.h (DARWIN_CRT1_SPEC, DARWIN_DYLIB1_SPEC):
4136         New to cater for Darwin10 Rosetta.
4138 2018-08-22  Iain Sandoe  <iain@sandoe.co.uk>
4140         * config/i386/i386.c (ix86_output_addr_diff_elt): Move the MACH-O
4141         specific test before the one for HAVE_AS_GOTOFF_IN_DATA.
4143 2018-08-22  Iain Sandoe  <iain@sandoe.co.uk>
4145         PR bootstrap/81033
4146         PR target/81733
4147         PR target/52795
4148         * gcc/dwarf2out.c (FUNC_SECOND_SECT_LABEL): New.
4149         (dwarf2out_switch_text_section): Generate a local label for the
4150         second function sub-section and apply it as the second FDE start
4151         label.
4152         * gcc/final.c (final_scan_insn_1): Emit second FDE label after the
4153         second sub-section start.
4155 2018-08-22  Richard Biener  <rguenther@suse.de>
4157         PR tree-optimization/86988
4158         * tree-vrp.c (vrp_prop::check_mem_ref): Bail out on VLAs.
4160 2018-08-22  Richard Biener  <rguenther@suse.de>
4162         PR tree-optimization/86945
4163         * tree-cfg.c (generate_range_test): Use unsigned arithmetic.
4165 2018-08-22  Alexandre Oliva <oliva@adacore.com>
4167         * config/rs6000/rs6000.c (SMALL_DATA_RELOC, SMALL_DATA_REG): Add
4168         a comment about how uses of r2 for .sdata2 come about.
4170 2018-08-22  Alexandre Oliva <aoliva@redhat.com>
4172         * tree-ssa-reassoc.c (is_reassociable_op): Fix cut&pasto.
4174 2018-08-21  Marek Polacek  <polacek@redhat.com>
4176         PR c++/86981, Implement -Wpessimizing-move.
4177         * doc/invoke.texi: Document -Wpessimizing-move.
4179 2018-08-21  Jan Hubicka  <jh@suse.cz>
4181         * tree.c (find_decls_types_r): Do not check for redundant typedefs.
4182         * tree.h (is_redundant_typedef): Remove.
4183         * dwarf2out.c (is_redundant_typedef): Turn into static function.
4185 2018-08-21  Jan Hubicka  <jh@suse.cz>
4187         * tree.c (free_lang_data_in_decl): Remove types from DECL_CONTEXT
4188         when possible.
4190 2018-08-21  Tamar Christina  <tamar.christina@arm.com>
4192         * expmed.c (extract_low_bits): Reject invalid subregs early.
4194 2018-08-21  Bernd Edlinger  <bernd.edlinger@hotmail.de>
4196         PR middle-end/86121
4197         * tree-ssa-strlen.c (adjust_last_stmt): Avoid folding away undefined
4198         behaviour.
4200 2018-08-21  Rasmus Villemoes  <rv@rasmusvillemoes.dk>
4202         * config/vxworks.h: Guard vxworks_asm_out_constructor and
4203         vxworks_asm_out_destructor by !HAVE_INITFINI_ARRAY_SUPPORT
4204         * config/vxworks.c: Likewise.
4206 2018-08-21  Rasmus Villemoes  <rv@rasmusvillemoes.dk>
4208         * config/vxworks.c: Set targetm.have_ctors_dtors
4209         if HAVE_INITFINI_ARRAY_SUPPORT.
4210         * config/vxworks.h: Set SUPPORTS_INIT_PRIORITY
4211         if HAVE_INITFINI_ARRAY_SUPPORT.
4213 2018-08-21  Rasmus Villemoes  <rv@rasmusvillemoes.dk>
4215         * config/vxworks.h: Add $(WIND_BASE)/target/h/wrn/coreip to
4216         default search path for VxWorks < 7.
4218 2018-08-21  Bernd Edlinger  <bernd.edlinger@hotmail.de>
4220         * gimple-ssa-sprintf.c (decl_constant_value): Remove.
4221         (get_format_string): Refer to c_getstr.
4223 2018-08-21  Tom de Vries  <tdevries@suse.de>
4225         * cgraph.h (debuginfo_early_init, debuginfo_init, debuginfo_fini)
4226         (debuginfo_start, debuginfo_stop, debuginfo_early_start)
4227         (debuginfo_early_stop): Declare.
4228         * cgraphunit.c (debuginfo_early_init, debuginfo_init, debuginfo_fini)
4229         (debuginfo_start, debuginfo_stop, debuginfo_early_start)
4230         (debuginfo_early_stop): New function.
4231         (symbol_table::finalize_compilation_unit): Call debuginfo_early_start
4232         and debuginfo_early_stop.
4233         * dwarf2out.c (dwarf2out_finish, dwarf2out_early_finish): Dump dwarf.
4234         * toplev.c (compile_file): Call debuginfo_start and debuginfo_stop.
4235         (general_init): Call debuginfo_early_init.
4236         (finalize): Call debuginfo_fini.
4237         (do_compile): Call debuginfo_init.
4238         * doc/invoke.texi (@gccoptlist): Add -fdump-debug and
4239         -fdump-early-debug.
4240         (@item -fdump-debug, @item -fdump-earlydebug): Add.
4242 2018-08-21  Tom de Vries  <tdevries@suse.de>
4244         * dwarf2out.c (print_dw_val, print_loc_descr, print_die): Handle
4245         flag_dump_noaddr and flag_dump_unnumbered.
4247 2018-08-21  Aldy Hernandez  <aldyh@redhat.com>
4249         * wide-int-range.cc (wide_int_range_abs): New.
4250         (wide_int_range_order_set): Rename from wide_int_range_min_max.
4251         * wide-int-range.h (wide_int_range_abs): New.
4252         (wide_int_range_min_max): New.
4253         * tree-vrp.c (extract_range_from_unary_expr): Rewrite ABS_EXPR
4254         case to call wide_int_range_abs.
4255         Rewrite MIN/MAX_EXPR to call wide_int_range_min_max.
4256         (extract_range_from_abs_expr): Delete.
4258 2018-08-20  Michael Meissner  <meissner@linux.ibm.com>
4260         PR target/87033
4261         * config/rs6000/rs6000.md (extendsi<mode>2): Change constraints
4262         from 'Y' to 'YZ' to enable the LWAX instruction to be generated
4263         for indexed loads.
4265 2018-08-20  Nathan Sidwell  <nathan@acm.org>
4266             Jeff Law <law@redhat.com>
4268         * config/s390/s390-c (s390_macro_to_expand): Use cpp_macro_p.
4269         * config/spu/spu-c.c (spu_macro_to_expand): Likewise.
4271 2018-08-20  David Malcolm  <dmalcolm@redhat.com>
4273         PR other/84889
4274         * attribs.c (diag_attr_exclusions): Add auto_diagnostic_group instance.
4275         (decl_attributes): Likewise.
4276         * calls.c (maybe_warn_nonstring_arg): Add auto_diagnostic_group
4277         instance.
4278         * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
4279         * diagnostic-core.h (class auto_diagnostic_group): New class.
4280         * diagnostic.c (diagnostic_initialize): Initialize the new fields.
4281         (diagnostic_report_diagnostic): Handle the first diagnostics within
4282         a group.
4283         (emit_diagnostic): Add auto_diagnostic_group instance.
4284         (inform): Likewise.
4285         (inform_n): Likewise.
4286         (warning): Likewise.
4287         (warning_at): Likewise.
4288         (warning_n): Likewise.
4289         (pedwarn): Likewise.
4290         (permerror): Likewise.
4291         (error): Likewise.
4292         (error_n): Likewise.
4293         (error_at): Likewise.
4294         (sorry): Likewise.
4295         (fatal_error): Likewise.
4296         (internal_error): Likewise.
4297         (internal_error_no_backtrace): Likewise.
4298         (auto_diagnostic_group::auto_diagnostic_group): New ctor.
4299         (auto_diagnostic_group::~auto_diagnostic_group): New dtor.
4300         * diagnostic.h (struct diagnostic_context): Add fields
4301         "diagnostic_group_nesting_depth",
4302         "diagnostic_group_emission_count", "begin_group_cb",
4303         "end_group_cb".
4304         * gimple-ssa-isolate-paths.c (find_implicit_erroneous_behavior):
4305         Add auto_diagnostic_group instance(s).
4306         (find_explicit_erroneous_behavior): Likewise.
4307         * gimple-ssa-warn-alloca.c (pass_walloca::execute): Likewise.
4308         * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Likewise.
4309         * gimplify.c (warn_implicit_fallthrough_r): Likewise.
4310         (gimplify_va_arg_expr): Likewise.
4311         * hsa-gen.c (HSA_SORRY_ATV): Likewise.
4312         (HSA_SORRY_AT): Likewise.
4313         * ipa-devirt.c (compare_virtual_tables): Likewise.
4314         (warn_odr): Likewise.
4315         * multiple_target.c (expand_target_clones): Likewise.
4316         * opts-common.c (cmdline_handle_error): Likewise.
4317         * reginfo.c (globalize_reg): Likewise.
4318         * substring-locations.c (format_warning_n_va): Likewise.
4319         * tree-inline.c (expand_call_inline): Likewise.
4320         * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Likewise.
4321         * tree-ssa-loop-niter.c
4322         (do_warn_aggressive_loop_optimizations): Likewise.
4323         * tree-ssa-uninit.c (warn_uninit): Likewise.
4324         * tree.c (warn_deprecated_use): Likewise.
4326 2018-08-20  H.J. Lu  <hongjiu.lu@intel.com>
4328         PR target/87014
4329         * config/i386/i386.md (eh_return): Always update EH return
4330         address in word_mode.
4332 2018-08-20  Chung-Lin Tang  <cltang@codesourcery.com>
4334         * targhooks.c (std_gimplify_va_arg_expr): Properly handle case of when
4335         TARGET_SPLIT_COMPLEX_ARG is defined.
4337 2018-08-20  Bernd Edlinger  <bernd.edlinger@hotmail.de>
4339         * expr.c (store_field): Change gcc_assert to gcc_checking_assert.
4341 2018-08-20  Bernd Edlinger  <bernd.edlinger@hotmail.de>
4343         PR target/86984
4344         * expr.c (expand_assignment): Assert that bitpos is positive.
4345         (store_field): Likewise
4346         (expand_expr_real_1): Make sure that bitpos is positive.
4347         * config/alpha/alpha.h (CONSTANT_ADDRESS_P): Avoid signed
4348         integer overflow.
4350 2018-08-20  Nathan Sidwell  <nathan@acm.org>
4352         * Makefile.in (CPP_ID_DATA_H): Delete.
4353         (CPP_INTERNAL_H): Don't add it.
4354         (GTFILES): Replace CPP_ID_DATA_H with CPPLIB_H.
4355         * gengtype.c (open_base_files): Replace cpp-id-data.h with cpplib.h
4357 2018-08-20  Richard Biener  <rguenther@suse.de>
4359         PR tree-optimization/78655
4360         * tree-vrp.c (extract_range_from_binary_expr_1): Make
4361         pointer + offset nonnull if either operand is nonnull work.
4363 2018-08-20  Tom de Vries  <tdevries@suse.de>
4365         * dwarf2out.c (add_scalar_info): Don't add reference to existing die
4366         unless the referenced die describes the added property using
4367         DW_AT_location or DW_AT_const_value.  Fall back to exprloc case.
4368         Otherwise, add a DW_AT_location to the referenced die.
4370 2018-08-19  Uros Bizjak  <ubizjak@gmail.com>
4372         PR target/86994
4373         * config/i386/i386.c (ix86_rtx_costs) [case SET]: Check source for
4374         register_operand when calling ix86_set_reg_reg_cost.
4375         [case CONST_INT, case CONST, case LABEL_REF, case SYMBOL_REF]:
4376         Set *total to 0 for operands that satisfy x86_64_immediate_operand
4377         predicate and to 1 otherwise.
4379 2018-08-18  Iain Sandoe  <iain@sandoe.co.uk>
4381         * config/darwin.c (darwin_override_options): If -gsplit-dwarf is set,
4382         emit a diagnostic that it is not supported and reset the option.
4383         * config/darwin.h (DRIVER_SELF_SPECS): Note that gsplit-dwarf is not
4384         supported and consume the option.  (ASM_FINAL_SPEC): New.
4386 2018-08-17  Segher Boessenkool  <segher@kernel.crashing.org>
4388         * doc/md.texi (Patterns): Use @ref instead of @xref in the middle of
4389         a sentence.
4391 2018-08-17  Sandra Loosemore  <sandra@codesourcery.com>
4393         C-SKY port: Documentation
4395         * doc/extend.texi (C-SKY Function Attributes): New section.
4396         * doc/invoke.texi (Option Summary): Add C-SKY options.
4397         (C-SKY Options): New section.
4398         * doc/md.texi (Machine Constraints): Document C-SKY constraints.
4400 2018-08-17  Jojo  <jijie_rong@c-sky.com>
4401             Huibin Wang  <huibin_wang@c-sky.com>
4402             Sandra Loosemore  <sandra@codesourcery.com>
4403             Chung-Lin Tang  <cltang@codesourcery.com>
4405         C-SKY port: Backend implementation
4407         * config/csky/*: New.
4408         * common/config/csky/*: New.
4410 2018-08-17  Jojo  <jijie_rong@c-sky.com>
4411             Huibin Wang  <huibin_wang@c-sky.com>
4412             Sandra Loosemore  <sandra@codesourcery.com>
4413             Chung-Lin Tang  <cltang@codesourcery.com>
4414             Andrew Jenner  <andrew@codesourcery.com>
4416         C-SKY port: Configury
4418         * config.gcc (csky-*-*): New.
4419         * configure.ac: Add csky to targets for dwarf2 debug_line support.
4420         * configure: Regenerated.
4422 2018-08-17  David Malcolm  <dmalcolm@redhat.com>
4424         * dump-context.h: Include "dumpfile.h".
4425         (dump_context::dump_printf_va): Convert final param from va_list
4426         to va_list *.  Convert from ATTRIBUTE_PRINTF to
4427         ATTRIBUTE_GCC_DUMP_PRINTF.
4428         (dump_context::dump_printf_loc_va): Likewise.
4429         * dumpfile.c: Include "stringpool.h".
4430         (make_item_for_dump_printf_va): Delete.
4431         (make_item_for_dump_printf): Delete.
4432         (class dump_pretty_printer): New class.
4433         (dump_pretty_printer::dump_pretty_printer): New ctor.
4434         (dump_pretty_printer::emit_items): New member function.
4435         (dump_pretty_printer::emit_any_pending_textual_chunks): New member
4436         function.
4437         (dump_pretty_printer::emit_item): New member function.
4438         (dump_pretty_printer::stash_item): New member function.
4439         (dump_pretty_printer::format_decoder_cb): New member function.
4440         (dump_pretty_printer::decode_format): New member function.
4441         (dump_context::dump_printf_va): Reimplement in terms of
4442         dump_pretty_printer.
4443         (dump_context::dump_printf_loc_va): Convert final param from va_list
4444         to va_list *.
4445         (dump_context::begin_scope): Reimplement call to
4446         make_item_for_dump_printf.
4447         (dump_printf): Update for change to dump_printf_va.
4448         (dump_printf_loc): Likewise.
4449         (selftest::test_capture_of_dump_calls): Convert "stmt" from
4450         greturn * to gimple *.  Add a test_decl.  Add tests of dump_printf
4451         with %T, %E, and %G.
4452         * dumpfile.h (ATTRIBUTE_GCC_DUMP_PRINTF): New macro.
4453         (dump_printf): Replace ATTRIBUTE_PRINTF_2 with
4454         ATTRIBUTE_GCC_DUMP_PRINTF (2, 3).
4455         (dump_printf_loc): Replace ATTRIBUTE_PRINTF_3 with
4456         ATTRIBUTE_GCC_DUMP_PRINTF (3, 0).
4457         * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Convert
4458         use of HOST_WIDE_INT_PRINT_DEC on unsigned HOST_WIDE_INT "count"
4459         within a dump_printf_loc call to "%wu".
4460         (vector_alignment_reachable_p): Merge two dump_printf[_loc] calls,
4461         converting a use of HOST_WIDE_INT_PRINT_DEC to "%wd".  Add a
4462         missing space after "=".
4463         * tree-vect-loop.c (vect_analyze_loop_2) Within a dump_printf
4464         call, convert use of HOST_WIDE_INT_PRINT_DEC to "%wd".
4465         * tree-vect-slp.c (vect_slp_bb): Within a dump_printf_loc call,
4466         convert use of HOST_WIDE_INT_PRINT_UNSIGNED to "%wu".
4467         * tree-vectorizer.c (try_vectorize_loop_1): Likewise.  Remove
4468         duplicate "vectorized" from message.
4470 2018-08-17  Szabolcs Nagy  <szabolcs.nagy@arm.com>
4472         * config/arm/arm-builtins.c (arm_init_simd_builtin_types): Clear
4473         polyNxK_t element's TYPE_STRING_FLAG.
4475 2018-08-17  Segher Boessenkool  <segher@kernel.crashing.org>
4477         * config/rs6000/rs6000.md (*cbranch, *creturn): Name these patterns
4478         (they were unnamed before).  Fix comments.
4480 2018-08-17  Nathan Sidwell  <nathan@acm.org>
4482         * cppbuiltin.c: Include "cpplib.h", not "cpp-id-data.h".
4484 2018-08-17  Richard Biener  <rguenther@suse.de>
4486         PR tree-optimization/86841
4487         * wide-int-range.cc (wide_int_range_lshift): Use to_uhwi.
4489 2018-08-17  Martin Liska  <mliska@suse.cz>
4491         * common.opt: Remove Warn, Init and Report for options with
4492         Ignore/Deprecated flag. Warning is done automatically for
4493         Deprecated flags.
4494         * config/i386/i386.opt: Likewise.
4495         * config/ia64/ia64.opt: Likewise.
4496         * config/rs6000/rs6000.opt: Likewise.
4497         * cppbuiltin.c (define_builtin_macros_for_compilation_flags):
4498         Remove usage of flag_check_pointer_bounds.
4499         * lto-wrapper.c (merge_and_complain): Do not handle
4500         OPT_fcheck_pointer_bounds.
4501         (append_compiler_options): Likewise.
4502         * opt-functions.awk: Do not handle Deprecated.
4503         * optc-gen.awk: Check that Var, Report and Init are not
4504         used for an option with Ignore/Deprecated flag.
4505         * opts-common.c (decode_cmdline_option): Do not report
4506         CL_ERR_DEPRECATED.
4507         (read_cmdline_option): Report warning for OPT_SPECIAL_deprecated
4508         options.
4509         * opts.h (struct cl_option): Remove cl_deprecated flag.
4510         (CL_ERR_DEPRECATED): Remove error enum value.
4512 2018-08-17  Richard Biener  <rguenther@suse.de>
4514         PR middle-end/86505
4515         * tree-inline.c (copy_bb): When inlining __builtin_va_arg_pack_len ()
4516         across a va-arg-pack using call adjust its return value accordingly.
4518 2018-08-16  Martin Sebor  <msebor@redhat.com>
4520         PR tree-optimization/86853
4521         * gimple-ssa-sprintf.c (struct format_result): Rename member.
4522         (struct fmtresult): Add member and initialize it in ctors.
4523         (format_character): Handle %C.  Extend range to NUL.  Set MAYFAIL.
4524         (format_string): Handle %S the same as %ls.  Set MAYFAIL.
4525         (format_directive): Set POSUNDER4K when MAYFAIL is set.
4526         (parse_directive): Handle %C same as %c.
4527         (sprintf_dom_walker::compute_format_length): Adjust.
4528         (is_call_safe): Adjust.
4530 2018-08-16  Bernd Edlinger  <bernd.edlinger@hotmail.de>
4532         * builtins.c (c_strlen): Add new parameter eltsize.  Use it
4533         for determining how to count the elements.
4534         * builtins.h (c_strlen): Adjust prototype.
4535         * expr.c (string_constant): Add new parameter mem_size.
4536         Set *mem_size appropriately.
4537         * expr.h (string_constant): Adjust protoype.
4538         * gimple-fold.c (get_range_strlen): Add new parameter eltsize.
4539         * gimple-fold.h (get_range_strlen): Adjust prototype.
4540         * gimple-ssa-sprintf.c (get_string_length): Add new parameter eltsize.
4541         (format_string): Call get_string_length with eltsize.
4543 2018-08-16  David Malcolm  <dmalcolm@redhat.com>
4545         * diagnostic.c (default_diagnostic_start_span_fn): Call pp_string
4546         to emit the span, rather than setting it as the prefix.
4548 2018-08-16  David Malcolm  <dmalcolm@redhat.com>
4550         * diagnostic-show-locus.c (layout::start_annotation_line): Add
4551         "margin_char" parameter, defaulting to space.  Use it in place
4552         of pp_space for the initial part of the margin.
4553         (layout::print_leading_fixits): Use '+' when filling the margin
4554         of line-insertion fix-it hints.
4556 2018-08-16  Segher Boessenkool  <segher@kernel.crashing.org>
4558         * config/rs6000/rs6000.md (two unnamed define_insn and define_split):
4559         Delete.
4561 2018-08-16  Segher Boessenkool  <segher@kernel.crashing.org>
4563         * config/rs6000/altivec.md: Don't set length attribute to the default
4564         value.
4565         * config/rs6000/darwin.md: Ditto.
4566         * config/rs6000/dfp.md: Ditto.
4567         * config/rs6000/htm.md: Ditto.
4568         * config/rs6000/rs6000.md: Ditto.
4569         * config/rs6000/sync.md: Ditto.
4570         * config/rs6000/vsx.md: Ditto.
4572 2018-08-16  Segher Boessenkool  <segher@kernel.crashing.org>
4574         * config/rs6000/altivec.md: Don't set length attribute to the default
4575         value, for branch instructions.
4576         * config/rs6000/darwin.md: Ditto.
4577         * config/rs6000/rs6000.md: Ditto.
4579 2018-08-16  Segher Boessenkool  <segher@kernel.crashing.org>
4581         * config/rs6000/rs6000.md (length): Always define as const_int 4.
4582         (unnamed conditional branch define_insn): Set length to 4 or 8
4583         depending on offset.
4584         (<bd>_<mode>): Similar, for alternative 0.
4585         (<bd>tf_<mode>): Ditto.
4587 2018-08-16  Tamar Christina  <tamar.christina@arm.com>
4589         * expr.c (copy_blkmode_to_reg): Perform larger copies when safe.
4591 2018-08-16  Matthew Malcomson  <matthew.malcomson@arm.com>
4593         * doc/rtl.texi: Replace old RTX class names with new names.
4596 2018-08-16  Vlad Lazar  <vlad.lazar@arm.com>
4598         * expmed.h (canonicalize_comparison): New declaration.
4599         * expmed.c (canonicalize_comparison, equivalent_cmp_code): New function.
4600         * expmed.c (emit_store_flag_1): Add call to canonicalize_comparison.
4601         * optabs.c (prepare_cmp_insn): Likewise.
4602         * rtl.h (unsigned_condition_p): New function which checks if a
4603         comparison operator is unsigned.
4605 2018-08-16  Nathan Sidwell  <nathan@acm.org>
4607         * config/rs6000/rs6000-c.c (rs6000_macro_to_expend): Use cpp_macro_p.
4608         * config/powerpcspc/powerpcspe-c.c (rs6000_macro_to_expend): Likewise.
4610 2018-08-16  Tamar Christina  <tamar.christina@arm.com>
4612         PR target/84711
4613         * config/arm/arm.c (arm_can_change_mode_class): Disallow subreg.
4614         * config/arm/neon.md (movv4hf, movv8hf): Refactored to..
4615         (mov<mov>): ..this and enable unconditionally.
4617 2018-08-16  Tamar Christina  <tamar.christina@arm.com>
4619         * config/arm/neon.md (*neon_mov<mode>): Remove reg-to-reg alternative.
4621 2018-08-16  Sam Tebbs  <sam.tebbs@arm.com>
4623         * config/aarch64/aarch64.opt (mlow-precision-recip-sqrt)
4624         (mlow-precision-sqrt, mlow-precision-div, mverbose-cost-dump): Replace
4625         "Common" with "Target".
4627 2018-08-15  Uros Bizjak  <ubizjak@gmail.com>
4629         * config/i386/i386.opt (mmitigate-rop): Mark as deprecated.
4630         * doc/invoke.texi (mmitigate-rop): Remove.
4631         * config/i386/i386.c: Do not include "regrename.h".
4632         (ix86_rop_should_change_byte_p, reg_encoded_number)
4633         (ix86_get_modrm_for_rop, set_rop_modrm_reg_bits, ix86_mitigate_rop):
4634         Remove.
4635         (ix86_reorg): Remove call to ix86_mitigate_rop.
4636         * config/i386/i386.md (attr "modrm_class"): Remove.
4637         (cmp<mode>_ccno_1, mov<mode>_xor, movstrict<mode>_xor)
4638         (x86_mov<mode>cc_0_m1. x86_mov<mode>cc_0_m1_se)
4639         (x86_mov<mode>cc_0_m1_neg): Remove modrm_class attribute override.
4641 2018-08-15  Will Schmidt  <will_schmidt@vnet.ibm.com>
4643         * config/rs6000/rs600.c (rs6000_gimple_fold_builtin): Add entries to
4644         allow folding of mergeh() and mergel() for the float and double types.
4645         (fold_mergehl_helper): Rework to handle building a permute tree
4646         for float vectors.
4648 2018-08-15  Uros Bizjak  <ubizjak@gmail.com>
4650         * config/i386/i386.c (expand_vec_perm_movs): Enable V4SFmode
4651         for TARGET_SSE.
4653 2018-08-15  David Malcolm  <dmalcolm@redhat.com>
4655         * common.opt (fdiagnostics-show-labels): New option.
4656         * diagnostic-show-locus.c (class layout_range): Add field
4657         "m_label".
4658         (class layout): Add field "m_show_labels_p".
4659         (layout_range::layout_range): Add param "label" and use it to
4660         initialize m_label.
4661         (make_range): Pass in NULL for new "label" param of layout_range's
4662         ctor.
4663         (layout::layout): Initialize m_show_labels_p.
4664         (layout::maybe_add_location_range): Pass in loc_range->m_label
4665         when constructing layout_range instances.
4666         (struct line_label): New struct.
4667         (layout::print_any_labels): New member function.
4668         (layout::print_line): Call it if label-printing is enabled.
4669         (selftest::test_one_liner_labels): New test.
4670         (selftest::test_diagnostic_show_locus_one_liner): Call it.
4671         * diagnostic.c (diagnostic_initialize): Initialize
4672         context->show_labels_p.
4673         * diagnostic.h (struct diagnostic_context): Add field
4674         "show_labels_p".
4675         * doc/invoke.texi (Diagnostic Message Formatting Options): Add
4676         -fno-diagnostics-show-labels.
4677         * dwarf2out.c (gen_producer_string): Add
4678         OPT_fdiagnostics_show_labels to the ignored options.
4679         * gcc-rich-location.c (gcc_rich_location::add_expr): Add "label"
4680         param.
4681         (gcc_rich_location::maybe_add_expr): Likewise.
4682         * gcc-rich-location.h (gcc_rich_location::gcc_rich_location): Add
4683         label" param, defaulting to NULL.
4684         (gcc_rich_location::add_expr): Add "label" param.
4685         (gcc_rich_location::maybe_add_expr): Likewise.
4686         (class text_range_label): New class.
4687         (class range_label_for_type_mismatch): New class.
4688         * gimple-ssa-sprintf.c (fmtwarn): Pass NULL for new label params
4689         of format_warning_va.
4690         (fmtwarn_n): Likewise for new params of format_warning_n_va.
4691         * lto-wrapper.c (merge_and_complain): Add
4692         OPT_fdiagnostics_show_labels to the "pick one setting" options.
4693         (append_compiler_options): Likewise to the dropped options.
4694         (append_diag_options): Likewise to the passed-on options.
4695         * opts.c (common_handle_option): Handle the new option.
4696         * selftest-diagnostic.c
4697         (test_diagnostic_context::test_diagnostic_context): Enable
4698         show_labels_p.
4699         * substring-locations.c: Include "gcc-rich-location.h".
4700         (format_warning_n_va): Add "fmt_label" and "param_label" params
4701         and use them as appropriate.
4702         (format_warning_va): Add "fmt_label" and "param_label" params,
4703         passing them on to format_warning_n_va.
4704         (format_warning_at_substring): Likewise.
4705         (format_warning_at_substring_n): Likewise.
4706         * substring-locations.h (format_warning_va): Add "fmt_label" and
4707         "param_label" params.
4708         (format_warning_n_va): Likewise.
4709         (format_warning_at_substring): Likewise.
4710         (format_warning_at_substring_n): Likewise.
4711         * toplev.c (general_init): Initialize global_dc->show_labels_p.
4713 2018-08-15  Qing Zhao  <qing.zhao@oracle.com>
4715         PR testsuite/86519
4716         * builtins.c (expand_builtin_memcmp): Do not expand the call
4717         when overflow is detected.
4719 2018-08-15  Martin Sebor  <msebor@redhat.com>
4721         PR tree-optimization/71625
4722         * config/aarch64/aarch64-builtins.c
4723         (aarch64_init_simd_builtin_types): Clear Poly8_t's TYPE_STRING_FLAG.
4725 2018-08-15  Ilya Leoshkevich  <iii@linux.ibm.com>
4727         * config/s390/s390.c (s390_reorg): Remove loop.
4729 2018-08-15  Iain Sandoe  <iain@sandoe.co.uk>
4731         * config/darwin.c
4732          (darwin_function_switched_text_sections): Delete.
4733         * gcc/config/darwin.h
4734          (TARGET_ASM_FUNCTION_SWITCHED_TEXT_SECTIONS): Likewise.
4736 2018-08-15  Iain Sandoe  <iain@sandoe.co.uk>
4738         PR target/81685
4739         * config/darwin.h: (DEBUG_STR_OFFSETS_SECTION, DEBUG_LOCLISTS_SECTION,
4740         DEBUG_RNGLISTS_SECTION) new macros.  (DEBUG_PUBNAMES_SECTION,
4741         DEBUG_PUBTYPES_SECTION) update to include GNU variant.
4743 2018-08-15  Martin Liska  <mliska@suse.cz>
4745         PR tree-optimization/86925
4746         * predict.c (expr_expected_value_1): When taking
4747         later predictor, assign also probability.
4748         Use fold_build2_initializer_loc in order to fold
4749         the expression in -frounding-math.
4751 2018-08-14  Allan Sandfeld Jensen <allan.jensen@qt.io>
4753         * config/i386/i386.c (expand_vec_perm_movs): New method matching movs
4754         patterns.
4755         (expand_vec_perm_1): Try the new method.
4757 2018-08-14  Ilya Leoshkevich  <iii@linux.ibm.com>
4759         PR target/86547
4760         * lra-lives.c (remove_some_program_points_and_update_live_ranges):
4761         Check whether lra_live_max_point is 0 before dividing.
4763 2018-08-14  Martin Sebor  <msebor@redhat.com>
4765         PR tree-optimization/86650
4766         * tree-vrp.c (vrp_prop::check_array_ref): Print an inform message.
4767         (vrp_prop::check_mem_ref): Same.
4769 2018-08-13  Liu Hao <lh_mouse@126.com>
4771         * pretty-print.c (eat_esc_sequence): Swap the foreground and
4772         background colors if the COMMON_LVB_REVERSE_VIDEO flag is set,
4773         and clear it thereafter, as it only works for DBCS.
4775 2018-08-13  Liu Hao <lh_mouse@126.com>
4777         * pretty-print.c (mingw_ansi_fputs): Do not call _close() on the
4778         handle returned by _get_osf_handle().
4780 2018-08-13  Will Schmidt  <will_schmidt@vnet.ibm.com>
4782         * gcc/config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support
4783         for folding vec_perm.
4785 2018-08-13  Will Schmidt  <will_schmidt@vnet.ibm.com>
4787         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin):
4788         Add support for gimple-folding of vec_pack() and vec_unpack()
4789         intrinsics.
4791 2018-08-13  Will Schmidt <will_schmidt@vnet.ibm.com>
4793         * config/rs6000/rs6000.c (rs6000_builtin_valid_without_lhs): Add
4794         vec_xst variants to the list.
4795         (rs6000_gimple_fold_builtin): Add support for folding unaligned
4796         vector loads and stores.
4798 2018-08-13  David Edelsohn  <dje.gcc@gmail.com>
4800         * config.gcc (rs6000-ibm-aix4.x): Delete.
4801         (rs6000-ibm-aix5.1): Delete.
4802         (rs6000-ibm-aix5.2): Delete.
4803         (rs6000-ibm-aix5.3): Delete.
4804         * config/rs6000/aix43.h: Delete.
4805         * config/rs6000/aix51.h: Delete.
4806         * config/rs6000/aix52.h: Delete.
4807         * config/rs6000/t-aix43: Delete.
4809 2018-08-13  Ilya Leoshkevich  <iii@linux.ibm.com>
4811         * config/s390/s390.c (s390_decompose_constant_pool_ref):
4812         New function.
4813         (s390_decompose_address): Factor out constant pool ref
4814         decomposition.
4816 2018-08-12  Chung-Ju Wu  <jasonwucj@gmail.com>
4818         * config/nds32/nds32-predicates.c
4819         (nds32_can_use_bclr_p): Change return type as bool.
4820         (nds32_can_use_bset_p): Ditto.
4821         (nds32_can_use_btgl_p): Ditto.
4822         (nds32_can_use_bitci_p): Ditto.
4823         * config/nds32/nds32-protos.h
4824         (nds32_can_use_bclr_p): Change declaration.
4825         (nds32_can_use_bset_p): Ditto.
4826         (nds32_can_use_btgl_p): Ditto.
4827         (nds32_can_use_bitci_p): Ditto.
4829 2018-08-12  Chung-Ju Wu  <jasonwucj@gmail.com>
4831         * config/nds32/nds32.c (nds32_expand_prologue, nds32_expand_epilogue):
4832         Support -msched-prolog-epilog option.
4833         * config/nds32/nds32.opt (msched-prolog-epilog): New option.
4835 2018-08-12  Chung-Ju Wu  <jasonwucj@gmail.com>
4837         * common/config/nds32/nds32-common.c
4838         (nds32_option_optimization_table): Enalbe -malways-align.
4840 2018-08-12  Chung-Ju Wu  <jasonwucj@gmail.com>
4842         * config.gcc (nds32*): Add nds32_isr.h and nds32_init.inc in
4843         extra_headers.
4844         * common/config/nds32/nds32-common.c (nds32_handle_option): Handle
4845         OPT_misr_secure_ case.
4846         * config/nds32/nds32-isr.c: Implementation of backward compatibility.
4847         * config/nds32/nds32-protos.h (nds32_isr_function_critical_p): New.
4848         * config/nds32/nds32.c (nds32_attribute_table): Add critical and
4849         secure attribute.
4850         * config/nds32/nds32.h (nds32_isr_nested_type): Add NDS32_CRITICAL.
4851         (nds32_isr_info): New field security_level.
4852         (TARGET_ISR_VECTOR_SIZE_4_BYTE): New macro.
4853         * config/nds32/nds32.md (return_internal): Consider critical attribute.
4854         * config/nds32/nds32.opt (misr-secure): New option.
4855         * config/nds32/nds32_init.inc: New file.
4856         * config/nds32/nds32_isr.h: New file.
4858 2018-08-11  John David Anglin  <danglin@gcc.gnu.org>
4860         * config/pa/pa.md (UNSPEC_MEMORY_BARRIER): New unspec enum.
4861         Update comment for atomic instructions.
4862         (atomic_storeqi, atomic_storehi, atomic_storesi, atomic_storesf,
4863         atomic_loaddf, atomic_loaddf_1, atomic_storedf, atomic_storedf_1):
4864         Remove.
4865         (atomic_loaddi): Revise fence expansion to only emit fence prior to
4866         load for __ATOMIC_SEQ_CST model.
4867         (atomic_loaddi_1): Remove float register target.
4868         (atomic_storedi): Handle CONST_INT values.
4869         (atomic_storedi_1): Remove float register source.  Add special case
4870         for zero value.
4871         (memory_barrier): New expander and insn.
4873 2018-08-11  Jakub Jelinek  <jakub@redhat.com>
4875         PR tree-optimization/86835
4876         * tree-ssa-math-opts.c (insert_reciprocals): Even when inserting
4877         new_stmt after def_gsi, make sure to insert new_square_stmt after
4878         that stmt, not 2 stmts before it.
4880 2018-08-10  Alexander Monakov  <amonakov@ispras.ru>
4882         PR target/82418
4883         * config/i386/i386.md (<s>mul<mode>3_highpart): Use DWIH mode iterator
4884         instead of SWI48.
4886 2018-08-10  Martin Liska  <mliska@suse.cz>
4888         PR target/83610
4889         * builtin-types.def (BT_FN_LONG_LONG_LONG_DOUBLE): Add new
4890         function type.
4891         * builtins.c (expand_builtin_expect_with_probability):
4892         New function.
4893         (expand_builtin_expect_with_probability): New function.
4894         (build_builtin_expect_predicate): Add new argumnet probability
4895         for BUILT_IN_EXPECT_WITH_PROBABILITY.
4896         (fold_builtin_expect):
4897         (fold_builtin_2):
4898         (fold_builtin_3):
4899         * builtins.def (BUILT_IN_EXPECT_WITH_PROBABILITY):
4900         * builtins.h (fold_builtin_expect): Set new argument.
4901         * doc/extend.texi: Document __builtin_expect_with_probability.
4902         * doc/invoke.texi: Likewise.
4903         * gimple-fold.c (gimple_fold_call): Pass new argument.
4904         * ipa-fnsummary.c (find_foldable_builtin_expect): Handle
4905         also BUILT_IN_EXPECT_WITH_PROBABILITY.
4906         * predict.c (get_predictor_value): New function.
4907         (expr_expected_value): Add new argument probability. Assume
4908         that predictor and probability are always non-null.
4909         (expr_expected_value_1): Likewise.  For __builtin_expect and
4910         __builtin_expect_with_probability set probability.  Handle
4911         combination in binary expressions.
4912         (tree_predict_by_opcode): Simplify code by simply calling
4913         get_predictor_value.
4914         (pass_strip_predict_hints::execute): Add handling of
4915         BUILT_IN_EXPECT_WITH_PROBABILITY.
4916         * predict.def (PRED_BUILTIN_EXPECT_WITH_PROBABILITY): Add
4917         new predictor.
4918         * tree.h (DECL_BUILT_IN_P): New function.
4920 2018-08-10  Martin Liska  <mliska@suse.cz>
4922         PR tree-optimization/85799
4923         * passes.def: Add argument for pass_strip_predict_hints.
4924         * predict.c (class pass_strip_predict_hints): Add new argument
4925         early_p.
4926         (strip_predictor_early): New function.
4927         (pass_strip_predict_hints::execute): Call the function to
4928         strip predictors.
4929         (strip_predict_hints): New function.
4930         * predict.def: Fix comment.
4932 2018-08-10  Thomas Preud'homme  <thomas.preudhomme@linaro.org>
4934         * Makefile.in: Clarify which tm.texi to copy over to assert the
4935         right to grant a GFDL license for all.
4937 2018-08-09  Jeff Law  <law@redhat.com>
4939         * config/m68k/m68k.c (m68k_adjust_decorated_operand): Remove
4940         unused variable.
4942 2018-08-09  Andreas Schwab  <schwab@linux-m68k.org>
4944         * config/m68k/m68k-protos.h (m68k_final_prescan_insn): Remove
4945         prototype.
4947 2018-08-09  Richard Sandiford  <richard.sandiford@arm.com>
4949         * tree-vect-loop.c (vectorizable_reduction): Allow inner-loop
4950         reductions for variable-length vectors.
4952 2018-08-09  David Malcolm  <dmalcolm@redhat.com>
4954         PR other/84889
4955         * common.opt (fdiagnostics-show-line-numbers): New option.
4956         * diagnostic-show-locus.c (class layout): Add fields
4957         "m_show_line_numbers_p" and "m_linenum_width";
4958         (num_digits): New function.
4959         (test_num_digits): New function.
4960         (layout::layout): Initialize new fields.  Update m_x_offset
4961         logic to handle any left margin.
4962         (layout::print_source_line): Print line number when requested.
4963         (layout::start_annotation_line): New member function.
4964         (layout::print_annotation_line): Call it.
4965         (layout::print_leading_fixits): Likewise.
4966         (layout::print_trailing_fixits): Likewise.  Update calls to
4967         move_to_column for new parameter.
4968         (layout::get_x_bound_for_row): Add "add_left_margin" param and use
4969         it to potentially call start_annotation_line.
4970         (layout::show_ruler): Call start_annotation_line.
4971         (selftest::test_line_numbers_multiline_range): New selftest.
4972         (selftest::diagnostic_show_locus_c_tests): Call test_num_digits
4973         and selftest::test_line_numbers_multiline_range.
4974         * diagnostic.c (diagnostic_initialize): Initialize
4975         show_line_numbers_p.
4976         * diagnostic.h (struct diagnostic_context): Add field
4977         "show_line_numbers_p".
4978         * doc/invoke.texi (Diagnostic Message Formatting Options): Add
4979         -fno-diagnostics-show-line-numbers.
4980         * dwarf2out.c (gen_producer_string): Add
4981         OPT_fdiagnostics_show_line_numbers to the ignored options.
4982         * lto-wrapper.c (merge_and_complain): Likewise to the "pick
4983         one setting" options.
4984         (append_compiler_options): Likewise to the dropped options.
4985         (append_diag_options): Likewise to the passed-on options.
4986         * opts.c (common_handle_option): Handle the new option.
4987         * toplev.c (general_init): Set up global_dc->show_line_numbers_p.
4989 2018-08-09  Kelvin Nilsen  <kelvin@gcc.gnu.org>
4991         * doc/extend.texi (PowerPC AltiVec Built-in Functions Available on
4992         ISA 2.07): Correct spelling of bcdsub to be __builtin_bcdsub.  Add
4993         third argument of type "const signed char" to descriptions of
4994         __builtin_bcdadd, __builtin_bcdadd_lt, __builtin_bcdadd_eq,
4995         __builtin_bcdadd_gt, __builtin_bcdadd_ov, __builtin_bcdsub,
4996         __builtin_bcdsub_lt, __builtin_bcdsub_eq, __builtin_bcdsub_gt,
4997         __builtin_bcdsub_ov functions.
4999 2018-08-09  Richard Sandiford  <richard.sandiford@arm.com>
5001         PR tree-optimization/86858
5002         * tree-vect-loop.c (vect_is_simple_reduction): Restore
5003         flow_bb_inside_loop_p calls.
5005 2018-08-09  Richard Sandiford  <richard.sandiford@arm.com>
5007         PR tree-optimization/86871
5008         * tree-vect-stmts.c (vect_transform_stmt): Use gimple_get_lhs
5009         instead of gimple_assign_lhs.
5011 2018-08-09  Richard Earnshaw  <rearnsha@arm.com>
5013         PR target/86887
5014         * config/aarch64/aarch64.md (add<mode>3_carryinC_zero): Add missing
5015         register constraint to operand 0.
5016         (add<mode>3_carryinC): Likewise.
5017         (add<mode>3_carryinV_zero, add<mode>3_carryinV): Likewise.
5019 2018-08-09  Martin Liska  <mliska@suse.cz>
5021         PR c/86895
5022         * common.opt: Remove extra line.
5024 2018-08-09  Martin Liska  <mliska@suse.cz>
5026         * params.def (PARAM_ALIGN_LOOP_ITERATIONS): Remove double dots
5027         at the end of a line, make first letter capital and end up
5028         a sentence with a dot.
5029         (PARAM_LOOP_INTERCHANGE_STRIDE_RATIO): Likewise.
5030         (PARAM_LOOP_BLOCK_TILE_SIZE): Likewise.
5031         (PARAM_GRAPHITE_MAX_NB_SCOP_PARAMS): Likewise.
5032         (PARAM_GRAPHITE_MAX_ARRAYS_PER_SCOP): Likewise.
5033         (PARAM_MAX_ISL_OPERATIONS): Likewise.
5034         (PARAM_GRAPHITE_ALLOW_CODEGEN_ERRORS): Likewise.
5035         (PARAM_PROFILE_FUNC_INTERNAL_ID): Likewise.
5036         (PARAM_INDIR_CALL_TOPN_PROFILE): Likewise.
5037         (PARAM_SLP_MAX_INSNS_IN_BB): Likewise.
5038         (PARAM_IPA_CP_EVAL_THRESHOLD): Likewise.
5039         (PARAM_IPA_CP_RECURSION_PENALTY): Likewise.
5040         (PARAM_IPA_CP_SINGLE_CALL_PENALTY): Likewise.
5041         (PARAM_IPA_CP_LOOP_HINT_BONUS): Likewise.
5042         (PARAM_IPA_CP_ARRAY_INDEX_HINT_BONUS): Likewise.
5043         (PARAM_TREE_REASSOC_WIDTH): Likewise.
5044         (PARAM_HSA_GEN_DEBUG_STORES): Likewise.
5045         (PARAM_MAX_SPECULATIVE_DEVIRT_MAYDEFS): Likewise.
5046         (PARAM_MAX_VRP_SWITCH_ASSERTIONS): Likewise.
5048 2018-08-09  Andreas Krebbel  <krebbel@linux.ibm.com>
5050         PR target/84332
5051         * config/s390/s390.c (s390_option_override_internal): Reduce the
5052         stack-clash-protection-probe-interval param if it would be too big
5053         for z900.
5055 2018-08-08  Andreas Schwab  <schwab@linux-m68k.org>
5057         PR target/46179
5058         * config/m68k/m68k.h (FINAL_PRESCAN_INSN): Don't define.
5059         * config/m68k/m68k.c (handle_move_double): Don't call
5060         m68k_final_prescan_insn.
5061         (m68k_adjust_decorated_operand): Renamed from
5062         m68k_final_prescan_insn, remove first and third operand and
5063         simplify.
5064         (print_operand): Call it.
5065         (print_operand_address): Call it.
5067 2018-08-08  Nathan Sidwell  <nathan@acm.org>
5069         * diagnostic.c (diagnostic_report_current_module): Use
5070         linemap_included_from & linemap_included_from_linemap.
5072 2018-08-08  Hongbo Zhang  <hongbo.zhang@linaro.org>
5074         * config/aarch64/aarch64-cores.def: Add phecda core.
5075         * config/aarch64/aarch64-tune.md: Regenerate.
5076         * doc/invoke.texi: Add phecda core.
5078 2018-08-08  Andreas Krebbel  <krebbel@linux.ibm.com>
5080         PR target/85295
5081         * config/s390/constraints.md ("NxHD0", "NxSD0"): New constraint
5082         definitions.
5083         * config/s390/s390.md ("movti"): Add more alternatives for
5084         constant to GPR copies.
5086 2018-08-08  Andreas Krebbel  <krebbel@linux.ibm.com>
5088         * config/s390/s390.c: Fix whitespace damage throughout the file.
5089         * config/s390/s390.h: Likewise.
5090         * config/s390/tpf.h: Likewise.
5092 2018-08-08  Ilya Leoshkevich  <iii@linux.ibm.com>
5094         * config/s390/s390.c (s390_loadrelative_operand_p):
5095         Remove TARGET_CPU_ZARCH usages.
5096         (s390_rtx_costs): Likewise.
5097         (s390_legitimate_constant_p): Likewise.
5098         (s390_cannot_force_const_mem): Likewise.
5099         (legitimate_reload_constant_p): Likewise.
5100         (s390_preferred_reload_class): Likewise.
5101         (legitimize_pic_address): Likewise.
5102         (legitimize_tls_address): Likewise.
5103         (s390_split_branches): Removed.
5104         (s390_add_execute): Removed.
5105         (s390_dump_pool): Remove TARGET_CPU_ZARCH usages.
5106         (s390_mainpool_start): Likewise.
5107         (s390_mainpool_finish): Likewise.
5108         (s390_mainpool_cancel): Removed.
5109         (s390_chunkify_start): Remove TARGET_CPU_ZARCH usages.
5110         (s390_chunkify_cancel): Likewise.
5111         (s390_return_addr_rtx): Likewise.
5112         (s390_register_info): Remove split_branches_pending_p uages.
5113         (s390_optimize_register_info): Likewise.
5114         (s390_init_frame_layout): Remove TARGET_CPU_ZARCH and
5115         split_branches_pending_p usages.
5116         (s390_can_eliminate): Remove TARGET_CPU_ZARCH usages.
5117         (s390_load_got): Likewise.
5118         (s390_expand_split_stack_prologue): Likewise.
5119         (output_asm_nops): Likewise.
5120         (s390_function_profiler): Likewise.
5121         (s390_emit_call): Likewise.
5122         (s390_conditional_register_usage): Likewise.
5123         (s390_optimize_prologue): Likewise.
5124         (s390_reorg): Remove TARGET_CPU_ZARCH and
5125         split_branches_pending_p usages.
5126         (s390_option_override_internal): Remove TARGET_CPU_ZARCH
5127         usages.
5128         (s390_output_indirect_thunk_function): Likewise.
5129         * config/s390/s390.h (TARGET_CPU_ZARCH): Removed.
5130         (TARGET_CPU_ZARCH_P): Removed.
5131         (struct machine_function): Remove split_branches_pending_p.
5132         * config/s390/s390.md: Remove TARGET_CPU_ZARCH usages.
5134 2018-08-08  Ilya Leoshkevich  <iii@linux.ibm.com>
5136         * common/config/s390/s390-common.c (processor_flags_table):
5137         Remove flags.
5138         * config.gcc: Remove with_arch/with_tune support.
5139         * config/s390/2064.md: Remove cpu attribute comparisons.
5140         * config/s390/driver-native.c (s390_host_detect_local_cpu):
5141         Remove MTN.
5142         * config/s390/linux.h (ASM_SPEC):
5143         Remove -march support.
5144         * config/s390/s390-c.c (s390_cpu_cpp_builtins_internal):
5145         Use a table to get an arch level.
5146         * config/s390/s390-opts.h (enum processor_type):
5147         Remove enum values.
5148         * config/s390/s390.c
5149         (processor_table): Remove entries, add arch_level values.
5150         (s390_issue_rate): Remove cases.
5151         (s390_option_override): Adjust
5152         s390_option_override_internal() call.
5153         (s390_option_override_internal): Remove deprecation warning.
5154         (s390_valid_target_attribute_tree): Adjust
5155         s390_option_override_internal() call.
5156         * config/s390/s390.h (struct s390_processor):
5157         Share with s390-c.c, add arch_level field.
5158         * config/s390/s390.md:
5159         Remove occurrences in cpu attribute.
5160         * config/s390/s390.opt: Remove -march/-mtune support.
5161         * config/s390/tpf.h (ASM_SPEC): Remove -march support.
5162         * doc/invoke.texi: Remove deprecation warning.
5164 2018-08-08  Luis Machado  <luis.machado@linaro.org>
5166         * config/aarch64/aarch64.c (qdf24xx_vector_cost): New static
5167         global.
5168         (qdf24xx_tunings): Set vector cost structure to
5169         qdf24xx_vector_cost.
5171         * config/aarch64/aarch64.c (qdf24xx_addrcost_table)
5172         <register_sextend>: Set to 3.
5174 2018-08-07  Richard Sandiford  <richard.sandiford@arm.com>
5176         PR target/86838
5177         * config/aarch64/iterators.md (FRECP, frecp_suffix): Delete.
5178         * config/aarch64/aarch64-simd.md
5179         (aarch64_frecp<FRECP:frecp_suffix><mode>): Fold FRECPE into...
5180         (@aarch64_frecpe<mode>): ...here and the move FRECPX to...
5181         (aarch64_frecpx<mode>): ...this new pattern.
5182         * config/aarch64/aarch64-simd-builtins.def: Remove comment
5183         about aarch64_frecp<FRECP:frecp_suffix><mode>.
5185 2018-08-07  Martin Liska  <mliska@suse.cz>
5187         PR middle-end/83023
5188         * predict.c (expr_expected_value_1): Handle DECL_IS_MALLOC,
5189         BUILT_IN_REALLOC and DECL_IS_OPERATOR_NEW.
5190         * predict.def (PRED_MALLOC_NONNULL): New predictor.
5191         * doc/extend.texi: Document that malloc attribute adds
5192         hit to compiler.
5194 2018-08-06  John David Anglin  <danglin@gcc.gnu.org>
5196         PR target/86785
5197         * config/pa/pa.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
5198         Define to speculation_safe_value_not_needed.
5200 2018-08-06  Jeff Law  <law@redhat.com>
5202         * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Pass down
5203         the vr_values instance to cprop_into_stmt.
5204         (cprop_into_stmt): Pass vr_values instance down to cprop_operand.
5205         (cprop_operand): Also query EVRP to determine if OP is a constant.
5207 2018-08-06  Nathan Sidwell  <nathan@acm.org>
5209         * diagnostic.c (diagnostic_report_current_module): Reroll
5210         included-at loop.  Translate text.
5212 2018-08-06  David Malcolm  <dmalcolm@redhat.com>
5214         * function-tests.c (selftest::test_expansion_to_rtl): Call
5215         free_after_compilation.
5217 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
5219         * config/aarch64/aarch64.md: Add clobber highs to tls_desc.
5221 2018-08-06  Andreas Krebbel  <krebbel@linux.ibm.com>
5223         * config/s390/s390.c (s390_loop_unroll_adjust): Prevent small
5224         loops with memory block operations from getting unrolled.
5226 2018-08-06  Ulrich Weigand  <uweigand@de.ibm.com>
5228         PR target/86807
5229         * config/spu/spu.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
5230         Define to speculation_safe_value_not_needed.
5232 2018-08-06  Jeff Law  <law@redhat.com>
5234         * reload1.c (forget_old_reloads_1): Adjust CLOBBER_HIGH
5235         assert.
5237 2018-08-06  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
5239         PR target/86662
5240         * gcc/tree.c (build_common_tree_nodes): Initialize integer_types array
5241         with all enabled __intN types.
5243         * gcc/testsuite/gcc.target/msp430/pr86662.c: New test.
5245 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
5247         * alias.c (record_set): Check for clobber high.
5248         * cfgexpand.c (expand_gimple_stmt): Likewise.
5249         * combine-stack-adj.c (single_set_for_csa): Likewise.
5250         * combine.c (find_single_use_1): Likewise.
5251         (set_nonzero_bits_and_sign_copies): Likewise.
5252         (get_combine_src_dest): Likewise.
5253         (is_parallel_of_n_reg_sets): Likewise.
5254         (try_combine): Likewise.
5255         (record_dead_and_set_regs_1): Likewise.
5256         (reg_dead_at_p_1): Likewise.
5257         (reg_dead_at_p): Likewise.
5258         * dce.c (deletable_insn_p): Likewise.
5259         (mark_nonreg_stores_1): Likewise.
5260         (mark_nonreg_stores_2): Likewise.
5261         * df-scan.c (df_find_hard_reg_defs): Likewise.
5262         (df_uses_record): Likewise.
5263         (df_get_call_refs): Likewise.
5264         * dwarf2out.c (mem_loc_descriptor): Likewise.
5265         * haifa-sched.c (haifa_classify_rtx): Likewise.
5266         * ira-build.c (create_insn_allocnos): Likewise.
5267         * ira-costs.c (scan_one_insn): Likewise.
5268         * ira.c (equiv_init_movable_p): Likewise.
5269         (rtx_moveable_p): Likewise.
5270         (interesting_dest_for_shprep): Likewise.
5271         * jump.c (mark_jump_label_1): Likewise.
5272         * postreload-gcse.c (record_opr_changes): Likewise.
5273         * postreload.c (reload_cse_simplify): Likewise.
5274         (struct reg_use): Add source expr.
5275         (reload_combine): Check for clobber high.
5276         (reload_combine_note_use): Likewise.
5277         (reload_cse_move2add): Likewise.
5278         (move2add_note_store): Likewise.
5279         * print-rtl.c (print_pattern): Likewise.
5280         * recog.c (decode_asm_operands): Likewise.
5281         (store_data_bypass_p): Likewise.
5282         (if_test_bypass_p): Likewise.
5283         * regcprop.c (kill_clobbered_value): Likewise.
5284         (kill_set_value): Likewise.
5285         * reginfo.c (reg_scan_mark_refs): Likewise.
5286         * reload1.c (maybe_fix_stack_asms): Likewise.
5287         (eliminate_regs_1): Likewise.
5288         (elimination_effects): Likewise.
5289         (mark_not_eliminable): Likewise.
5290         (scan_paradoxical_subregs): Likewise.
5291         (forget_old_reloads_1): Likewise.
5292         * reorg.c (find_end_label): Likewise.
5293         (try_merge_delay_insns): Likewise.
5294         (redundant_insn): Likewise.
5295         (own_thread_p): Likewise.
5296         (fill_simple_delay_slots): Likewise.
5297         (fill_slots_from_thread): Likewise.
5298         (dbr_schedule): Likewise.
5299         * resource.c (update_live_status): Likewise.
5300         (mark_referenced_resources): Likewise.
5301         (mark_set_resources): Likewise.
5302         * rtl.c (copy_rtx): Likewise.
5303         * rtlanal.c (reg_referenced_p): Likewise.
5304         (single_set_2): Likewise.
5305         (noop_move_p): Likewise.
5306         (note_stores): Likewise.
5307         * sched-deps.c (sched_analyze_reg): Likewise.
5308         (sched_analyze_insn): Likewise.
5310 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
5312         * cse.c (invalidate_reg): New function extracted from...
5313         (invalidate): ...here.
5314         (canonicalize_insn): Check for clobber high.
5315         (invalidate_from_clobbers): invalidate clobber highs.
5316         (invalidate_from_sets_and_clobbers): Likewise.
5317         (count_reg_usage): Check for clobber high.
5318         (insn_live_p): Likewise.
5319         * cselib.c (cselib_expand_value_rtx_1):Likewise.
5320         (cselib_invalidate_regno): Check for clobber in setter.
5321         (cselib_invalidate_rtx): Pass through setter.
5322         (cselib_invalidate_rtx_note_stores):
5323         (cselib_process_insn): Check for clobber high.
5324         * cselib.h (cselib_invalidate_rtx): Add operand.
5326 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
5328         * lra-eliminations.c (lra_eliminate_regs_1): Check for clobber high.
5329         (mark_not_eliminable): Likewise.
5330         * lra-int.h (struct lra_insn_reg): Add clobber high marker.
5331         * lra-lives.c (process_bb_lives): Check for clobber high.
5332         * lra.c (new_insn_reg): Remember clobber highs.
5333         (collect_non_operand_hard_regs): Check for clobber high.
5334         (lra_set_insn_recog_data): Likewise.
5335         (add_regs_to_insn_regno_info): Likewise.
5336         (lra_update_insn_regno_info): Likewise.
5338 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
5340         * rtl.h (reg_is_clobbered_by_clobber_high): Add declarations.
5341         * rtlanal.c (reg_is_clobbered_by_clobber_high): Add function.
5343 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
5345         * emit-rtl.c (verify_rtx_sharing): Check for CLOBBER_HIGH.
5346         (copy_insn_1): Likewise.
5347         (gen_hard_reg_clobber_high): New gen function.
5348         * genconfig.c (walk_insn_part): Check for CLOBBER_HIGH.
5349         * genemit.c (gen_exp): Likewise.
5350         (gen_emit_seq): Pass through info.
5351         (gen_insn): Check for CLOBBER_HIGH.
5352         (gen_expand): Pass through info.
5353         (gen_split): Likewise.
5354         (output_add_clobbers): Likewise.
5355         * genrecog.c (validate_pattern): Check for CLOBBER_HIGH.
5356         (remove_clobbers): Likewise.
5357         * rtl.h (gen_hard_reg_clobber_high): New declaration.
5359 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
5361         * doc/rtl.texi (clobber_high): Add.
5362         (parallel): Add in clobber high
5363         * rtl.c (rtl_check_failed_code3): Add function.
5364         * rtl.def (CLOBBER_HIGH): Add expression.
5365         * rtl.h (RTL_CHECKC3): Add macro.
5366         (rtl_check_failed_code3): Add declaration.
5367         (XC3EXP): Add macro.
5369 2018-08-05  H.J. Lu  <hongjiu.lu@intel.com>
5371         PR target/86386
5372         * config/i386/i386.c (ix86_finalize_stack_frame_flags): Set
5373         cfun->machine->max_used_stack_alignment if needed.
5375 2018-08-04  Martin Sebor  <msebor@redhat.com>
5377         PR tree-optimization/86571
5378         * gimple-ssa-sprintf.c (format_floating): Extend upper bound of
5379         NaN output to 4.
5381 2018-08-03  Sandra Loosemore  <sandra@codesourcery.com>
5383         PR target/86799
5384         * config/nios2/nios2.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
5385         Define.
5387 2018-08-03  Jeff Law  <law@redhat.com>
5389         PR target/86795
5390         * config/mn10300/mn10300.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
5391         Define to speculation_safe_value_not_needed.
5393 2018-08-03  David Malcolm  <dmalcolm@redhat.com>
5395         * doc/gcov.texi (-x): Remove duplicate "to".
5396         * doc/invoke.texi (-Wnoexcept-type): Remove duplicate "calls".
5397         (-Wif-not-aligned): Remove duplicate "is".
5398         (-flto): Remove duplicate "the".
5399         (MicroBlaze Options): In examples of "-mcpu=cpu-type", remove
5400         duplicate "v5.00.b".
5401         (MSP430 Options): Remove duplicate "and" from the description
5402         of "-mgprel-sec=regexp".
5403         (x86 Options): Remove duplicate copies of "vmldLog102" and
5404         vmlsLog104 from description of "-mveclibabi=type".
5406 2018-08-03  Richard Sandiford  <richard.sandiford@arm.com>
5408         * internal-fn.h (first_commutative_argument): Declare.
5409         * internal-fn.c (first_commutative_argument): New function.
5410         * tree-vect-slp.c (vect_get_and_check_slp_defs): Remove extra
5411         restrictions for pattern statements.  Use first_commutative_argument
5412         to look for commutative operands in calls to internal functions.
5414 2018-08-03  Aldy Hernandez  <aldyh@redhat.com>
5416         * Makefile.in (wide-int-range.o): New.
5417         * tree-vrp.c: Move all the wide_int_* functions to...
5418         * wide-int-range.cc: ...here.
5419         * tree-vrp.h: Move all the wide_int_* prototypes to...
5420         * wide-int-range.h: ...here.
5422 2018-08-03  Tom de Vries  <tdevries@suse.de>
5424         * common/config/nvptx/nvptx-common.c (nvptx_except_unwind_info): Return
5425         UI_NONE.
5426         * config/nvptx/nvptx.c (TARGET_ASM_BYTE_OP): Remove define.
5427         * except.c (output_function_exception_table): Do early exit if
5428         targetm_common.except_unwind_info (&global_options) == UI_NONE.
5430 2018-08-03  Martin Liska  <mliska@suse.cz>
5432         * predict.c (dump_prediction): Change to 2 digits
5433         in fraction part.
5435 2018-08-03  Siddhesh Poyarekar  <siddhesh@sourceware.org>
5437         * config/aarch64/falkor.md (falkor_am_1_vxvy_vxvy): Move
5438         neon_dup_q to...
5439         (falkor_am_1_gtov_gtov): ... a new insn reservation.
5441 2018-07-19  Ilya Leoshkevich  <iii@linux.ibm.com>
5443         * config/nds32/nds32.c (nds32_hard_regno_mode_ok): Replace > with >=.
5444         * df-problems.c (df_remove_dead_eq_notes): Replace > with >=.
5445         * dwarf2out.c (mem_loc_descriptor): Replace > with >=.
5446         * lra-constraints.c (spill_hard_reg_in_range): Replace <= with <.
5447         * lra-remat.c (call_used_input_regno_present_p): Replace <= with <.
5449 2018-08-02  David Malcolm  <dmalcolm@redhat.com>
5451         * diagnostic-show-locus.c (diagnostic_show_locus): Use
5452         pp_take_prefix when saving the existing prefix.
5453         * diagnostic.c (diagnostic_append_note): Likewise.
5454         * langhooks.c (lhd_print_error_function): Likewise.
5455         * pretty-print.c (pp_set_prefix): Drop the "const" from "prefix"
5456         param's type.  Free the existing prefix.
5457         (pp_take_prefix): New function.
5458         (pretty_printer::pretty_printer): Drop the prefix parameter.
5459         Rename the length parameter to match the comment.
5460         (pretty_printer::~pretty_printer): Free the prefix.
5461         * pretty-print.h (pretty_printer::pretty_printer): Drop the prefix
5462         parameter.
5463         (struct pretty_printer): Drop the "const" from "prefix" field's
5464         type and clarify memory management.
5465         (pp_set_prefix): Drop the "const" from the 2nd param.
5466         (pp_take_prefix): New decl.
5468 2018-08-02  Aaron Sawdey  <acsawdey@linux.ibm.com>
5470         * config/rs6000/rs6000-string.c (select_block_compare_mode): Move test
5471         for word_mode_ok here instead of passing as argument.
5472         (expand_block_compare): Change select_block_compare_mode() call.
5473         (expand_strncmp_gpr_sequence): New function.
5474         (expand_strn_compare): Make use of expand_strncmp_gpr_sequence.
5476 2018-08-02  Jeff Law  <law@redhat.com>
5478         PR target/86790
5479         * config/m68k/m68k.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
5480         Define to speculation_safe_value_not_needed.
5482         PR target/86784
5483         * config/h8300/h8300.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
5484         Define to speculation_safe_value_not_needed.
5486 2018-08-02  Tom de Vries  <tdevries@suse.de>
5488         PR target/86660
5489         * common/config/nvptx/nvptx-common.c (nvptx_except_unwind_info): New
5490         function.  Return UI_TARGET unconditionally.
5491         (TARGET_EXCEPT_UNWIND_INFO): Redefine to nvptx_except_unwind_info.
5492         * config/nvptx/nvptx.c (TARGET_ASM_BYTE_OP): Emit commented out '.byte'.
5494 2018-08-02  Richard Sandiford  <richard.sandiford@arm.com>
5496         * genemit.c (print_overload_test): Fix typo.
5498 2018-08-02  Richard Biener  <rguenther@suse.de>
5500         PR tree-optimization/86816
5501         * tree-ssa-tail-merge.c (tail_merge_valueize): New function
5502         which checks for value availability before querying it.
5503         (gvn_uses_equal): Use it.
5504         (same_succ_hash): Likewise.
5505         (gimple_equal_p): Likewise.
5507 2018-08-02  Nick Clifton  <nickc@redhat.com>
5509         PR target/86813
5510         * config/stormy16/stormy16.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
5511         Define to speculation_safe_value_not_needed.
5513         PR target/86810
5514         * config/v850/v850.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
5515         Define to speculation_safe_value_not_needed.
5517         PR target/86810
5518         * config/v850/v850.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
5519         Define to speculation_safe_value_not_needed.
5521         PR target/86803
5522         * config/rx/rx.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
5523         Define to speculation_safe_value_not_needed.
5525         PR target/86797
5526         * config/msp430/msp430.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
5527         Define to speculation_safe_value_not_needed.
5529         PR target/86791
5530         * config/mcore/mcore.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
5531         Define to speculation_safe_value_not_needed.
5533         PR target/86789
5534         * config/m32r/m32r.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
5535         Define to speculation_safe_value_not_needed.
5537         PR target/86787
5538         * config/iq2000/iq2000.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
5539         Define to speculation_safe_value_not_needed.
5541         PR target/86782
5542         * config/frv/frv.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define to
5543         speculation_safe_value_not_needed.
5545         PR target/86781
5546         * config/fr30/fr30.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define
5547         to speculation_safe_value_not_needed.
5549 2018-08-02  Richard Sandiford  <richard.sandiford@arm.com>
5551         * doc/md.texi: Expand the documentation of instruction names
5552         to mention port-local uses.  Document '@' in pattern names.
5553         * read-md.h (overloaded_instance, overloaded_name): New structs.
5554         (mapping): Declare.
5555         (md_reader::handle_overloaded_name): New member function.
5556         (md_reader::get_overloads): Likewise.
5557         (md_reader::m_first_overload): New member variable.
5558         (md_reader::m_next_overload_ptr): Likewise.
5559         (md_reader::m_overloads_htab): Likewise.
5560         * read-md.c (md_reader::md_reader): Initialize m_first_overload,
5561         m_next_overload_ptr and m_overloads_htab.
5562         * read-rtl.c (iterator_group): Add "type" and "get_c_token" fields.
5563         (get_mode_token, get_code_token, get_int_token): New functions.
5564         (map_attr_string): Add an optional argument that passes back
5565         the associated iterator.
5566         (overloaded_name_hash, overloaded_name_eq_p, named_rtx_p):
5567         (md_reader::handle_overloaded_name, add_overload_instance): New
5568         functions.
5569         (apply_iterators): Handle '@' names.  Report an error if '@'
5570         is used without iterators.
5571         (initialize_iterators): Initialize the new iterator_group fields.
5572         * genopinit.c (handle_overloaded_code_for)
5573         (handle_overloaded_gen): New functions.
5574         (main): Use them to print declarations of maybe_code_for_* and
5575         maybe_gen_* functions, and inline definitions of code_for_* and gen_*.
5576         * genemit.c (print_overload_arguments, print_overload_test)
5577         (handle_overloaded_code_for, handle_overloaded_gen): New functions.
5578         (main): Use it to print definitions of maybe_code_for_* and
5579         maybe_gen_* functions.
5580         * config/aarch64/aarch64.c (aarch64_split_128bit_move): Use
5581         gen_aarch64_mov{low,high}_di and gen_aarch64_movdi_{low,high}
5582         instead of explicit mode checks.
5583         (aarch64_split_simd_combine): Likewise gen_aarch64_simd_combine.
5584         (aarch64_split_simd_move): Likewise gen_aarch64_split_simd_mov.
5585         (aarch64_emit_load_exclusive): Likewise gen_aarch64_load_exclusive.
5586         (aarch64_emit_store_exclusive): Likewise gen_aarch64_store_exclusive.
5587         (aarch64_expand_compare_and_swap): Likewise
5588         gen_aarch64_compare_and_swap and gen_aarch64_compare_and_swap_lse
5589         (aarch64_gen_atomic_cas): Likewise gen_aarch64_atomic_cas.
5590         (aarch64_emit_atomic_swap): Likewise gen_aarch64_atomic_swp.
5591         (aarch64_constant_pool_reload_icode): Delete.
5592         (aarch64_secondary_reload): Use code_for_aarch64_reload_movcp
5593         instead of aarch64_constant_pool_reload_icode.  Use
5594         code_for_aarch64_reload_mov instead of explicit mode checks.
5595         (rsqrte_type, get_rsqrte_type, rsqrts_type, get_rsqrts_type): Delete.
5596         (aarch64_emit_approx_sqrt): Use gen_aarch64_rsqrte instead of
5597         get_rsqrte_type and gen_aarch64_rsqrts instead of gen_rqrts_type.
5598         (recpe_type, get_recpe_type, recps_type, get_recps_type): Delete.
5599         (aarch64_emit_approx_div): Use gen_aarch64_frecpe instead of
5600         get_recpe_type and gen_aarch64_frecps instead of get_recps_type.
5601         (aarch64_atomic_load_op_code): Delete.
5602         (aarch64_emit_atomic_load_op): Likewise.
5603         (aarch64_gen_atomic_ldop): Use UNSPECV_ATOMIC_* instead of
5604         aarch64_atomic_load_op_code.  Use gen_aarch64_atomic_load
5605         instead of aarch64_emit_atomic_load_op.
5606         * config/aarch64/aarch64.md (aarch64_reload_movcp<GPF_TF:mode><P:mode>)
5607         (aarch64_reload_movcp<VALL:mode><P:mode>, aarch64_reload_mov<mode>)
5608         (aarch64_movdi_<mode>low, aarch64_movdi_<mode>high)
5609         (aarch64_mov<mode>high_di, aarch64_mov<mode>low_di): Add a '@'
5610         character before the pattern name.
5611         * config/aarch64/aarch64-simd.md (aarch64_split_simd_mov<mode>)
5612         (aarch64_rsqrte<mode>, aarch64_rsqrts<mode>)
5613         (aarch64_simd_combine<mode>, aarch64_frecpe<mode>)
5614         (aarch64_frecps<mode>): Likewise.
5615         * config/aarch64/atomics.md (atomic_compare_and_swap<mode>)
5616         (aarch64_compare_and_swap<mode>, aarch64_compare_and_swap<mode>_lse)
5617         (aarch64_load_exclusive<mode>, aarch64_store_exclusive<mode>)
5618         (aarch64_atomic_swp<mode>, aarch64_atomic_cas<mode>)
5619         (aarch64_atomic_load<atomic_ldop><mode>): Likewise.
5621 2018-08-02  Richard Sandiford  <richard.sandiford@arm.com>
5623         * config/aarch64/aarch64.c (aarch64_float_const_representable_p):
5624         Allow HFmode constants if TARGET_FP_F16INST.
5626 2018-08-02  Jackson Woodruff  <jackson.woodruff@arm.com>
5628         PR target/86014
5629         * config/aarch64/aarch64.c (aarch64_operands_adjust_ok_for_ldpstp):
5630         No longer check last store for clobber of address register.
5632 2018-08-02  Martin Liska  <mliska@suse.cz>
5634         PR gcov-profile/86817
5635         * gcov.c (process_all_functions): New function.
5636         (main): Call it.
5637         (process_file): Move functions processing to
5638         process_all_functions.
5640 2018-08-02  David Malcolm  <dmalcolm@redhat.com>
5642         * dumpfile.c (dump_user_location_t::dump_user_location_t): Add
5643         "const" to the "gimple *" and "rtx_insn *" parameters.
5644         * dumpfile.h (dump_user_location_t::dump_user_location_t):
5645         Likewise.
5646         (dump_location_t::dump_location_t): Likewise.
5648 2018-08-01  Martin Sebor  <msebor@redhat.com>
5650         PR tree-optimization/86650
5651         * gimple-pretty-print.c (percent_G_format): Accept a "gimple *"
5652         rather than a "gcall *".  Directly pass the data of interest
5653         to percent_K_format, rather than building a temporary CALL_EXPR
5654         to hold it.
5655         * gimple-fold.c (gimple_fold_builtin_strncpy): Adjust.
5656         (gimple_fold_builtin_strncat): Adjust.
5657         * gimple-ssa-warn-restrict.h (check_bounds_or_overlap): Replace
5658         gcall* argument with gimple*.
5659         * gimple-ssa-warn-restrict.c (check_call): Same.
5660         (wrestrict_dom_walker::before_dom_children): Same.
5661         (builtin_access::builtin_access): Same.
5662         (check_bounds_or_overlap): Same
5663         (maybe_diag_overlap): Same.
5664         (maybe_diag_offset_bounds): Same.
5665         * tree-diagnostic.c (default_tree_printer): Move usage of
5666         EXPR_LOCATION (t) and TREE_BLOCK (t) from within percent_K_format
5667         to this callsite.
5668         * tree-pretty-print.c (percent_K_format): Add argument.
5669         * tree-pretty-print.h: Add argument.
5670         * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Adjust.
5671         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Adjust.
5672         (maybe_diag_stxncpy_trunc): Same.
5673         (handle_builtin_stxncpy): Same.
5674         (handle_builtin_strcat): Same.
5676 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
5678         * match.pd: Optimise pointer range checks.
5680 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
5682         PR tree-optimization/86758
5683         * tree-vect-stmts.c (vectorizable_simd_clone_call): Don't try
5684         to remove pattern statements.
5686 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
5688         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Use the
5689         result of dfs_enumerate_from when constructing stmt_vec_infos,
5690         instead of additionally calling get_loop_body.
5692 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
5694         * tree-vectorizer.h (vect_create_data_ref_ptr): Remove inv_p
5695         parameter.
5696         * tree-vect-data-refs.c (vect_create_data_ref_ptr): Likewise.
5697         When creating an iv, assert that the step is not known to be zero.
5698         (vect_setup_realignment): Update call accordingly.
5699         * tree-vect-stmts.c (vectorizable_store): Likewise.
5700         (vectorizable_load): Likewise.  Handle VMAT_INVARIANT separately.
5702 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
5704         * tree-vectorizer.h (vect_stmt_to_vectorize): New function.
5705         * tree-vect-loop.c (vect_update_vf_for_slp): Use it.
5706         (vectorizable_reduction): Likewise.
5707         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
5708         (vect_detect_hybrid_slp_stmts): Likewise.
5709         * tree-vect-stmts.c (vect_is_simple_use): Likewise.
5711 2018-08-01  Aldy Hernandez  <aldyh@redhat.com>
5713         * tree-vrp (zero_nonzero_bits_from_bounds): Rename to...
5714         (wide_int_set_zero_nonzero_bits): ...this.
5715         (zero_nonzero_bits_from_vr): Rename to...
5716         (vrp_set_zero_nonzero_bits): ...this.
5717         (extract_range_from_multiplicative_op_1): Abstract wide int
5718         code...
5719         (wide_int_range_multiplicative_op): ...here.
5720         (extract_range_from_binary_expr_1): Extract wide int binary
5721         operations into their own functions.
5722         (wide_int_range_lshift): New.
5723         (wide_int_range_can_optimize_bit_op): New.
5724         (wide_int_range_shift_undefined_p): New.
5725         (wide_int_range_bit_xor): New.
5726         (wide_int_range_bit_ior): New.
5727         (wide_int_range_bit_and): New.
5728         (wide_int_range_trunc_mod): New.
5729         (extract_range_into_wide_ints): New.
5730         (vrp_shift_undefined_p): New.
5731         (extract_range_from_multiplicative_op): New.
5732         (vrp_can_optimize_bit_op): New.
5733         * tree-vrp.h (value_range::dump): New.
5734         (wide_int_range_multiplicative_op): New.
5735         (wide_int_range_lshift):New.
5736         (wide_int_range_shift_undefined_p): New.
5737         (wide_int_range_bit_xor): New.
5738         (wide_int_range_bit_ior): New.
5739         (wide_int_range_bit_and): New.
5740         (wide_int_range_trunc_mod): New.
5741         (zero_nonzero_bits_from_bounds): Rename to...
5742         (wide_int_set_zero_nonzero_bits): ...this.
5743         (zero_nonzero_bits_from_vr): Rename to...
5744         (vrp_set_zero_nonzero_bits): ...this.
5745         (range_easy_mask_min_max): Rename to...
5746         (wide_int_range_can_optimize_bit_op): this.
5747         * vr-values.c (simplify_bit_ops_using_ranges): Rename
5748         zero_nonzero_bits_from_vr into vrp_set_zero_nonzero_bits.
5750 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
5752         * tree-vectorizer.h (vect_orig_stmt): New function.
5753         * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Use it.
5754         * tree-vect-loop.c (vect_model_reduction_cost): Likewise.
5755         (vect_create_epilog_for_reduction): Likewise.
5756         (vectorizable_live_operation): Likewise.
5757         * tree-vect-slp.c (vect_find_last_scalar_stmt_in_slp): Likewise.
5758         (vect_detect_hybrid_slp_stmts, vect_schedule_slp): Likewise.
5759         * tree-vect-stmts.c (vectorizable_call): Likewise.
5760         (vectorizable_simd_clone_call, vect_remove_stores): Likewise.
5762 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
5764         * tree-vectorizer.h (vect_transform_stmt): Remove grouped_store
5765         argument.
5766         * tree-vect-stmts.c (vect_transform_stmt): Likewise.
5767         * tree-vect-loop.c (vect_transform_loop_stmt): Update call accordingly.
5768         (vect_transform_loop): Likewise.
5769         * tree-vect-slp.c (vect_schedule_slp_instance): Likewise.
5771 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
5773         * tree-vectorizer.h (vect_schedule_slp): Return void.
5774         * tree-vect-slp.c (vect_schedule_slp_instance): Likewise.
5775         (vect_schedule_slp): Likewise.
5777 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
5779         * tree-vect-loop.c (vect_transform_loop_stmt): Remove slp_scheduled
5780         argument.
5781         (vect_transform_loop): Update calls accordingly.  Schedule SLP
5782         instances before the main loop, if any exist.
5784 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
5786         PR tree-optimization/86749
5787         * tree-vect-patterns.c (vect_determine_min_output_precision_1):
5788         If the lhs is used in a COND_EXPR, check that it is being used
5789         as the "then" or "else" value.
5791 2018-08-01  Tom de Vries  <tdevries@suse.de>
5793         PR target/86800
5794         * config/nvptx/nvptx.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define to
5795         speculation_safe_value_not_needed.
5797 2018-08-01  Richard Biener  <rguenther@suse.de>
5799         * tree-ssa-sccvn.c (visit_phi): Compare invariant addresses
5800         as base and offset.
5802 2018-08-01  Martin Liska  <mliska@suse.cz>
5804         * value-prof.c (gimple_divmod_fixed_value_transform): Unify
5805         format how successful transformation is dumped.
5806         (gimple_mod_pow2_value_transform): Likewise.
5807         (gimple_mod_subtract_transform): Likewise.
5808         (gimple_stringops_transform): Likewise.
5810 2018-08-01  Martin Liska  <mliska@suse.cz>
5812         PR value-prof/35543
5813         * value-prof.c (interesting_stringop_to_profile_p):
5814         Simplify the code and add BUILT_IN_MEMMOVE.
5815         (gimple_stringops_transform): Likewise.
5817 2018-08-01  Sam Tebbs  <sam.tebbs@arm.com>
5819         * config/aarch64/aarch64-simd.md
5820         (*aarch64_get_lane_zero_extendsi<mode>): Rename to...
5821         (*aarch64_get_lane_zero_extend<GPI:mode><VDQQH:mode>): ... This and
5822         use GPI iterator instead of SI mode.
5824 2018-08-01  Richard Earnshaw  <rearnsha@arm.com>
5826         * config/rs6000/rs6000.md (speculation_barrier): Renamed from
5827         rs6000_speculation_barrier.
5828         * config/rs6000/rs6000.c (rs6000_expand_builtin): Adjust for
5829         new barrier pattern name.
5831 2018-08-01  Richard Earnshaw  <rearnsha@arm.com>
5833         * config/i386/i386.md (unspecv): Add UNSPECV_SPECULATION_BARRIER.
5834         (speculation_barrier): New insn.
5836 2018-08-01  Richard Biener  <rguenther@suse.de>
5838         PR bootstrap/86724
5839         * graphite.h: Include isl/id.h and isl/space.h to allow build
5840         with ISL 0.20.
5842 2018-08-01  Jan Willem Jagersma  <jwjagersma@gmail.com>
5844         PR target/86651
5845         * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
5846         mode for COFF targets.
5847         * defaults.h (TARGET_COFF): Define.
5848         * config/i386/djgpp.h (TARGET_ASM_LTO_START, TARGET_ASM_LTO_END,
5849         TARGET_COFF): Define.
5850         (i386_djgpp_asm_lto_start, i386_djgpp_asm_lto_end): Declare.
5851         * config/i386/djgpp.c (saved_debug_info_level): New static variable.
5852         (i386_djgpp_asm_lto_start, i386_djgpp_asm_lto_end): New functions.
5854 2018-07-31  Alexandre Oliva <oliva@adacore.com>
5856         * gimple-streamer-in.c (input_bb): Restore BB discriminator.
5857         * gimple-streamer-out.c (output_bb): Save it.
5858         * lto-streamer-in.c (input_struct_function_base): Restore
5859         instance discriminator if available.  Create map on demand.
5860         * lto-streamer-out.c (output_struct_function_base): Save it if
5861         available.
5862         * final.c (decl_to_instance_map): Document LTO strategy.
5864 2018-07-31  Alexandre Oliva  <oliva@adacore.com>
5865             Olivier Hainque  <hainque@adacore.com>
5867         * debug.h (decl_to_instance_map_t): New type.
5868         (decl_to_instance_map): Declare.
5869         (maybe_create_decl_to_instance_map): New inline function.
5870         * final.c (bb_discriminator, last_bb_discriminator): New statics,
5871         to track basic block discriminators.
5872         (final_start_function_1): Initialize them.
5873         (final_scan_insn_1): On NOTE_INSN_BASIC_BLOCK, track
5874         bb_discriminator.
5875         (decl_to_instance_map): New variable.
5876         (map_decl_to_instance, maybe_set_discriminator): New functions.
5877         (notice_source_line): Set discriminator.
5879 2018-07-31  Ian Lance Taylor  <iant@golang.org>
5881         * targhooks.c (default_have_speculation_safe_value): Add
5882         ATTRIBUTE_UNUSED.
5884 2018-07-31  David Malcolm  <dmalcolm@redhat.com>
5886         * dump-context.h: Include "pretty-print.h".
5887         (dump_context::refresh_dumps_are_enabled): New decl.
5888         (dump_context::emit_item): New decl.
5889         (class dump_context): Add fields "m_test_pp" and
5890         "m_test_pp_flags".
5891         (temp_dump_context::temp_dump_context): Add param "test_pp_flags".
5892         (temp_dump_context::get_dumped_text): New decl.
5893         (class temp_dump_context): Add field "m_pp".
5894         * dumpfile.c (refresh_dumps_are_enabled): Convert to...
5895         (dump_context::refresh_dumps_are_enabled): ...and add a test for
5896         m_test_pp.
5897         (set_dump_file): Update for above change.
5898         (set_alt_dump_file): Likewise.
5899         (dump_loc): New overload, taking a pretty_printer *.
5900         (dump_context::dump_loc): Call end_any_optinfo.  Dump the location
5901         to any test pretty-printer.
5902         (make_item_for_dump_gimple_stmt): New function, adapted from
5903         optinfo::add_gimple_stmt.
5904         (dump_context::dump_gimple_stmt): Call it, and use the result,
5905         eliminating the direct usage of dump_file and alt_dump_file in
5906         favor of indirectly using them via emit_item.
5907         (make_item_for_dump_gimple_expr): New function, adapted from
5908         optinfo::add_gimple_expr.
5909         (dump_context::dump_gimple_expr): Call it, and use the result,
5910         eliminating the direct usage of dump_file and alt_dump_file in
5911         favor of indirectly using them via emit_item.
5912         (make_item_for_dump_generic_expr): New function, adapted from
5913         optinfo::add_tree.
5914         (dump_context::dump_generic_expr): Call it, and use the result,
5915         eliminating the direct usage of dump_file and alt_dump_file in
5916         favor of indirectly using them via emit_item.
5917         (make_item_for_dump_printf_va): New function, adapted from
5918         optinfo::add_printf_va.
5919         (make_item_for_dump_printf): New function.
5920         (dump_context::dump_printf_va): Call make_item_for_dump_printf_va,
5921         and use the result, eliminating the direct usage of dump_file and
5922         alt_dump_file in favor of indirectly using them via emit_item.
5923         (make_item_for_dump_dec): New function.
5924         (dump_context::dump_dec): Call it, and use the result,
5925         eliminating the direct usage of dump_file and alt_dump_file in
5926         favor of indirectly using them via emit_item.
5927         (make_item_for_dump_symtab_node): New function, adapted from
5928         optinfo::add_symtab_node.
5929         (dump_context::dump_symtab_node): Call it, and use the result,
5930         eliminating the direct usage of dump_file and alt_dump_file in
5931         favor of indirectly using them via emit_item.
5932         (dump_context::begin_scope): Reimplement, avoiding direct usage
5933         of dump_file and alt_dump_file in favor of indirectly using them
5934         via emit_item.
5935         (dump_context::emit_item): New member function.
5936         (temp_dump_context::temp_dump_context): Add param "test_pp_flags".
5937         Set up test pretty-printer on the underlying context.  Call
5938         refresh_dumps_are_enabled.
5939         (temp_dump_context::~temp_dump_context): Call
5940         refresh_dumps_are_enabled.
5941         (temp_dump_context::get_dumped_text): New member function.
5942         (selftest::verify_dumped_text): New function.
5943         (ASSERT_DUMPED_TEXT_EQ): New macro.
5944         (selftest::test_capture_of_dump_calls): Run all tests twice, with
5945         and then without optinfo enabled.  Add uses of
5946         ASSERT_DUMPED_TEXT_EQ to all tests.  Add test of nested scopes.
5947         * dumpfile.h: Update comment for the dump_* API.
5948         * optinfo-emit-json.cc
5949         (selftest::test_building_json_from_dump_calls): Update for new
5950         param for temp_dump_context ctor.
5951         * optinfo.cc (optinfo_item::optinfo_item): Remove "owned" param
5952         and "m_owned" field.
5953         (optinfo_item::~optinfo_item): Likewise.
5954         (optinfo::add_item): New member function.
5955         (optinfo::emit): Update comment.
5956         (optinfo::add_string): Delete.
5957         (optinfo::add_printf): Delete.
5958         (optinfo::add_printf_va): Delete.
5959         (optinfo::add_gimple_stmt): Delete.
5960         (optinfo::add_gimple_expr): Delete.
5961         (optinfo::add_tree): Delete.
5962         (optinfo::add_symtab_node): Delete.
5963         (optinfo::add_dec): Delete.
5964         * optinfo.h (class dump_context): New forward decl.
5965         (optinfo::add_item): New decl.
5966         (optinfo::add_string): Delete.
5967         (optinfo::add_printf): Delete.
5968         (optinfo::add_printf_va): Delete.
5969         (optinfo::add_gimple_stmt): Delete.
5970         (optinfo::add_gimple_expr): Delete.
5971         (optinfo::add_tree): Delete.
5972         (optinfo::add_symtab_node): Delete.
5973         (optinfo::add_dec): Delete.
5974         (optinfo::add_poly_int): Delete.
5975         (optinfo_item::optinfo_item): Remove "owned" param.
5976         (class optinfo_item): Remove field "m_owned".
5978 2018-07-31  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
5980         PR middle-end/86705
5981         * gcc/cfgexpand.c (set_parm_rtl): Use the alignment of Pmode when
5982         MAX_SUPPORTED_STACK_ALIGNMENT would otherwise be exceeded by the
5983         requested variable alignment.
5984         (expand_one_ssa_partition): Likewise.
5985         (expand_one_var): Likewise.
5987 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
5989         * config/pdp11/pdp11.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
5990         to speculation_safe_value_not_needed.
5992 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
5994         * targhooks.h (speculation_safe_value_not_needed): New prototype.
5995         * targhooks.c (speculation_safe_value_not_needed): New function.
5996         * target.def (have_speculation_safe_value): Update documentation.
5997         * doc/tm.texi: Regenerated.
5999 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
6001         * config/aarch64/iterators.md (ALLI_TI): New iterator.
6002         * config/aarch64/aarch64.md (despeculate_copy<ALLI_TI:mode>): New
6003         expand.
6004         (despeculate_copy<ALLI:mode>_insn): New insn.
6005         (despeculate_copyti_insn): New insn.
6006         (despeculate_simple<ALLI:mode>): New insn
6007         (despeculate_simpleti): New insn.
6008         * config/aarch64/aarch64.c (aarch64_speculation_safe_value): New
6009         function.
6010         (TARGET_SPECULATION_SAFE_VALUE): Redefine to
6011         aarch64_speculation_safe_value.
6012         (aarch64_print_operand): Handle const0_rtx in modifier 'H'.
6014 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
6016         * config/aarch64/aarch64-speculation.cc: New file.
6017         * config/aarch64/aarch64-passes.def (pass_track_speculation): Add
6018         before pass_reorder_blocks.
6019         * config/aarch64/aarch64-protos.h (make_pass_track_speculation): Add
6020         prototype.
6021         * config/aarch64/aarch64.c (aarch64_conditional_register_usage): Fix
6022         X14 and X15 when tracking speculation.
6023         * config/aarch64/aarch64.md (register name constants): Add
6024         SPECULATION_TRACKER_REGNUM and SPECULATION_SCRATCH_REGNUM.
6025         (unspec): Add UNSPEC_SPECULATION_TRACKER.
6026         (speculation_barrier): New insn attribute.
6027         (cmp<mode>): Allow SP in comparisons.
6028         (speculation_tracker): New insn.
6029         (speculation_barrier): Add speculation_barrier attribute.
6030         * config/aarch64/t-aarch64: Add make rule for aarch64-speculation.o.
6031         * config.gcc (aarch64*-*-*): Add aarch64-speculation.o to extra_objs.
6032         * doc/invoke.texi (AArch64 Options): Document -mtrack-speculation.
6034 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
6036         * config/aarch64/aarch64.md (cb<optab><mode>1): Disable when
6037         aarch64_track_speculation is true.
6038         (tb<optab><mode>1): Likewise.
6039         * config/aarch64/aarch64.c (aarch64_split_compare_regs): Do not
6040         generate CB[N]Z when tracking speculation.
6041         (aarch64_split_compare_and_swap): Likewise.
6042         (aarch64_split_atomic_op): Likewise.
6044 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
6046         * config/aarch64/aarch64.opt (mtrack-speculation): New target option.
6048 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
6050         * config/aarch64.md (unspecv): Add UNSPECV_SPECULAION_BARRIER.
6051         (speculation_barrier): New insn.
6053 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
6055         * config/arm/unspecs.md (unspecv): Add VUNSPEC_SPECULATION_BARRIER.
6056         * config/arm/arm.md (speculation_barrier): New expand.
6057         (speculation_barrier_insn): New pattern.
6059 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
6061         * builtin-types.def (BT_FN_PTR_PTR_VAR): New function type.
6062         (BT_FN_I1_I1_VAR, BT_FN_I2_I2_VAR, BT_FN_I4_I4_VAR): Likewise.
6063         (BT_FN_I8_I8_VAR, BT_FN_I16_I16_VAR): Likewise.
6064         * builtin-attrs.def (ATTR_NOVOPS_NOTHROW_LEAF_LIST): New attribute
6065         list.
6066         * builtins.def (BUILT_IN_SPECULATION_SAFE_VALUE_N): New builtin.
6067         (BUILT_IN_SPECULATION_SAFE_VALUE_PTR): New internal builtin.
6068         (BUILT_IN_SPECULATION_SAFE_VALUE_1): Likewise.
6069         (BUILT_IN_SPECULATION_SAFE_VALUE_2): Likewise.
6070         (BUILT_IN_SPECULATION_SAFE_VALUE_4): Likewise.
6071         (BUILT_IN_SPECULATION_SAFE_VALUE_8): Likewise.
6072         (BUILT_IN_SPECULATION_SAFE_VALUE_16): Likewise.
6073         * builtins.c (expand_speculation_safe_value): New function.
6074         (expand_builtin): Call it.
6075         * doc/cpp.texi: Document predefine __HAVE_SPECULATION_SAFE_VALUE.
6076         * doc/extend.texi: Document __builtin_speculation_safe_value.
6077         * doc/md.texi: Document "speculation_barrier" pattern.
6078         * doc/tm.texi.in: Pull in TARGET_SPECULATION_SAFE_VALUE and
6079         TARGET_HAVE_SPECULATION_SAFE_VALUE.
6080         * doc/tm.texi: Regenerated.
6081         * target.def (have_speculation_safe_value, speculation_safe_value): New
6082         hooks.
6083         * targhooks.c (default_have_speculation_safe_value): New function.
6084         (default_speculation_safe_value): New function.
6085         * targhooks.h (default_have_speculation_safe_value): Add prototype.
6086         (default_speculation_safe_value): Add prototype.
6088 2018-07-31  David Malcolm  <dmalcolm@redhat.com>
6090         * dump-context.h (dump_context::dump_loc): New decl.
6091         * dumpfile.c (dump_context::dump_loc): New member function.
6092         (dump_context::dump_gimple_stmt_loc): Reimplement using dump_loc
6093         and dump_gimple_stmt.
6094         (dump_context::dump_gimple_expr_loc): Likewise, using
6095         dump_gimple_expr.
6096         (dump_context::dump_generic_expr_loc): Likewise, using
6097         dump_generic_expr.
6098         (dump_context::dump_printf_loc_va): Likewise, using
6099         dump_printf_va.
6100         (dump_context::begin_scope): Explicitly using the global function
6101         "dump_loc", rather than the member function.
6103 2018-07-31  Martin Sebor  <msebor@redhat.com>
6105         PR tree-optimization/86741
6106         * tree-vrp.c (vrp_prop::check_mem_ref): Avoid incomplete types.
6108 2018-07-31  Andreas Krebbel  <krebbel@linux.ibm.com>
6110         * config/s390/s390.c (s390_expand_setmem): Make the unrolling to
6111         depend on whether prefetch instructions will be emitted or not.
6112         Use TARGET_SETMEM_PFD for checking whether prefetch instructions
6113         will be emitted or not.
6114         * config/s390/s390.h (TARGET_SETMEM_PREFETCH_DISTANCE)
6115         (TARGET_SETMEM_PFD): New macros.
6117 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
6119         * tree-vectorizer.h (stmt_vec_info): Turn back into a typedef.
6120         (NULL_STMT_VEC_INFO): Delete.
6121         (stmt_vec_info::operator*): Likewise.
6122         (stmt_vec_info::operator gimple *): Likewise.
6123         * tree-vect-loop.c (vectorizable_reduction): Use NULL instead
6124         of NULL_STMT_VEC_INFO.
6125         * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
6126         (vect_reassociating_reduction_p): Likewise.
6127         * tree-vect-stmts.c (vect_build_gather_load_calls): Likewise.
6128         (vectorizable_store): Likewise.
6129         * tree-vectorizer.c (vec_info::set_vinfo_for_stmt): Likewise.
6130         (vec_info::free_stmt_vec_infos): Likewise.
6132 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
6134         * tree-vectorizer.h (vect_stmt_in_region_p): Delete.
6135         * tree-vectorizer.c (vect_stmt_in_region_p): Likewise.
6137 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
6139         * tree-vectorizer.h (vec_info::new_vinfo_for_stmt)
6140         (vec_info::set_vinfo_for_stmt, vec_info::free_stmt_vec_infos)
6141         (vec_info::free_stmt_vec_info): New private member functions.
6142         (set_stmt_vec_info_vec, free_stmt_vec_infos, vinfo_for_stmt)
6143         (set_vinfo_for_stmt, new_stmt_vec_info, free_stmt_vec_info): Delete.
6144         * tree-parloops.c (gather_scalar_reductions): Remove calls to
6145         set_stmt_vec_info_vec and free_stmt_vec_infos.
6146         * tree-vect-loop.c (_loop_vec_info): Remove call to
6147         set_stmt_vec_info_vec.
6148         * tree-vect-stmts.c (new_stmt_vec_info, set_stmt_vec_info_vec)
6149         (free_stmt_vec_infos, free_stmt_vec_info): Delete in favor of...
6150         * tree-vectorizer.c (vec_info::new_stmt_vec_info)
6151         (vec_info::set_vinfo_for_stmt, vec_info::free_stmt_vec_infos)
6152         (vec_info::free_stmt_vec_info): ...these new functions.  Remove
6153         assignments in {vec_info::,}new_stmt_vec_info that are redundant
6154         with the clearing in the xcalloc.
6155         (stmt_vec_info_vec): Delete.
6156         (vec_info::vec_info): Don't call set_stmt_vec_info_vec.
6157         (vectorize_loops): Likewise.
6158         (vec_info::~vec_info): Remove argument from call to
6159         free_stmt_vec_infos.
6160         (vec_info::add_stmt): Remove vinfo argument from call to
6161         new_stmt_vec_info.
6163 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
6165         * tree-vectorizer.h (free_stmt_vec_info): Take a stmt_vec_info
6166         rather than a gimple stmt.
6167         * tree-vect-stmts.c (free_stmt_vec_info): Likewise.  Don't free
6168         information for pattern statements when passed the original
6169         statement; instead wait to be passed the pattern statement itself.
6170         Don't call set_vinfo_for_stmt here.
6171         (free_stmt_vec_infos): Update call to free_stmt_vec_info.
6172         * tree-vect-loop.c (_loop_vec_info::~loop_vec_info): Don't free
6173         stmt_vec_infos here.
6174         * tree-vect-slp.c (_bb_vec_info::~bb_vec_info): Likewise.
6175         * tree-vectorizer.c (vec_info::remove_stmt): Nullify the statement's
6176         stmt_vec_infos entry.
6178 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
6180         * tree-vectorizer.h (vec_info::replace_stmt): Declare.
6181         * tree-vectorizer.c (vec_info::replace_stmt): New function.
6182         * tree-vect-slp.c (vect_remove_slp_scalar_calls): Use it.
6183         * tree-vect-stmts.c (vectorizable_call): Likewise.
6184         (vectorizable_simd_clone_call): Likewise.
6186 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
6188         * tree-vectorizer.h (vec_info::remove_stmt): Declare.
6189         * tree-vectorizer.c (vec_info::remove_stmt): New function.
6190         * tree-vect-loop-manip.c (vect_set_loop_condition): Use it.
6191         * tree-vect-loop.c (vect_transform_loop): Likewise.
6192         * tree-vect-slp.c (vect_schedule_slp): Likewise.
6193         * tree-vect-stmts.c (vect_remove_stores): Likewise.
6195 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
6197         * tree-vectorizer.h (vec_info::lookup_dr): New member function.
6198         (vect_dr_stmt): Delete.
6199         * tree-vectorizer.c (vec_info::lookup_dr): New function.
6200         * tree-vect-loop-manip.c (vect_update_inits_of_drs): Use it instead
6201         of DR_VECT_AUX.
6202         * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
6203         (vect_analyze_data_ref_dependence, vect_record_base_alignments)
6204         (vect_verify_datarefs_alignment, vect_peeling_supportable)
6205         (vect_analyze_data_ref_accesses, vect_prune_runtime_alias_test_list)
6206         (vect_analyze_data_refs): Likewise.
6207         (vect_slp_analyze_data_ref_dependence): Likewise.  Take a vec_info
6208         argument.
6209         (vect_find_same_alignment_drs): Likewise.
6210         (vect_slp_analyze_node_dependences): Update calls accordingly.
6211         (vect_analyze_data_refs_alignment): Likewise.  Use vec_info::lookup_dr
6212         instead of DR_VECT_AUX.
6213         (vect_get_peeling_costs_all_drs): Take a loop_vec_info instead
6214         of a vector data references.  Use vec_info::lookup_dr instead of
6215         DR_VECT_AUX.
6216         (vect_peeling_hash_get_lowest_cost): Update calls accordingly.
6217         (vect_enhance_data_refs_alignment): Likewise.  Use vec_info::lookup_dr
6218         instead of DR_VECT_AUX.
6220 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
6222         * tree-vectorizer.h (_loop_vec_info::unaligned_dr): Change to
6223         dr_vec_info.
6224         * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Update
6225         accordingly.
6226         * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
6227         * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
6228         (vect_gen_prolog_loop_niters): Likewise.
6230 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
6232         * tree-vectorizer.h (set_dr_misalignment, dr_misalignment)
6233         (DR_TARGET_ALIGNMENT, aligned_access_p, known_alignment_for_access_p)
6234         (vect_known_alignment_in_bytes, vect_dr_behavior)
6235         (vect_get_scalar_dr_size): Take references as dr_vec_infos
6236         instead of data_references.  Update calls to other routines for
6237         which the same change has been made.
6238         * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Take
6239         dr_vec_infos instead of stmt_vec_infos.
6240         (vect_analyze_data_ref_dependence): Update call accordingly.
6241         (vect_slp_analyze_data_ref_dependence)
6242         (vect_record_base_alignments): Use DR_VECT_AUX.
6243         (vect_calculate_target_alignment, vect_compute_data_ref_alignment)
6244         (vect_update_misalignment_for_peel, verify_data_ref_alignment)
6245         (vector_alignment_reachable_p, vect_get_data_access_cost)
6246         (vect_peeling_supportable, vect_analyze_group_access_1)
6247         (vect_analyze_group_access, vect_analyze_data_ref_access)
6248         (vect_vfa_segment_size, vect_vfa_access_size, vect_vfa_align)
6249         (vect_compile_time_alias, vect_small_gap_p)
6250         (vectorizable_with_step_bound_p, vect_duplicate_ssa_name_ptr_info):
6251         (vect_supportable_dr_alignment): Take references as dr_vec_infos
6252         instead of data_references.  Update calls to other routines for
6253         which the same change has been made.
6254         (vect_verify_datarefs_alignment, vect_get_peeling_costs_all_drs)
6255         (vect_find_same_alignment_drs, vect_analyze_data_refs_alignment)
6256         (vect_slp_analyze_and_verify_node_alignment)
6257         (vect_analyze_data_ref_accesses, vect_prune_runtime_alias_test_list)
6258         (vect_create_addr_base_for_vector_ref, vect_create_data_ref_ptr)
6259         (vect_setup_realignment): Use dr_vec_infos.  Update calls after
6260         above changes.
6261         (_vect_peel_info::dr): Replace with...
6262         (_vect_peel_info::dr_info): ...this new field.
6263         (vect_peeling_hash_get_most_frequent)
6264         (vect_peeling_hash_choose_best_peeling): Update accordingly.
6265         (vect_peeling_hash_get_lowest_cost):
6266         (vect_enhance_data_refs_alignment): Likewise.  Update calls to other
6267         routines for which the same change has been made.
6268         (vect_peeling_hash_insert): Likewise.  Take a dr_vec_info instead of a
6269         data_reference.
6270         * tree-vect-loop-manip.c (get_misalign_in_elems)
6271         (vect_gen_prolog_loop_niters): Use dr_vec_infos.  Update calls after
6272         above changes.
6273         * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
6274         * tree-vect-stmts.c (vect_get_store_cost, vect_get_load_cost)
6275         (vect_truncate_gather_scatter_offset, compare_step_with_zero)
6276         (get_group_load_store_type, get_negative_load_store_type)
6277         (vect_get_data_ptr_increment, vectorizable_store)
6278         (vectorizable_load): Likewise.
6279         (ensure_base_align): Take a dr_vec_info instead of a data_reference.
6280         Update calls to other routines for which the same change has been made.
6282 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
6284         * tree-vectorizer.h (vec_info::move_dr): New member function.
6285         (dataref_aux): Rename to...
6286         (dr_vec_info): ...this and add "dr" and "stmt" fields.
6287         (_stmt_vec_info::dr_aux): Update accordingly.
6288         (_stmt_vec_info::data_ref_info): Delete.
6289         (STMT_VINFO_GROUPED_ACCESS, DR_GROUP_FIRST_ELEMENT)
6290         (DR_GROUP_NEXT_ELEMENT, DR_GROUP_SIZE, DR_GROUP_STORE_COUNT)
6291         (DR_GROUP_GAP, DR_GROUP_SAME_DR_STMT, REDUC_GROUP_FIRST_ELEMENT):
6292         (REDUC_GROUP_NEXT_ELEMENT, REDUC_GROUP_SIZE): Use dr_aux.dr instead
6293         of data_ref.
6294         (STMT_VINFO_DATA_REF): Likewise.  Turn into an lvalue.
6295         (STMT_VINFO_DR_INFO): New macro.
6296         (DR_VECT_AUX): Use STMT_VINFO_DR_INKFO and vect_dr_stmt.
6297         (set_dr_misalignment): Update after rename of dataref_aux.
6298         (vect_dr_stmt): Move earlier in file.  Return dr_aux.stmt.
6299         * tree-vect-stmts.c (new_stmt_vec_info): Remove redundant
6300         initialization of STMT_VINFO_DATA_REF.
6301         * tree-vectorizer.c (vec_info::move_dr): New function.
6302         * tree-vect-patterns.c (vect_recog_bool_pattern)
6303         (vect_recog_mask_conversion_pattern)
6304         (vect_recog_gather_scatter_pattern): Use it.
6305         * tree-vect-data-refs.c (vect_analyze_data_refs): Initialize
6306         the "dr" and "stmt" fields of dr_vec_info instead of
6307         STMT_VINFO_DATA_REF.
6309 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
6311         * tree-vectorizer.h (_stmt_vec_info::pattern_stmt_p): New field.
6312         (is_pattern_stmt_p): Use it.
6313         * tree-vect-patterns.c (vect_init_pattern_stmt): Set pattern_stmt_p
6314         on pattern statements.
6316 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
6318         * tree-vect-patterns.c (vect_mark_pattern_stmts): Take the
6319         original stmt as a stmt_vec_info rather than a gimple stmt.
6320         (vect_pattern_recog_1): Take the statement directly as a
6321         stmt_vec_info, rather than via a gimple_stmt_iterator.
6322         Update call to vect_mark_pattern_stmts.
6323         (vect_pattern_recog): Update calls accordingly.
6325 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
6327         * tree-vectorizer.h (vect_get_vec_defs_for_stmt_copy)
6328         (vect_get_vec_def_for_stmt_copy): Take a vec_info rather than
6329         a vect_def_type for the first argument.
6330         * tree-vect-stmts.c (vect_get_vec_defs_for_stmt_copy): Likewise.
6331         (vect_get_vec_def_for_stmt_copy): Likewise.  Return the original
6332         operand if it isn't defined by a vectorized statement.
6333         (vect_build_gather_load_calls): Remove the mask_dt argument and
6334         update calls to vect_get_vec_def_for_stmt_copy.
6335         (vectorizable_bswap): Likewise the dt argument.
6336         (vectorizable_call): Update calls to vectorizable_bswap and
6337         vect_get_vec_def_for_stmt_copy.
6338         (vectorizable_simd_clone_call, vectorizable_assignment)
6339         (vectorizable_shift, vectorizable_operation, vectorizable_condition)
6340         (vectorizable_comparison): Update calls to
6341         vect_get_vec_def_for_stmt_copy.
6342         (vectorizable_store): Likewise.  Remove now-unnecessary calls to
6343         vect_is_simple_use.
6344         (vect_get_loop_based_defs): Remove dt argument and update call
6345         to vect_get_vec_def_for_stmt_copy.
6346         (vectorizable_conversion): Update calls to vect_get_loop_based_defs
6347         and vect_get_vec_def_for_stmt_copy.
6348         (vectorizable_load): Update calls to vect_build_gather_load_calls
6349         and vect_get_vec_def_for_stmt_copy.
6350         * tree-vect-loop.c (vect_create_epilog_for_reduction)
6351         (vectorizable_reduction, vectorizable_live_operation): Update calls
6352         to vect_get_vec_def_for_stmt_copy.
6354 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
6356         * tree-vect-data-refs.c (vect_record_base_alignment): Replace vec_info
6357         and gimple stmt arguments with a stmt_vec_info.
6358         (vect_record_base_alignments): Update calls accordingly.
6359         * tree-vect-slp.c (vect_record_max_nunits): Replace vec_info
6360         and gimple stmt arguments with a stmt_vec_info.
6361         (vect_build_slp_tree_1): Remove vinfo argument and update call
6362         to vect_record_max_nunits.
6363         (vect_build_slp_tree_2): Update calls to vect_build_slp_tree_1
6364         and vect_record_max_nunits.
6366 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
6368         * tree-vectorizer.h (nested_in_vect_loop_p): Move further down
6369         file and take a stmt_vec_info instead of a gimple stmt.
6370         (supportable_widening_operation, vect_finish_replace_stmt)
6371         (vect_finish_stmt_generation, vect_get_store_rhs)
6372         (vect_get_vec_def_for_operand_1, vect_get_vec_def_for_operand)
6373         (vect_get_vec_defs, vect_init_vector, vect_transform_stmt)
6374         (vect_remove_stores, vect_analyze_stmt, vectorizable_condition)
6375         (vect_get_smallest_scalar_type, vect_check_gather_scatter)
6376         (vect_create_data_ref_ptr, bump_vector_ptr)
6377         (vect_permute_store_chain, vect_setup_realignment)
6378         (vect_transform_grouped_load, vect_record_grouped_load_vectors)
6379         (vect_create_addr_base_for_vector_ref, vectorizable_live_operation)
6380         (vectorizable_reduction, vectorizable_induction)
6381         (get_initial_def_for_reduction, is_simple_and_all_uses_invariant)
6382         (vect_get_place_in_interleaving_chain): Take stmt_vec_infos rather
6383         than gimple stmts as arguments.
6384         * tree-vect-data-refs.c (vect_get_smallest_scalar_type)
6385         (vect_preserves_scalar_order_p, vect_slp_analyze_node_dependences)
6386         (can_group_stmts_p, vect_check_gather_scatter)
6387         (vect_create_addr_base_for_vector_ref, vect_create_data_ref_ptr)
6388         (bump_vector_ptr, vect_permute_store_chain, vect_setup_realignment)
6389         (vect_permute_load_chain, vect_shift_permute_load_chain)
6390         (vect_transform_grouped_load)
6391         (vect_record_grouped_load_vectors): Likewise.
6392         * tree-vect-loop.c (vect_fixup_reduc_chain)
6393         (get_initial_def_for_reduction, vect_create_epilog_for_reduction)
6394         (vectorize_fold_left_reduction, is_nonwrapping_integer_induction)
6395         (vectorizable_reduction, vectorizable_induction)
6396         (vectorizable_live_operation, vect_loop_kill_debug_uses): Likewise.
6397         * tree-vect-patterns.c (type_conversion_p, adjust_bool_stmts)
6398         (vect_get_load_store_mask): Likewise.
6399         * tree-vect-slp.c (vect_get_place_in_interleaving_chain)
6400         (vect_analyze_slp_instance, vect_mask_constant_operand_p): Likewise.
6401         * tree-vect-stmts.c (vect_mark_relevant)
6402         (is_simple_and_all_uses_invariant)
6403         (exist_non_indexing_operands_for_use_p, process_use)
6404         (vect_init_vector_1, vect_init_vector, vect_get_vec_def_for_operand_1)
6405         (vect_get_vec_def_for_operand, vect_get_vec_defs)
6406         (vect_finish_stmt_generation_1, vect_finish_replace_stmt)
6407         (vect_finish_stmt_generation, vect_truncate_gather_scatter_offset)
6408         (compare_step_with_zero, vect_get_store_rhs, get_group_load_store_type)
6409         (get_negative_load_store_type, get_load_store_type)
6410         (vect_check_load_store_mask, vect_check_store_rhs)
6411         (vect_build_gather_load_calls, vect_get_strided_load_store_ops)
6412         (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
6413         (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
6414         (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
6415         (get_group_alias_ptr_type, vectorizable_store, hoist_defs_of_uses)
6416         (vectorizable_load, vectorizable_condition, vectorizable_comparison)
6417         (vect_analyze_stmt, vect_transform_stmt, vect_remove_stores)
6418         (supportable_widening_operation): Likewise.
6420 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
6422         * tree-vect-data-refs.c (vect_describe_gather_scatter_call): Take
6423         a stmt_vec_info instead of a gcall.
6424         (vect_check_gather_scatter): Update call accordingly.
6425         * tree-vect-loop-manip.c (iv_phi_p): Take a stmt_vec_info instead
6426         of a gphi.
6427         (vect_can_advance_ivs_p, vect_update_ivs_after_vectorizer)
6428         (slpeel_update_phi_nodes_for_loops):): Update calls accordingly.
6429         * tree-vect-loop.c (vect_transform_loop_stmt): Take a stmt_vec_info
6430         instead of a gimple stmt.
6431         (vect_transform_loop): Update calls accordingly.
6432         * tree-vect-slp.c (vect_split_slp_store_group): Take and return
6433         stmt_vec_infos instead of gimple stmts.
6434         (vect_analyze_slp_instance): Update use accordingly.
6435         * tree-vect-stmts.c (read_vector_array, write_vector_array)
6436         (vect_clobber_variable, vect_stmt_relevant_p, permute_vec_elements)
6437         (vect_use_strided_gather_scatters_p, vect_build_all_ones_mask)
6438         (vect_build_zero_merge_argument, vect_get_gather_scatter_ops)
6439         (vect_gen_widened_results_half, vect_get_loop_based_defs)
6440         (vect_create_vectorized_promotion_stmts, can_vectorize_live_stmts):
6441         Take a stmt_vec_info instead of a gimple stmt and pass stmt_vec_infos
6442         down to subroutines.
6444 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
6446         * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Change the type
6447         of the worklist from a vector of gimple stmts to a vector of
6448         stmt_vec_infos.
6449         * tree-vect-stmts.c (vect_mark_relevant, process_use)
6450         (vect_mark_stmts_to_be_vectorized): Likewise
6452 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
6454         * tree-vect-loop.c (vect_analyze_loop_operations): Look up the
6455         statement before passing it to vect_analyze_stmt.
6456         (vect_create_epilog_for_reduction): Use a stmt_vec_info to walk
6457         the chain of phi vector definitions.  Track the exit phi via its
6458         stmt_vec_info.
6459         (vectorizable_reduction): Set cond_stmt_vinfo directly from the
6460         STMT_VINFO_REDUC_DEF.
6461         * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Use
6462         stmt_vec_infos to handle the statement chains.
6463         (vect_get_slp_defs): Record the first statement in the node
6464         using a stmt_vec_info.
6465         * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Look up
6466         statements here and pass their stmt_vec_info down to subroutines.
6467         (vect_init_vector_1): Hoist call to vinfo_for_stmt and pass it
6468         down to vect_finish_stmt_generation.
6469         (vect_init_vector, vect_get_vec_defs, vect_finish_replace_stmt)
6470         (vect_finish_stmt_generation): Call vinfo_for_stmt and pass
6471         stmt_vec_infos to subroutines.
6472         (vect_remove_stores): Use stmt_vec_infos to handle the statement
6473         chains.
6475 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
6477         * tree-vect-data-refs.c (vect_slp_analyze_node_dependences):
6478         (vect_check_gather_scatter, vect_create_data_ref_ptr, bump_vector_ptr)
6479         (vect_permute_store_chain, vect_setup_realignment)
6480         (vect_permute_load_chain, vect_shift_permute_load_chain)
6481         (vect_transform_grouped_load): Use stmt_vec_info rather than gimple
6482         stmts internally, and when passing values to other vectorizer routines.
6483         * tree-vect-loop-manip.c (vect_can_advance_ivs_p): Likewise.
6484         * tree-vect-loop.c (vect_analyze_scalar_cycles_1)
6485         (vect_analyze_loop_operations, get_initial_def_for_reduction)
6486         (vect_create_epilog_for_reduction, vectorize_fold_left_reduction)
6487         (vectorizable_reduction, vectorizable_induction)
6488         (vectorizable_live_operation, vect_transform_loop_stmt)
6489         (vect_transform_loop): Likewise.
6490         * tree-vect-patterns.c (vect_reassociating_reduction_p)
6491         (vect_recog_widen_op_pattern, vect_recog_mixed_size_cond_pattern)
6492         (vect_recog_bool_pattern, vect_recog_gather_scatter_pattern): Likewise.
6493         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
6494         (vect_slp_analyze_node_operations_1): Likewise.
6495         * tree-vect-stmts.c (vect_mark_relevant, process_use)
6496         (exist_non_indexing_operands_for_use_p, vect_init_vector_1)
6497         (vect_mark_stmts_to_be_vectorized, vect_get_vec_def_for_operand)
6498         (vect_finish_stmt_generation_1, get_group_load_store_type)
6499         (get_load_store_type, vect_build_gather_load_calls)
6500         (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
6501         (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
6502         (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
6503         (vectorizable_store, vectorizable_load, vectorizable_condition)
6504         (vectorizable_comparison, vect_analyze_stmt, vect_transform_stmt)
6505         (supportable_widening_operation): Likewise.
6506         (vect_get_vector_types_for_stmt): Likewise.
6507         * tree-vectorizer.h (vect_dr_behavior): Likewise.
6509 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
6511         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
6512         (vect_slp_analyze_node_dependences, vect_analyze_data_ref_accesses)
6513         (vect_permute_store_chain, vect_permute_load_chain)
6514         (vect_shift_permute_load_chain, vect_transform_grouped_load): Avoid
6515         repeated stmt_vec_info lookups.
6516         * tree-vect-loop-manip.c (vect_can_advance_ivs_p): Likewise.
6517         (vect_update_ivs_after_vectorizer): Likewise.
6518         * tree-vect-loop.c (vect_is_simple_reduction): Likewise.
6519         (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
6520         * tree-vect-patterns.c (adjust_bool_stmts): Likewise.
6521         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
6522         (vect_bb_slp_scalar_cost): Likewise.
6523         * tree-vect-stmts.c (get_group_alias_ptr_type): Likewise.
6525 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
6527         * tree-vect-data-refs.c (vect_check_gather_scatter): Pass the
6528         gcall rather than the generic gimple stmt to gimple_call_internal_fn.
6529         (vect_get_smallest_scalar_type, can_group_stmts_p): Use dyn_cast
6530         to get gassigns and gcalls, rather than operating on generc gimple
6531         stmts.
6532         * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p)
6533         (vect_mark_stmts_to_be_vectorized, vectorizable_store)
6534         (vectorizable_load, vect_analyze_stmt): Likewise.
6535         * tree-vect-loop.c (vectorizable_reduction): Likewise gphi.
6537 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
6539         * tree-vectorizer.h (get_earlier_stmt, get_later_stmt): Take and
6540         return stmt_vec_infos rather than gimple stmts.  Do not accept
6541         null arguments.
6542         (vect_find_last_scalar_stmt_in_slp): Return a stmt_vec_info instead
6543         of a gimple stmt.
6544         * tree-vect-slp.c (vect_find_last_scalar_stmt_in_slp): Likewise.
6545         Update use of get_later_stmt.
6546         (vect_get_constant_vectors): Update call accordingly.
6547         (vect_schedule_slp_instance): Likewise
6548         * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Likewise.
6549         (vect_slp_analyze_instance_dependence): Likewise.
6550         (vect_preserves_scalar_order_p): Update use of get_earlier_stmt.
6552 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
6554         * tree-vectorizer.h (stmt_info_for_cost::stmt): Replace with...
6555         (stmt_info_for_cost::stmt_info): ...this new field.
6556         (add_stmt_costs): Update accordingly.
6557         * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost)
6558         (vect_get_known_peeling_cost): Likewise.
6559         (vect_estimate_min_profitable_iters): Likewise.
6560         * tree-vect-stmts.c (record_stmt_cost): Likewise.
6562 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
6564         * tree-vectorizer.h (_loop_vec_info::may_misalign_stmts): Change
6565         from an auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
6566         * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Update
6567         accordingly.
6568         * tree-vect-loop-manip.c (vect_create_cond_for_align_checks): Likewise.
6570 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
6572         * tree-vectorizer.h (_stmt_vec_info::same_dr_stmt): Change from
6573         a gimple stmt to a stmt_vec_info.
6574         * tree-vect-stmts.c (vectorizable_load): Update accordingly.
6576 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
6578         * tree-vectorizer.h (vec_info::grouped_stores): Change from
6579         an auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
6580         (_loop_vec_info::reduction_chains): Likewise.
6581         * tree-vect-loop.c (vect_fixup_scalar_cycles_with_patterns): Update
6582         accordingly.
6583         * tree-vect-slp.c (vect_analyze_slp): Likewise.
6585 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
6587         * tree-vectorizer.h (_stmt_vec_info::first_element): Change from
6588         a gimple stmt to a stmt_vec_info.
6589         (_stmt_vec_info::next_element): Likewise.
6590         * tree-vect-data-refs.c (vect_update_misalignment_for_peel)
6591         (vect_slp_analyze_and_verify_node_alignment)
6592         (vect_analyze_group_access_1, vect_analyze_group_access)
6593         (vect_small_gap_p, vect_prune_runtime_alias_test_list)
6594         (vect_create_data_ref_ptr, vect_record_grouped_load_vectors)
6595         (vect_supportable_dr_alignment): Update accordingly.
6596         * tree-vect-loop.c (vect_fixup_reduc_chain): Likewise.
6597         (vect_fixup_scalar_cycles_with_patterns, vect_is_slp_reduction)
6598         (vect_is_simple_reduction, vectorizable_reduction): Likewise.
6599         * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
6600         * tree-vect-slp.c (vect_build_slp_tree_1)
6601         (vect_attempt_slp_rearrange_stmts, vect_supported_load_permutation_p)
6602         (vect_split_slp_store_group, vect_analyze_slp_instance)
6603         (vect_analyze_slp, vect_transform_slp_perm_load): Likewise.
6604         * tree-vect-stmts.c (vect_model_store_cost, vect_model_load_cost)
6605         (get_group_load_store_type, get_load_store_type)
6606         (get_group_alias_ptr_type, vectorizable_store, vectorizable_load)
6607         (vect_transform_stmt, vect_remove_stores): Likewise.
6609 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
6611         * tree-vectorizer.h (vect_dr_stmt): Return a stmt_vec_info rather
6612         than a gimple stmt.
6613         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
6614         (vect_slp_analyze_data_ref_dependence, vect_record_base_alignments)
6615         (vect_calculate_target_alignmentm, vect_compute_data_ref_alignment)
6616         (vect_update_misalignment_for_peel, vect_verify_datarefs_alignment)
6617         (vector_alignment_reachable_p, vect_get_data_access_cost)
6618         (vect_get_peeling_costs_all_drs, vect_peeling_hash_get_lowest_cost)
6619         (vect_peeling_supportable, vect_enhance_data_refs_alignment)
6620         (vect_find_same_alignment_drs, vect_analyze_data_refs_alignment)
6621         (vect_analyze_group_access_1, vect_analyze_group_access)
6622         (vect_analyze_data_ref_access, vect_analyze_data_ref_accesses)
6623         (vect_vfa_access_size, vect_small_gap_p, vect_analyze_data_refs)
6624         (vect_supportable_dr_alignment): Remove vinfo_for_stmt from the
6625         result of vect_dr_stmt and use the stmt_vec_info instead of
6626         the associated gimple stmt.
6627         * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
6628         (vect_gen_prolog_loop_niters): Likewise.
6629         * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
6631 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
6633         * tree-vectorizer.h (_slp_tree::stmts): Change from a vec<gimple *>
6634         to a vec<stmt_vec_info>.
6635         * tree-vect-slp.c (vect_free_slp_tree): Update accordingly.
6636         (vect_create_new_slp_node): Take a vec<gimple *> instead of a
6637         vec<stmt_vec_info>.
6638         (_slp_oprnd_info::def_stmts): Change from a vec<gimple *>
6639         to a vec<stmt_vec_info>.
6640         (bst_traits::value_type, bst_traits::value_type): Likewise.
6641         (bst_traits::hash): Update accordingly.
6642         (vect_get_and_check_slp_defs): Change the stmts parameter from
6643         a vec<gimple *> to a vec<stmt_vec_info>.
6644         (vect_two_operations_perm_ok_p, vect_build_slp_tree_1): Likewise.
6645         (vect_build_slp_tree): Likewise.
6646         (vect_build_slp_tree_2): Likewise.  Update uses of
6647         SLP_TREE_SCALAR_STMTS.
6648         (vect_print_slp_tree): Update uses of SLP_TREE_SCALAR_STMTS.
6649         (vect_mark_slp_stmts, vect_mark_slp_stmts_relevant)
6650         (vect_slp_rearrange_stmts, vect_attempt_slp_rearrange_stmts)
6651         (vect_supported_load_permutation_p, vect_find_last_scalar_stmt_in_slp)
6652         (vect_detect_hybrid_slp_stmts, vect_slp_analyze_node_operations_1)
6653         (vect_slp_analyze_node_operations, vect_slp_analyze_operations)
6654         (vect_bb_slp_scalar_cost, vect_slp_analyze_bb_1)
6655         (vect_get_constant_vectors, vect_get_slp_defs)
6656         (vect_transform_slp_perm_load, vect_schedule_slp_instance)
6657         (vect_remove_slp_scalar_calls, vect_schedule_slp): Likewise.
6658         (vect_analyze_slp_instance): Build up a vec of stmt_vec_infos
6659         instead of gimple stmts.
6660         * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Change
6661         the stores parameter for a vec<gimple *> to a vec<stmt_vec_info>.
6662         (vect_slp_analyze_instance_dependence): Update uses of
6663         SLP_TREE_SCALAR_STMTS.
6664         (vect_slp_analyze_and_verify_node_alignment): Likewise.
6665         (vect_slp_analyze_and_verify_instance_alignment): Likewise.
6666         * tree-vect-loop.c (neutral_op_for_slp_reduction): Likewise.
6667         (get_initial_defs_for_reduction): Likewise.
6668         (vect_create_epilog_for_reduction): Likewise.
6669         (vectorize_fold_left_reduction): Likewise.
6670         * tree-vect-stmts.c (vect_prologue_cost_for_slp_op): Likewise.
6671         (vect_model_simple_cost, vectorizable_shift, vectorizable_load)
6672         (can_vectorize_live_stmts): Likewise.
6674 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
6676         * tree-vectorizer.h (_loop_vec_info::reductions): Change from an
6677         auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
6678         (vect_force_simple_reduction): Take and return stmt_vec_infos rather
6679         than gimple stmts.
6680         * tree-parloops.c (valid_reduction_p): Take a stmt_vec_info instead
6681         of a gimple stmt.
6682         (gather_scalar_reductions): Update after above interface changes.
6683         * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Likewise.
6684         (vect_is_simple_reduction): Take and return stmt_vec_infos rather
6685         than gimple stmts.
6686         (vect_force_simple_reduction): Likewise.
6687         * tree-vect-patterns.c (vect_pattern_recog_1): Update use of
6688         LOOP_VINFO_REDUCTIONS.
6689         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
6691 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
6693         * tree-vectorizer.h (_stmt_vec_info::reduc_def): Change from
6694         a gimple stmt to a stmt_vec_info.
6695         * tree-vect-loop.c (vect_active_double_reduction_p)
6696         (vect_force_simple_reduction, vectorizable_reduction): Update
6697         accordingly.
6699 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
6701         * tree-vectorizer.h (_slp_tree::vec_stmts): Change from a
6702         vec<gimple *> to a vec<stmt_vec_info>.
6703         * tree-vect-loop.c (vect_create_epilog_for_reduction): Change
6704         the reduction_phis argument from a vec<gimple *> to a
6705         vec<stmt_vec_info>.
6706         (vectorizable_reduction): Likewise the phis local variable that
6707         is passed to vect_create_epilog_for_reduction.  Update for new type
6708         of SLP_TREE_VEC_STMTS.
6709         (vectorizable_induction): Update for new type of SLP_TREE_VEC_STMTS.
6710         (vectorizable_live_operation): Likewise.
6711         * tree-vect-slp.c (vect_get_slp_vect_defs): Likewise.
6712         (vect_transform_slp_perm_load, vect_schedule_slp_instance): Likewise.
6714 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
6716         * tree-vectorizer.h (_stmt_vec_info::vectorized_stmt): Change from
6717         a gimple stmt to a stmt_vec_info.
6718         (vectorizable_condition, vectorizable_live_operation)
6719         (vectorizable_reduction, vectorizable_induction): Pass back the
6720         vectorized statement as a stmt_vec_info.
6721         * tree-vect-data-refs.c (vect_record_grouped_load_vectors): Update
6722         use of STMT_VINFO_VEC_STMT.
6723         * tree-vect-loop.c (vect_create_epilog_for_reduction): Likewise,
6724         accumulating the inner phis that feed the STMT_VINFO_VEC_STMT
6725         as stmt_vec_infos rather than gimple stmts.
6726         (vectorize_fold_left_reduction): Change vec_stmt from a gimple stmt
6727         to a stmt_vec_info.
6728         (vectorizable_live_operation): Likewise.
6729         (vectorizable_reduction, vectorizable_induction): Likewise,
6730         updating use of STMT_VINFO_VEC_STMT.
6731         * tree-vect-stmts.c (vect_get_vec_def_for_operand_1): Update use
6732         of STMT_VINFO_VEC_STMT.
6733         (vect_build_gather_load_calls, vectorizable_bswap, vectorizable_call)
6734         (vectorizable_simd_clone_call, vectorizable_conversion)
6735         (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
6736         (vectorizable_store, vectorizable_load, vectorizable_condition)
6737         (vectorizable_comparison, can_vectorize_live_stmts): Change vec_stmt
6738         from a gimple stmt to a stmt_vec_info.
6739         (vect_transform_stmt): Update use of STMT_VINFO_VEC_STMT.  Pass a
6740         pointer to a stmt_vec_info to the vectorizable_* routines.
6742 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
6744         * tree-vectorizer.h (_stmt_vec_info::related_stmt): Change from
6745         a gimple stmt to a stmt_vec_info.
6746         (is_pattern_stmt_p): Update accordingly.
6747         * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Likewise.
6748         (vect_record_grouped_load_vectors): Likewise.
6749         * tree-vect-loop.c (vect_determine_vf_for_stmt): Likewise.
6750         (vect_fixup_reduc_chain, vect_update_vf_for_slp): Likewise.
6751         (vect_model_reduction_cost): Likewise.
6752         (vect_create_epilog_for_reduction): Likewise.
6753         (vectorizable_reduction, vectorizable_induction): Likewise.
6754         * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
6755         Return the stmt_vec_info for the pattern statement.
6756         (vect_set_pattern_stmt): Update use of STMT_VINFO_RELATED_STMT.
6757         (vect_split_statement, vect_mark_pattern_stmts): Likewise.
6758         * tree-vect-slp.c (vect_detect_hybrid_slp_stmts): Likewise.
6759         (vect_detect_hybrid_slp, vect_get_slp_defs): Likewise.
6760         * tree-vect-stmts.c (vect_mark_relevant): Likewise.
6761         (vect_get_vec_def_for_operand_1, vectorizable_call): Likewise.
6762         (vectorizable_simd_clone_call, vect_analyze_stmt, new_stmt_vec_info)
6763         (free_stmt_vec_info, vect_is_simple_use): Likewise.
6765 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
6767         * tree-vectorizer.h (vect_finish_replace_stmt): Return a stmt_vec_info
6768         (vect_finish_stmt_generation): Likewise.
6769         * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
6770         (vect_finish_replace_stmt, vect_finish_stmt_generation): Likewise.
6771         (vect_build_gather_load_calls): Use the return value of the above
6772         functions instead of a separate call to vinfo_for_stmt.  Use narrow
6773         scopes for the input gimple stmt and wider scopes for the associated
6774         stmt_vec_info.  Use vec_info::lookup_def when setting these
6775         stmt_vec_infos from an SSA_NAME definition.
6776         (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
6777         (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
6778         (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
6779         (vectorizable_store, vectorizable_load, vectorizable_condition)
6780         (vectorizable_comparison): Likewise.
6781         * tree-vect-loop.c (vectorize_fold_left_reduction): Likewise.
6782         (vectorizable_reduction): Likewise.
6784 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
6786         * tree-vectorizer.h (vect_is_simple_use): Add an optional
6787         stmt_vec_info * parameter before the optional gimple **.
6788         * tree-vect-stmts.c (vect_is_simple_use): Likewise.
6789         (process_use, vect_get_vec_def_for_operand_1): Update callers.
6790         (vect_get_vec_def_for_operand, vectorizable_shift): Likewise.
6791         * tree-vect-loop.c (vectorizable_reduction): Likewise.
6792         (vectorizable_live_operation): Likewise.
6793         * tree-vect-patterns.c (type_conversion_p): Likewise.
6794         (vect_look_through_possible_promotion): Likewise.
6795         (vect_recog_rotate_pattern): Likewise.
6796         * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
6798 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
6800         * tree-vectorizer.h (stmt_vec_info): Temporarily change from
6801         a typedef to a wrapper class.
6802         (NULL_STMT_VEC_INFO): New macro.
6803         (vec_info::stmt_infos): Change to vec<stmt_vec_info>.
6804         (stmt_vec_info::operator*): New function.
6805         (stmt_vec_info::operator gimple *): Likewise.
6806         (set_vinfo_for_stmt): Use NULL_STMT_VEC_INFO.
6807         (add_stmt_costs): Likewise.
6808         * tree-vect-loop-manip.c (iv_phi_p): Likewise.
6809         * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost)
6810         (vect_get_known_peeling_cost): Likewise.
6811         (vect_estimate_min_profitable_iters): Likewise.
6812         * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
6813         * tree-vect-slp.c (vect_remove_slp_scalar_calls): Likewise.
6814         * tree-vect-stmts.c (vect_build_gather_load_calls): Likewise.
6815         (vectorizable_store, free_stmt_vec_infos): Likewise.
6816         (new_stmt_vec_info): Change return type of xcalloc to
6817         _stmt_vec_info *.
6819 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
6821         * tree-vectorizer.h (vec_info::lookup_single_use): Declare.
6822         * tree-vectorizer.c (vec_info::lookup_single_use): New function.
6823         * tree-vect-loop.c (vectorizable_reduction): Use it instead of
6824         a single_imm_use-based sequence.
6825         * tree-vect-stmts.c (supportable_widening_operation): Likewise.
6827 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
6829         * tree-vectorizer.h (vec_info::lookup_def): Declare.
6830         * tree-vectorizer.c (vec_info::lookup_def): New function.
6831         * tree-vect-patterns.c (vect_get_internal_def): Use it.
6832         (vect_widened_op_tree): Likewise.
6833         * tree-vect-stmts.c (vect_is_simple_use): Likewise.
6834         * tree-vect-loop.c (vect_analyze_loop_operations): Likewise.
6835         (vectorizable_reduction): Likewise.
6836         (vect_valid_reduction_input_p): Take a stmt_vec_info instead
6837         of a gimple *.
6838         (vect_is_slp_reduction): Update calls accordingly.  Use
6839         vec_info::lookup_def.
6840         (vect_is_simple_reduction): Likewise
6841         * tree-vect-slp.c (vect_detect_hybrid_slp_1): Use vec_info::lookup_def.
6843 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
6845         * tree-vectorizer.h (vec_info::lookup_stmt): Declare.
6846         * tree-vectorizer.c (vec_info::lookup_stmt): New function.
6847         * tree-vect-loop.c (vect_determine_vf_for_stmt): Use it instead
6848         of vinfo_for_stmt.
6849         (vect_determine_vectorization_factor, vect_analyze_scalar_cycles_1)
6850         (vect_compute_single_scalar_iteration_cost, vect_analyze_loop_form)
6851         (vect_update_vf_for_slp, vect_analyze_loop_operations)
6852         (vect_is_slp_reduction, vectorizable_induction)
6853         (vect_transform_loop_stmt, vect_transform_loop): Likewise.
6854         * tree-vect-patterns.c (vect_init_pattern_stmt):
6855         (vect_determine_min_output_precision_1, vect_determine_precisions)
6856         (vect_pattern_recog): Likewise.
6857         * tree-vect-stmts.c (vect_analyze_stmt, vect_transform_stmt): Likewise.
6858         * config/powerpcspe/powerpcspe.c (rs6000_density_test): Likewise.
6859         * config/rs6000/rs6000.c (rs6000_density_test): Likewise.
6860         * tree-vect-slp.c (vect_detect_hybrid_slp_stmts): Likewise.
6861         (vect_detect_hybrid_slp_1, vect_detect_hybrid_slp_2)
6862         (vect_detect_hybrid_slp): Likewise.  Change the walk_stmt_info
6863         info field from a loop to a loop_vec_info.
6865 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
6867         * tree-vectorizer.h (stmt_vec_info): Move typedef earlier in file.
6868         (vec_info::add_stmt): Declare.
6869         * tree-vectorizer.c (vec_info::add_stmt): New function.
6870         * tree-vect-data-refs.c (vect_create_data_ref_ptr): Use it.
6871         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Likewise.
6872         (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
6873         (vectorizable_induction): Likewise.
6874         * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Likewise.
6875         * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
6876         (vectorizable_simd_clone_call, vectorizable_store): Likewise.
6877         (vectorizable_load): Likewise.
6878         * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
6879         (vect_recog_bool_pattern, vect_recog_mask_conversion_pattern)
6880         (vect_recog_gather_scatter_pattern): Likewise.
6881         (append_pattern_def_seq): Likewise.  Remove a check that is
6882         performed by add_stmt itself.
6884 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
6886         * tree-vect-loop.c (vectorizable_reduction): Fix an instance in
6887         which make_ssa_name was called with new_stmt before new_stmt
6888         had been created.
6890 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
6892         * tree-vect-loop.c (vect_valid_reduction_input_p): New function,
6893         split out from...
6894         (vect_is_slp_reduction): ...here...
6895         (vect_is_simple_reduction): ...and here.  Remove repetition of tests
6896         that are already known to be false.
6898 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
6900         * tree-vectorizer.h (vect_free_slp_instance): Add a final_p parameter.
6901         * tree-vect-slp.c (vect_free_slp_tree): Likewise.  Don't update
6902         STMT_VINFO_NUM_SLP_USES when it's true.
6903         (vect_free_slp_instance): Add a final_p parameter and pass it to
6904         vect_free_slp_tree.
6905         (vect_build_slp_tree_2): Update call to vect_free_slp_instance.
6906         (vect_analyze_slp_instance): Likewise.
6907         (vect_slp_analyze_operations): Likewise.
6908         (vect_slp_analyze_bb_1): Likewise.
6909         * tree-vectorizer.c (vec_info): Likewise.
6910         * tree-vect-loop.c (vect_transform_loop): Likewise.
6912 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
6914         * tree-vect-loop.c (vectorizable_reduction): Assert that the
6915         function is not called for second and subsequent members of
6916         a reduction group.
6918 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
6920         * tree-vect-loop.c (get_initial_def_for_reduction): Move special
6921         cases for nested loops from here to ...
6922         (vect_create_epilog_for_reduction): ...here.  Only call
6923         vect_is_simple_use for inner-loop reductions.
6925 2018-07-31  Martin Liska  <mliska@suse.cz>
6927         PR gcov-profile/85338
6928         PR gcov-profile/85350
6929         PR gcov-profile/85372
6930         * profile.c (struct location_triplet): New.
6931         (struct location_triplet_hash): Likewise.
6932         (output_location): Do not output a BB that
6933         is already recorded for a line.
6934         (branch_prob): Use streamed_locations.
6936 2018-07-31  Martin Liska  <mliska@suse.cz>
6938         PR gcov-profile/85370
6939         * coverage.c (coverage_begin_function): Do not mark target
6940         clones as artificial functions.
6942 2018-07-31  Martin Liska  <mliska@suse.cz>
6944         PR gcov-profile/83813
6945         PR gcov-profile/84758
6946         PR gcov-profile/85217
6947         PR gcov-profile/85332
6948         * profile.c (branch_prob): Do not record GOTO expressions
6949         for GIMPLE statements which locations are already streamed.
6951 2018-07-31  Olivier Hainque  <hainque@adacore.com>
6953         * gcc.c (handle_spec_function): Accept a soft_matched_part
6954         argument, as do_spec_1.  Pass it down to ...
6955         (eval_spec_function): Accept a soft_matched_part argument,
6956         and pass it down to ...
6957         (do_spec_2): Accept a soft_matched_part argument, and pass
6958         it down to do_spec_1.
6959         (do_spec_1): Pass soft_matched_part to handle_spec_function.
6960         (handle_braces): Update call to handle_spec_function.
6961         (driver::set_up_specs): Update calls to do_spec_2.
6962         (compare_debug_dump_opt_spec_function): Likewise.
6963         (compare_debug_self_opt_spec_function): Likewise.
6965 2018-07-31  Olivier Hainque  <hainque@adacore.com>
6967         * common.opt (nolibc): New option.
6968         * doc/invoke.texi (Link Options): Document it.
6969         * gcc.c (LINK_GCC_C_SEQUENCE_SPEC): Honor nolibc.
6970         * config/alpha/linux.h: Likewise.
6971         * config/arc/elf.h: Likewise.
6972         * config/arm/uclinux-elf.h: Likewise.
6973         * config/arm/unknown-elf.h: Likewise.
6974         * config/avr/avrlibc.h: Likewise.
6975         * config/bfin/bfin.h: Likewise.
6976         * config/bfin/linux.h: Likewise.
6977         * config/bfin/uclinux.h: Likewise.
6978         * config/darwin.h: Likewise.
6979         * config/darwin10.h: Likewise.
6980         * config/darwin12.h: Likewise.
6981         * config/gnu-user.h: Likewise.
6982         * config/lm32/uclinux-elf.h: Likewise.
6983         * config/pa/pa-hpux11.h: Likewise.
6984         * config/pa/pa64-hpux.h: Likewise.
6985         * config/sparc/sparc.h: Likewise.
6987 2018-07-31  Olivier Hainque  <hainque@adacore.com>
6989         * gcc.c (getenv_spec_function): Prepend '/' to value for allowed
6990         undefined variables.
6992 2018-07-30  Segher Boessenkool  <segher@kernel.crashing.org>
6994         PR target/86640
6995         * config/arm/arm.c (arm_block_set_aligned_vect): Use gen_int_mode
6996         instead of GEN_INT.
6998 2018-07-30  Bernd Edlinger  <bernd.edlinger@hotmail.de>
7000         * tree-ssa-forwprop.c (simplify_builtin_call): Don't create a not NUL
7001         terminated string literal.
7003 2018-07-30  Segher Boessenkool  <segher@kernel.crashing.org>
7005         PR rtl-optimization/85160
7006         * combine.c (is_just_move): New function.
7007         (try_combine): Allow combining two instructions into two if neither of
7008         the original instructions was a move.
7010 2018-07-30  Alexander Monakov  <amonakov@ispras.ru>
7012         PR target/86673
7013         * doc/extend.texi (Global Register Variables): Discourage use of type
7014         qualifiers.
7015         (Local Register Variables): Likewise.
7017 2018-07-30  Richard Sandiford  <richard.sandiford@arm.com>
7019         PR tree-optimization/86506
7020         * hwint.h (ceil_log2): Resync with hwint.c implementation.
7022 2018-07-30  Ilya Leoshkevich  <iii@linux.ibm.com>
7024         PR target/86547
7025         * lra-constraints.c (spill_hard_reg_in_range): When selecting the
7026         hard_regno, make sure no insn between `from` and `to` clobbers it.
7028 2018-07-30  Cesar Philippidis  <cesar@codesourcery.com>
7029             Tom de Vries  <tdevries@suse.de>
7031         * config/nvptx/nvptx.c (PTX_GANG_DEFAULT): Rename to ...
7032         (PTX_DEFAULT_RUNTIME_DIM): ... this.
7033         (nvptx_goacc_validate_dims): Set default worker and gang dims to
7034         PTX_DEFAULT_RUNTIME_DIM.
7035         (nvptx_dim_limit): Ignore GOMP_DIM_WORKER.
7037 2018-07-29  John David Anglin  <danglin@gcc.gnu.org>
7039         * config/pa/pa.c (pa_output_addr_vec): Align address table.
7040         * config/pa/pa.h (JUMP_TABLES_IN_TEXT_SECTION): Revise comment.
7041         * config/pa/pa32-linux.h (JUMP_TABLES_IN_TEXT_SECTION): Define.
7043 2018-07-27  Michael Meissner  <meissner@linux.ibm.com>
7045         * config/rs6000/constraints.md (wG constraint): Delete, no longer
7046         used.
7047         * config/rs6000/predicates.md (p9_fusion_reg_operand): Rename
7048         predicate to reflect toc fusion has been deleted.
7049         (toc_fusion_mem_raw): Delete, no longer used.
7050         (toc_fusion_mem_wrapped): Likewise.
7051         * config/rs6000/rs6000-cpus.def (POWERPC_MASKS): Delete toc
7052         fusion mask bit.
7053         * config/rs6000/rs6000-protos.h (fusion_wrap_memory_address):
7054         Delete, no longer used.
7055         * config/rs6000/rs6000.c (struct rs6000_reg_addr): Delete fields
7056         meant to be used for toc fusion.
7057         (rs6000_debug_print_mode): Delete toc fusion debugging.
7058         (rs6000_debug_reg_global): Likewise.
7059         (rs6000_init_hard_regno_mode_ok): Delete setting up fields for toc
7060         fusion and secondary reload support that were never used.
7061         (rs6000_option_override_internal): Delete TOC fusion, that was only
7062         partially defined, and it did not work unless you also used the
7063         -mcmodel= switch.
7064         (rs6000_legitimate_address_p): Delete TOC fusion support.
7065         (rs6000_opt_masks): Likewise.
7066         (fusion_wrap_memory_address): Delete function, no longer used.
7067         (fusion_split_address); Delete TOC fusion support.
7068         * config/rs6000/rs6000.h (TARGET_TOC_FUSION_INT): Delete, no
7069         longer used with toc fusion being deleted.
7070         (TARGET_TOC_FUSION_FP): Likewise.
7071         * config/rs6000/rs6000.md (UNSPEC_FUSION_ADDIS): Delete TOC fusion
7072         UNSPEC.
7073         (toc fusion spliter): Delete TOC fusion support.
7074         (toc_fusionload_<mode>): Likewise.
7075         (toc_fusionload_di): Likewise.
7076         (fusion_gpr_load_<mode>): Delete generator function, this insn no
7077         longer needs to be named.  Rename predicate to delete TOC fusion.
7078         (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_load): Likewise.
7079         (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_store): Likewise.
7080         (fusion_vsx_<P:mode>_<GPR_FUSION:mode>_load): Likewise.
7081         (fusion_vsx_<P:mode>_<GPR_FUSION:mode>_store): Likewise.
7082         (p9 fusion peephole2s): Rename predicate to delete TOC fusion.
7084 2018-07-27  Kelvin Nilsen  <kelvin@gcc.gnu.org>
7086         * doc/extend.texi (Basic PowerPC Built-in Functions Available on
7087         ISA 2.05):  Replace __uint128_t with __uint128 and __int128_t with
7088         __int128 in built-in function prototypes.
7089         (PowerPC AltiVec Built-in Functions on ISA 2.07): Likewise.
7090         (PowerPC AltiVec Built-in Functions on ISA 3.0): Likewise.
7092 2018-07-27  Martin Sebor  <msebor@redhat.com>
7094         PR tree-optimization/86696
7095         * tree-ssa-strlen.c (get_min_string_length): Handle all integer
7096         types, including enums.
7097         (handle_char_store): Be prepared for the above function to fail.
7099 2018-07-26  Qing Zhao  <qing.zhao@oracle.com>
7101         * builtins.c (inline_expand_builtin_string_cmp): Disable inlining
7102         when optimization level is lower than 2 or optimize for size.
7104 2018-07-26  Martin Sebor  <msebor@redhat.com>
7106         PR tree-optimization/86043
7107         PR tree-optimization/86042
7108         * tree-ssa-strlen.c (handle_builtin_memcpy): Handle strict overlaps.
7109         (get_string_cst_length): Rename...
7110         (get_min_string_length): ...to this.  Add argument.
7111         (handle_char_store): Extend to handle multi-character stores by
7112         MEM_REF.
7113         * tree.c (initializer_zerop): Use new argument.  Handle MEM_REF.
7114         * tree.h (initializer_zerop): Add argument.
7116 2018-07-26  Jakub Jelinek  <jakub@redhat.com>
7118         PR middle-end/86660
7119         * omp-low.c (scan_sharing_clauses): Don't ignore map clauses for
7120         declare target to variables if they have always,{to,from,tofrom} map
7121         kinds.
7123 2018-07-26  Martin Liska  <mliska@suse.cz>
7125         PR lto/86548
7126         * lto-wrapper.c: Add linker_output as prefix
7127         for ltrans_output_file.
7129 2018-07-26  Segher Boessenkool  <segher@kernel.crashing.org>
7131         PR rtl-optimization/85805
7132         * combine.c (reg_nonzero_bits_for_combine): Only use the last set
7133         value for hard registers if that was written in the same mode.
7135 2018-07-26  Martin Liska  <mliska@suse.cz>
7137         PR gcov-profile/86536
7138         * gcov.c (format_gcov): Use printf format %.*f directly
7139         and do not handle special values.
7141 2018-07-25  Claudiu Zissulescu  <claziss@synopsys.com>
7143         * common/config/arc/arc-common.c (arc_option_optimization_table):
7144         Update default optimizations for size.
7146 2018-07-25  Claudiu Zissulescu  <claziss@synopsys.com>
7148         * config/arc/arc.md (movsf_insn): Add short instruction selection.
7149         * config/arc/constraints.md (CfZ): New constraint.
7150         * config/arc/fpu.md (addssf3_fpu): Use CfZ constraint.
7151         (subsf3_fpu): Likewise.
7152         (cmpsf_fpu): Likewise.
7153         (cmpsf_fpu_uneq): Likewise.
7155 2018-07-25  Claudiu Zissulescu  <claziss@synopsys.com>
7157         * config/arc/arc.c (compact_memory_operand_p): Check for uncached
7158         accesses as well.
7159         (arc_is_uncached_mem_p): uncached applies to both the variable and
7160         the pointer.
7162 2018-07-25  Claudiu Zissulescu  <claziss@synopsys.com>
7164         * config/arc/arc.h (ADDITIONAL_REGISTER_NAMES): Add additional
7165         register names.
7167 2018-07-25  David Malcolm  <dmalcolm@redhat.com>
7169         * optinfo-emit-json.cc (class optrecord_json_writer): Convert
7170         field "m_scopes" from vec to auto_vec.
7172 2018-07-25  Martin Liska  <mliska@suse.cz>
7174         * config/powerpcspe/powerpcspe-protos.h (rs6000_loop_align): Fix
7175         return type.
7177 2018-07-25  Richard Biener  <rguenther@suse.de>
7179         PR debug/86654
7180         * dwarf2out.c (dwarf2out_decl): Do not handle nested functions
7181         special wrt context_die late.
7182         (gen_subprogram_die): Re-use DIEs in local scope.
7184 2018-07-25  Richard Sandiford  <richard.sandiford@arm.com>
7186         PR tree-optimization/86644
7187         * hwint.c (ceil_log2): Fix comment.  Return 0 for 0.
7189 2018-07-25  Martin Liska  <mliska@suse.cz>
7191         PR middle-end/86645
7192         * dumpfile.c: And excluded values with TDF_ALL_VALUES.
7193         * dumpfile.h (enum dump_flag): Defince TDF_ALL_VALUES.
7195 2018-07-25  Martin Liska  <mliska@suse.cz>
7197         PR sanitizer/79635
7198         * params.def: Explain ASan abbreviation and provide
7199         a documentation link.
7201 2018-07-24  Martin Sebor  <msebor@redhat.com>
7203         PR tree-optimization/86622
7204         PR tree-optimization/86532
7205         * builtins.h (string_length): Declare.
7206         * builtins.c (c_strlen): Correct handling of non-constant offsets.
7207         (check_access): Be prepared for non-constant length ranges.
7208         (string_length): Make extern.
7209         * expr.c (string_constant): Only handle the minor non-constant
7210         array index.  Use string_constant to compute the length of
7211         a generic string constant.
7213 2018-07-24  Richard Sandiford  <richard.sandiford@arm.com>
7215         PR tree-optimization/86618
7216         * tree-vect-stmts.c (vectorizable_call): Don't take the address
7217         of LOOP_VINFO_MASKS (loop_vinfo) when loop_vinfo is null.
7219 2018-07-24  David Malcolm  <dmalcolm@redhat.com>
7221         PR tree-optimization/86636
7222         * json.cc (json::object::set): Fix comment.  Add assertions.
7223         (json::array::append): Move here from json.h.  Add comment and an
7224         assertion.
7225         (json::string::string): Likewise.
7226         * json.h (json::array::append): Move to json.cc.
7227         (json::string::string): Likewise.
7228         * optinfo-emit-json.cc
7229         (optrecord_json_writer::impl_location_to_json): Assert that we
7230         aren't attempting to write out UNKNOWN_LOCATION, or an ad-hoc
7231         wrapper around it.  Expand the location once, rather than three
7232         times.
7233         (optrecord_json_writer::inlining_chain_to_json): Fix the check for
7234         UNKNOWN_LOCATION, to use LOCATION_LOCUS to look through ad-hoc
7235         wrappers.
7236         (optrecord_json_writer::optinfo_to_json): Likewise, in four
7237         places.  Fix some overlong lines.
7239 2018-07-24  Matthew Malcomson  <matthew.malcomson@arm.com>
7241         * config/aarch64/aarch64-simd.md
7242         (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w<mode>): Split into...
7243         (aarch64_<ANY_EXTEND:su>subw<mode>): ... This...
7244         (aarch64_<ANY_EXTEND:su>addw<mode>): ... And this.
7245         (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w<mode>_internal): Split into...
7246         (aarch64_<ANY_EXTEND:su>subw<mode>_internal): ... This...
7247         (aarch64_<ANY_EXTEND:su>addw<mode>_internal): ... And this.
7248         (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w2<mode>_internal): Split into...
7249         (aarch64_<ANY_EXTEND:su>subw2<mode>_internal): ... This...
7250         (aarch64_<ANY_EXTEND:su>addw2<mode>_internal): ... And this.
7252 2018-07-24  Jakub Jelinek  <jakub@redhat.com>
7254         PR middle-end/86627
7255         * expmed.c (expand_divmod): Punt if d == HOST_WIDE_INT_MIN
7256         and size > HOST_BITS_PER_WIDE_INT.  For size > HOST_BITS_PER_WIDE_INT
7257         and abs_d == d, do the power of two handling if profitable.
7259 2018-07-24  Richard Biener  <rguenther@suse.de>
7261         * match.pd: Add BIT_FIELD_REF canonicalizations.
7263 2018-07-23  Bernd Edlinger  <bernd.edlinger@hotmail.de>
7265         PR c/86617
7266         * genmatch.c (dt_operand::gen_match_op): Avoid folding volatile values.
7268 2018-07-23  Bernd Edlinger  <bernd.edlinger@hotmail.de>
7270         * gimple-fold.c (gimple_fold_builtin_printf): Don't create a not NUL
7271         terminated STRING_CST object.
7273 2018-07-23  Bernd Edlinger  <bernd.edlinger@hotmail.de>
7275         hsa-dump.c (dump_hsa_symbol): Avoid out of scope access to buf.
7277 2018-07-23  Segher Boessenkool  <segher@kernel.crashing.org>
7279         * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Adjust.
7280         * config/rs6000/rs6000-protos.h (rs6000_split_v4si_init): Delete.
7281         * config/rs6000/rs6000.c (rs6000_expand_vector_init): Always force
7282         the elements into a register.
7283         (rs6000_split_v4si_init_di_reg): Delete.
7284         (rs6000_split_v4si_init): Delete.
7285         * config/rs6000/vsx.md (unspec): Delete UNSPEC_VSX_VEC_INIT.
7286         (vsx_init_v4si): Rewrite as a define_expand.
7288 2018-07-23  Segher Boessenkool  <segher@kernel.crashing.org>
7290         * config/rs6000/rs6000.md (splitters for rldimi and rlwimi with the
7291         zero_extend argument from memory): New.
7293 2018-07-22  Martin Sebor  <msebor@redhat.com>
7295         PR bootstrap/86621
7296         * gimple-ssa-warn-alloca.c (alloca_call_type_by_arg): Avoid
7297         diagnosing calls with unknown arguments unless -Walloca-larger-than
7298         is restricted to less than PTRDIFF_MAX bytes.
7300 2018-07-22  Gerald Pfeifer  <gerald@pfeifer.com>
7302         * doc/gcov.texi (Invoking Gcov): Editorial changes.
7304 2018-07-20  David Malcolm  <dmalcolm@redhat.com>
7306         * pretty-print.c (text_info::set_location): Remove redundant
7307         "line_table" parameter from call to rich_location::set_range.
7309 2018-07-20  Martin Sebor  <msebor@redhat.com>
7311         PR middle-end/82063
7312         * builtins.c (expand_builtin_alloca): Adjust.
7313         * calls.c (alloc_max_size): Simplify.
7314         * cgraphunit.c (cgraph_node::expand): Adjust.
7315         * common.opt (larger_than_size, warn_frame_larger_than): Remove
7316         variables.
7317         (frame_larger_than_size): Same.
7318         (-Wframe-larger-than, -Wlarger-than, -Wstack-usage): Change options
7319         to take a HOST_WIDE_INT argument and accept a byte-size suffix.
7320         Initialize.
7321         * doc/invoke.texi (GCC Command Options): Document option arguments.
7322         Explain byte-size arguments and suffixes.
7323         (-Wvla-larger-than, -Wno-alloc-size-larger-than): Update.
7324         (-Wno-alloca-larger-than, -Wno-vla-larger-than): Same.
7325         (-Wframe-larger-than, -Wlarger-than, -Wstack-usage): Same.
7326         * doc/options.texi (UInteger): Expand.
7327         (Host_Wide_Int, ByteSize): Document new properties.
7328         * final.c (final_start_function_1): Include sizes in an error message.
7329         * function.c (frame_offset_overflow): Same.
7330         * gimple-ssa-warn-alloca.c (pass_walloca::gate): Adjust.
7331         (alloca_call_type_by_arg): Change function argument to HOST_WIDE_INT.
7332         Diagnose unbounded alloca calls only for limits of less than
7333         PTRDIFF_MAX.
7334         (alloca_call_type): Adjust.  Diagnose possibly out-of-bounds alloca
7335         calls and VLA size only for limits of less than PTRDIFF_MAX.  Same
7336         for alloca(0).
7337         (pass_walloca::execute): Adjust.  Diagnose alloca calls in loops
7338         only for limits of less than PTRDIFF_MAX.
7339         * langhooks-def.h (lhd_handle_option): Change function argument
7340         to HOST_WIDE_INT.
7341         * langhooks.c (lhd_handle_option): Same.
7342         * langhooks.h (handle_option): Same.
7343         * opt-functions.awk (switch_bit_fields): Handle Host_Wide_Int and
7344         ByteSize flags.
7345         (var_type, var_type_struct): Same.
7346         (var_set): Handle ByteSize flag.
7347         * optc-gen.awk: Add comments to output to ease debugging.  Make
7348         use of HOST_WIDE_INT where appropriate.
7349         * opts-gen-save.awk:  Use %lx to format unsigned long.
7350         * opth-gen.awk: Change function argument to HOST_WIDE_INT.
7351         * opts-common.c (integral_argument): Return HOST_WIDE_INT and add
7352         arguments.  Parse bytes-size suffixes.
7353         (enum_arg_to_value): Change function argument to HOST_WIDE_INT.
7354         (enum_value_to_arg): Same.
7355         (decode_cmdline_option): Handle cl_host_wide_int.  Adjust.
7356         (handle_option): Adjust.
7357         (generate_option): Change function argument to HOST_WIDE_INT.
7358         (cmdline_handle_error): Adjust.
7359         (read_cmdline_option): Change function argument to HOST_WIDE_INT.
7360         (set_option): Change function argument to HOST_WIDE_INT.
7361         (option_enabled): Handle cl_host_wide_int.
7362         (get_option_state): Handle CLVC_SIZE.
7363         (control_warning_option): Same.
7364         * opts.c (common_handle_option): Change function argument to
7365         HOST_WIDE_INT.  Remove handling of OPT_Walloca_larger_than_ and
7366         OPT_Wvla_larger_than_.
7367         * opts.h (enum cl_var_type): Add an enumerator.
7368         * stor-layout.c (layout_decl): Print a more meaningful warning.
7369         * toplev.c (output_stack_usage): Adjust.
7371 2018-07-20  Qing Zhao  <qing.zhao@oracle.com>
7373         * builtins.c (expand_builtin_memcmp): Delete the last parameter for
7374         call to inline_expand_builtin_string_cmp.
7375         (expand_builtin_strcmp): Likewise.
7376         (expand_builtin_strncmp): Likewise.
7377         (inline_string_cmp): Delete the last parameter, change char_type_node
7378         to unsigned_char_type_node for strcmp/strncmp, add conversions to the
7379         two operands.
7380         (inline_expand_builtin_string_cmp): Delete the last parameter, give up
7381         the inlining expansion on target where the type of the call has same or
7382         narrower precision than unsigned char.
7384 2018-07-20  David Malcolm  <dmalcolm@redhat.com>
7386         * Makefile.in (OBJS): Add json.o and optinfo-emit-json.o.
7387         (CFLAGS-optinfo-emit-json.o): Define TARGET_NAME.
7388         * common.opt (fsave-optimization-record): New option.
7389         * coretypes.h (struct kv_pair): Move here from dumpfile.c.
7390         * doc/invoke.texi (-fsave-optimization-record): New option.
7391         * dumpfile.c: Include "optinfo-emit-json.h".
7392         (struct kv_pair): Move to coretypes.h.
7393         (optgroup_options): Make non-static.
7394         (dump_context::end_scope): Call
7395         optimization_records_maybe_pop_dump_scope.
7396         * dumpfile.h (optgroup_options): New decl.
7397         * json.cc: New file.
7398         * json.h: New file.
7399         * optinfo-emit-json.cc: New file.
7400         * optinfo-emit-json.h: New file.
7401         * optinfo.cc: Include "optinfo-emit-json.h".
7402         (optinfo::emit): Call optimization_records_maybe_record_optinfo.
7403         (optinfo_enabled_p): Check optimization_records_enabled_p.
7404         (optinfo_wants_inlining_info_p): Likewise.
7405         * optinfo.h: Update comment.
7406         * profile-count.c (profile_quality_as_string): New function.
7407         * profile-count.h (profile_quality_as_string): New decl.
7408         (profile_count::quality): New accessor.
7409         * selftest-run-tests.c (selftest::run_tests): Call json_cc_tests
7410         and optinfo_emit_json_cc_tests.
7411         * selftest.h (selftest::json_cc_tests): New decl.
7412         (selftest::optinfo_emit_json_cc_tests): New decl.
7413         * toplev.c: Include "optinfo-emit-json.h".
7414         (compile_file): Call optimization_records_finish.
7415         (do_compile): Call optimization_records_start.
7416         * tree-ssa-live.c: Include optinfo.h.
7417         (remove_unused_scope_block_p): Retain inlining information if
7418         optinfo_wants_inlining_info_p returns true.
7420 2018-07-20  Richard Biener  <rguenther@suse.de>
7422         PR debug/86585
7423         * dwarf2out.c (dwarf2out_die_ref_for_decl): Test in_lto_p
7424         to cover -flto-partition=none.
7426 2018-07-20  Martin Liska  <mliska@suse.cz>
7428         * tree.h (DECL_LOCATION_RANGE): Remove unused macro.
7429         (get_decl_source_range): Remove unused function.
7431 2018-07-20  Richard Biener  <rguenther@suse.de>
7433         * tree-ssa-sccvn.h (struct vn_nary_op_s): Add next member.
7434         (struct vn_phi_s): Likewise.
7435         (struct vn_reference_s): Likewise.
7436         * tree-ssa-sccvn.c (vn_nary_op_hasher::equal): Add shortcut
7437         for searching the slot of an entry known to be in the hash itself.
7438         (vn_phi_hasher::equal): Likewise.
7439         (vn_reference_hasher::equal): Likewise.
7440         (last_inserted_ref, last_inserted_phi, last_inserted_nary): New
7441         globals.
7442         (optimistic_info, current_info): Remove, keeping only valid_info.
7443         (vn_reference_lookup_1): Remove fallback lookup.
7444         (vn_reference_lookup_2): Likewise.
7445         (vn_nary_op_lookup_1): Likewise.
7446         (vn_phi_lookup): Likewise.
7447         (vn_nary_build_or_lookup_1): Make sure to not chain the built
7448         hash element.
7449         (vn_reference_insert): Adjust, chain the inserted hash element
7450         at last_inserted_ref.
7451         (vn_reference_insert_pieces): Likewise.
7452         (visit_reference_op_call): Likewise.
7453         (vn_nary_op_insert_into): Chain the inserted hash element at
7454         last_inserted_nary.
7455         (vn_nary_op_insert_pieces): Adjust.
7456         (vn_nary_op_insert): Likewise.
7457         (vn_nary_op_insert_stmt): Likewise.
7458         (vn_phi_insert): Adjust, chain the inserted hash element at
7459         last_inserted_phi.
7460         (process_scc): Remove clearing and copying the optimistic
7461         table.  Instead remove elements inserted during an optimistic
7462         iteration from the single table we maintain.
7463         (init_scc_vn): Adjust.
7464         (free_scc_vn): Likewise.
7465         (sccvn_dom_walker::record_cond): Likewise.
7466         (sccvn_dom_walker::after_dom_children): Likewise.
7468 2018-07-19  Martin Sebor  <msebor@redhat.com>
7470         PR tree-optimization/84047
7471         PR tree-optimization/83776
7472         * tree-vrp.c (vrp_prop::check_mem_ref): New function.
7473         (check_array_bounds): Call it.
7475 2018-07-19  Martin Sebor  <msebor@redhat.com>
7477         * align.h (align_flags): Use member initialization.
7479 2018-07-19  David Malcolm  <dmalcolm@redhat.com>
7481         * Makefile.in (OBJS): Add optinfo.o.
7482         * coretypes.h (class symtab_node): New forward decl.
7483         (struct cgraph_node): New forward decl.
7484         (class varpool_node): New forward decl.
7485         * dump-context.h: New file.
7486         * dumpfile.c: Include "optinfo.h", "dump-context.h", "cgraph.h",
7487         "tree-pass.h".
7488         (refresh_dumps_are_enabled): Use optinfo_enabled_p.
7489         (set_dump_file): Call dumpfile_ensure_any_optinfo_are_flushed.
7490         (set_alt_dump_file): Likewise.
7491         (dump_context::~dump_context): New dtor.
7492         (dump_gimple_stmt): Move implementation to...
7493         (dump_context::dump_gimple_stmt): ...this new member function.
7494         Add the stmt to any pending optinfo, creating one if need be.
7495         (dump_gimple_stmt_loc): Move implementation to...
7496         (dump_context::dump_gimple_stmt_loc): ...this new member function.
7497         Start a new optinfo and add the stmt to it.
7498         (dump_gimple_expr): Move implementation to...
7499         (dump_context::dump_gimple_expr): ...this new member function.
7500         Add the stmt to any pending optinfo, creating one if need be.
7501         (dump_gimple_expr_loc): Move implementation to...
7502         (dump_context::dump_gimple_expr_loc): ...this new member function.
7503         Start a new optinfo and add the stmt to it.
7504         (dump_generic_expr): Move implementation to...
7505         (dump_context::dump_generic_expr): ...this new member function.
7506         Add the tree to any pending optinfo, creating one if need be.
7507         (dump_generic_expr_loc): Move implementation to...
7508         (dump_context::dump_generic_expr_loc): ...this new member
7509         function.  Add the tree to any pending optinfo, creating one if
7510         need be.
7511         (dump_printf): Move implementation to...
7512         (dump_context::dump_printf_va): ...this new member function.  Add
7513         the text to any pending optinfo, creating one if need be.
7514         (dump_printf_loc): Move implementation to...
7515         (dump_context::dump_printf_loc_va): ...this new member function.
7516         Start a new optinfo and add the stmt to it.
7517         (dump_dec): Move implementation to...
7518         (dump_context::dump_dec): ...this new member function.  Add the
7519         value to any pending optinfo, creating one if need be.
7520         (dump_context::dump_symtab_node): New member function.
7521         (dump_context::get_scope_depth): New member function.
7522         (dump_context::begin_scope): New member function.
7523         (dump_context::end_scope): New member function.
7524         (dump_context::ensure_pending_optinfo): New member function.
7525         (dump_context::begin_next_optinfo): New member function.
7526         (dump_context::end_any_optinfo): New member function.
7527         (dump_context::s_current): New global.
7528         (dump_context::s_default): New global.
7529         (dump_scope_depth): Delete global.
7530         (dumpfile_ensure_any_optinfo_are_flushed): New function.
7531         (dump_symtab_node): New function.
7532         (get_dump_scope_depth): Reimplement in terms of dump_context.
7533         (dump_begin_scope): Likewise.
7534         (dump_end_scope): Likewise.
7535         (selftest::temp_dump_context::temp_dump_context): New ctor.
7536         (selftest::temp_dump_context::~temp_dump_context): New dtor.
7537         (selftest::verify_item): New function.
7538         (ASSERT_IS_TEXT): New macro.
7539         (ASSERT_IS_TREE): New macro.
7540         (ASSERT_IS_GIMPLE): New macro.
7541         (selftest::test_capture_of_dump_calls): New test.
7542         (selftest::dumpfile_c_tests): Call it.
7543         * dumpfile.h (dump_printf, dump_printf_loc, dump_basic_block)
7544         (dump_generic_expr_loc, dump_generic_expr, dump_gimple_stmt_loc)
7545         (dump_gimple_stmt, dump_dec): Gather these related decls and add a
7546         descriptive comment.
7547         (dump_function, print_combine_total_stats, enable_rtl_dump_file)
7548         (dump_node, dump_bb): Move these unrelated decls.
7549         (class dump_manager): Add leading comment.
7550         * optinfo.cc: New file.
7551         * optinfo.h: New file.
7553 2018-07-19  Michael Collison  <michael.collison@arm.com>
7554             Richard Henderson <rth@redhat.com>
7556         * config/aarch64/aarch64.md (subv<GPI>4, usubv<GPI>4): New patterns.
7557         (subti): Handle op1 zero.
7558         (subvti4, usub4ti4): New.
7559         (*sub<GPI>3_compare1_imm): New.
7560         (sub<GPI>3_carryinCV): New.
7561         (*sub<GPI>3_carryinCV_z1_z2, *sub<GPI>3_carryinCV_z1): New.
7562         (*sub<GPI>3_carryinCV_z2, *sub<GPI>3_carryinCV): New.
7564 2018-07-19  Michael Collison  <michael.collison@arm.com>
7565             Richard Henderson <rth@redhat.com>
7567         * config/aarch64/aarch64.md: (addv<GPI>4, uaddv<GPI>4): New.
7568         (addti3): Create simpler code if low part is already known to be 0.
7569         (addvti4, uaddvti4): New.
7570         (*add<GPI>3_compareC_cconly_imm): New.
7571         (*add<GPI>3_compareC_cconly): New.
7572         (*add<GPI>3_compareC_imm): New.
7573         (*add<GPI>3_compareC): Rename from add<GPI>3_compare1; do not
7574         handle constants within this pattern..
7575         (*add<GPI>3_compareV_cconly_imm): New.
7576         (*add<GPI>3_compareV_cconly): New.
7577         (*add<GPI>3_compareV_imm): New.
7578         (add<GPI>3_compareV): New.
7579         (add<GPI>3_carryinC, add<GPI>3_carryinV): New.
7580         (*add<GPI>3_carryinC_zero, *add<GPI>3_carryinV_zero): New.
7581         (*add<GPI>3_carryinC, *add<GPI>3_carryinV): New.
7582         ((*add<GPI>3_compareC_cconly_imm): Replace 'ne' operator
7583         with 'comparison' operator.
7584         (*add<GPI>3_compareV_cconly_imm): Ditto.
7585         (*add<GPI>3_compareV_cconly): Ditto.
7586         (*add<GPI>3_compareV_imm): Ditto.
7587         (add<GPI>3_compareV): Ditto.
7588         (add<mode>3_carryinC): Ditto.
7589         (*add<mode>3_carryinC_zero): Ditto.
7590         (*add<mode>3_carryinC): Ditto.
7591         (add<mode>3_carryinV): Ditto.
7592         (*add<mode>3_carryinV_zero): Ditto.
7593         (*add<mode>3_carryinV): Ditto.
7595 2018-07-19  Michael Collison  <michael.collison@arm.com>
7596             Richard Henderson <rth@redhat.com>
7598         * config/aarch64/aarch64-modes.def (CC_V): New.
7599         * config/aarch64/aarch64-protos.h
7600         (aarch64_addti_scratch_regs): Declare
7601         (aarch64_subvti_scratch_regs): Declare.
7602         (aarch64_expand_subvti): Declare.
7603         (aarch64_gen_unlikely_cbranch): Declare
7604         * config/aarch64/aarch64.c (aarch64_select_cc_mode): Test
7605         for signed overflow using CC_Vmode.
7606         (aarch64_get_condition_code_1): Handle CC_Vmode.
7607         (aarch64_gen_unlikely_cbranch): New function.
7608         (aarch64_addti_scratch_regs): New function.
7609         (aarch64_subvti_scratch_regs): New function.
7610         (aarch64_expand_subvti): New function.
7612 2018-07-19  Andre Vieira  <andre.simoesdiasvieira@arm.com>
7614         * config/aarch64/aarch64-option-extensions.def: New entry for profile
7615         extension.
7616         * config/aarch64/aarch64.h (AARCH64_FL_PROFILE): New.
7617         * doc/invoke.texi (aarch64-feature-modifiers): New entry for profile
7618         extension.
7620 2018-07-19  Andre Vieira  <andre.simoesdiasvieira@arm.com>
7622         PR target/83009
7623         * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
7624         address check not strict.
7626 2018-07-19  Andre Vieira  <andre.simoesdiasvieira@arm.com>
7628         * config/aarch64/aarch64-simd.md (aarch64_simd_mov<VQ:mode>): Replace
7629         Umq with Umn.
7630         (store_pair_lanes<mode>): Likewise.
7631         * config/aarch64/aarch64-protos.h (aarch64_addr_query_type): Add new
7632         enum value 'ADDR_QUERY_LDP_STP_N'.
7633         * config/aarch64/aarch64.c (aarch64_addr_query_type): Likewise.
7634         (aarch64_print_address_internal): Add declaration.
7635         (aarch64_print_ldpstp_address): Remove.
7636         (aarch64_classify_address): Adapt mode for 'ADDR_QUERY_LDP_STP_N'.
7637         (aarch64_print_operand): Change printing of 'y'.
7638         * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Use
7639         new enum value 'ADDR_QUERY_LDP_STP_N', don't hardcode mode and use
7640         'true' rather than '1'.
7641         * config/aarch64/constraints.md (Uml): Likewise.
7642         (Uml): Rename to Umn.
7643         (Umq): Remove.
7645 2018-07-19  Richard Biener  <rguenther@suse.de>
7647         * tree-ssa-sccvn.h (struct vn_phi_s): Make phiargs member
7648         a trailing array.
7649         * tree-ssa-sccvn.c: Remove alloc-pool.h use.
7650         (vn_phi_hasher): Derive from nofree_ptr_hash and remove remove method.
7651         (vn_reference_hasher): Likewise.
7652         (struct vn_tables_s): Remove obstack and alloc-pool members.
7653         (vn_tables_obstack, vn_tables_insert_obstack): New global obstacks.
7654         (vn_nary_build_or_lookup_1): Manually build in vn_tables_insert_obstack.
7655         (vn_reference_insert): Allocate from obstack instead of from alloc-pool.
7656         (vn_reference_insert_pieces): Likewise.
7657         (alloc_vn_nary_op_noinit): Adjust.
7658         (vn_nary_op_insert_stmt): Allocate phiargs in-place.
7659         (vn_phi_eq): Adjust.
7660         (shared_lookup_phiargs): Remove.
7661         (vn_phi_lookup): Allocate temporary vn_phi_s on the stack.
7662         (vn_phi_insert): Allocate from obstack instead of from alloc-pool.
7663         (visit_reference_op_call): Likewise.
7664         (copy_nary, copy_phi, copy_reference): Remove.
7665         (process_scc): Rewind the obstack when iterating.  Do not
7666         copy the elements to valid_info but just move them from one
7667         hashtable to the other.
7668         (allocate_vn_table): Adjust.
7669         (free_vn_table): Likewise.
7670         (init_scc_vn): Likewise.
7671         (free_scc_vn): Likewise.
7673 2018-07-19  H.J. Lu  <hongjiu.lu@intel.com>
7675         PR target/86560
7676         * config/i386/i386.c (rest_of_insert_endbranch): Lookup
7677         indirect_return as function type attribute.
7678         (ix86_attribute_table): Change indirect_return to function
7679         type attribute.
7680         * doc/extend.texi: Update indirect_return attribute.
7682 2018-07-19  Aldy Hernandez  <aldyh@redhat.com>
7684         * wide-int.h (widest2_int): New.
7685         * gimple-fold.c (arith_overflowed_p): Use it.
7686         * tree.h (widest2_int_cst): New.
7687         * tree-vrp.c (wide_int_binop_overflow): Rename from
7688         vrp_int_const_binop.
7689         Rewrite to work on trees.
7690         (extract_range_from_multiplicative_op_1): Abstract code to...
7691         (wide_int_range_min_max): ...here.
7692         (wide_int_range_cross_product): ...and here.
7693         (extract_range_from_binary_expr_1): Abstract overflow code to...
7694         (wide_int_range_mult_wrapping): ...here.
7695         * tree-vrp.h (wide_int_range_cross_product): New.
7696         (wide_int_range_mult_wrapping): New.
7698 2018-07-19  Andrew Senkevich  <andrew.senkevich@intel.com>
7699             Julia Koval  <julia.koval@intel.com>
7701         * config/i386/x86-tune-costs.h (skylake_memcpy,
7702         skylake_memset): Replace rep_prefix with unrolling for size 512.
7704 2018-07-18  Kugan Vivekanandarajah  <kuganv@linaro.org>
7706         PR middle-end/86544
7707         * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): Handle
7708         comparision with EQ_EXPR in last stmt.
7710 2018-07-18  Kelvin Nilsen  <kelvin@gcc.gnu.org>
7712         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Rename
7713         this subsection to "PowerPC AltiVec/VSX Built-in Functions".
7714         (PowerPC AltiVec/VSX Built-in Functions): New name for subsection
7715         previously known as "PowerPC AltiVec Built-in Functions".  Move
7716         some material to new subsubsections "PowerPC AltiVec Built-in
7717         Functions on ISA 2.06" and "PowerPC AltiVec Built-in Functions on
7718         ISA 2.07".
7719         (PowerPC Altivec Built-in Functions on ISA 2.05): New subsubsection.
7720         (PowerPC Altivec Built-in Functions on ISA 2.06): Likewise.
7721         (PowerPC Altivec Built-in Functions on ISA 2.07): Likewise.
7722         (PowerPC Altivec Built-in Functions on ISA 3.0): Likewise.
7724 2018-07-18  Richard Biener  <rguenther@suse.de>
7726         PR tree-optimization/86557
7727         * tree-vect-patterns.c (vect_recog_divmod_pattern): Also handle
7728         EXACT_DIV_EXPR.
7730 2018-07-18  Ilya Leoshkevich  <iii@linux.ibm.com>
7732         * config/s390/s390.c (s390_function_profiler): Generate CFI.
7734 2018-07-17  Jeff Law  <law@redhat.com>
7736         * config/arm/arm.c (get_label_padding): Update for recent
7737         changes to label_to_alignment.
7739         PR tree-optimization/86010
7740         * tree-ssa-dse.c (compute_trims): Fix typo/thinko.
7742         * config/mips/mips.c (vr4130_align_insns): Update for recent
7743         changes to label_to_alignment.
7745         * config/frv/frv.c (frv_label_align): Update for recent changes
7746         to label_to_alignment.
7748         * config/nios2/nios2.c (nios2_label_align): Update for recent
7749         changes which dropped ALIGN_LABELS_LOG.
7751 2018-07-17  Andreas Schwab  <schwab@linux-m68k.org>
7753         * config/m68k/m68k.md (umulsi3_highpart+1, const_umulsi3_highpart)
7754         (smulsi3_highpart+1, const_smulsi3_highpart): Add CC_STATUS_INIT.
7756 2018-07-17  Claudiu Zissulescu  <claziss@synopsys.com>
7758         * config/arc/arc.c (arc_label_align): Use align_labels instead of
7759         deprecated align_labels_log.
7761 2018-07-17  Richard Biener  <rguenther@suse.de>
7763         PR lto/86456
7764         * dwarf2out.c (init_sections_and_labels): Always generate
7765         a debug_line_str_section for early LTO debug.
7766         (dwarf2out_finish): Reset debug_line_str_hash output early.
7767         Bump counter for extra dwarf5 .debug_loc labels to not conflict
7768         with fat LTO part.
7769         (dwarf2out_early_finish): Output debug_line_str.
7771 2018-07-17  Robin Dapp  <rdapp@linux.ibm.com>
7773         * config/s390/s390.c (preferred_la_operand_p): Do not use LA with
7774         index register on z196 or later.
7776 2018-07-17  Robin Dapp  <rdapp@linux.ibm.com>
7778         * config/s390/s390.c (s390_default_align): Set default function
7779         alignment to 16.
7780         (s390_override_options_after_change): Call s390_default align.
7781         (s390_option_override_internal): Call s390_default_align.
7782         (TARGET_OVERRIDE_OPTIONS_AFTER_CHANGE): Define.
7784 2018-07-17  Jakub Jelinek  <jakub@redhat.com>
7786         PR middle-end/86542
7787         * omp-low.c (create_task_copyfn): Copy over also fields corresponding
7788         to _looptemp_ clauses, other than the first two.
7790 2018-07-17  Martin Liska  <mliska@suse.cz>
7792         * opts.c: Do not enable OPT_falign_* for -Os.
7794 2018-07-17  Martin Liska  <mliska@suse.cz>
7796         * align.h (MAX_CODE_ALIGN): New.
7797         (MAX_CODE_ALIGN_VALUE): New.
7798         * common/config/i386/i386-common.c (ix86_handle_option):
7799         (MAX_CODE_ALIGN): Moved to align.h.
7800         * final.c (MAX_CODE_ALIGN): Likewise.
7801         * opts.c (parse_and_check_align_values):
7802         (MAX_CODE_ALIGN): Likewise.
7803         (MAX_CODE_ALIGN_VALUE): Likewise.
7805 2018-07-17  Martin Liska  <mliska@suse.cz>
7807         * config/i386/att.h (ASM_OUTPUT_ALIGN): Fix spacing
7808         in order to fulfil coding style.
7809         * config/i386/cygming.h (ASM_OUTPUT_ALIGN): Likewise.
7810         * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
7811         * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
7812         * config/iq2000/iq2000.h (ASM_OUTPUT_ALIGN): Likewise.
7813         * config/pa/pa.h (ASM_OUTPUT_ALIGN): Likewise.
7814         * config/sparc/sol2.h (ASM_OUTPUT_ALIGN_WITH_NOP): Likewise.
7815         * config/sparc/sparc.h (ASM_OUTPUT_ALIGN): Likewise.
7816         * config/visium/visium.h (ASM_OUTPUT_ALIGN): Likewise.
7817         (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
7819 2018-07-17  Martin Liska  <mliska@suse.cz>
7821         * align.h: New file.
7822         * config/alpha/alpha.c (alpha_align_insns_1): Use align_functions
7823         directly.
7824         * config/i386/i386.c (ix86_avoid_jump_mispredicts): Use new return type
7825         align_flags of label_to_alignment.
7826         * config/m32r/m32r.h (LOOP_ALIGN): Wrap returned values into
7827         align_flags class.
7828         * config/m68k/m68k.c: Do not use removed align_labels_value and
7829         align_loops_value.
7830         * config/nds32/nds32.h (JUMP_ALIGN): Wrap result into align_flags class.
7831         (LOOP_ALIGN): Likewise.
7832         (LABEL_ALIGN): Likewise.
7833         * config/powerpcspe/powerpcspe.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
7834         Remove not used macro.
7835         (rs6000_loop_align): Change return type to align_flags.
7836         (rs6000_loop_align_max_skip): Remove.
7837         * config/rs6000/rs6000-protos.h (rs6000_loop_align):
7838         Change return type to align_flags.
7839         * config/rs6000/rs6000.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
7840         Remove not used macro.
7841         (rs6000_loop_align):  Change return type to align_flags.
7842         (rs6000_loop_align_max_skip): Remove.
7843         * config/rx/rx.h (JUMP_ALIGN): Wrap integer values
7844         * config/rx/rx-protos.h (rx_align_for_label): Make it
7845         static function.
7846         * config/rx/rx.c (rx_align_for_label): Change return type
7847         to align_flags.
7848         (rx_max_skip_for_label): Remove TARGET_ASM_*_ALIGN_MAX_SKIP
7849         macro definitions.
7850         into align_flags class.
7851         (LABEL_ALIGN): Likewise.
7852         (LOOP_ALIGN): Likewise.
7853         * config/s390/s390.c (s390_label_align): Use align_flags
7854         class member.
7855         (s390_asm_output_function_label): Likewise.
7856         * config/sh/sh.c (sh_override_options_after_change):
7857         Use align_flags class directly without macros.
7858         (find_barrier): Likewise.
7859         (barrier_align): Likewise.
7860         (sh_loop_align): Likewise.
7861         * config/spu/spu.c (spu_option_override):
7862         Use align_flags_tuple::get_value instead of removed macros.
7863         (spu_sched_init): Likewise.
7864         * config/spu/spu.h (GTY): Likewise.
7865         * config/visium/visium.c (visium_option_override):
7866         Set "8" as default secondary alignment.
7867         * config/visium/visium.h (SUBALIGN_LOG): Define to 3
7868         in order to guarantee secondary alignment of 8.
7869         * coretypes.h: Include align.h header file.
7870         * doc/tm.texi: Remove TARGET_ASM_JUMP_ALIGN_MAX_SKIP,
7871         TARGET_ASM_LOOP_ALIGN_MAX_SKIP, TARGET_ASM_LABEL_ALIGN_MAX_SKIP
7872         and TARGET_ASM_LABEL_ALIGN_AFTER_BARRIER_MAX_SKIP macros.
7873         * doc/tm.texi.in: Likewise.
7874         * final.c (struct label_alignment): Remove not used structure.
7875         (LABEL_ALIGN): Change type to align_flags.
7876         (LOOP_ALIGN): Likewise.
7877         (JUMP_ALIGN): Likewise.
7878         (default_loop_align_max_skip): Remove.
7879         (default_label_align_max_skip): Likewise.
7880         (default_jump_align_max_skip): Likewise.
7881         (default_label_align_after_barrier_max_skip):
7882         (LABEL_TO_ALIGNMENT): Change to access label_align vector.
7883         (LABEL_TO_MAX_SKIP): Remove.
7884         (label_to_alignment): Return align_flags type instead of integer.
7885         (label_to_max_skip): Remove.
7886         (align_fuzz): Use align_flags type.
7887         (compute_alignments): Use align_flags type and use align_flags::max
7888         to combine multiple alignments.
7889         (grow_label_align): Grow vec instead of C array.
7890         (update_alignments): Assign just LABEL_TO_ALIGNMENT.
7891         (shorten_branches):  Use align_flags type and use align_flags::max
7892         to combine multiple alignments.
7893         (final_scan_insn_1): Remove usage of secondary alignment that comes
7894         from label alignment, but instead use proper secondary alignment
7895         which is computed in grow_label_align.
7896         * flags.h (struct align_flags_tuple): Move to align.h.
7897         (struct align_flags): Likewise.
7898         (state_align_loops): Rename to align_loops.
7899         (state_align_jumps): Rename to align_jumps.
7900         (state_align_labels): Rename to align_labels.
7901         (state_align_functions): Rename to align_functions.
7902         (align_loops_log): Remove.
7903         (align_jumps_log): Remove.
7904         (align_labels_log): Remove.
7905         (align_functions_log): Remove.
7906         (align_loops_max_skip): Remove.
7907         (align_jumps_max_skip): Remove.
7908         (align_labels_max_skip): Remove.
7909         (align_functions_max_skip): Remove.
7910         (align_loops_value): Remove.
7911         (align_jumps_value): Remove.
7912         (align_labels_value): Remove.
7913         (align_functions_value): Remove.
7914         * output.h (label_to_alignment): Change return type to align_flags.
7915         (label_to_max_skip): Remove.
7916         * target.def: Remove loop_align_max_skip, label_align_max_skip,
7917         jump_align_max_skip macros.
7918         * targhooks.h (default_loop_align_max_skip): Remove.
7919         (default_label_align_max_skip): Likewise.
7920         (default_jump_align_max_skip): Likewise.
7921         (default_label_align_after_barrier_max_skip): Remove.
7922         * toplev.c (read_log_maxskip): Use ::normalize function.
7923         (parse_N_M): Remove not used argument and also call ::normalize.
7924         (parse_alignment_opts): Do not pass unused arguments.
7925         * varasm.c (assemble_start_function): Use directly align_functions
7926         instead of removed macros.
7927         * system.h: Do not poison removed macros.
7929 2018-07-17  Jakub Jelinek  <jakub@redhat.com>
7931         PR middle-end/86539
7932         * gimplify.c (gimplify_omp_for): Ensure taskloop firstprivatized init
7933         and cond temporaries don't have reference type if iterator has
7934         pointer type.  For init use &for_pre_body instead of pre_p if
7935         for_pre_body is non-empty.
7937 2018-07-16  Segher Boessenkool  <segher@kernel.crashing.org>
7939         * config/rs6000/rs6000.md (trunc<mode>sf2): Expand truncates of
7940         double-double modes to SFmode directly directly.
7941         (trunc<mode>sf2_fprs): Delete.
7943 2018-07-16  Segher Boessenkool  <segher@kernel.crashing.org>
7945         * config/rs6000/rs6000.c (init_float128_ibm): Use the correct names
7946         for conversions between IFmode and the decimal floating point modes.
7947         (init_float128_ieee): Use the correct names for conversions between
7948         KFmode and the decimal floating point modes.
7950 2018-07-16  Segher Boessenkool  <segher@kernel.crashing.org>
7952         * config/rs6000/rs6000.c (init_float128_ibm): Use more correct names
7953         for the conversions between TDmode and IFmode.
7954         (init_float128_ieee): Use more correct names for the conversions
7955         between TDmode and KFmode.
7957 2018-07-16  Jakub Jelinek  <jakub@redhat.com>
7959         PR tree-optimization/86526
7960         * builtins.c (expand_builtin_memcmp): Formatting fixes.
7961         (inline_expand_builtin_string_cmp): Likewise.
7962         (inline_string_cmp): Likewise.  Use c_readstr instead of
7963         builtin_memcpy_read_str.  Add unit_mode temporary.
7965 2018-07-16  Bernd Edlinger  <bernd.edlinger@hotmail.de>
7967         PR middle-end/86528
7968         * builtins.c (check_access): Bail out if range[0] is no INTEGER_CST.
7969         * expr.c (string_constant): Fix the element size of ARRAY_TYPE.
7971 2018-07-16  Kelvin Nilsen  <kelvin@gcc.gnu.org>
7973         * doc/extend.texi (PowerPC AltiVec Built-in Functions):
7974         Alphabetize prototypes of built-in functions, separating out
7975         built-in functions that are listed in this section but should be
7976         described elsewhere.
7978 2018-07-16  Uros Bizjak  <ubizjak@gmail.com>
7980         PR target/86511
7981         * expmed.c (emit_store_flag): Do not emit setcc followed by a
7982         conditional move when trapping comparison was split to a
7983         non-trapping one (and vice versa).
7985 2018-07-16  Ilya Leoshkevich  <iii@linux.ibm.com>
7987         * config/s390/s390.c (s390_function_profiler): Generate nops
7988         instead of profiler call sequences.
7989         * config/s390/s390.opt: Add the new option.
7991 2018-07-16  Ilya Leoshkevich  <iii@linux.ibm.com>
7993         * config/s390/s390.c (s390_function_profiler): Generate
7994         __mcount_loc section.
7995         * config/s390/s390.opt: Add the new option.
7997 2018-07-16  Ilya Leoshkevich  <iii@linux.ibm.com>
7999         * common.opt: Add the new warning.
8000         * config/s390/s390.c (s390_function_profiler): Emit "brasl
8001         %r0,__fentry__" when -mfentry is specified.
8002         (s390_option_override_internal): Disallow -mfentry for 31-bit
8003         CPUs.
8004         * config/s390/s390.opt: Add the new option.
8006 2018-07-16  Richard Biener  <rguenther@suse.de>
8008         PR lto/86523
8009         * dwarf2out.c (dwarf2out_register_external_die): Assign DIE parents
8010         for function-local FUNCTION_DECL and RESULT_DECL immediately.
8012 2018-07-16  Martin Liska  <mliska@suse.cz>
8014         PR ipa/86529
8015         * ipa-pure-const.c (malloc_candidate_p): Revert ::get
8016         to ::get_create.
8018 2017-07-16  Claudiu Zissulescu  <claziss@synopsys.com>
8020         * config/arc/arcHS.md: Update ARCHS scheduling rules.
8022 2017-07-16  Claudiu Zissulescu  <claziss@synopsys.com>
8024         * config/arc/arc-arch.h (arc_tune_attr): Add new tune parameters
8025         for ARCHS4x.
8026         * config/arc/arc-cpus.def (hs4x): New cpu.
8027         (hs4xd): Likewise.
8028         * config/arc/arc-tables.opt: Regenerate.
8029         * config/arc/arc.c (arc_sched_issue_rate): New function.
8030         (TARGET_SCHED_ISSUE_RATE): Define.
8031         (TARGET_SCHED_EXPOSED_PIPELINE): Likewise.
8032         * config/arc/arc.md (attr type): Add fpu_fuse, fpu_sdiv, fpu_ddiv,
8033         fpu_cvt.
8034         (attr tune): Add ARCHS4x tune values.
8035         (attr tune_dspmpy): Define.
8036         (*tst): Correct instruction type.
8037         * config/arc/arcHS.md: Don't use this automaton for ARCHS4x cpus.
8038         * config/arc/arcHS4x.md: New file.
8039         * config/arc/fpu.md: Update instruction type attributes.
8040         * config/arc/t-multilib: Regenerate.
8042 2018-07-16  Tom de Vries  <tdevries@suse.de>
8044         PR debug/86455
8045         * var-tracking.c (vt_initialize): Fix pre_dec handling.
8047 2018-07-16  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
8049         * config/aarch64/atomics.md (aarch64_store_execlusive<mode>): Add
8050         early clobber.
8052 2018-07-16  Eric Botcazou  <ebotcazou@adacore.com>
8054         PR tree-optimization/86514
8055         * tree-ssa-reassoc.c (init_range_entry) <CASE_CONVERT>: Return for a
8056         conversion to a boolean type from a type with greater precision.
8058 2018-07-16  Tom de Vries  <tdevries@suse.de>
8060         * var-tracking.c (vt_initialize): Print adjusted insn slim if
8061         dump_flags request TDF_SLIM.
8063 2018-07-16  Aldy Hernandez  <aldyh@redhat.com>
8065         * fold-const.c (int_const_binop_1): Abstract...
8066         (wide_int_binop): ...wide int code here.
8067         (poly_int_binop): ...poly int code here.
8068         Abstract the rest of int_const_binop_1 into int_const_binop.
8069         * fold-const.h (wide_int_binop): New.
8070         * tree-vrp.c (vrp_int_const_binop): Call wide_int_binop.
8071         Remove useless PLUS/MINUS_EXPR case.
8072         (zero_nonzero_bits_from_vr): Move wide int code...
8073         (zero_nonzero_bits_from_bounds): ...here.
8074         (extract_range_from_binary_expr_1): Move mask optimization code...
8075         (range_easy_mask_min_max): ...here.
8076         * tree-vrp.h (zero_nonzero_bits_from_bounds): New.
8077         (range_easy_mask_min_max): New.
8079 2018-07-15  Jeff Law  <law@redhat.com>
8081         PR target/85993
8082         * config/sh/sh.c (output_mi_thunk): Remove dead conditional
8083         block.
8085 2018-07-14  Jim Wilson  <jimw@sifive.com>
8087         * config/riscv/linux.h (TARGET_ASM_FILE_END): New.
8089 2018-07-14  Paul Koning  <ni1d@arrl.net>
8091         * config/pdp11/pdp11.c (pdp11_rtx_costs): Bugfixes.
8093 2018-07-13  Jan Hubicka  <hubicka@ucw.cz>
8095         * lto-streamer-out.c (copy_function_or_variable): Dump info about
8096         copying section.
8098 2018-07-13  Bill Schmidt  <wschmidt@linux.ibm.com>
8099             Steve Munroe  <munroesj52@gmail.com>
8101         * config/rs6000/emmintrin.h (_mm_and_si128): New function.
8102         (_mm_andnot_si128): Likewise.
8103         (_mm_or_si128): Likewise.
8104         (_mm_xor_si128): Likewise.
8106 2018-07-13  Qing Zhao  <qing.zhao@oracle.com>
8108         PR middle-end/78809
8109         * builtins.c (expand_builtin_memcmp): Inline the calls first
8110         when result_eq is false.
8111         (expand_builtin_strcmp): Inline the calls first.
8112         (expand_builtin_strncmp): Likewise.
8113         (inline_string_cmp): New routine. Expand a string compare
8114         call by using a sequence of char comparison.
8115         (inline_expand_builtin_string_cmp): New routine. Inline expansion
8116         a call to str(n)cmp/memcmp.
8117         * doc/invoke.texi (--param builtin-string-cmp-inline-length):
8118         New option.
8119         * params.def (BUILTIN_STRING_CMP_INLINE_LENGTH): New.
8121 2018-07-13  Richard Earnshaw  <rearnsha@arm.com>
8123         * config/arm/driver-arm.c: Include arm-native.h.
8124         (host_detect_local_cpu): Use auto-generated data tables.
8125         (vendors, arm_cpu_table): Delete.  Move part information to ...
8126         * config/arm/arm-cpus.in: ... here.
8127         * config/arm/parsecpu.awk (gen_native): New function.
8128         (vendor, part): New CPU fields.
8129         (END): Add support for building the native CPU detection tables.
8130         * config/arm/t-arm (arm-native.h): Add build rule.
8131         (driver-arm.o): Add dependency on arm-native.h.
8133 2018-07-13  Richard Biener  <rguenther@suse.de>
8135         PR middle-end/85974
8136         * match.pd (addr1 - addr2): Allow either of the operand to
8137         have a conversion.
8139 2018-07-13  Tom de Vries  <tdevries@suse.de>
8141         * tree-inline.c (remap_ssa_name): Save and reuse debug exprs generated
8142         in remap_ssa_name.
8144 2018-07-13  Jackson Woodruff  <jackson.woodruff@arm.com>
8146         * config/aarch64/aarch64.c (aarch64_operands_adjust_ok_for_ldpstp): Use
8147         arrays instead of numbered variables.
8149 2018-07-13  Eric Botcazou  <ebotcazou@adacore.com>
8151         * config/sparc/sparc-protos.h (sparc_compute_frame_size): Delete.
8152         * config/sparc/sparc.c (sparc_compute_frame_size): Make static.
8154 2018-07-13  Richard Biener  <rguenther@suse.de>
8156         PR debug/86452
8157         * dwarf2out.c (gen_type_die_with_usage): Use scope_die_for
8158         instead of get_context_die.
8160 2018-07-13  Kugan Vivekanandarajah  <kuganv@linaro.org>
8161             Richard Biener  <rguenther@suse.de>
8163         PR middle-end/86489
8164         * tree-ssa-loop-niter.c (number_of_iterations_popcount): Check
8165         that the loop latch destination where phi is defined.
8167 2018-07-12  Kito Cheng  <kito.cheng@gmail.com>
8169         * config/riscv/riscv.c (enum riscv_privilege_levels): Add UNKNOWN_MODE.
8170         (riscv_expand_epilogue): Add assertion to check interrupt mode.
8171         (riscv_set_current_function): Extract getting interrupt type to new
8172         function.
8173         (riscv_get_interrupt_type): New function.
8174         (riscv_merge_decl_attributes): New function, checking interrupt type is
8175         same.
8176         (TARGET_MERGE_DECL_ATTRIBUTES): Define.
8178 2018-07-12  Paul Koning  <ni1d@arrl.net>
8180         * config/pdp11/pdp11.c (pdp11_output_def): Fix typo in .set
8181         directive.
8183 2018-07-12  Paul Koning  <ni1d@arrl.net>
8185         * doc/rtl.texi (REG_NONNEG): Remove decrement and branch until
8186         zero reference, add doloop_end instead.
8187         * doc/md.texi (decrement_and_branch_until_zero): Remove.
8188         (Looping patterns): Remove decrement_and_branch_until_zero.  Add
8189         detail for doloop_end.
8191 2018-07-12  Martin Sebor  <msebor@redhat.com>
8193         PR c/86453
8194         * attribs.c (decl_attributes): Reject conflicting attributes before
8195         calling attribute handlers.
8197 2018-07-12  Jan Hubicka  <hubicka@ucw.cz>
8199         * dumpfile.c (gcc::dump_manager::get_dump_file_name): Add PART
8200          parameter.
8201         (gcc::dump_manager::get_dump_file_name): likewise.
8202         (dump_begin): Likewise.
8203         * dumpfile.h (dump_begin): Update prototype.
8204         (gcc::dump_manager::get_dump_file_name,
8205         gcc::dump_manager::get_dump_file_name): Update prototype.
8207 2018-07-12  Richard Sandiford  <richard.sandiford@linaro.org>
8209         * internal-fn.h (vectorizable_internal_fn_p): New function.
8210         * tree-vect-slp.c (compatible_calls_p): Likewise.
8211         (vect_build_slp_tree_1): Remove nops argument.  Handle calls
8212         to internal functions.
8213         (vect_build_slp_tree_2): Update call to vect_build_slp_tree_1.
8215 2018-07-12  Richard Sandiford  <richard.sandiford@linaro.org>
8217         * fold-const.h (inverse_conditions_p): Declare.
8218         * fold-const.c (inverse_conditions_p): New function.
8219         * match.pd: Use inverse_conditions_p.  Add folds of view_converts
8220         that test the inverse condition of a conditional internal function.
8221         * internal-fn.h (vectorized_internal_fn_supported_p): Declare.
8222         * internal-fn.c (internal_fn_mask_index): Handle conditional
8223         internal functions.
8224         (vectorized_internal_fn_supported_p): New function.
8225         * tree-if-conv.c: Include internal-fn.h and fold-const.h.
8226         (any_pred_load_store): Replace with...
8227         (need_to_predicate): ...this new variable.
8228         (redundant_ssa_names): New variable.
8229         (ifcvt_can_use_mask_load_store): Move initial checks to...
8230         (ifcvt_can_predicate): ...this new function.  Handle tree codes
8231         for which a conditional internal function exists.
8232         (if_convertible_gimple_assign_stmt_p): Use ifcvt_can_predicate
8233         instead of ifcvt_can_use_mask_load_store.  Update after variable
8234         name change.
8235         (predicate_load_or_store): New function, split out from
8236         predicate_mem_writes.
8237         (check_redundant_cond_expr): New function.
8238         (value_available_p): Likewise.
8239         (predicate_rhs_code): Likewise.
8240         (predicate_mem_writes): Rename to...
8241         (predicate_statements): ...this.  Use predicate_load_or_store
8242         and predicate_rhs_code.
8243         (combine_blocks, tree_if_conversion): Update after above name changes.
8244         (ifcvt_local_dce): Handle redundant_ssa_names.
8245         * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): Handle
8246         general conditional functions.
8247         * tree-vect-stmts.c (vectorizable_call): Likewise.
8249 2018-07-12  Richard Sandiford  <richard.sandiford@linaro.org>
8250             Alan Hayward  <alan.hayward@arm.com>
8251             David Sherwood  <david.sherwood@arm.com>
8253         * internal-fn.h (can_interpret_as_conditional_op_p): Declare.
8254         * internal-fn.c (can_interpret_as_conditional_op_p): New function.
8255         * tree-ssa-math-opts.c (convert_mult_to_fma_1): Handle conditional
8256         plus and minus and convert them into IFN_COND_FMA-based sequences.
8257         (convert_mult_to_fma): Handle conditional plus and minus.
8259 2018-07-12  Richard Sandiford  <richard.sandiford@linaro.org>
8261         * doc/md.texi (cond_fma, cond_fms, cond_fnma, cond_fnms): Document.
8262         * optabs.def (cond_fma_optab, cond_fms_optab, cond_fnma_optab)
8263         (cond_fnms_optab): New optabs.
8264         * internal-fn.def (COND_FMA, COND_FMS, COND_FNMA, COND_FNMS): New
8265         internal functions.
8266         (FMA): Use DEF_INTERNAL_FLT_FN rather than DEF_INTERNAL_FLT_FLOATN_FN.
8267         * internal-fn.h (get_conditional_internal_fn): Declare.
8268         (get_unconditional_internal_fn): Likewise.
8269         * internal-fn.c (cond_ternary_direct): New macro.
8270         (expand_cond_ternary_optab_fn): Likewise.
8271         (direct_cond_ternary_optab_supported_p): Likewise.
8272         (FOR_EACH_COND_FN_PAIR): Likewise.
8273         (get_conditional_internal_fn): New function.
8274         (get_unconditional_internal_fn): Likewise.
8275         * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 5.
8276         (gimple_match_op::gimple_match_op): Add a new overload for 5
8277         operands.
8278         (gimple_match_op::set_op): Likewise.
8279         (gimple_resimplify5): Declare.
8280         * genmatch.c (decision_tree::gen): Generate simplifications for
8281         5 operands.
8282         * gimple-match-head.c (gimple_simplify): Define an overload for
8283         5 operands.  Handle calls with 5 arguments in the top-level overload.
8284         (convert_conditional_op): Handle conversions from unconditional
8285         internal functions to conditional ones.
8286         (gimple_resimplify5): New function.
8287         (build_call_internal): Pass a fifth operand.
8288         (maybe_push_res_to_seq): Likewise.
8289         (try_conditional_simplification): Try converting conditional
8290         internal functions to unconditional internal functions.
8291         Handle 3-operand unconditional forms.
8292         * match.pd (UNCOND_TERNARY, COND_TERNARY): Operator lists.
8293         Define ternary equivalents of the current rules for binary conditional
8294         internal functions.
8295         * config/aarch64/aarch64.c (aarch64_preferred_else_value): Handle
8296         ternary operations.
8297         * config/aarch64/iterators.md (UNSPEC_COND_FMLA, UNSPEC_COND_FMLS)
8298         (UNSPEC_COND_FNMLA, UNSPEC_COND_FNMLS): New unspecs.
8299         (optab): Handle them.
8300         (SVE_COND_FP_TERNARY): New int iterator.
8301         (sve_fmla_op, sve_fmad_op): New int attributes.
8302         * config/aarch64/aarch64-sve.md (cond_<optab><mode>)
8303         (*cond_<optab><mode>_2, *cond_<optab><mode_4)
8304         (*cond_<optab><mode>_any): New SVE_COND_FP_TERNARY patterns.
8306 2018-07-12  Richard Sandiford  <richard.sandiford@linaro.org>
8308         * target.def (preferred_else_value): New target hook.
8309         * doc/tm.texi.in (TARGET_PREFERRED_ELSE_VALUE): New hook.
8310         * doc/tm.texi: Regenerate.
8311         * targhooks.h (default_preferred_else_value): Declare.
8312         * targhooks.c (default_preferred_else_value): New function.
8313         * internal-fn.h (conditional_internal_fn_code): Declare.
8314         * internal-fn.c (FOR_EACH_CODE_MAPPING): New macro.
8315         (get_conditional_internal_fn): Use it.
8316         (conditional_internal_fn_code): New function.
8317         * gimple-match.h (gimple_match_cond): New struct.
8318         (gimple_match_op): Add a cond member function.
8319         (gimple_match_op::gimple_match_op): Update all forms to take a
8320         gimple_match_cond.
8321         * genmatch.c (expr::gen_transform): Use the same condition as res_op
8322         for the suboperation, but don't specify a particular else_value.
8323         * tree-ssa-sccvn.c (vn_nary_simplify, vn_reference_lookup_3)
8324         (visit_nary_op, visit_reference_op_load): Pass
8325         gimple_match_cond::UNCOND to the gimple_match_op constructor.
8326         * gimple-match-head.c: Include tree-eh.h
8327         (convert_conditional_op): New function.
8328         (maybe_resimplify_conditional_op): Likewise.
8329         (gimple_resimplify1): Call maybe_resimplify_conditional_op.
8330         (gimple_resimplify2): Likewise.
8331         (gimple_resimplify3): Likewise.
8332         (gimple_resimplify4): Likewise.
8333         (maybe_push_res_to_seq): Return null for conditional operations.
8334         (try_conditional_simplification): New function.
8335         (gimple_simplify): Call it.  Pass conditions to the gimple_match_op
8336         constructor.
8337         * match.pd: Fold VEC_COND_EXPRs of an IFN_COND_* call to a new
8338         IFN_COND_* call.
8339         * config/aarch64/aarch64.c (aarch64_preferred_else_value): New
8340         function.
8341         (TARGET_PREFERRED_ELSE_VALUE): Redefine.
8343 2018-07-12  Jan Hubicka  <hubicka@ucw.cz>
8345         * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
8346         DECL_FCONTEXT
8347         (hash_tree): Do not hash DECL_FCONTEXT
8348         * tree-streamer-in.c (lto_input_ts_field_decl_tree_pointers):
8349         Do not stream DECL_FCONTEXT.
8350         * tree-streamer-out.c (write_ts_field_decl_tree_pointers): Likewise.
8351         * tree.c (free_lang_data_in_decl): Free DECL_FCONTEXT.
8353 2018-07-12  Richard Biener  <rguenther@suse.de>
8355         PR debug/86462
8356         * dwarf2out.c (gen_block_die): Only output blocks when they have
8357         at least one !DECL_IGNORED_P variable.
8359 2018-07-12  Richard Biener  <rguenther@suse.de>
8361         PR target/84829
8362         * config/gnu-user.h (GNU_USER_TARGET_NO_PTHREADS_LIB_SPEC):
8363         Remove -mieee-fp handling.
8365 2018-07-12  Richard Biener  <rguenther@suse.de>
8367         * tree-ssa-sccvn.c (vn_lookup_simplify_result): Remove bogus
8368         left-over from last patch.
8370 2018-07-12  Jakub Jelinek  <jakub@redhat.com>
8372         PR tree-optimization/86492
8373         * gimple-ssa-store-merging.c
8374         (imm_store_chain_info::coalesce_immediate_stores): Call
8375         check_no_overlap even for the merge_overlapping case.  Formatting fix.
8377 2018-07-12  Richard Biener  <rguenther@suse.de>
8379         PR middle-end/86479
8380         * fold-const.c (fold_binary_op_with_conditional_arg): Do not
8381         move possibly trapping operations into the conditional.
8383 2018-07-12  Richard Biener  <rguenther@suse.de>
8385         * tree-ssa-sccvn.c (mprts_hook_cnt): Remove.
8386         (vn_lookup_simplify_result): Remove recursion limit applied
8387         here.
8388         (vn_nary_build_or_lookup_1): Adjust.
8389         (try_to_simplify): Likewise.
8390         * gimple-match-head.c (gimple_resimplify1): Instead apply one
8391         here.
8392         (gimple_resimplify2): Likewise.
8393         (gimple_resimplify3): Likewise.
8394         (gimple_resimplify4): Likewise.
8396 2018-07-11  Jakub Jelinek  <jakub@redhat.com>
8398         * config/i386/avx512bitalgintrin.h (_mm512_mask_bitshuffle_epi64_mask):
8399         Use __mmask64 type instead of __mmask8 for __M argument.
8400         * config/i386/avx512fintrin.h (_mm512_mask_xor_epi64,
8401         _mm512_maskz_xor_epi64): Use __mmask8 type instead of __mmask16 for
8402         __U argument.
8403         (_mm512_mask_cmpneq_epi64_mask): Use __mmask8 type instead of
8404         __mmask16 for __M argument.
8405         (_mm512_maskz_insertf32x4, _mm512_maskz_inserti32x4,
8406         _mm512_mask_insertf32x4, _mm512_mask_inserti32x4): Cast last argument
8407         to __mmask16 instead of __mmask8.
8408         * config/i386/avx512vlintrin.h (_mm_mask_add_ps, _mm_maskz_add_ps,
8409         _mm256_mask_add_ps, _mm256_maskz_add_ps, _mm_mask_sub_ps,
8410         _mm_maskz_sub_ps, _mm256_mask_sub_ps, _mm256_maskz_sub_ps,
8411         _mm256_maskz_cvtepi32_ps, _mm_maskz_cvtepi32_ps): Use __mmask8 type
8412         instead of __mmask16 for __U argument.
8413         * config/i386/avx512vlbwintrin.h (_mm_mask_cmp_epi8_mask): Use
8414         __mmask16 instead of __mmask8 for __U argument.
8415         (_mm256_mask_cmp_epi8_mask): Use __mmask32 instead of __mmask16 for
8416         __U argument.
8417         (_mm256_cmp_epi8_mask): Use __mmask32 return type instead of
8418         __mmask16.
8419         (_mm_mask_cmp_epu8_mask): Use __mmask16 instead of __mmask8 for __U
8420         argument.
8421         (_mm256_mask_cmp_epu8_mask): Use __mmask32 instead of __mmask16 for
8422         __U argument.
8423         (_mm256_cmp_epu8_mask): Use __mmask32 return type instead of
8424         __mmask16.
8425         (_mm_mask_cmp_epi16_mask): Cast last argument to __mmask8 instead
8426         of __mmask16.
8427         (_mm256_mask_cvtepi8_epi16): Use __mmask16 instead of __mmask32 for
8428         __U argument.
8429         (_mm_mask_cvtepi8_epi16): Use __mmask8 instead of __mmask32 for
8430         __U argument.
8431         (_mm256_mask_cvtepu8_epi16): Use __mmask16 instead of __mmask32 for
8432         __U argument.
8433         (_mm_mask_cvtepu8_epi16): Use __mmask8 instead of __mmask32 for
8434         __U argument.
8435         (_mm256_mask_cmpneq_epu8_mask, _mm256_mask_cmplt_epu8_mask,
8436         _mm256_mask_cmpge_epu8_mask, _mm256_mask_cmple_epu8_mask): Change
8437         return type as well as __M argument type and all casts from __mmask8
8438         to __mmask32.
8439         (_mm256_mask_cmpneq_epu16_mask, _mm256_mask_cmplt_epu16_mask,
8440         _mm256_mask_cmpge_epu16_mask, _mm256_mask_cmple_epu16_mask): Change
8441         return type as well as __M argument type and all casts from __mmask8
8442         to __mmask16.
8443         (_mm256_mask_cmpneq_epi8_mask, _mm256_mask_cmplt_epi8_mask,
8444         _mm256_mask_cmpge_epi8_mask, _mm256_mask_cmple_epi8_mask): Change
8445         return type as well as __M argument type and all casts from __mmask8
8446         to __mmask32.
8447         (_mm256_mask_cmpneq_epi16_mask, _mm256_mask_cmplt_epi16_mask,
8448         _mm256_mask_cmpge_epi16_mask, _mm256_mask_cmple_epi16_mask): Change
8449         return type as well as __M argument type and all casts from __mmask8
8450         to __mmask16.
8451         * config/i386/avx512vbmi2vlintrin.h (_mm_mask_shrdi_epi32,
8452         _mm_mask_shldi_epi32): Cast last argument to __mmask8 instead of
8453         __mmask16.
8455 2018-07-11  Grazvydas Ignotas  <notasas@gmail.com>
8457         * config/i386/avx512bwintrin.h: (_mm512_mask_cmp_epi8_mask,
8458         _mm512_mask_cmp_epu8_mask): Use __mmask64 type instead of __mmask32
8459         for __U argument.
8461 2018-07-11  Paul Koning  <ni1d@arrl.net>
8463         * doc/md.texi (define_subst): Document how multiple occurrences of
8464         the same argument in the replacement pattern are handled.
8466 2018-07-11  Paul Koning  <ni1d@arrl.net>
8468         * doc/extend.texi (Common Variable Attributes): Move "mode" into
8469         alphabetical order.
8470         (Common Type Attributes): Add "mode" attribute.
8472 2018-07-11  Jan Hubicka  <hubicka@ucw.cz>
8474         * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not
8475         stream DECL_ORIGINAL_TYPE.
8476         (DFS::DFS_write_tree_body): Drop hack handling local external decls.
8477         (hash_tree): Do not walk DECL_ORIGINAL_TYPE.
8478         * tree-streamer-in.c (lto_input_ts_decl_non_common_tree_pointers):
8479         Do not walk original type.
8480         * tree-streamer-out.c (streamer_write_chain): Drop hack handling
8481         external decls.
8482         (write_ts_decl_non_common_tree_pointers): Do not stream
8483         DECL_ORIGINAL_TYPE
8484         * tree.c (free_lang_data_in_decl): Clear DECL_ORIGINAL_TYPE.
8485         (find_decls_types_r): Do not walk DEC_ORIGINAL_TYPE.
8487 2018-07-11  Aldy Hernandez  <aldyh@redhat.com>
8489         * tree-ssa-threadupdate.c (thread_through_all_blocks): Do not jump
8490         thread twice from the same starting edge.
8492 2018-07-11  Aldy Hernandez  <aldyh@redhat.com>
8494         * vr-values.c (gimple_stmt_nonzero_p): Abstract common code to...
8495         * gimple.c (gimple_call_nonnull_result_p): ...here...
8496         (gimple_call_nonnull_arg): ...and here.
8497         * gimple.h (gimple_call_nonnull_result_p): New.
8498         (gimple_call_nonnull_arg): New.
8500 2018-07-11  Richard Earnshaw  <rearnsha@arm.com>
8502         * config/arm/arm-cpus.in: Move information from fpu field of each
8503         cpu definition to the isa field.
8504         * config/arm/parsecpu.awk (fpu): Delete match rule.
8505         (gen_comm_data): Don't add bits from the CPU's FPU entry.
8507 2018-07-11  Richard Biener  <rguenther@suse.de>
8509         PR debug/86457
8510         * dwarf2out.c (init_sections_and_labels): Use
8511         output_asm_line_debug_info consistently.
8512         (dwarf2out_early_finish): Likewise.
8513         (dwarf2out_finish): Remove DW_AT_stmt_list from early generated
8514         type units.
8516 2018-07-11  Richard Biener  <rguenther@suse.de>
8518         * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1):
8519         Rework father_bb setting in a way to avoid propagating constants
8520         multiple times on a loop body.
8522 2018-07-10  Mark Wielaard  <mark@klomp.org>
8524         PR debug/86459
8525         * dwarf2out.c (output_macinfo_op): Fix dwarf_FORM typo in gcc_assert.
8527 2018-07-10  Richard Biener  <rguenther@suse.de>
8529         * hash-map.h (hash_map::iterator::operator*): Return
8530         references to key and value.
8532 2018-07-10  Jakub Jelinek  <jakub@redhat.com>
8534         PR c++/86443
8535         * gimplify.c (find_combined_omp_for): Add DATA argument, in addition
8536         to finding the inner OMP_FOR/OMP_SIMD stmt find non-trivial wrappers,
8537         BLOCKs with BLOCK_VARs, OMP_PARALLEL in between, OMP_FOR in between.
8538         (gimplify_omp_for): For composite loops, move outer
8539         OMP_{DISTRIBUTE,TASKLOOP,FOR,PARALLEL} right around innermost
8540         OMP_FOR/OMP_SIMD if there are any non-trivial wrappers.  For class
8541         iterators add any needed clauses.  Allow OMP_FOR_ORIG_DECLS to contain
8542         TREE_LIST for both the original class iterator and the "last" helper
8543         var.  Gimplify OMP_FOR_PRE_BODY before the outermost composite
8544         loop, remember has_decl_expr from outer composite loops for the
8545         innermost OMP_SIMD in TREE_PRIVATE bit on OMP_FOR_INIT.
8547 2018-07-09  Martin Sebor  <msebor@redhat.com>
8549         PR middle-end/77357
8550         PR middle-end/86428
8551         * builtins.c (c_strlen): Avoid out-of-bounds warnings when
8552         accessing implicitly initialized array elements.
8553         * expr.c (string_constant): Handle string initializers of
8554         character arrays within aggregates.
8555         * gimple-fold.c (fold_array_ctor_reference): Add argument.
8556         Store element offset.  As a special case, handle zero size.
8557         (fold_nonarray_ctor_reference): Same.
8558         (fold_ctor_reference): Add argument.  Store subobject offset.
8559         * gimple-fold.h (fold_ctor_reference): Add argument.
8561 2018-07-09  Paul Koning  <ni1d@arrl.net>
8563         * config/pdp11/pdp11.c (pdp11_addr_cost): New function.
8564         (pdp11_insn_cost): New function.
8565         (pdp11_md_asm_adjust): New function.
8566         (TARGET_INVALID_WITHIN_DOLOOP): Define.
8567         (pdp11_rtx_costs): Update to match machine better.
8568         (output_addr_const_pdp11): Correct format mismatch warnings.
8569         * config/pdp11/pdp11.h (SLOW_BYTE_ACCESS): Correct definition.
8570         * config/pdp11/pdp11.md: General change to add base_cost and/or
8571         length attributes for use by new pdp11_insn_cost function.
8572         (MIN_BRANCH): Correct definition.
8573         (MIN_SOB): Ditto.
8574         (doloop_end): Use standard pattern name for looping pattern.
8575         (doloop_end_nocc): New.
8576         (movsf): Add another constraint alternative.
8577         (zero_extendqihi2): Add constraint alternatives for not in place
8578         extend.
8579         (zero_extendhisi2): Remove.
8580         (shift patterns): Add CC handling variants.
8581         (bswaphi2): New.
8582         (bswapsi2): New.
8583         (rothi3): New.
8584         (define_peephole2): New peephole to recognize mov that sets CC for
8585         subsequent test.
8587 2018-07-09  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
8589         * config/sparc/sparc.c (sparc_fold_builtin) <SPARC_BUILTIN_PDIST,
8590         SPARC_BUILTIN_PDISTN>: Adapt for signature change of wi::neg,
8591         wi::add.
8593 2018-07-09  Jakub Jelinek  <jakub@redhat.com>
8595         PR c/86420
8596         * real.c (real_nextafter): Return true if result is denormal.
8598 2018-07-09  Martin Liska  <mliska@suse.cz>
8600         * common.opt: Add back wrongly removed attribute.
8602 2018-07-09  Richard Biener  <rguenther@suse.de>
8604         PR debug/86413
8605         * dwarf2out.c (gen_block_die): For an early generated DIE
8606         always output high/low PC attributes.
8608 2018-07-09  Tom de Vries  <tdevries@suse.de>
8610         * cfgexpand.c (expand_debug_source_expr): Handle VAR_DECL.
8611         * tree-inline.c (remap_ssa_name): Handle default def ssa_name mapping
8612         onto VAR_DECL with abstract origin.
8614 2018-07-07  Jim Wilson  <jimw@sifive.com>
8616         * config/riscv/riscv.c (TARGET_CUSTOM_FUNCTION_DESCRIPTORS): New.
8618 2018-07-07  Tom de Vries  <tdevries@suse.de>
8620         * tree-dfa.c (dump_enumerated_decls): Handle cfun->cfg == NULL.
8622 2018-07-07  Aldy Hernandez  <aldyh@redhat.com>
8624         * tree-vrp.c (vrp_int_const_binop): Change overflow type to
8625         overflow_type.
8626         (combine_bound): Use wide-int overflow calculation instead of
8627         rolling our own.
8628         * calls.c (maybe_warn_alloc_args_overflow): Change overflow type to
8629         overflow_type.
8630         * fold-const.c (int_const_binop_2): Same.
8631         (extract_muldiv_1): Same.
8632         (fold_div_compare): Same.
8633         (fold_abs_const): Same.
8634         * match.pd: Same.
8635         * poly-int.h (add): Same.
8636         (sub): Same.
8637         (neg): Same.
8638         (mul): Same.
8639         * predict.c (predict_iv_comparison): Same.
8640         * profile-count.c (slow_safe_scale_64bit): Same.
8641         * simplify-rtx.c (simplify_const_binary_operation): Same.
8642         * tree-chrec.c (tree_fold_binomial): Same.
8643         * tree-data-ref.c (split_constant_offset_1): Same.
8644         * tree-if-conv.c (idx_within_array_bound): Same.
8645         * tree-scalar-evolution.c (iv_can_overflow_p): Same.
8646         * tree-ssa-phiopt.c (minmax_replacement): Same.
8647         * tree-vect-loop.c (is_nonwrapping_integer_induction): Same.
8648         * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Same.
8649         * vr-values.c (vr_values::adjust_range_with_scev): Same.
8650         * wide-int.cc (wi::add_large): Same.
8651         (wi::mul_internal): Same.
8652         (wi::sub_large): Same.
8653         (wi::divmod_internal): Same.
8654         * wide-int.h: Change overflow type to overflow_type for neg, add,
8655         mul, smul, umul, div_trunc, div_floor, div_ceil, div_round,
8656         mod_trunc, mod_ceil, mod_round, add_large, sub_large,
8657         mul_internal, divmod_internal.
8658         (overflow_type): New enum.
8659         (accumulate_overflow): New.
8661 2018-07-06  Kugan Vivekanandarajah  <kugan.vivekanandarajah@linaro.org>
8663         * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): New.
8664         (tree_ssa_phiopt_worker): Call cond_removal_in_popcount_pattern.
8666 2018-07-06  Kugan Vivekanandarajah  <kuganv@linaro.org>
8668         * tree-ssa-loop-niter.c (number_of_iterations_popcount): If popcount
8669         argument is checked for zero before entering loop, avoid checking again.
8671 2018-07-06  Kugan Vivekanandarajah  <kuganv@linaro.org>
8673         * gimplify.h (generic_expr_could_trap_p): Set as global function.
8674         * gimplify.h (generic_expr_could_trap_p): Likwise.
8675         * tree-scalar-evolution.c (expression_expensive_p): Handle COND_EXPR.
8677 2018-07-06  Jakub Jelinek  <jakub@redhat.com>
8679         PR tree-optimization/86401
8680         * fold-const.c (fold_binary_loc) <case BIT_AND_EXPR>: Move the
8681         ((A & N) + B) & M -> (A + B) & M etc. optimization into ...
8682         (fold_bit_and_mask): ... here.  New helper function for match.pd.
8683         * fold-const.h (fold_bit_and_mask): Declare.
8684         * match.pd (((A & N) + B) & M -> (A + B) & M): New optimization.
8686 2018-07-06  Peter Bergner  <bergner@linux.ibm.com>
8688         PR target/86324
8689         * target.def (translate_mode_attribute): New hook.
8690         * targhooks.h (default_translate_mode_attribute): Declare.
8691         * targhooks.c (default_translate_mode_attribute): New function.
8692         * doc/tm.texi.in (TARGET_TRANSLATE_MODE_ATTRIBUTE): New hook.
8693         * doc/tm.texi: Regenerate.
8694         * config/rs6000/rs6000.c (TARGET_TRANSLATE_MODE_ATTRIBUTE): Define.
8695         (rs6000_translate_mode_attribute): New function.
8697 2018-07-06  Paul Koning  <ni1d@arrl.net>
8699         * doc/md.texi (define_split): Document DONE and FAIL.
8700         (define_peephole2): Ditto.
8702 2018-07-05  Jeff Law  <law@redhat.com>
8704         PR tree-optimization/86010
8705         * tree-ssa-dse.c (compute_trims): More aggressively trim at
8706         both the head and tail of mem* and str* calls.
8708 2018-07-05  Jim Wilson  <jimw@sifive.com>
8710         * config.gcc (riscv*-*-*): When setting xlen, handle riscv-*.
8712 2018-07-05  Indu Bhagat  <indu.bhagat@oracle.com>
8714         * config/aarch64/aarch64-simd.md: correct flags text for
8715         MIN_EXPR replacement.
8717 2018-07-05  James Clarke  <jrtc27@jrtc27.com>
8719         * configure: Regenerated.
8721 2018-07-05  Carl Love  <cel@us.ibm.com>
8723         * config/rs6000/rs6000-c.c: Map ALTIVEC_BUILTIN_VEC_UNPACKH for
8724         float argument to VSX_BUILTIN_DOUBLEH_V4SF.
8725         Map ALTIVEC_BUILTIN_VEC_UNPACKL for float argument to
8726         VSX_BUILTIN_DOUBLEL_V4SF.
8728 2018-07-05  Martin Sebor  <msebor@redhat.com>
8730         PR c++/86400
8731         * tree-ssa-strlen.c (maybe_set_strlen_range): Use type size rather
8732         than its domain to compute its the upper bound of a char array.
8734 2018-07-05  Nathan Sidwell  <nathan@acm.org>
8736         Replace NO_IMPLICIT_EXTERN_C with SYSTEM_IMPLICIT_EXTERN_C.
8737         * doc/cpp.texi: Update comment.
8738         * doc/tm.texi: Rebuilt.
8739         * doc/tm.texi.in (NO_IMPLICIT_EXTERN_C): Replace with ...
8740         (SYSTEM_IMPLICIT_EXTERN_C): ... this, opposite sense.
8741         * doc/extend.texi (Backwards Compatibility): Clarify it is system
8742         headers affected by extern "C".
8743         * system.h: Poison NO_IMPLICIT_EXTERN_C.
8744         * config/alpha/alpha.h, config/arm/uclinux-elf.h,
8745         config/bfin/elf.h, config/cris/cris.h, config/darwin.h,
8746         config/dragonfly.h, config/freebsd.h, config/gnu-user.h,
8747         config/i386/cygming.h, config/i386/djgpp.h, config/i386/nto.h,
8748         config/ia64/hpux.h, config/lm32/lm32.h, config/lm32/uclinux-elf.h,
8749         config/lynx.h, config/mips/elf.h, config/mmix/mmix.h,
8750         config/netbsd.h, config/pa/pa-hpux.h, config/powerpcspe/sysv4.h,
8751         config/riscv/elf.h, config/rs6000/sysv4.h, config/rtems.h,
8752         config/s390/tpf.h, config/sh/newlib.h, config/sol2.h,
8753         config/sparc/openbsd64.h, config/sparc/sp-elf.h,
8754         config/sparc/sp64-elf.h, config/spu/spu.h,
8755         config/stormy16/stormy16.h, config/v850/v850.h,
8756         config/visium/visium.h, config/vx-common.h, config/xtensa/elf.h: Don't
8757         define NO_IMPLICIT_EXTERN_C.
8758         * config/rs6000/aix.h: Set SYSTEM_IMPLICIT_EXTERN_C.
8760 2018-07-05  Tamar Christina  <tamar.christina@arm.com>
8762         PR target/84711
8763         * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
8764         instead of GET_MODE_SIZE when comparing Units.
8766 2018-07-05  Tamar Christina  <tamar.christina@arm.com>
8768         PR target/84711
8769         * rtlanal.c (set_noop_p): Constrain on mode change,
8770         include hard-reg-set.h
8772 2018-07-05  Tamar Christina  <tamar.christina@arm.com>
8774         * config/aarch64/aarch64.c (aarch64_expand_movmem): Fix mode size.
8776 2018-07-05  Jakub Jelinek  <jakub@redhat.com>
8778         Revert
8779         2018-07-04  Maxim Ostapenko  <m.ostapenko@samsung.com>
8781         PR sanitizer/84250
8782         * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
8783         libasan.
8784         * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
8786 2018-07-04  Maxim Ostapenko  <m.ostapenko@samsung.com>
8788         PR sanitizer/84250
8789         * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
8790         libasan.
8791         * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
8793 2018-07-04  Eric Botcazou  <ebotcazou@adacore.com>
8795         PR middle-end/86380
8796         * expmed.c (choose_multiplier): Fix incorrect comparison with mask.
8798 2018-07-04  Aldy Hernandez  <aldyh@redhat.com>
8800         * tree-vrp.c (extract_range_from_binary_expr_1): Initialize
8801         neg_*_op* variables.
8803 2018-07-04  Martin Liska  <mliska@suse.cz>
8805         * tree-switch-conversion.c: Define
8806         max_ratio_for_speed and max_ratio_for_size constants.
8808 2018-07-04  Denys Vlasenko  <dvlasenk@redhat.com>
8809             Martin Liska  <mliska@suse.cz>
8811         PR middle-end/66240
8812         PR target/45996
8813         PR c/84100
8814         * common.opt: Rename align options with 'str_' prefix.
8815         * common/config/i386/i386-common.c (set_malign_value): New
8816         function.
8817         (ix86_handle_option): Use it to set -falign-* options/
8818         * config/aarch64/aarch64-protos.h (struct tune_params): Change
8819         type from int to string.
8820         * config/aarch64/aarch64.c: Update default values from int
8821         to string.
8822         * config/alpha/alpha.c (alpha_override_options_after_change):
8823         Likewise.
8824         * config/arm/arm.c (arm_override_options_after_change_1): Likewise.
8825         * config/i386/dragonfly.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
8826         max skip conditionally.
8827         * config/i386/freebsd.h (SUBALIGN_LOG): New.
8828         (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
8829         max skip conditionally.
8830         * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
8831         max skip conditionally.
8832         * config/i386/gnu-user.h (SUBALIGN_LOG): New.
8833         (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
8834         max skip conditionally.
8835         * config/i386/i386.c (struct ptt): Change type from int to
8836         string.
8837         (ix86_default_align): Set default values.
8838         * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Print
8839         max skip conditionally.
8840         * config/i386/iamcu.h (SUBALIGN_LOG): New.
8841         (ASM_OUTPUT_MAX_SKIP_ALIGN):
8842         * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN):
8843         * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
8844         max skip conditionally.
8845         * config/i386/openbsdelf.h (SUBALIGN_LOG): New.
8846         (ASM_OUTPUT_MAX_SKIP_ALIGN) Print max skip conditionally.:
8847         * config/i386/x86-64.h (SUBALIGN_LOG): New.
8848         (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
8849         max skip conditionally.
8850         (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
8851         * config/ia64/ia64.c (ia64_option_override): Set default values
8852         for alignment options.
8853         * config/m68k/m68k.c: Handle new str_align_* options.
8854         * config/mips/mips.c (mips_set_compression_mode): Change
8855         type of constants.
8856         (mips_option_override): Set default values for options.
8857         * config/powerpcspe/powerpcspe.c (rs6000_option_override_internal):
8858         Likewise.
8859         * config/rs6000/rs6000.c (rs6000_option_override_internal):
8860         Likewise.
8861         * config/rx/rx.c (rx_option_override): Likewise.
8862         * config/rx/rx.h (JUMP_ALIGN): Use align_jumps_log.
8863         (LABEL_ALIGN): Use align_labels_log.
8864         (LOOP_ALIGN): Use align_loops_align.
8865         * config/s390/s390.c (s390_asm_output_function_label): Use new
8866         macros.
8867         * config/sh/sh.c (sh_override_options_after_change):
8868         Change type of constants.
8869         * config/spu/spu.c (spu_sched_init): Likewise.
8870         * config/sparc/sparc.c (sparc_option_override): Set default
8871         values for options.
8872         * config/visium/visium.c (visium_option_override): Likewise.
8873         * config/visium/visium.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Do not
8874         emit p2align format with last argument if it's not needed.
8875         * doc/invoke.texi: Document extended format of -falign-*.
8876         * final.c: Use align_labels alignment.
8877         * flags.h (struct target_flag_state): Change type to use
8878         align_flags.
8879         (struct align_flags_tuple): New.
8880         (struct align_flags): Likewise.
8881         (align_loops_log): Redefine macro to use new types.
8882         (align_loops_max_skip): Redefine macro to use new types.
8883         (align_jumps_log): Redefine macro to use new types.
8884         (align_jumps_max_skip): Redefine macro to use new types.
8885         (align_labels_log): Redefine macro to use new types.
8886         (align_labels_max_skip): Redefine macro to use new types.
8887         (align_functions_log): Redefine macro to use new types.
8888         (align_loops): Redefine macro to use new types.
8889         (align_jumps): Redefine macro to use new types.
8890         (align_labels): Redefine macro to use new types.
8891         (align_functions): Redefine macro to use new types.
8892         (align_functions_max_skip): Redefine macro to use new types.
8893         (align_loops_value): New macro.
8894         (align_jumps_value): New macro.
8895         (align_labels_value): New macro.
8896         (align_functions_value): New macro.
8897         * function.c (invoke_set_current_function_hook): Propagate
8898         alignment values from flags to global variables default in
8899         topleev.h.
8900         * ipa-icf.c (sem_function::equals_wpa): Use
8901         cl_optimization_option_eq instead of memcmp.
8902         * lto-streamer.h (cl_optimization_stream_out): Support streaming
8903         of string types.
8904         (cl_optimization_stream_in): Likewise.
8905         * optc-save-gen.awk: Support strings in cl_optimization.
8906         * opth-gen.awk: Likewise.
8907         * opts.c (finish_options): Remove error checking of invalid
8908         value ranges.
8909         (MAX_CODE_ALIGN): Remove.
8910         (MAX_CODE_ALIGN_VALUE): Likewise.
8911         (parse_and_check_align_values): New function.
8912         (check_alignment_argument): Likewise.
8913         (common_handle_option): Use check_alignment_argument.
8914         * opts.h (parse_and_check_align_values): Declare.
8915         * toplev.c (init_alignments): Remove.
8916         (read_log_maxskip): New.
8917         (parse_N_M): Likewise.
8918         (parse_alignment_opts): Likewise.
8919         (backend_init_target): Remove usage of init_alignments.
8920         * toplev.h (parse_alignment_opts): Declare.
8921         * tree-streamer-in.c (streamer_read_tree_bitfields): Add new
8922         argument.
8923         * tree-streamer-out.c (streamer_write_tree_bitfields): Likewise.
8924         * tree.c (cl_option_hasher::equal): New.
8925         * varasm.c: Use new global macros.
8927 2018-07-04  Denys Vlasenko  <dvlasenk@redhat.com>
8929         * config/i386/dragonfly.h: (ASM_OUTPUT_MAX_SKIP_ALIGN):
8930         Use a simpler align directive also if MAXSKIP = ALIGN-1.
8931         * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
8932         * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
8933         * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
8934         * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
8935         * config/i386/freebsd.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Remove "If N
8936         is large, do at least 8 byte alignment" code. Add SUBALIGN_LOG
8937         define. Use a simpler align directive also if MAXSKIP = ALIGN-1.
8938         * config/i386/gnu-user.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
8939         * config/i386/iamcu.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
8940         * config/i386/openbsdelf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
8941         * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
8943 2018-07-04  Martin Liska  <mliska@suse.cz>
8944             Jonathan Wakely  <jwakely@redhat.com>
8946         * coverage.c: Use correct type.
8947         * doc/invoke.texi: Language correction.
8949 2018-07-03  H.J. Lu  <hongjiu.lu@intel.com>
8951         PR target/85620
8952         * config/i386/i386.c (rest_of_insert_endbranch): Also generate
8953         ENDBRANCH for non-tail call which may return via indirect branch.
8954         * doc/extend.texi: Document indirect_return attribute.
8956 2018-07-03  Martin Sebor  <msebor@redhat.com>
8958         PR tree-optimization/86274
8959         * gimple-ssa-sprintf.c (fmtresult::type_max_digits): Verify
8960         precondition.
8961         (format_floating): Correct handling of infinities and NaNs.
8963 2018-07-03  Martin Sebor  <msebor@redhat.com>
8965         * print-tree.c (print_real_cst): New function.
8966         (print_node_brief): Call it.
8967         (print_node): Ditto.
8969 2018-07-03  Jeff Law  <law@redhat.com>
8971         * config/h8300/h8300.md (logical<mode>3_sn, logical<mode>3): Merge
8972         into a single pattern.
8974         * config/h8300/h8300.md (ors code_iterator): New.
8975         (bsetqi_msx, bnotqi_msx patterns and splitters): Consolidate into
8976         a single pattern and single splitter.
8977         (bsethi_msx, bnothi_msx patterns): Consolidate into a single pattern.
8978         (iorqi3_1, xorqi3_1): Likewise.
8979         (iorqi3, xorqi3 expanders): Similarly.
8981         * config/h8300/h8300.md (movmd_internal_normal): Consolidated with
8982         (movmd_internal) into a single pattern using the P mode iterator.
8983         (movmd splitters): Similarly.
8984         (stpcpy_internal_normal, stpcpy_internal): Similarly for thes patterns.
8985         (movsd splitters): Similarly.
8987         * config/h8300/h8300.c (h8300_insn_length_from_table): Consolidate
8988         ADDB, ADDW and ADDL into a single ADD attribute which selects the
8989         right table based on the size of the operand.
8990         * config/h8300/h8300.md (length_table): Corresponding changes. All
8991         references to "addb", "addw" and "addl" changed to "add".
8992         (btst patterns): Merge two variants into a single pattern.
8993         (tstqi, tsthi): Likewise.
8994         (addhi3_incdec, addsi3_incdec): Likewise.
8995         (subhi3_h8300hs, subsi3_h8300hs): Likewise.
8996         (mulhi3, mulsi3): Likewise.
8997         (udivhi3, udivsi3): Likewise.
8998         (divhi3, divsi3): Likewise.
8999         (andorqi3, andorhi3, andorsi3): Likewise.
9001 2018-07-03  Uros Bizjak  <ubizjak@gmail.com>
9003         PR target/85694
9004         * config/i386/sse.md (uavg<mode>3_ceil): New expander.
9005         (<sse2_avx2>_uavg<mode>3<mask_name>): Simplify expander.
9007 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
9009         PR tree-optimization/85694
9010         * config/aarch64/iterators.md (HADD, RHADD): New int iterators.
9011         (u): Handle UNSPEC_SHADD, UNSPEC_UHADD, UNSPEC_SRHADD and
9012         UNSPEC_URHADD.
9013         * config/aarch64/aarch64-simd.md (<u>avg<mode>3_floor)
9014         (<u>avg<mode>3_ceil): New patterns.
9016 2018-07-03  David Malcolm  <dmalcolm@redhat.com>
9018         * gcc.dg/vect/slp-perm-1.c: Remove "note: " prefix from
9019         scan-tree-dump directive.
9020         * gcc.dg/vect/slp-perm-2.c: Likewise.
9021         * gcc.dg/vect/slp-perm-3.c: Likewise.
9022         * gcc.dg/vect/slp-perm-5.c: Likewise.
9023         * gcc.dg/vect/slp-perm-6.c: Likewise.
9024         * gcc.dg/vect/slp-perm-7.c: Likewise.
9025         * gcc.dg/vect/slp-perm-8.c: Likewise.
9027 2018-07-03  Marek Polacek  <polacek@redhat.com>
9029         PR middle-end/86202
9030         * gimple-fold.c (size_must_be_zero_p): Check the type of the size.
9032 2018-07-03  Richard Biener  <rguenther@suse.de>
9034         PR ipa/86389
9035         * tree-ssa-structalias.c (find_func_clobbers): Properly
9036         handle indirect calls.
9038 2018-07-03  Jeff Law  <law@redhat.com>
9040         * config/h8300/h8300.md (HSI, QHSI, QHSIF): New mode iterators.
9041         (shifts): New code iterator.
9042         (movqi, movhi, movsi, movsf expanders): Consolidate into a single
9043         expander.  Fix HImode handling on H8/SX.
9044         (addqi3, addhi3, addsi3 expanders): Consolidate into a single expander.
9045         (subqi3, subhi3, subsi3 expanders): Likewise.
9046         (andqi3, andhi3, andsi3 expanders): Likewise.
9047         (iorqi3, iorhi3, iorsi3 expanders): Likewise.
9048         (xorqi3, xorhi3, xorsi3 expanders): Likewise.
9049         (negqi2, neghi2, negsi2, negsf2 expanders): Likewise.
9050         (one_cmplqi2, one_cmplhi2, one_cmplsi2): Likewise.
9051         (zero_extendqihi2, zero_extendqisi2): Likewise.
9052         (extendqihi2, extendqisi2): Likewise.
9053         (rotlqi3, rotlhi3, rotlsi3): Likewise.
9054         (neghi2_h8300, negsi2_h8300): Likewise for these patterns.
9055         (rotlqi3_1, rotlhi3_1): Likewise.
9056         (logicalhi3_sn, logicalsi3_sn): Likewise.
9057         (logicalhi3, logicalsi3): Likewise.
9059 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
9061         * tree-vect-patterns.c (vect_recog_rotate_pattern)
9062         (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
9063         (vect_recog_mixed_size_cond_pattern, adjust_bool_pattern_cast)
9064         (adjust_bool_pattern, vect_recog_bool_pattern): Pass the vector
9065         type to append_pattern_def_seq instead of creating a stmt_vec_info
9066         directly.
9067         (build_mask_conversion): Likewise.  Remove vinfo argument.
9068         (vect_add_conversion_to_patterm): Likewise, renaming to...
9069         (vect_add_conversion_to_pattern): ...this.
9070         (vect_recog_mask_conversion_pattern): Update call to
9071         build_mask_conversion.  Pass the vector type to
9072         append_pattern_def_seq here too.
9073         (vect_recog_gather_scatter_pattern): Update call to
9074         vect_add_conversion_to_pattern.
9076 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
9078         * tree-vect-patterns.c (new_pattern_def_seq): Delete.
9079         (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
9080         (vect_recog_widen_op_pattern, vect_recog_over_widening_pattern)
9081         (vect_recog_rotate_pattern, vect_synth_mult_by_constant): Don't set
9082         STMT_VINFO_PATTERN_DEF_SEQ to null here.
9083         (vect_recog_pow_pattern, vect_recog_vector_vector_shift_pattern)
9084         (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern): Use
9085         append_pattern_def_seq instead of new_pattern_def_seq.
9086         (vect_recog_divmod_pattern): Do both of the above.
9087         (vect_pattern_recog_1): Assert that STMT_VINO_PATTERN_DEF_SEQ
9088         is null.
9090 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
9092         * tree-vect-patterns.c (vect_recog_dot_prod_pattern):
9093         (vect_recog_sad_pattern, vect_recog_widen_op_pattern)
9094         (vect_recog_widen_mult_pattern, vect_recog_pow_pattern):
9095         (vect_recog_widen_sum_pattern, vect_recog_over_widening_pattern)
9096         (vect_recog_average_pattern, vect_recog_cast_forwprop_pattern)
9097         (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
9098         (vect_recog_vector_vector_shift_pattern, vect_synth_mult_by_constant)
9099         (vect_recog_mult_pattern, vect_recog_divmod_pattern)
9100         (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
9101         (vect_recog_mask_conversion_pattern): Replace vec<gimple *>
9102         parameter with a single stmt_vec_info.
9103         (vect_recog_func_ptr): Likewise.
9104         (vect_recog_gather_scatter_pattern): Likewise, folding in...
9105         (vect_try_gather_scatter_pattern): ...this.
9106         (vect_pattern_recog_1): Remove stmts_to_replace and just pass
9107         the stmt_vec_info of the statement to be matched.  Don't clear
9108         STMT_VINFO_RELATED_STMT.
9109         (vect_pattern_recog): Update call accordingly.
9111 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
9113         PR tree-optimization/85694
9114         * doc/md.texi (avgM3_floor, uavgM3_floor, avgM3_ceil)
9115         (uavgM3_ceil): Document new optabs.
9116         * doc/sourcebuild.texi (vect_avg_qi): Document new target selector.
9117         * internal-fn.def (IFN_AVG_FLOOR, IFN_AVG_CEIL): New internal
9118         functions.
9119         * optabs.def (savg_floor_optab, uavg_floor_optab, savg_ceil_optab)
9120         (savg_ceil_optab): New optabs.
9121         * tree-vect-patterns.c (vect_recog_average_pattern): New function.
9122         (vect_vect_recog_func_ptrs): Add it.
9123         * tree-vect-stmts.c (vectorizable_call): Get the type of the zero
9124         constant directly from the associated lhs.
9126 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
9128         * tree-vect-patterns.c (vect_split_statement): New function.
9129         (vect_convert_input): Use it to try to split an existing cast.
9131 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
9133         * poly-int.h (print_hex): New function.
9134         * dumpfile.h (dump_dec, dump_hex): Declare.
9135         * dumpfile.c (dump_dec, dump_hex): New poly_wide_int functions.
9136         * tree-vectorizer.h (_stmt_vec_info): Add min_output_precision,
9137         min_input_precision, operation_precision and operation_sign.
9138         * tree-vect-patterns.c (vect_get_range_info): New function.
9139         (vect_same_loop_or_bb_p, vect_single_imm_use)
9140         (vect_operation_fits_smaller_type): Delete.
9141         (vect_look_through_possible_promotion): Add an optional
9142         single_use_p parameter.
9143         (vect_recog_over_widening_pattern): Rewrite to use new
9144         stmt_vec_info infomration.  Handle one operation at a time.
9145         (vect_recog_cast_forwprop_pattern, vect_narrowable_type_p)
9146         (vect_truncatable_operation_p, vect_set_operation_type)
9147         (vect_set_min_input_precision): New functions.
9148         (vect_determine_min_output_precision_1): Likewise.
9149         (vect_determine_min_output_precision): Likewise.
9150         (vect_determine_precisions_from_range): Likewise.
9151         (vect_determine_precisions_from_users): Likewise.
9152         (vect_determine_stmt_precisions, vect_determine_precisions): Likewise.
9153         (vect_vect_recog_func_ptrs): Put over_widening first.
9154         Add cast_forwprop.
9155         (vect_pattern_recog): Call vect_determine_precisions.
9157 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
9159         * tree-vect-patterns.c (vect_mark_pattern_stmts): Remove pattern
9160         statements that have been replaced by further pattern statements.
9161         (vect_pattern_recog_1): Clear STMT_VINFO_PATTERN_DEF_SEQ on failure.
9163 2018-07-03  Richard Biener  <rguenther@suse.de>
9165         * tree-vect-stmts.c (vect_is_simple_use): Consolidate dumping,
9166         always set *dt.  Dump vectype in vectype overload.
9167         * dumpfile.h (dump_gimple_expr): New function.
9168         (dump_gimple_expr_loc): Likewise.
9169         * dumpfile.c (dump_gimple_expr): New function.
9170         (dump_gimple_expr_loc): Likewise.
9172 2018-07-02  Jeff Law  <law@redhat.com>
9174         * config/h8300/h8300.md (movqi_h8300, movqi_h8300hs): Consolidate
9175         the H8/300, H8/300H and H8/S variants into a single pattern.
9176         (movhi_h8300, movqi_h8300hs): Similarly.
9177         (pushqi_h8300hs, pushhi_h8300hs): Consolidate into a single pattern.
9178         (QHI mode iterator): New.
9180         * config/h8300/h8300.md: Remove trailing whitespace.
9182 2018-07-02  Jim Wilson  <jimw@sifive.com>
9184         * config/riscv/riscv.c (riscv_expand_epilogue): Use emit_jump_insn
9185         instead of emit_insn for interrupt returns.
9186         * config/riscv/riscv.md (riscv_met): Add (return) to rtl.
9187         (riscv_sret, riscv_uret): Likewise.
9189 2018-07-02  David Malcolm  <dmalcolm@redhat.com>
9191         * pretty-print.c (selftest::test_pp_format): Move save and restore
9192         of quotes to class auto_fix_quotes, and add an instance.
9193         * selftest.c: Include "intl.h".
9194         (selftest::auto_fix_quotes::auto_fix_quotes): New ctor.
9195         (selftest::auto_fix_quotes::~auto_fix_quotes): New dtor.
9196         * selftest.h (selftest::auto_fix_quotes): New class.
9198 2018-07-02  Richard Henderson  <richard.henderson@linaro.org>
9200         * config/aarch64/aarch64-protos.h, config/aarch64/aarch64.c
9201         (aarch64_sve_prepare_conditional_op): Remove.
9202         * config/aarch64/aarch64-sve.md (cond_<SVE_INT_BINARY><SVE_I>):
9203         Allow aarch64_simd_reg_or_zero as select operand; remove
9204         the aarch64_sve_prepare_conditional_op call.
9205         (cond_<SVE_INT_BINARY_SD><SVE_SDI>): Likewise.
9206         (cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
9207         (*cond_<SVE_INT_BINARY><SVE_I>_z): New pattern.
9208         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_z): New pattern.
9209         (*cond_<SVE_COND_FP_BINARY><SVE_F>_z): New pattern.
9210         (*cond_<SVE_INT_BINARY><SVE_I>_any): New pattern.
9211         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_any): New pattern.
9212         (*cond_<SVE_COND_FP_BINARY><SVE_F>_any): New pattern
9213         and a splitters to match all of the *_any patterns.
9214         * config/aarch64/predicates.md (aarch64_sve_any_binary_operator): New.
9216         * config/aarch64/iterators.md (SVE_INT_BINARY_REV): Remove.
9217         (SVE_COND_FP_BINARY_REV): Remove.
9218         (sve_int_op_rev, sve_fp_op_rev): New.
9219         * config/aarch64/aarch64-sve.md (*cond_<SVE_INT_BINARY><SVE_I>_0): New.
9220         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_0): New.
9221         (*cond_<SVE_COND_FP_BINARY><SVE_F>_0): New.
9222         (*cond_<SVE_INT_BINARY><SVE_I>_2): Rename, add movprfx alternative.
9223         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_2): Similarly.
9224         (*cond_<SVE_COND_FP_BINARY><SVE_F>_2): Similarly.
9225         (*cond_<SVE_INT_BINARY><SVE_I>_3): Similarly; use sve_int_op_rev.
9226         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_3): Similarly.
9227         (*cond_<SVE_COND_FP_BINARY><SVE_F>_3): Similarly; use sve_fp_op_rev.
9229         * config/aarch64/aarch64-sve.md (cond_<SVE_COND_FP_BINARY><SVE_F>):
9230         Remove match_dup 1 from the inner unspec.
9231         (*cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
9233         * config/aarch64/aarch64.md (movprfx): New attr.
9234         (length): Default movprfx to 8.
9235         * config/aarch64/aarch64-sve.md (*mul<SVE_I>3): Add movprfx alt.
9236         (*madd<SVE_I>, *msub<SVE_I): Likewise.
9237         (*<su>mul<SVE_I>3_highpart): Likewise.
9238         (*<SVE_INT_BINARY_SD><SVE_SDI>3): Likewise.
9239         (*v<ASHIFT><SVE_I>3): Likewise.
9240         (*<su><MAXMIN><SVE_I>3): Likewise.
9241         (*<su><MAXMIN><SVE_F>3): Likewise.
9242         (*fma<SVE_F>4, *fnma<SVE_F>4): Likewise.
9243         (*fms<SVE_F>4, *fnms<SVE_F>4): Likewise.
9244         (*div<SVE_F>4): Likewise.
9246 2018-07-02  Richard Sandiford  <richard.sandiford@arm.com>
9248         * tree-vect-patterns.c (vect_recog_widen_shift_pattern): Fix typo
9249         in dump string.
9251 2018-07-02  Richard Biener  <rguenther@suse.de>
9253         PR tree-optimization/86363
9254         * tree-ssa-sccvn.c (vn_reference_lookup_3): Check the
9255         memset argument refers to a non-variable address.
9257 2018-07-02  Aldy Hernandez  <aldyh@redhat.com>
9259         * tree-vrp.c (extract_range_from_binary_expr_1): Abstract a lot of the
9260         {PLUS,MINUS}_EXPR code to...
9261         (adjust_symbolic_bound): ...here,
9262         (combine_bound): ...here,
9263         (set_value_range_with_overflow): ...and here.
9265 2018-07-02  Aldy Hernandez  <aldyh@redhat.com>
9267         * tree-vrp.c (extract_range_from_unary_expr): Abstract ABS_EXPR
9268         code...
9269         (extract_range_from_abs_expr): ...here.
9271 2018-07-02  Eric Botcazou  <ebotcazou@adacore.com>
9273         * config/i386/i386.c (ix86_finalize_stack_frame_flags): Do not overrule
9274         -fno-omit-frame-pointer when not optimizing.
9276 2018-07-02  Martin Liska  <mliska@suse.cz>
9278         PR ipa/86279
9279         * ipa-pure-const.c (malloc_candidate_p): Revert usage of ::get.
9280         (propagate_nothrow): Likewise.
9282 2018-07-02  Martin Liska  <mliska@suse.cz>
9284         PR ipa/86323
9285         * ipa-inline.c (early_inliner): Revert wrongly added ::get call.
9287 2018-07-02  David Malcolm  <dmalcolm@redhat.com>
9289         * dumpfile.c (dump_generic_expr_loc): Undo removal of this
9290         function in r262149, changing "loc" param from source_location to
9291         const dump_location_t &.
9292         * dumpfile.h (dump_generic_expr_loc): Undo removal of this
9293         declaration, as above.
9295 2018-07-01  Paul Koning  <ni1d@arrl.net>
9297         * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
9298         -munit-asm, -mgnu-asm, -mdec-asm.
9299         * config/pdp11/pdp11-protos.h (pdp11_gen_int_label): New.
9300         (pdp11_output_labelref): New.
9301         (pdp11_output_def): New.
9302         (pdp11_output_addr_vec_elt): New.
9303         * config/pdp11/pdp11.c: Use tab between opcode and operands.  Use
9304         %# and %@ format codes.
9305         (pdp11_option_override): New.
9306         (TARGET_ASM_FILE_START_FILE_DIRECTIVE): Define.
9307         (pdp11_output_ident): New.
9308         (pdp11_asm_named_section): New.
9309         (pdp11_asm_init_sections): New.
9310         (pdp11_file_start): New.
9311         (pdp11_file_end): New.
9312         (output_ascii): Use .ascii/.asciz for -mdec-asm.
9313         (pdp11_asm_print_operand): Update %# and %$ for -mdec-asm.  Add
9314         %o, like %c but octal.
9315         (pdp11_option_override): New.
9316         * config/pdp11/pdp11.h (TEXT_SECTION_ASM_OP): Update for
9317         -mdec-asm.
9318         (DATA_SECTION_ASM_OP): Ditto.
9319         (READONLY_DATA_SECTION_ASM_OP): New.
9320         (IS_ASM_LOGICAL_LINE_SEPARATOR): New.
9321         (ASM_GENERATE_INTERNAL_LABEL): Use new function.
9322         (ASM_OUTPUT_LABELREF): Ditto.
9323         (ASM_OUTPUT_DEF): Ditto.
9324         (ASM_OUTPUT_EXTERNAL): New.
9325         (ASM_OUTPUT_SOURCE_FILENAME): New.
9326         (ASM_OUTPUT_ADDR_VEC_ELT): Use new function.
9327         (ASM_OUTPUT_SKIP): Update for -mdec-asm.
9328         * config/pdp11/pdp11.md: Use tab between opcode and operands.  Use
9329         %# and %@ format codes.
9330         * config/pdp11/pdp11.opt (mgnu-asm): New.
9331         (mdec-asm): Conflicts with -mgnu-asm and -munix-asm.
9332         (munix-asm): Conflicts with -mdec-asm and -mgnu-asm.
9333         * doc/invoke.txt (PDP-11 Options): Add -mgnu-asm.
9335 2018-07-01  Aldy Hernandez  <aldyh@redhat.com>
9337         * tree-ssa-threadupdate.c (mark_threaded_blocks): Avoid
9338         dereferencing path[] beyond its length.
9339         (debug_path): New.
9340         (debug_all_paths): New.
9341         (rewire_first_differing_edge): New.
9342         (adjust_paths_after_duplication): New.
9343         (duplicate_thread_path): Call adjust_paths_after_duplication.
9344         Add new argument.
9345         (thread_through_all_blocks): Add new argument to
9346         duplicate_thread_path.
9348 2018-06-30  Jim Wilson  <jimw@sifive.com>
9350         * config/riscv/predicates.md (p2m1_shift_operand): New.
9351         (high_mask_shift_operand): New.
9352         * config/riscv/riscv.md (lshrsi3_zero_extend_3+1): New combiner
9353         pattern using p2m1_shift_operand.
9354         (lshsi3_zero_extend_3+2): New combiner pattern using
9355         high_mask_shift_operand.
9357 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
9359         * tree-vect-patterns.c (vect_get_external_def_edge): New function,
9360         split out from...
9361         (vect_recog_rotate_pattern): ...here.
9362         (vect_convert_input): Try to insert casts of invariants in the
9363         preheader.
9364         * tree-vect-loop-manip.c (vect_loop_versioning): Don't require the
9365         preheader to be empty.
9367 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
9369         * tree-vect-patterns.c (append_pattern_def_seq): Take an optional
9370         vector type.  If given, install it in the new statement's
9371         STMT_VINFO_VECTYPE.
9372         (vect_element_precision): New function.
9373         (vect_unpromoted_value): New struct.
9374         (vect_unpromoted_value::vect_unpromoted_value): New function.
9375         (vect_unpromoted_value::set_op): Likewise.
9376         (vect_look_through_possible_promotion): Likewise.
9377         (vect_joust_widened_integer, vect_joust_widened_type): Likewise.
9378         (vect_widened_op_tree, vect_convert_input): Likewise.
9379         (vect_convert_inputs, vect_convert_output): Likewise.
9380         (vect_recog_dot_prod_pattern): Use vect_look_through_possible_promotion
9381         to handle the optional cast of the multiplication result and
9382         vect_widened_op_tree to detect the widened multiplication itself.
9383         Do not require the input and output of promotion casts to have
9384         the same sign, but base the signedness of the operation on the
9385         input rather than the result.  If the pattern includes two
9386         promotions, check that those promotions have the same sign.
9387         Do not restrict the MULT_EXPR handling to a double-width result;
9388         handle quadruple-width results and wider.  Use vect_convert_inputs
9389         to convert the inputs to the common type.
9390         (vect_recog_sad_pattern):  Use vect_look_through_possible_promotion
9391         to handle the optional cast of the ABS result.  Also allow a sign
9392         change or a sign extension between the ABS and MINUS.
9393         Use vect_widened_op_tree to detect the widened subtraction and use
9394         vect_convert_inputs to convert the inputs to the common type.
9395         (vect_handle_widen_op_by_const): Delete.
9396         (vect_recog_widen_op_pattern): New function.
9397         (vect_recog_widen_mult_pattern): Use it.
9398         (vect_recog_widen_shift_pattern): Likewise.
9399         (vect_recog_widen_sum_pattern): Use
9400         vect_look_through_possible_promotion to handle the promoted
9401         PLUS_EXPR operand.
9403 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
9405         * gimple-iterator.c (gsi_for_stmt): Add a new overload that takes
9406         the containing gimple_seq *.
9407         * gimple-iterator.h (gsi_for_stmt): Declare it.
9408         * tree-vect-patterns.c (vect_recog_dot_prod_pattern)
9409         (vect_recog_sad_pattern, vect_recog_widen_sum_pattern)
9410         (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
9411         (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
9412         (vect_recog_mask_conversion_pattern): Remove STMT_VINFO_IN_PATTERN_P
9413         checks.
9414         (vect_init_pattern_stmt, vect_set_pattern_stmt): New functions,
9415         split out from...
9416         (vect_mark_pattern_stmts): ...here.  Handle cases in which the
9417         statement being replaced is part of an existing pattern
9418         definition sequence, inserting the new pattern statements before
9419         the original one.
9420         (vect_pattern_recog_1): Don't return a bool.  If the statement
9421         is already part of a pattern, instead apply pattern matching
9422         to the pattern definition statements.  Don't clear the
9423         STMT_VINFO_RELATED_STMT if is_pattern_stmt_p.
9424         (vect_pattern_recog): Don't break after the first match;
9425         continue processing the pattern definition statements instead.
9426         Don't bail out for STMT_VINFO_IN_PATTERN_P here.
9428 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
9430         * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
9431         (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
9432         (vect_recog_widen_sum_pattern): Use it.
9434 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
9436         * tree-vect-loop.c (vectorizable_reduction): Assert that the
9437         phi is not a pattern statement and has not been replaced by
9438         a pattern statement.
9439         * tree-vect-patterns.c (type_conversion_p): Don't check
9440         STMT_VINFO_IN_PATTERN_P.
9441         (vect_recog_vector_vector_shift_pattern): Likewise.
9442         (vect_recog_dot_prod_pattern): Expect vect_is_simple_use to return
9443         the pattern statement rather than the original statement; check
9444         directly for a WIDEN_MULT_EXPR here.
9445         * tree-vect-slp.c (vect_get_and_check_slp_defs): Expect
9446         vect_is_simple_use to return the pattern statement rather
9447         than the original statement; use is_pattern_stmt_p to check
9448         for such a pattern statement.
9449         * tree-vect-stmts.c (process_use): Expect vect_is_simple_use
9450         to return the pattern statement rather than the original statement;
9451         don't do the same transformation here.
9452         (vect_is_simple_use): If the defining statement has been replaced
9453         by a pattern statement, return the pattern statement instead.
9454         Remove the corresponding (local) transformation from the vectype
9455         overload.
9457 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
9459         * tree-vectorizer.h (vect_is_simple_use): Move the gimple ** to the
9460         end and default to null.
9461         * tree-vect-loop.c (vect_create_epilog_for_reduction)
9462         (vectorizable_reduction): Update calls accordingly, dropping the
9463         gimple ** argument if the passed-back statement isn't needed.
9464         * tree-vect-patterns.c (vect_get_internal_def, type_conversion_p)
9465         (vect_recog_rotate_pattern): Likewise.
9466         (vect_recog_mask_conversion_pattern): Likewise.
9467         * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
9468         (vect_mask_constant_operand_p): Likewise.
9469         * tree-vect-stmts.c (is_simple_and_all_uses_invariant, process_use):
9470         (vect_model_simple_cost, vect_get_vec_def_for_operand): Likewise.
9471         (get_group_load_store_type, get_load_store_type): Likewise.
9472         (vect_check_load_store_mask, vect_check_store_rhs): Likewise.
9473         (vectorizable_call, vectorizable_simd_clone_call): Likewise.
9474         (vectorizable_conversion, vectorizable_assignment): Likewise.
9475         (vectorizable_shift, vectorizable_operation): Likewise.
9476         (vectorizable_store, vect_is_simple_cond): Likewise.
9477         (vectorizable_condition, vectorizable_comparison): Likewise.
9478         (get_same_sized_vectype, vect_get_mask_type_for_stmt): Likewise.
9479         (vect_is_simple_use): Rename the def_stmt argument to def_stmt_out
9480         and move it to the end.  Cope with null def_stmt_outs.
9482 2018-06-30  Bernd Edlinger  <bernd.edlinger@hotmail.de>
9484         * Makefile.in (FLAGS_TO_PASS): Add RANLIB_FOR_TARGET.
9486 2018-06-29  Jeff Law  <law@redhat.com>
9488         * config/v850/v850.c (v850_legitimate_address_p): Handle large
9489         displacements for TARGET_V850E2V3 and newer.
9490         (TARGET_LRA_P): Remove.  Defaults to LRA now.
9491         * config/v850/v850.md (sign23byte_load): Remove.
9492         (unsign23byte_load, sign23hword_load, unsign23hword_load): Likewise.
9493         (23word_load, 23byte_store, 23hword_store, 23word_store): Likewise.
9495 2018-06-29  Martin Liska  <mliska@suse.cz>
9497         PR lto/85759
9498         * coverage.c (coverage_init): Mangle full path name.
9499         * doc/invoke.texi: Document the change.
9500         * gcov-io.c (mangle_path): New.
9501         * gcov-io.h (mangle_path): Likewise.
9502         * gcov.c (mangle_name): Use mangle_path for path mangling.
9504 2018-06-29  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
9506         * config/arm/arm.c (output_move_double): Don't allow STRD instructions
9507         if starting source register is not even.
9509 2018-06-29  Martin Liska  <mliska@suse.cz>
9511         PR tree-optimization/86263
9512         * tree-switch-conversion.c (switch_decision_tree::try_switch_expansion):
9513         Make edge redirection.
9515 2018-06-29  David Malcolm  <dmalcolm@redhat.com>
9517         * dumpfile.c (dump_loc): Add indentation based on scope depth.
9518         (dump_scope_depth): New variable.
9519         (get_dump_scope_depth): New function.
9520         (dump_begin_scope): New function.
9521         (dump_end_scope): New function.
9522         * dumpfile.h (get_dump_scope_depth): New declaration.
9523         (dump_begin_scope): New declaration.
9524         (dump_end_scope): New declaration.
9525         (class auto_dump_scope): New class.
9526         (AUTO_DUMP_SCOPE): New macro.
9527         * tree-vectorizer.h (DUMP_VECT_SCOPE): Reimplement in terms of
9528         AUTO_DUMP_SCOPE.
9530 2018-06-29  Richard Biener  <rguenther@suse.de>
9532         * tree-vect-data-refs.c (vect_analyze_data_ref_dependences): Assert
9533         compute_all_dependences succeeds.
9534         * tree-vect-loop.c (vect_get_datarefs_in_loop): Fail early if we
9535         exceed --param loop-max-datarefs-for-datadeps.
9537 2018-06-29  Jakub Jelinek  <jakub@redhat.com>
9539         * config/rs6000/t-rs6000: Append rs6000-modes.h to TM_H.
9541 2018-06-28  Uros Bizjak  <ubizjak@gmail.com>
9543         PR target/86348
9544         * config/i386/sse.md (*vec_extractv4si_0_zext_sse4): Use
9545         alternative 0 in preferred_for_speed attribute.
9547 2018-06-28  Paul Koning  <ni1d@arrl.net>
9549         * config/pdp11/pdp11-protos.h (pdp11_shift_length): New function.
9550         * config/pdp11/pdp11.c (pdp11_shift_length): New function.
9551         * config/pdp11/pdp11.h (ADJUST_INSN_LENGTH): Remove.
9552         * config/pdp11/pdp11.md: Correct "length" attribute calculation
9553         for shift insn patterns.
9555 2018-06-28  David Malcolm  <dmalcolm@redhat.com>
9557         * cgraph.c (cgraph_node::get_body): Replace assignments to
9558         "dump_file" with calls to set_dump_file.
9559         * dumpfile.c (alt_dump_file): Make static, and group with...
9560         (alt_flags): ...this definition.
9561         (dumps_are_enabled): New variable.
9562         (refresh_dumps_are_enabled): New function.
9563         (set_dump_file): New function.
9564         (set_alt_dump_file): New function.
9565         (gcc::dump_manager::dump_start): Replace assignments to
9566         "dump_file" and "alt_dump_file" with calls to set_dump_file and
9567         set_alt_dump_file.
9568         (gcc::dump_manager::dump_finish): Likewise.
9569         * dumpfile.h (alt_dump_file): Delete decl.
9570         (dumps_are_enabled): New variable decl.
9571         (set_dump_file): New function decl.
9572         (dump_enabled_p): Rewrite in terms of new "dumps_are_enabled"
9573         global.
9574         * tree-nested.c (lower_nested_functions): Replace assignments to
9575         "dump_file" with calls to set_dump_file.
9577 2018-06-28  Eric Botcazou  <ebotcazou@adacore.com>
9579         * tree-cfg.c (verify_gimple_in_cfg): Call verify_location on the
9580         goto_locus of each outgoing edge of each basic block.
9582 2018-06-28  Richard Biener  <rguenther@suse.de>
9584         * dwarf2out.c (decl_scope_table): Remove.
9585         (push_decl_scope): Likewise.
9586         (pop_decl_scope): Likewise.
9587         (gen_type_die_for_member): Do not call push/pop_decl_scope.
9588         (gen_struct_or_union_type_die): Likewise.
9589         (gen_tagged_type_die): Likewise.
9590         (dwarf2out_init): Do not initialize decl_scope_table.
9591         (dwarf2out_c_finalize): Do not free it.
9593 2018-06-28  Richard Biener  <rguenther@suse.de>
9595         * dwarf2out.c (gen_subprogram_die): Use is_unit_die when
9596         deciding whether to not re-use a DIE.
9598 2018-06-28  Richard Biener  <rguenther@suse.de>
9600         * dwarf2out.c (gen_subprogram_die): Always re-use DIEs with an
9601         DW_AT_abstract_origin attribute.
9603 2018-06-28  Martin Liska  <mliska@suse.cz>
9605         * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
9606         Use newly introduced constants.
9607         * tree-switch-conversion.h (struct jump_table_cluster):
9608         Define max_ratio_for_size and max_ratio_for_speed.
9610 2018-06-28  Martin Liska  <mliska@suse.cz>
9612         * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
9613         Add new checking assert to catch invalid state.
9614         (jump_table_cluster::can_be_handled): Handle single case
9615         clusters.
9616         (jump_table_cluster::is_beneficial): Bail out for such case.
9617         (bit_test_cluster::find_bit_tests):
9618         Add new checking assert to catch invalid state.
9619         (bit_test_cluster::can_be_handled): Handle single case
9620         clusters.
9621         (bit_test_cluster::is_beneficial): Bail out for such case.
9622         (switch_decision_tree::analyze_switch_statement):
9623         Fix comment.
9625 2018-06-28  Martin Liska  <mliska@suse.cz>
9627         * common.opt: Introduce -completion option.
9628         * gcc.c (driver_handle_option): Handle it.
9629         (driver::main): Print completions if completion
9630         is set.
9631         * opt-suggestions.c (option_proposer::get_completions):
9632         New function.
9633         (option_proposer::suggest_completion): Likewise.
9634         (option_proposer::find_param_completions): Likewise.
9635         (verify_autocompletions): Likewise.
9636         (test_completion_valid_options): Likewise.
9637         (test_completion_valid_params): Likewise.
9638         (in_completion_p): Likewise.
9639         (empty_completion_p): Likewise.
9640         (test_completion_partial_match): Likewise.
9641         (test_completion_garbage): Likewise.
9642         (opt_proposer_c_tests): Likewise.
9643         * opt-suggestions.h: Declare new functions.
9644         * opts.c (common_handle_option): Handle OPT__completion_.
9645         * selftest-run-tests.c (selftest::run_tests): Add
9646         opt_proposer_c_tests.
9647         * selftest.c (assert_str_startswith): New.
9648         * selftest.h (assert_str_startswith): Likewise.
9649         (opt_proposer_c_tests): New.
9650         (ASSERT_STR_STARTSWITH): Likewise.
9652 2018-06-28  Martin Liska  <mliska@suse.cz>
9654         * Makefile.in: Add opt-suggestions.o.
9655         * gcc-main.c: Include opt-suggestions.h.
9656         * gcc.c (driver::driver): Likewise.
9657         (driver::~driver): Remove m_option_suggestions.
9658         (driver::build_option_suggestions): Moved to option_proposer.
9659         (driver::suggest_option): Likewise.
9660         (driver::handle_unrecognized_options): Use option_proposer.
9661         * gcc.h (class driver): Add new memver m_option_proposer.
9662         * opt-suggestions.c: New file.
9663         * opt-suggestions.h: New file.
9665 2018-06-28  Martin Liska  <mliska@suse.cz>
9667         * vec.h (class auto_string_vec): New (moved from auto_argvec).
9668         (auto_string_vec::~auto_string_vec): Likewise.
9670 2018-06-28  Eric Botcazou  <ebotcazou@adacore.com>
9672         * tree-inline.h (struct copy_body_data): Move remapping_type_depth and
9673         prevent_decl_creation_for_types fields up and add reset_location field.
9674         * tree-inline.c (remap_gimple_stmt): Force input_location on the new
9675         statement if id->reset_location is true.
9676         (copy_edges_for_bb): Do not set goto_locus on the new edges if
9677         id->reset_location is true.
9678         (copy_phis_for_bb): Force input_location on the arguments if
9679         id->reset_location is true.
9680         (expand_call_inline): Set id->reset_location if DECL_IGNORED_P
9681         is set on the function to be inlined.
9683 2018-06-27  Stephan Bergmann  <sbergman@redhat.com>
9685         * doc/invoke.texi (Debugging Options): Add -gsplit-dwarf.
9687 2018-06-27  Dimitar Dimitrov  <dimitar@dinux.eu>
9689         * lra-eliminations.c (update_reg_eliminate): Mark all spanning hard
9690         registers for Pmode.
9691         * lra-lives.c (check_pseudos_live_through_calls): Mark all spanning
9692         hard registers for the clobbered pseudo.
9694 2018-06-27  Paul Koning  <ni1d@arrl.net>
9696         * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
9697         mutually exclusive options.
9698         * config/pdp11/constraints.md (h): New constraint.
9699         (O): Update definition to match shift code generation.
9700         (D): New constraint.
9701         * config/pdp11/pdp11-modes.def (CCNZ): Define mode.
9702         (CCFP): Remove.
9703         * config/pdp11/pdp11-protos.h (int_no_side_effect_operand): New
9704         function.
9705         (output_jump): Change arguments.
9706         (pdp11_fixed_cc_regs): New function.
9707         (pdp11_cc_mode): Ditto.
9708         (pdp11_expand_shift): Ditto.
9709         (pdp11_assemble_shift): Ditto.
9710         (pdp11_small_shift): Ditto.
9711         (pdp11_branch_cost): Remove.
9712         * config/pdp11/pdp11.c (pdp11_assemble_integer): Remove comments
9713         from output.
9714         (pdp11_register_move_cost): Update for CC registers.
9715         (pdp11_rtx_costs): Add case for LSHIFTRT.
9716         (pdp11_output_jump): Add CCNZ mode conditional branches.
9717         (notice_update_cc_on_set): Remove.
9718         (pdp11_cc_mode): New function.
9719         (simple_memory_operand): Correct pre/post decrement case.
9720         (no_side_effect_operand): New function.
9721         (pdp11_regno_reg_class): Add CC_REGS class.
9722         (pdp11_fixed_cc_regs): New function.
9723         (pdp11_small_shift): New function.
9724         (pdp11_expand_shift): New function to expand shift insns.
9725         (pdp11_assemble_shift): New function to output shifts.
9726         (pdp11_branch_cost): Remove.
9727         (pdp11_modes_tieable_p): Make QI/HI modes tieable.
9728         * config/pdp11/pdp11.h (SIZE_TYPE): Ensure 16-bit type.
9729         (WCHAR_TYPE): Ditto.
9730         (PTRDIFF_TYPE): Ditto.
9731         (ADJUST_INSN_LENGTH): New macro.
9732         (FIXED_REGISTERS): Add CC registers.
9733         (CALL_USED_REGISTERS): Ditto.
9734         (reg_class): Ditto.
9735         (REG_CLASS_NAMES): Ditto.
9736         (REG_CLASS_CONTENTS): Ditto.
9737         (SELECT_CC_MODE): Use new function.
9738         (TARGET_FLAGS_REGNUM): New macro.
9739         (TARGET_FIXED_CONDITION_CODE_REGS): Ditto.
9740         (cc0_reg_rtx): Remove.
9741         (CC_STATUS_MDEP): Remove.
9742         (CC_STATUS_MDEFP_INIT): Remove.
9743         (CC_IN_FPU): Remove.
9744         (NOTICE_UPDATE_CC): Remove.
9745         (REGISTER_NAMES): Add CC registers.
9746         (BRANCH_COST): Change to constant 1.
9747         * config/pdp11/pdp11.md: Rewrite for CCmode condition code
9748         handling.
9749         * config/pdp11/pdp11.opt (mbcopy): Remove.
9750         (mbcopy-builtin): Remove.
9751         (mbranch-cheap): Remove.
9752         (mbranch-expensive): Remove.
9753         * config/pdp11/predicates.md (expand_shift_operand): Update to
9754         match shift code generation.
9755         (ccnz_operator): New predicate.
9756         * doc/invoke.texi (PDP-11 Options): Remove deleted options
9757         -mbcopy, -mbcopy-builtin, -mbranch-cheap, -mbranch-expensive.
9758         Remove non-existent option -mabshi, -mno-abshi.  Document mutually
9759         exclusive options.
9760         * doc/md.texi (PDP-11): Document new D and h constraints.  Update
9761         description of O constraint.
9763 2018-06-27  Jeff Law  <law@redhat.com>
9764             Austin Law  <austinklaw@gmail.com>
9766         * config/v850/v850.md (addsi3_set_flags): New pattern.
9767         (subsi3_set_flags, negsi2_set_flags, andsi3_set_flags): Likewise.
9768         (iorsi3_set_flags, xorsi3_set_flags, one_cmplsi2_set_flags): Likewise.
9769         (zero_extendhisi2_v850_set_flags): Likewise.
9770         (zero_extendqisi2_v850_set_flags): Likewise.
9771         (ashlsi3_set_flags, ashlsi3_v850e2_set_flags): Likewise.
9772         (lshrsi3_set_flags, lshrsi3_v850e2_set_flags): Likewise.
9773         (ashrsi3_set_flags, ashrsi3_v850e2_set_flags): Likewise.
9775         * config/v850/v850-protos.h (notice_update_cc): Remove.
9776         * config/v850/v850.c (v850_compare_op0, v850_compare_op1): Remove.
9777         (v850_print_operand): Handle 'D' and "d".
9778         (v850_select_cc_mode): Remove ATTRIBUTE_UNUSED for last argument.
9779         Add handling of arithmetic/logical operations compared against zero.
9780         (v850_gen_float_compare): Remove ATTRIBUTE_UNUSED for last argument.
9781         Do not look at v850_compare_op, instead get mode from last argument.
9782         (v850_gen_compare): Remove
9783         (increment_stack): Use addsi3_clobber_flags to avoid splitting failure
9784         after reload for prologue insns.
9785         (expand_prologue): Account for CLOBBER of CC_REGNUM in various
9786         patterns.
9787         (construct_save_jarl): Likewise.
9788         (TARGET_FLAGS_REGNUM): Define.
9789         * config/v850/v850.h (v850_compare_op0, v850_compare_op1): Remove.
9790         (NOTICE_UPDATE_CC): Remove.
9791         * config/v850/v850.md (v850_tst1): Use (reg:CCZ CC_REGNUM) rather
9792         than cc0.  Conditionalize on reload_completed.
9793         (cmpsi_insn, setfcc_insn): Likewise.
9794         (tst1 splitter): Turn into define_and_split which sets the flags
9795         after reload.
9796         (cstoresi4, cbranchsf4, cbranchdf4, cbranchsi4_insn): Likewise.
9797         (cbranchsi4, branch_normal, branch_invert): Do not expose cc0 here.
9798         (cstoresf4, cstoredf4): Clobber the flags.
9799         (cmpsi, cmpsf, cmpdf): Remove expanders.
9800         (setf_insn): Remove pattern.
9801         (addsi3): Turn into define_and_split which clobbers the flags after
9802         reload and a suitable pattern (addsi3_clobber_flags) for use after
9803         reload.
9804         (subsi3, negsi2, andsi3, iorsi3, xorsi3, one_cmplsi2) Likewise.
9805         (ashlsi3, ashlsi3_v850e2, lshrsi3, lsh4si3_v850e2): Likewise.
9806         (ashrsi3, ashrsi3_v850e2): Likewise.
9807         (bins): Clobber the flags.
9808         (movsicc_normal_cc, movsicc_normal, movsicc_tst1): Likewise.
9809         (movsicc_tst1_revesed, sasf, swap and rotate patterns): Likewise.
9810         (fix_loop_counter, call_internal_short, call_internal_long): Likewise.
9811         (call_value_internal_short, call_value_internal_long): Likewise.
9812         (callt_save_interrupt, callt_return_interrupt): Likewise.
9813         (save_interrupt, return_interrupt): Likewise.
9814         (callt_save_all_interrupt, save_all_interrupt): Likewise.
9815         (_save_all_interrupt, callt_restore_all_interrupt): Likewise.
9816         (restore_all_interrupt, _restore_all_interrupt): Likewise.
9817         (All FP comparisons): Only allow after reload has completed.
9818         (trfsr): Likewise.
9819         (divh, divhu): Tweak output template.
9820         (branch_z_normal, branch_z_invert): Remove
9821         (branch_nz_normal, branch_nz_invert): Likewise.
9822         (extendhisi_insn, extendqisi_insn): Do not clobber flags.
9824         * config/v850/v850-modes.def (CCZ, CCNZ): Add new modes.
9825         * config/v850/v850.c (notice_update_cc): Remove.
9826         * config/v850/v850.h  (CC_OVERFLOW_UNUSABLE): Remove
9827         (CC_NO_CARRY): Likewise.
9828         (NOTICE_UPDATE_CC): Define to nothing.
9829         * config/v850/v850.md: Remove block comment on cc0 handling
9830         Remove "cc" attribute from all patterns.  Remove cc_status handling
9831         from all patterns.  Minor formatting fixes.
9833 2018-06-27  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
9835         * config/aarch64/aarch64-cores.def (cortex-a76): New entry.
9836         (cortex-a76.cortex-a55): Likewise.
9837         * config/aarch64/aarch64-tune.md: Regenerate.
9838         * doc/invoke.texi (AArch64 Options): Document cortex-a76 and
9839         cortex-a76.cortex-a55.
9841 2018-06-27  Jeff Law  <law@redhat.com>
9843         * config/v850/t-v850 (MULTILIB_OPTIONS): Remove 8byte-align.
9844         (MULTILIB_DIRNAMES): Similarly.
9846 2018-06-27  Eric Botcazou  <ebotcazou@adacore.com>
9848         * gimple.h (gimple_return_retbnd): Delete.
9849         (gimple_return_set_retbnd): Likewise.
9850         * cgraphunit.c (cgraph_node::expand_thunk): Remove call to
9851         gimple_return_set_retbnd.
9852         * gimple-pretty-print.c (dump_gimple_return): Remove call to
9853         gimple_return_retbnd and adjust.
9854         * tree-inline.h (struct copy_body_data): Remove retbnd field.
9855         * tree-inline.c (remap_gimple_stmt): Remove handling of retbnd.
9856         Explicitly return NULL in a couple more cases.  Move assertion
9857         on debug statements and remove unreachable code.
9858         (reset_debug_binding): Do not test id->retbnd.
9859         (expand_call_inline): Do not set it.
9861 2018-06-27  Rasmus Villemoes  <rv@rasmusvillemoes.dk>
9863         * configure.ac: Add --disable-gcov option.
9864         * configure: Regenerate.
9865         * Makefile.in: Honour @enable_gcov@.
9866         * doc/install.texi: Document --disable-gcov.
9868 2018-06-27  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
9870         * config/arm/arm-cpus.in (cortex-a76): New entry.
9871         (cortex-a76.cortex-a55): Likewise.
9872         * config/arm/arm-tables.opt: Regenerate.
9873         * config/arm/arm-tune.md: Likewise.
9874         * config/arm/driver-arm.c (arm_cpu_table): Add Cortex-A76 entry.
9875         * doc/invoke.texi (ARM Options): Document cortex-a76 and
9876         cortex-a76.cortex-a55.
9878 2018-06-27  Tamar Christina  <tamar.christina@arm.com>
9880         PR target/85769
9881         * config/aarch64/aarch64.md (*movhf_aarch64): Add dup v0.4h pattern.
9883 2018-06-27  Siddhesh Poyarekar  <siddhesh@sourceware.org>
9885         * config/aarch64/aarch64.h (CALL_USE_REGISTERS): Fix obsolete
9886         comment.
9887         (EPILOGUE_USES): Likewise.
9889 2018-06-26  Eric Botcazou  <ebotcazou@adacore.com>
9891         * tree-inline.c (remap_location): New function extracted from...
9892         (copy_edges_for_bb): Add ID parameter.  Remap goto_locus.
9893         (copy_phis_for_bb): ...here.  Call remap_location.
9894         (copy_cfg_body): Adjust call to copy_edges_for_bb.
9896 2018-06-26  Aaron Sawdey  <acsawdey@linux.ibm.com>
9898         * config/rs6000/rs6000-string.c (expand_block_clear): Don't use
9899         unaligned vsx for 16B memset.
9901 2018-06-26  Segher Boessenkool  <segher@kernel.crashing.org>
9903         PR target/86285
9904         * config/rs6000/rs6000.c (rs6000_init_builtins): Do not set
9905         ieee128_float_type_node to long_double_type_node unless
9906         TARGET_LONG_DOUBLE_128 is set.
9908 2018-06-26  David Malcolm  <dmalcolm@redhat.com>
9910         * cfgloop.c (get_loop_location): Convert return type from
9911         location_t to dump_user_location_t, replacing INSN_LOCATION lookups
9912         by implicit construction from rtx_insn *, and using
9913         dump_user_location_t::from_function_decl for the fallback case.
9914         * cfgloop.h (get_loop_location): Convert return type from
9915         location_t to dump_user_location_t.
9916         * cgraphunit.c (walk_polymorphic_call_targets): Update call to
9917         dump_printf_loc to pass in a dump_location_t rather than a
9918         location_t, via the gimple stmt.
9919         * coverage.c (get_coverage_counts): Update calls to
9920         dump_printf_loc to pass in dump_location_t rather than a
9921         location_t.
9922         * doc/optinfo.texi (Dump types): Convert example of
9923         dump_printf_loc from taking "locus" to taking "insn".  Update
9924         description of the "_loc" calls to cover dump_location_t.
9925         * dumpfile.c: Include "backend.h", "gimple.h", "rtl.h", and
9926         "selftest.h".
9927         (dump_user_location_t::dump_user_location_t): New constructors,
9928         from gimple *stmt and rtx_insn *.
9929         (dump_user_location_t::from_function_decl): New function.
9930         (dump_loc): Make static.
9931         (dump_gimple_stmt_loc): Convert param "loc" from location_t to
9932         const dump_location_t &.
9933         (dump_generic_expr_loc): Delete.
9934         (dump_printf_loc): Convert param "loc" from location_t to
9935         const dump_location_t &.
9936         (selftest::test_impl_location): New function.
9937         (selftest::dumpfile_c_tests): New function.
9938         * dumpfile.h: Include "profile-count.h".
9939         (class dump_user_location_t): New class.
9940         (struct dump_impl_location_t): New struct.
9941         (class dump_location_t): New class.
9942         (dump_printf_loc): Convert 2nd param from source_location to
9943         const dump_location_t &.
9944         (dump_generic_expr_loc): Delete.
9945         (dump_gimple_stmt_loc): Convert 2nd param from source_location to
9946         const dump_location_t &.
9947         * gimple-fold.c (fold_gimple_assign): Update call to
9948         dump_printf_loc to pass in a dump_location_t rather than a
9949         location_t, via the gimple stmt.
9950         (gimple_fold_call): Likewise.
9951         * gimple-loop-interchange.cc
9952         (loop_cand::analyze_iloop_reduction_var): Update for change to
9953         check_reduction_path.
9954         (tree_loop_interchange::interchange): Update for change to
9955         find_loop_location.
9956         * graphite-isl-ast-to-gimple.c (scop_to_isl_ast): Update for
9957         change in return-type of find_loop_location.
9958         (graphite_regenerate_ast_isl): Likewise.
9959         * graphite-optimize-isl.c (optimize_isl): Likewise.
9960         * graphite.c (graphite_transform_loops): Likewise.
9961         * ipa-devirt.c (ipa_devirt): Update call to dump_printf_loc to
9962         pass in a dump_location_t rather than a location_t, via the
9963         gimple stmt.
9964         * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
9965         * ipa.c (walk_polymorphic_call_targets): Likewise.
9966         * loop-unroll.c (report_unroll): Convert "locus" param from
9967         location_t to dump_location_t.
9968         (decide_unrolling): Update for change to get_loop_location's
9969         return type.
9970         * omp-grid.c (struct grid_prop): Convert field "target_loc" from
9971         location_t to dump_user_location_t.
9972         (grid_find_single_omp_among_assignments_1): Updates calls to
9973         dump_printf_loc to pass in a dump_location_t rather than a
9974         location_t, via the gimple stmt.
9975         (grid_parallel_clauses_gridifiable): Convert "tloc" from
9976         location_t to dump_location_t.  Updates calls to dump_printf_loc
9977         to pass in a dump_location_t rather than a location_t, via the
9978         gimple stmt.
9979         (grid_inner_loop_gridifiable_p): Likewise.
9980         (grid_dist_follows_simple_pattern): Likewise.
9981         (grid_gfor_follows_tiling_pattern): Likewise.
9982         (grid_target_follows_gridifiable_pattern): Likewise.
9983         (grid_attempt_target_gridification): Convert initialization
9984         of local "grid" from memset to zero-initialization; FIXME: does
9985         this require C++11?  Update call to dump_printf_loc to pass in a
9986         optinfo_location rather than a location_t, via the gimple stmt.
9987         * profile.c (read_profile_edge_counts): Updates call to
9988         dump_printf_loc to pass in a dump_location_t rather than a
9989         location_t
9990         (compute_branch_probabilities): Likewise.
9991         * selftest-run-tests.c (selftest::run_tests): Call
9992         dumpfile_c_tests.
9993         * selftest.h (dumpfile_c_tests): New decl.
9994         * tree-loop-distribution.c (pass_loop_distribution::execute):
9995         Update for change in return type of find_loop_location.
9996         * tree-parloops.c (parallelize_loops): Likewise.
9997         * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Convert
9998         "locus" from location_t to dump_user_location_t.
9999         (canonicalize_loop_induction_variables): Likewise.
10000         * tree-ssa-loop-ivopts.c (tree_ssa_iv_optimize_loop): Update
10001         for change in return type of find_loop_location.
10002         * tree-ssa-loop-niter.c (number_of_iterations_exit): Update call
10003         to dump_printf_loc to pass in a dump_location_t rather than a
10004         location_t, via the stmt.
10005         * tree-ssa-sccvn.c (eliminate_dom_walker::before_dom_children):
10006         Likewise.
10007         * tree-vect-loop-manip.c (find_loop_location): Convert return
10008         type from source_location to dump_user_location_t.
10009         (vect_do_peeling): Update for above change.
10010         (vect_loop_versioning): Update for change in type of
10011         vect_location.
10012         * tree-vect-loop.c (check_reduction_path): Convert "loc" param
10013         from location_t to dump_user_location_t.
10014         (vect_estimate_min_profitable_iters): Update for change in type
10015         of vect_location.
10016         * tree-vect-slp.c (vect_print_slp_tree): Convert param "loc" from
10017         location_t to dump_location_t.
10018         (vect_slp_bb): Update for change in type of vect_location.
10019         * tree-vectorizer.c (vect_location): Convert from source_location
10020         to dump_user_location_t.
10021         (try_vectorize_loop_1): Update for change in vect_location's type.
10022         (vectorize_loops): Likewise.
10023         (increase_alignment): Likewise.
10024         * tree-vectorizer.h (vect_location): Convert from source_location
10025         to dump_user_location_t.
10026         (find_loop_location): Convert return type from source_location to
10027         dump_user_location_t.
10028         (check_reduction_path): Convert 1st param from location_t to
10029         dump_user_location_t.
10030         * value-prof.c (check_counter): Update call to dump_printf_loc to
10031         pass in a dump_user_location_t rather than a location_t; update
10032         call to error_at for change in type of "locus".
10033         (check_ic_target): Update call to dump_printf_loc to
10034         pass in a dump_user_location_t rather than a location_t, via the
10035         call_stmt.
10037 2018-06-26  Robin Dapp  <rdapp@linux.vnet.ibm.com>
10039         * config/s390/s390.h (enum processor_flags): Do not use
10040         default tune parameter when -march was specified.
10042 2018-06-26  Jakub Jelinek  <jakub@redhat.com>
10044         PR target/86314
10045         * config/i386/i386.md (setcc + movzbl to xor + setcc peephole2s):
10046         Check reg_overlap_mentioned_p in addition to reg_set_p with the same
10047         operands.
10049 2018-06-26  Richard Biener  <rguenther@suse.de>
10051         PR tree-optimization/86287
10052         PR bootstrap/86316
10053         * tree-vect-loop.c (vect_transform_loop_stmt): Fix read-after-free.
10054         (vect_analyze_loop): Initialize n_stmts.
10056 2018-06-26  Richard Biener  <rguenther@suse.de>
10058         PR middle-end/86271
10059         * fold-const.c (fold_convertible_p): Pointer extension
10060         isn't valid.
10062 2018-06-26  Alexandre Oliva <aoliva@redhat.com>
10064         PR debug/86064
10065         * dwarf2out.c (loc_list_has_views): Adjust comments.
10066         (dw_loc_list): Split single cross-partition range with
10067         nonzero locview.
10069 2018-06-25  Jeff Law  <law@redhat.com>
10071         * common/config/v850/v850-common.c (TARGET_DEFAULT_TARGET_FLAGS): Turn
10072         on -mbig-switch by default.
10074         * config/v850/predicates.md (const_float_1_operand): Fix match_code
10075         test.
10076         (const_float_0_operand): Remove unused predicate.
10077         * config/v850/v850.md (define_constants): Remove UNSPEC_LOOP.
10078         (define_c_enum unspec): Add LOOP, RCP and RSQRT constants.
10079         (recipsf2): New expander.  Original pattern now called
10080         (recipsf2_insn).
10081         (recipdf2, recipdf2_insn): Similarly.
10082         (rsqrtsf2, rsqrtsf2_insn): Similarly
10083         (rsqrtdf2, rsqrtdf2_insn): Similarly
10085 2018-06-26  Gerald Pfeifer  <gerald@pfeifer.com>
10087         * ginclude/stddef.h: Remove an obsolete comment on FreeBSD 5.
10088         Simplify logic for FreeBSD (twice).
10090 2018-06-25  Martin Sebor  <msebor@redhat.com>
10092         PR tree-optimization/86204
10093         * tree-ssa-strlen.c (handle_builtin_strlen): Avoid storing
10094         a strnlen result if it's less than the length of the string.
10096 2018-06-25  Martin Sebor  <msebor@redhat.com>
10098         PR tree-optimization/85700
10099         * gimple-fold.c (gimple_fold_builtin_strncat): Adjust comment.
10100         * tree-ssa-strlen.c (is_strlen_related_p): Handle integer subtraction.
10101         (maybe_diag_stxncpy_trunc): Distinguish strncat from strncpy.
10103 2018-06-25  Martin Sebor  <msebor@redhat.com>
10105         * doc/extend.texi (Zero-length arrays): Update and clarify.
10107 2018-06-25  Michael Meissner  <meissner@linux.ibm.com>
10109         * config.gcc (powerpc64le*): Revert January 16th, 2018 patch that
10110         added IEEE/IBM long double multilib support on PowerPC little
10111         endian Linux systems.
10112         * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Likewise.
10113         (MULTILIB_DEFAULTS): Likewise.
10114         * config/rs6000/rs6000.c (rs6000_option_override_internal):
10115         Likewise.
10116         * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Likewise.
10117         * config/rs6000/t-ldouble-linux64le-ibm: Delete, no longer used.
10118         * config/rs6000/t-ldouble-linux64le-ieee: Delete, no longer used.
10120 2018-06-25  Alexander Monakov  <amonakov@ispras.ru>
10122         PR middle-end/86311
10123         * sort.cc (REORDER_23): Avoid memcpy with same destination and source.
10124         (REORDER_45): Likewise.
10126 2018-06-25  Jeff Law  <law@redhat.com>
10128         * config/v850/v850.md (divmodhi4): Make sure to sign extend the
10129         dividend to 32 bits.  Adjust length.
10130         (udivmodhi4): Cleanup output template.  Fix length.
10132 2018-06-25  Carl Love  <cel@us.ibm.com>
10134         * config/rs6000/vsx.md: Change word selector to prefered location.
10136 2018-06-25  Richard Biener  <rguenther@suse.de>
10138         PR tree-optimization/86304
10139         * tree-vectorizer.c (vectorize_loops): Walk over new possibly
10140         epilogue-if-converted loops as well.
10142 2018-06-25  Jan Hubicka  <hubicka@ucw.cz>
10144         * lto-section-out.c (lto_begin_section): Do not print section
10145         name for noaddr and unnumbered dumps.
10147 2018-06-25  Richard Biener  <rguenther@suse.de>
10149         * tree-vectorizer.h (struct vec_info_shared): New structure
10150         with parts split out from struct vec_info and loop_nest from
10151         struct _loop_vec_info.
10152         (struct vec_info): Adjust accordingly.
10153         (struct _loop_vec_info): Likewise.
10154         (LOOP_VINFO_LOOP_NEST): Adjust.
10155         (LOOP_VINFO_DATAREFS): Likewise.
10156         (LOOP_VINFO_DDRS): Likewise.
10157         (struct _bb_vec_info): Likewise.
10158         (BB_VINFO_DATAREFS): Likewise.
10159         (BB_VINFO_DDRS): Likewise.
10160         (struct _stmt_vec_info): Add dr_aux member.
10161         (DR_VECT_AUX): Adjust to refer to member of DR_STMTs vinfo.
10162         (DR_MISALIGNMENT_UNINITIALIZED): New.
10163         (set_dr_misalignment): Adjust.
10164         (dr_misalignment): Assert misalign isn't DR_MISALIGNMENT_UNINITIALIZED.
10165         (vect_analyze_loop): Adjust prototype.
10166         (vect_analyze_loop_form): Likewise.
10167         * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
10168         Compute dependences lazily.
10169         (vect_record_base_alignments): Use shared datarefs/ddrs.
10170         (vect_verify_datarefs_alignment): Likewise.
10171         (vect_analyze_data_refs_alignment): Likewise.
10172         (vect_analyze_data_ref_accesses): Likewise.
10173         (vect_analyze_data_refs): Likewise.
10174         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Add
10175         constructor parameter for shared part.
10176         (vect_analyze_loop_form): Pass in shared part and adjust.
10177         (vect_analyze_loop_2): Pass in storage for the number of
10178         stmts.  Move loop nest finding to the caller.  Compute
10179         datarefs lazily.
10180         (vect_analyze_loop): Pass in shared part.
10181         (vect_transform_loop): Verify shared datarefs are unchanged.
10182         * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Add
10183         constructor parameter for shared part.
10184         (vect_slp_analyze_bb_1): Pass in shared part and adjust.
10185         (vect_slp_bb): Verify shared datarefs are unchanged before
10186         transform.
10187         * tree-vect-stmts.c (ensure_base_align): Adjust for DR_AUX
10188         change.
10189         (new_stmt_vec_info): Initialize DR_AUX misalignment to
10190         DR_MISALIGNMENT_UNINITIALIZED.
10191         * tree-vectorizer.c (vec_info::vec_info): Add constructor
10192         parameter for shared part.
10193         (vec_info::~vec_info): Adjust.
10194         (vec_info_shared::vec_info_shared): New.
10195         (vec_info_shared::~vec_info_shared): Likewise.
10196         (vec_info_shared::save_datarefs): Likewise.
10197         (vec_info_shared::check_datarefs): Likewise.
10198         (try_vectorize_loop_1): Construct shared part live for analyses
10199         of a single loop for multiple vector sizes.
10200         * tree-parloops.c (gather_scalar_reductions): Adjust.
10202 2018-06-25  Richard Biener  <rguenther@suse.de>
10204         * tree-vect-data-refs.c (vect_find_stmt_data_reference): Modify
10205         DR for SIMD lane accesses here and mark DR with (void *)-1 aux.
10206         (vect_analyze_data_refs): Remove similar code from here and
10207         simplify accordingly.
10209 2018-06-25  Richard Biener  <rguenther@suse.de>
10211         * tree-vect-data-refs.c (vect_check_gather_scatter): Fail
10212         for reverse storage order accesses rather than asserting
10213         they cannot happen here.
10215 2018-06-25  Tom de Vries  <tdevries@suse.de>
10217         PR debug/86257
10218         * config/i386/i386.md (define_insn "*tls_global_dynamic_64_<mode>"):
10219         Use data16 instead of .byte for insn prefix.
10221 2018-06-25  Andreas Krebbel  <krebbel@linux.ibm.com>
10223         PR C++/86082
10224         * parser.c (make_char_string_pack): Pass this literal chars
10225         through cpp_interpret_string.
10226         (cp_parser_userdef_numeric_literal): Check the result of
10227         make_char_string_pack.
10229 2018-06-24  Maya Rashish  <coypu@sdf.org>
10231         * ginclude/stddef.h: Simplify conditions around avoiding
10232         re-definition of __size_t.
10234 2018-06-22  Jan Hubicka  <hubicka@ucw.cz>
10236         * lto-streamer-out.c (tree_is_indexable): Make LABEL_DECL nonindexable
10237         unless it is forced or nonlocal; assert that we stream no IMPORTED_DECL.
10239 2018-06-22  Maya Rashish  <coypu@sdf.org>
10241         * doc/invoke.texi (mno-fancy-math-387): Update for changes
10242         made to OpenBSD and NetBSD through the years.
10244 2018-06-22  Kelvin Nilsen  <kelvin@gcc.gnu.org>
10246         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
10247         behavior of vec_pack (vector double, vector double) to match
10248         behavior of vec_float2 (vector double, vector double).
10250 2018-06-22  Olivier Hainque  <hainque@adacore.com>
10252         * gimplify.c (gimplify_function_tree): Prevent creation
10253         of a trampoline for the address of the current function
10254         passed to entry/exit instrumentation hooks.
10256 2018-06-22  Aaron Sawdey  <acsawdey@linux.ibm.com>
10258         PR target/86222
10259         * config/rs6000/rs6000-string.c (expand_strn_compare): Handle -m32
10260         correctly.
10262 2018-06-22  Martin Liska  <mliska@suse.cz>
10264         PR tree-optimization/86263
10265         * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
10266         Bail out if is_enabled is false.
10267         * tree-switch-conversion.h (jump_table_cluster::is_enabled):
10268         New declaration.
10269         (jump_table_cluster::is_enabled): New function.
10271 2018-06-22  Jan Hubicka  <hubicka@ucw.cz>
10273         * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
10274         BINFO_BASE_ACCESSES and BINFO_VPTR_FIELD.
10275         * tree-streamer-in.c (streamer_read_tree_bitfields): Likewise.
10276         (lto_input_ts_binfo_tree_pointers): Likewise.
10277         * tree-streamer-out.c (streamer_write_tree_bitfields,
10278         write_ts_binfo_tree_pointers): Likewise.
10279         * tree.c (free_lang_data_in_binfo): Clear BINFO_VPTR_FIELD.
10281 2018-06-22  Jan Hubicka  <hubicka@ucw.cz>
10283         * tree.c (free_lang_data_in_type): Free all TYPE_VFIELDs.
10285 2018-06-22  Martin Liska  <mliska@suse.cz>
10287         * symbol-summary.h (get): Make it pure and inline move
10288         functionality from ::get function.
10289         (get): Remove and inline into ::get and ::get_create.
10290         (get_create): Move code from ::get function.
10292 2018-06-22  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
10294         PR target/85994
10295         * config/i386/sol2.h (CPP_SPEC): Don't pass -P for
10296         -x assembler-with-cpp.
10298 2018-06-22  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
10300         * config/sol2.h (TARGET_OS_CPP_BUILTINS): Define
10301         _FILE_OFFSET_BITS=64 for C++.
10303 2018-06-21  Michael Meissner  <meissner@linux.ibm.com>
10305         * config/rs6000/rs6000.md (extendtfif2): Add missing 128-bit
10306         conversion insn that shows up when pr85657-3.c is compiled using
10307         IEEE 128-bit long double.
10308         (neg<mode>2_internal): Use the correct mode to check whether the
10309         mode is IBM extended.
10310         * config/rs6000/rs6000.c (init_float128_ieee): Prevent complex
10311         multiply and divide external functions from being created more
10312         than once.
10314 2018-06-21  Eric Botcazou  <ebotcazou@adacore.com>
10316         * cfgrtl.c (fixup_reorder_chain): Do not emit NOPs in DECL_IGNORED_P
10317         functions.
10318         (rtl_merge_blocks): Likewise.  Do not emit a NOP if the location of
10319         the edge can be forwarded.
10320         (cfg_layout_merge_blocks): Likewise.
10322 2018-06-21  Eric Botcazou  <ebotcazou@adacore.com>
10324         * except.c (finish_eh_generation): Commit edge insertions only after
10325         the EH edges have been redirected from post-landing to landing pads.
10327 2018-06-21  Eric Botcazou  <ebotcazou@adacore.com>
10329         * tree-nested.c (get_frame_type): Use create_tmp_var_raw instead of
10330         create_tmp_var_for to create the FRAME decl.
10331         (finalize_nesting_tree_1): Do not unchain the FRAME decl.
10333 2018-06-21  Eric Botcazou  <ebotcazou@adacore.com>
10335         * tree-inline.c (copy_edges_for_bb): Minor tweak.
10336         (maybe_move_debug_stmts_to_successors): Also reset the locus of the
10337         debug statement when resetting its value.
10338         (expand_call_inline): Copy the locus of the call onto the assignment
10339         of the return value, if any.  Use local variable in more cases.
10341 2018-06-21  Martin Liska  <mliska@suse.cz>
10343         * ipa-pure-const.c (propagate_nothrow): Use
10344         funct_state_summaries->get.
10345         (dump_malloc_lattice): Likewise.
10346         (propagate_malloc): Likewise.
10348 2018-06-21  Richard Biener  <rguenther@suse.de>
10350         * lto-streamer-out.c (DFS::DFS_write_tree_body): Update outdated
10351         comment.  Follow BLOCK_ABSTRACT_ORIGIN unconditionally.
10352         * tree-streamer-in.c (lto_input_ts_block_tree_pointers): Update
10353         comment.
10354         * tree-streamer-out.c (write_ts_block_tree_pointers): Stream
10355         BLOCK_ABSTRACT_ORIGIN unconditionally.
10357 2018-06-21  David Malcolm  <dmalcolm@redhat.com>
10359         * ipa-cp.c (ipcp_driver): Set edge_clone_summaries to NULL after
10360         deleting it.
10361         * ipa-reference.c (ipa_reference_c_finalize): Delete
10362         ipa_ref_opt_sum_summaries and set it to NULL.
10364 2018-06-21  Tom de Vries  <tdevries@suse.de>
10366         PR tree-optimization/85859
10367         * tree-ssa-tail-merge.c (stmt_local_def): Copy gimple_is_call
10368         test with comment from bb_no_side_effects_p.
10370 2018-06-21  Richard Biener  <rguenther@suse.de>
10372         PR tree-optimization/86232
10373         * tree-ssa-loop-niter.c (number_of_iterations_popcount): Adjust
10374         max for constant niter.
10376 2018-06-21  Andre Vieira  <andre.simoesdiasvieira@arm.com>
10378         * config/aarch64/aarch64-simd.md
10379         (*aarch64_crypto_aes<aes_op>v16qi_xor_combine): New.
10381 2018-06-21  Andre Vieira  <andre.simoesdiasvieira@arm.com>
10383         * config/aarch64/aarch64-simd.md (aarch64_crypto_aes<aes_op>v16qi):
10384         Make opernads of the unspec commutative.
10386 2018-06-21  Richard Biener  <rguenther@suse.de>
10388         * tree-data-ref.c (dr_step_indicator): Handle NULL DR_STEP.
10389         * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr):
10390         Avoid calling vect_mark_for_runtime_alias_test with gathers or scatters.
10391         (vect_analyze_data_ref_dependence): Re-order checks to deal with
10392         NULL DR_STEP.
10393         (vect_record_base_alignments): Do not record base alignment
10394         for gathers or scatters.
10395         (vect_compute_data_ref_alignment): Drop return value that is always
10396         true.  Bail out early for gathers or scatters.
10397         (vect_enhance_data_refs_alignment): Bail out early for gathers
10398         or scatters.
10399         (vect_find_same_alignment_drs): Likewise.
10400         (vect_analyze_data_refs_alignment): Remove dead code.
10401         (vect_slp_analyze_and_verify_node_alignment): Likewise.
10402         (vect_analyze_data_refs): For possible gathers or scatters do
10403         not create an alternate DR, just check their possible validity
10404         and mark them.  Adjust DECL_NONALIASED handling to not rely
10405         on DR_BASE_ADDRESS.
10406         * tree-vect-loop-manip.c (vect_update_inits_of_drs): Do not
10407         update inits of gathers or scatters.
10408         * tree-vect-patterns.c (vect_recog_mask_conversion_pattern):
10409         Also copy gather/scatter flag to pattern vinfo.
10411 2018-06-20  Kelvin Nilsen  <kelvin@gcc.gnu.org>
10413         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
10414         behavior of vec_packsu (vector unsigned long long, vector unsigned
10415         long long) to match behavior of vec_packs with same signature.
10417 2018-06-20  Chung-Lin Tang <cltang@codesourcery.com>
10418             Thomas Schwinge <thomas@codesourcery.com>
10419             Cesar Philippidis  <cesar@codesourcery.com>
10421         * gimplify.c (gimplify_scan_omp_clauses): Add support for
10422         OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
10423         (gimplify_adjust_omp_clauses): Likewise.
10424         (gimplify_oacc_declare_1): Add support for GOMP_MAP_RELEASE, remove
10425         support for GOMP_MAP_FORCE_{ALLOC,TO,FROM,TOFROM}.
10426         (gimplify_omp_target_update): Update handling of acc update and
10427         enter/exit data.
10428         * omp-low.c (install_var_field): Remove unused parameter
10429         base_pointers_restrict.
10430         (scan_sharing_clauses): Remove base_pointers_restrict parameter.
10431         Update call to install_var_field. Handle OMP_CLAUSE_{IF_PRESENT,
10432         FINALIZE}
10433         (omp_target_base_pointers_restrict_p): Delete.
10434         (scan_omp_target): Update call to scan_sharing_clauses.
10435         * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_{IF_PRESENT,
10436         FINALIZE}.
10437         * tree-nested.c (convert_nonlocal_omp_clauses): Handle
10438         OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
10439         (convert_local_omp_clauses): Likewise.
10440         * tree-pretty-print.c (dump_omp_clause): Likewise.
10441         * tree.c (omp_clause_num_ops): Add entries for  OMP_CLAUSE_{IF_PRESENT,
10442         FINALIZE}.
10443         (omp_clause_code_name): Likewise.
10445 2018-06-20  Jakub Jelinek  <jakub@redhat.com>
10447         PR debug/86194
10448         * var-tracking.c (use_narrower_mode_test): Check if shift amount can
10449         be narrowed.
10451         PR tree-optimization/86231
10452         * tree-vrp.c (union_ranges): For (  [  )  ] or (   )[   ] range and
10453         anti-range don't overwrite *vr0min before using it to compute *vr0max.
10455 2018-06-20  Tom de Vries  <tdevries@suse.de>
10457         PR tree-optimization/86097
10458         * tree-ssa-loop-manip.c (canonicalize_loop_ivs): Also convert *nit to
10459         iv type if signedness of iv type is not the same as that of *nit.
10461 2018-06-20  Jakub Jelinek  <jakub@redhat.com>
10463         * cfgrtl.c (rtl_verify_edges): Formatting fix.  If bb->preds has any
10464         EDGE_EH edges, verify they are all EDGE_EH.
10466 2018-06-20  Maya Rashish  <coypu@sdf.org>
10468         * ginclude/stddef.h: Limit #include <machine/ansi.h> to NetBSD.
10470 2018-06-20  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
10472         * config/aarch64/aarch64-tuning-flags.def (no_ldp_stp_qregs): New.
10473         * config/aarch64/aarch64.c (xgene1_tunings): Add
10474         AARCH64_EXTRA_TUNE_NO_LDP_STP_QREGS to tune_flags.
10475         (aarch64_mode_valid_for_sched_fusion_p):
10476         Allow 16-byte modes.
10477         (aarch64_classify_address): Allow 16-byte modes for load_store_pair_p.
10478         * config/aarch64/aarch64-ldpstp.md: Add peepholes for LDP STP of
10479         128-bit modes.
10480         * config/aarch64/aarch64-simd.md (load_pair<VQ:mode><VQ2:mode>):
10481         New pattern.
10482         (vec_store_pair<VQ:mode><VQ2:mode>): Likewise.
10483         * config/aarch64/iterators.md (VQ2): New mode iterator.
10485 2018-06-20  Martin Liska  <mliska@suse.cz>
10487         * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
10488         Change default ratio from 10 to 8.
10490 2018-06-20  Martin Liska  <mliska@suse.cz>
10492         * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
10493         New.
10494         (bit_test_cluster::find_bit_tests): Likewise.
10495         (switch_decision_tree::analyze_switch_statement): Find clusters.
10496         * tree-switch-conversion.h (struct jump_table_cluster): Document
10497         hierarchy.
10499 2018-06-20  Martin Liska  <mliska@suse.cz>
10501         * tree-switch-conversion.c (switch_conversion::collect):
10502         Record m_uniq property.
10503         (switch_conversion::expand): Bail out for special conditions.
10504         (group_cluster::~group_cluster): New.
10505         (group_cluster::group_cluster): Likewise.
10506         (group_cluster::dump): Likewise.
10507         (jump_table_cluster::emit): New.
10508         (switch_decision_tree::fix_phi_operands_for_edges): New.
10509         (struct case_node): Remove struct.
10510         (jump_table_cluster::can_be_handled): New.
10511         (case_values_threshold): Moved to header.
10512         (reset_out_edges_aux): Likewise.
10513         (jump_table_cluster::is_beneficial): New.
10514         (bit_test_cluster::can_be_handled): Likewise.
10515         (add_case_node): Remove.
10516         (bit_test_cluster::is_beneficial): New.
10517         (case_bit_test::cmp): New.
10518         (bit_test_cluster::emit): New.
10519         (expand_switch_as_decision_tree_p): Remove.
10520         (bit_test_cluster::hoist_edge_and_branch_if_true): New.
10521         (fix_phi_operands_for_edge): Likewise.
10522         (switch_decision_tree::analyze_switch_statement): New.
10523         (compute_cases_per_edge): Move ...
10524         (switch_decision_tree::compute_cases_per_edge): ... here.
10525         (try_switch_expansion): Likewise.
10526         (switch_decision_tree::try_switch_expansion): Likewise.
10527         (record_phi_operand_mapping): Likewise.
10528         (switch_decision_tree::record_phi_operand_mapping): Likewise.
10529         (emit_case_decision_tree): Likewise.
10530         (switch_decision_tree::emit): Likewise.
10531         (balance_case_nodes): Likewise.
10532         (switch_decision_tree::balance_case_nodes): Likewise.
10533         (dump_case_nodes): Likewise.
10534         (switch_decision_tree::dump_case_nodes): Likewise.
10535         (emit_jump): Likewise.
10536         (switch_decision_tree::emit_jump): Likewise.
10537         (emit_cmp_and_jump_insns): Likewise.
10538         (switch_decision_tree::emit_cmp_and_jump_insns): Likewise.
10539         (emit_case_nodes): Likewise.
10540         (switch_decision_tree::emit_case_nodes): Likewise.
10541         (conditional_probability): Remove.
10542         * tree-switch-conversion.h (enum cluster_type): New.
10543         (PRINT_CASE): New.
10544         (struct cluster): Likewise.
10545         (cluster::cluster): Likewise.
10546         (struct simple_cluster): Likewise.
10547         (simple_cluster::simple_cluster): Likewise.
10548         (struct group_cluster): Likewise.
10549         (struct jump_table_cluster): Likewise.
10550         (struct bit_test_cluster): Likewise.
10551         (struct min_cluster_item): Likewise.
10552         (struct case_tree_node): Likewise.
10553         (case_tree_node::case_tree_node): Likewise.
10554         (jump_table_cluster::case_values_threshold): Likewise.
10555         (struct case_bit_test): Likewise.
10556         (struct switch_decision_tree): Likewise.
10557         (struct switch_conversion): Likewise.
10558         (switch_decision_tree::reset_out_edges_aux): Likewise.
10560 2018-06-20  Martin Liska  <mliska@suse.cz>
10562         * tree-switch-conversion.c (MAX_CASE_BIT_TESTS): Remove.
10563         (hoist_edge_and_branch_if_true): Likewise.
10564         (expand_switch_using_bit_tests_p): Likewise.
10565         (struct case_bit_test): Likewise.
10566         (case_bit_test_cmp): Likewise.
10567         (emit_case_bit_tests): Likewise.
10568         (switch_conversion::switch_conversion): New class.
10569         (struct switch_conv_info): Remove old struct.
10570         (collect_switch_conv_info): More to ...
10571         (switch_conversion::collect): ... this.
10572         (check_range): Likewise.
10573         (switch_conversion::check_range): Likewise.
10574         (check_all_empty_except_final): Likewise.
10575         (switch_conversion::check_all_empty_except_final): Likewise.
10576         (check_final_bb): Likewise.
10577         (switch_conversion::check_final_bb): Likewise.
10578         (create_temp_arrays): Likewise.
10579         (switch_conversion::create_temp_arrays): Likewise.
10580         (free_temp_arrays): Likewise.
10581         (gather_default_values): Likewise.
10582         (switch_conversion::gather_default_values): Likewise.
10583         (build_constructors): Likewise.
10584         (switch_conversion::build_constructors): Likewise.
10585         (constructor_contains_same_values_p): Likewise.
10586         (switch_conversion::contains_same_values_p): Likewise.
10587         (array_value_type): Likewise.
10588         (switch_conversion::array_value_type): Likewise.
10589         (build_one_array): Likewise.
10590         (switch_conversion::build_one_array): Likewise.
10591         (build_arrays): Likewise.
10592         (switch_conversion::build_arrays): Likewise.
10593         (gen_def_assigns): Likewise.
10594         (switch_conversion::gen_def_assigns): Likewise.
10595         (prune_bbs): Likewise.
10596         (switch_conversion::prune_bbs): Likewise.
10597         (fix_phi_nodes): Likewise.
10598         (switch_conversion::fix_phi_nodes): Likewise.
10599         (gen_inbound_check): Likewise.
10600         (switch_conversion::gen_inbound_check): Likewise.
10601         (process_switch): Use the newly created class.
10602         (switch_conversion::expand): New.
10603         (switch_conversion::~switch_conversion): New.
10604         * tree-switch-conversion.h: New file.
10606 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
10608         * tree-vectorizer.h (NUM_PATTERNS, vect_recog_func_ptr): Move to
10609         tree-vect-patterns.c.
10610         * tree-vect-patterns.c (vect_supportable_direct_optab_p): New function.
10611         (vect_recog_dot_prod_pattern): Use it.  Remove the type_in argument.
10612         (vect_recog_sad_pattern): Likewise.
10613         (vect_recog_widen_sum_pattern): Likewise.
10614         (vect_recog_pow_pattern): Likewise.  Check for a null vectype.
10615         (vect_recog_widen_shift_pattern): Remove the type_in argument.
10616         (vect_recog_rotate_pattern): Likewise.
10617         (vect_recog_mult_pattern): Likewise.
10618         (vect_recog_vector_vector_shift_pattern): Likewise.
10619         (vect_recog_divmod_pattern): Likewise.
10620         (vect_recog_mixed_size_cond_pattern): Likewise.
10621         (vect_recog_bool_pattern): Likewise.
10622         (vect_recog_mask_conversion_pattern): Likewise.
10623         (vect_try_gather_scatter_pattern): Likewise.
10624         (vect_recog_widen_mult_pattern): Likewise.  Check for a null vectype.
10625         (vect_recog_over_widening_pattern): Likewise.
10626         (vect_recog_gather_scatter_pattern): Likewise.
10627         (vect_recog_func_ptr): Move from tree-vectorizer.h
10628         (vect_vect_recog_func_ptrs): Move further down the file.
10629         (vect_recog_func): Likewise.  Remove the third argument.
10630         (NUM_PATTERNS): Define based on vect_vect_recog_func_ptrs.
10631         (vect_pattern_recog_1): Expect the pattern function to do any
10632         necessary target tests.  Also expect it to provide a vector type.
10633         Remove the type_in handling.
10635 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
10637         * tree-vect-patterns.c (vect_pattern_detected): New function.
10638         (vect_recog_dot_prod_patternm, vect_recog_sad_pattern)
10639         (vect_recog_widen_mult_pattern, vect_recog_widen_sum_pattern)
10640         (vect_recog_over_widening_pattern, vect_recog_widen_shift_pattern
10641         (vect_recog_rotate_pattern, vect_recog_vector_vector_shift_pattern)
10642         (vect_recog_mult_pattern, vect_recog_divmod_pattern)
10643         (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
10644         (vect_recog_mask_conversion_pattern)
10645         (vect_try_gather_scatter_pattern): Likewise.
10647 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
10649         * tree-vect-patterns.c (vect_get_internal_def): New function.
10650         (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
10651         (vect_recog_vector_vector_shift_pattern, check_bool_pattern)
10652         (search_type_for_mask_1): Use it.
10654 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
10656         * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
10657         redundant WIDEN_SUM_EXPR handling.
10658         (vect_recog_sad_pattern): Likewise.
10660 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
10662         * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
10663         redundant check that the types of a PLUS_EXPR or MULT_EXPR agree.
10664         (vect_recog_sad_pattern): Likewise PLUS_EXPR, ABS_EXPR and MINUS_EXPR.
10665         (vect_recog_widen_mult_pattern): Likewise MULT_EXPR.
10666         (vect_recog_widen_sum_pattern): Likewise PLUS_EXPR.
10668 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
10670         * tree-vect-stmts.c (vectorizable_call): Make sure that we
10671         use the stmt_vec_info of the original bb statement for the
10672         new zero assignment, even if the call is part of a pattern.
10674 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
10676         * tree-vectorizer.h (_stmt_vec_info): Note above pattern_def_seq
10677         that the sequence is attached to the original statement rather
10678         than the pattern statement.
10679         * tree-vect-loop.c (vect_determine_vf_for_stmt): Take the
10680         PATTERN_DEF_SEQ from the original statement rather than
10681         the main pattern statement.
10682         * tree-vect-stmts.c (free_stmt_vec_info): Likewise.
10683         * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Likewise.
10684         (vect_mark_pattern_stmts): Don't copy the PATTERN_DEF_SEQ.
10686 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
10688         * tree-vect-stmts.c (vect_analyze_stmt): Move the handling of pattern
10689         definition statements before the early exit for statements that aren't
10690         live or relevant.
10691         * tree-vect-loop.c (vect_transform_loop_stmt): New function,
10692         split out from...
10693         (vect_transform_loop): ...here.  Process pattern definition
10694         statements without first checking whether the main pattern
10695         statement is live or relevant.
10697 2018-06-19  Eric Botcazou  <ebotcazou@adacore.com>
10699         * tree-cfgcleanup.c (tree_forwarder_block_p): Do not return false at
10700         -O0 if the locus represent UNKNOWN_LOCATION but have different values.
10702 2018-06-19  Aaron Sawdey  <acsawdey@linux.ibm.com>
10704         * config/rs6000/rs6000-string.c (select_block_compare_mode): Check
10705         TARGET_EFFICIENT_OVERLAPPING_UNALIGNED here instead of in caller.
10706         (do_and3, do_and3_mask, do_compb3, do_rotl3): New functions.
10707         (expand_block_compare): Change select_block_compare_mode call.
10708         (expand_strncmp_align_check): Use new functions, fix comment.
10709         (emit_final_str_compare_gpr): New function.
10710         (expand_strn_compare): Refactor and clean up code.
10711         * config/rs6000/vsx.md (vsx_mov<mode>_64bit): Remove *.
10713 2018-06-19  Tony Reix  <tony.reix@atos.com>
10714             Damien Bergamini  <damien.bergamini@atos.com>
10715             David Edelsohn  <dje.gcc@gmail.com>
10717         * collect2.c (static_obj): New variable.
10718         (static_libs): New variable.
10719         (is_in_list): Uncomment declaration.
10720         (main): Track AIX libraries linked statically.
10721         (is_in_list): Uncomment definition.
10722         (scan_prog_file): Don't add AIX shared libraries initializer
10723         to constructor list if linking statically.
10725 2018-06-19  Max Filippov  <jcmvbkbc@gmail.com>
10727         * config/xtensa/xtensa.md (UNSPEC_FRAME_BLOCKAGE): New unspec
10728         constant.
10729         (allocate_stack, frame_blockage, *frame_blockage): New patterns.
10731 2018-06-19  Jan Hubicka  <hubicka@ucw.cz>
10733         * tree.c (find_decls_types_r): Remove all non-VAR_DECLs from
10734         blocks.
10736 2018-06-19  Martin Liska  <mliska@suse.cz>
10738         * config/i386/i386.c (ix86_can_inline_p): Do not use
10739         ipa_fn_summaries::get_create.
10740         * ipa-cp.c (ipcp_cloning_candidate_p): Replace get_create with
10741         get.
10742         (devirtualization_time_bonus): Likewise.
10743         (ipcp_propagate_stage): Likewise.
10744         * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
10745         (edge_set_predicate): Likewise.
10746         (evaluate_conditions_for_known_args): Likewise.
10747         (evaluate_properties_for_edge): Likewise.
10748         (ipa_call_summary::reset): Tranform to ...
10749         (ipa_call_summary::~ipa_call_summary): ... this.
10750         (ipa_fn_summary::reset): Transform to ...
10751         (ipa_fn_summary::~ipa_fn_summary): ... this.
10752         (ipa_fn_summary_t::remove): Rename to ...
10753         (ipa_fn_summary_t::remove_callees): ... this.
10754         (ipa_fn_summary_t::duplicate): Use placement new
10755         instead of memory copy.
10756         (ipa_call_summary_t::duplicate): Likewise.
10757         (ipa_call_summary_t::remove): Remove.
10758         (dump_ipa_call_summary): Change get_create to get.
10759         (ipa_dump_fn_summary): Dump only when summary exists.
10760         (analyze_function_body): Use symbol_summary::get instead
10761         of get_create.
10762         (compute_fn_summary): Likewise.
10763         (estimate_edge_devirt_benefit): Likewise.
10764         (estimate_edge_size_and_time): Likewise.
10765         (inline_update_callee_summaries): Likewise.
10766         (remap_edge_change_prob): Likewise.
10767         (remap_edge_summaries): Likewise.
10768         (ipa_merge_fn_summary_after_inlining): Likewise.
10769         (write_ipa_call_summary): Likewise.
10770         (ipa_fn_summary_write): Likewise.
10771         (ipa_free_fn_summary): Likewise.
10772         * ipa-fnsummary.h (struct GTY): Add new ctor and copy ctor.
10773         (struct ipa_call_summary): Likewise.
10774         * ipa-icf.c (sem_function::merge): Use symbol_summary::get instead
10775         of get_create.
10776         * ipa-inline-analysis.c (do_estimate_edge_time): Likewise.
10777         (estimate_size_after_inlining): Likewise.
10778         (estimate_growth): Likewise.
10779         (growth_likely_positive): Likewise.
10780         * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
10781         (inline_call): Likewise.
10782         * ipa-inline.c (caller_growth_limits): Likewise.
10783         (can_inline_edge_p): Likewise.
10784         (can_inline_edge_by_limits_p): Likewise.
10785         (compute_uninlined_call_time): Likewise.
10786         (compute_inlined_call_time): Likewise.
10787         (want_inline_small_function_p): Likewise.
10788         (edge_badness): Likewise.
10789         (update_caller_keys): Likewise.
10790         (update_callee_keys): Likewise.
10791         (inline_small_functions): Likewise.
10792         (inline_to_all_callers_1): Likewise.
10793         (dump_overall_stats): Likewise.
10794         (early_inline_small_functions): Likewise.
10795         (early_inliner): Likewise.
10796         * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
10797         * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
10798         * ipa-pure-const.c (malloc_candidate_p): Likewise.
10799         * ipa-split.c (execute_split_functions): Likewise.
10800         * symbol-summary.h: Likewise.
10801         * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
10803 2018-06-19  Richard Biener  <rguenther@suse.de>
10805         * tree-vectorizer.c (try_vectorize_loop_1): Split out of ...
10806         (vectorize_loops): ... here.  Fix dbgcnt handling.
10807         (try_vectorize_loop): Wrap try_vectorize_loop_1.
10809 2018-06-19  Segher Boessenkool  <segher@kernel.crashing.org>
10811         PR target/86197
10812         * config/rs6000/rs6000.md (rs6000_discover_homogeneous_aggregate): An
10813         ieee128 argument takes up only one (vector) register, not two (floating
10814         point) registers.
10816 2018-06-19  Eric Botcazou  <ebotcazou@adacore.com>
10818         * gimplify.c (gimplify_init_constructor): Really never clear for an
10819         incomplete constructor if CONSTRUCTOR_NO_CLEARING is set.
10821 2018-06-19  Richard Biener  <rguenther@suse.de>
10823         PR tree-optimization/86179
10824         * tree-vect-patterns.c (vect_pattern_recog_1): Clean up
10825         after failed recognition.
10827 2018-06-18  Martin Sebor  <msebor@redhat.com>
10829         PR middle-end/85602
10830         * calls.c (maybe_warn_nonstring_arg): Handle strncat.
10831         * tree-ssa-strlen.c (is_strlen_related_p): Make extern.
10832         Handle integer subtraction.
10833         (maybe_diag_stxncpy_trunc): Handle nonstring source arguments.
10834         * tree-ssa-strlen.h (is_strlen_related_p): Declare.
10836 2018-06-18  David Malcolm  <dmalcolm@redhat.com>
10838         * config/frv/frv-protos.h (frv_ifcvt_modify_insn): Strengthen 3rd
10839         param from rtx to rtx_insn *.
10840         * config/frv/frv.c (frv_ifcvt_add_insn): Likewise for "insn"
10841         param.
10842         (frv_ifcvt_modify_insn): Likwise.
10843         (frv_ifcvt_modify_final): Likwise for local "existing_insn",
10844         adding an as_a <rtx_insn *> cast.  Likewise for local "insn".
10845         * config/mips/mips.c (r10k_insert_cache_barriers): Add an
10846         as_a <rtx_insn *> cast to local "unprotected_region" once
10847         it's been established that it's not NULL or pc_rtx.
10848         * config/nds32/nds32-relax-opt.c (nds32_group_insns): Strengthen
10849         param "sethi" from rtx to rtx_insn *.
10850         (nds32_group_float_insns): Likewise for param "insn".
10851         * config/vax/vax-protos.h (vax_output_int_add): Likewise for 1st
10852         param.
10853         (vax_output_int_subtract): Likewise.
10854         * config/vax/vax.c (vax_output_int_add): Likewise for param
10855         "insn".
10856         (vax_output_int_subtract): Likewise.
10857         * emit-rtl.c (set_insn_deleted): Likewise, removing cast.
10858         (emit_pattern_after): Likewise for param "after".
10859         (emit_insn_after): Likewise.
10860         (emit_jump_insn_after): Likewise.
10861         (emit_call_insn_after): Likewise.
10862         (emit_debug_insn_after): Likewise.
10863         (emit_pattern_before): Likewise for param "before".
10864         (emit_insn_before): Likewise.
10865         (emit_jump_insn_before): Likewise.
10866         * final.c (get_insn_template): Likewise for param "insn", removing
10867         a cast.
10868         * output.h (get_insn_template): Likewise for 2nd param.
10869         * rtl.h (emit_insn_before): Likewise.
10870         (emit_jump_insn_before): Likewise.
10871         (emit_debug_insn_before_noloc): Likewise.
10872         (emit_insn_after): Likewise.
10873         (emit_jump_insn_after): Likewise.
10874         (emit_call_insn_after): Likewise.
10875         (emit_debug_insn_after): Likewise.
10876         (set_insn_deleted): Likewise for param.
10878 2018-06-18  Michael Meissner  <meissner@linux.ibm.com>
10880         PR target/85358
10881         * config/rs6000/rs6000-modes.def (toplevel): Rework the 128-bit
10882         floating point modes, so that IFmode is numerically greater than
10883         TFmode, which is greater than KFmode using FRACTIONAL_FLOAT_MODE
10884         to declare the ordering.  This prevents IFmode from being
10885         converted to TFmode when long double is IEEE 128-bit on an ISA 3.0
10886         machine.  Include rs6000-modes.h to share the fractional values
10887         between genmodes* and the rest of the compiler.
10888         (IFmode): Likewise.
10889         (KFmode): Likewise.
10890         (TFmode): Likewise.
10891         * config/rs6000/rs6000-modes.h: New file.
10892         * config/rs6000/rs6000.c (rs6000_debug_reg_global): Change the
10893         meaning of rs6000_long_double_size so that 126..128 selects an
10894         appropriate 128-bit floating point type.
10895         (rs6000_option_override_internal): Likewise.
10896         * config/rs6000/rs6000.h (toplevel): Include rs6000-modes.h.
10897         (TARGET_LONG_DOUBLE_128): Change the meaning of
10898         rs6000_long_double_size so that 126..128 selects an appropriate
10899         128-bit floating point type.
10900         (LONG_DOUBLE_TYPE_SIZE): Update comment.
10901         * config/rs6000/rs6000.md (trunciftf2): Correct the modes of the
10902         source and destination to match the standard usage.
10903         (truncifkf2): Likewise.
10904         (copysign<mode>3, IEEE iterator): Rework copysign of float128 on
10905         ISA 2.07 to use an explicit clobber, instead of passing in a
10906         temporary.
10907         (copysign<mode>3_soft): Likewise.
10909 2018-06-18  David Malcolm  <dmalcolm@redhat.com>
10911         * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
10912         Replace dump_printf_loc call with DUMP_VECT_SCOPE.
10913         (vect_slp_analyze_instance_dependence): Likewise.
10914         (vect_enhance_data_refs_alignment): Likewise.
10915         (vect_analyze_data_refs_alignment): Likewise.
10916         (vect_slp_analyze_and_verify_instance_alignment
10917         (vect_analyze_data_ref_accesses): Likewise.
10918         (vect_prune_runtime_alias_test_list): Likewise.
10919         (vect_analyze_data_refs): Likewise.
10920         * tree-vect-loop-manip.c (vect_update_inits_of_drs): Likewise.
10921         * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
10922         (vect_analyze_scalar_cycles_1): Likewise.
10923         (vect_get_loop_niters): Likewise.
10924         (vect_analyze_loop_form_1): Likewise.
10925         (vect_update_vf_for_slp): Likewise.
10926         (vect_analyze_loop_operations): Likewise.
10927         (vect_analyze_loop): Likewise.
10928         (vectorizable_induction): Likewise.
10929         (vect_transform_loop): Likewise.
10930         * tree-vect-patterns.c (vect_pattern_recog): Likewise.
10931         * tree-vect-slp.c (vect_analyze_slp): Likewise.
10932         (vect_make_slp_decision): Likewise.
10933         (vect_detect_hybrid_slp): Likewise.
10934         (vect_slp_analyze_operations): Likewise.
10935         (vect_slp_bb): Likewise.
10936         * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Likewise.
10937         (vectorizable_bswap): Likewise.
10938         (vectorizable_call): Likewise.
10939         (vectorizable_simd_clone_call): Likewise.
10940         (vectorizable_conversion): Likewise.
10941         (vectorizable_assignment): Likewise.
10942         (vectorizable_shift): Likewise.
10943         (vectorizable_operation): Likewise.
10944         * tree-vectorizer.h (DUMP_VECT_SCOPE): New macro.
10946 2018-06-18  Martin Sebor  <msebor@redhat.com>
10948         PR tree-optimization/81384
10949         * builtin-types.def (BT_FN_SIZE_CONST_STRING_SIZE): New.
10950         * builtins.c (expand_builtin_strnlen): New function.
10951         (expand_builtin): Call it.
10952         (fold_builtin_n): Avoid setting TREE_NO_WARNING.
10953         * builtins.def (BUILT_IN_STRNLEN): New.
10954         * calls.c (maybe_warn_nonstring_arg): Handle BUILT_IN_STRNLEN.
10955         Warn for bounds in excess of maximum object size.
10956         * tree-ssa-strlen.c (maybe_set_strlen_range): Return tree representing
10957         single-value ranges.  Handle strnlen.
10958         (handle_builtin_strlen): Handle strnlen.
10959         (strlen_check_and_optimize_stmt): Same.
10960         * doc/extend.texi (Other Builtins): Document strnlen.
10962 2018-06-18  Maya Rashish  <coypu@sdf.org>
10964         * config/alpha/openbsd.h (TARGET_DEFAULT): Define.
10965         (LINK_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): Likewise.
10966         (INTMAX_TYPE, UINTMAX_TYPE, WINT_TYPE): Likewise.
10968         * config/alpha/elf.h (STARTFILE_SPEC, ENDFILE_SPEC): Move from
10969         here to ...
10970         * config/alpha/linux.h (STARTFILE_SPEC, ENDFILE_SPEC): Here.
10972 2018-06-18  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
10974         * tree.c (escaped_string::escape): Replace cast to char * by
10975         const_cast<char *> (unescaped).
10977 2018-06-18  Nick Clifton  <nickc@redhat.com>
10979         PR 84195
10980         * tree.c (escaped_string): New class.  Converts an unescaped
10981         string into its escaped equivalent.
10982         (warn_deprecated_use): Use the new class to convert the
10983         deprecation message, if present.
10984         (test_escaped_strings): New self test.
10985         (test_c_tests): Add test_escaped_strings.
10986         * doc/extend.texi (deprecated): Add a note that the
10987         deprecation message is affected by the -fmessage-length
10988         option, and that control characters will be escaped.
10989         (#pragma GCC error): Document this pragma.
10990         (#pragma GCC warning): Likewise.
10991         * doc/invoke.texi (-fmessage-length): Document this option's
10992         effect on the #warning and #error preprocessor directives and
10993         the deprecated attribute.
10995 2018-06-18  Eric Botcazou  <ebotcazou@adacore.com>
10997         * tree.c (decl_value_expr_lookup): Revert latest change.
10998         (decl_value_expr_insert): Likewise.
11000 2018-06-17  Eric Botcazou  <ebotcazou@adacore.com>
11002         * gimplify.c (nonlocal_vlas): Delete.
11003         (nonlocal_vla_vars): Likewise.
11004         (gimplify_var_or_parm_decl): Do not add debug VAR_DECLs for non-local
11005         referenced VLAs.
11006         (gimplify_body): Do not create and destroy nonlocal_vlas.
11007         * tree-nested.c: Include diagnostic.h.
11008         (use_pointer_in_frame): Tweak.
11009         (lookup_field_for_decl): Add assertion and declare the transformation.
11010         (convert_nonlocal_reference_op) <PARM_DECL>: Rework and issue an
11011         internal error when the reference is in a wrong context.  Do not
11012         create a debug decl by default.
11013         (note_nonlocal_block_vlas): Delete.
11014         (convert_nonlocal_reference_stmt) <GIMPLE_BIND>: Do not call it.
11015         (convert_local_reference_op) <PARM_DECL>: Skip the frame decl.  Do not
11016         create a debug decl by default.
11017         (convert_gimple_call) <GIMPLE_CALL>: Issue an internal error when the
11018         call is in a wrong context.
11019         (fixup_vla_decls): New function.
11020         (finalize_nesting_tree_1): Adjust comment.  Call fixup_vla_decls if no
11021         debug variables were created.
11022         * tree.c (decl_value_expr_lookup): Add checking assertion.
11023         (decl_value_expr_insert): Likewise.
11025 2018-06-16  Kugan Vivekanandarajah  <kuganv@linaro.org>
11027         PR middle-end/82479
11028         * ipa-fnsummary.c (will_be_nonconstant_expr_predicate): Handle CALL_EXPR.
11029         * tree-scalar-evolution.c (interpret_expr): Likewise.
11030         (expression_expensive_p): Likewise.
11031         * tree-ssa-loop-ivopts.c (contains_abnormal_ssa_name_p): Likewise.
11032         * tree-ssa-loop-niter.c (number_of_iterations_popcount): New.
11033         (number_of_iterations_exit_assumptions): Use number_of_iterations_popcount.
11034         (ssa_defined_by_minus_one_stmt_p): New.
11036 2018-06-16  Kugan Vivekanandarajah  <kuganv@linaro.org>
11038         PR middle-end/64946
11039         * cfgexpand.c (expand_debug_expr): Hande ABSU_EXPR.
11040         * config/i386/i386.c (ix86_add_stmt_cost): Likewise.
11041         * dojump.c (do_jump): Likewise.
11042         * expr.c (expand_expr_real_2): Check operand type's sign.
11043         * fold-const.c (const_unop): Handle ABSU_EXPR.
11044         (fold_abs_const): Likewise.
11045         * gimple-pretty-print.c (dump_unary_rhs): Likewise.
11046         * gimple-ssa-backprop.c (backprop::process_assign_use): Likesie.
11047         (strip_sign_op_1): Likesise.
11048         * match.pd: Add new pattern to generate ABSU_EXPR.
11049         * optabs-tree.c (optab_for_tree_code): Handle ABSU_EXPR.
11050         * tree-cfg.c (verify_gimple_assign_unary): Likewise.
11051         * tree-eh.c (operation_could_trap_helper_p): Likewise.
11052         * tree-inline.c (estimate_operator_cost): Likewise.
11053         * tree-pretty-print.c (dump_generic_node): Likewise.
11054         * tree-vect-patterns.c (vect_recog_sad_pattern): Likewise.
11055         * tree.def (ABSU_EXPR): New.
11057 2018-06-16  Jakub Jelinek  <jakub@redhat.com>
11059         PR middle-end/86095
11060         * common.opt (Wunsafe-loop-optimizations): Add Ignore, remove Var,
11061         documented as preserved for backward compatibility only.
11062         * doc/invoke.texi: Remove -Wunsafe-loop-optimizations documentation.
11064         PR rtl-optimization/86108
11065         * bb-reorder.c (create_forwarder_block): Renamed to ...
11066         (create_eh_forwarder_block): ... this.  Split OLD_BB after labels and
11067         jump from new landing pad to the second part.
11068         (sjlj_fix_up_crossing_landing_pad, dw2_fix_up_crossing_landing_pad):
11069         Adjust callers.
11071 2018-06-15  Jakub Jelinek  <jakub@redhat.com>
11073         PR middle-end/85878
11074         * expr.c (expand_assignment): Remove now redundant COMPLEX_MODE_P
11075         check from first store_expr, use to_mode instead of GET_MODE (to_rtx).
11076         Only call store_expr for halves if the mode is the same.
11078         PR middle-end/86123
11079         * match.pd ((X / Y) == 0 -> X < Y): Don't transform complex divisions.
11080         Fix up comment formatting.
11082 2018-06-15  Bernd Edlinger  <bernd.edlinger@hotmail.de>
11084         * typed-splay-tree.h (typed_splay_tree::remove): New function.
11085         (typed_splay_tree::closure,
11086         typed_splay_tree::inner_foreach_fn, typed_splay_tree::m_inner): Deleted.
11087         (typed_splay_tree::typed_splay_tree,
11088         typed_splay_tree::operator =): Declared private.
11089         (typed_splay_tree::splay_tree_key, typed_splay_tree::splay_tree_value,
11090         typed_splay_tree::splay_tree_node_s, typed_splay_tree::KDEL,
11091         typed_splay_tree::VDEL, typed_splay_tree::splay_tree_delete_helper,
11092         typed_splay_tree::rotate_left, typed_splay_tree::rotate_right,
11093         typed_splay_tree::splay_tree_splay,
11094         typed_splay_tree::splay_tree_foreach_helper,
11095         typed_splay_tree::splay_tree_insert,
11096         typed_splay_tree::splay_tree_remove,
11097         typed_splay_tree::splay_tree_lookup,
11098         typed_splay_tree::splay_tree_predecessor,
11099         typed_splay_tree::splay_tree_successor,
11100         typed_splay_tree::splay_tree_min,
11101         typed_splay_tree::splay_tree_max): Took over from splay-tree.c/.h.
11102         (typed_splay_tree::root, typed_splay_tree::comp,
11103         typed_splay_tree::delete_key,
11104         typed_splay_tree::delete_value): New data members.
11105         * typed-splay-tree.c (selftest::test_str_to_int): Add a test for
11106         typed_splay_tree::remove.
11108 2018-06-15  Matthew Fortune  <matthew.fortune@mips.com>
11110         * config/mips/mips.h (ASM_SPEC): Pass through -mcrc, -mno-crc,
11111         -mginv and -mno-ginv to the assembler.
11112         * config/mips/mips.opt (-mcrc): New option.
11113         (-mginv): Likewise.
11114         * doc/invoke.text (-mcrc): Document.
11115         (-mginv): Likewise.
11117 2018-06-15  Nick Clifton  <nickc@redhat.com>
11119         PR 84195
11120         * tree.c (escaped_string): New class.  Converts an unescaped
11121         string into its escaped equivalent.
11122         (warn_deprecated_use): Use the new class to convert the
11123         deprecation message, if present.
11124         (test_escaped_strings): New self test.
11125         (test_c_tests): Add test_escaped_strings.
11126         * doc/extend.texi (deprecated): Add a note that the
11127         deprecation message is affected by the -fmessage-length
11128         option, and that control characters will be escaped.
11129         (#pragma GCC error): Document this pragma.
11130         (#pragma GCC warning): Likewise.
11131         * doc/invoke.texi (-fmessage-length): Document this option's
11132         effect on the #warning and #error preprocessor directives and
11133         the deprecated attribute.
11135 2018-06-15  Richard Biener  <rguenther@suse.de>
11137         * tree-vect-slp.c (vect_slp_bb): Dump MSG_OPTIMIZED_LOCATIONS
11138         here, also noting vector size used.
11139         * tree-vectorizer.c (vectorize_loops): Adjust.  Note vector
11140         size used in MSG_OPTIMIZED_LOCATIONS dump.
11141         (pass_slp_vectorize::execute): Adjust.
11143 2018-06-15  Claudiu Zissulescu  <claziss@synopsys.com>
11145         PR target/85968
11146         * config/arc/arc.c (arc_return_address_register): Fix
11147         if-condition.
11149 2018-06-15  Richard Biener  <rguenther@suse.de>
11151         PR middle-end/86159
11152         * tree-cfg.c (gimplify_build3): Do not strip sign conversions,
11153         leave useless conversion stripping to force_gimple_operand_gsi.
11154         (gimplify_build2): Likewise.
11155         (gimplify_build1): Likewise.
11157 2018-06-15  Richard Biener  <rguenther@suse.de>
11159         PR middle-end/86076
11160         * tree-cfg.c (move_stmt_op): unshare invariant addresses
11161         before adjusting their block.
11163 2018-06-15  Sebastian Huber  <sebastian.huber@embedded-brains.de>
11165         * config.gcc (riscv*-*-elf* | riscv*-*-rtems*): Use custom
11166         multilibs for *-*-rtems*.
11167         * config/riscv/t-rtems: New file.
11169 2018-06-14  Jakub Jelinek  <jakub@redhat.com>
11171         PR middle-end/86122
11172         * match.pd ((A +- CST1) +- CST2): Punt if last resort
11173         unsigned_type_for returns NULL.
11175         PR target/85945
11176         * lower-subreg.c (find_decomposable_subregs): Don't decompose float
11177         subregs of multi-word pseudos unless the float mode has word size.
11179 2018-06-14  Richard Biener  <rguenther@suse.de>
11181         PR middle-end/86139
11182         * tree-vect-generic.c (build_word_mode_vector_type): Remove
11183         duplicate and harmful type_hash_canon.
11184         * tree.c (type_hash_canon): Assert we didn't find ourselves.
11186 2018-06-14  Richard Biener  <rguenther@suse.de>
11188         PR ipa/86124
11189         * tree-ssa-struct-alias.c (create_variable_info_for): Handle
11190         NULL cgraph_node.
11192 2018-06-14  Sebastian Huber  <sebastian.huber@embedded-brains.de>
11194         * config/rtems.h (STDINT_LONG32): Define.
11196 2018-06-13  Matthew Fortune  <matthew.fortune@mips.com>
11197             Prachi Godbole  <prachi.godbole@imgtec.com>
11199         * config/mips/mips-cpus.def: Define P6600.
11200         * config/mips/mips-tables.opt: Regenerate.
11201         * config/mips/mips.c (mips_ucbranch_type): New enum.
11202         (mips_rtx_cost_data): Add support for P6600.
11203         (mips_issue_rate): Likewise.
11204         (mips_multipass_dfa_lookahead): Likewise.
11205         (mips_avoid_hazard): Likewise.
11206         (mips_reorg_process_insns): Likewise.
11207         (mips_classify_branch_p6600): New function.
11208         * config/mips/mips.h (TUNE_P6600): New define.
11209         (MIPS_ISA_LEVEL_SPEC): Infer mips64r6 from p6600.
11210         (ENABLE_LD_ST_PAIRS): Enable load/store bonding for p6600.
11211         * config/mips/mips.md: Include p6600.md.
11212         (processor): Add p6600.
11213         * config/mips/p6600.md: New file.
11214         * doc/invoke.texi: Add p6600 to supported architectures.
11216 2018-06-13  Martin Sebor  <msebor@redhat.com>
11218         PR tree-optimization/86114
11219         * gimple-fold.c (gimple_fold_builtin_strlen): Only handle LHS
11220         of integer types.
11221         * tree-ssa-strlen.c (maybe_set_strlen_range): Same.
11223 2018-06-13  Richard Biener  <rguenther@suse.de>
11225         * tree-vect-patterns.c (vect_recog_vector_vector_shift_pattern):
11226         Properly set vector type of the intermediate stmt.
11227         * tree-vect-stmts.c (vectorizable_operation): The destination
11228         var always has vectype_out type.
11230 2018-06-13  Jeff Law  <law@redhat.com>
11232         * config/rl78/rl78.c (move_elim_pass): Use TDF_NONE rather than
11233         integer 0 for argument to print_rtl_with_bb.
11234         (rl78_reorg): Likewise.
11236 2018-06-13  David Malcolm  <dmalcolm@redhat.com>
11238         * config/arc/arc.c (hwloop_optimize): Strengthen local "end_label"
11239         from rtx to rtx_insn *.
11240         * config/bfin/bfin.c (hwloop_optimize): Likewise for local
11241         "label".
11242         (add_sched_insns_for_speculation): Likewise for local "target",
11243         converting usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
11244         * config/c6x/c6x.c (reorg_split_calls): Strengthen param "call_labels"
11245         from rtx_insn ** to rtx_code_label **.
11246         (reorg_emit_nops): Likewise.
11247         (c6x_reorg): Likewise for local "call_labels".
11248         * config/sh/sh-protos.h (get_dest_uid): Strengthen 1st param from
11249         rtx to rtx_insn *.
11250         * config/sh/sh.c (dump_table): Strengthen local "lab" from rtx to
11251         rtx_code_label *, adding safe_as_a <rtx_code_label *> casts to
11252         the loops over LABEL_REFS.
11253         (fixup_addr_diff_vecs): Add as_a <rtx_insn *> to usage of
11254         braf_label.
11255         (barrier_align): Convert usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
11256         (get_dest_uid): Strengthen param "label" from rtx to rtx_insn *.
11257         (split_branches): Strengthen local "olabel" from rtx to
11258         rtx_insn *, adding a safe_as_a cast.
11259         * emit-rtl.c (next_real_insn): Strengthen param from "rtx"
11260         to "rtx_insn *".
11261         (add_insn_after): Likewise for first two params.
11262         (add_insn_before): Likewise.
11263         (remove_insn): Likewise for param.
11264         (emit_pattern_before_noloc): Likewise for second and third params.
11265         (emit_jump_insn_before_noloc): Convert NULL_RTX to NULL.
11266         (emit_call_insn_before_noloc): Likewise.
11267         (emit_debug_insn_before_noloc): Strengthen "before" param from "rtx"
11268         to "rtx_insn *".
11269         (emit_barrier_before): Likewise.
11270         (emit_label_before): Strengthen "label" param from "rtx" to
11271         "rtx_code_label *".  Strengthen "before" param from "rtx" to
11272         "rtx_insn *".
11273         (emit_insn_after_1): Strengthen "after" param from "rtx" to
11274         "rtx_insn *".
11275         (emit_pattern_after_noloc): Likewise.
11276         (emit_insn_after_noloc): Likewise.
11277         (emit_jump_insn_after_noloc): Likewise.
11278         (emit_call_insn_after_noloc): Likewise.
11279         (emit_debug_insn_after_noloc): Likewise.
11280         (emit_barrier_after): Likewise.
11281         (emit_label_after): Likewise for both params.
11282         (emit_pattern_after_setloc): Likewise for "after" param.  Convert
11283         "loc" param from "int" to "location_t".
11284         (emit_insn_after_setloc): Likewise.
11285         (emit_jump_insn_after_setloc): Likewise.
11286         (emit_call_insn_after_setloc): Likewise.
11287         (emit_debug_insn_after_setloc): Likewise.
11288         (emit_pattern_before_setloc): Likewise for "before" param.  Convert
11289         "loc" param from "int" to "location_t".
11290         (emit_pattern_before): Convert NULL_RTX to NULL.
11291         (emit_insn_before_setloc): Convert "loc" param from "int" to
11292         "location_t".
11293         (emit_jump_insn_before_setloc): Likewise.
11294         (emit_call_insn_before_setloc): Likewise.
11295         (emit_debug_insn_before_setloc): Strengthen "before" param from rtx to
11296         rtx_insn *.  Convert "loc" param from "int" to "location_t".
11297         * rtl.h (emit_insn_before_setloc, emit_jump_insn_before_setloc,
11298         emit_call_insn_before_setloc, emit_debug_insn_before_setloc):
11299         Convert 3rd param from "int" to "location_t".
11300         (emit_barrier_before, emit_barrier_after, next_real_insn):
11301         Strengthen param from rtx to rtx_insn *.
11302         (emit_label_before): Strengthen 1st param from "rtx" to
11303         "rtx_code_label *".  Strengthen 2nd param from "rtx" to
11304         "rtx_insn *".
11305         (emit_insn_after_noloc, emit_jump_insn_after_noloc,
11306         emit_call_insn_after_noloc, emit_debug_insn_after_noloc):
11307         Strengthen 2nd param from "rtx" to "rtx_insn *".
11308         (emit_insn_after_setloc, emit_jump_insn_after_setloc)
11309         emit_call_insn_after_setloc, emit_debug_insn_after_setloc):
11310         Likewise. Convert 3rd param from "int" to "location_t".
11311         (emit_label_after): Strengthen 1st param from "rtx" to
11312         "rtx_code_label *".
11313         (next_real_insn, remove_insn): Strengthen param from "rtx" to
11314         "rtx_insn *".
11315         (add_insn_before, add_insn_after): Strengthen 1st and 2nd params
11316         from "rtx" to "rtx_insn *".
11318 2018-06-13  Jan Hubicka  <hubicka@gcc.gnu.org>
11320         * cgraph.c (cgraph_node::get_untransformed_body): Dump function
11321         bodies streamed in with -Q.
11322         * dumpfile.c (dump_files): Add lto-stream-out dump file.
11323         * dumpfile.h (tree_dump_index): Add lto_stream_out.
11324         * gimple-streamer-out.c: Include gimple-pretty-print.h
11325         (output_bb): Dump stmts streamed.
11326         * lto-section-out.c: Include print-tree.h
11327         (lto_begin_section): Dump sections created.
11328         (lto_output_decl_index): Dump decl encoded.
11329         * lto-streamer-out.c: Include print-tree.h
11330         (create_output_block): Dump output block created.
11331         (DFS::DFS_write_tree_body): Dump DFS SCCs streamed.
11332         (output_function): Dump function output.
11333         (output_constructor): Dump constructor streamed.
11334         (write_global_stream): Output indexes encoded.
11335         (produce_asm_for_decls): Dump streams encoded.
11336         * lto-streamer.c (streamer_dump_file): New global var.
11337         * lto-streamer.h (streamer_dump_file): Declare.
11338         * passes.c (ipa_write_summaries): Initialize streamer dump.
11339         * varpool.c (varpool_node::get_constructor): Dump constructors streamed
11340         in.
11342 2018-06-13  Eric Botcazou  <ebotcazou@adacore.com>
11344         PR target/86048
11345         * config/i386/winnt.c (i386_pe_seh_cold_init): Do not emit negative
11346         offsets for register save directives.  Emit a second batch of save
11347         directives, if need be, when the function accesses prior frames.
11349 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
11351         * config/arc/fpu.md (fmasf4): Force operand to register.
11352         (fnmasf4): Likewise.
11354 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
11356         * config/arc/arc-protos.h (arc_pad_return): Remove.
11357         * config/arc/arc.c (machine_function): Remove force_short_suffix
11358         and size_reason.
11359         (arc_print_operand): Adjust printing of '&'.
11360         (arc_verify_short): Remove conditional printing of short suffix.
11361         (arc_final_prescan_insn): Remove reference to size_reason.
11362         (pad_return): New function.
11363         (arc_reorg): Call pad_return.
11364         (arc_pad_return): Remove.
11365         (arc_init_machine_status): Remove reference to force_short_suffix.
11366         * config/arc/arc.md (vunspec): Add VUNSPEC_ARC_BLOCKAGE.
11367         (attr length): When attribute iscompact is true force to 2
11368         regardless; in the case of maybe check if we want to force the
11369         instruction to have 4 bytes length.
11370         (nopv): Change it to generate 4 byte long nop as well.
11371         (blockage): New pattern.
11372         (simple_return): Remove call to arc_pad_return.
11373         (p_return_i): Likewise.
11375 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
11377         * config/arc/elf.h (LINK_GCC_C_SEQUENCE_SPEC): Define.
11379 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
11381         * config/arc/builtins.def (SYNC): SYNC instruction is valid on all
11382         ARC cores.
11384 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
11386         * config/arc/arc.c (atomic_exchangesi): EX instruction is default
11387         for ARC700 and ARCv2.
11389 2018-06-13  Chenghua Xu <paul.hua.gm@gmail.com>
11391         PR target/86076
11392         * config/mips/loongson.md (vec_setv4hi): Gen_lowpart for
11393         operands[2] instead of operands[1].
11396 2018-06-12  Richard Sandiford  <richard.sandiford@linaro.org>
11398         * lra-constraints.c (simplify_operand_subreg): In the paradoxical
11399         case, check whether the outer register overlaps an unallocatable
11400         register, not just whether it fits the required class.
11402 2018-06-12  Richard Sandiford  <richard.sandiford@linaro.org>
11404         * poly-int.h (can_div_trunc_p): Add new overload in which all values
11405         are poly_ints.
11406         * alias.c (get_addr): Extend CONST_INT handling to poly_int_rtx_p.
11407         (memrefs_conflict_p): Likewise.
11408         (init_alias_analysis): Likewise.
11409         * cfgexpand.c (expand_debug_expr): Likewise.
11410         * combine.c (combine_simplify_rtx, force_int_to_mode): Likewise.
11411         * cse.c (fold_rtx): Likewise.
11412         * explow.c (adjust_stack, anti_adjust_stack): Likewise.
11413         * expr.c (emit_block_move_hints): Likewise.
11414         (clear_storage_hints, push_block, emit_push_insn): Likewise.
11415         (store_expr_with_bounds, reduce_to_bit_field_precision): Likewise.
11416         (emit_group_load_1): Use rtx_to_poly_int64 for group offsets.
11417         (emit_group_store): Likewise.
11418         (find_args_size_adjust): Use strip_offset.  Use rtx_to_poly_int64
11419         to read the PRE/POST_MODIFY increment.
11420         * calls.c (store_one_arg): Use strip_offset.
11421         * rtlanal.c (rtx_addr_can_trap_p_1): Extend CONST_INT handling to
11422         poly_int_rtx_p.
11423         (set_noop_p): Use rtx_to_poly_int64 for the elements selected
11424         by a VEC_SELECT.
11425         * simplify-rtx.c (avoid_constant_pool_reference): Use strip_offset.
11426         (simplify_binary_operation_1): Extend CONST_INT handling to
11427         poly_int_rtx_p.
11428         * var-tracking.c (compute_cfa_pointer): Take a poly_int64 rather
11429         than a HOST_WIDE_INT.
11430         (hard_frame_pointer_adjustment): Change from HOST_WIDE_INT to
11431         poly_int64.
11432         (adjust_mems, add_stores): Update accodingly.
11433         (vt_canonicalize_addr): Track polynomial offsets.
11434         (emit_note_insn_var_location): Likewise.
11435         (vt_add_function_parameter): Likewise.
11436         (vt_initialize): Likewise.
11438 2018-06-12  Jeff Law  <law@redhat.com>
11440         * config.gcc (alpha*-*-freebsd*): Remove.
11441         * config/alpha/freebsd.h: Remove.
11443 2018-06-12  David Malcolm  <dmalcolm@redhat.com>
11445         PR other/69968
11446         * spellcheck-tree.c (levenshtein_distance): Rename to...
11447         (get_edit_distance): ...this, and update for underlying renaming.
11448         * spellcheck-tree.h (levenshtein_distance): Rename to...
11449         (get_edit_distance): ...this.
11450         * spellcheck.c (levenshtein_distance): Rename to...
11451         (get_edit_distance): ...this.  Convert from Levenshtein distance
11452         to Damerau-Levenshtein distance by supporting transpositions of
11453         adjacent characters.  Rename "v1" to "v_next" and "v0" to
11454         "v_one_ago".
11455         (selftest::levenshtein_distance_unit_test_oneway): Rename to...
11456         (selftest::test_edit_distance_unit_test_oneway): ...this, and
11457         update for underlying renaming.
11458         (selftest::levenshtein_distance_unit_test): Rename to...
11459         (selftest::test_get_edit_distance_unit): ...this, and update for
11460         underlying renaming.
11461         (selftest::test_find_closest_string): Add example from PR 69968
11462         where transposition helps
11463         (selftest::test_metric_conditions): Update for renaming.
11464         (selftest::test_metric_conditions): Likewise.
11465         (selftest::spellcheck_c_tests): Likewise.
11466         * spellcheck.h (levenshtein_distance): Rename both overloads to...
11467         (get_edit_distance): ...this.
11468         (best_match::consider): Update for renaming.
11470 2018-06-12  Martin Sebor  <msebor@redhat.com>
11472         PR tree-optimization/85259
11473         * builtins.c (compute_objsize): Handle constant offsets.
11474         * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Return
11475         true iff a warning has been issued.
11476         * gimple.h (gimple_nonartificial_location): New function.
11477         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Call
11478         gimple_nonartificial_location and handle -Wno-system-headers.
11479         (handle_builtin_stxncpy): Same.
11481 2018-06-12  Martin Sebor  <msebor@redhat.com>
11483         PR c/85931
11484         * fold-const.c (operand_equal_p): Handle SAVE_EXPR.
11486 2018-06-12  Will Schmidt  <will_schmidt@vnet.ibm.com>
11488         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
11489         BUILTIN_VEC_XST entries for pointer to double and long long.
11491 2018-06-12  H.J. Lu  <hongjiu.lu@intel.com>
11493         PR target/85990
11494         * config/i386/gnu-user.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
11495         Update comments.
11496         * config/i386/gnu-user64.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
11497         Likewise.
11499 2018-06-12  Martin Liska  <mliska@suse.cz>
11501         * doc/options.texi: Document IntegerRange.
11503 2018-06-12  Martin Liska  <mliska@suse.cz>
11505         * config/i386/i386.opt: Make MPX-related options as Deprecated.
11506         * opt-functions.awk: Handle Deprecated flag.
11507         * opts-common.c (decode_cmdline_option): Handle cl_deprecated
11508         and report error.
11509         (read_cmdline_option): Report warning for a deprecated option.
11510         * opts.h (struct cl_option): Add new field cl_deprecated.
11511         (CL_ERR_DEPRECATED): New.
11513 2018-06-12  Martin Liska  <mliska@suse.cz>
11515         * doc/options.texi: Document Deprecated option flag.
11517 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
11519         * config/arc/arc-arch.h (arc_extras): New enum.
11520         (arc_cpu_t):Add field extra.
11521         (arc_cpu_types): Consider the extras.
11522         * config/arc/arc-cpus.def: Add extras info.
11523         * config/arc/arc-opts.h (processor_type): Consider extra field.
11524         * config/arc/arc.c (arc_override_options): Handle extra field.
11526 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
11528         * config/arc/arc-arch.h: Update ARC_OPTX macro.
11529         * config/arc/arc-options.def (ARC_OPTX): Introduce a new doc
11530         field.
11531         * config/arc/arc.c (arc_init): Update pic warning.
11532         (irq_range): Update irq range parsing warnings.
11533         (arc_override_options): Update various warning messages.
11534         (arc_handle_aux_attribute): Likewise.
11536 2018-06-12  Robert Suchanek  <robert.suchanek@mips.com>
11538         * config/mips/i6400.md (i6400_fpu_fadd): Remove frint.
11540 2018-06-12  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
11542         * doc/sourcebuild.texi: Document usage of line number 0 in verify
11543         compiler messages directives.
11545 2018-06-12  Matthew Fortune  <mfortune@gmail.com>
11547         * config/mips/mips-cpus.def: New MIPS_CPU for i6500.
11548         * config/mips/mips-tables.opt: Regenerate.
11549         * config/mips/mips.h (MIPS_ISA_LEVEL_SPEC): Mark i6500 as
11550         mips64r6.
11551         * doc/invoke.texi: Document -march=i6500.
11553 2018-06-12  Prachi Godbole  <prachi.godbole@imgtec.com>
11555         * config/mips/i6400.md (i6400_gpmuldiv): Remove cpu_unit.
11556         (i6400_gpmul): Add cpu_unit.
11557         (i6400_gpdiv): Likewise.
11558         (i6400_msa_add_d): Update reservations.
11559         (i6400_msa_int_add) Likewise.
11560         (i6400_msa_short_logic3) Likewise.
11561         (i6400_msa_short_logic2) Likewise.
11562         (i6400_msa_short_logic) Likewise.
11563         (i6400_msa_move) Likewise.
11564         (i6400_msa_cmp) Likewise.
11565         (i6400_msa_short_float2) Likewise.
11566         (i6400_msa_div_d) Likewise.
11567         (i6400_msa_long_logic1) Likewise.
11568         (i6400_msa_long_logic2) Likewise.
11569         (i6400_msa_mult) Likewise.
11570         (i6400_msa_long_float2) Likewise.
11571         (i6400_msa_long_float4) Likewise.
11572         (i6400_msa_long_float5) Likewise.
11573         (i6400_msa_long_float8) Likewise.
11574         (i6400_fpu_fadd): Include frint type.
11575         (i6400_fpu_store): New define_insn_reservation.
11576         (i6400_fpu_load): Likewise.
11577         (i6400_fpu_move): Likewise.
11578         (i6400_fpu_fcmp): Likewise.
11579         (i6400_fpu_fmadd): Likewise.
11580         (i6400_int_mult): Include imul3nc type and update reservation.
11581         (i6400_int_div): Include idiv3 type and update reservation.
11582         (i6400_int_load): Update to check type not move_type.
11583         (i6400_int_store): Likewise.
11584         (i6400_int_prefetch): Set zero latency.
11586 2018-06-12  Eric Botcazou  <ebotcazou@adacore.com>
11588         * gcc.c: Document new %@{...} sequence.
11589         (LINK_COMMAND_SPEC): Use it for the -L switches.
11590         (cpp_unique_options): Use it for the -I switches.
11591         (at_file_argbuf): New global variable.
11592         (in_at_file): Likewise.
11593         (alloc_args): Create at_file_argbuf.
11594         (clear_args): Truncate at_file_argbuf.
11595         (store_arg): If in_at_file, push the argument onto at_file_argbuf.
11596         (open_at_file): New function.
11597         (close_at_file): Likewise.
11598         (create_at_file): Delete.
11599         (do_spec_1) <'i'>: Use open_at_file/close_at_file.
11600         <'o'>: Likewise.
11601         <'@'>: New case.
11602         (validate_switches_from_spec): Deal with %@{...} sequence.
11603         (validate_switches): Likewise.
11604         (driver::finalize): Call clear_args.
11606 2018-06-11  Rasmus Villemoes  <rasmus.villemoes@prevas.dk>
11608         * config/vx-common.h (USE_TM_CLONE_REGISTRY): #define to 0.
11610 2018-06-11  Martin Sebor  <msebor@redhat.com>
11612         * doc/invoke.texi (-Wall): List -Wc++17-compat.
11613         (Wno-class-memaccess): Add @opindex.
11614         (Wno-templates, Wno-multiple-inheritance): Same.
11615         (Wno-virtual-inheritance, Wno-namespaces): Same.
11616         (Wno-pedantic, Wno-chkp, Wcoverage-mismatch): Same.
11617         (Wno-format-overflow, Wno-format-truncation): Same.
11618         (Wno-unused-local-typedefs, Walloc-size-larger-than): Same.
11619         (Wno-alloc-size-larger-than, Wframe-larger-than): Same
11620         (Wno-stack-usage, Wno-c++-compat, Wno-c++11-compat): Same.
11621         (Wno-c++14-compat, Wno-c++17-compat, Wno-openmp-simd): Same.
11622         (Wno-unsuffixed-float-constants, Wno-addr-space-convert): Same.
11623         (Wno-misspelled-isr): Same.
11625 2018-06-11  Martin Sebor  <msebor@redhat.com>
11627         * PR tree-optimization/86083
11628         * tree-ssa-strlen.c (handle_char_store): Use tree_expr_nonzero_p.
11630 2018-06-11  Zhouyi Zhou <zhouzhouyi@gmail.com>
11632         * tree-eh.c (lower_eh_constructs_2): Add a comma to comment.
11634 2018-06-11  Segher Boessenkool  <segher@kernel.crashing.org>
11636         PR target/85755
11637         * config/rs6000/rs6000.md (*movdi_internal32): Put constraint modifiers
11638         on the correct operand.
11639         (*movdi_internal64): Ditto.
11641 2018-06-11  Martin Liska  <mliska@suse.cz>
11643         PR tree-optimization/86089
11644         * tree-ssa-strlen.c (get_string_length): Move back removed hunk.
11646 2018-06-11  Julia Koval  <julia.koval@intel.com>
11648         * config/i386/vaesintrin.h (_mm_aesdec_epi128, _mm_aesdeclast_epi128)
11649         _mm_aesenc_epi128, _mm_aesenclast_epi128: Remove.
11650         * config/i386/vpclmulqdqintrin.h (_mm_clmulepi64_epi128): Remove.
11652 2018-06-11  Olivier Hainque  <hainque@adacore.com>
11654         * dwarf2out.c (gen_compile_unit_die): Fallout to DW_LANG_Ada83
11655         for Ada with strict dwarf2.
11657 2018-06-08  Peter Bergner  <bergner@vnet.ibm.com>
11659         PR target/85755
11660         * config/rs6000/rs6000.c (mem_operand_gpr): Enable PRE_INC and PRE_DEC
11661         addresses.
11663 2018-06-08  Jan Hubicka  <hubicka@ucw.cz>
11665         * dumpfile.c (FIRST_ME_AUTO_NUMBERED_DUMP): Bump to 4.
11667 2018-06-08  David Edelsohn  <dje.gcc@gmail.com>
11669         * config/rs6000/rs6000.c (rs6000_passes_ieee128): Protect with #if
11670         TARGET_ELF.
11672 2018-06-08  Martin Liska  <mliska@suse.cz>
11674         * tree-cfg.h (debug_function): Fix argument type to match
11675         implementation.
11677 2018-06-08  Martin Liska  <mliska@suse.cz>
11679         * config/powerpcspe/powerpcspe.c (rs6000_xcoff_visibility):
11680         Remove usage of MPX-related (and removed) fields.
11681         * config/rs6000/rs6000.c (rs6000_xcoff_visibility): Likewise.
11683 2018-06-08  David Malcolm  <dmalcolm@redhat.com>
11685         * cfg.c (debug): Use TDF_NONE rather than 0.
11686         * cfghooks.c (debug): Likewise.
11687         * dumpfile.c (DUMP_FILE_INFO): Likewise; also for OPTGROUP.
11688         (struct dump_option_value_info): Convert to...
11689         (struct kv_pair): ...this template type.
11690         (dump_options): Convert to kv_pair<dump_flags_t>; use TDF_NONE
11691         rather than 0.
11692         (optinfo_verbosity_options): Likewise.
11693         (optgroup_options): Convert to kv_pair<optgroup_flags_t>; use
11694         OPTGROUP_NONE.
11695         (gcc::dump_manager::dump_register): Use optgroup_flags_t rather
11696         than int for "optgroup_flags" param.
11697         (dump_generic_expr_loc): Use dump_flags_t rather than int for
11698         "dump_kind" param.
11699         (dump_dec): Likewise.
11700         (dump_finish): Use TDF_NONE rather than 0.
11701         (gcc::dump_manager::opt_info_enable_passes): Use optgroup_flags_t
11702         rather than int for "optgroup_flags" param.  Use TDF_NONE rather
11703         than 0.  Update for change to option_ptr.
11704         (opt_info_switch_p_1): Convert "optgroup_flags" param from int *
11705         to optgroup_flags_t *.  Use TDF_NONE and OPTGROUP_NONE rather than
11706         0.  Update for changes to optinfo_verbosity_options and
11707         optgroup_options.
11708         (opt_info_switch_p): Convert optgroup_flags from int to
11709         optgroup_flags_t.
11710         (dump_basic_block): Use dump_flags_t rather than int
11711         for "dump_kind" param.
11712         * dumpfile.h (TDF_ADDRESS, TDF_SLIM, TDF_RAW, TDF_DETAILS,
11713         TDF_STATS, TDF_BLOCKS, TDF_VOPS, TDF_LINENO, TDF_UID)
11714         TDF_STMTADDR, TDF_GRAPH, TDF_MEMSYMS, TDF_RHS_ONLY, TDF_ASMNAME,
11715         TDF_EH, TDF_NOUID, TDF_ALIAS, TDF_ENUMERATE_LOCALS, TDF_CSELIB,
11716         TDF_SCEV, TDF_GIMPLE, TDF_FOLDING, MSG_OPTIMIZED_LOCATIONS,
11717         MSG_MISSED_OPTIMIZATION, MSG_NOTE, MSG_ALL, TDF_COMPARE_DEBUG,
11718         TDF_NONE): Convert from macros to...
11719         (enum dump_flag): ...this new enum.
11720         (dump_flags_t): Update to use enum.
11721         (operator|, operator&, operator~, operator|=, operator&=):
11722         Implement for dump_flags_t.
11723         (OPTGROUP_NONE, OPTGROUP_IPA, OPTGROUP_LOOP, OPTGROUP_INLINE,
11724         OPTGROUP_OMP, OPTGROUP_VEC, OPTGROUP_OTHER, OPTGROUP_ALL):
11725         Convert from macros to...
11726         (enum optgroup_flag): ...this new enum.
11727         (optgroup_flags_t): New typedef.
11728         (operator|, operator|=): Implement for optgroup_flags_t.
11729         (struct dump_file_info): Convert field "alt_flags" to
11730         dump_flags_t.  Convert field "optgroup_flags" to
11731         optgroup_flags_t.
11732         (dump_basic_block): Use dump_flags_t rather than int for param.
11733         (dump_generic_expr_loc): Likewise.
11734         (dump_dec): Likewise.
11735         (dump_register): Convert param "optgroup_flags" to
11736         optgroup_flags_t.
11737         (opt_info_enable_passes): Likewise.
11738         * early-remat.c (early_remat::dump_edge_list): Use TDF_NONE rather
11739         than 0.
11740         * gimple-pretty-print.c (debug): Likewise.
11741         * gimple-ssa-store-merging.c (bswap_replace): Likewise.
11742         (merged_store_group::apply_stores): Likewise.
11743         * gimple-ssa-strength-reduction.c (insert_initializers): Likewise.
11744         * gimple.c (verify_gimple_pp): Likewise.
11745         * graphite-poly.c (print_pbb_body): Likewise.
11746         * passes.c (pass_manager::register_one_dump_file): Convert
11747         local "optgroup_flags" to optgroup_flags_t.
11748         * print-tree.c (print_node): Use TDF_NONE rather than 0.
11749         (debug): Likewise.
11750         (debug_body): Likewise.
11751         * tree-pass.h (struct pass_data): Convert field "optgroup_flags"
11752         to optgroup_flags_t.
11753         * tree-pretty-print.c (print_struct_decl): Use TDF_NONE rather
11754         than 0.
11755         * tree-ssa-math-opts.c (convert_mult_to_fma_1): Likewise.
11756         (convert_mult_to_fma): Likewise.
11757         * tree-ssa-reassoc.c (undistribute_ops_list): Likewise.
11758         * tree-ssa-sccvn.c (vn_eliminate): Likewise.
11759         * tree-vect-data-refs.c (dump_lower_bound): Convert param
11760         "dump_kind" to dump_flags_t.
11762 2018-06-08  Segher Boessenkool  <segher@kernel.crashing.org>
11764         * config/rs6000/rs6000.c (min, max): Delete.
11766 2018-06-08  Segher Boessenkool  <segher@kernel.crashing.org>
11768         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete mention of
11769         -mabi=spe and -mabi=no-spe.
11771 2018-06-08  Martin Liska  <mliska@suse.cz>
11773         * ipa-pure-const.c (propagate_pure_const): Use ::get at places
11774         where we expect an existing summary.
11776 2018-06-08  Martin Liska  <mliska@suse.cz>
11778         * ipa-inline-analysis.c (simple_edge_hints): Use ::get method.
11779         * ipa-inline.h (estimate_edge_growth): Likewise.
11781 2018-06-08  Martin Liska  <mliska@suse.cz>
11783         * cgraph.c (function_version_hasher::hash): Use
11784         cgraph_node::get_uid ().
11785         (function_version_hasher::equal):
11786         * cgraph.h (cgraph_node::get_uid): New method.
11787         * ipa-inline.c (update_caller_keys): Use
11788         cgraph_node::get_uid ().
11789         (update_callee_keys): Likewise.
11790         * ipa-utils.c (searchc): Likewise.
11791         (ipa_reduced_postorder): Likewise.
11792         * lto-cgraph.c (input_node): Likewise.
11793         * passes.c (is_pass_explicitly_enabled_or_disabled): Likewise.
11794         * symbol-summary.h (symtab_insertion): Likewise.
11795         (symtab_removal): Likewise.
11796         (symtab_duplication): Likewise.
11797         * tree-pretty-print.c (dump_function_header): Likewise.
11798         * tree-sra.c (convert_callers_for_node): Likewise.
11800 2018-06-08  Martin Liska  <mliska@suse.cz>
11802         * cgraph.c (symbol_table::create_edge): Always assign a new
11803         unique number.
11804         (symbol_table::free_edge): Do not recycle numbers.
11805         * cgraph.h (cgraph_edge::get): New method.
11806         * symbol-summary.h (symtab_removal): Use it.
11807         (symtab_duplication): Likewise.
11808         (call_summary::hashable_uid): Remove.
11810 2018-06-08  Martin Liska  <mliska@suse.cz>
11812         * ipa-inline-analysis.c (inline_edge_removal_hook): Remove.
11813         (initialize_growth_caches): Remove.
11814         (free_growth_caches): Likewise.
11815         (do_estimate_edge_time): Use edge_growth_cache.
11816         (do_estimate_edge_size): Likewise.
11817         (do_estimate_edge_hints): Likewise.
11818         * ipa-inline.c (reset_edge_caches): Likewise.
11819         (recursive_inlining): Likewise.
11820         (inline_small_functions): Likewise.
11821         * ipa-inline.h (initialize_growth_caches): Remove.
11822         (estimate_edge_size): Likewise.
11823         (estimate_edge_time): Likewise.
11824         (estimate_edge_hints): Likewise.
11825         (reset_edge_growth_cache): Likewise.
11826         * symbol-summary.h (call_summary::remove): New method.
11828 2018-06-08  Martin Liska  <mliska@suse.cz>
11830         * ipa-cp.c (class edge_clone_summary): New summary.
11831         (grow_edge_clone_vectors): Remove.
11832         (ipcp_edge_duplication_hook): Remove.
11833         (class edge_clone_summary_t): New call_summary class.
11834         (ipcp_edge_removal_hook): Remove.
11835         (edge_clone_summary_t::duplicate): New function.
11836         (get_next_cgraph_edge_clone): Use edge_clone_summaries.
11837         (create_specialized_node): Likewise.
11838         (ipcp_driver): Initialize edge_clone_summaries and do not
11839         register hooks.
11841 2018-06-08  Martin Liska  <mliska@suse.cz>
11843         * symbol-summary.h (get): New function.
11844         (call_summary::m_initialize_when_cloning): New class member.
11846 2018-06-08  Martin Liska  <mliska@suse.cz>
11848         * cgraph.c (cgraph_node::remove): Do not recycle uid.
11849         * cgraph.h (symbol_table::release_symbol): Do not pass uid.
11850         (symbol_table::allocate_cgraph_symbol): Do not set uid.
11851         * passes.c (uid_hash_t): Record removed_nodes by their uids.
11852         (remove_cgraph_node_from_order): Use the removed_nodes set.
11853         (do_per_function_toporder): Likwise.
11854         * symbol-summary.h (symtab_insertion): Use cgraph_node::uid
11855         instead of summary_uid.
11856         (symtab_removal): Likewise.
11857         (symtab_duplication): Likewise.
11859 2018-06-08  Martin Liska  <mliska@suse.cz>
11861         * ipa-cp.c (ipcp_store_bits_results): Use
11862         ipcp_transformation_sum.
11863         (ipcp_store_vr_results): Likewise.
11864         * ipa-prop.c (ipcp_grow_transformations_if_necessary): Renamed
11865         to ...
11866         (ipcp_transformation_initialize): ... this.
11867         (ipa_set_node_agg_value_chain):
11868         (ipa_node_params_t::duplicate): Use ipcp_transformation_sum.
11869         (write_ipcp_transformation_info): Likewise.
11870         (read_ipcp_transformation_info): Likewise.
11871         (ipcp_update_bits): Likewise.
11872         (ipcp_update_vr): Likewise.
11873         (ipcp_transform_function): Likewise.
11874         * ipa-prop.h: Rename ipcp_transformation_summary to
11875         ipcp_transformation.
11876         (class ipcp_transformation_t): New function summary.
11877         (ipcp_get_transformation_summary): Use ipcp_transformation_sum.
11878         (ipa_get_agg_replacements_for_node): Likewise.
11880 2018-06-08  Martin Liska  <mliska@suse.cz>
11882         * ipa-pure-const.c (struct funct_state_d): Do it class instead
11883         of struct.
11884         (class funct_state_summary_t): New function_summary class.
11885         (has_function_state): Remove.
11886         (get_function_state): Likewise.
11887         (set_function_state): Likewise.
11888         (add_new_function): Likewise.
11889         (funct_state_summary_t::insert): New function.
11890         (duplicate_node_data): Remove.
11891         (remove_node_data): Remove.
11892         (funct_state_summary_t::duplicate): New function.
11893         (register_hooks): Create new funct_state_summaries.
11894         (pure_const_generate_summary): Use it.
11895         (pure_const_write_summary): Likewise.
11896         (pure_const_read_summary): Likewise.
11897         (propagate_pure_const): Likewise.
11898         (propagate_nothrow): Likewise.
11899         (dump_malloc_lattice): Likewise.
11900         (propagate_malloc): Likewise.
11901         (execute): Do not register hooks, just remove summary
11902         instead.
11903         (pass_ipa_pure_const::pass_ipa_pure_const): Simplify
11904         constructor.
11906 2018-06-08  Martin Liska  <mliska@suse.cz>
11908         * ipa-reference.c (remove_node_data): Remove.
11909         (duplicate_node_data): Likewise.
11910         (class ipa_ref_var_info_summary_t): New class.
11911         (class ipa_ref_opt_summary_t): Likewise.
11912         (get_reference_vars_info): Use ipa_ref_var_info_summaries.
11913         (get_reference_optimization_summary): Use
11914         ipa_ref_opt_sum_summaries.
11915         (set_reference_vars_info): Remove.
11916         (set_reference_optimization_summary): Likewise.
11917         (ipa_init): Create summaries.
11918         (init_function_info): Use function summary.
11919         (ipa_ref_opt_summary_t::duplicate): New function.
11920         (ipa_ref_opt_summary_t::remove): New function.
11921         (get_read_write_all_from_node): Fix GNU coding style.
11922         (propagate): Use function summary.
11923         (write_node_summary_p): Fix GNU coding style.
11924         (stream_out_bitmap): Likewise.
11925         (ipa_reference_read_optimization_summary): Use function summary.
11926         (ipa_reference_c_finalize): Do not release hooks.
11928 2018-06-08  Martin Liska  <mliska@suse.cz>
11930         * ipa-fnsummary.c (dump_ipa_call_summary): Use ::get method.
11931         (analyze_function_body): Extract multiple calls of get_create.
11932         * ipa-inline-analysis.c (simple_edge_hints): Likewise.
11933         * ipa-inline.c (recursive_inlining): Use ::get method.
11934         * ipa-inline.h (estimate_edge_growth): Likewise.
11936 2018-06-08  Martin Liska  <mliska@suse.cz>
11938         * hsa-common.h (enum hsa_function_kind): Rename HSA_NONE to
11939         HSA_INVALID.
11940         (hsa_function_summary::hsa_function_summary): Use the new enum
11941         value.
11942         (hsa_gpu_implementation_p): Use hsa_summaries::get.
11943         * hsa-gen.c (hsa_get_host_function): Likewise.
11944         (get_brig_function_name): Likewise.
11945         * ipa-hsa.c (process_hsa_functions): Likewise.
11946         (ipa_hsa_write_summary): Likewise.
11947         * symbol-summary.h (symtab_duplication): Use ::get function/
11948         (get): New function.
11950 2018-06-08  Martin Liska  <mliska@suse.cz>
11952         * config/i386/i386.c (ix86_can_inline_p): Use get_create instead
11953         of get.
11954         * hsa-common.c (hsa_summary_t::link_functions): Likewise.
11955         (hsa_register_kernel): Likewise.
11956         * hsa-common.h (hsa_gpu_implementation_p): Likewise.
11957         * hsa-gen.c (hsa_get_host_function): Likewise.
11958         (get_brig_function_name): Likewise.
11959         (generate_hsa): Likewise.
11960         (pass_gen_hsail::execute): Likewise.
11961         * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
11962         (devirtualization_time_bonus): Likewise.
11963         (ipcp_propagate_stage): Likewise.
11964         * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
11965         (edge_set_predicate): Likewise.
11966         (evaluate_conditions_for_known_args): Likewise.
11967         (evaluate_properties_for_edge): Likewise.
11968         (ipa_fn_summary::reset): Likewise.
11969         (ipa_fn_summary_t::duplicate): Likewise.
11970         (dump_ipa_call_summary): Likewise.
11971         (ipa_dump_fn_summary): Likewise.
11972         (analyze_function_body): Likewise.
11973         (compute_fn_summary): Likewise.
11974         (estimate_edge_devirt_benefit): Likewise.
11975         (estimate_edge_size_and_time): Likewise.
11976         (estimate_calls_size_and_time): Likewise.
11977         (estimate_node_size_and_time): Likewise.
11978         (inline_update_callee_summaries): Likewise.
11979         (remap_edge_change_prob): Likewise.
11980         (remap_edge_summaries): Likewise.
11981         (ipa_merge_fn_summary_after_inlining): Likewise.
11982         (ipa_update_overall_fn_summary): Likewise.
11983         (read_ipa_call_summary): Likewise.
11984         (inline_read_section): Likewise.
11985         (write_ipa_call_summary): Likewise.
11986         (ipa_fn_summary_write): Likewise.
11987         (ipa_free_fn_summary): Likewise.
11988         * ipa-hsa.c (process_hsa_functions): Likewise.
11989         (ipa_hsa_write_summary): Likewise.
11990         (ipa_hsa_read_section): Likewise.
11991         * ipa-icf.c (sem_function::merge): Likewise.
11992         * ipa-inline-analysis.c (simple_edge_hints): Likewise.
11993         (do_estimate_edge_time): Likewise.
11994         (estimate_size_after_inlining): Likewise.
11995         (estimate_growth): Likewise.
11996         (growth_likely_positive): Likewise.
11997         * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
11998         (inline_call): Likewise.
11999         * ipa-inline.c (caller_growth_limits): Likewise.
12000         (can_inline_edge_p): Likewise.
12001         (can_inline_edge_by_limits_p): Likewise.
12002         (compute_uninlined_call_time): Likewise.
12003         (compute_inlined_call_time): Likewise.
12004         (want_inline_small_function_p): Likewise.
12005         (edge_badness): Likewise.
12006         (update_caller_keys): Likewise.
12007         (update_callee_keys): Likewise.
12008         (recursive_inlining): Likewise.
12009         (inline_small_functions): Likewise.
12010         (inline_to_all_callers_1): Likewise.
12011         (dump_overall_stats): Likewise.
12012         (early_inline_small_functions): Likewise.
12013         (early_inliner): Likewise.
12014         * ipa-inline.h (estimate_edge_growth): Likewise.
12015         * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
12016         * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
12017         * ipa-prop.h (IPA_NODE_REF): Likewise.
12018         (IPA_EDGE_REF): Likewise.
12019         * ipa-pure-const.c (malloc_candidate_p): Likewise.
12020         (propagate_malloc): Likewise.
12021         * ipa-split.c (execute_split_functions): Likewise.
12022         * symbol-summary.h: Rename get to get_create.
12023         (get): Likewise.
12024         (get_create): Likewise.
12025         * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
12027 2018-06-08  Martin Liska  <mliska@suse.cz>
12029         * symbol-summary.h (release): Move definition out of class
12030         declaration.
12031         (symtab_removal): Likewise.
12032         (symtab_duplication): Likewise.
12034 2018-06-08  Martin Liska  <mliska@suse.cz>
12036         * symbol-summary.h (function_summary): Move constructor
12037         implementation out of class declaration.
12038         (release): Likewise.
12039         (symtab_insertion): Likewise.
12040         (symtab_removal): Likewise.
12041         (symtab_duplication): Likewise.
12042         (get): Likewise.
12044 2018-06-08  Martin Liska  <mliska@suse.cz>
12046         * Makefile.in: Remove support for MPX (macros, related functions,
12047         fields in cgraph_node, ...).
12048         * builtin-types.def (BT_BND): Likewise.
12049         (BT_FN_BND_CONST_PTR): Likewise.
12050         (BT_FN_CONST_PTR_BND): Likewise.
12051         (BT_FN_VOID_PTR_BND): Likewise.
12052         (BT_FN_BND_CONST_PTR_SIZE): Likewise.
12053         (BT_FN_VOID_CONST_PTR_BND_CONST_PTR): Likewise.
12054         * builtins.c (expand_builtin_memcpy_with_bounds): Likewise.
12055         (expand_builtin_mempcpy_with_bounds): Likewise.
12056         (expand_builtin_memset_with_bounds): Likewise.
12057         (expand_builtin_memset_args): Likewise.
12058         (std_expand_builtin_va_start): Likewise.
12059         (expand_builtin): Likewise.
12060         (expand_builtin_with_bounds): Likewise.
12061         * builtins.def (DEF_BUILTIN_CHKP): Likewise.
12062         (DEF_LIB_BUILTIN_CHKP): Likewise.
12063         (DEF_EXT_LIB_BUILTIN_CHKP): Likewise.
12064         (DEF_CHKP_BUILTIN): Likewise.
12065         (BUILT_IN_MEMCPY): Likewise.
12066         (BUILT_IN_MEMMOVE): Likewise.
12067         (BUILT_IN_MEMPCPY): Likewise.
12068         (BUILT_IN_MEMSET): Likewise.
12069         (BUILT_IN_STPCPY): Likewise.
12070         (BUILT_IN_STRCAT): Likewise.
12071         (BUILT_IN_STRCHR): Likewise.
12072         (BUILT_IN_STRCPY): Likewise.
12073         (BUILT_IN_STRLEN): Likewise.
12074         (BUILT_IN_MEMCPY_CHK): Likewise.
12075         (BUILT_IN_MEMMOVE_CHK): Likewise.
12076         (BUILT_IN_MEMPCPY_CHK): Likewise.
12077         (BUILT_IN_MEMSET_CHK): Likewise.
12078         (BUILT_IN_STPCPY_CHK): Likewise.
12079         (BUILT_IN_STRCAT_CHK): Likewise.
12080         (BUILT_IN_STRCPY_CHK): Likewise.
12081         * calls.c (store_bounds): Likewise.
12082         (emit_call_1): Likewise.
12083         (special_function_p): Likewise.
12084         (maybe_warn_nonstring_arg): Likewise.
12085         (initialize_argument_information): Likewise.
12086         (finalize_must_preallocate): Likewise.
12087         (compute_argument_addresses): Likewise.
12088         (expand_call): Likewise.
12089         * cfgexpand.c (expand_call_stmt): Likewise.
12090         (expand_return): Likewise.
12091         (expand_gimple_stmt_1): Likewise.
12092         (pass_expand::execute): Likewise.
12093         * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
12094         (cgraph_node::remove): Likewise.
12095         (cgraph_node::dump): Likewise.
12096         (cgraph_node::verify_node): Likewise.
12097         * cgraph.h (chkp_function_instrumented_p): Likewise.
12098         (symtab_node::get_alias_target): Likewise.
12099         (cgraph_node::can_remove_if_no_direct_calls_and_refs_p): Likewise.
12100         (cgraph_local_p): Likewise.
12101         * cgraphbuild.c (cgraph_edge::rebuild_edges): Likewise.
12102         (cgraph_edge::rebuild_references): Likewise.
12103         * cgraphunit.c (varpool_node::finalize_decl): Likewise.
12104         (walk_polymorphic_call_targets): Likewise.
12105         (cgraph_node::expand_thunk): Likewise.
12106         (symbol_table::output_weakrefs): Likewise.
12107         * common/config/i386/i386-common.c (OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET): Likewise.
12108         (ix86_handle_option): Likewise.
12109         * config/i386/constraints.md: Likewise.
12110         * config/i386/i386-builtin-types.def (BND): Likewise.
12111         (VOID): Likewise.
12112         (PVOID): Likewise.
12113         (ULONG): Likewise.
12114         * config/i386/i386-builtin.def (BDESC_END): Likewise.
12115         (BDESC_FIRST): Likewise.
12116         (BDESC): Likewise.
12117         * config/i386/i386-c.c (ix86_target_macros_internal): Likewise.
12118         * config/i386/i386-protos.h (ix86_bnd_prefixed_insn_p): Likewise.
12119         * config/i386/i386.c (enum reg_class): Likewise.
12120         (ix86_target_string): Likewise.
12121         (ix86_option_override_internal): Likewise.
12122         (ix86_conditional_register_usage): Likewise.
12123         (ix86_valid_target_attribute_inner_p): Likewise.
12124         (ix86_set_indirect_branch_type): Likewise.
12125         (ix86_set_current_function): Likewise.
12126         (ix86_function_arg_regno_p): Likewise.
12127         (init_cumulative_args): Likewise.
12128         (ix86_function_arg_advance): Likewise.
12129         (ix86_function_arg): Likewise.
12130         (ix86_pass_by_reference): Likewise.
12131         (ix86_function_value_regno_p): Likewise.
12132         (ix86_function_value_1): Likewise.
12133         (ix86_function_value_bounds): Likewise.
12134         (ix86_return_in_memory): Likewise.
12135         (ix86_setup_incoming_vararg_bounds): Likewise.
12136         (ix86_va_start): Likewise.
12137         (indirect_thunk_need_prefix): Likewise.
12138         (print_reg): Likewise.
12139         (ix86_print_operand): Likewise.
12140         (ix86_expand_call): Likewise.
12141         (ix86_output_function_return): Likewise.
12142         (reg_encoded_number): Likewise.
12143         (BDESC_VERIFYS): Likewise.
12144         (ix86_init_mpx_builtins): Likewise.
12145         (ix86_init_builtins): Likewise.
12146         (ix86_emit_cmove): Likewise.
12147         (ix86_emit_move_max): Likewise.
12148         (ix86_expand_builtin): Likewise.
12149         (ix86_builtin_mpx_function): Likewise.
12150         (ix86_get_arg_address_for_bt): Likewise.
12151         (ix86_load_bounds): Likewise.
12152         (ix86_store_bounds): Likewise.
12153         (ix86_load_returned_bounds): Likewise.
12154         (ix86_store_returned_bounds): Likewise.
12155         (ix86_class_likely_spilled_p): Likewise.
12156         (ix86_hard_regno_mode_ok): Likewise.
12157         (x86_order_regs_for_local_alloc): Likewise.
12158         (ix86_mitigate_rop): Likewise.
12159         (ix86_bnd_prefixed_insn_p): Likewise.
12160         (ix86_mpx_bound_mode): Likewise.
12161         (ix86_make_bounds_constant): Likewise.
12162         (ix86_initialize_bounds): Likewise.
12163         (TARGET_LOAD_BOUNDS_FOR_ARG): Likewise.
12164         (TARGET_STORE_BOUNDS_FOR_ARG): Likewise.
12165         (TARGET_LOAD_RETURNED_BOUNDS): Likewise.
12166         (TARGET_STORE_RETURNED_BOUNDS): Likewise.
12167         (TARGET_CHKP_BOUND_MODE): Likewise.
12168         (TARGET_BUILTIN_CHKP_FUNCTION): Likewise.
12169         (TARGET_CHKP_FUNCTION_VALUE_BOUNDS): Likewise.
12170         (TARGET_CHKP_MAKE_BOUNDS_CONSTANT): Likewise.
12171         (TARGET_CHKP_INITIALIZE_BOUNDS): Likewise.
12172         * config/i386/i386.h (TARGET_MPX): Likewise.
12173         (TARGET_MPX_P): Likewise.
12174         (VALID_BND_REG_MODE): Likewise.
12175         (FIRST_BND_REG): Likewise.
12176         (LAST_BND_REG): Likewise.
12177         (enum reg_class): Likewise.
12178         (BND_REG_P): Likewise.
12179         (BND_REGNO_P): Likewise.
12180         (BNDmode): Likewise.
12181         (ADJUST_INSN_LENGTH): Likewise.
12182         * config/i386/i386.md: Likewise.
12183         * config/i386/i386.opt: Likewise.
12184         * config/i386/linux-common.h (LIBMPX_LIBS): Likewise.
12185         (defined): Likewise.
12186         (LINK_MPX): Likewise.
12187         (MPX_SPEC): Likewise.
12188         (LIBMPX_SPEC): Likewise.
12189         (LIBMPXWRAPPERS_SPEC): Likewise.
12190         (CHKP_SPEC): Likewise.
12191         * config/i386/predicates.md: Likewise.
12192         * dbxout.c (dbxout_type): Likewise.
12193         * doc/extend.texi: Likewise.
12194         * doc/invoke.texi: Likewise.
12195         * doc/md.texi: Likewise.
12196         * doc/tm.texi: Likewise.
12197         * doc/tm.texi.in: Likewise.
12198         * dwarf2out.c (is_base_type): Likewise.
12199         (gen_formal_types_die): Likewise.
12200         (gen_subprogram_die): Likewise.
12201         (gen_type_die_with_usage): Likewise.
12202         (gen_decl_die): Likewise.
12203         (dwarf2out_late_global_decl): Likewise.
12204         * expr.c (expand_assignment): Likewise.
12205         (emit_storent_insn): Likewise.
12206         (store_expr_with_bounds): Likewise.
12207         (store_expr): Likewise.
12208         (expand_expr_real_1): Likewise.
12209         * expr.h (store_expr_with_bounds): Likewise.
12210         * function.c (use_register_for_decl): Likewise.
12211         (struct bounds_parm_data): Likewise.
12212         (assign_parms_augmented_arg_list): Likewise.
12213         (assign_parm_find_entry_rtl): Likewise.
12214         (assign_parm_is_stack_parm): Likewise.
12215         (assign_parm_load_bounds): Likewise.
12216         (assign_bounds): Likewise.
12217         (assign_parms): Likewise.
12218         (expand_function_start): Likewise.
12219         * gcc.c (CHKP_SPEC): Likewise.
12220         * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
12221         * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Likewise.
12222         (wrestrict_dom_walker::check_call): Likewise.
12223         * gimple.c (gimple_build_call_from_tree): Likewise.
12224         * gimple.h (enum gf_mask): Likewise.
12225         (gimple_call_with_bounds_p): Likewise.
12226         (gimple_call_set_with_bounds): Likewise.
12227         * gimplify.c (gimplify_init_constructor): Likewise.
12228         * ipa-cp.c (initialize_node_lattices): Likewise.
12229         (propagate_constants_across_call): Likewise.
12230         (find_more_scalar_values_for_callers_subset): Likewise.
12231         * ipa-hsa.c (process_hsa_functions): Likewise.
12232         * ipa-icf-gimple.c (func_checker::compare_gimple_call): Likewise.
12233         * ipa-icf.c (sem_function::merge): Likewise.
12234         * ipa-inline.c (early_inliner): Likewise.
12235         * ipa-pure-const.c (warn_function_noreturn): Likewise.
12236         (warn_function_cold): Likewise.
12237         (propagate_pure_const): Likewise.
12238         * ipa-ref.h (enum GTY): Likewise.
12239         * ipa-split.c (find_retbnd): Likewise.
12240         (consider_split): Likewise.
12241         (split_function): Likewise.
12242         * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
12243         * ipa.c (walk_polymorphic_call_targets): Likewise.
12244         (symbol_table::remove_unreachable_nodes): Likewise.
12245         (process_references): Likewise.
12246         (cgraph_build_static_cdtor_1): Likewise.
12247         * lto-cgraph.c (lto_output_node): Likewise.
12248         (output_refs): Likewise.
12249         (compute_ltrans_boundary): Likewise.
12250         (input_overwrite_node): Likewise.
12251         (input_node): Likewise.
12252         (input_cgraph_1): Likewise.
12253         * params.def (PARAM_CHKP_MAX_CTOR_SIZE): Likewise.
12254         * passes.c (pass_manager::execute_early_local_passes): Likewise.
12255         (class pass_chkp_instrumentation_passes): Likewise.
12256         (make_pass_chkp_instrumentation_passes): Likewise.
12257         * passes.def: Likewise.
12258         * rtl.h (struct GTY): Likewise.
12259         (CALL_EXPR_WITH_BOUNDS_P): Likewise.
12260         * stor-layout.c (layout_type): Likewise.
12261         * symtab.c: Likewise.
12262         * target.def: Likewise.
12263         * targhooks.c (default_chkp_bound_type): Likewise.
12264         (default_chkp_bound_mode): Likewise.
12265         (default_builtin_chkp_function): Likewise.
12266         (default_chkp_function_value_bounds): Likewise.
12267         (default_chkp_make_bounds_constant): Likewise.
12268         (default_chkp_initialize_bounds): Likewise.
12269         * targhooks.h (default_chkp_bound_type): Likewise.
12270         (default_chkp_bound_mode): Likewise.
12271         (default_builtin_chkp_function): Likewise.
12272         (default_chkp_function_value_bounds): Likewise.
12273         (default_chkp_make_bounds_constant): Likewise.
12274         (default_chkp_initialize_bounds): Likewise.
12275         * toplev.c (compile_file): Likewise.
12276         (process_options): Likewise.
12277         * tree-core.h (DEF_BUILTIN): Likewise.
12278         (DEF_BUILTIN_CHKP): Likewise.
12279         * tree-inline.c (declare_return_variable): Likewise.
12280         (remap_gimple_stmt): Likewise.
12281         (copy_bb): Likewise.
12282         (initialize_inlined_parameters): Likewise.
12283         (expand_call_inline): Likewise.
12284         * tree-pass.h (make_pass_ipa_chkp_versioning): Likewise.
12285         (make_pass_ipa_chkp_early_produce_thunks): Likewise.
12286         (make_pass_ipa_chkp_produce_thunks): Likewise.
12287         (make_pass_chkp): Likewise.
12288         (make_pass_chkp_opt): Likewise.
12289         (make_pass_chkp_instrumentation_passes): Likewise.
12290         * tree-pretty-print.c (dump_generic_node): Likewise.
12291         * tree-ssa-ccp.c (insert_clobber_before_stack_restore): Likewise.
12292         * tree-ssa-dce.c (propagate_necessity): Likewise.
12293         (eliminate_unnecessary_stmts): Likewise.
12294         * tree-ssa-pre.c (create_expression_by_pieces): Likewise.
12295         * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
12296         * tree-ssa-sccvn.h: Likewise.
12297         * tree-ssa-strlen.c (get_string_length): Likewise.
12298         (valid_builtin_call): Likewise.
12299         (adjust_last_stmt): Likewise.
12300         (handle_builtin_strchr): Likewise.
12301         (handle_builtin_strcpy): Likewise.
12302         (handle_builtin_stxncpy): Likewise.
12303         (handle_builtin_memcpy): Likewise.
12304         (handle_builtin_strcat): Likewise.
12305         (strlen_check_and_optimize_stmt): Likewise.
12306         * tree-stdarg.c (expand_ifn_va_arg_1): Likewise.
12307         * tree-streamer-in.c: Likewise.
12308         * tree-streamer.c (record_common_node): Likewise.
12309         * tree.c (tree_code_size): Likewise.
12310         (wide_int_to_tree_1): Likewise.
12311         (type_contains_placeholder_1): Likewise.
12312         (build_common_tree_nodes): Likewise.
12313         * tree.def (POINTER_BOUNDS_TYPE): Likewise.
12314         * tree.h (POINTER_BOUNDS_TYPE_P): Likewise.
12315         (POINTER_BOUNDS_P): Likewise.
12316         (BOUNDED_TYPE_P): Likewise.
12317         (BOUNDED_P): Likewise.
12318         (CALL_WITH_BOUNDS_P): Likewise.
12319         (pointer_bounds_type_node): Likewise.
12320         * value-prof.c (gimple_ic): Likewise.
12321         * var-tracking.c (vt_add_function_parameters): Likewise.
12322         * varasm.c (make_decl_rtl): Likewise.
12323         (assemble_start_function): Likewise.
12324         (output_constant): Likewise.
12325         (maybe_assemble_visibility): Likewise.
12326         * varpool.c (ctor_for_folding): Likewise.
12327         * chkp-builtins.def: Remove.
12328         * ipa-chkp.c: Remove.
12329         * ipa-chkp.h: Remove.
12330         * rtl-chkp.c: Remove.
12331         * rtl-chkp.h: Remove.
12332         * tree-chkp-opt.c: Remove.
12333         * tree-chkp.c: Remove.
12334         * tree-chkp.h: Remove.
12336 2018-06-07  Carl Love  <cel@us.ibm.com>
12338         * config/rs6000/vsx.md (vextract_fp_from_shorth,
12339         vextract_fp_from_shortl): Add BE support.
12341 2018-06-07  Paul Koning  <ni1d@arrl.net>
12343         * compare-elim.c (try_merge_compare): Don't merge compare if
12344         address contains a side effect.
12345         (try_eliminate_compare): Likewise.
12347 2018-06-07  Olga Makhotina  <olga.makhotina@intel.com>
12349         * config.gcc: Support "tremont".
12350         * config/i386/driver-i386.c (host_detect_local_cpu): Detect "tremont".
12351         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
12352         PROCESSOR_TREMONT.
12353         * config/i386/i386.c (m_TREMONT): Define.
12354         (processor_target_table): Add "tremont".
12355         (PTA_TREMONT): Define.
12356         (ix86_lea_outperforms): Add TARGET_TREMONT.
12357         (get_builtin_code_for_version): Handle PROCESSOR_TREMONT.
12358         (fold_builtin_cpu): Add M_INTEL_TREMONT, replace M_INTEL_GOLDMONT
12359         and M_INTEL_GOLDMONT_PLUS.
12360         (fold_builtin_cpu): Add "tremont".
12361         (ix86_add_stmt_cost): Add TARGET_TREMONT.
12362         (ix86_option_override_internal): Add "tremont".
12363         * config/i386/i386.h (processor_costs): Define TARGET_TREMONT.
12364         (processor_type): Add PROCESSOR_TREMONT.
12365         * config/i386/x86-tune.def: Add m_TREMONT.
12366         * doc/invoke.texi: Add tremont as x86 -march=/-mtune= CPU type.
12368 2018-06-07  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
12370         * config/msp430/msp430.c (msp430_mcu_name): Set the "i" in the
12371         symbol defined for msp430i* devices to be lower case.
12373 2018-06-07  Richard Biener  <rguenther@suse.de>
12375         * graphite-sese-to-poly.c (extract_affine): Avoid unneded
12376         wrapping.  Properly wrap the result of a BIT_NOT_EXPR.
12377         Properly wrap signed arithmetic if overflow wraps.
12379 2018-06-07  Jakub Jelinek  <jakub@redhat.com>
12381         PR tree-optimization/69615
12382         * tree-ssa-reassoc.c (optimize_range_tests_var_bound): If rhs2 is lhs
12383         of a cast from a same precision integral SSA_NAME in a bb dominated
12384         by first_bb, retry with rhs2 set to the rhs1 of the cast.  Don't emit
12385         cast to utype if rhs2 has already a compatible type.
12387 2018-06-07  Richard Biener  <rguenther@suse.de>
12389         PR tree-optimization/85935
12390         * graphite-scop-detection.c (find_params_in_bb): Analyze
12391         condition operands with respect to the correct loop.  Assert
12392         the analysis doesn't fail.
12394 2018-06-04  Carl Love  <cel@us.ibm.com>
12396         * config/rs6000/vsx.md (first_match_index_<mode>): Calculate index
12397         using natural element order.  Use gen_lshrsi3 instead of gen_ashrsi3
12398         as it is slightly cheaper.
12399         (first_match_or_eos_index_<mode>):
12400         Calculate index using natural element order.
12401         (first_match_index_<mode>):
12402         Calculate index using natural element order.
12403         (first_match_or_eos_index_<mode>):
12404         Calculate index using natural order.
12405         (define_insn vclzlsbb): Change to define_insn vclzlsbb_<mode>.
12406         for BE and LE modes.
12407         * config/rs6000/rs6000-c.c: Rename P9V_BUILTIN_VCLZLSBB,
12408         P9V_BUILTIN_VCLZLSBB_V16QI.
12409         * config/rs6000/rs6000-builtin.def: Make VCLZLSBB mode
12410         specific.
12412 2018-06-06  Kelvin Nilsen  <kelvin@gcc.gnu.org>
12414         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Adjust
12415         indentation and line wrap for many prototypes.  Add missing
12416         @smallexample directives around block of prototypes for vec_xl and
12417         vec_xst.
12419 2018-06-05  Michael Meissner  <meissner@linux.ibm.com>
12421         * config/rs6000/rs6000.c (rs6000_passes_ieee128): New boolean to
12422         track if we pass or return IEEE 128-bit floating point.
12423         (ieee128_mangling_gcc_8_1): New boolean to say whether to generate
12424         C++ mangling that is compatible with GCC 8.1.
12425         (TARGET_ASM_GLOBALIZE_DECL_NAME): Override target hook.
12426         (init_cumulative_args): Note if we pass or return IEEE 128-bit
12427         floating point types.
12428         (rs6000_function_arg_advance_1): Likewise.
12429         (rs6000_mangle_type): Optionally generate mangled names that match
12430         what GCC 8.1 generated for IEEE 128-bit floating point types.
12431         (rs6000_globalize_decl_name): If we have an external function that
12432         passes or returns IEEE 128-bit types, generate a weak reference
12433         from the mangled name used in GCC 8.1 to the current mangled
12434         name.
12435         (rs6000_init_builtins): Make __ibm128 use the long double type if
12436         long double is IBM extended double.  Make __float128 use the long
12437         double type if long double is IEEE 128-bit.
12439         PR target/85657
12440         * config/rs6000/rs6000-builtin.def (BU_IBM128_2): New helper
12441         macro for __ibm128 built-in functions.
12442         (PACK_IF): Add __ibm128 pack/unpack functions.
12443         (UNPACK_IF): Likewise.
12444         * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Do not
12445         enable long double built-in functions if long double is IEEE
12446         128-bit floating point.
12447         (rs6000_invalid_builtin): Update long double built-in function
12448         error message.
12449         (rs6000_expand_builtin): For PACK_IF and UNPACK_IF built-in
12450         functions, adjust the built-in function to use the long double
12451         built-in function if __ibm128 and long double are the same type.
12452         * doc/extend.texi (PowerPC builtins): Update documention for
12453         __builtin_{,un}pack_longdouble.  Add documentation for
12454         __builtin_{,un}pack_ibm128.
12456 2018-06-06  Jim Wilson  <jimw@sifive.com>
12458         * config/riscv/riscv.c (enum riscv_privilege_levels): New.
12459         (struct machine_function): New field interrupt_mode.
12460         (riscv_handle_type_attribute): New function.  Add forward declaration.
12461         (riscv_attribute_table) <interrupt>: Use riscv_handle_type_attribute.
12462         (riscv_expand_epilogue): Check interrupt_mode field.
12463         (riscv_set_current_function): Check interrupt attribute args and
12464         set interrupt_mode field.
12465         * config/riscv/riscv.md (UNSPECV_SRET, UNSPECV_URET): New.
12466         (riscv_sret, riscv_uret): New.
12467         * doc/extend.texi (RISC-V Function Attributes) <interrupt>: Document
12468         new arguments to interrupt attribute.
12470 2018-06-06  Peter Bergner  <bergner@vnet.ibm.com>
12472         PR target/63177
12473         * /config/rs6000/rs6000.h (ASM_CPU_SPEC): Add support for -mpower9.
12474         Don't handle -mcpu=power8 if -mpower9-vector is also used.
12476 2018-06-06  Kelvin Nilsen  <kelvin@gcc.gnu.org>
12478         * config/rs6000/rs6000-builtin.def (VSX_BUILTIN_VEC_LD,
12479         VSX_BUILTIN_VEC_ST): Add comment to explain non-traditional uses.
12480         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
12481         several redundant entries.
12483 2018-06-06  David Malcolm  <dmalcolm@redhat.com>
12485         * config/i386/i386-protos.h (ix86_expand_call): Strengthen return
12486         type from "rtx" to "rtx_insn *".
12487         * config/i386/i386.c (ix86_expand_split_stack_prologue): Likewise
12488         for local "call_insn", removing cast.
12489         (ix86_expand_call): Likewise, introducing a "call_insn" local.
12491 2018-06-06  Eric Botcazou  <ebotcazou@adacore.com>
12493         PR tree-optimization/86066
12494         * gimple-ssa-store-merging.c (process_store): Do not bypass BIT_NOT_EXPR
12495         for BIT_INSERT_EXPR stores.
12497 2018-06-06  Richard Biener  <rguenther@suse.de>
12499         PR tree-optimization/86062
12500         * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
12501         component refs ontop
12502         of to be offsetted base.
12504 2018-06-06  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
12506         * config/msp430/msp430.c (msp430_attr): Allow interrupt handlers
12507         to be static and remove check on interrupt attribute name.
12509 2018-06-05  Kelvin Nilsen  <kelvin@gcc.gnu.org>
12511         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
12512         volatile qualifier from vec_lvsl and vec_lvsr argument prototypes.
12514 2018-06-05  Steve Ellcey  <sellcey@cavium.com>
12516         PR target/79924
12517         * config/aarch64/aarch64-protos.h (aarch64_err_no_fpadvsimd): Remove
12518         second argument.
12519         * config/aarch64/aarch64-protos..c (aarch64_err_no_fpadvsimd):
12520         Remove second argument, change how error is called.
12521         (aarch64_layout_arg): Remove second argument from
12522         aarch64_err_no_fpadvsimd call.
12523         (aarch64_init_cumulative_args): Ditto.
12524         (aarch64_gimplify_va_arg_expr): Ditto.
12525         * config/aarch64/aarch64.md (mov<mode>): Ditto.
12527 2018-06-05  Uros Bizjak  <ubizjak@gmail.com>
12529         * config/i386/i386.md (simple_return_indirect_internal): New expander.
12530         (*simple_return_indirect_internal<mode>): Rename from
12531         simple_return_indirect_internal.  Use W mode iterator.
12532         (rstorssp): New expander.
12533         (*rstorssp<mode>): Rename from rstorssp.  Use P mode iterator.
12534         (clrssbsy): New expander.
12535         (*clrssbsy<mode>): Rename from clrssbsy.  Use P mode iterator.
12537 2018-06-05  Andre Vieira  <andre.simoesdiasvieira@arm.com>
12539         * config/arm/arm_cmse.h (cmse_nsfptr_create): Change typeof to
12540         __typeof__.
12541         (cmse_check_pointed_object): Likewise.
12543 2018-06-05  Martin Liska  <mliska@suse.cz>
12545         PR gcov-profile/47618
12546         * doc/invoke.texi: Document how -fprofile-dir format
12547         is extended.
12549 2018-06-05  Richard Biener  <rguenther@suse.de>
12551         * tree-cfgcleanup.c (cleanup_control_flow_pre): For edge
12552         removal pretend DOM info isn't available so we do not update
12553         it and only remove edges, not dominated blocks.  Actually free
12554         DOM info in case we removed something.  Remove unreachable blocks.
12555         (mfb_keep_latches): Work with either DOM info or marked backedges.
12556         (cleanup_tree_cfg_noloop): Do not remove unreachable blocks
12557         first.  Mark backedges if DOM info isn't available.
12558         (Re-)compute DOM info after cleanup_control_flow_pre.
12560 2018-06-05  Richard Biener  <rguenther@suse.de>
12562         * tree-cfg.c (struct locus_discrim_map): Store line, not location.
12563         (locus_discrim_hasher::hash): Adjust.
12564         (locus_discrim_hasher::equal): Likewise.
12565         (next_discriminator_for_locus): Work on line directly.
12566         (same_line_p): Pass in expanded locus1 as well.
12567         (assign_discriminators): Avoid redundant location expansions.
12569 2018-06-05  Richard Biener  <rguenther@suse.de>
12571         PR tree-optimization/86046
12572         * tree-ssa.c (maybe_optimize_var): Clear DECL_GIMPLE_REG_P
12573         if required after clearing TREE_ADDRESSABLE.
12575 2018-06-05  Richard Biener  <rguenther@suse.de>
12577         PR tree-optimization/86047
12578         * tree-ssa-loop.c (for_each_index): Glob handling of all
12579         decls and constants and really handle all of them.
12581 2018-06-05  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
12583         PR target/81497
12584         * config/arm/arm-builtins.c (arm_type_qualifiers): Add
12585         qualifier_void_pointer and qualifier_const_void_pointer.
12586         (arm_ldc_qualifiers, arm_stc_qualifiers): Use the above.
12587         (arm_init_builtins): Handle the above.
12588         * config/arm/arm_acle.h (__arm_cdp, __arm_ldc, __arm_ldcl, __arm_stc,
12589         __arm_stcl, __arm_mcr, __arm_cdp2, __arm_ldc2, __arm_ldcl2, __arm_stc2,
12590         __arm_stcl2,__arm_mcr2, __arm_mcrr, __arm_mcrr2): Remove return for
12591         void intrinsics.
12593 2018-06-05  Martin Liska  <mliska@suse.cz>
12595         * auto-profile.c (read_autofdo_file): Do not use
12596         gcov_ctr_summary struct.
12597         (afdo_callsite_hot_enough_for_early_inline): Likewise.
12598         * coverage.c (struct counts_entry): Likewise.
12599         (read_counts_file): Read just single summary entry.
12600         (get_coverage_counts): Use gcov_summary struct.
12601         * coverage.h (get_coverage_counts): Likewise.
12602         * gcov-dump.c (dump_working_sets): Likewise.
12603         (tag_summary): Dump just single summary.
12604         * gcov-io.c (gcov_write_summary): Write just histogram
12605         summary.
12606         (gcov_read_summary): Read just single summary.
12607         (compute_working_sets): Use gcov_summary struct.
12608         * gcov-io.h (GCOV_TAG_SUMMARY_LENGTH): Remove usage
12609         of GCOV_COUNTERS_SUMMABLE.
12610         (GCOV_COUNTERS_SUMMABLE): Remove.
12611         (GCOV_FIRST_VALUE_COUNTER): Replace with
12612         GCOV_COUNTER_V_INTERVAL.
12613         (struct gcov_ctr_summary): Remove.
12614         (struct gcov_summary): Directly use fields of former
12615         gcov_ctr_summary.
12616         (compute_working_sets): Use gcov_summary struct.
12617         * gcov.c (read_count_file): Do not use ctrs fields.
12618         * lto-cgraph.c (merge_profile_summaries): Use gcov_summary
12619         struct.
12620         * lto-streamer.h (struct GTY): Make profile_info gcov_summary
12621         struct.
12622         * profile.c: Likewise.
12623         * profile.h: Likewise.
12625 2018-06-05  Martin Liska  <mliska@suse.cz>
12627         PR gcov-profile/84846
12628         * gcov.c (output_lines): Print working directory only
12629         in intermediate format.
12631 2018-06-05  Andreas Krebbel  <krebbel@linux.ibm.com>
12633         * config/s390/s390-builtin-types.def: Add void function type.
12634         * config/s390/s390-builtins.def: Use the function type for the
12635         tbeginc builtin.
12637 2018-06-04  Jim Wilson  <jimw@sifive.com>
12639         * config/riscv/riscv-protos.h (riscv_expand_epilogue): Change bool arg
12640         to int.
12641         * config/riscv/riscv.c (riscv_for_each_saved_reg): New args epilogue
12642         and maybe_eh_return.  Change regno to unsigned int.  Use new args to
12643         handle EH_RETURN_DATA_REGNO registers properly.
12644         (riscv_expand_prologue): Pass new args to riscv_for_each_saved_reg.
12645         (riscv_expand_epilogue): Update comment.  Change argument name and
12646         type.  Update code to use new name and type.  Pass new args to
12647         riscv_for_each_saved_reg.  Only use EH_RETURN_STACKADJ_RTX when
12648         EXCEPTION_RETURN.
12649         * config/riscv/riscv.md (NORMAL_RETURN): New.
12650         (SIBCALL_RETURN, EXCEPTION_RETURN): New.
12651         (epilogue, sibcall_epilogue): Update riscv_expand_epilogue arg.
12652         (eh_return): Call gen_eh_return_internal and emit barrier.
12653         (eh_return_internal): Call riscv_expand_epilogue.
12655 2018-06-04  Eric Botcazou  <ebotcazou@adacore.com>
12657         * gimple-ssa-store-merging.c (struct merged_store_group): Move up
12658         bit_insertion field and declare can_be_merged_into method.
12659         (merged_store_group::can_be_merged_into): New method.
12660         (imm_store_chain_info::coalesce_immediate): Call it to decide whether
12661         consecutive non-overlapping stores can be merged.  Turn MEM_REF stores
12662         into BIT_INSERT_EXPR stores if the group contains a non-MEM_REF store.
12664 2018-06-04  Richard Biener  <rguenther@suse.de>
12666         PR tree-optimization/85955
12667         * builtins.c (fold_builtin_sincos): Convert pointers to
12668         destination to appropriate type before dereferencing.
12670 2018-06-04  Segher Boessenkool  <segher@kernel.crashing.org>
12672         * config/rs6000/rs6000.md (abs<mode>2 for FLOAT128): Handle IFmode.
12674 2018-06-04  Richard Sandiford  <richard.sandiford@linaro.org>
12676         * expr.c (expand_expr_real_1): Force the operand into memory if
12677         its TYPE_MODE is BLKmode and if there is no integer mode for
12678         the number of bits being extracted.
12680 2018-06-04  Jakub Jelinek  <jakub@redhat.com>
12682         PR target/85832
12683         PR target/86036
12684         * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
12685         Use vptestnm rather than vptestm in (=Yc,v,C) variant.
12687 2018-06-04  Richard Biener  <rguenther@suse.de>
12689         * tree-cfgcleanup.c (cleanup_tree_cfg_1): Fold into...
12690         (cleanup_tree_cfg_noloop): ... single caller.  Do
12691         start_recording_case_labels later.
12693 2018-06-04  Sebastian Peryt  <sebastian.peryt@intel.com>
12695         * config/i386/cldemoteintrin.h: Change define from _X86INTRIN_H_INCLUDED
12696         to _IMMINTRIN_H_INCLUDED.
12697         * config/i386/pconfigintrin.h: Ditto.
12698         * config/i386/waitpkgintrin.h: Ditto.
12699         * config/i386/immintrin.h: Add includes for sgxintrin.h,
12700         pconfigintrin.h, waitpkgintrin.h and cldemoteintrin.h.
12701         * config/i386/x86intrin.h: Remove includes for mintrin.h, xmmintrin.h,
12702         emmintrin.h, pmmintrin.h, tmmintrin.h, smmintrin.h, wmmintrin.h,
12703         bmiintrin.h, bmi2intrin.h, lzcntintrin.h, sgxintrin.h, pconfigintrin.h,
12704         waitpkgintrin.h and cldemoteintrin.h.
12706 2018-06-04  Richard Biener  <rguenther@suse.de>
12708         PR tree-optimization/86038
12709         * tracer.c (find_best_successor): Check probability for
12710         being initialized, bail out if not.
12712 2018-06-04  Richard Earnshaw  <rearnsha@arm.com>
12714         PR target/86003
12715         * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
12716         of bits to ignore when comparing architectures.
12718 2018-06-04  Jakub Jelinek  <jakub@redhat.com>
12720         PR tree-optimization/69615
12721         * fold-const.c (merge_ranges): If range1 is - [x, x] and x is the
12722         maximum or minimum of the type, try to merge it also as if
12723         range1 is + [-, x - 1] or + [x + 1, -].
12725         PR c++/86025
12726         * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
12728 2018-06-03  Eric Botcazou  <ebotcazou@adacore.com>
12730         PR tree-optimization/86034
12731         * gimple-ssa-store-merging.c (output_merged_store): Convert the RHS to
12732         the unsigned bitfield type in a bit insertion sequence if it does not
12733         have a larger precision than the bitfield size.
12734         (process_store): Also bypass widening conversions for BIT_INSERT_EXPR.
12736 2018-06-03  Kito Cheng  <kito.cheng@gmail.com>
12738         * config/nds32/nds32-peephole2.md: Add new patterns for code size.
12740 2018-06-03  Chung-Ju Wu  <jasonwucj@gmail.com>
12742         * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3J.
12743         * config/nds32/nds32.c (nds32_option_override): Consider ARCH_V3J.
12744         * config/nds32/nds32.h (TARGET_ISA_V3): Add ARCH_V3J checking.
12745         * config/nds32/nds32.opt (march): Add enum value Value(ARCH_V3J).
12747 2018-06-03  Chung-Ju Wu  <jasonwucj@gmail.com>
12749         * common/config/nds32/nds32-common.c (nds32_option_optimization_table):
12750         Disable -fdelete-null-pointer-checks for ELF toolchain.
12752 2018-06-02  Chung-Ju Wu  <jasonwucj@gmail.com>
12753             Kito Cheng  <kito.cheng@gmail.com>
12755         * config.gcc (nds32*): Use nds32-linux.opt and nds32-elf.opt.
12756         (nds32le-*-*, nds32be-*-*): Integrate checking process.
12757         (nds32*-*-*): Add glibc and uclibc conditions.
12758         * common/config/nds32/nds32-common.c (nds32_except_unwind_info): New.
12759         (TARGET_EXCEPT_UNWIND_INFO): Define.
12760         * config/nds32/elf.h: New file.
12761         * config/nds32/linux.h: New file.
12762         * config/nds32/nds32-elf.opt: New file.
12763         * config/nds32/nds32-linux.opt: New file.
12764         * config/nds32/nds32-fp-as-gp.c
12765         (pass_nds32_fp_as_gp::gate): Consider TARGET_LINUX_ABI.
12766         * config/nds32/nds32.c (nds32_conditional_register_usage): Consider
12767         TARGET_LINUX_ABI.
12768         (nds32_asm_file_end): Ditto.
12769         (nds32_print_operand): Ditto.
12770         (nds32_insert_attributes): Ditto.
12771         (nds32_init_libfuncs): New function.
12772         (TARGET_HAVE_TLS): Define.
12773         (TARGET_INIT_LIBFUNCS): Define.
12774         * config/nds32/nds32.h (TARGET_DEFAULT_RELAX): Apply different relax
12775         spec content.
12776         (TARGET_ELF): Apply different mcmodel setting.
12777         (LINK_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): The content has
12778         been migrated into elf.h and linux.h files.
12779         * config/nds32/nds32.md (add_pc): Consider TARGET_LINUX_ABI.
12780         * config/nds32/nds32.opt (mvh): Consider TARGET_LINUX_ABI.
12781         (mcmodel): The content has been migrated into nds32-elf.opt and
12782         nds32-linux.opt files.
12783         * config/nds32/t-elf: New file.
12784         * config/nds32/t-linux: New file.
12786 2018-06-02  Chung-Ju Wu  <jasonwucj@gmail.com>
12787             Shiva Chen  <shiva0217@gmail.com>
12789         * config/nds32/constants.md (unspec_volatile_element): Add
12790         UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END.
12791         * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp
12792         optimization.
12793         * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare.
12794         (make_pass_nds32_fp_as_gp): Declare.
12795         * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one
12796         optmization pass.
12797         (nds32_asm_function_end_prologue): Remove unused asm output.
12798         (nds32_asm_function_begin_epilogue): Remove unused asm output.
12799         (nds32_asm_file_start): Output necessary fp_as_gp information.
12800         (nds32_option_override): Adjust register usage.
12801         (nds32_expand_prologue): Consider fp_as_gp situation.
12802         (nds32_expand_prologue_v3push): Consider fp_as_gp situation.
12803         * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p.
12804         (epilogue): Ditto.
12805         (return): Ditto.
12806         (simple_return): Ditto.
12807         (omit_fp_begin): Output special directive for fp_as_gp.
12808         (omit_fp_end): Output special directive for fp_as_gp.
12809         * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp,
12810         mforbid-fp-as-gp): New options.
12812 2018-06-01  Mark Wielaard  <mark@klomp.org>
12814         * dwarf2out.c (dwarf2out_finish): Remove generation of
12815         DW_AT_loclists_base.
12817 2018-06-01  Eric Botcazou  <ebotcazou@adacore.com>
12819         * gimple-ssa-store-merging.c: Include gimple-fold.h.
12820         (struct store_immediate_info): Document BIT_INSERT_EXPR stores.
12821         (struct merged_store_group): Add bit_insertion field.
12822         (dump_char_array): Use standard hexadecimal format.
12823         (merged_store_group::merged_store_group): Set bit_insertion to false.
12824         (merged_store_group::apply_stores): Use optimal buffer size.  Deal
12825         with BIT_INSERT_EXPR stores.  Move up code updating the mask and
12826         also print the mask in the dump file.
12827         (pass_store_merging::gate): Minor tweak.
12828         (imm_store_chain_info::coalesce_immediate): Fix wrong association
12829         of stores with groups in dump.  Allow coalescing of BIT_INSERT_EXPR
12830         stores with INTEGER_CST stores.
12831         (count_multiple_uses) <BIT_INSERT_EXPR>: New case.
12832         (imm_store_chain_info::output_merged_store): Add try_bitpos variable
12833         and use it throughout.  Generate bit insertion sequences if need be.
12834         (pass_store_merging::process_store): Remove redundant condition.
12835         Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR.
12837 2018-06-01  Segher Boessenkool  <segher@kernel.crashing.org>
12839         * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
12840         the 128-bit floating point types.  Fix function comment.
12842 2018-06-01  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
12844         * config/aarch64/aarch64-simd.md
12845         (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler
12846         mnemonics.
12847         (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
12848         mnemonics.
12850 2018-06-01  Richard Sandiford  <richard.sandiford@linaro.org>
12852         PR tree-optimization/85989
12853         * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
12854         variable.
12855         (backprop::intersect_uses): Check it when deciding whether this
12856         is a backedge reference.
12857         (backprop::process_block): Add each phi to m_visited_phis
12858         after visiting it, then clear it at the end.
12860 2018-06-01  Richard Biener  <rguenther@suse.de>
12862         * tree-vectorizer.h (vect_dr_stmt): New function.
12863         (vect_get_load_cost): Adjust.
12864         (vect_get_store_cost): Likewise.
12865         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
12866         Use vect_dr_stmt instead of DR_SMTT.
12867         (vect_record_base_alignments): Likewise.
12868         (vect_calculate_target_alignment): Likewise.
12869         (vect_compute_data_ref_alignment): Likewise and make static.
12870         (vect_update_misalignment_for_peel): Likewise.
12871         (vect_verify_datarefs_alignment): Likewise.
12872         (vector_alignment_reachable_p): Likewise.
12873         (vect_get_data_access_cost): Likewise.  Pass down
12874         vinfo to vect_get_load_cost/vect_get_store_cost instead of DR.
12875         (vect_get_peeling_costs_all_drs): Likewise.
12876         (vect_peeling_hash_get_lowest_cost): Likewise.
12877         (vect_enhance_data_refs_alignment): Likewise.
12878         (vect_find_same_alignment_drs): Likewise.
12879         (vect_analyze_data_refs_alignment): Likewise.
12880         (vect_analyze_group_access_1): Likewise.
12881         (vect_analyze_group_access): Likewise.
12882         (vect_analyze_data_ref_access): Likewise.
12883         (vect_analyze_data_ref_accesses): Likewise.
12884         (vect_vfa_segment_size): Likewise.
12885         (vect_small_gap_p): Likewise.
12886         (vectorizable_with_step_bound_p): Likewise.
12887         (vect_prune_runtime_alias_test_list): Likewise.
12888         (vect_analyze_data_refs): Likewise.
12889         (vect_supportable_dr_alignment): Likewise.
12890         * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
12891         (vect_gen_prolog_loop_niters): Likewise.
12892         * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
12893         * tree-vect-patterns.c (vect_recog_bool_pattern): Do not
12894         modify DR_STMT.
12895         (vect_recog_mask_conversion_pattern): Likewise.
12896         (vect_try_gather_scatter_pattern): Likewise.
12897         * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info
12898         to vect_get_store_cost.
12899         (vect_get_store_cost): Get stmt_info instead of DR.
12900         (vect_model_load_cost): Pass stmt_info to vect_get_load_cost.
12901         (vect_get_load_cost): Get stmt_info instead of DR.
12903 2018-06-01  Richard Biener  <rguenther@suse.de>
12905         PR middle-end/86017
12906         * gimple-fold.c (var_decl_component_p): Also allow offsetted
12907         vars wrapped in MEM_REFs.
12909 2018-06-01  Richard Sandiford  <richard.sandiford@linaro.org>
12911         * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
12912         Fix subreg tests so that we only return a choice between
12913         GENERAL_REGS and FP_REGS if the original classes included both.
12915 2018-06-01  Richard Biener  <rguenther@suse.de>
12917         PR ipa/85960
12918         * tree-ssa-structalias.c (get_function_part_constraint):
12919         Handle NULL fi->decl.
12920         (find_func_aliases_for_call): Properly handle indirect
12921         fi from direct call.
12922         (find_func_clobbers): Likewise.
12923         (ipa_pta_execute): Likewise.
12924         (create_variable_info_for): For functions that are ifunc_resolver
12925         resolve to a varinfo that contains the result of the resolver call.
12926         (associate_varinfo_to_alias): Do not treat ifunc resolvers as
12927         aliases.
12929 2018-05-31  Michael Collison  <michael.collison@arm.com>
12931         * config/aarch64/aarch64.md:
12932         (*fix_to_zero_extenddfdi2): New pattern.
12933         * gcc.target/aarch64/fix_extend1.c: New testcase.
12935 2018-05-31  Qing Zhao <qing.zhao@oracle.com>
12937         PR middle-end/78809
12938         PR middle-end/83026
12939         * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
12940         and BUILT_IN_STRNCMP_EQ.
12941         * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
12942         BUILT_IN_STRNCMP_EQ.
12943         * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
12944         handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
12945         (gimple_fold_builtin): Likewise.
12946         * tree-ssa-strlen.c (compute_string_length): New function.
12947         (determine_min_obsize): New function.
12948         (handle_builtin_string_cmp): New function to handle calls to
12949         string compare functions.
12950         (strlen_optimize_stmt): Add handling to builtin string compare
12951         calls.
12952         * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
12953         Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
12954         * tree.c (build_common_builtin_nodes): Add new defines of
12955         BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
12957 2018-05-31  Jakub Jelinek  <jakub@redhat.com>
12959         PR target/85984
12960         * bb-reorder.c (pass_partition_blocks::gate): Return false for
12961         functions with naked attribute.
12963 2018-05-31  Uros Bizjak  <ubizjak@gmail.com>
12965         * config/i386/sse.md (avx_vec_concat<mode>):
12966         Substitute concat_tg_mode mode attribute with xtg_mode.
12967         (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
12968         (concat_tg_mode): Remove mode attribute.
12970 2018-05-31  Martin Sebor  <msebor@redhat.com>
12972         PR c/82063
12973         * calls.c (alloc_max_size): Correct a logic error/typo.
12974         Treat excessive arguments as infinite.  Warn for invalid arguments.
12975         * doc/invoke.texi (-Walloc-size-larger-than): Update.
12977 2018-05-31  H.J. Lu  <hongjiu.lu@intel.com>
12979         PR target/85829
12980         * config/i386/x86-tune.def: Re-enable partial_reg_dependency
12981         and movx for Haswell.
12983 2018-05-31  Chung-Lin Tang  <cltang@codesourcery.com>
12984             Cesar Philippidis  <cesar@codesourcery.com>
12986         PR middle-end/85879
12987         * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
12988         when emitting error on private/firstprivate reductions.
12989         * omp-low.c (lower_omp_target): Avoid reference-type processing
12990         on pointers for firstprivate clause.
12992 2018-05-31  Sameera Deshpande <sameera.deshpande@linaro.org>
12994         * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
12995         (st1x2): Likewise.
12996         (st1x3): Likewise.
12997         * config/aarch64/aarch64-simd.md
12998         (aarch64_ld1x3<VALLDIF:mode>): New pattern.
12999         (aarch64_ld1_x3_<mode>): Likewise
13000         (aarch64_st1x2<VALLDIF:mode>): Likewise
13001         (aarch64_st1_x2_<mode>): Likewise
13002         (aarch64_st1x3<VALLDIF:mode>): Likewise
13003         (aarch64_st1_x3_<mode>): Likewise
13004         * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
13005         (vld1_s8_x3): Likewise.
13006         (vld1_u16_x3): Likewise.
13007         (vld1_s16_x3): Likewise.
13008         (vld1_u32_x3): Likewise.
13009         (vld1_s32_x3): Likewise.
13010         (vld1_u64_x3): Likewise.
13011         (vld1_s64_x3): Likewise.
13012         (vld1_f16_x3): Likewise.
13013         (vld1_f32_x3): Likewise.
13014         (vld1_f64_x3): Likewise.
13015         (vld1_p8_x3): Likewise.
13016         (vld1_p16_x3): Likewise.
13017         (vld1_p64_x3): Likewise.
13018         (vld1q_u8_x3): Likewise.
13019         (vld1q_s8_x3): Likewise.
13020         (vld1q_u16_x3): Likewise.
13021         (vld1q_s16_x3): Likewise.
13022         (vld1q_u32_x3): Likewise.
13023         (vld1q_s32_x3): Likewise.
13024         (vld1q_u64_x3): Likewise.
13025         (vld1q_s64_x3): Likewise.
13026         (vld1q_f16_x3): Likewise.
13027         (vld1q_f32_x3): Likewise.
13028         (vld1q_f64_x3): Likewise.
13029         (vld1q_p8_x3): Likewise.
13030         (vld1q_p16_x3): Likewise.
13031         (vld1q_p64_x3): Likewise.
13032         (vst1_s64_x2): Likewise.
13033         (vst1_u64_x2): Likewise.
13034         (vst1_f64_x2): Likewise.
13035         (vst1_s8_x2): Likewise.
13036         (vst1_p8_x2): Likewise.
13037         (vst1_s16_x2): Likewise.
13038         (vst1_p16_x2): Likewise.
13039         (vst1_s32_x2): Likewise.
13040         (vst1_u8_x2): Likewise.
13041         (vst1_u16_x2): Likewise.
13042         (vst1_u32_x2): Likewise.
13043         (vst1_f16_x2): Likewise.
13044         (vst1_f32_x2): Likewise.
13045         (vst1_p64_x2): Likewise.
13046         (vst1q_s8_x2): Likewise.
13047         (vst1q_p8_x2): Likewise.
13048         (vst1q_s16_x2): Likewise.
13049         (vst1q_p16_x2): Likewise.
13050         (vst1q_s32_x2): Likewise.
13051         (vst1q_s64_x2): Likewise.
13052         (vst1q_u8_x2): Likewise.
13053         (vst1q_u16_x2): Likewise.
13054         (vst1q_u32_x2): Likewise.
13055         (vst1q_u64_x2): Likewise.
13056         (vst1q_f16_x2): Likewise.
13057         (vst1q_f32_x2): Likewise.
13058         (vst1q_f64_x2): Likewise.
13059         (vst1q_p64_x2): Likewise.
13060         (vst1_s64_x3): Likewise.
13061         (vst1_u64_x3): Likewise.
13062         (vst1_f64_x3): Likewise.
13063         (vst1_s8_x3): Likewise.
13064         (vst1_p8_x3): Likewise.
13065         (vst1_s16_x3): Likewise.
13066         (vst1_p16_x3): Likewise.
13067         (vst1_s32_x3): Likewise.
13068         (vst1_u8_x3): Likewise.
13069         (vst1_u16_x3): Likewise.
13070         (vst1_u32_x3): Likewise.
13071         (vst1_f16_x3): Likewise.
13072         (vst1_f32_x3): Likewise.
13073         (vst1_p64_x3): Likewise.
13074         (vst1q_s8_x3): Likewise.
13075         (vst1q_p8_x3): Likewise.
13076         (vst1q_s16_x3): Likewise.
13077         (vst1q_p16_x3): Likewise.
13078         (vst1q_s32_x3): Likewise.
13079         (vst1q_s64_x3): Likewise.
13080         (vst1q_u8_x3): Likewise.
13081         (vst1q_u16_x3): Likewise.
13082         (vst1q_u32_x3): Likewise.
13083         (vst1q_u64_x3): Likewise.
13084         (vst1q_f16_x3): Likewise.
13085         (vst1q_f32_x3): Likewise.
13086         (vst1q_f64_x3): Likewise.
13087         (vst1q_p64_x3): Likewise.
13089 2018-05-30  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
13091         * config/msp430/msp430.c (msp430_output_labelref): Prepend
13092         user_label_prefix to name.
13094         * tree-core.h: Update comment about the format of NAME string
13095         passed to handler in attribute_spec.
13097         * config/msp430/msp430.md: Remove erroneous subreg expression from
13098         zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
13099         zero_extend{q,h}isi2.
13101 2018-05-30  Borislav Petkov  <bp@suse.de>
13103         * doc/extend.texi: Document some architecture specific
13104         constraints and sort entries.
13106 2018-05-30  Martin Sebor  <msebor@redhat.com>
13108         PR middle-end/85369
13109         * builtins.c (expand_builtin_stpcpy_1): New function.
13110         (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
13111         only if the former succeeds.
13113 2018-05-31  Sameera Deshpande <sameera.deshpande@linaro.org>
13115         * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
13116         in saphira.
13118 2018-05-30  Jan Hubicka  <hubicka@ucw.cz>
13120         * doc/invoke.texi (-flinker-output): Document
13122 2018-05-30  Jan Hubicka  <hubicka@ucw.cz>
13124         * passes.c (ipa_write_summaries): Only modify statements if body
13125         is in memory.
13126         * cgraphunit.c (ipa_passes): Also produce intermeidate code when
13127         incrementally linking.
13128         (ipa_passes): Likewise.
13129         * lto-cgraph.c (lto_output_node): When incrementally linking do not
13130         pass down resolution info.
13131         * common.opt (flag_incremental_link): Update info.
13132         * gcc.c (plugin specs): Turn flinker-output=* to
13133         -plugin-opt=-linker-output-known
13134         * toplev.c (compile_file): Also cut compilation when doing incremental
13135         link.
13136         * flag-types. (enum lto_partition_model): Add
13137         LTO_LINKER_OUTPUT_NOLTOREL.
13138         (invoke.texi): Add -flinker-output docs.
13139         * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
13140         link same way as WPA; do not stream in dead initializers.
13142         * dwarf2out.c (dwarf2out_die_ref_for_decl,
13143         darf2out_register_external_decl): Support incremental link.
13145 2018-05-30  Jan Hubicka  <hubicka@ucw.cz>
13147         * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
13149 2018-05-30  Jan Hubicka  <hubicka@ucw.cz>
13151         * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
13152         it down to simple_object_copy_lto_debug_sections.
13153         (run_gcc): Determine incremental LTO link time and configure
13154         lto1 into non-wpa mode, disable renaming of debug sections.
13156 2018-05-30  Kelvin Nilsen  <kelvin@gcc.gnu.org>
13158         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
13159         descriptions of various incorrectly documented functions.
13161 2018-05-30  Andre Vieira  <andre.simoesdiasvieira@arm.com>
13163         Revert:
13164         * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
13165         address check not strict.
13167 2018-05-30  Richard Biener  <rguenther@suse.de>
13169         PR tree-optimization/85964
13170         * tracer.c (better_p): Drop initialized count check, we only
13171         call the function with initialized counts now.
13172         (find_best_successor): Do find a best edge if one
13173         has uninitialized count.
13174         (find_best_predecessor): Likewise.  Do BB frequency check only
13175         if count is initialized.
13177 2017-05-30  Jackson Woodruff  <jackson.woodruff@arm.com>
13179         * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
13180         (aarch64_ldrstr_offset_compare): New.
13181         (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
13182         load/store orderings.
13183         (aarch64_gen_adjusted_ldpstp): Likewise.
13185 2018-05-30  Wilco Dijkstra  <wdijkstr@arm.com>
13187         * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
13188         Check for subset of GENERAL_REGS and FP_REGS.
13189         * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
13190         r=w alternative.
13192 2018-05-30  Richard Sandiford  <richard.sandiford@linaro.org>
13194         * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
13195         and wi::to_poly_offset.  Add the current offset and then check
13196         whether the sum fits, rather than using an unchecked addition of
13197         a checked term.  Check for a shwi rather than a uhwi.
13198         * expr.c (get_bit_range): Use tree_to_poly_uint64.
13199         (store_constructor): Use poly_int_tree_p.
13200         (expand_expr_real_1): Likewise.
13201         * function.c (assign_temp): Likewise.
13202         * fold-const.c (const_binop): Use poly_int_tree_p and
13203         wi::to_poly_offset.
13204         (fold_indirect_ref_1): Likewise.  Use multiple_p to attempt an exact
13205         division.
13206         * ipa-icf-gimple.c (func_checker::compare_operand): Use
13207         to_poly_offset for MEM offsets.
13208         * ipa-icf.c (sem_variable::equals): Likewise.
13209         * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
13210         * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
13211         wi::to_poly_offset for BIT_FIELD_REF offsets.
13212         (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
13213         wi::to_poly_offset.
13214         * var-tracking.c (emit_note_insn_var_location): Use
13215         tree_to_poly_uint64.
13217 2018-05-29  Jim Wilson  <jimw@sifive.com>
13219         * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
13221 2018-05-29  Uros Bizjak  <ubizjak@gmail.com>
13223         PR target/85950
13224         * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
13225         Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
13226         sequence.
13227         (sse4_1_round<mode>2): Use nonimmediate_operand
13228         for operand 1 predicate.
13230 2018-05-29  Martin Sebor  <msebor@redhat.com>
13231             Richard Biener  <rguenther@suse.de>
13233         PR testsuite/85888
13234         * calls.c (get_size_range): Call determine_value_range instead
13235         of get_value_range..
13236         * tree-vrp.h (determine_value_range): Declared new function.
13237         * tree-vrp.c (determine_value_range_1, determine_value_range): New.
13239 2018-05-29  Richard Biener  <rguenther@suse.de>
13241         * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
13242         sure to use non-pattern stmts for get_earlier_stmt arguments.
13243         * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
13244         called on pattern stmts.
13245         (get_later_stmt): Likewise.
13247 2018-05-29  Martin Liska  <mliska@suse.cz>
13249         PR gcov-profile/85759
13250         * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
13251         env variables.
13253 2018-05-29  Jakub Jelinek  <jakub@redhat.com>
13255         * tree-cfg.c (verify_gimple_assign_unary): Add checking for
13256         VEC_UNPACK_*_EXPR.
13257         (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
13258         VEC_PACK_*_EXPR.
13260         PR target/85918
13261         * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
13262         VEC_PACK_FLOAT_EXPR): New tree codes.
13263         * tree-pretty-print.c (op_code_prio): Handle
13264         VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
13265         (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
13266         VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
13267         * tree-inline.c (estimate_operator_cost): Likewise.
13268         * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
13269         * fold-const.c (const_binop): Likewise.
13270         (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
13271         VEC_UNPACK_FIX_TRUNC_LO_EXPR.
13272         * tree-cfg.c (verify_gimple_assign_unary): Likewise.
13273         (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
13274         * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
13275         VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
13276         * expr.c (expand_expr_real_2): Likewise.
13277         * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
13278         vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
13279         vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
13280         optabs.
13281         * optabs.c (expand_widen_pattern_expr): For
13282         VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
13283         sign from result type rather than operand's type.
13284         (expand_binop_directly): For vec_packu_float_optab and
13285         vec_packs_float_optab allow result type to be different from operand's
13286         type.
13287         * optabs-tree.c (optab_for_tree_code): Handle
13288         VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
13289         VEC_PACK_FLOAT_EXPR.  Formatting fixes.
13290         * tree-vect-generic.c (expand_vector_operations_1):  Handle
13291         VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
13292         VEC_PACK_FLOAT_EXPR.
13293         * tree-vect-stmts.c (supportable_widening_operation): Handle
13294         FIX_TRUNC_EXPR.
13295         (supportable_narrowing_operation): Handle FLOAT_EXPR.
13296         * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
13297         * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
13298         (float<floatunssuffix>v2div2sf2): ... this.  Formatting fix.
13299         (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
13300         mode attributes.
13301         (vec_pack<floatprefix>_float_<mode>): New expander.
13302         (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
13303         attributes.
13304         (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
13305         vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
13306         * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
13307         vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
13308         vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
13309         Document.
13310         * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
13311         VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
13312         (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
13313         VEC_PACK_FLOAT_EXPR): Document.
13315 2018-05-29  Richard Biener  <rguenther@suse.de>
13317         * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
13318         member.
13319         (stmt_vec_info_vec): Make pointer.
13320         (init_stmt_vec_info_vec): Remove.
13321         (free_stmt_vec_info_vec): Likewise.
13322         (set_stmt_vec_info_vec): New function.
13323         (free_stmt_vec_infos): Likewise.
13324         (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
13325         (set_vinfo_for_stmt): Likewise.
13326         (get_earlier_stmt): Likewise.
13327         (get_later_stmt): Likewise.
13328         * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
13329         (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
13330         (vec_info::~vec_info): Free stmt_vec_infos.
13331         (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
13332         Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
13333         (pass_slp_vectorize::execute): Likewise.
13334         * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
13335         (free_stmt_vec_info_vec): Likewise.
13336         (set_stmt_vec_info_vec): New function.
13337         (free_stmt_vec_infos): Likewise.
13338         * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
13339         the global stmt_vec_info_vec.
13340         * tree-parloops.c (gather_scalar_reductions): Use
13341         set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
13342         vector.
13344 2018-05-29  Richard Biener  <rguenther@suse.de>
13346         * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
13348 2018-05-29  Martin Liska  <mliska@suse.cz>
13349             David Malcolm  <dmalcolm@redhat.com>
13351         * vec.c (test_reverse): New.
13352         (vec_c_tests): Add new test.
13353         * vec.h (vl_ptr>::reverse): New function.
13355 2018-05-29  Gerald Pfeifer  <gerald@pfeifer.com>
13357         * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
13359         * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
13360         and later.
13362 2018-05-28  Bernd Edlinger  <bernd.edlinger@hotmail.de>
13364         * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
13366 2018-05-28  Richard Biener  <rguenther@suse.de>
13368         PR tree-optimization/85933
13369         * tree-vect-data-refs.c (vect_record_base_alignments): Only
13370         look at stmts marked as vectorizable.
13372 2018-05-28  Richard Biener  <rguenther@suse.de>
13374         PR tree-optimization/85934
13375         * tree-vect-generic.c (expand_vector_operations_1): Hoist
13376         vector boolean check before scalar optimization.
13378 2018-05-28  Jakub Jelinek  <jakub@redhat.com>
13380         * doc/invoke.texi (ARM Options): Use @item instead of @itemx
13381         for armv5te.
13383 2018-05-28  Mark Wielaard  <mark@klomp.org>
13385         * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
13386         if it is an expression containing a minus sign.
13388 2018-05-27  John David Anglin  <danglin@gcc.gnu.org>
13390         * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
13392 2018-05-27  Paul Koning  <ni1d@arrl.net>
13394         * config/pdp11/pdp11.md (truncsihi2): Remove.
13396 2018-05-27  Monk Chiang  <sh.chiang04@gmail.com>
13397             Chung-Ju Wu  <jasonwucj@gmail.com>
13399         * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
13400         implementation.
13401         (unaligned_store_dw): Ditto.
13402         * config/nds32/nds32-memory-manipulation.c
13403         (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
13404         (nds32_gen_dup_4_byte_to_word_value): Rename to ...
13405         (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
13406         (emit_setmem_word_loop): Rename to ...
13407         (emit_setmem_doubleword_loop): ... this.
13408         (nds32_gen_dup_4_byte_to_word_value): New function.
13409         (nds32_gen_dup_8_byte_to_double_word_value): New function.
13410         (nds32_expand_setmem_loop): Refine implementation.
13411         (nds32_expand_setmem_loop_v3m): Ditto.
13412         * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
13413         pattern.
13415 2018-05-27  Chung-Ju Wu  <jasonwucj@gmail.com>
13417         * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
13419 2018-05-27  Chung-Ju Wu  <jasonwucj@gmail.com>
13421         * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
13422         (nds32_init_machine_status): Initialize machine->attr_naked_p and
13423         machine->attr_no_prologue_p.
13424         (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
13425         (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
13426         (nds32_expand_epilogue): Consider attr_naked_p.
13427         (nds32_expand_epilogue_v3pop): Likewise.
13428         (nds32_can_use_return_insn): Likewise.
13429         * config/nds32/nds32.h (machine_function): Add attr_naked_p and
13430         attr_no_prologue_p fields.
13431         * config/nds32/nds32.opt (mret-in-naked-func): New option.
13433 2018-05-27  Jakub Jelinek  <jakub@redhat.com>
13435         PR target/85918
13436         * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
13437         attributes.
13438         * config/i386/sse.md
13439         (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
13440         Rename to ...
13441         (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
13442         ... this.
13443         (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
13444         Rename to ...
13445         (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
13446         ... this.
13447         (*<floatsuffix>floatv2div2sf2): Rename to ...
13448         (*float<floatunssuffix>v2div2sf2): ... this.
13449         (<floatsuffix>floatv2div2sf2_mask): Rename to ...
13450         (float<floatunssuffix>v2div2sf2_mask): ... this.
13451         (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
13452         (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
13453         (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
13454         to ...
13455         (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
13456         ... this.
13457         (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
13458         Rename to ...
13459         (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
13460         ... this.
13461         (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
13462         Rename to ...
13463         (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
13464         ... this.
13465         (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
13466         (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
13467         (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
13468         gen_ufix_truncv8dfv8si2.
13469         * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
13470         __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
13471         __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
13472         __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
13473         __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
13474         __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
13475         __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
13476         Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
13478 2018-05-24  H.J. Lu  <hongjiu.lu@intel.com>
13480         PR target/85900
13481         PR target/85345
13482         * varasm.c (assemble_alias): Lookup ifunc attribute on error.
13484 2018-05-25  Jim Wilson  <jimw@sifive.com>
13486         * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
13487         * config/riscv/riscv.c (struct machine_function): Add
13488         interrupt_handler_p and attribute_checked_p fields.
13489         (riscv_attribute_table): Add interrupt.
13490         (riscv_interrupt_type_p): New.
13491         (riscv_save_reg_p): Save extra regs for interrupt handler.
13492         (riscv_use_save_libcall): Return false  for interrupt handler.
13493         (riscv_first_stack_step): Add forward declaration.
13494         (riscv_compute_frame_info): New local interrupt_save_t1.  Set it
13495         for interrupt handler with large frame.  Use it for saved reg list.
13496         (riscv_expand_prologue): Move flag_stack_usage_info support to
13497         eliminate duplication.
13498         (riscv_expand_epilogue): Generate mret for interrupt handler.
13499         (riscv_epilogue_uses): New.
13500         (riscv_can_use_return_insn): Return false for interrupt handler.
13501         (riscv_function_ok_for_sibcall): Likewise.
13502         (riscv_set_current_function): Add interrupt handler support.
13503         * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
13504         * config/riscv/riscv.md (UNSPECV_MRET): New.
13505         (GP_REGNUM): New.
13506         (riscv_frflags, riscv_fsflags): Use tab after opcode.
13507         (riscv_mret): New.
13508         * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
13510 2018-05-25  Bill Schmidt  <wschmidt@linux.ibm.com>
13512         PR tree-optimization/85712
13513         * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
13514         this candidate has already been replaced in-situ by a copy.
13516 2018-05-25  Jason Merrill  <jason@redhat.com>
13518         PR c++/80485 - inline function non-zero address.
13519         * symtab.c (nonzero_address): Check DECL_COMDAT.
13521 2018-05-25  Uros Bizjak  <ubizjak@gmail.com>
13523         PR target/83628
13524         * config/alpha/alpha.md (ashlsi3): New insn pattern.
13525         (*ashlsi_se): Rename from *ashldi_se.  Define as sign
13526         extension of SImode operation.  Use const123_operand predicate.
13527         (*saddsi_1): Remove.
13528         (*saddl_se_1): Ditto.
13529         (*ssubsi_1): Ditto.
13530         (*ssubl_se_1): Ditto.
13531         * config/alpha/predicates.md (const123_operand): New predicate.
13532         * config/alpha/constraints.md (P): Use IN_RANGE.
13534 2018-05-25  Richard Biener  <rguenther@suse.de>
13536         * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
13537         defaulted to true.
13538         (ref_maybe_used_by_stmt_p): Likewise.
13539         (stmt_may_clobber_ref_p): Likewise.
13540         (stmt_may_clobber_ref_p_1): Likewise.
13541         * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
13542         and pass it along.
13543         (ref_maybe_used_by_stmt_p): Likewise.
13544         (stmt_may_clobber_ref_p): Likewise.
13545         (stmt_may_clobber_ref_p_1): Likewise.
13546         * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
13547         the alias oracle to disambiguate DRs with stmts DR analysis
13548         couldn't handle.
13549         (vect_analyze_data_refs): Do not give up on not analyzable
13550         DRs for BB vectorization.  Remove code truncating the dataref
13551         vector.
13553 2018-05-25  Jakub Jelinek  <jakub@redhat.com>
13555         PR target/85832
13556         * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
13557         Add (=Yk,v,C) variant using vptestm insn.  Use TARGET_AVX512BW
13558         in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
13560 2018-05-25  Richard Biener  <rguenther@suse.de>
13562         * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
13563         function, combining stmt data ref gathering and fatal analysis
13564         parts.
13565         (vect_analyze_data_refs): Remove now redudnant code and simplify.
13566         * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
13567         vect_analyze_loop_2 and use vect_find_stmt_data_reference.
13568         * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
13569         * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
13571 2018-05-25  Bin Cheng  <bin.cheng@arm.com>
13573         PR tree-optimization/85720
13574         * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
13575         SCC if all partitions are builtins.
13576         (version_loop_by_alias_check): New parameter.  Generate cancelable
13577         runtime alias check if all partitions are builtins.
13578         (distribute_loop): Update call to above function.
13580 2018-05-25  Bin Cheng  <bin.cheng@arm.com>
13582         * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
13583         (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
13584         (parm_default_def_partition_arg): Ditto.
13585         (set_parm_default_def_partition): Ditto.
13586         (get_parm_default_def_partitions): Ditto and make it static.
13587         (get_undefined_value_partitions): Ditto and make it static.
13588         (remove_ssa_form): Refactor call to init_var_map here.
13589         * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
13590         computation for loop region.
13591         (coalesce_partitions, compute_optimized_partition_bases): Ditto.
13592         (register_default_def): Delete.
13593         (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
13594         (parm_default_def_partition_arg): Ditto.
13595         (set_parm_default_def_partition): Ditto.
13596         (get_parm_default_def_partitions): Ditto and make it static.
13597         (get_undefined_value_partitions): Ditto and make it static.
13598         (coalesce_with_default, coalesce_with_default): Update comment.
13599         (create_coalesce_list_for_region): New func factored out from
13600         create_outofssa_var_map.
13601         (populate_coalesce_list_for_outofssa): New func factored out from
13602         create_outofssa_var_map and coalesce_ssa_name.
13603         (create_outofssa_var_map): Delete.
13604         (coalesce_ssa_name): Refactor to support live range computation.
13605         * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
13606         (get_parm_default_def_partitions): Delete.
13607         (get_undefined_value_partitions): Ditto.
13608         * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
13609         computation for loop region.
13610         (new_tree_live_info, loe_visit_block): Ditto.
13611         (live_worklist, set_var_live_on_entry): Ditto.
13612         (calculate_live_on_exit, verify_live_on_entry): Ditto.
13613         * tree-ssa-live.h (struct _var_map): New fields.
13614         (init_var_map): Change decl.
13615         (region_contains_p): New.
13617 2018-05-25  Bin Cheng  <bin.cheng@arm.com>
13619         * tree-ssa-live.h (live_merge_and_clear): Delete.
13621 2018-05-25  Richard Biener  <rguenther@suse.de>
13623         PR c++/85912
13624         * tree-dump.c (dequeue_and_dump): Remove access to removed
13625         operand 2 of a SWITCH_EXPR.
13627 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
13629         * doc/sourcebuild.texi (vect_double_cond_arith): Include
13630         multiplication and division.
13631         * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
13632         (cond_udiv@var{m}, cond_umod@var{m}): Document.
13633         * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
13634         (cond_udiv_optab, cond_umod_optab): New optabs.
13635         * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
13636         (IFN_COND_RDIV): New internal functions.
13637         * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
13638         TRUNC_MOD_EXPR and RDIV_EXPR.
13639         * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
13640         * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
13641         New unspecs.
13642         (SVE_INT_BINARY): Include mult.
13643         (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
13644         (optab, sve_int_op): Handle mult.
13645         (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
13646         UNSPEC_COND_DIV.
13647         * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
13648         for SVE_INT_BINARY_SD.
13650 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
13652         * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
13653         (optab, sve_int_op): Handle div and udiv.
13654         * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
13655         for SVE_INT_BINARY_SD.
13656         (*<optab><mode>3): New insn for the same.
13658 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
13660         * tree-vect-patterns.c: Include predict.h.
13661         (vect_recog_divmod_pattern): Restrict check for division support
13662         to when optimizing for size.
13664 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
13666         * doc/sourcebuild.texi (vect_double_cond_arith: Document.
13667         * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
13668         (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
13669         (gimple_match_op::set_op): Likewise.
13670         (gimple_resimplify4): Declare.
13671         * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
13672         (expr::gen_transform): Likewise.
13673         (decision_tree::gen): Generate a simplification routine for 4 operands.
13674         * gimple-match-head.c (gimple_simplify): Add an overload for
13675         4 operands.  In the top-level function, handle up to 4 call
13676         arguments and call gimple_resimplify4.
13677         (gimple_resimplify4): New function.
13678         (build_call_internal): Pass a fourth operand.
13679         (maybe_push_to_seq): Likewise.
13680         * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
13681         Fold VEC_COND_EXPRs of an operation and a default value into
13682         an IFN_COND_* function if possible.
13683         * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
13684         New unspecs.
13685         (SVE_COND_FP_BINARY): Include them.
13686         (optab, sve_fp_op): Handle them.
13687         (SVE_INT_BINARY_REV): New code iterator.
13688         (SVE_COND_FP_BINARY_REV): New int iterator.
13689         (commutative): New int attribute.
13690         * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
13691         Declare.
13692         * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
13693         function.
13694         * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
13695         (*cond_<optab><mode>): New patterns for reversed operands.
13697 2018-05-25  Richard Biener  <rguenther@suse.de>
13699         * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
13700         (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
13701         (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
13702         (STMT_VINFO_GROUPED_ACCESS): Adjust.
13703         * tree-vect-data-refs.c (everywhere): Adjust users.
13704         * tree-vect-loop.c (everywhere): Likewise.
13705         * tree-vect-slp.c (everywhere): Likewise.
13706         * tree-vect-stmts.c (everywhere): Likewise.
13707         * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
13709 2018-05-25  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
13711         * configure.ac (gcc_cv_as_section_has_e): Move to common section.
13712         Rename to...
13713         (gcc_cv_as_section_exclude): ... this.
13714         Try Solaris as #exclude syntax.
13715         * configure: Regenerate.
13716         * config.in: Regenerate.
13717         * config/i386/i386.c (i386_solaris_elf_named_section): Handle
13718         SECTION_EXCLUDE.
13719         * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
13720         [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
13722         * varasm.c (default_elf_asm_named_section): Don't check if
13723         HAVE_GAS_SECTION_EXCLUDE is defined.
13725 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
13727         * doc/md.texi: Update the documentation of the cond_* optabs
13728         to mention the new final operand.  Fix GET_MODE_NUNITS call.
13729         Describe the scalar case too.
13730         * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
13731         * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
13732         instead of 2.
13733         (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
13734         (get_conditional_internal_fn): Update comment.
13735         * tree-vect-loop.c (vectorizable_reduction): Pass the original
13736         accumulator value as a final argument to conditional functions.
13737         * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
13738         a define_expand and add an "else" operand.  Assert for now that
13739         the else operand is equal to operand 2.  Use SVE_INT_BINARY and
13740         SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
13741         (*cond_<optab><mode>): New patterns.
13742         * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
13743         (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
13744         (UNSPEC_COND_EOR): Delete.
13745         (optab): Remove associated mappings.
13746         (SVE_INT_BINARY): New code iterator.
13747         (sve_int_op): Remove int attribute and add "minus" to the code
13748         attribute.
13749         (SVE_COND_INT_OP): Delete.
13750         (SVE_COND_FP_OP): Rename to...
13751         (SVE_COND_FP_BINARY): ...this.
13753 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
13755         * optabs.c (can_reuse_operands_p): New function.
13756         (maybe_legitimize_operands): Try to reuse the results for
13757         earlier operands.
13759 2018-05-24  Uros Bizjak  <ubizjak@gmail.com>
13761         * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
13762         Add {q} suffix to insn mnemonic.
13764 2018-05-23  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
13766         * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
13767         (msp430_warn_func_return): New.
13769 2018-05-24  Roger Sayle  <roger@nextmovesoftware.com>
13771         * fold-const.c (tree_nonzero_bits): New function.
13772         * fold-const.h (tree_nonzero_bits): Likewise.
13773         * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
13774         friends.  POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
13776 2018-05-24  H.J. Lu  <hongjiu.lu@intel.com>
13778         PR target/85900
13779         PR target/85345
13780         * varasm.c (assemble_alias): Check ifunc_resolver only on
13781         FUNCTION_DECL.
13783 2018-05-24  Uros Bizjak  <ubizjak@gmail.com>
13785         PR target/85903
13786         * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
13787         when memory input operand is handled.
13789 2018-05-24  Luis Machado  <luis.machado@linaro.org>
13791         * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
13792         global.
13793         (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
13795 2018-05-24  Richard Sandiford  <richard.sandiford@linaro.org>
13797         * match.pd: Delay FMA folds until after vectorization.
13799 2018-05-24  Andre Vieira  <andre.simoesdiasvieira@arm.com>
13801         PR target/83009
13802         * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
13803         address check not strict.
13805 2018-05-24  Richard Sandiford  <richard.sandiford@linaro.org>
13807         * gimple-match.h (gimple_match_op): New class.
13808         (mprts_hook): Replace parameters with a gimple_match_op *.
13809         (maybe_build_generic_op): Likewise.
13810         (gimple_simplified_result_is_gimple_val): Replace parameters with
13811         a const gimple_match_op *.
13812         (gimple_simplify): Replace code_helper * and tree * parameters with
13813         a gimple_match_op * parameter.
13814         (gimple_resimplify1): Replace code_helper *, tree and tree *
13815         parameters with a gimple_match_op * parameter.
13816         (gimple_resimplify2): Likewise.
13817         (gimple_resimplify3): Likewise.
13818         (maybe_push_res_to_seq): Replace code_helper, tree and tree *
13819         parameters with a gimple_match_op * parameter.
13820         * gimple-match-head.c (gimple_simplify): Change prototypes of
13821         auto-generated functions to take a gimple_match_op * instead of
13822         separate code_helper * and tree * parameters.  Make the same
13823         change in the top-level overload and update calls to the
13824         gimple_resimplify routines.  Update calls to the auto-generated
13825         functions and to maybe_push_res_to_seq in the publicly-facing
13826         operation-specific gimple_simplify overloads.
13827         (gimple_match_op::MAX_NUM_OPS): Define.
13828         (gimple_resimplify1): Replace rcode and ops with a single res_op
13829         parameter.  Update call to gimple_simplify.
13830         (gimple_resimplify2): Likewise.
13831         (gimple_resimplify3): Likewise.
13832         (mprts_hook): Replace parameters with a gimple_match_op *.
13833         (maybe_build_generic_op): Likewise.
13834         (build_call_internal): Replace type, nargs and ops with
13835         a gimple_match_op *.
13836         (maybe_push_res_to_seq): Replace res_code, type and ops parameters
13837         with a single gimple_match_op *.  Update calls to mprts_hook,
13838         build_call_internal and gimple_simplified_result_is_gimple_val.
13839         Factor out code that is common to the tree_code and combined_fn cases.
13840         * genmatch.c (expr::gen_transform): Replace tem_code and
13841         tem_ops with a gimple_match_op called tem_op.  Update calls
13842         to the gimple_resimplify functions and maybe_push_res_to_seq.
13843         (dt_simplify::gen_1): Manipulate res_op instead of res_code and
13844         res_ops.  Update call to the gimple_resimplify functions.
13845         (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
13846         (decision_tree::gen): Make the functions take a gimple_match_op *
13847         called res_op instead of separate res_code and res_ops parameters.
13848         Update call accordingly.
13849         * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
13850         and ops with a single res_op parameter.  Update calls to
13851         maybe_build_generic_op and maybe_push_res_to_seq.
13852         (fold_stmt_1): Update calls to gimple_simplify and
13853         replace_stmt_with_simplification.
13854         (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
13855         and gimple_simplified_result_is_gimple_val.
13856         * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
13857         gimple_simplify.
13858         * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
13859         with a gimple_match_op *.
13860         (vn_nary_build_or_lookup): Likewise.  Update call to
13861         vn_nary_build_or_lookup_1.
13862         (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
13863         gimple_match_op *.  Update calls to the gimple_resimplify routines
13864         and to gimple_simplified_result_is_gimple_val.
13865         (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
13866         Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
13867         (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
13868         (visit_nary_op): Likewise.
13869         (visit_reference_op_load): Likewise.
13871 2018-05-23  Luis Machado  <luis.machado@linaro.org>
13873         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
13874         modifier for printing the step amount.
13876 2018-05-23  Jozef Lawrynowicz  <jozef.l@somniumtech.com>
13878         PR target/78849
13879         * tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
13880         types.
13882 2018-05-23  Segher Boessenkool  <segher@kernel.crashing.org>
13884         * doc/sourcebuild.texi (Endianness): New subsubsection.
13886 2018-05-23  Luis Machado  <luis.machado@linaro.org>
13888         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
13889         <prefetch_dynamic_strides>: New const bool field.
13890         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
13891         prefetch_dynamic_strides.
13892         (exynosm1_prefetch_tune): Likewise.
13893         (thunderxt88_prefetch_tune): Likewise.
13894         (thunderx_prefetch_tune): Likewise.
13895         (thunderx2t99_prefetch_tune): Likewise.
13896         (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
13897         false.
13898         (aarch64_override_options_internal): Update to set
13899         PARAM_PREFETCH_DYNAMIC_STRIDES.
13900         * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
13901         * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
13902         * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
13903         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
13904         prefetch-dynamic-strides setting.
13906 2018-05-23  Luis Machado  <luis.machado@linaro.org>
13908         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
13909         <minimum_stride>: New const int field.
13910         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
13911         minimum_stride field defaulting to -1.
13912         (exynosm1_prefetch_tune): Likewise.
13913         (thunderxt88_prefetch_tune): Likewise.
13914         (thunderx_prefetch_tune): Likewise.
13915         (thunderx2t99_prefetch_tune): Likewise.
13916         (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
13917         <default_opt_level>: Set to 3.
13918         (aarch64_override_options_internal): Update to set
13919         PARAM_PREFETCH_MINIMUM_STRIDE.
13920         * doc/invoke.texi (prefetch-minimum-stride): Document new option.
13921         * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
13922         * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
13923         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
13924         stride is constant and is below the minimum stride threshold.
13926 2018-05-23  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
13928         * config/arm/arm-cpus.in (mode26): Delete.
13929         (armv4): Delete mode26 reference.
13930         * config/arm/arm.c (arm_configure_build_target): Delete use of
13931         isa_bit_mode26.
13933 2018-05-23  Uros Bizjak  <ubizjak@gmail.com>
13935         * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
13936         New insn pattern.
13937         (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
13938         Rewrite expander pattern.  Emit gen_floatunssi<mode>2_i387_with_xmm
13939         for non-SSE modes.
13940         (floatunsdisf2): Rewrite expander pattern.  Hanlde TARGET_AVX512F.
13941         (floatunsdidf2): Ditto.
13943 2018-05-23  Uros Bizjak  <ubizjak@gmail.com>
13945         * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
13946         (fixuns_trunc<mode>si2_avx512f): Ditto.
13947         (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
13948         (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
13949         Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
13951 2018-05-23  Alexander Monakov  <amonakov@ispras.ru>
13953         PR rtl-optimization/79985
13954         * df-scan.c (df_insn_refs_collect): Remove special case for
13955         global registers and asm statements.
13957 2018-05-23  Alexander Monakov  <amonakov@ispras.ru>
13959         * extend.texi (Global Register Variables): Rewrite the bullet list.
13960         Note that the register is available for allocation. Note that access
13961         via inline asm must use constraints. Add note about async-signal
13962         handlers. Remove paragraph about automagic register selection.
13964 2018-05-23  Richard Biener  <rguenther@suse.de>
13966         * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
13967         of fixed offset from memset VN.
13969 2018-05-23  Bill Schmidt  <wschmidt@linux.ibm.com>
13971         * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
13972         first_interp field.
13973         (alloc_cand_and_find_basis): Initialize first_interp field.
13974         (slsr_process_mul): Modify first_interp field.
13975         (slsr_process_add): Likewise.
13976         (slsr_process_cast): Modify first_interp field for each new
13977         interpretation.
13978         (slsr_process_copy): Likewise.
13979         (dump_candidate): Dump first_interp field.
13980         (replace_mult_candidate): Process all interpretations, not just
13981         subsequent ones.
13982         (replace_rhs_if_not_dup): Likewise.
13983         (replace_one_candidate): Likewise.
13985 2018-05-23  Wilco Dijkstra  <wdijkstr@arm.com>
13987         * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
13988         Add new boolean.
13989         (aarch64_needs_frame_chain): New function.
13990         (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
13992 2018-05-23  Sudakshina Das  <sudi.das@arm.com>
13994         PR target/84882
13995         * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
13996         Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
13997         * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
13998         * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
13999         as true for strict-align.
14000         (aarch64_can_inline_p): Perform checks even when callee has no
14001         attributes to check for strict alignment.
14002         * doc/extend.texi (AArch64 Function Attributes): Document
14003         no-strict-align.
14004         * doc/invoke.texi: (AArch64 Options): Likewise.
14006 2018-05-23  Richard Sandiford  <richard.sandiford@linaro.org>
14008         PR tree-optimization/85853
14009         * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
14010         the handling of the root of the node to...
14011         (vect_slp_analyze_node_operations_1): ...this new function,
14012         and run the whole thing with the child nodes' def types
14013         set according to their SLP node's def type.
14015 2018-05-23  Richard Biener  <rguenther@suse.de>
14017         PR middle-end/85874
14018         * tree-data-ref.c (create_runtime_alias_checks): Defer
14019         and ignore overflow warnings.
14021 2018-05-23  Yury Gribov  <tetra2005@gmail.com>
14023         PR tree-optimization/85822
14024         * tree-vrp.c (is_masked_range_test): Fix handling of negative
14025         constants.
14027 2018-05-23  Richard Biener  <rguenther@suse.de>
14029         * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
14030         memset constants via native_interpret_expr.
14032 2018-05-22  H.J. Lu  <hongjiu.lu@intel.com>
14034         PR target/85345
14035         * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
14036         attribute.
14037         (cgraph_node::create_alias): Likewise.
14038         (cgraph_node::get_availability): Check ifunc_resolver instead
14039         of looking up ifunc attribute.
14040         * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
14041         * varasm.c (do_assemble_alias): Likewise.
14042         (assemble_alias): Likewise.
14043         (default_binds_local_p_3): Likewise.
14044         * cgraph.h (cgraph_node): Add ifunc_resolver.
14045         (cgraph_node::only_called_directly_or_aliased_p): Return false
14046         for IFUNC resolver.
14047         * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
14048         attribute.
14049         * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
14050         is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
14051         (symtab_node::binds_to_current_def_p): Check ifunc_resolver
14052         instead of looking up ifunc attribute.
14054 2018-05-22  Luis Machado  <luis.machado@linaro.org>
14056         * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
14058 2018-05-22  Martin Sebor  <msebor@redhat.com>
14060         PR middle-end/85359
14061         * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
14062         only when expasion succeeds.
14063         (expand_builtin_strcmp): Same.
14064         (expand_builtin_strncmp): Same.
14066 2018-05-22  Martin Sebor  <msebor@redhat.com>
14068         * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
14070 2018-05-22  Jackson Woodruff  <jackson.woodruff@arm.com>
14071             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
14073         * config/aarch64/aarch64-ldpstp.md: Replace uses of
14074         aarch64_mem_pair_operand with memory_operand and delete operand swapping
14075         code.
14076         * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
14077         Add check for legitimate_address.
14078         (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
14079         (aarch64_swap_ldrstr_operands): New.
14080         * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
14081         Define prototype.
14083 2018-05-22  Jackson Woodruff  <jackson.woodruff@arm.com>
14084             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
14086         * config/aarch64/aarch64.md: New patterns to generate stp
14087         and ldp.
14088         (store_pair_sw, store_pair_dw): New patterns to generate stp for
14089         single words and double words.
14090         (load_pair_sw, load_pair_dw): Likewise.
14091         (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
14092         Delete.
14093         (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
14094         Delete.
14095         * config/aarch64/aarch64-ldpstp.md: Modify peephole
14096         for different mode ldpstp and add peephole for merged zero stores.
14097         Likewise for loads.
14098         * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
14099         Add size check.
14100         (aarch64_gen_store_pair): Rename calls to match new patterns.
14101         (aarch64_gen_load_pair): Rename calls to match new patterns.
14102         * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
14103         (load_pair<DREG:mode><DREG2:mode>): ... This.
14104         (store_pair<mode>): Rename to...
14105         (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
14106         * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
14107         New mode iterators.
14108         (V_INT_EQUIV): Handle SImode.
14109         * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
14110         New predicate.
14112 2018-05-22  Martin Sebor  <msebor@redhat.com>
14114         PR c/85623
14115         * calls.c (maybe_warn_nonstring_arg): Use string length to set
14116         or ajust the presumed bound on an operation to avoid unnecessary
14117         warnings.
14119 2018-05-22  Martin Sebor  <msebor@redhat.com>
14121         PR tree-optimization/85826
14122         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
14123         assuming that a DECL necesarily has a constant size.
14125 2018-05-22  Richard Sandiford  <richard.sandiford@linaro.org>
14127         PR middle-end/85862
14128         * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
14130 2018-05-22  Richard Biener  <rguenther@suse.de>
14132         PR tree-optimization/85834
14133         * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
14134         non-constant and non-zero memset arguments.
14136 2018-05-22  Martin Liska  <mliska@suse.cz>
14138         PR ipa/85607
14139         * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
14141 2018-05-22  Richard Biener  <rguenther@suse.de>
14143         PR tree-optimization/85863
14144         * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
14145         comparisons when vectype is specified.
14146         (vectorizable_condition): Do not specify vectype for
14147         vect_is_simple_cond when SLP vectorizing.
14149 2018-05-21  Michael Meissner  <meissner@linux.ibm.com>
14151         PR target/85657
14152         * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
14153         define __ibm128 as long double.
14154         * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
14155         as a distinct type when IEEE 128-bit support is enabled.
14156         (init_float128_ieee): Fix up conversions between IFmode and IEEE
14157         128-bit types to use the correct functions.
14158         (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
14159         convert between 128-bit floating point types that have different
14160         modes but the same representation, instead of using gen_lowpart to
14161         makean alias.
14162         * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
14163         KFmode.
14164         (IFKF_reg): New attributes to give the register constraints for
14165         IFmode and KFmode.
14166         (extend<mode>tf2_internal): New insns to mark an explicit
14167         conversion between 128-bit floating point types that have a
14168         different mode but share the same representation.
14170 2018-05-21  Richard Sandiford  <richard.sandiford@linaro.org>
14172         PR tree-optimization/85814
14173         * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
14174         a null return from get_strinfo when unsharing the next
14175         strinfo in the chain.
14177 2018-05-21  Vladimir Mezentsev  <vladimir.mezentsev@oracle.com>
14179         PR gcc/84923
14180         * varasm.c (weak_finish): Clean up weak_decls.
14182 2018-05-21  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
14184         * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
14185         UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
14186         UNSPEC_UADALP values.
14187         * config/aarch64/iterators.md (ABAL): New int iterator.
14188         (ABDL2): Likewise.
14189         (ADALP): Likewise.
14190         (sur): Add mappings for the above.
14191         * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
14192         New define_insn.
14193         (aarch64_<sur>abal<mode>_4): Likewise.
14194         (aarch64_<sur>adalp<mode>_3): Likewise.
14195         (<sur>sadv16qi): New define_expand.
14197 2018-05-21  Alexander Nesterovskiy  <alexander.nesterovskiy@intel.com>
14199         * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
14200         (*movdf_internal): Ditto.
14201         (*rcpsf2_sse): Ditto.
14202         (*rsqrtsf2_sse): Ditto.
14203         (*sqrt<mode>2_sse): Ditto.
14205 2018-05-21  Tamar Christina  <tamar.christina@arm.com>
14207         * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
14208         eor3q<mode>4.
14209         (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
14210         * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
14211         veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
14212         vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
14213         vbcaxq_s64): New.
14214         * config/aarch64/arm_neon.h: Likewise.
14215         * config/aarch64/iterators.md (VQ_I): New.
14217 2018-05-21  Alexey Brodkin <abrodkin@synopsys.com>
14219         * config.gcc: Add arc/t-multilib-linux to tmake_file for
14220         arc*-*-linux*.
14221         * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
14222         MULTILIB_DIRNAMES
14224 2018-05-20  Chung-Ju Wu  <jasonwucj@gmail.com>
14226         * config/nds32/constraints.md (S): New constraint.
14227         * config/nds32/nds32.md (call_internal): Use constraint S.
14228         (call_value_internal): Likewise.
14229         (sibcall_internal): Likewise.
14230         (sibcall_value_internal): Likewise.
14232 2018-05-20  Kito Cheng  <kito.cheng@gmail.com>
14233             Chung-Ju Wu  <jasonwucj@gmail.com>
14235         * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
14236         into consideration.
14238 2018-05-20  Kito Cheng  <kito.cheng@gmail.com>
14239             Chung-Ju Wu  <jasonwucj@gmail.com>
14241         * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
14242         (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
14243         (nds32_rtx_costs_impl): Simplify.
14244         (nds32_address_cost_impl): Simplify.
14245         (nds32_init_rtx_costs): New function.
14246         (nds32_rtx_costs_speed_prefer): Likewise.
14247         (nds32_rtx_costs_size_prefer): Likewise.
14248         (nds32_address_cost_speed_prefer): Likewise.
14249         (nds32_address_cost_speed_fwprop): Likewise.
14250         (nds32_address_cost_size_prefer): Likewise.
14251         * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
14252         * config/nds32/nds32.c (nds32_option_override): Use
14253         nds32_init_rtx_costs function.
14255 2018-05-20  Chung-Ju Wu  <jasonwucj@gmail.com>
14257         * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
14258         * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
14259         (TARGET_PIPELINE_N8): Likewise.
14260         (TARGET_PIPELINE_N10): Likewise.
14261         (TARGET_PIPELINE_N13): Likewise.
14262         (TARGET_PIPELINE_GRAYWOLF): Likewise.
14264 2018-05-19  Monk Chiang  <sh.chiang04@gmail.com>
14266         * config/nds32/nds32-fpu.md: Update copyright year.
14268 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
14270         * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
14272 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
14274         * config/nds32/nds32.c
14275         (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
14276         * config/nds32/nds32.opt (minline-asm-r15): New option.
14278 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
14280         * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
14281         MASK_HW_ABS.
14282         * config/nds32/nds32.md (abssi2): New pattern.
14284 2018-05-19  Uros Bizjak  <ubizjak@gmail.com>
14286         * config/i386/i386.md (rex64namesuffix): New mode attribute.
14287         * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
14288         Merge insn pattern from sse_cvtsi2ss<round_name> and
14289         sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
14290         (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
14291         from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
14292         using SWI48 mode iterator.
14293         (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
14294         sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
14295         (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
14296         pattern from sse_cvttss2si<round_saeonly_name>
14297         and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
14298         (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
14299         from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
14300         using SWI48 mode iterator.
14301         (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
14302         insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
14303         avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
14304         (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
14305         from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
14306         using SWI48 mode iterator.
14307         (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
14308         insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
14309         avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
14310         (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
14311         sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
14312         SWI48 mode iterator.
14313         (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
14314         sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
14315         (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
14316         pattern from sse_cvttsd2si<round_saeonly_name>
14317         and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
14319 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
14321         * config/nds32/nds32-md-auxiliary.c
14322         (nds32_valid_smw_lwm_base_p): Refine.
14323         (nds32_output_smw_single_word): Refine.
14324         (nds32_output_smw_double_word): New.
14325         * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
14327 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
14329         * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
14330         (nds32_output_stack_pop): Refine.
14331         (nds32_expand_unaligned_load): Refine.
14332         (nds32_expand_unaligned_store): Refine.
14334 2018-05-19  Kuan-Lin Chen <kuanlinchentw@gmail.com>
14335             Chung-Ju Wu  <jasonwucj@gmail.com>
14337         * config/nds32/constants.md: Add TP_REGNUM constant.
14338         (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
14339         UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
14340         UNSPEC_ADD32.
14341         * config/nds32/nds32-doubleword.md: Consider flag_pic.
14342         * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
14343         * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
14344         * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
14345         and PIC code generation.
14346         * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
14347         code generation.
14348         * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
14349         optimization.
14350         * config/nds32/nds32.md: Support TLS and PIC.
14351         * config/nds32/nds32.c: Support TLS and PIC.
14352         * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
14353         * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
14354         predicate.
14356 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
14358         * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
14359         mode with E_ prefix.
14361 2018-05-19  Kuan-Lin Chen <kuanlinchentw@gmail.com>
14362             Chung-Ju Wu  <jasonwucj@gmail.com>
14364         * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
14365         * config/nds32/nds32-md-auxiliary.c
14366         (symbolic_reference_mentioned_p): New.
14367         (nds32_legitimize_ict_address): New.
14368         (nds32_expand_ict_move): New.
14369         (nds32_indirect_call_referenced_p): New.
14370         (nds32_symbol_binds_local_p): Delete.
14371         (nds32_long_call_p): Modify.
14372         * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
14373         * config/nds32/nds32-protos.h
14374         (symbolic_reference_mentioned_p): Declare.
14375         (nds32_legitimize_ict_address): Declare.
14376         (nds32_expand_ict_move): Declare.
14377         (nds32_indirect_call_referenced_p): Declare.
14378         * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
14379         (nds32_relax_group): Use nds32_ict_const_p as condition.
14380         * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
14381         (nds32_asm_file_start): Output ict_model directive in asm code.
14382         (nds32_legitimate_address_p): Consider indirect call.
14383         (nds32_print_operand): Consider indirect call.
14384         (nds32_print_operand_address): Consider indirect call.
14385         (nds32_insert_attributes): Handle "indirect_call" attribute.
14386         (TARGET_LEGITIMATE_ADDRESS_P): Define.
14387         (TARGET_LEGITIMATE_CONSTANT_P): Define.
14388         (TARGET_CANNOT_FORCE_CONST_MEM): Define.
14389         (TARGET_DELEGITIMIZE_ADDRESS): Define.
14390         (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
14391         * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
14392         (TARGET_ICT_MODEL_SMALL): Define.
14393         (TARGET_ICT_MODEL_LARGE): Define.
14394         * config/nds32/nds32.md (movsi): Consider ict model.
14395         (call, call_value): Consider ict model.
14396         (sibcall, sibcall_value): Consider ict model.
14397         * config/nds32/nds32.opt (mict-model): New option.
14398         * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
14399         model.
14401 2018-05-18  Kito Cheng  <kito.cheng@gmail.com>
14402             Monk Chiang  <sh.chiang04@gmail.com>
14403             Jim Wilson <jimw@sifive.com>
14405         * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
14406         Add support to parse rv32e*.  Clear MASK_RVE for rv32i and rv64i.
14407         * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
14408         * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
14409         __riscv_32e when TARGET_RVE.  Handle ABI_ILP32E as soft-float ABI.
14410         * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
14411         * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
14412         compute save_libcall_adjustment properly.
14413         (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
14414         (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
14415         * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
14416         (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
14417         (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
14418         (ABI_SPEC): Handle mabi=ilp32e.
14419         * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
14420         (RVE): Add RVE mask.
14421         * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
14422         <-march>: Add rv32e as an example.
14424 2018-05-18  Marc Glisse  <marc.glisse@inria.fr>
14426         PR c++/82899
14427         * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
14428         (intra_create_variable_infos): Handle C++ constructors.
14430 2018-05-18  Martin Liska  <mliska@suse.cz>
14432         * passes.def: Remove a redundant pass.
14434 2018-05-18  Eric Botcazou  <ebotcazou@adacore.com>
14436         PR bootstrap/85838
14437         * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
14439 2018-05-18  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
14441         * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
14442         (ARMv4): Update.
14443         (ARMv2, ARMv3, ARMv3m): Delete fgroups.
14444         (ARMv6m): Update.
14445         (armv2, armv2a, armv3, armv3m): Delete architectures.
14446         (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
14447         arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
14448         arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
14449         Delete cpus.
14450         * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
14451         (*mulsidi3adddi): Likewise.
14452         (mulsidi3): Likewise.
14453         (*mulsidi3_nov6): Likewise.
14454         (umulsidi3): Likewise.
14455         (umulsidi3_nov6): Likewise.
14456         (umaddsidi4): Likewise.
14457         (*umulsidi3adddi): Likewise.
14458         (smulsi3_highpart): Likewise.
14459         (*smulsi3_highpart_nov6): Likewise.
14460         (umulsi3_highpart): Likewise.
14461         (*umulsi3_highpart_nov6): Likewise.
14462         * config/arm/arm.h (arm_arch3m): Delete.
14463         * config/arm/arm.c (arm_arch3m): Delete.
14464         (arm_option_override_internal): Update armv3-related comment.
14465         (arm_configure_build_target): Delete use of isa_bit_mode32.
14466         (arm_option_reconfigure_globals): Delete set of arm_ach3m.
14467         (arm_rtx_costs_internal): Delete check of arm_arch3m.
14468         * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
14469         (mulsa3): Likewise.
14470         (mulusa3): Likewise.
14471         * config/arm/arm-protos.h (arm_arch3m): Delete.
14472         * config/arm/arm-tables.opt: Regenerate.
14473         * config/arm/arm-tune.md: Likewise.
14474         * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
14475         deleted architectures.
14477 2018-05-18  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
14479         * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
14480         (armv5t, armv5te): New features.
14481         (ARMv5, ARMv5e): Delete fgroups.
14482         (ARMv5t, ARMv5te): Adjust for above changes.
14483         (ARMv6m): Likewise.
14484         (armv5, armv5e): Delete arches.
14485         * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
14486         arm_arch5.
14487         (*call_reg_arm): Likewise.
14488         (*call_value_reg_armv5): Likewise.
14489         (*call_value_reg_arm): Likewise.
14490         (*call_symbol): Likewise.
14491         (*call_value_symbol): Likewise.
14492         (*sibcall_insn): Likewise.
14493         (*sibcall_value_insn): Likewise.
14494         (clzsi2): Likewise.
14495         (prefetch): Likewise.
14496         (define_split and define_peephole2 dependent on arm_arch5):
14497         Likewise.
14498         * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
14499         arm_arch5e.
14500         (TARGET_ARM_QBIT): Likewise.
14501         (TARGET_DSP_MULTIPLY): Likewise.
14502         (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
14503         (arm_arch5, arm_arch5e): Delete.
14504         (arm_arch5t, arm_arch5te): Declare.
14505         * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
14506         (arm_arch5t): Declare.
14507         (arm_option_reconfigure_globals): Update for the above.
14508         (arm_options_perform_arch_sanity_checks): Update comment, replace
14509         use of arm_arch5 with arm_arch5t.
14510         (use_return_insn): Likewise.
14511         (arm_emit_call_insn): Likewise.
14512         (output_return_instruction): Likewise.
14513         (arm_final_prescan_insn): Likewise.
14514         (arm_coproc_builtin_available): Likewise.
14515         * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
14516         arm_arch5e with arm_arch5t and arm_arch5te.
14517         * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
14518         (arm_arch5t, arm_arch5te): Declare.
14519         * config/arm/arm-tables.opt: Regenerate.
14520         * config/arm/t-arm-elf: Remove references to armv5, armv5e.
14521         * config/arm/t-multilib: Likewise.
14522         * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
14523         instead of arm_arch5.
14524         (*call_reg_thumb1): Likewise.
14525         (*call_value_reg_thumb1_v5): Likewise.
14526         (*call_value_reg_thumb1): Likewise.
14527         * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
14528         unreachable path.
14529         * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
14531 2018-05-18  Martin Liska  <mliska@suse.cz>
14533         PR gcov-profile/84846
14534         * doc/gcov.texi: Document -t option of gcov tool.
14536 2018-05-18  Martin Liska  <mliska@suse.cz>
14538         PR gcov-profile/84846
14539         * gcov.c (print_usage): Add new -t option.
14540         (process_args): Handle the option.
14541         (generate_results): Use stdout as output when requested by
14542         the option.
14544 2018-05-18  Martin Liska  <mliska@suse.cz>
14546         PR gcov-profile/84846
14547         * coverage.c (coverage_init): Write PWD to .gcno file.
14548         * doc/gcov.texi: Document how working directory is printed.
14549         * gcov-dump.c (dump_gcov_file): Print PWD.
14550         * gcov.c (output_intermediate_file): Likewise.
14551         (read_graph_file): Read PWD string.
14552         (output_lines): Print PWD.
14554 2018-05-18  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
14556         PR middle-end/85817
14557         * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
14558         for retval and return false if all args to phi are zero.
14560 2018-05-18  Richard Biener  <rguenther@suse.de>
14562         * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
14563         method.
14564         (evrp_dom_walker::before_dom_children): Call it.
14566 2018-05-18  Richard Biener  <rguenther@suse.de>
14568         * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
14569         results when processing array refs with variable index.
14571 2018-05-18  Toon Moene  <toon@moene.org>
14573         * doc/invoke.texi: Move -floop-unroll-and-jam documentation
14574         directly after that of -floop-interchange. Indicate that both
14575         options are enabled by default when specifying -O3.
14577 2018-05-18  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
14579         * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
14580         iterator.  Delete separate integer-mode vec_set<mode> expander.
14581         (aarch64_simd_vec_setv2di): Delete.
14582         (vec_setv2di): Delete.
14583         (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
14584         Use VALL_F16 mode iterator.  Add LD1 alternative and use vwcore for
14585         the "w, r" alternative.
14587 2018-05-18  Martin Liska  <mliska@suse.cz>
14589         * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
14590         * tree-pass.h (make_pass_lower_switch_O0): New function.
14591         * tree-switch-conversion.c (node_has_low_bound): Remove.
14592         (node_has_high_bound): Likewise.
14593         (node_is_bounded): Likewise.
14594         (class pass_lower_switch): Make it a template type and create
14595         two instances.
14596         (pass_lower_switch::execute): Add template argument.
14597         (make_pass_lower_switch): New function.
14598         (make_pass_lower_switch_O0): New function.
14599         (do_jump_if_equal): Remove.
14600         (emit_case_nodes): Simplify to just handle all 3 cases and leave
14601         all the hard work to tree optimization passes.
14603 2018-05-18  Martin Liska  <mliska@suse.cz>
14605         * dbgcnt.c (limit_low): Renamed from limit.
14606         (limit_high): New variable.
14607         (dbg_cnt_is_enabled): Check for upper limit.
14608         (dbg_cnt): Adjust dumping.
14609         (dbg_cnt_set_limit_by_index): Add new argument for high
14610         value.
14611         (dbg_cnt_set_limit_by_name): Likewise.
14612         (dbg_cnt_process_single_pair): Parse new format.
14613         (dbg_cnt_process_opt): Use strtok.
14614         (dbg_cnt_list_all_counters): Remove 'value' and add
14615         'limit_high'.
14616         * doc/invoke.texi: Document changes.
14618 2018-05-18  Richard Sandiford  <richard.sandiford@linaro.org>
14620         * doc/sourcebuild.texi (scalar_all_fma): Document.
14621         * tree.def (FMA_EXPR): Delete.
14622         * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
14623         * internal-fn.c (ternary_direct): New macro.
14624         (expand_ternary_optab_fn): Likewise.
14625         (direct_ternary_optab_supported_p): Likewise.
14626         * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
14627         * builtins.c (fold_builtin_fma): Delete.
14628         (fold_builtin_3): Don't call it.
14629         * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
14630         * expr.c (expand_expr_real_2): Likewise.
14631         * fold-const.c (operand_equal_p): Likewise.
14632         (fold_ternary_loc): Likewise.
14633         * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
14634         * gimple.c (DEFTREECODE): Likewise.
14635         * gimplify.c (gimplify_expr): Likewise.
14636         * optabs-tree.c (optab_for_tree_code): Likewise.
14637         * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
14638         * tree-eh.c (operation_could_trap_p): Likewise.
14639         (stmt_could_throw_1_p): Likewise.
14640         * tree-inline.c (estimate_operator_cost): Likewise.
14641         * tree-pretty-print.c (dump_generic_node): Likewise.
14642         (op_code_prio): Likewise.
14643         * tree-ssa-loop-im.c (stmt_cost): Likewise.
14644         * tree-ssa-operands.c (get_expr_operands): Likewise.
14645         * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
14646         * fold-const-call.h (fold_fma): Delete.
14647         * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
14648         CFN_FNMA and CFN_FNMS.
14649         (fold_fma): Delete.
14650         * genmatch.c (combined_fn): New enum.
14651         (commutative_ternary_tree_code): Remove FMA_EXPR handling.
14652         (commutative_op): New function.
14653         (commutate): Use it.  Handle more than 2 operands.
14654         (dt_operand::gen_gimple_expr): Use commutative_op.
14655         (parser::parse_expr): Allow :c to be used with non-binary
14656         operators if the commutative operand is known.
14657         * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
14658         CFN_FMS, CFN_FNMA and CFN_FNMS.
14659         (backprop::process_assign_use): Remove FMA_EXPR handling.
14660         * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
14661         (gen_hsa_fma): New function.
14662         (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
14663         IFN_FNMA and IFN_FNMS.
14664         * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
14665         * gimple-fold.h (follow_all_ssa_edges): Declare.
14666         * gimple-fold.c (follow_all_ssa_edges): New function.
14667         * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
14668         gimple_build interface and use follow_all_ssa_edges to fold the result.
14669         (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
14670         instead of checking for optabs directly.
14671         * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
14672         rather than FMA_EXPRs.
14673         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
14674         call to IFN_FMA instead of an FMA_EXPR.
14676 2018-05-17  Jim Wilson  <jimw@sifive.com>
14678         * expr.c (do_tablejump): When converting index to Pmode, if we have a
14679         sign extended promoted subreg, and the range does not have the sign bit
14680         set, then do a sign extend.
14682         * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
14683         test, check for sign extended subreg and/or constant operands, and
14684         do a sign extend in that case.
14686 2018-05-17  Steve Ellcey  <sellcey@cavium.com>
14688         * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
14689         (thunderx2t99_multiple): Delete psuedo-units from used cpus.
14690         Add untyped.
14691         (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
14692         Change logics_shift_reg to logics_shift_imm.
14693         (thunderx2t99_fp_loadpair_basic): Delete.
14694         (thunderx2t99_fp_storepair_basic): Delete.
14695         (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
14696         (thunderx2t99_asimd_polynomial): Delete.
14697         (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
14698         and neon_fp_mul_d_scalar_q.
14699         (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
14700         (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
14701         (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
14702         (thunderx2t99_asimd_lut): Add missing tbl types.
14703         (thunderx2t99_asimd_ext): Delete.
14704         (thunderx2t99_asimd_load1_1_mult): Delete.
14705         (thunderx2t99_asimd_load1_2_mult): Delete.
14706         (thunderx2t99_asimd_load1_ldp): New.
14707         (thunderx2t99_asimd_load1): New.
14708         (thunderx2t99_asimd_load2): Add missing *load2* types.
14709         (thunderx2t99_asimd_load3): New.
14710         (thunderx2t99_asimd_load4): New.
14711         (thunderx2t99_asimd_store1_1_mult): Delete.
14712         (thunderx2t99_asimd_store1_2_mult): Delete.
14713         (thunderx2t99_asimd_store2_mult): Delete.
14714         (thunderx2t99_asimd_store2_onelane): Delete.
14715         (thunderx2t99_asimd_store_stp): New.
14716         (thunderx2t99_asimd_store1): New.
14717         (thunderx2t99_asimd_store2): New.
14718         (thunderx2t99_asimd_store3): New.
14719         (thunderx2t99_asimd_store4): New.
14721 2018-05-17  Jerome Lambourg  <lambourg@adacore.com>
14723         * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
14724         #include <stdint.h>.  Replace intptr_t with __INTPTR_TYPE__.
14726 2018-05-17  Pat Haugen  <pthaugen@us.ibm.com>
14727             Segher Boessenkool  <segher@kernel.crashing.org>
14729         PR target/85698
14730         * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
14731         operand.
14733 2018-05-17  Richard Biener  <rguenther@suse.de>
14735         * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
14736         for pruning loop and prune defs feeding only already visited PHIs.
14738 2018-05-17  Richard Biener  <rguenther@suse.de>
14740         * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
14742 2018-05-17  Bin Cheng  <bin.cheng@arm.com>
14743             Richard Biener  <rguenther@suse.de>
14745         PR tree-optimization/85793
14746         * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
14747         for VMAT_ELEMENTWISE.
14749 2018-05-17  Richard Sandiford  <richard.sandiford@linaro.org>
14751         * internal-fn.h (lookup_internal_fn): Declare
14752         * internal-fn.c (lookup_internal_fn): New function.
14753         * gimple.c (gimple_build_call_from_tree): Handle calls to
14754         internal functions.
14755         * gimple-pretty-print.c (dump_gimple_call): Print "." before
14756         internal function names.
14757         * tree-pretty-print.c (dump_generic_node): Likewise.
14758         * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
14760 2018-05-17  Richard Sandiford  <richard.sandiford@linaro.org>
14762         * gimple-fold.h (gimple_build): Make the function forms take
14763         combined_fn rather than built_in_function.
14764         (gimple_simplify): Likewise.
14765         * gimple-match-head.c (gimple_simplify): Likewise.
14766         * gimple-fold.c (gimple_build): Likewise.
14767         * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
14768         rather than gimple_build_call_internal.
14769         (get_initial_defs_for_reduction): Likewise.
14770         (vect_create_epilog_for_reduction): Likewise.
14771         (vectorizable_live_operation): Likewise.
14773 2018-05-17  Martin Liska  <mliska@suse.cz>
14775         * gimple-ssa-sprintf.c (format_directive): Do not use
14776         space in between 'G_' and '('.
14778 2018-05-17  Jakub Jelinek  <jakub@redhat.com>
14780         PR target/85323
14781         * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
14782         even if the mask is not all ones.
14784         PR target/85323
14785         * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
14786         vector.
14787         (ix86_gimple_fold_builtin): Likewise.
14789         PR target/85323
14790         * config/i386/i386.c: Include tree-vector-builder.h.
14791         (ix86_vector_shift_count): New function.
14792         (ix86_fold_builtin): Fold shift builtins by scalar count.
14793         (ix86_gimple_fold_builtin): Likewise.
14795         * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
14796         _mm512_setzero): New intrinsics.
14798 2018-05-17  James Greenhalgh  <james.greenhalgh@arm.com>
14799             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
14801         * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
14802         code generation for cases where splatting a value is not useful.
14803         * simplify-rtx.c (simplify_ternary_operation): Simplify
14804         vec_merge across a vec_duplicate and a paradoxical subreg forming
14805         a vector mode to a vec_concat.
14807 2018-05-17  Olga Makhotina  <olga.makhotina@intel.com>
14809         * config.gcc: Support "goldmont-plus".
14810         * config/i386/driver-i386.c (host_detect_local_cpu): Detect
14811         "goldmont-plus".
14812         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
14813         PROCESSOR_GOLDMONT_PLUS.
14814         * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
14815         (processor_target_table): Add "goldmont-plus".
14816         (PTA_GOLDMONT_PLUS): Define.
14817         (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
14818         (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
14819         (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
14820         (fold_builtin_cpu): Add "goldmont-plus".
14821         (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
14822         (ix86_option_override_internal): Add "goldmont-plus".
14823         * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
14824         (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
14825         * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
14826         * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
14828 2018-05-17  Richard Biener  <rguenther@suse.de>
14830         PR tree-optimization/85757
14831         * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
14832         remove defs that only feed that PHI from further processing.
14834 2018-05-16  Jim Wilson  <jimw@sifive.com>
14836         * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
14837         asterisk to name.
14838         (<optab>di3_mask, <optab>di3_mask_1): Likewise.
14840 2018-05-16  Mark Wielaard  <mark@klomp.org>
14842         * dwarf2out.c (count_index_strings): New function.
14843         (output_indirect_strings): Call count_index_strings and generate
14844         header for dwarf_version >= 5.
14846 2018-05-16  Mark Wielaard  <mark@klomp.org>
14848         * dwarf2out.c (dwarf_FORM): New function.
14849         (set_indirect_string): Use dwarf_FORM.
14850         (reset_indirect_string): Likewise.
14851         (size_of_die): Likewise.
14852         (value_format): Likewise.
14853         (output_die): Likewise.
14854         (add_skeleton_AT_string): Likewise.
14855         (output_macinfo_op): Likewise.
14856         (index_string): Likewise.
14857         (output_index_string_offset): Likewise.
14858         (output_index_string): Likewise.
14859         (count_index_strings): Likewise.
14861 2018-05-16  Carl Love  <cel@us.ibm.com>
14863         * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
14864         dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
14866 2018-05-16  Martin Jambor  <mjambor@suse.cz>
14868         * ipa-prop.c (ipa_free_all_edge_args): Remove.
14869         * ipa-prop.h (ipa_free_all_edge_args): Likewise.
14871 2018-05-16  Wilco Dijkstra  <wdijkstr@arm.com>
14873         * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
14874         (fnma<mode>4): Likewise.
14875         (fms<mode>4): Likewise.
14876         (fnms<mode>4): Likewise.
14877         (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
14878         (aarch64_fnma<mode>4): Likewise.
14879         (aarch64_fms<mode>4): Likewise.
14880         (aarch64_fnms<mode>4): Likewise.
14881         (aarch64_fnmadd<mode>4): Likewise.
14883 2018-05-16  Jason Merrill  <jason@redhat.com>
14885         * tree.c (warn_deprecated_use): Return bool.  Simplify logic.
14887 2018-05-16  Richard Biener  <rguenther@suse.de>
14889         * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
14890         (dump_stmt_cost): Declare.
14891         (add_stmt_cost): Dump cost we add.
14892         (add_stmt_costs): New function.
14893         (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
14894         No longer exported.
14895         (vect_analyze_stmt): Adjust prototype.
14896         (vectorizable_condition): Likewise.
14897         (vectorizable_live_operation): Likewise.
14898         (vectorizable_reduction): Likewise.
14899         (vectorizable_induction): Likewise.
14900         * tree-vect-loop.c (vect_analyze_loop_operations): Create local
14901         cost vector to pass to vectorizable_ and record afterwards.
14902         (vect_model_reduction_cost): Take cost vector argument and adjust.
14903         (vect_model_induction_cost): Likewise.
14904         (vectorizable_reduction): Likewise.
14905         (vectorizable_induction): Likewise.
14906         (vectorizable_live_operation): Likewise.
14907         * tree-vect-slp.c (vect_create_new_slp_node): Initialize
14908         SLP_TREE_NUMBER_OF_VEC_STMTS.
14909         (vect_analyze_slp_cost_1): Remove.
14910         (vect_analyze_slp_cost): Likewise.
14911         (vect_slp_analyze_node_operations): Take visited args and
14912         a target cost vector.  Avoid processing already visited stmt sets.
14913         (vect_slp_analyze_operations): Use a local cost vector to gather
14914         costs and register those of non-discarded instances.
14915         (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
14916         (vect_schedule_slp_instance): Remove copying of
14917         SLP_TREE_NUMBER_OF_VEC_STMTS.  Instead assert that it is not
14918         zero.
14919         * tree-vect-stmts.c (record_stmt_cost): Remove path directly
14920         adding cost.  Record cost entry location.
14921         (vect_prologue_cost_for_slp_op): Function to compute cost of
14922         a constant or invariant generated for SLP vect in the prologue,
14923         split out from vect_analyze_slp_cost_1.
14924         (vect_model_simple_cost): Make static.  Adjust for SLP costing.
14925         (vect_model_promotion_demotion_cost): Likewise.
14926         (vect_model_store_cost): Likewise, make static.
14927         (vect_model_load_cost): Likewise.
14928         (vectorizable_bswap): Add cost vector arg and adjust.
14929         (vectorizable_call): Likewise.
14930         (vectorizable_simd_clone_call): Likewise.
14931         (vectorizable_conversion): Likewise.
14932         (vectorizable_assignment): Likewise.
14933         (vectorizable_shift): Likewise.
14934         (vectorizable_operation): Likewise.
14935         (vectorizable_store): Likewise.
14936         (vectorizable_load): Likewise.
14937         (vectorizable_condition): Likewise.
14938         (vectorizable_comparison): Likewise.
14939         (can_vectorize_live_stmts): Likewise.
14940         (vect_analyze_stmt): Likewise.
14941         (vect_transform_stmt): Adjust calls to vectorizable_*.
14942         * tree-vectorizer.c: Include gimple-pretty-print.h.
14943         (dump_stmt_cost): New function.
14945 2018-05-16  Richard Biener  <rguenther@suse.de>
14947         * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
14948         * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
14949         * tree-ssa-dse.c: Include tree-ssa-loop.h.
14950         (check_name): New callback.
14951         (dse_classify_store): Track cycles via a visited bitmap of PHI
14952         defs and simplify handling of in-loop and across loop dead stores
14953         and properly fail for loop-variant refs.  Handle byte-tracking with
14954         multiple defs.  Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
14955         limiting the walk.
14957 2018-05-16  Richard Sandiford  <richard.sandiford@linaro.org>
14959         * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
14960         (vect_get_mask_type_for_stmt): Likewise.
14961         * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
14962         split out from...
14963         (vect_build_slp_tree_1): ...here.  Use vect_get_vector_types_for_stmt
14964         to determine the statement's vector type and the vector type that
14965         should be used for calculating nunits.  Deal with cases in which
14966         the type has to be deferred.
14967         (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
14968         and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
14969         * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
14970         (vect_determine_vf_for_stmt): New functions, split out from...
14971         (vect_determine_vectorization_factor): ...here.
14972         * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
14973         (vect_get_mask_type_for_stmt): New functions, split out from
14974         vect_determine_vectorization_factor.
14976 2018-05-16  Richard Biener  <rguenther@suse.de>
14978         * tree-cfg.c (verify_gimple_assign_ternary): Properly
14979         verify the [VEC_]COND_EXPR embedded comparison.
14981 2018-05-15  Martin Sebor  <msebor@redhat.com>
14983         PR tree-optimization/85753
14984         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
14985         RECORD_TYPE in addition to ARRAY_TYPE.
14987 2018-05-15  Martin Sebor  <msebor@redhat.com>
14989         PR middle-end/85643
14990         * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
14992 2018-05-15  Richard Biener  <rguenther@suse.de>
14994         * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
14995         add by_clobber_p one.  Change algorithm to collect all defs
14996         representing uses we need to walk and try reducing them to
14997         a single one before failing.
14998         (dse_dom_walker::dse_optimize_stmt): Adjust.
15000 2018-05-13  Mark Wielaard  <mark@klomp.org>
15002         * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
15003         (size_of_loc_descr): Likewise.
15004         (output_loc_operands): Likewise.
15005         (output_loc_operands_raw): Likewise.
15006         (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
15007         (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
15008         (hash_loc_operands): Likewise.
15009         (compare_loc_operands): Likewise.
15011 2018-05-14  Mark Wielaard  <mark@klomp.org>
15013         * dwarf2out.c (count_index_addrs): New function.
15014         (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
15016 2018-05-15  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
15018         PR tree-optimization/83648
15019         * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
15020         return value as malloc candidate.
15022 2018-05-15  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
15024         PR ipa/85734
15025         * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite
15026         param as true in call to suggest_attribute.
15028 2018-05-14  Segher Boessenkool  <segher@kernel.crashing.org>
15030         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
15031         -mreadonly-in-sdata.
15033 2018-05-14  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
15035         * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
15036         New pattern.
15037         (aarch64_crypto_aesd_fused): Likewise.
15039 2018-05-14  Wilco Dijkstra  <wdijkstr@arm.com>
15041         * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
15042         (movsi_aarch64): Likewise.
15043         (load_pairsi): Likewise.
15044         (load_pairdi): Likewise.
15045         (store_pairsi): Likewise.
15046         (store_pairdi): Likewise.
15047         (load_pairsf): Likewise.
15048         (load_pairdf): Likewise.
15049         (store_pairsf): Likewise.
15050         (store_pairdf): Likewise.
15051         (zero_extend): Likewise.
15052         (trunc): Swap alternatives.
15053         (fcvt_target): Add '?' to prefer w over r.
15055 2018-05-14  Jakub Jelinek  <jakub@redhat.com>
15057         PR target/85756
15058         * config/i386/i386.md: Disallow non-commutative arithmetics in
15059         last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
15060         optimization.  Use COMMUTATIVE_ARITH_P test rather than != MINUS
15061         in the peephole2 before it.
15063 2018-05-14  Sebastian Peryt  <sebastian.peryt@intel.com>
15065         * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
15066         OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
15067         (ix86_handle_option): Handle -mcldemote.
15068         * config.gcc: New header.
15069         * config/i386/cldemoteintrin.h: New file.
15070         * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
15071         * config/i386/driver-i386.c (host_detect_local_cpu): Detect
15072         -mcldemote.
15073         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
15074         OPTION_MASK_ISA_CLDEMOTE.
15075         * config/i386/i386.c (ix86_target_string): Add -mcldemote.
15076         (ix86_valid_target_attribute_inner_p): Ditto.
15077         (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
15078         (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
15079         (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
15080         * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
15081         * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
15082         (cldemote): New.
15083         * config/i386/i386.opt: Add -mcldemote.
15084         * config/i386/x86intrin.h: New header.
15085         * doc/invoke.texi: Add -mcldemote.
15087 2018-05-14  Richard Biener  <rguenther@suse.de>
15089         * doc/match-and-simplify.texi: Adjust :s documentation.
15091 2018-05-14  Alexander Monakov  <amonakov@ispras.ru>
15093         * sort.cc (REORDER_23): Pass the type for the temporaries instead of
15094         intended memcpy size.
15095         (REORDER_45): Likewise.
15097 2018-05-13  Alexander Monakov  <amonakov@ispras.ru>
15099         * sort.cc: New file.
15100         * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
15101         * vec.c (qsort_chk): Use gcc_qsort.
15102         * Makefile.in (OBJS-libcommon): Add sort.o.
15103         (build/sort.o): New target.  Use it...
15104         (BUILD_RTL): ... here, and...
15105         (build/gencfn-macros): ... here, and...
15106         (build/genmatch): ... here.
15108 2018-05-13  Kito Cheng  <kito.cheng@gmail.com>
15109             Chung-Ju Wu  <jasonwucj@gmail.com>
15111         * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
15112         * config/nds32/nds32-graywolf.md: New file.
15113         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
15114         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
15115         pipeline.
15116         * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
15117         * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
15118         * config/nds32/nds32.md (pipeline_model): Add graywolf.
15119         * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
15120         * config/nds32/pipelines.md: Include n15 settings.
15122 2018-05-13  Kito Cheng  <kito.cheng@gmail.com>
15123             Chung-Ju Wu  <jasonwucj@gmail.com>
15125         * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
15126         * config/nds32/nds32-n13.md: New file.
15127         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
15128         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
15129         pipeline.
15130         * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
15131         * config/nds32/nds32.md (pipeline_model): Add n13.
15132         * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
15133         * config/nds32/pipelines.md: Include n13 settings.
15135 2018-05-13  Kito Cheng  <kito.cheng@gmail.com>
15136             Chung-Ju Wu  <jasonwucj@gmail.com>
15138         * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
15139         * config/nds32/nds32-n10.md: New file.
15140         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
15141         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
15142         pipeline.
15143         * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
15144         * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
15145         * config/nds32/nds32.md (pipeline_model): Add n10.
15146         * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
15147         * config/nds32/pipelines.md: Include n10 settings.
15149 2018-05-13  Monk Chiang  <sh.chiang04@gmail.com>
15150             Kito Cheng  <kito.cheng@gmail.com>
15151             Chung-Ju Wu  <jasonwucj@gmail.com>
15153         * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
15154         * config/nds32/constants.md (unspec_element, unspec_volatile_element):
15155         Add enum values for DSP extension instructions.
15156         * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
15157         New constraints.
15158         * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
15159         sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
15160         New code iterators.
15161         (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
15162         * config/nds32/nds32-dspext.md: New file for DSP implementation.
15163         * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
15164         * config/nds32/nds32-intrinsic.md: Likewise.
15165         * config/nds32/nds32_intrinsic.h: Likewise.
15166         * config/nds32/nds32-md-auxiliary.c: Likewise.
15167         * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
15168         * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
15169         (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
15170         (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
15171         * config/nds32/nds32-protos.h: New declarations for DSP extension.
15172         * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
15173         TYPE_DMAC in switch statement.
15174         * config/nds32/nds32.c: New checking and implementation for DSP
15175         extension instructions.
15176         * config/nds32/nds32.h: Likewise.
15177         * config/nds32/nds32.md: Likewise.
15178         * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
15179         * config/nds32/predicates.md: Implement new predicates for DSP
15180         extension.
15182 2018-05-11  Michael Meissner  <meissner@linux.vnet.ibm.com>
15184         * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
15185         Reformat alternatives and attributes so it is easier to identify
15186         which constraints/attributes go with which instruction.
15187         (mov<mode>_hardfloat32, FMOVE64): Likewise.
15188         (mov<mode>_softfloat32, FMOVE64): Likewise.
15189         (mov<mode>_hardfloat64, FMOVE64): Likewise.
15190         (mov<mode>_softfloat64, FMOVE64): Likewise.
15192 2018-05-11  Kelvin Nilsen  <kelvin@gcc.gnu.org>
15194         * doc/extend.texi (PowerPC Built-in Functions): Rename this
15195         subsection.
15196         (Basic PowerPC Built-in Functions): The new name of the
15197         subsection previously known as "PowerPC Built-in Functions".
15198         (Basic PowerPC Built-in Functions Available on all Configurations):
15199         New subsubsection.
15200         (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
15201         (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
15202         (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
15203         (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
15205 2018-05-11  Martin Jambor  <mjambor@suse.cz>
15207         PR ipa/85655
15208         * ipa-cp.c (intersect_with_plats): Check that the lattice contains
15209         single const.
15211 2018-05-11  Richard Earnshaw  <rearnsha@arm.com>
15213         PR target/85733
15214         * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
15216 2018-05-11  Sebastian Peryt  <sebastian.peryt@intel.com>
15218         * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
15219         OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
15220         (ix86_handle_option): Handle -mwaitpkg.
15221         * config.gcc: New header.
15222         * config/i386/cpuid.h (bit_WAITPKG): New bit.
15223         * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
15224         * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
15225         function type.
15226         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
15227         OPTION_MASK_ISA_WAITPKG.
15228         * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
15229         (ix86_option_override_internal): Add PTA_WAITPKG.
15230         (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
15231         (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
15232         IX86_BUILTIN_TPAUSE.
15233         (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
15234         __builtin_ia32_umwait and __builtin_ia32_tpause.
15235         (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
15236         IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
15237         * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
15238         * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
15239         UNSPECV_TPAUSE): New.
15240         (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
15241         * config/i386/i386.opt: Add -mwaitpkg.
15242         * config/i386/waitpkgintrin.h: New file.
15243         * config/i386/x86intrin.h: New header.
15244         * doc/invoke.texi: Add -mwaitpkg.
15246 2018-05-11  Richard Earnshaw  <rearnsha@arm.com>
15248         PR target/85606
15249         * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
15250         equivalent.
15251         (cortex-m0): Use armv6s-m isa.
15252         (cortex-m0plus): Likewise.
15253         (cortex-m1): Likewise.
15254         (cortex-m0.small-multiply): Likewise.
15255         (cortex-m0plus.small-multiply): Likewise.
15256         (cortex-m1.small-multiply): Likewise.
15258 2018-05-11  Allan Sandfeld Jensen  <allan.jensen@qt.io>
15259             Jakub Jelinek  <jakub@redhat.com>
15261         PR tree-optimization/85692
15262         * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
15263         source permute as well.
15265 2018-05-11  Martin Liska  <mliska@suse.cz>
15267         PR sanitizer/85556
15268         * doc/extend.texi: Document LLVM style format for no_sanitize
15269         attribute.
15271 2018-05-10  Michael Meissner  <meissner@linux.vnet.ibm.com>
15273         * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
15274         mode_supports_vsx_dform_quad to mode_supports_dq_form.
15275         (mode_supports_vsx_dform_quad): Likewise.
15276         (mode_supports_vmx_dform): Move these functions to be next to the
15277         other mode_supports functions.
15278         (mode_supports_dq_form): Likewise.
15279         (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
15280         mode_supports_dq_form.
15281         (reg_offset_addressing_ok_p): Likewise.
15282         (offsettable_ok_by_alignment): Likewise.
15283         (rs6000_legitimate_offset_address_p): Likewise.
15284         (legitimate_lo_sum_address_p): Likewise.
15285         (rs6000_legitimize_address): Likewise.
15286         (rs6000_legitimize_reload_address): Likewise.
15287         (rs6000_secondary_reload_inner): Likewise.
15288         (rs6000_preferred_reload_class): Likewise.
15289         (rs6000_output_move_128bit): Likewise.
15291 2018-05-10  Uros Bizjak  <ubizjak@gmail.com>
15293         * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
15294         Generate SImode target register for null target.
15295         <case IX86_BUILTIN_XGETBV>: Ditto.
15296         <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
15297         * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
15299 2018-05-10  Carl Love  <cel@us.ibm.com>
15301         * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
15302         dcbtt and dcbtstt if operands[2] is 0.
15304 2018-05-10  Uros Bizjak  <ubizjak@gmail.com>
15306         PR target/85693
15307         * config/i386/sse.md (usadv64qi): New expander.
15309 2018-05-10  Segher Boessenkool  <segher@kernel.crashing.org>
15311         * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
15312         altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
15313         -maltivec=be support.
15314         (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
15315         vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
15316         vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
15317         vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
15318         vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
15319         vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
15320         altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
15321         altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
15322         altivec_vsumsws): Adjust.
15323         (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
15324         *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
15325         altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
15326         support.
15327         (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
15328         altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
15329         altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
15330         (altivec_lve<VI_char>x): Delete expand.
15331         (*altivec_lve<VI_char>x_internal): Rename to...
15332         (altivec_lve<VI_char>x): ... this.
15333         (altivec_lvxl_<mode>): Delete expand.
15334         (*altivec_lvxl_<mode>_internal): Rename to ...
15335         (altivec_lvxl_<mode>): ... this.
15336         (altivec_stvxl_<mode>): Delete expand.
15337         (*altivec_stvxl_<mode>_internal): Rename to ...
15338         (altivec_stvxl_<mode>): ... this.
15339         (altivec_stve<VI_char>x): Delete expand.
15340         (*altivec_stve<VI_char>x_internal): Rename to ...
15341         (altivec_stve<VI_char>x): ... this.
15342         (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
15343         doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
15344         reduc_plus_scal_<mode>): Adjust.
15345         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
15346         comment.
15347         (rs6000_cpu_cpp_builtins): Adjust.
15348         (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
15349         * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
15350         altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
15351         * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
15352         -maltivec=be support.
15353         (rs6000_split_vec_extract_var): Adjust.
15354         (rs6000_split_v4si_init): Adjust.
15355         (swap_selector_for_mode): Delete.
15356         (altivec_expand_lvx_be, altivec_expand_stvx_be,
15357         altivec_expand_stvex_be): Delete.
15358         (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
15359         -maltivec=be support.
15360         (rs6000_gimple_fold_builtin): Ditto.
15361         (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
15362         Adjust.
15363         * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
15364         (TARGET_DIRECT_MOVE_64BIT): Adjust.
15365         * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
15366         * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
15367         * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
15368         unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
15369         vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
15370         *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
15371         *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
15372         *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
15373         anonymous split): Adjust.
15374         (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
15375         (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
15377 2018-05-10  Eric Botcazou  <ebotcazou@adacore.com>
15379         * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
15380         when --with-gxx-include-dir is also specified.
15381         * configure: Regenerate.
15383 2018-05-09  Jim Wilson  <jimw@sifive.com>
15385         PR target/84797
15386         * config.gcc (riscv*-*-*): Handle --with-multilib-list.
15387         * config/riscv/t-withmultilib: New.
15388         * config/riscv/withmultilib.h: New.
15389         * doc/install.texi: Document RISC-V --with-multilib-list support.
15391 2018-05-09  Richard Biener  <rguenther@suse.de>
15393         * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
15394         vector.
15395         (vect_bb_vectorization_profitable_p): Adjust.  Compute
15396         actual scalar cost using the cost vector and the add_stmt_cost
15397         machinery.
15399 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
15401         PR rtl-optimization/85645
15402         * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
15403         in the REG_CFA_REGISTER note for LR, don't leave it empty.
15405 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
15407         PR rtl-optimization/85645
15408         * shrink-wrap.c (spread_components): Return a boolean saying if
15409         anything was changed.
15410         (try_shrink_wrapping_separate): Iterate spread_components until
15411         nothing changes anymore.
15413 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
15415         PR rtl-optimization/85645
15416         * regrename.c (build_def_use): Also kill the chains that include the
15417         destination of a REG_CFA_REGISTER note.
15419 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
15421         PR rtl-optimization/85645
15422         *  regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
15423         insn that has a REG_CFA_REGISTER note.
15425 2018-05-09  Richard Sandiford  <richard.sandiford@linaro.org>
15427         * cfgexpand.c (expand_clobber): New function.
15428         (expand_gimple_stmt_1): Use it.
15429         * tree-vect-stmts.c (vect_clobber_variable): New function,
15430         split out from...
15431         (vectorizable_simd_clone_call): ...here.
15432         (vectorizable_store): Emit a clobber either side of an
15433         IFN_STORE_LANES sequence.
15434         (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
15436 2018-05-09  Tom de Vries  <tom@codesourcery.com>
15438         PR target/85626
15439         * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
15440         (define_insn "trap_if_false"): Add exit after trap.
15442 2018-05-09  Eric Botcazou  <ebotcazou@adacore.com>
15444         PR rtl-optimization/85638
15445         * bb-reorder.c: Include common/common-target.h.
15446         (create_forwarder_block): New function extracted from...
15447         (fix_up_crossing_landing_pad): ...here.  Rename into...
15448         (dw2_fix_up_crossing_landing_pad): ...this.
15449         (sjlj_fix_up_crossing_landing_pad): New function.
15450         (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
15451         call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
15452         from both partitions and exit the loop after one iteration.
15454 2018-05-08  Kelvin Nilsen  <kelvin@gcc.gnu.org>
15456         Revert:
15457         * doc/extend.texi (PowerPC Built-in Functions): Rename this
15458         subsection.
15459         (Basic PowerPC Built-in Functions): The new name of the
15460         subsection previously known as "PowerPC Built-in Functions".
15461         (Basic PowerPC Built-in Functions Available on all Configurations):
15462         New subsubsection.
15463         (Basic PowerPC Built-in Functions Available on ISA 2.05): New
15464         subsubsection.
15465         (Basic PowerPC Built-in Functions Available on ISA 2.06): New
15466         subsubsection.
15467         (Basic PowerPC Built-in Functions Available on ISA 2.07): New
15468         subsubsection.
15469         (Basic PowerPC Built-in Functions Available on ISA 3.0): New
15470         subsubsection.
15472 2018-05-08  Jim Wilson  <jimw@sifive.com>
15474         * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
15475         (LD_EMUL_SUFFIX): New.
15476         (LINK_SPEC): Use it.
15478 2018-05-08  Kelvin Nilsen  <kelvin@gcc.gnu.org>
15480         * doc/extend.texi (PowerPC Built-in Functions): Rename this
15481         subsection.
15482         (Basic PowerPC Built-in Functions): The new name of the
15483         subsection previously known as "PowerPC Built-in Functions".
15484         (Basic PowerPC Built-in Functions Available on all Configurations):
15485         New subsubsection.
15486         (Basic PowerPC Built-in Functions Available on ISA 2.05): New
15487         subsubsection.
15488         (Basic PowerPC Built-in Functions Available on ISA 2.06): New
15489         subsubsection.
15490         (Basic PowerPC Built-in Functions Available on ISA 2.07): New
15491         subsubsection.
15492         (Basic PowerPC Built-in Functions Available on ISA 3.0): New
15493         subsubsection.
15495 2018-05-08  Jakub Jelinek  <jakub@redhat.com>
15497         PR target/85683
15498         * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
15499         after cmpelim optimization.
15501 2018-05-08  Olga Makhotina  <olga.makhotina@intel.com>
15503         * config.gcc: Support "goldmont".
15504         * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
15505         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
15506         PROCESSOR_GOLDMONT.
15507         * config/i386/i386.c (m_GOLDMONT): Define.
15508         (processor_target_table): Add "goldmont".
15509         (PTA_GOLDMONT): Define.
15510         (ix86_lea_outperforms): Add TARGET_GOLDMONT.
15511         (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
15512         (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
15513         (fold_builtin_cpu): Add "goldmont".
15514         (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
15515         (ix86_option_override_internal): Add "goldmont".
15516         * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
15517         (processor_type): Add PROCESSOR_GOLDMONT.
15518         * config/i386/i386.md: Add CPU "glm".
15519         * config/i386/glm.md: New file.
15520         * config/i386/x86-tune.def: Add m_GOLDMONT.
15521         * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
15523 2018-05-08  Jakub Jelinek  <jakub@redhat.com>
15525         PR target/85572
15526         * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
15527         E_V4DImode.
15528         * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
15529         VI1248_AVX512VL_AVX512BW.  Handle V2DImode and V4DImode if not
15530         TARGET_AVX512VL using ix86_expand_sse2_abs.  Formatting fixes.
15532         PR target/85317
15533         * config/i386/i386.c (ix86_fold_builtin): Handle
15534         IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
15536         PR target/85480
15537         * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
15538         (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
15540 2018-05-08  Richard Earnshaw  <rearnsha@arm.com>
15542         PR target/85658
15543         * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
15544         (check_arch): Likewise.
15545         (check_fpu): Return the result rather than printing it.
15546         (end arch): Fix operator precedence.
15547         (end cpu): Likewise.
15548         (END): Print the result from check_fpu.
15550 2018-05-08  Richard Sandiford  <richard.sandiford@linaro.org>
15551             Alan Hayward  <alan.hayward@arm.com>
15552             David Sherwood  <david.sherwood@arm.com>
15554         * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
15555         (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
15556         (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
15557         (*fcmuo<mode>_and): New patterns.
15559 2018-05-08  Richard Sandiford  <richard.sandiford@linaro.org>
15561         * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
15562         (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
15563         (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
15564         (cmp_op, sve_imm_con): New code attributes.
15565         (SVE_COND_INT_CMP, imm_con): Delete.
15566         (cmp_op): Remove above unspecs from int attribute.
15567         * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
15568         to...
15569         (*cmp<cmp_op><mode>): ...this.  Use UNSPEC_MERGE_PTRUE instead of
15570         comparison-specific unspecs.
15571         (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
15572         (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
15573         (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
15574         (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
15575         (*vec_fcm<cmp_op><mode>): Rename to...
15576         (*fcm<cmp_op><mode>): ...this and adjust likewise.
15577         (*vec_fcmuo<mode>): Rename to...
15578         (*fcmuo<mode>): ...this and adjust likewise.
15579         (*pred_fcm<cmp_op><mode>): New pattern.
15580         * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
15581         (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
15582         functions.
15583         (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
15584         and UNORDERED.
15585         (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
15586         (aarch64_emit_sve_predicated_cond): New function.
15587         (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
15588         (aarch64_emit_unspec_cond_or): Replace with...
15589         (aarch64_emit_sve_or_conds): ...this new function.  Use
15590         aarch64_emit_sve_ptrue_op for the individual comparisons and
15591         aarch64_emit_binop to OR them together.
15592         (aarch64_emit_inverted_unspec_cond): Replace with...
15593         (aarch64_emit_sve_inverted_cond): ...this new function.  Use
15594         aarch64_emit_sve_ptrue_op for the comparison and
15595         aarch64_emit_unop to invert the result.
15596         (aarch64_expand_sve_vec_cmp_float): Update after the above
15597         changes.  Use aarch64_emit_sve_ptrue_op for native comparisons.
15599 2018-05-07  Nathan Sidwell  <nathan@acm.org>
15601         * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
15602         * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
15603         (Backwards Compatibility): Likewise.
15605 2018-05-07  Luis Machado  <luis.machado@linaro.org>
15607         PR bootstrap/85681
15608         Revert:
15609         2018-05-07  Luis Machado  <luis.machado@linaro.org>
15611         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
15612         <prefetch_dynamic_strides>: New const bool field.
15613         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
15614         prefetch_dynamic_strides.
15615         (exynosm1_prefetch_tune): Likewise.
15616         (thunderxt88_prefetch_tune): Likewise.
15617         (thunderx_prefetch_tune): Likewise.
15618         (thunderx2t99_prefetch_tune): Likewise.
15619         (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
15620         to false.
15621         (aarch64_override_options_internal): Update to set
15622         PARAM_PREFETCH_DYNAMIC_STRIDES.
15623         * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
15624         * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
15625         * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
15626         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
15627         prefetch-dynamic-strides setting.
15629         2018-05-07  Luis Machado  <luis.machado@linaro.org>
15631         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
15632         <minimum_stride>: New const int field.
15633         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
15634         minimum_stride field.
15635         (exynosm1_prefetch_tune): Likewise.
15636         (thunderxt88_prefetch_tune): Likewise.
15637         (thunderx_prefetch_tune): Likewise.
15638         (thunderx2t99_prefetch_tune): Likewise.
15639         (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
15640         (aarch64_override_options_internal): Update to set
15641         PARAM_PREFETCH_MINIMUM_STRIDE.
15642         * doc/invoke.texi (prefetch-minimum-stride): Document new option.
15643         * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
15644         * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
15645         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
15646         stride is constant and is below the minimum stride threshold.
15648 2018-05-07  Luis Machado  <luis.machado@linaro.org>
15650         * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
15651         to 512.
15653 2018-05-07  Luis Machado  <luis.machado@linaro.org>
15655         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
15656         <prefetch_dynamic_strides>: New const bool field.
15657         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
15658         prefetch_dynamic_strides.
15659         (exynosm1_prefetch_tune): Likewise.
15660         (thunderxt88_prefetch_tune): Likewise.
15661         (thunderx_prefetch_tune): Likewise.
15662         (thunderx2t99_prefetch_tune): Likewise.
15663         (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
15664         to false.
15665         (aarch64_override_options_internal): Update to set
15666         PARAM_PREFETCH_DYNAMIC_STRIDES.
15667         * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
15668         * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
15669         * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
15670         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
15671         prefetch-dynamic-strides setting.
15673 2018-05-07  Luis Machado  <luis.machado@linaro.org>
15675         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
15676         <minimum_stride>: New const int field.
15677         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
15678         minimum_stride field.
15679         (exynosm1_prefetch_tune): Likewise.
15680         (thunderxt88_prefetch_tune): Likewise.
15681         (thunderx_prefetch_tune): Likewise.
15682         (thunderx2t99_prefetch_tune): Likewise.
15683         (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
15684         (aarch64_override_options_internal): Update to set
15685         PARAM_PREFETCH_MINIMUM_STRIDE.
15686         * doc/invoke.texi (prefetch-minimum-stride): Document new option.
15687         * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
15688         * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
15689         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
15690         stride is constant and is below the minimum stride threshold.
15692 2018-05-06  Jakub Jelinek  <jakub@redhat.com>
15694         PR c++/85659
15695         * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
15696         the type is addressable.  Don't force op into register if it has
15697         BLKmode.
15699 2018-05-05  Roland McGrath  <mcgrathr@google.com>
15701         PR other/77609
15702         * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
15703         any section for which we don't know a specific type it should have,
15704         regardless of name.  Previously this was done only for the exact
15705         names ".init_array", ".fini_array", and ".preinit_array".
15706         (default_elf_asm_named_section): Add comment about
15707         relationship with default_section_type_flags and SECTION_NOTYPE.
15708         (get_section): Don't consider it a type conflict if one side has
15709         SECTION_NOTYPE and the other doesn't, as long as neither has the
15710         SECTION_BSS et al used in the default_section_type_flags logic.
15712 2018-05-05  Tom de Vries  <tom@codesourcery.com>
15714         PR target/85653
15715         * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
15716         (workaround_barsyncs): New function.
15717         (nvptx_reorg): Use workaround_barsyncs.
15718         * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
15719         (define_expand "nvptx_membar_cta"): New define_expand.
15720         (define_insn "*nvptx_membar_cta"): New insn.
15722 2018-05-04  Pekka Jääskeläinen  <pekka.jaaskelainen@parmance.com>
15724         * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
15725         To improve optimization opportunities.
15726         * builtin-types.def: The new needed builtin types for the above.
15728 2018-05-04  Richard Biener  <rguenther@suse.de>
15730         * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
15731         * gimple-ssa-store-merging.c
15732         (imm_store_chain_info::output_merged_store): Remove redundant create,
15733         release split_store vector contents on failure.
15734         * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
15735         scalar stmt vector on cache hit.
15737 2018-05-04  Segher Boessenkool  <segher@kernel.crashing.org>
15739         * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
15740         Xilinx FP support.
15741         * config.gcc (powerpc-xilinx-eabi*): Remove.
15742         * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
15743         support.
15744         (fusion_addis_mem_combo_load): Ditto.
15745         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
15746         FP support.
15747         (rs6000_cpu_cpp_builtins): Ditto.
15748         * config/rs6000/rs6000-linux.c
15749         (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
15750         * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
15751         * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
15752         support.
15753         (rs6000_setup_reg_addr_masks): Ditto.
15754         (rs6000_init_hard_regno_mode_ok): Ditto.
15755         (rs6000_option_override_internal): Ditto.
15756         (legitimate_lo_sum_address_p): Ditto.
15757         (rs6000_legitimize_address): Ditto.
15758         (rs6000_legitimize_reload_address): Ditto.
15759         (rs6000_legitimate_address_p): Ditto.
15760         (abi_v4_pass_in_fpr): Ditto.
15761         (setup_incoming_varargs): Ditto.
15762         (rs6000_gimplify_va_arg): Ditto.
15763         (rs6000_split_multireg_move): Ditto.
15764         (rs6000_savres_strategy): Ditto.
15765         (rs6000_emit_prologue_components): Ditto.
15766         (rs6000_emit_epilogue_components): Ditto.
15767         (rs6000_emit_prologue): Ditto.
15768         (rs6000_emit_epilogue): Ditto.
15769         (rs6000_elf_file_end): Ditto.
15770         (rs6000_function_value): Ditto.
15771         (rs6000_libcall_value): Ditto.
15772         * config/rs6000/rs6000.h: Ditto.
15773         (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
15774         (TARGET_MINMAX): ... this.  New.
15775         (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
15776         * config/rs6000/rs6000.md: Remove Xilinx FP support.
15777         (*movsi_internal1_single): Delete.
15778         * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
15779         mfpu=, mxilinx-fpu): Delete.
15780         * config/rs6000/singlefp.h: Delete.
15781         * config/rs6000/sysv4.h: Remove Xilinx FP support.
15782         * config/rs6000/t-rs6000: Ditto.
15783         * config/rs6000/t-xilinx: Delete.
15784         * config/rs6000/titan.md: Adjust for fp_type removal.
15785         * config/rs6000/vsx.md: Remove Xilinx FP support.
15786         (VStype_simple): Delete.
15787         (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
15788         * config/rs6000/xfpu.h: Delete.
15789         * config/rs6000/xfpu.md: Delete.
15790         * config/rs6000/xilinx.h: Delete.
15791         * config/rs6000/xilinx.opt: Delete.
15792         * doc/invoke.texi (RS/6000 and PowerPC Options): Remove
15793         -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
15795 2018-05-04  Tom de Vries  <tom@codesourcery.com>
15797         PR libgomp/85639
15798         * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
15799         if ignore == 0.
15801 2018-05-04  Richard Biener  <rguenther@suse.de>
15803         PR middle-end/85627
15804         * tree-complex.c (update_complex_assignment): We are always in SSA form.
15805         (expand_complex_div_wide): Likewise.
15806         (expand_complex_operations_1): Likewise.
15807         (expand_complex_libcall): Preserve EH info of the original stmt.
15808         (tree_lower_complex): Handle removed blocks.
15809         * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
15810         on complex multiplication and division libcall builtins.
15812 2018-05-04  Richard Biener  <rguenther@suse.de>
15814         PR middle-end/85574
15815         * fold-const.c (negate_expr_p): Restrict negation of operand
15816         zero of a division to when we know that can happen without
15817         overflow.
15818         (fold_negate_expr_1): Likewise.
15820 2018-05-04  Jakub Jelinek  <jakub@redhat.com>
15822         PR libstdc++/85466
15823         * real.h (real_nextafter): Declare.
15824         * real.c (real_nextafter): New function.
15825         * fold-const-call.c (fold_const_nextafter): New function.
15826         (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
15827         CASE_CFN_NEXTTOWARD.
15828         (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
15829         even when arg1_mode is different from arg0_mode.
15831 2018-05-03  Nathan Sidwell  <nathan@acm.org>
15833         * doc/extend.texi (Deprecated Features): Remove
15834         -ffriend-injection.
15835         (Backwards Compatibility): Likewise.
15836         * doc/invoke.texi (C++ Language Options): Likewise.
15837         (C++ Dialect Options): Likewise.
15839 2018-05-03  Jakub Jelinek  <jakub@redhat.com>
15841         PR target/85530
15842         * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
15843         _mm512_mask_mullox_epi64): New intrinsics.
15845 2018-05-03  Tom de Vries  <tom@codesourcery.com>
15847         PR testsuite/85106
15848         * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
15849         dump files): Add offload-tree.
15851 2018-05-03  Richard Biener  <rguenther@suse.de>
15853         PR tree-optimization/85615
15854         * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
15855         to loops not nested in BBs loop father to avoid creating multi-entry
15856         loops.
15858 2018-05-03  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
15860         PR tree-optimization/70291
15861         * tree-complex.c (expand_complex_libcall): Add type, inplace_p
15862         arguments.  Change return type to tree.  Emit libcall as a new
15863         statement rather than replacing existing one when inplace_p is true.
15864         (expand_complex_multiplication_components): New function.
15865         (expand_complex_multiplication): Expand floating-point complex
15866         multiplication using the above.
15867         (expand_complex_division): Rename inner_type parameter to type.
15868         Update expand_complex_libcall call-site.
15869         (expand_complex_operations_1): Update expand_complex_multiplication
15870         and expand_complex_division call-sites.
15872 2018-05-02  Jakub Jelinek  <jakub@redhat.com>
15874         PR target/85582
15875         * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
15876         *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
15877         *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
15878         the highest significant bit of the shift count mask is clear.  In
15879         check whether and[sq]i3 is needed verify that all significant bits
15880         of the shift count other than the highest are set.
15882 2018-05-02  Tom de Vries  <tom@codesourcery.com>
15884         PR libgomp/82428
15885         * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
15886         * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
15887         (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
15888         * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
15889         (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
15890         * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
15891         __builtin_goacc_parlevel_size.
15893 2018-05-02  Richard Biener  <rguenther@suse.de>
15895         PR tree-optimization/85597
15896         * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
15897         do not use split vect_get_vec_defs call but call vect_get_slp_defs
15898         directly.
15900 2018-05-02  Tom de Vries  <tom@codesourcery.com>
15902         PR testsuite/85106
15903         * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
15904         dump files): Add ltrans-tree.
15906 2018-05-02  Tom de Vries  <tom@codesourcery.com>
15908         PR testsuite/85106
15909         * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
15910         dump files): Add wpa-ipa.
15912 2018-05-02  Segher Boessenkool  <segher@kernel.crashing.org>
15914         * config.gcc (powerpc*-*-*): Remove paired.h.  Unsupport the
15915         powerpc*-*-linux*paired* target.
15916         * config/rs6000/750cl.h: Delete.
15917         * config/rs6000/paired.h: Delete.
15918         * config/rs6000/paired.md: Delete.
15919         * config/rs6000/predicates.md (easy_vector_constant): Remove paired
15920         float support.
15921         * config/rs6000/rs6000-builtin.def: Remove paired float support.
15922         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
15923         comment.  Remove paired float support.
15924         * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
15925         * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
15926         VECTOR_PAIRED.
15927         * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
15928         paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
15929         declarations.
15930         * config/rs6000/rs6000.c: Remove paired float support.
15931         (paired_expand_vector_init, paired_expand_vector_move,
15932         paired_emit_vector_compare, paired_emit_vector_cond_expr,
15933         (paired_expand_lv_builtin, paired_expand_stv_builtin,
15934         paired_expand_builtin, paired_expand_predicate_builtin,
15935         paired_init_builtins): Delete.
15936         * config/rs6000/rs6000.h: Remove paired float support.
15937         * config/rs6000/rs6000.md: Remove paired float support.
15938         (move_from_CR_ov_bit): Delete.
15939         * config/rs6000/rs6000.opt (mpaired): Delete.
15940         * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
15941         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
15943 2018-05-02  Richard Biener  <rguenther@suse.de>
15945         PR middle-end/85567
15946         * gimplify.c (gimplify_save_expr): When in SSA form allow
15947         SAVE_EXPRs to compute to SSA vars.
15949 2018-05-02  Jakub Jelinek  <jakub@redhat.com>
15951         PR target/85582
15952         * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
15953         *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
15954         *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
15955         clobber operands[2], instead use a new pseudo.  Formatting fixes.
15957 2018-05-02  Richard Sandiford  <richard.sandiford@linaro.org>
15959         PR tree-optimization/85586
15960         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
15961         exit early for statements in the same group if the accesses are
15962         not strided.
15964 2018-05-02  Tom de Vries  <tom@codesourcery.com>
15966         PR lto/85451
15967         * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
15968         error message.
15970 2018-05-01  Marc Glisse  <marc.glisse@inria.fr>
15972         PR tree-optimization/85143
15973         * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
15975 2018-05-01  Tom de Vries  <tom@codesourcery.com>
15977         PR lto/85451
15978         * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
15979         not found" error message.
15981 2018-05-01  Tom de Vries  <tom@codesourcery.com>
15983         PR other/83786
15984         * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
15985         * vec.c (test_ordered_remove_if): New function.
15986         (vec_c_tests): Call test_ordered_remove_if.
15987         * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
15988         * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
15989         * tree-vect-patterns.c (vect_pattern_recog_1): Use
15990         VEC_ORDERED_REMOVE_IF.
15992 2018-05-01  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
15994         PR tree-optimization/82665
15995         * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
15996         pointer subtraction where arguments come from a memchr call.
15998 2018-05-01  Jakub Jelinek  <jakub@redhat.com>
16000         * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
16001         --push-state --as-needed and --pop-state instead of --as-needed and
16002         --no-as-needed if ld supports it.
16003         * configure: Regenerated.
16005         PR web/85578
16006         * doc/install.texi2html: Replace _002d with - and _002a with * in
16007         generated html files using sed.
16009 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
16011         PR c++/85523
16012         * gcc-rich-location.c (blank_line_before_p): New function.
16013         (use_new_line): New function.
16014         (gcc_rich_location::add_fixit_insert_formatted): New function.
16015         * gcc-rich-location.h
16016         (gcc_rich_location::add_fixit_insert_formatted): New function.
16018 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
16020         * selftest.c (assert_streq): Rename "expected" and "actual" to
16021         "val1" and "val2".  Extend NULL-handling to cover both inputs
16022         symmetrically, while still requiring both to be non-NULL for a pass.
16023         * selftest.h (assert_streq): Rename "expected" and "actual" to
16024         "val1" and "val2".
16025         (ASSERT_EQ): Likewise.
16026         (ASSERT_EQ_AT): Likewise.
16027         (ASSERT_KNOWN_EQ): Likewise.
16028         (ASSERT_KNOWN_EQ_AT): Likewise.
16029         (ASSERT_NE): Likewise.
16030         (ASSERT_MAYBE_NE): Likewise.
16031         (ASSERT_MAYBE_NE_AT): Likewise.
16032         (ASSERT_STREQ): Likewise.  Clarify that both must be non-NULL for
16033         the assertion to pass.
16034         (ASSERT_STREQ_AT): Likewise.
16036 2018-04-30  Jonathan Wakely  <jwakely@redhat.com>
16038         * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
16039         interaction with -pie.
16041 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
16043         * selftest.h: Fix alphabetization of per-source-file selftest
16044         declarations.
16046 2018-04-30  Jason Merrill  <jason@redhat.com>
16048         PR c++/61982 - dead stores to destroyed objects.
16049         * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
16050         of clobber.
16052 2018-04-30  Jason Merrill  <jason@redhat.com>
16054         * tree.c (build_clobber): New.
16055         * tree.h: Declare it.
16056         * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
16058 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
16060         * diagnostic-show-locus.c (layout::layout): Update for
16061         location_get_source_line returning a char_span.
16062         (struct char_span): Move to input.h.
16063         (struct correction): Update for fields in char_span becoming
16064         private.
16065         (struct source_line): Update for location_get_source_line
16066         returning a char_span.
16067         (layout::print_line): Likewise.
16068         * edit-context.c (edited_file::print_content): Likewise.
16069         (edited_file::print_diff_hunk): Likewise.
16070         (edited_file::print_run_of_changed_lines): Likewise.
16071         (edited_file::get_num_lines): Likewise.
16072         (edited_line::edited_line): Likewise.
16073         * final.c (asm_show_source): Likewise.
16074         * input.c (location_get_source_line): Convert return type
16075         from const char * to char_span, losing the final "line_len"
16076         param.
16077         (dump_location_info): Update for the above.
16078         (get_substring_ranges_for_loc): Likewise.  Use a char_span
16079         when handling the literal within the line.
16080         (test_reading_source_line): Update for location_get_source_line
16081         returning a char_span.
16082         * input.h (class char_span): Move here from
16083         diagnostic-show-locus.c, converting from a struct to a class.
16084         Make data members private.
16085         (char_span::operator bool): New.
16086         (char_span::length): New.
16087         (char_span::get_buffer): New.
16088         (char_span::operator[]): New.
16089         (char_span::subspan): Make const.
16090         (char_span::xstrdup): New.
16091         (location_get_source_line): Convert return type from const char *
16092         to char_span, losing the final "line_size" param.
16094 2018-04-30  Jan Hubicka  <jh@suse.cz>
16096         * lto-wrapper.c (ltrans_priorities): New static var.
16097         (cmp_priority): New.
16098         (run_gcc): Read priorities and if doing parallel build order
16099         the Makefile by them.
16101 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
16103         * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
16105 2018-04-30  Richard Biener  <rguenther@suse.de>
16107         * tree-cfg.c (verify_address): Remove base argument, add
16108         flag whether to check TREE_ADDRESSABLE and do that.
16109         (verify_expr): Remove.
16110         (verify_types_in_gimple_reference): Add pieces from verify_expr.
16111         (verify_gimple_assign_single): Likewise.
16112         (verify_gimple_switch): Likewise.
16113         (verify_expr_location_1): Dereference tp once.  Add (disabled)
16114         piece from verify_expr.
16115         (verify_gimple_in_cfg): Do not call verify_expr on all ops.
16117 2018-04-30  Claudiu Zissulescu  <claziss@synopsys.com>
16119         * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
16121 2018-04-30  Claudiu Zissulescu  <claziss@synopsys.com>
16123         * config/arc/arc-protos.h (prepare_extend_operands): Remove.
16124         (small_data_pattern): Likewise.
16125         (arc_rewrite_small_data): Likewise.
16126         * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
16127         (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
16128         (get_symbol_alignment): New function.
16129         (legitimate_small_data_address_p): Likewise.
16130         (legitimate_scaled_address): Update, call
16131         legitimate_small_data_address_p.
16132         (output_sdata): New static variable.
16133         (arc_print_operand): Update how we handle small data operands.
16134         (arc_print_operand_address): Likewise.
16135         (arc_legitimate_address_p): Update, use
16136         legitimate_small_data_address_p.
16137         (arc_rewrite_small_data_p): Remove.
16138         (arc_rewrite_small_data_1): Likewise.
16139         (arc_rewrite_small_data): Likewise.
16140         (small_data_pattern): Likewise.
16141         (compact_sda_memory_operand): Update to use
16142         legitimate_small_data_address_p and get_symbol_alignment.
16143         (prepare_move_operands): Don't rewite sdata pattern.
16144         (prepare_extend_operands): Remove.
16145         * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
16146         pattern.
16147         (zero_extendqisi2): Likewise.
16148         (zero_extendhisi2): Likewise.
16149         (extendqihi2): Likewise.
16150         (extendqisi2): Likewise.
16151         (extendhisi2): Likewise.
16152         (addsi3): Likewise.
16153         (subsi3): Likewise.
16154         (andsi3): Likewise.
16155         * config/arc/constraints.md (Usd): Change it to memory constraint.
16157 2018-04-30  Claudiu Zissulescu  <claziss@synopsys.com>
16159         * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
16160         as source of std instructions.
16161         * config/arc/arc.md (movsi_insn): Update pattern predicate to
16162         allow 6-bit constants as source for store instructions.
16163         (movdi_insn): Update instruction pattern to allow 6-bit constants
16164         as source for store instructions.
16166 2018-04-30  Jonathan Wakely  <jwakely@redhat.com>
16168         * doc/invoke.texi (-fdebug-types-section): Fix grammar.
16170 2018-04-30  Nathan Sidwell  <nathan@acm.org>
16171             Sandra Loosemore <sandra@codesourcery.com>
16173         * dumpfile.c (dump_open): Allow '-' for stdout.
16174         * doc/invoke.texi (Developer Options): Document dump filename
16175         determination early.  Document stdin/stdout selection.
16177 2018-04-30  Andrew Sadek  <andrew.sadek.se@gmail.com>
16179         Microblaze Target: PIC data text relative
16181         * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
16182         * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
16183         Add declaration.
16184         * config/microblaze/microblaze.h (microblaze_constant_address_p):
16185         CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
16186         * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
16187         New addressing mode for data-text relative position indepenedent code.
16188         (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
16189         'ADDRESS_SYMBOLIC_TXT_REL'.
16190         (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
16191         (microblaze_legitimate_pic_operand): Exclude function calls from
16192         pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
16193         (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
16194         addresses cases.
16195         (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
16196         (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
16197         (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
16198         for 'address + offset'.
16199         (microblaze_expand_prologue): Add new function prologue call for
16200         'r20' assignation.
16201         (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
16202         'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
16203         table in case of TARGET_PIC_DATA_TEXT_REL.
16204         (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
16205         * config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
16206         Add new macros 'UNSPEC_TEXT',
16207         'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
16208         + exclude function calls from 'UNSPEC_PLT' in case of data text
16209         relative mode.
16210         * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
16211         new target hook for generating address diff vector tables in case of
16212         flag_pic.
16213         * doc/tm.texi : Regenerate.
16214         * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
16215         'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
16216         of addr diff vector generation.
16217         * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
16218         target hook definition.
16219         * targhooks.h, targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
16220         Add default function for generate_pic_addr_diff_vec -> flag_pic.
16221         * doc/invoke.texi (Add new pic option): Add new microblaze pic
16222         option for data text relative.
16224 2018-04-30  Richard Biener  <rguenther@suse.de>
16226         * tree-chrec.h (evolution_function_is_constant_p): Remove
16227         redundant check.
16228         * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
16230 2018-04-30  Richard Biener  <rguenther@suse.de>
16232         PR bootstrap/85571
16233         * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
16235 2018-04-30  Richard Biener  <rguenther@suse.de>
16237         PR tree-optimization/28364
16238         PR tree-optimization/85275
16239         * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
16240         copying first exit test.
16242 2018-04-28  Mark Wielaard  <mark@klomp.org>
16244         * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
16245         dwarf_version >= 5.
16246         (dwarf_AT): Handle DW_AT_addr_base.
16247         (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
16249 2018-04-28  Uros Bizjak  <ubizjak@gmail.com>
16251         PR target/84431
16252         * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
16253         (*ashl<dwi>3_doubleword_mask_1): Ditto.
16254         (*<shift_insn><dwi>3_doubleword_mask): Ditto.
16255         (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
16257 2018-04-28  Richard Biener  <rguenther@suse.de>
16259         * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
16260         (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
16261         to reflect use.  Only add interesting stmts.
16263 2018-04-27  Martin Jambor  <mjambor@suse.cz>
16265         PR ipa/85549
16266         * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
16267         the jump function allows for passing through aggregate values.
16269 2018-04-27  David Malcolm  <dmalcolm@redhat.com>
16271         * input.h (in_system_header_at): Convert from macro to inline
16272         function.
16273         (from_macro_expansion_at): Likewise.
16274         (from_macro_definition_at): Likewise.
16276 2018-04-27  Jeff Law  <law@redhat.com>
16278         * config.gcc: Mark tile* targets as deprecated/obsolete.
16280 2018-04-27  Richard Biener  <rguenther@suse.de>
16282         * config/aarch64/aarch64.c: Simplify ap.__stack advance and
16283         fix for ILP32.
16285 2018-04-27  Richard Biener  <rguenther@suse.de>
16287         * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
16289 2018-04-27  Uros Bizjak  <ubizjak@gmail.com>
16291         * config/i386/i386.md (*movti_internal): Substitute Ye constraint
16292         with Yd constraint. Set "preferred_for_speed" attribute from
16293         TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
16294         with Yd constraint.
16295         (*movdi_internal): Ditto.
16296         (movti_interunit splitters): Remove
16297         TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
16298         (movdi_interunit splitters): Ditto.
16299         * config/i386/constraints.md (Ye): Remove.
16300         (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
16302 2018-04-27  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
16304         PR target/85512
16305         * config/aarch64/constraints.md (Usg): Limit to 31.
16306         (Usj): Limit to 63.
16308 2018-04-27  Jakub Jelinek  <jakub@redhat.com>
16310         PR tree-optimization/85529
16311         * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
16312         argument.  Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
16313         rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
16314         zero extension or masking of the MSB bit.
16315         (optimize_range_tests): Add FIRST_BB argument, pass it through
16316         to optimize_range_tests_var_bound.
16317         (maybe_optimize_range_tests, reassociate_bb): Adjust
16318         optimize_range_tests callers.
16320 2018-04-26  Richard Biener  <rguenther@suse.de>
16321             Jakub Jelinek  <jakub@redhat.com>
16323         * cgraph.h (symbol_table): Just declare debug method here.
16324         * symtab.c (symbol_table::debug): Define.
16326 2018-04-26  Eric Botcazou  <ebotcazou@adacore.com>
16328         * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
16330 2018-04-26  Uros Bizjak  <ubizjak@gmail.com>
16332         * config/i386/i386.md ("isa" attribute): Add x64_sse2.
16333         ("enabled" attribute): Handle x64_sse2 "isa" attribute.
16334         (*movdi_internal): Substitute Yi and Yj constraint with x
16335         and Ym and Yn constraint with y constraint.  Update "isa"
16336         attribute and set "preferred_for_speed" attribute from
16337         TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
16338         (*movsi_internal): Ditto.
16339         (*movdf_internal): Ditto.
16340         (*movsf_internal): Ditto.
16341         (*zero_extendsidi2): Ditto.
16342         * config/i386/sse.md (vec_set<mode>_0): Ditto.
16343         (sse2_loadld): Ditto.
16344         (*vec_extract<ssevecmodelower>_0): Ditto.
16345         (*vec_extractv4si_0_zext_sse4): Ditto.
16346         (vec_concatv2di): Ditto.
16347         (*vec_dup<mode>): Ditto.
16348         * config/i386/mmx.md (*mov<mode>_internal): Ditto.
16349         * config/i386/constraints.md (Yi): Remove.
16350         (Yj): Remove.
16351         (Ym): Remove.
16352         (Yn): Remove.
16354 2018-04-26  Nathan Sidwell  <nathan@acm.org>
16356         * dumpfile.c (dump_open): New.
16357         (dump_open_alternate_stream, dump_start, dump_begin): Call it.
16358         (dump_finish): Detect stdio/stderr by value not name.
16360 2018-04-26  Jonathan Wakely  <jwakely@redhat.com>
16362         * doc/invoke.texi (-Wreturn-type): Document default status for C++.
16364 2018-04-26  Tom de Vries  <tom@codesourcery.com>
16366         PR target/84952
16367         * config/nvptx/nvptx.c (verify_neutering_jumps)
16368         (verify_neutering_labels): New function
16369         (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
16371 2018-04-26  Tom de Vries  <tom@codesourcery.com>
16373         PR target/84025
16374         * config/nvptx/nvptx.c (needs_neutering_p): New function.
16375         (nvptx_single): Use needs_neutering_p to skip over insns that do not
16376         need neutering.
16378 2018-04-26  Richard Biener <rguenther@suse.de>
16379             Tom de Vries  <tom@codesourcery.com>
16381         PR lto/85422
16382         * lto-streamer-out.c (output_function): Fixup loops if required to match
16383         discovery done in the reader.
16385 2018-04-26  Richard Biener  <rguenther@suse.de>
16387         PR tree-optimization/85116
16388         * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
16389         have a loop exit from the single latch predecessor.  Remove
16390         case of header with just condition.
16391         (ch_base::copy_headers): Exclude infinite loops from any
16392         processing.
16393         (pass_ch::execute): Record exits.
16395 2018-04-26  Richard Biener  <rguenther@suse.de>
16397         * tree-vect-data-refs.c (vect_get_data_access_cost): Get
16398         prologue cost vector and pass it to vect_get_load_cost.
16399         (vect_get_peeling_costs_all_drs): Likewise.
16400         (vect_peeling_hash_get_lowest_cost): Likewise.
16401         (vect_enhance_data_refs_alignment): Likewise.
16403 2018-04-26  Richard Biener  <rguenther@suse.de>
16405         PR middle-end/85450
16406         * tree-cfg.c (verify_gimple_assign_unary): Restore proper
16407         checking of integer<->pointer conversions.
16408         * omp-expand.c (expand_omp_for_static_nochunk): Avoid
16409         sign-/zero-extending pointer types.
16410         (expand_omp_for_static_chunk): Likewise.
16412 2018-03-22  Hans-Peter Nilsson  <hp@axis.com>
16413             Jean Lee  <xiaoyur347@gmail.com>
16415         * config/mips/mips.c (mips_asan_shadow_offset): New function.
16416         (TARGET_ASAN_SHADOW_OFFSET): Define.
16417         * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
16418         true for -fsanitize=address.
16420 2018-04-25  Mark Wielaard  <mark@klomp.org>
16422         * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
16423         shorter ones.
16425 2018-04-25  Jakub Jelinek  <jakub@redhat.com>
16427         * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
16428         than "alu", remove explicit "memory" and "imm_disp" attributes.
16429         (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
16431         PR middle-end/85414
16432         * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
16433         case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
16434         gen_lowpart_no_emit.
16436 2018-04-25  Sebastian Peryt  <sebastian.peryt@intel.com>
16438         PR target/85473
16439         * config/i386/i386.c (ix86_expand_builtin): Change memory
16440         operand to XI, extend p0 to Pmode.
16441         * config/i386/i386.md: Change unspec volatile and operand
16442         1 mode to XI, change operand 0 mode to P.
16444 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
16446         * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
16447         GET_MODE_MASK before any checking.
16448         (nds32_can_use_bset_p): Likewise.
16449         (nds32_can_use_btgl_p): Likewise.
16451 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
16453         * config/nds32/nds32-doubleword.md: New define_split pattern for
16454         illegal register number.
16456 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
16458         * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
16460 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
16462         * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
16464 2018-04-25  Richard Biener  <rguenther@suse.de>
16466         * lto-streamer.h (LTO_major_version): Bump to 8.
16468 2018-04-25  Jakub Jelinek  <jakub@redhat.com>
16470         * BASE-VER: Set to 9.0.0.
16472 2018-04-24  Segher Boessenkool  <segher@kernel.crashing.org>
16474         * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
16475         in __abskf2 and __powikf2.
16477 2018-04-24  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
16479         PR target/85512
16480         * config/aarch64/constraints.md (Usg, Usj): New constraints.
16481         * config/aarch64/iterators.md (cmode_simd): New mode attribute.
16482         * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
16483         Use the above on operand 2.  Reindent.
16484         (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
16486 2018-04-24  H.J. Lu  <hongjiu.lu@intel.com>
16488         PR target/85485
16489         * common/config/i386/i386-common.c (ix86_handle_option): Don't
16490         handle OPT_mcet.
16491         * config/i386/i386.opt (mcet): Removed.
16492         * doc/install.texi: Remove -mcet documentation.
16493         * doc/invoke.texi: Likewise.
16495 2018-04-24  H.J. Lu  <hongjiu.lu@intel.com>
16497         PR target/85485
16498         * doc/install.texi: Remove -mcet from bootstrap-cet.
16500 2018-04-24  Jakub Jelinek  <jakub@redhat.com>
16502         PR target/85511
16503         * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
16504         __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
16505         if TARGET_64BIT.
16507         PR target/85503
16508         * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
16509         const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
16510         containing a CONST_VECTOR.
16512 2018-04-24  Cesar Philippidis  <cesar@codesourcery.com>
16514         * doc/install.texi: Update newlib dependency for nvptx.
16516 2018-04-24  Jakub Jelinek  <jakub@redhat.com>
16518         PR target/85508
16519         * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
16520         instead of INTVAL when shifting x left.
16522 2018-04-24  Andreas Krebbel  <krebbel@linux.ibm.com>
16524         PR tree-optimization/85478
16525         * tree-vect-loop.c (vect_analyze_loop_2): Do not call
16526         vect_grouped_store_supported for single element vectors.
16528 2018-04-24  Richard Biener  <rguenther@suse.de>
16530         PR target/85491
16531         * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
16532         load cost increase to the case of non-constant step.
16534 2018-04-24  Jakub Jelinek  <jakub@redhat.com>
16536         PR target/84828
16537         * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
16538         destination if any_malformed_asm.
16540 2018-04-23  Eric Botcazou  <ebotcazou@adacore.com>
16542         PR middle-end/85496
16543         * expr.c (store_field): In the bitfield case, if the value comes from
16544         a function call and is returned in registers by means of a PARALLEL,
16545         do not change the mode of the temporary unless BLKmode and VOIDmode.
16547 2018-04-23  Andrey Belevantsev  <abel@ispras.ru>
16549         PR rtl-optimization/85423
16550         * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
16551         dependencies to debug insns when the previous insn is non-debug.
16553 2018-04-23  Claudiu Zissulescu  <claziss@synopsys.com>
16555         * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
16556         enums into a single definition.
16557         (fls): Fix predicates and printing.
16558         (seti): Likewise.
16560 2018-04-23  Claudiu Zissulescu  <claziss@synopsys.com>
16562         * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
16563         * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
16564         and short u6 immediate.
16565         (check_if_valid_sleep_operand): Remove.
16566         * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
16568 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
16570         * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
16571         flag_always_save_lp condition.
16572         * config/nds32/nds32.opt (malways-save-lp): New option.
16574 2018-04-22  Shiva Chen  <shiva0217@gmail.com>
16576         * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
16577         * config/nds32/nds32.c (nds32_use_load_post_increment): New.
16578         * config/nds32/nds32.h
16579         (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
16580         (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
16582 2018-04-22  Shiva Chen  <shiva0217@gmail.com>
16584         * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
16585         * config/nds32/nds32.c (nds32_ls_333_p): Remove.
16587 2018-04-22  Shiva Chen  <shiva0217@gmail.com>
16588             Chung-Ju Wu  <jasonwucj@gmail.com>
16590         * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
16591         Declare.
16592         * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
16593         * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
16595 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
16597         * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
16599 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
16601         * config/nds32/nds32-protos.h (nds32_data_alignment,
16602         nds32_local_alignment): Declare.
16603         * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
16604         nds32_local_alignment): New functions.
16605         (TARGET_CONSTANT_ALIGNMENT): Define.
16606         * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
16608 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
16610         * config/nds32/nds32.c
16611         (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
16612         (TARGET_MODES_TIEABLE_P): Likewise.
16614 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
16616         * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
16617         level Ofast and Og.
16619 2018-04-22  Monk Chiang  <sh.chiang04@gmail.com>
16620             Chung-Ju Wu  <jasonwucj@gmail.com>
16622         * config/nds32/constants.md (unspec_volatile_element): Add enum values
16623         for unaligned access.
16624         * config/nds32/nds32-intrinsic.c: Implementation of expanding
16625         unaligned access.
16626         * config/nds32/nds32-intrinsic.md: Likewise.
16627         * config/nds32/nds32_intrinsic.h: Likewise.
16628         * config/nds32/nds32.h (nds32_builtins): Likewise.
16629         * config/nds32/nds32.opt (munaligned-access): New option.
16630         * config/nds32/nds32.c (nds32_asm_file_start): Display
16631         flag_unaligned_access status.
16633 2018-04-20  Kito Cheng  <kito.cheng@gmail.com>
16635         * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
16636         -mno-relax is present.
16637         * config/riscv/linux.h (LINK_SPEC): Ditto.
16639 2018-04-20  Martin Sebor  <msebor@redhat.com>
16641         PR c/85365
16642         * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
16643         for null pointers.
16644         (gimple_fold_builtin_stxcpy_chk): Same.
16645         * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
16647 2018-04-20  Michael Meissner  <meissner@linux.ibm.com>
16649         PR target/85456
16650         * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
16651         __powikf2 when long double is IEEE 128-bit.
16653 2018-04-20  Kito Cheng  <kito.cheng@gmail.com>
16655         * config/riscv/riscv.c (riscv_first_stack_step): Round up min
16656         step to make sure stack always aligned.
16658 2018-04-20  Carl Love  <cel@us.ibm.com>
16660         PR target/83402
16661         * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
16662         size check for arg0.
16664 2018-04-20  Nathan Sidwell  <nathan@codesourcery.com>
16665             Tom de Vries  <tom@codesourcery.com>
16667         PR target/85445
16668         * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
16669         Emit insns for calls too.
16670         (nvptx_find_par): Always look for worker-level predecessor insn.
16671         (nvptx_propagate): Add is_call parm, return bool.  Copy frame for
16672         calls.
16673         (nvptx_vpropagate, nvptx_wpropagate): Adjust.
16674         (nvptx_process_pars): Propagate frames for calls.
16676 2018-04-20  H.J. Lu  <hongjiu.lu@intel.com>
16678         PR target/85469
16679         * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
16680         Removed.
16681         (OPTION_MASK_ISA_IBT_UNSET): Likewise.
16682         (ix86_handle_option): Don't handle OPT_mibt.
16683         * config/i386/cet.h: Check __CET__ instead of __IBT__ and
16684         __SHSTK__.
16685         * config/i386/driver-i386.c (host_detect_local_cpu): Remove
16686         has_ibt and ibt.
16687         * config/i386/i386-c.c (ix86_target_macros_internal): Don't
16688         check OPTION_MASK_ISA_IBT nor flag_cf_protection.
16689         (ix86_target_macros): Define __CET__ with flag_cf_protection
16690         for -fcf-protection.
16691         * config/i386/i386.c (isa2_opts): Remove -mibt.
16692         * config/i386/i386.h (TARGET_IBT): Removed.
16693         (TARGET_IBT_P): Likewise.
16694         (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
16695         * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
16696         * config/i386/i386.opt (mcet): Update help message.
16697         (mshstk): Likewise.
16698         (mibt): Removed.
16699         * doc/invoke.texi: Remove -mibt.  Document __CET__.  Document
16700         -mcet as an alias for -mshstk.
16702 2018-04-20  Richard Biener <rguenther@suse.de>
16704         PR middle-end/85475
16705         * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
16706         complexity by forcing a single use of the multiply operand.
16708 2018-04-20  Martin Jambor  <mjambor@suse.cz>
16710         ipa/85449
16711         * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
16712         recursion dependency to only apply to non-clones.
16714 2018-04-20  Martin Jambor  <mjambor@suse.cz>
16716         ipa/85447
16717         * ipa-cp.c (create_specialized_node): Check that clones of
16718         self-recursive edges exist during IPA-CP.
16720 2018-04-19  Toon Moene  <toon@moene.org>
16722         * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
16723         by -O3.
16725 2018-04-19  Jakub Jelinek  <jakub@redhat.com>
16727         PR tree-optimization/85467
16728         * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
16729         VECTOR_TYPE_P macro.  If type is vector type, VIEW_CONVERT_EXPR the
16730         VECTOR_CST element to type.
16732 2018-04-19  H.J. Lu  <hongjiu.lu@intel.com>
16734         PR target/85397
16735         * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
16736         * config/i386/i386.md (builtin_setjmp_setup): Removed.
16737         (builtin_longjmp): Likewise.
16738         (save_stack_nonlocal): New pattern.
16739         (restore_stack_nonlocal): Likewise.
16741 2018-04-19  H.J. Lu  <hongjiu.lu@intel.com>
16743         PR target/85404
16744         * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
16745         Replace ASM_OUTPUT_LABEL with fprintf.
16747 2018-04-19  H.J. Lu  <hongjiu.lu@intel.com>
16749         PR target/85417
16750         * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
16751         Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
16752         * config/i386/i386-c.c (ix86_target_macros_internal): Also
16753         define __IBT__ and __SHSTK__ for -fcf-protection.
16754         * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
16755         TARGET_IBT.
16756         (ix86_trampoline_init): Likewise.
16757         (x86_output_mi_thunk): Likewise.
16758         (ix86_notrack_prefixed_insn_p): Likewise.
16759         (ix86_option_override_internal): Don't disallow -fcf-protection.
16760         * config/i386/i386.md (rdssp<mode>): Also enable for
16761         -fcf-protection.
16762         (incssp<mode>): Likewise.
16763         (nop_endbr): Likewise.
16764         * config/i386/i386.opt (mcet): Change help message to built-in
16765         functions only.
16766         (mibt): Likewise.
16767         (mshstk): Likewise.
16768         * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
16769         on -fcf-protection.  Change -mcet, -mibt and -mshstk to only
16770         enable CET built-in functions.
16772 2018-04-19  Sebastian Peryt  <sebastian.peryt@intel.com>
16774         * common/config/i386/i386-common.c
16775         (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
16776         OPTION_MASK_ISA_MOVDIRI_UNSET,
16777         OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
16778         (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
16779         * config.gcc (movdirintrin.h): New header.
16780         * config/i386/cpuid.h (bit_MOVDIRI,
16781         bit_MOVDIR64B): New bits.
16782         * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
16783         and -mmvodir64b.
16784         * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
16785         (VOID, PVOID, PCVOID)): New function types.
16786         * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
16787         __builtin_ia32_directstoreu_u64,
16788         __builtin_ia32_movdir64b): New builtins.
16789         * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
16790         * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
16791         and -mmovdiri.
16792         (ix86_valid_target_attribute_inner_p): Ditto.
16793         (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
16794         and VOID_FTYPE_PUNSIGNED_UNSIGNED.
16795         (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
16796         * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
16797         TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
16798         * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
16799         (movdiri<mode>, movdir64b_<mode>): New.
16800         * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
16801         * config/i386/immintrin.h: Include movdirintrin.h.
16802         * config/i386/movdirintrin.h: New file.
16803         * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
16805 2018-04-19  Richard Biener  <rguenther@suse.de>
16807         PR middle-end/85455
16808         * cfg.c (clear_bb_flags): When loop state says we have
16809         marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
16811 2018-04-19  Richard Biener  <rguenther@suse.de>
16813         PR tree-optimization/84737
16814         * tree-vect-data-refs.c (vect_copy_ref_info): New function
16815         copying restrict info.
16816         (vect_setup_realignment): Use it.
16817         * tree-vectorizer.h (vect_copy_ref_info): Declare.
16818         * tree-vect-stmts.c (vectorizable_store): Copy ref info from
16819         the first DR to all generated stores.
16820         (vectorizable_load): Likewise for loads.
16822 2018-04-19  Jakub Jelinek  <jakub@redhat.com>
16824         PR tree-optimization/85446
16825         * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
16826         the integral and pointer types to have the same precision.
16828         * doc/install.texi: Document --disable-cet being the default and
16829         --enable-cet=auto.
16831 2018-04-18  Martin Liska  <mliska@suse.cz>
16833         * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
16834         style.
16836 2018-04-18  Martin Liska  <mliska@suse.cz>
16838         Revert
16839         2018-03-02  Eric Botcazou  <ebotcazou@adacore.com>
16841         PR ipa/83983
16842         * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
16843         arguments if they are comparable.
16845 2018-04-18  Martin Liska  <mliska@suse.cz>
16847         Revert
16848         2018-03-13  Eric Botcazou  <ebotcazou@adacore.com>
16850         PR lto/84805
16851         * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
16852         incomplete types.
16854 2018-04-18  H.J. Lu  <hongjiu.lu@intel.com>
16856         PR target/85388
16857         * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
16858         ENDBR after calling __morestack.
16860 2018-04-18  David Malcolm  <dmalcolm@redhat.com>
16862         PR jit/85384
16863         * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
16864         by using gcc_base_ver to generate a gcc_driver_version, and use
16865         it when generating GCC_DRIVER_NAME.
16866         * configure: Regenerate.
16868 2018-04-18  Jakub Jelinek  <jakub@redhat.com>
16870         PR target/81084
16871         * config.gcc: Obsolete powerpc*-*-*spe*.
16873 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
16875         PR debug/84637
16876         * dbxout.c (dbxout_int): Perform negation in unsigned int type.
16877         (stabstr_D): Change type of unum from unsigned int to
16878         unsigned HOST_WIDE_INT.  Perform negation in unsigned HOST_WIDE_INT
16879         type.
16881 2018-04-17  Jim Wilson  <jimw@sifive.com>
16883         PR 84856
16884         * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
16885         RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
16886         Set arg_pointer_offset after using pretend_args_size.
16888 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
16890         PR rtl-optimization/85431
16891         * dse.c (record_store): Ignore zero width stores.
16893         PR sanitizer/85230
16894         * asan.c (handle_builtin_stack_restore): Adjust comment.  Emit
16895         __asan_allocas_unpoison call and last_alloca_addr = new_sp before
16896         __builtin_stack_restore rather than after it.
16897         * builtins.c (expand_asan_emit_allocas_unpoison): Pass
16898         arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
16899         argument instead of virtual_dynamic_stack_rtx.
16901 2018-04-17  Kelvin Nilsen  <kelvin@gcc.gnu.org>
16903         * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
16904         New prototype.
16905         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
16906         Add note to error message to explain internal mapping of overloaded
16907         built-in function name to non-overloaded built-in function name.
16908         * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
16909         function.
16911 2018-04-17  Michael Meissner  <meissner@linux.vnet.ibm.com>
16913         PR target/85424
16914         * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
16915         where the inputs overlap with the output.
16917 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
16919         * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
16920         (=v, v) alternative and explicit "memory" attribute.
16921         (vec_extract_lo_<mode><mask_name>): Likewise.  Also add
16922         "type", "prefix", "prefix_extra", "length_immediate" and "mode"
16923         attributes.
16924         (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
16925         "sselog1" type instead of "sselog".
16926         (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
16927         "sselog".  Remove explicit "memory" attribute.
16928         (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
16929         "type", "prefix", "prefix_extra", "length_immediate" and "mode"
16930         attributes.
16931         (vec_extract_hi_v32hi): Merge all alternatives into one, use
16932         "sselog1" type instead of "sselog".  Remove explicit "memory"
16933         attribute.
16934         (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
16935         use "sselog1" type instead of "sselog".  Remove explicit "memory"
16936         attribute.
16937         (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
16938         "type", "prefix", "prefix_extra", "length_immediate" and "mode"
16939         attributes.
16940         (vec_extract_hi_v64qi): Merge all alternatives into one, use
16941         "sselog1" type instead of "sselog".  Remove explicit "memory"
16942         attribute.
16943         (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
16944         use "sselog1" type instead of "sselog".  Remove explicit "memory"
16945         attribute.
16947         PR target/85430
16948         * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
16950         PR middle-end/85414
16951         * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
16952         on a SUBREG.
16954 2018-04-17  Martin Jambor  <mjambor@suse.cz>
16956         PR ipa/85421
16957         * ipa-cp.c (create_specialized_node): Call
16958         expand_all_artificial_thunks if necessary.
16960 2018-04-17  Martin Liska  <mliska@suse.cz>
16962         PR lto/85405
16963         * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
16964         in message, remote space in between '_G' and '('.
16966 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
16968         PR target/85281
16969         * config/i386/sse.md (reduces<mode><mask_scalar_name>,
16970         avx512f_vmcmp<mode>3<round_saeonly_name>,
16971         avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
16972         avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
16973         avx512f_rndscale<mode><round_saeonly_name>,
16974         avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
16975         avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
16976         Use %<iptr>2 instead of %2 for -masm=intel.
16977         (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
16978         avx512f_vcvttss2usi<round_saeonly_name>,
16979         avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
16980         -masm=intel.
16981         (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
16982         avx512f_vcvttsd2usi<round_saeonly_name>,
16983         avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
16984         Use %q1 instead of %1 for -masm=intel.
16985         (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
16986         avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
16987         of %3 for -masm=intel.
16988         (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
16989         -masm=intel.
16990         (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
16991         -masm=intel.
16992         (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
16993         -masm=intel.
16994         (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
16995         %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
16996         %g1.
16997         (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
16998         -masm=intel.
16999         (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
17000         %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
17001         %g1 and one with %0 and %1.
17002         (avx512er_vmrcp28<mode><round_saeonly_name>,
17003         avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
17004         %1 for -masm=intel.
17005         (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
17006         avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
17007         avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
17008         of %0 and %{%4%} for -masm=intel.
17009         (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
17010         avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
17011         avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
17012         order of %0 and %{%5%}%{z%} for -masm=intel.
17014 2018-04-17  Jan Hubicka  <jh@suse.cz>
17016         PR lto/85405
17017         * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
17019 2018-04-17  Martin Liska  <mliska@suse.cz>
17021         PR ipa/85329
17022         * multiple_target.c (create_dispatcher_calls): Set apostrophes
17023         for target_clone error message.  Make default implementation
17024         clone to be a local declaration.
17025         (separate_attrs): Add new argument and check for an empty
17026         string.
17027         (expand_target_clones): Handle it.
17028         (ipa_target_clone): Make redirection just for target_clones
17029         functions.
17031 2018-04-16  Cesar Philippidis  <cesar@codesourcery.com>
17032             Tom de Vries  <tom@codesourcery.com>
17034         PR middle-end/84955
17035         * omp-expand.c (expand_oacc_for): Add dummy false branch for
17036         tiled basic blocks without omp continue statements.
17038 2018-04-16  Aaron Sawdey  <acsawdey@linux.ibm.com>
17040         PR target/83660
17041         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
17042         vec_extract expression as having side effects to make sure it gets
17043         a cleanup point.
17045 2018-04-16  H.J. Lu  <hongjiu.lu@intel.com>
17047         PR target/85403
17048         * config/i386/i386.c (get_builtin_code_for_version): Check
17049         error_mark_node.
17051 2018-04-16  Olga Makhotina  <olga.makhotina@intel.com>
17053         PR target/84331
17054         * config.gcc: Support "skylake".
17055         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
17056         PROCESSOR_SKYLAKE.
17057         * config/i386/i386.c (m_SKYLAKE): Define.
17058         (processor_target_table): Add "skylake".
17059         (ix86_option_override_internal): Add "skylake".
17060         (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
17061         PROCESSOR_CANNONLAKE.
17062         (get_builtin_code_for_version): Fix priority for
17063         PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
17064         PROCESSOR_SKYLAKE-AVX512.
17065         * config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
17066         (processor_type): Add PROCESSOR_SKYLAKE.
17068 2018-04-16  Paolo Carlini  <paolo.carlini@oracle.com>
17069             Jason Merrill  <jason@redhat.com>
17071         PR c++/85112
17072         * convert.c (convert_to_integer_1): Use direct recursion for
17073         enumeral types and types with a precision less than the number
17074         of bits in their mode.
17076 2018-04-16  Julia Koval  <julia.koval@intel.com>
17078         PR target/84413
17079         * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
17080         X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
17082 2018-04-14  Segher Boessenkool  <segher@kernel.crashing.org>
17084         PR target/85293
17085         * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
17086         * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
17087         and -mno-direct-move.
17089 2018-04-13  Paul A. Clarke  <pc@us.ibm.com>
17091         PR target/83402
17092         * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
17093         Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
17094         Ensure negative shifts result in {0}.
17096 2018-04-13  Vladimir Makarov  <vmakarov@redhat.com>
17098         PR rtl-optimization/79916
17099         * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
17100         regs (if any) to define how to gnerate SD moves when LRA is in
17101         progress.
17103 2018-04-13  Jakub Jelinek  <jakub@redhat.com>
17105         PR rtl-optimization/85393
17106         * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
17107         * except.c (expand_dw2_landing_pad_for_region): Make static.
17108         * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
17109         a label and unconditional jump to old_bb, rather than
17110         expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
17111         basic block.
17113         PR rtl-optimization/85376
17114         * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
17115         zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
17116         instead of a specific value.
17118 2018-04-13  Jan Hubicka  <hubicka@ucw.cz>
17119             Bin Cheng  <bin.cheng@arm.com>
17121         PR tree-optimization/82965
17122         PR tree-optimization/83991
17123         * cfgloopanal.c (expected_loop_iterations_unbounded): Add
17124         by_profile_only parameter.
17125         * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
17126         information if the loop was predicted to iterate too many times.
17127         * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
17129 2018-04-13  Jan Hubicka  <hubicka@ucw.cz>
17131         PR lto/71991
17132         * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
17133         always inline.
17135 2018-04-13  Martin Liska  <mliska@suse.cz>
17136             Jakub Jelinek  <jakub@redhat.com>
17138         PR middle-end/81657
17139         * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
17140         * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
17141         * builtins.c (expand_builtin_memory_copy_args): Use
17142         BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
17143         handle dest_addr == pc_rtx.
17145 2018-04-12  Segher Boessenkool  <segher@kernel.crashing.org>
17147         PR target/85291
17148         * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
17149         asked to not generate direct moves.
17150         (fix_trunc<mode>si2_stfiwx): Similar.
17151         (fix_trunc<mode>si2_internal): Similar.
17153 2018-04-12  Jakub Jelinek  <jakub@redhat.com>
17155         PR debug/83157
17156         * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
17157         * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
17158         lookup if dest in some wider mode is known to be const0_rtx and
17159         if so, record permanent equivalence for it to be ZERO_EXTEND of
17160         the narrower mode destination.
17162 2018-04-12  Cesar Philippidis  <cesar@codesourcery.com>
17164         * lto-streamer-out.c (output_function): Revert 259346.
17165         * omp-expand.c (expand_oacc_for): Likewise.
17167 2018-04-12  Alexander Monakov  <amonakov@ispras.ru>
17169         PR rtl-optimization/85354
17170         * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
17171         * sel-sched.c (sel_global_init): ... here.
17173 2018-04-12  Eric Botcazou  <ebotcazou@adacore.com>
17175         PR target/85238
17176         * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
17177         * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
17178         mode for PE-COFF targets.
17179         * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
17180         (i386_pe_asm_lto_end): Likewise.
17181         * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
17182         (TARGET_ASM_LTO_END): Likewise.
17183         * config/i386/winnt.c (saved_debug_info_level): New static variable.
17184         (i386_pe_asm_lto_start): New function.
17185         (i386_pe_asm_lto_end): Likewise.
17187 2018-04-12  Cesar Philippidis  <cesar@codesourcery.com>
17188             Richard Biener  <rguenther@suse.de>
17190         PR middle-end/84955
17191         * lto-streamer-out.c (output_function): Fix CFG loop state before
17192         streaming out.
17193         * omp-expand.c (expand_oacc_for): Handle calls to internal
17194         functions like regular functions.
17196 2018-04-12  Richard Biener  <rguenther@suse.de>
17198         PR lto/85371
17199         * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
17200         for the early LTO debug to properly generate references to it
17201         during DIE emission.  Do not re-use that for the skeleton for
17202         split-dwarf.
17203         (dwarf2out_early_finish): Likewise.
17205 2018-04-12  Jakub Jelinek  <jakub@redhat.com>
17207         PR target/85328
17208         * config/i386/sse.md
17209         (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
17210         <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
17211         vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
17212         vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
17213         and output is a reg, avoid creating invalid lowpart subreg, but
17214         instead split into a 512-bit move.  Don't split if not AVX512VL,
17215         input is xmm16+ reg and output is a mem.
17216         (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
17217         vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
17218         xmm16+ reg and output is a mem.
17220 2018-04-12  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
17222         * config/s390/s390.c (s390_output_indirect_thunk_function): Check
17223         also for flag_dwarf2_cfi_asm.
17225 2018-04-12  Jakub Jelinek  <jakub@redhat.com>
17227         PR rtl-optimization/85342
17228         * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
17229         a bool scalar var inside of the loop instead.  Don't try to update
17230         recog_data.operand after failed apply_change_group.
17232 2018-04-12  Tom de Vries  <tom@codesourcery.com>
17234         PR target/85296
17235         * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
17236         (nvptx_assemble_decl_begin): Add undefined param.  Declare undefined
17237         array with flexible array member as array without given dimension.
17238         (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
17239         argument for undefined param to true.
17241 2018-04-11  Aaron Sawdey  <acsawdey@linux.ibm.com>
17243         PR target/85321
17244         * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
17245         -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
17246         from PowerPC section.
17247         * config/rs6000/sysv4.opt (mcall-): Improve help text.
17248         * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
17249         help text that is too long.
17250         * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
17251         help text that is too long.
17252         * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
17253         help text that is too long.
17255 2018-04-11  Uros Bizjak  <ubizjak@gmail.com>
17257         * config/alpha/alpha.md (stack_probe_internal): Rename
17258         from "probe_stack".  Update all callers.
17260 2018-04-11  Alexander Monakov  <amonakov@ispras.ru>
17262         PR rtl-optimization/84566
17263         * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
17264         sched_macro_fuse_insns.
17266 2018-04-11  Alexander Monakov  <amonakov@ispras.ru>
17268         PR target/84301
17269         * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
17270         (compute_block_dependences): ... from here.
17272 2018-04-11  Jakub Jelinek  <jakub@redhat.com>
17274         PR tree-optimization/85331
17275         * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
17276         from int to HOST_WIDE_INT.
17278 2018-04-11  Martin Jambor  <mjambor@suse.cz>
17280         PR ipa/84149
17281         * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
17282         (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
17283         not the same as the source val.
17284         (cgraph_edge_brings_value_p): New parameter.
17285         (gather_edges_for_value): Pass destination value to
17286         cgraph_edge_brings_value_p.
17287         (perhaps_add_new_callers): Likewise.
17288         (get_info_about_necessary_edges): Likewise and exclude values brought
17289         only by self-recursive edges.
17290         (create_specialized_node): Redirect only clones of self-calling edges.
17291         (+self_recursive_pass_through_p): New function.
17292         (find_more_scalar_values_for_callers_subset): Use it.
17293         (find_aggregate_values_for_callers_subset): Likewise.
17294         (known_aggs_to_agg_replacement_list): Removed.
17295         (decide_whether_version_node): Re-calculate known constants for all
17296         remaining context clones.
17298 2018-04-11  Richard Biener  <rguenther@suse.de>
17300         PR lto/85339
17301         * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
17302         from early DWARF output.
17303         (dwarf2out_early_finish): Output line info unconditionally into
17304         early DWARF and add reference to it.
17306 2018-04-11  Jakub Jelinek  <jakub@redhat.com>
17308         PR target/85281
17309         * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
17310         (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
17311         other than V2DFmode using iptr mode attribute.
17312         (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
17314 2018-04-11  Alexander Monakov  <amonakov@ispras.ru>
17316         PR rtl-optimization/84659
17317         * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
17319 2018-04-11  Jakub Jelinek  <jakub@redhat.com>
17321         PR debug/85302
17322         * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
17323         SIZEP is NULL.
17324         (output_loc_list): Pass address of a dummy size variable even in the
17325         locview handling loop.
17326         (index_location_lists): Add comment on why skip_loc_list_entry can't
17327         call size_of_locs.
17329 2018-04-11  Thomas Preud'homme  <thomas.preudhomme@arm.com>
17331         PR target/85261
17332         * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
17333         into register.
17335 2018-04-10  Aaron Sawdey  <acsawdey@linux.ibm.com>
17337         PR target/85321
17338         * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
17339         -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
17340         and -mstring-compare-inline-limit.
17342 2018-04-10  Segher Boessenkool  <segher@kernel.crashing.org>
17344         PR target/85287
17345         * config/rs6000/rs6000.md (allocate_stack): Put the residual size
17346         for stack clash protection in a register whenever we need it to be in
17347         a register.
17349 2018-04-10  Segher Boessenkool  <segher@kernel.crashing.org>
17351         * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
17352         Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
17354 2018-04-10  Segher Boessenkool  <segher@kernel.crashing.org>
17356         PR target/85321
17357         * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
17358         the help text.
17359         (mlong-double-): Ditto.
17360         * config/rs6000/sysv4.opt (msdata=): Ditto.
17361         (mtls-size=): Ditto.
17363 2018-04-10  Kelvin Nilsen  <kelvin@gcc.gnu.org>
17365         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
17366         erroneous entries for
17367         "vector int vec_ldl (int, long int *)", and
17368         "vector unsigned int vec_ldl (int, unsigned long int *)".
17369         Add comments and entries for
17370         "vector bool char vec_ldl (int, bool char *)",
17371         "vector bool short vec_ldl (int, bool short *)",
17372         "vector bool int vec_ldl (int, bool int *)",
17373         "vector bool long long vec_ldl (int, bool long long *)",
17374         "vector pixel vec_ldl (int, pixel *)",
17375         "vector long long vec_ldl (int, long long *)",
17376         "vector unsigned long long vec_ldl (int, unsigned long long *)".
17377         * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
17378         type tree bool_long_long_type_node and correct definition of
17379         bool_V2DI_type_node to make reference to this new type tree.
17380         (rs6000_mangle_type): Replace erroneous reference to
17381         bool_long_type_node with bool_long_long_type_node.
17382         * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
17383         comments to emphasize sign distinctions for char and int types and
17384         replace RS6000_BTI_bool_long constant with
17385         RS6000_BTI_bool_long_long constant.  Also add comment to restrict
17386         use of RS6000_BTI_pixel.
17387         (bool_long_type_node): Remove this macro definition.
17388         (bool_long_long_type_node): New macro definition
17390 2018-04-10  Jakub Jelinek  <jakub@redhat.com>
17392         PR rtl-optimization/85300
17393         * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
17394         into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
17395         simplify_unary_operation fails.
17397 2018-04-10  Martin Liska  <mliska@suse.cz>
17399         * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
17400         cgraph_edge and ipa_ref.
17402 2018-04-10  Jakub Jelinek  <jakub@redhat.com>
17404         PR target/85177
17405         PR target/85255
17406         * config/i386/sse.md
17407         (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
17408         computation of the VEC_MERGE selector from mask.
17409         (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
17410         Fix decoding of the VEC_MERGE selector into mask.
17412 2018-04-10  Richard Sandiford  <richard.sandiford@linaro.org>
17414         PR tree-optimization/85286
17415         * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
17417 2018-04-10  Richard Sandiford  <richard.sandiford@linaro.org>
17419         * final.c (final_1): Set insn_last_address as well as
17420         insn_current_address.
17422 2018-04-10  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
17424         PR target/85173
17425         * explow.c (emit_stack_probe): Call validize_mem on memory location
17426         before passing it to gen_probe_stack.  Create address operand and
17427         legitimize it for the probe_stack_address case.
17429 2018-04-09  Jan Hubicka  <jh@suse.cz>
17431         PR lto/85078
17432         * ipa-devirt.c (rebuild_type_inheritance-hash): New.
17433         * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
17434         * tree.c (free_lang_data_in_type): Fix handling of binfos;
17435         walk basetypes.
17436         (free_lang_data): Rebuild type inheritance graph.
17438 2018-04-09  Martin Sebor  <msebor@redhat.com>
17440         * invoke.texi (-finline-small-functions): Mention other optimization
17441         options.
17442         (-findirect-inlining, -fpartial-inlining): Same.
17443         (-finline-functions-called-once): Same.
17444         (-freorder-blocks-and-partition): Same.
17446 2018-04-09  Jan Hubicka  <jh@suse.cz>
17448         PR rtl/84058
17449         * cfgcleanup.c (try_forward_edges): Do not give up on crossing
17450         jumps; choose last target that matches the criteria (i.e.
17451         no partition changes for non-crossing jumps).
17452         * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
17453         support for redirecting crossing jumps to non-crossing.
17455 2018-04-09  Alexey Brodkin  <abrodkin@synopsys.com>
17457         * config/arc/arc.c (arc_expand_prologue): Set stack usage info
17458         also for naked functions.
17460 2018-04-09  Claudiu Zissulescu  <claziss@synopsys.com>
17462         * config/arc/arc.md (add_shift): New pattern.
17463         (add_shift2): Likewise.
17464         (sub_shift): Likewise.
17465         (sub_shift_cmp0_noout): Likewise.
17466         (compare_si_ashiftsi): Likewise.
17467         (xbfu_cmp0_noout): New combine pattern.
17468         (xbfu_cmp0"): Likewise.
17469         (movsi_set_cc_insn): Place the predicable variant first.
17470         (commutative_binary_cmp0_noout): Remove clobber.
17471         (commutative_binary_cmp0): New pattern.
17472         (noncommutative_binary_cmp0): Likewise.
17473         (noncommutative_binary_cmp0_noout): Likewise.
17474         (noncommutative_binary_comparison_result_used): Removed.
17475         (rsub_cmp0): New pattern.
17476         (rsub_cmp0_noout): Likewise.
17477         (extzvsi): Changed, keep only meaningful variants.
17478         (SQH, SEZ): New iterators.
17479         (SQH_postfix): New mode attribute.
17480         (SEZ_prefix): New code attribute.
17481         (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
17482         (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
17483         * config/arc/predicates.md (cc_set_register): Use CC_REG instead
17484         of numerical value.
17485         (noncommutative_operator): Check the availability of barrel
17486         shifter option.
17488 2018-04-09  Richard Biener  <rguenther@suse.de>
17490         PR tree-optimization/85284
17491         * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
17492         Only use the niter constraining form of simple_iv when the exit
17493         is always executed.
17495 2018-04-09  Tom de Vries  <tom@codesourcery.com>
17497         PR target/84041
17498         * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
17499         (define_expand "*memory_barrier"): New define_expand.
17500         (define_insn "memory_barrier"): New insn.
17502 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
17504         PR rtl-optimization/80463
17505         PR rtl-optimization/83972
17506         PR rtl-optimization/83480
17508         * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
17509         correct producer for the insn.
17510         (tidy_control_flow): Fixup seqnos in case of debug insns.
17512 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
17514         PR rtl-optimization/83913
17516         * sel-sched-ir.c (merge_expr_data): Choose the middle between two
17517         different sched-times when merging exprs.
17519 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
17521         PR rtl-optimization/83962
17523         * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
17524         tidy_fallthru_edge and tidy_control_flow.
17526 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
17528         PR rtl-optimization/83530
17530         * sel-sched.c (force_next_insn): New global variable.
17531         (remove_insn_for_debug): When force_next_insn is true, also leave only
17532         next insn in the ready list.
17533         (sel_sched_region): When the region wasn't scheduled, make another pass
17534         over it with force_next_insn set to 1.
17536 2018-04-08  Monk Chiang  <sh.chiang04@gmail.com>
17538         * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
17539         into tm_file.
17540         * config/nds32/constants.md (unspec_volatile_element): Add enum values
17541         for interrupt control.
17542         * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
17543         functions for interrupt control.
17544         * config/nds32/nds32-intrinsic.md: Likewise.
17545         * config/nds32/nds32_intrinsic.h: Likewise.
17546         * config/nds32/nds32.h (nds32_builtins): Likewise.
17548 2018-04-08  Chung-Ju Wu  <jasonwucj@gmail.com>
17550         * config/nds32/nds32.c (nds32_init_machine_status,
17551         nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
17552         strict_aligned_p field.
17553         (nds32_expand_to_rtl_hook): New function.
17554         (TARGET_EXPAND_TO_RTL_HOOK): Define.
17555         * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
17557 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
17558             Chung-Ju Wu  <jasonwucj@gmail.com>
17560         * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
17561         * config/nds32/nds32-n7.md: New file.
17562         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
17563         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
17564         pipeline.
17565         * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
17566         * config/nds32/nds32.md (pipeline_model): Add n7.
17567         * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
17568         * config/nds32/pipelines.md: Include n7 settings.
17570 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
17571             Chung-Ju Wu  <jasonwucj@gmail.com>
17573         * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
17574         * config/nds32/nds32-e8.md: New file.
17575         * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
17576         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
17577         pipeline.
17578         * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
17579         * config/nds32/nds32.md (pipeline_model): Add e8.
17580         * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
17581         * config/nds32/pipelines.md: Include e8 settings.
17583 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
17584             Chung-Ju Wu  <jasonwucj@gmail.com>
17586         * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
17587         * config/nds32/nds32-n8.md: New file.
17588         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
17589         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
17590         pipeline.
17591         * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
17592         * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
17593         * config/nds32/nds32.md (pipeline_model): Add n8.
17594         * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
17595         * config/nds32/pipelines.md: Include n8 settings.
17597 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
17598             Chung-Ju Wu  <jasonwucj@gmail.com>
17600         * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
17601         * config/nds32/nds32-n9-2r1w.md: New file.
17602         * config/nds32/nds32-n9-3r2w.md: New file.
17603         * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
17604         nds32_register_ports): New or modify for cpu n9.
17605         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
17606         pipeline.
17607         * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
17608         * config/nds32/nds32-utils.c: New file.
17609         * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
17610         TARGET_MUL_SLOW): Define.
17611         * config/nds32/nds32.md (pipeline_model): New attribute.
17612         * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
17613         New options that support cpu n9.
17614         * config/nds32/pipelines.md: Include n9 settings.
17615         * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
17617 2018-04-08  Chung-Ju Wu  <jasonwucj@gmail.com>
17619         * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
17620         information if necessary.
17621         (output_cond_branch_compare_zero): Likewise.
17622         * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
17623         (nds32_target_alignment): Refine for alignment.
17624         * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
17625         (FUNCTION_BOUNDARY): Modify.
17626         * config/nds32/nds32.md (call_internal, call_value_internal): Consider
17627         align case.
17628         * config/nds32/nds32.opt (malways-align, malign-functions): New.
17630 2018-04-08  Monk Chiang  <sh.chiang04@gmail.com>
17632         * config/nds32/constants.md (unspec_volatile_element): Add values for
17633         TLB operation and data prefetch.
17634         * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
17635         functions for TLB operation and data prefetch.
17636         * config/nds32/nds32-intrinsic.md: Likewise.
17637         * config/nds32/nds32_intrinsic.h: Likewise.
17638         * config/nds32/nds32.c (nds32_dpref_names): Likewise.
17639         (nds32_print_operand): Likewise.
17640         * config/nds32/nds32.h (nds32_builtins): Likewise.
17642 2018-04-07  Thomas Koenig  <tkoenig@gcc.gnu.org>
17643         Andrew Pinski <pinsika@gcc.gnu.org>
17645         PR middle-end/82976
17646         * match.pd: Use constant_boolean_node of correct type instead of
17647         boolean_true_node or boolean_false_node for simplifying
17648         pointer comparisons to zero.
17650 2018-04-07  Jakub Jelinek  <jakub@redhat.com>
17652         PR tree-optimization/80021
17653         * tree.c (verify_type_variant): Make error call in verify_variant_match
17654         translatable and remove final full stop.
17656 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
17658         * config/nds32/constants.md (unspec_volatile_element): Add
17659         UNSPEC_VOLATILE_EH_RETURN.
17660         * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
17661         nds32_output_stack_pop): Support dwarf exception handling process.
17662         * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
17663         * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
17664         exception handling process.
17665         (nds32_compute_stack_frame): Likewise.
17666         (nds32_return_addr_rtx): Likewise.
17667         (nds32_initial_elimination_offset): Likewise.
17668         (nds32_expand_prologue): Likewise.
17669         (nds32_expand_epilogue): Likewise.
17670         (nds32_dynamic_chain_address): New function.
17671         * config/nds32/nds32.h (machine_function): Add fields for dwarf
17672         exception handling.
17673         (DYNAMIC_CHAIN_ADDRESS): Define.
17674         (EH_RETURN_DATA_REGNO): Define.
17675         (EH_RETURN_STACKADJ_RTX): Define.
17676         * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
17677         patterns for dwarf exception handling.
17679 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
17681         * config/nds32/nds32.h: Clean up obsolete macros.
17683 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
17685         * config/nds32/constants.md (unspec_element, unspec_volatile_element):
17686         Add enum values for particular instructions.
17687         * config/nds32/nds32-intrinsic.c: Implementation of expanding
17688         particular intrinsic functions.
17689         * config/nds32/nds32-intrinsic.md: Likewise.
17690         * config/nds32/nds32_intrinsic.h: Likewise.
17691         * config/nds32/nds32.h (nds32_builtins): Likewise.
17692         * config/nds32/nds32.md (type): Add pbsad and pbsada.
17693         (btst, ave): New patterns for particular instructions.
17695 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
17697         * config/nds32/constants.md (unspec_element, unspec_volatile_element):
17698         Add enum values for atomic load/store and memory sync.
17699         * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
17700         and memory sync.
17701         * config/nds32/nds32-intrinsic.md: Likewise.
17702         * config/nds32/nds32_intrinsic.h: Likewise.
17703         * config/nds32/nds32.h (nds32_builtins): Likewise.
17705 2018-04-07  Jakub Jelinek  <jakub@redhat.com>
17707         PR tree-optimization/85257
17708         * fold-const.c (native_encode_vector): If not all elts could fit
17709         and off is -1, return 0 rather than offset.
17710         * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
17711         (offseti - offset2) / BITS_PER_UNIT as 4th argument to
17712         native_encode_expr.  Verify len * BITS_PER_UNIT >= maxsizei.  Don't
17713         adjust buffer in native_interpret_expr call.
17715 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
17717         * config/nds32/constants.md (unspec_volatile_element): Add cache
17718         control enum values.
17719         * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
17720         * config/nds32/nds32-intrinsic.md: Add cache control patterns.
17721         * config/nds32/nds32.c (nds32_cctl_names): New.
17722         (nds32_print_operand): Handle cache control register names.
17723         * config/nds32/nds32.h (nds32_builtins): New enum values.
17724         * config/nds32/nds32_intrinsic.h: Add cache control enum types and
17725         macros.
17726         * config/nds32/nds32.md (type): Add mmu.
17727         * config/nds32/pipelines.md (simple_insn): Add mmu.
17729 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
17731         * config/nds32/nds32.md (type): Remove call.
17732         * config/nds32/pipelines.md (simple_insn): Likewise.
17734 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
17736         * config/nds32/constants.md (unspec_volatile_element): Add
17737         UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
17738         UNSPEC_VOLATILE_FMFCFG.
17739         * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
17740         description for fmfcfg and fmfcsr.
17741         (bdesc_1arg): Add fmtcsr.
17742         (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
17743         (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
17744         * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
17745         unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
17746         unspec_fmfcfg): New patterns.
17747         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
17748         NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
17749         NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
17750         * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
17751         __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
17752         __nds32__fmfcfg): Define.
17754 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
17756         * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
17757         intrinsic register names.
17758         * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
17759         intrinsic register enum values and macros.
17761 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
17763         * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
17764         for load/store addressing form.
17765         (nds32_print_operand_address): Likewise.
17767 2018-04-06  Eric Botcazou  <ebotcazou@adacore.com>
17769         PR target/85196
17770         * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
17771         based on LABEL_REF.  Remove useless assertion.
17772         (pic_address_needs_scratch): Fix formatting.
17773         (sparc_legitimize_pic_address): Minor tweaks.
17774         (sparc_delegitimize_address): Adjust assertion accordingly.
17775         * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
17776         into symbolic_operand.
17777         (movsi_high_pic_label_ref): Likewise.
17778         (movsi_lo_sum_pic_label_ref): Likewise.
17779         (movdi_pic_label_ref): Likewise.
17780         (movdi_high_pic_label_ref): Likewise.
17781         (movdi_lo_sum_pic_label_ref): Likewise.
17783 2018-04-06  Amaan Cheval  <amaan.cheval@gmail.com>
17785         * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
17786         custom LIB_SPEC setup.
17788 2018-04-06  Ruslan Bukin  <br@bsdpad.com>
17789             Kito Cheng  <kito.cheng@gmail.com>
17791         * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
17792         * config/riscv/freebsd.h: New.
17794 2018-04-06  Chung-Ju Wu  <jasonwucj@gmail.com>
17796         * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
17797         * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
17798         file.
17800 2018-04-06  Chung-Ju Wu  <jasonwucj@gmail.com>
17801             Kito Cheng  <kito.cheng@gmail.com>
17803         * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
17804         nds32_output_call, nds32_symbol_binds_local_p): New functions.
17805         * config/nds32/nds32-protos.h (nds32_output_call,
17806         nds32_output_return): Declare.
17807         * config/nds32/nds32.md: Refine all the call and return patterns.
17809 2018-04-06  Jakub Jelinek  <jakub@redhat.com>
17811         PR debug/85252
17812         * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
17813         build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
17815         PR rtl-optimization/84872
17816         * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
17817         nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
17818         EDGE_CROSSING edge.
17820 2018-04-06  Tamar Christina  <tamar.christina@arm.com>
17822         * expr.c (copy_blkmode_to_reg): Revert 254862.
17823         * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
17825 2018-04-06  Richard Biener  <rguenther@suse.de>
17827         PR middle-end/85244
17828         * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
17829         after seeing a component reference with an adjacent field.  Treat
17830         refs to arrays at struct end of external decls similar to
17831         refs to unconstrained commons.
17833 2018-04-06  Jakub Jelinek  <jakub@redhat.com>
17835         PR sanitizer/85213
17836         * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
17837         look through SAVE_EXPRs with non-side-effects argument.  Adjust
17838         recursive calls.
17839         (fold_comparison): Adjust twoval_comparison_p caller, don't handle
17840         save_p here.
17842 2018-04-06  Richard Biener  <rguenther@suse.de>
17844         PR middle-end/85180
17845         * alias.c (find_base_term): New wrapper around find_base_term
17846         unwinding CSELIB_VAL_PTR changes.
17847         (find_base_term): Do not restore CSELIB_VAL_PTR during the
17848         recursion.
17850 2018-04-06  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
17852         * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
17853         instructions.
17854         * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
17855         constant definitions.
17856         ("nop"): lr 0,0 -> nopr r0
17857         ("nop_lr0", "nop_lr1"): New insn definitions.
17859 2018-04-06  Chung-Ju Wu  <jasonwucj@gmail.com>
17861         * config/nds32/nds32.md (*stack_push, *stack_pop): Use
17862         NDS32_V3PUSH_AVAILABLE_P macro.
17864 2018-04-06  Monk Chiang  <sh.chiang04@gmail.com>
17865             Chung-Ju Wu  <jasonwucj@gmail.com>
17867         * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
17868         (nds32*-*-*): Add float and fpu_config into supported_defaults.
17869         * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
17870         Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
17871         * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
17872         UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
17873         * config/nds32/constraints.md: New constraints and checking for hard
17874         float configuration.
17875         * config/nds32/iterators.md: New mode iterator and attribute for hard
17876         float configuration.
17877         * config/nds32/nds32-doubleword.md: Use hard float alternatives and
17878         patterns.
17879         * config/nds32/nds32-fpu.md: New file.
17880         * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
17881         deal with hard float code generation.
17882         * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
17883         ARCH_V3S.
17884         (abi_type, float_reg_number): New enum type.
17885         * config/nds32/nds32-predicates.c: New predicates for hard float.
17886         * config/nds32/nds32-protos.h: Declare functions for hard float.
17887         * config/nds32/nds32.c: Implementation for hard float configuration.
17888         * config/nds32/nds32.h: Definitions for hard float configuration.
17889         * config/nds32/nds32.md: Include hard float machine description and
17890         modify patterns for hard float configuration.
17891         * config/nds32/nds32.opt: New options for hard float configuration.
17892         * config/nds32/predicates.md: New predicates for hard float
17893         configuration.
17895 2018-04-06  Kuan-Lin Chen  <kuanlinchentw@gmail.com>
17897         * common/config/nds32/nds32-common.c
17898         (nds32_option_optimization_table): Enable -mreleax-hint by default.
17900 2018-04-05  Jakub Jelinek  <jakub@redhat.com>
17902         PR middle-end/85195
17903         * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
17904         CONSTRUCTOR_ELT (ctor, ...)->value.
17906 2018-04-05  Uros Bizjak  <ubizjak@gmail.com>
17908         PR target/85193
17909         * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
17911 2018-04-05  Tom de Vries  <tom@codesourcery.com>
17913         PR target/85204
17914         * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
17915         cond jump.
17917 2018-04-05  Shiva Chen  <shiva0217@gmail.com>
17918             Kito Cheng  <kito.cheng@gmail.com>
17920         * config/nds32/constraints.md (U33): Fine-tune checking condition.
17921         * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
17922         * config/nds32/nds32.h (nds32_16bit_address_type): Add
17923         ADDRESS_POST_MODIFY_LO_REG_IMM3U.
17925 2018-04-05  Shiva Chen  <shiva0217@gmail.com>
17926             Kito Cheng  <kito.cheng@gmail.com>
17928         * config/nds32/constraints.md (Ufe): New memory constraint.
17929         * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
17930         nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
17931         * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
17932         operands.
17933         * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
17934         * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
17936 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
17938         * config/nds32/nds32.md: Use optimize_size in the condition for
17939         alu-shift instructions.
17941 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
17943         * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
17945 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
17947         * config/nds32/nds32.md (negsi2): Refine pattern.
17949 2018-04-05  Kito Cheng  <kito.cheng@gmail.com>
17950             Chung-Ju Wu  <jasonwucj@gmail.com>
17952         * config/nds32/iterators.md (shift_rotate): New code iterator.
17953         (shift): New code attribute.
17954         * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
17955         * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
17956         * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
17957         * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
17958         bit-wise operations.
17959         (andsi3, *andsi3): Ditto.
17960         (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
17961         (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
17962         (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
17963         * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
17964         nds32_ior_operand, nds32_xor_operand): New predicates.
17966 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
17968         * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
17969         (addsi3, subsi3): ... this.
17971 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
17973         * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
17975 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
17977         * config/nds32/nds32.md: Adjust indention.
17979 2018-04-05  Kito Cheng  <kito.cheng@gmail.com>
17981         * config/nds32/nds32.md (feature): New attribute.
17983 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
17985         * config/nds32/nds32.md (subtype): New attribute.
17987 2018-04-04  Thomas Preud'homme  <thomas.preudhomme@arm.com>
17989         PR target/85203
17990         * config/arm/arm-builtins.c (arm_expand_builtin): Change
17991         expansion to perform a bitwise AND of the argument followed by a
17992         boolean negation of the result.
17994 2018-04-04  Peter Bergner  <bergner@vnet.ibm.com>
17996         PR rtl-optimization/84878
17997         * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
17998         the basic block.  Assert the use reference is not artificial and that
17999         it has an associated insn.
18001 2018-04-04  Michael Matz  <matz@suse.de>
18003         * builtins.c (compute_objsize): Pass correct operand
18004         to array_at_struct_end_p.
18006 2018-04-04  Richard Biener  <rguenther@suse.de>
18008         PR lto/85176
18009         * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
18010         from contexts for DINFO_LEVEL_TERSE and below.
18012 2018-04-04  Kito Cheng  <kito.cheng@gmail.com>
18014         * config/nds32/nds32-doubleword.md (move_<mode>): Require
18015         resiter_operand condition.
18016         * config/nds32/nds32.md (*move<mode>): Ditto.
18018 2018-04-04  Kito Cheng  <kito.cheng@gmail.com>
18019             Monk Chiang  <sh.chiang04@gmail.com>
18021         * config/nds32/nds32.md (movmisalign<mode>): New pattern.
18023 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
18025         * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
18027 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
18028             Kito Cheng  <kito.cheng@gmail.com>
18030         * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
18031         nds32_cond_code_str, output_cond_branch,
18032         output_cond_branch_compare_zero, nds32_expand_cbranch,
18033         nds32_expand_cstore, nds32_expand_movcc,
18034         nds32_output_cbranchsi4_equality_zero,
18035         nds32_output_cbranchsi4_equality_reg,
18036         nds32_output_cbranchsi4_equality_reg_or_const_int,
18037         nds32_output_cbranchsi4_greater_less_zero: New functions.
18038         * config/nds32/nds32-protos.h (nds32_expand_cbranch,
18039         nds32_expand_cstore, nds32_expand_movcc,
18040         nds32_output_cbranchsi4_equality_zero,
18041         nds32_output_cbranchsi4_equality_reg,
18042         nds32_output_cbranchsi4_equality_reg_or_const_int,
18043         nds32_output_cbranchsi4_greater_less_zero): Declare.
18044         * config/nds32/predicates.md (nds32_movecc_comparison_operator,
18045         nds32_rimm11s_operand): New predicates.
18046         * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
18047         * config/nds32/nds32.md: Rewrite all the branch and conditional move
18048         patterns.
18050 2018-04-04  Kito Cheng  <kito.cheng@gmail.com>
18052         * config/nds32/nds32-doubleword.md: Refine all the instruction type.
18053         * config/nds32/nds32.md: Ditto.
18054         * config/nds32/pipelines.md: Ditto.
18056 2018-04-04  Richard Biener  <rguenther@suse.de>
18058         PR tree-optimization/85168
18059         * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
18060         propagating abnormals.
18062 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
18064         * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
18066 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
18067             Kito Cheng  <kito.cheng@gmail.com>
18069         * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
18070         * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
18071         * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
18072         (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
18073         * config/nds32/nds32.md (sibcall_internal): New.
18074         (sibcall_register): Remove.
18075         (sibcall_immediate): Remove.
18076         (sibcall_value_internal): New.
18077         (sibcall_value_register): Remove.
18078         (sibcall_value_immediate): Remove.
18079         * config/nds32/predicates.md (nds32_general_register_operand): New.
18080         (nds32_call_address_operand): New.
18082 2018-04-03  Jakub Jelinek  <jakub@redhat.com>
18084         PR rtl-optimization/85167
18085         * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
18086         bb_defs if *split_p, instead preinitialize it to NULL.
18088         PR tree-optimization/85156
18089         * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
18090         evaluating the argument multiple times.
18092 2018-04-03  Bill Schmidt  <wschmidt@linux.ibm.com>
18094         * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
18095         than vector.
18096         (_mm_cvtpd_ps): Likewise.
18097         (_mm_cvttpd_epi32): Likewise.
18098         * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
18099         * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
18100         vector, pixel, and bool following altivec.h include.
18102 2018-04-03  Martin Sebor  <msebor@redhat.com>
18104         * doc/extend.texi (Common Function Attributes): Clarify.
18105         (const attribute): Likewise.
18106         (pure attribute): Likewise.
18108 2018-04-03  Jakub Jelinek  <jakub@redhat.com>
18110         PR target/85169
18111         * config/i386/i386.c (ix86_expand_vector_set): Use
18112         HOST_WIDE_INT_1U << elt instead of 1 << elt.  Formatting fix.
18114 2018-04-03  Uros Bizjak  <ubizjak@gmail.com>
18116         * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
18117         instructions when changing rounding bits to preserve precision bits
18118         in the x87 control word.
18120 2018-04-03  Martin Liska  <mliska@suse.cz>
18122         PR tree-optimization/82491
18123         * rtl.h (strip_offset_and_add): Replace += suboffset with
18124         poly_uint64 () + suboffset.
18126 2018-03-29  Martin Liska  <mliska@suse.cz>
18127             Martin Jambor  <mjambor@suse.cz>
18129         PR ipa/84947
18130         * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
18131         param_type is not an integral or pointer type.
18133 2018-04-03  Richard Biener  <rguenther@suse.de>
18135         * sese.h (recompute_all_dominators): Remove.
18137 2018-04-02  Martin Sebor  <msebor@redhat.com>
18139         * doc/invoke.texi (-Wrestrict): Fix typos.
18141 2018-04-02  Jim Wilson  <jimw@sifive.com>
18143         * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
18144         * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
18145         (<optab>di3, <optab>si3_extend): Likewise.
18146         (<optab>si3_mask, <optab>si3_mask_1): New.
18147         (<optab>di3_mask, <optab>di3_mask_1): New.
18148         (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
18149         (lshrsi3_zero_extend_1): Use VOIDmode shift count.
18150         * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
18152 2018-04-02  Gerald Pfeifer  <gerald@pfeifer.com>
18154         * doc/cpp.texi (Variadic Macros): Fix line continuation in an
18155         example.
18157 2018-04-02  Chung-Ju Wu  <jasonwucj@gmail.com>
18159         * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
18160         (nds32_canonicalize_comparison): New function.
18162 2018-04-01  Chung-Ju Wu  <jasonwucj@gmail.com>
18163             Kito Cheng  <kito.cheng@gmail.com>
18164             Kuan-Lin Chen  <kuanlinchentw@gmail.com>
18166         * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
18167         * config/nds32/constants.md (unspec_volatile_element): Add
18168         UNSPEC_VOLATILE_RELAX_GROUP.
18169         * config/nds32/nds32-relax-opt.c: New file.
18170         * config/nds32/nds32-predicates.c
18171         (nds32_symbol_load_store_p): New function.
18172         * config/nds32/nds32-protos.h
18173         (nds32_symbol_load_store_p): Declare function.
18174         (make_pass_nds32_relax_opt): Declare new rtl pass function.
18175         * config/nds32/nds32.c
18176         (nds32_register_pass): New function to register pass.
18177         (nds32_register_passes): New function to register passes.
18178         * config/nds32/nds32.md (relax_group): New pattern.
18179         * config/nds32/nds32.opt (mrelax-hint): New option.
18180         * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
18182 2018-04-01  Kito Cheng  <kito.cheng@gmail.com>
18184         * config/nds32/t-nds32: Modify files dependency.
18186 2018-04-01  Chung-Ju Wu  <jasonwucj@gmail.com>
18188         * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
18189         (PROFILE_HOOK): Define its implementation.
18191 2018-04-01  Chung-Ju Wu  <jasonwucj@gmail.com>
18193         * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
18194         type and 32-bit size.
18196 2018-04-01  Jakub Jelinek  <jakub@redhat.com>
18198         PR middle-end/85090
18199         * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
18200         (V_128_256): New mode iterator.
18201         (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
18202         (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
18203         (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
18204         of V.
18205         * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
18206         V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
18208 2018-03-31  Segher Boessenkool  <segher@kernel.crashing.org>
18210         PR target/83315
18211         * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
18212         NaN inputs correctly.
18214 2018-03-30  Peter Bergner  <bergner@vnet.ibm.com>
18216         PR target/80546
18217         * config/rs6000/vsx.md (??r): New mode attribute.
18218         (*vsx_mov<mode>_64bit): Use it.
18219         (*vsx_mov<mode>_32bit): Likewise.
18221 2018-03-30  Martin Sebor  <msebor@redhat.com>
18223         PR tree-optimization/84818
18224         * builtins.c (check_access): Use warning_n.
18226 2018-03-30  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
18228         PR target/83822
18229         * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
18230         condition.
18231         * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
18232         condition.
18234 2018-03-30  Julia Koval  <julia.koval@intel.com>
18236         PR target/84413
18237         * x86-tune.def (movx, partial_reg_dependency): Enable for
18238         m_SKYLAKE_AVX512.
18240 2018-03-29  Vladimir Makarov  <vmakarov@redhat.com>
18242         PR inline-asm/84985
18243         * lra-constraints.c (process_alt_operands): Move setting
18244         this_alternative_matches below.
18246 2018-03-29  Martin Liska  <mliska@suse.cz>
18248         PR lto/84995.
18249         * doc/invoke.texi: Document how LTO works with debug info.
18250         Describe auto-load support of binutils.  Mention 'x86-64'
18251         as valid option value of -march option.
18253 2018-03-29  Jakub Jelinek  <jakub@redhat.com>
18255         * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
18257         PR c/85094
18258         * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
18259         For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
18260         OEP_NO_HASH_CHECK for recursive call, to avoid exponential
18261         checking.
18263 2018-03-28  Peter Bergner  <bergner@vnet.ibm.com>
18265         PR target/84912
18266         * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
18267         (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
18268         * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
18269         for RS6000_BTM_POWERPC64.
18270         (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
18271         (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
18272         * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
18273         definition.
18274         (DIVDE): Use it.
18275         (DIVDEU): Likewise.
18277 2018-03-28  Carl Love  <cel@us.ibm.com>
18279         Revert
18280         2017-09-27  Carl Love  <cel@us.ibm.com>
18282         * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
18283         (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
18284         * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
18285         fctiw instruction.
18287 2018-03-28  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
18289         * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
18290         instead of __vector bool.
18291         (_mm_max_pu8): Likewise.
18292         (_mm_min_pi16): Likewise.
18294 2018-03-28  Peter Bergner  <bergner@vnet.ibm.com>
18296         PR target/84912
18297         * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
18298         (DIVWEUO): Likewise.
18299         (DIVDEO): Likewise.
18300         (DIVDEUO): Likewise.
18301         * config/rs6000/rs6000.c (builtin_function_type): Remove support for
18302         DIVWEUO and DIVDEUO.
18303         * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
18304         (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
18305         (div_extend): Likewise.
18306         * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
18307         builtin function.
18308         (__builtin_divweuo): Likewise.
18309         (__builtin_divdeo): Likewise.
18310         (__builtin_divdeuo): Likewise.
18312 2018-03-28  Jakub Jelinek  <jakub@redhat.com>
18314         PR target/85095
18315         * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
18316         *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
18318         PR tree-optimization/82004
18319         * gimple-match-head.c (optimize_pow_to_exp): New function.
18320         * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
18321         Don't fold to exp if optimize_pow_to_exp is false.
18323 2018-03-28  Martin Liska  <mliska@suse.cz>
18325         PR other/84819
18326         * calls.c (initialize_argument_information): Fix trailing space.
18327         * common.opt: Fix typo and provide better explanation for
18328         -fsanitize-coverage option.
18329         * config/i386/i386.opt: Fix typo.
18331 2018-03-28  Jakub Jelinek  <jakub@redhat.com>
18332             Martin Liska  <mliska@suse.cz>
18334         PR sanitizer/85081
18335         * gimplify.c (asan_poison_variable): Don't do the check for
18336         gimplify_omp_ctxp here.
18337         (gimplify_decl_expr): Do it here.
18338         (gimplify_target_expr): Likewise.
18340 2018-03-28  Martin Liska  <mliska@suse.cz>
18342         PR target/84988
18343         * config/i386/i386.c (ix86_function_arg_advance): Do not call
18344         chkp_type_bounds_count if MPX is not enabled.
18346 2018-03-27  Chung-Ju Wu  <jasonwucj@gmail.com>
18348         * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
18350 2018-03-27  Michael Meissner  <meissner@linux.vnet.ibm.com>
18352         PR target/84914
18353         * config/rs6000/rs6000.c (create_complex_muldiv): New helper
18354         function to create the function decl for complex long double
18355         multiply and divide for -mabi=ieeelongdouble.
18356         (init_float128_ieee): Call it.
18358 2018-03-27  H.J. Lu  <hongjiu.lu@intel.com>
18360         PR target/85044
18361         * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
18362         -fcf-protection=branch -mibt.
18363         * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
18365 2018-03-27  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
18367         PR target/81863
18368         * config/arm/arm.c (arm_valid_symbolic_address): Handle
18369         arm_word_relocations.
18371 2018-03-27  Cesar Philippidis  <cesar@codesourcery.com>
18373         PR target/85056
18374         * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
18375         extern array declarations.
18377 2018-03-27  Richard Biener  <rguenther@suse.de>
18379         PR middle-end/84067
18380         * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
18381         explicit single_use checks.
18383 2018-03-27  Richard Biener  <rguenther@suse.de>
18385         PR tree-optimization/85082
18386         * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
18387         Valueize the VUSE.
18389 2018-03-27  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
18391         * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
18392         * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
18393         Turn on fasynchronous-unwind-tables and funwind-tables.
18395 2018-03-26  Uros Bizjak  <ubizjak@gmail.com>
18397         PR target/85073
18398         * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
18399         (*bmi_blsr_<mode>_ccz): Ditto.
18401 2018-03-26  Tom de Vries  <tom@codesourcery.com>
18403         PR tree-optimization/85063
18404         * omp-general.c (offloading_function_p): New function.  Factor out
18405         of ...
18406         * omp-offload.c (pass_omp_target_link::gate): ... here.
18407         * omp-general.h (offloading_function_p): Declare.
18408         * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
18409         with attribute omp declare target for offloading functions.
18411 2018-03-24  Richard Sandiford  <richard.sandiford@linaro.org>
18413         PR tree-optimization/84005
18414         * tree-data-ref.h (get_base_for_alignment): Declare.
18415         * tree-data-ref.c (get_base_for_alignment_1): New function.
18416         (get_base_for_alignment): Likewise.
18417         * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
18418         get_base_for_alignment to find a suitable base object, instead
18419         of always using drb->base_address.
18421 2018-03-23  Jakub Jelinek  <jakub@redhat.com>
18423         PR inline-asm/85022
18424         * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
18425         known size by default.
18427 2018-03-23  Vladimir Makarov  <vmakarov@redhat.com>
18429         PR inline-asm/85030
18430         * lra-constraints.c (process_alt_operands): Don't match BLKmode
18431         and non BLKmode operands.
18433 2018-03-23  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
18435         PR target/85026
18436         * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
18437         Clean up attributes.
18439 2018-03-23  Richard Biener  <rguenther@suse.de>
18441         PR debug/85020
18442         * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
18443         we are going to emit early debug for LTO.
18445 2018-03-23  Jakub Jelinek  <jakub@redhat.com>
18447         PR inline-asm/85034
18448         * function.c (match_asm_constraints_1): Don't optimize if input
18449         doesn't satisfy general_operand predicate for output's mode.
18451         PR inline-asm/85022
18452         * alias.c (write_dependence_p): Don't require for x_canonicalized
18453         non-VOIDmode if x has VOIDmode.
18455         PR sanitizer/85029
18456         * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
18457         just don't try to optimize it rather than assert it never happens.
18459 2018-03-22  Kelvin Nilsen  <kelvin@gcc.gnu.org>
18461         * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
18462         macro expansions for definition of ST_INTERNAL_<mode> and
18463         LD_INTERNAL_<mode> builtins.
18464         * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
18465         Remove prototype.
18466         * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
18467         function.
18468         (altivec_expand_st_builtin): Likewise.
18469         (altivec_expand_builtin): Remove calls to deleted functions.
18470         (rs6000_address_for_altivec): Delete this function.
18471         * config/rs6000/vector.md: Remove expands for
18472         vector_altivec_load_<mode> and vector_altivec_store_<mode>.
18474 2018-03-22  Sudakshina Das  <sudi.das@arm.com>
18476         PR target/84826
18477         * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
18478         * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
18479         re-computing once computed.
18480         (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
18481         (arm_init_machine_status): Initialize
18482         machine->static_chain_stack_bytes.
18484 2018-03-22  Kelvin Nilsen  <kelvin@gcc.gnu.org>
18486         PR target/84760
18487         * doc/extend.texi: Add four new prototypes for vec_ld.
18488         * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
18489         definitions for more logical presentation.
18490         * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
18491         entries for V1TI variants of __builtin_altivec_ld builtin.
18492         * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
18493         handling of V1TI variant of LVX icode pattern.
18494         (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
18495         (rs6000_gimple_fold_builtin): Likewise.
18496         (altivec_init_builtins): Add code to define
18497         __builtin_altivec_lvx_v1ti function.
18499 2018-03-22  Jakub Jelinek  <jakub@redhat.com>
18501         PR inline-asm/84941
18502         * function.c (match_asm_constraints_1): Don't do the optimization
18503         if input isn't a REG, SUBREG, MEM or constant.
18505 2018-03-22  Tom de Vries  <tom@codesourcery.com>
18507         PR tree-optimization/84956
18508         * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
18509         bb_has_abnormal_pred.
18511 2018-03-22  Jakub Jelinek  <jakub@redhat.com>
18513         PR sanitizer/85018
18514         * dwarf2asm.c (dw2_output_indirect_constant_1): Set
18515         DECL_INITIAL (decl) to decl at the end.
18516         * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
18517         adjust the comment.
18519 2018-03-21  Joseph Myers  <joseph@codesourcery.com>
18521         * doc/extend.texi (__builtin_tgmath): Document when complex
18522         integer types are treated as _Complex _Float64.
18524 2018-03-21  Tom de Vries  <tom@codesourcery.com>
18526         * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
18528 2018-03-21  Jakub Jelinek  <jakub@redhat.com>
18530         PR tree-optimization/84960
18531         * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
18532         if it is ENTRY block, move them into single succ of ENTRY in that case.
18534 2018-03-21  Richard Sandiford  <richard.sandiford@linaro.org>
18536         PR tree-optimization/84811
18537         * poly-int.h (poly_span_traits): Remove the T3 parameter and
18538         promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
18539         (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
18540         (known_subrange_p): Update accordingly.  Cast each value involved
18541         in the size comparison, rather than casting the result of the
18542         subtraction.
18544 2018-03-21  Jakub Jelinek  <jakub@redhat.com>
18546         PR tree-optimization/84982
18547         * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
18548         by flipping the least significant bit rather than all bits from
18549         bitpos to bitpos + bitsize - 1.
18551 2018-03-21  Nathan Sidwell  <nathan@acm.org>
18553         * doc/extend.texi (Deprecated Features): Remove mention of
18554         long-deleted deprecations.
18556 2018-03-21  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
18558         PR jit/84288
18559         * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
18560         * configure: Regenerate.
18562 2018-03-21  Tom de Vries  <tom@codesourcery.com>
18564         PR tree-optimization/83126
18565         * tree-parloops.c (num_phis): New function.
18566         (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
18568 2018-03-21  Nathan Sidwell  <nathan@acm.org>
18570         * doc/extend.texi (Deprecated Features): Update deprecated flags,
18571         mention anon-struct/union members and trailing attributes.
18573 2018-03-21  Bin Cheng  <bin.cheng@arm.com>
18575         PR tree-optimization/84969
18576         * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
18577         builtin memset partitions if they set different rhs values.
18579 2018-03-21  Jakub Jelinek  <jakub@redhat.com>
18581         PR rtl-optimization/84989
18582         * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
18583         VEC_DUPLICATE with scalar result mode.
18585 2018-03-21  Martin Liska  <mliska@suse.cz>
18587         PR ipa/84963
18588         * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
18589         not intended return statement.
18591 2018-03-21  Martin Liska  <mliska@suse.cz>
18593         PR target/84988
18594         * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
18595         (chkp_find_bound_slots_1): Limit number of iterations.
18597 2018-03-20  David H. Gutteridge  <dhgutteridge@sympatico.ca>
18599         PR target/84838
18600         * Minor grammar fixes for x86 options.
18602 2018-03-20  Jakub Jelinek  <jakub@redhat.com>
18604         PR debug/84875
18605         * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
18606         holding REG_CFA_RESTORE notes, instead turn them into a USE.
18608 2018-03-20  Peter Bergner  <bergner@vnet.ibm.com>
18610         PR target/83789
18611         * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
18612         (altivec_lvx_<mode>_1op): Likewise.
18613         (altivec_stvx_<mode>_2op): Likewise.
18614         (altivec_stvx_<mode>_1op): Likewise.
18615         (altivec_lvx_<VM2:mode>): New define_expand.
18616         (altivec_stvx_<VM2:mode>): Likewise.
18617         (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
18618         (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
18619         (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
18620         (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
18621         * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
18622         (rs6000_gen_lvx): Likewise.
18623         * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
18624         (altivec_expand_stv_builtin): Likewise.
18625         (altivec_expand_builtin): Likewise.
18626         * config/rs6000/vector.md: Likewise.
18628 2018-03-20  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
18630         PR target/82518
18631         * config/arm/arm.c (arm_array_mode_supported_p): Return false for
18632         BYTES_BIG_ENDIAN.
18634 2018-03-20  Richard Biener  <rguenther@suse.de>
18636         PR target/84986
18637         * config/i386/i386.c (ix86_add_stmt_cost): Only cost
18638         sign-conversions as zero, fall back to standard scalar_stmt
18639         cost for the rest.
18641 2018-03-20  Martin Liska  <mliska@suse.cz>
18643         PR ipa/84825
18644         * predict.c (rebuild_frequencies): Handle case when we have
18645         PROFILE_ABSENT, but flag_guess_branch_prob is false.
18647 2018-03-20  Jakub Jelinek  <jakub@redhat.com>
18649         PR target/84990
18650         * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
18651         flag_section_anchors.
18652         * varasm.c (use_blocks_for_decl_p): Remove hack for
18653         dw2_force_const_mem.
18655         PR target/84845
18656         * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
18657         to ...
18658         (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this.  If pseudos can't
18659         be created, use lowpart_subreg of operands[0] rather than operands[0]
18660         itself.
18661         (*aarch64_reg_<mode>3_minus_mask): Rename to ...
18662         (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
18663         (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
18664         and n constraint instead of aarch64_shift_imm_di and Usd.
18665         (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
18666         (*aarch64_<optab>_reg_minus<mode>3): ... this.
18668 2018-03-20  Sudakshina Das  <sudi.das@arm.com>
18670         PR target/82989
18671         * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
18672         to favor GPR over NEON registers.
18673         (<shift>di3_neon): Likewise.
18675 2018-03-20  Tom de Vries  <tom@codesourcery.com>
18677         PR target/84952
18678         * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
18679         (nvptx_process_pars): Emit bar.sync asap and alap.
18681 2018-03-20  Tom de Vries  <tom@codesourcery.com>
18683         PR target/84954
18684         * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
18685         seen_label if seen_label is already set.
18687 2018-03-20  Jakub Jelinek  <jakub@redhat.com>
18689         PR target/84945
18690         * config/i386/i386.c (fold_builtin_cpu): For features above 31
18691         use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
18692         Use 1U instead of 1.  Formatting fixes.
18694         PR c/84953
18695         * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
18696         instead of TREE_TYPE (s1) for the return value.
18698 2018-03-19  Jakub Jelinek  <jakub@redhat.com>
18700         PR tree-optimization/84946
18701         * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
18702         bitsize + bitsize in poly_uint64 rather than poly_int64.
18704         PR sanitizer/78651
18705         * dwarf2asm.c: Include fold-const.c.
18706         (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
18707         of decl rather than decl itself.
18709         PR rtl-optimization/84643
18710         * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
18712 2018-03-19  Maxim Ostapenko  <m.ostapenko@samsung.com>
18714         PR sanitizer/78651
18715         * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
18716         calling assemble_variable.
18718 2018-03-19  Sudakshina Das  <sudi.das@arm.com>
18720         PR target/81647
18721         * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
18722         instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
18724 2018-03-19  Jim Wilson  <jimw@sifive.com>
18726         PR bootstrap/84856
18727         * config/riscv/riscv.c (riscv_function_arg_boundary): Use
18728         PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
18729         (riscv_first_stack_step): Likewise.
18730         (riscv_option_override): Use STACK_BOUNDARY instead of
18731         MIN_STACK_BOUNDARY.
18732         * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
18733         MIN_STACK_BOUNDARY.
18734         (BIGGEST_ALIGNMENT): Set to 128.
18735         (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
18736         (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
18737         STACK_BOUNDARY.
18739 2018-03-19  Richard Biener  <rguenther@suse.de>
18741         PR tree-optimization/84933
18742         * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
18743         values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
18745 2018-03-19  Richard Biener  <rguenther@suse.de>
18747         PR tree-optimization/84859
18748         * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
18749         (cond_if_else_store_replacement): Perform sinking operation on
18750         single-store BBs regardless of MAX_STORES_TO_SINK setting.
18751         Generalize what a BB with a single eligible store is.
18753 2018-03-19  Richard Biener  <rguenther@suse.de>
18755         PR tree-optimization/84929
18756         * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
18757         chrec_is_positive against non-chrec arg.
18759 2018-03-19  Tamar Christina  <tamar.christina@arm.com>
18761         PR target/84711
18762         * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
18764 2018-03-18  Martin Liska  <mliska@suse.cz>
18766         PR rtl-optimization/84635
18767         * regrename.c (build_def_use): Use matches_mode only when
18768         matches >= 0.
18770 2018-03-18  Richard Sandiford  <richard.sandiford@linaro.org>
18772         PR tree-optimization/84913
18773         * tree-vect-loop.c (vectorizable_reduction): Don't try to
18774         vectorize chains of COND_EXPRs.
18776 2018-03-18  Chung-Ju Wu  <jasonwucj@gmail.com>
18778         * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
18780 2018-03-18  Chung-Ju Wu  <jasonwucj@gmail.com>
18782         * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
18784 2018-03-18  Chung-Ju Wu  <jasonwucj@gmail.com>
18786         * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
18788 2018-03-17  Chung-Ju Wu  <jasonwucj@gmail.com>
18789             Kito Cheng  <kito.cheng@gmail.com>
18791         * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
18792         * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
18793         (nds32_adjust_reg_alloc_order): New function.
18794         * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
18796 2018-03-17  Kito Cheng  <kito.cheng@gmail.com>
18798         * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
18799         nds32_print_operand, nds32_print_operand_address): Use
18800         HOST_WIDE_INT_PRINT_DEC instead.
18802 2018-03-17  Chung-Ju Wu  <jasonwucj@gmail.com>
18804         * config/nds32/nds32.c (nds32_register_priority): Modify cost.
18806 2018-03-17  Jakub Jelinek  <jakub@redhat.com>
18808         PR target/84902
18809         * config/i386/i386.c (initial_ix86_tune_features,
18810         initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
18811         unsigned long long.
18812         (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
18813         to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
18814         rather than 1u << ix86_tune.  Formatting fix.
18815         (ix86_option_override_internal): Change ix86_arch_mask from
18816         unsigned int to unsigned HOST_WIDE_INT, initialize to
18817         HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
18818         (ix86_function_specific_restore): Likewise.
18820 2018-03-16  Jakub Jelinek  <jakub@redhat.com>
18822         PR target/84899
18823         * postreload.c (reload_combine_recognize_pattern): Perform
18824         INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
18825         truncate_int_for_mode the result for the destination's mode.
18827         PR c/84909
18828         * hsa-gen.c (mem_type_for_type): Fix comment typo.
18829         * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
18830         Likewise.
18831         * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
18832         Likewise.
18834 2018-03-16  Vladimir Makarov  <vmakarov@redhat.com>
18836         PR target/84876
18837         * lra-assigns.c (lra_split_hard_reg_for): Don't use
18838         regno_allocno_class_array and sorted_pseudos.
18839         * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
18840         insns where regno is used.
18842 2018-03-16  Martin Liska  <mliska@suse.cz>
18844         PR ipa/84833
18845         * multiple_target.c (create_dispatcher_calls): Redirect
18846         reference in the symbol table.
18848 2018-03-16  Martin Liska  <mliska@suse.cz>
18850         PR ipa/84722
18851         * multiple_target.c (create_dispatcher_calls): Redirect also
18852         an alias.
18854 2018-03-16  Jakub Jelinek  <jakub@redhat.com>
18856         PR c++/79937
18857         PR c++/82410
18858         * tree.h (TARGET_EXPR_NO_ELIDE): Define.
18859         * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
18860         TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
18862 2018-03-16  Julia Koval  <julia.koval@intel.com>
18864         * doc/invoke.texi (Skylake Server): Add CLWB.
18865         Cannonlake): Remove CLWB.
18867 2018-03-16  Jakub Jelinek  <jakub@redhat.com>
18869         PR tree-optimization/84841
18870         * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
18871         1 << 3.
18872         (FLOAT_ONE_CONST_TYPE): Define.
18873         (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
18874         (sort_by_operand_rank): Put entries with higher constant_type last
18875         rather than first to match comments.
18877 2018-03-15  Sandra Loosemore  <sandra@codesourcery.com>
18879         * config/nios2/nios2.md (movsi_internal): Fix thinko in
18880         split predicate.
18882 2018-03-15  Jakub Jelinek  <jakub@redhat.com>
18884         PR c++/79085
18885         * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
18886         check and use address of target always.
18888 2018-03-15  H.J. Lu  <hongjiu.lu@intel.com>
18890         PR target/84574
18891         * config/i386/i386.c (indirect_thunk_needed): Update comments.
18892         (indirect_thunk_bnd_needed): Likewise.
18893         (indirect_thunks_used): Likewise.
18894         (indirect_thunks_bnd_used): Likewise.
18895         (indirect_return_needed): New.
18896         (indirect_return_bnd_needed): Likewise.
18897         (output_indirect_thunk_function): Add a bool argument for
18898         function return.
18899         (output_indirect_thunk_function): Don't generate alias for
18900         function return thunk.
18901         (ix86_code_end): Call output_indirect_thunk_function to generate
18902         function return thunks.
18903         (ix86_output_function_return): Set indirect_return_bnd_needed
18904         and indirect_return_needed instead of indirect_thunk_bnd_needed
18905         and indirect_thunk_needed.
18907 2018-03-15  Olga Makhotina  <olga.makhotina@intel.com>
18909         * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
18910         (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
18911         (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
18913 2018-03-15  David Malcolm  <dmalcolm@redhat.com>
18914             Paul Hua <paul.hua.gm@gmail.com>
18916         PR c/84852
18917         * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
18919 2018-03-15  Segher Boessenkool  <segher@kernel.crashing.org>
18921         * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
18922         TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
18923         resp. SFmode cases.
18925 2018-03-15  Tamar Christina  <tamar.christina@arm.com>
18927         PR target/84711
18928         * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
18929         instead of GET_MODE_SIZE when comparing Units.
18931 2018-03-15  Vladimir Mezentsev  <vladimir.mezentsev@oracle.com>
18933         PR target/68256
18934         * varasm.c (hash_section): Return an unchangeble hash value
18935         * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
18936         Return !aarch64_can_use_per_function_literal_pools_p ().
18938 2018-03-15  Jakub Jelinek  <jakub@redhat.com>
18940         PR target/84860
18941         * optabs.c (emit_conditional_move): Pass address of cmode's copy
18942         rather than address of cmode as last argument to prepare_cmp_insn.
18944 2018-03-15  Julia Koval  <julia.koval@intel.com>
18946         * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
18947         F_AVX512VNNI, F_AVX512BITALG): New.
18949 2018-03-14  John David Anglin  <danglin@gcc.gnu.org>
18951         PR target/83451
18952         * config/pa/pa.c (pa_emit_move_sequence):  Always emit secondary reload
18953         insn for floating-point loads and stores.
18955 2018-03-14  Carl Love  <cel@us.ibm.com>
18957         * config/rs6000/rs6000-c.c: Add macro definitions for
18958         ALTIVEC_BUILTIN_VEC_PERMXOR.
18959         * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
18960         * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
18961         * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
18962         * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
18963         UNSPEC_VPERMXOR.
18964         * config/doc/extend.texi: Add prototypes for vec_permxor.
18966 2018-03-14  David Malcolm  <dmalcolm@redhat.com>
18968         PR c/84852
18969         * diagnostic-show-locus.c (class layout_point): Convert m_line
18970         from int to linenum_type.
18971         (line_span::comparator): Use linenum "compare" function when
18972         comparing line numbers.
18973         (test_line_span): New function.
18974         (layout_range::contains_point): Convert param "row" from int to
18975         linenum_type.
18976         (layout_range::intersects_line_p): Likewise.
18977         (layout::will_show_line_p): Likewise.
18978         (layout::print_source_line): Likewise.
18979         (layout::should_print_annotation_line_p): Likewise.
18980         (layout::print_annotation_line): Likewise.
18981         (layout::print_leading_fixits): Likewise.
18982         (layout::annotation_line_showed_range_p): Likewise.
18983         (struct line_corrections): Likewise for field m_row.
18984         (line_corrections::line_corrections): Likewise for param "row".
18985         (layout::print_trailing_fixits): Likewise.
18986         (layout::get_state_at_point): Likewise.
18987         (layout::get_x_bound_for_row): Likewise.
18988         (layout::print_line): Likewise.
18989         (diagnostic_show_locus): Likewise for locals "last_line" and "row".
18990         (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
18991         * input.c (selftest::test_linenum_comparisons): New function.
18992         (selftest::input_c_tests): Call it.
18993         * selftest.c (selftest::test_assertions): Test ASSERT_GT,
18994         ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
18995         * selftest.h (ASSERT_GT): New macro.
18996         (ASSERT_GT_AT): New macro.
18997         (ASSERT_LT): New macro.
18998         (ASSERT_LT_AT): New macro.
19000 2018-03-14  Segher Boessenkool  <segher@kernel.crashing.org>
19002         PR rtl-optimization/84780
19003         * combine.c (distribute_links): Don't make a link based on pc_rtx.
19005 2018-03-14  Martin Liska  <mliska@suse.cz>
19007         * tree.c (record_node_allocation_statistics): Use
19008         get_stats_node_kind.
19009         (get_stats_node_kind): New function extracted from
19010         record_node_allocation_statistics.
19011         (free_node): Use get_stats_node_kind.
19013 2018-03-14  Richard Biener  <rguenther@suse.de>
19015         * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
19016         that the value-set of ANTIC_IN doesn't grow.
19018         Revert
19019         * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
19020         member.
19021         (BB_VISITED_WITH_VISITED_SUCCS): New define.
19022         (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
19024 2018-03-14  Julia Koval  <julia.koval@intel.com>
19026         * config.gcc (icelake-client, icelake-server): New.
19027         (icelake): Remove.
19028         * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
19029         (initial_ix86_arch_features): Ditto.
19030         (PTA_SKYLAKE): Add SGX.
19031         (PTA_ICELAKE): Remove.
19032         (PTA_ICELAKE_CLIENT): New.
19033         (PTA_ICELAKE_SERVER): New.
19034         (ix86_option_override_internal): Split up icelake on icelake client and
19035         icelake server.
19036         (get_builtin_code_for_version): Ditto.
19037         (fold_builtin_cpu): Ditto.
19038         * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
19039         * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
19040         * config/i386/i386.h (processor_type): Ditto.
19041         * doc/invoke.texi: Ditto.
19043 2018-03-14  Jakub Jelinek  <jakub@redhat.com>
19045         PR sanitizer/83392
19046         * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
19047         INTEGER_CST offset, add it together with bitpos / 8 and
19048         sign extend based on POINTER_SIZE.
19050         PR target/84844
19051         Revert
19052         2017-04-20  Uros Bizjak  <ubizjak@gmail.com>
19054         PR target/78090
19055         * config/i386/constraints.md (Yc): New register constraint.
19056         * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
19057         Use Yc constraint for alternative 2 of operand 0.  Remove
19058         preferred_for_speed attribute.
19060 2018-03-14  Richard Biener  <rguenther@suse.de>
19062         PR tree-optimization/84830
19063         * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
19064         with the old one to avoid oscillations.
19066 2018-03-13  Vladimir Makarov  <vmakarov@redhat.com>
19068         PR target/83712
19069         * lra-assigns.c (find_all_spills_for): Ignore uninteresting
19070         pseudos.
19071         (assign_by_spills): Return a flag of reload assignment failure.
19072         Do not process the reload assignment failures.  Do not spill other
19073         reload pseudos if they has the same reg class.  Update n if
19074         necessary.
19075         (lra_assign): Add a return arg.  Set up from the result of
19076         assign_by_spills call.
19077         (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
19078         * lra-constraints.c (split_reg): Add a new arg.  Use it instead of
19079         usage_insns if it is not NULL.
19080         (spill_hard_reg_in_range): New function.
19081         (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
19082         * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
19083         function prototypes.
19084         (lra_assign): Change prototype.
19085         * lra.c (lra): Add code to deal with fails by splitting hard reg
19086         live ranges.
19088 2018-03-01  Palmer Dabbelt  <palmer@sifive.com>
19090         * config/riscv/riscv.opt (mrelax): New option.
19091         * config/riscv/riscv.c (riscv_file_start): Emit ".option
19092         "norelax" when riscv_mrelax is disabled.
19093         * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
19095 2018-03-13  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
19097         PR target/84743
19098         * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
19099         reassociation for int modes.
19101 2018-03-13  Richard Sandiford  <richard.sandiford@linaro.org>
19103         * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
19104         Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
19105         for big-endian.
19106         * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
19107         * config/aarch64/aarch64-sve.md
19108         (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
19109         (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
19110         (*extend<mode><Vwide>2): Rename to...
19111         (aarch64_sve_extend<mode><Vwide>2): ...this.
19112         (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
19113         renaming the old pattern to...
19114         (aarch64_sve_punpk<perm_hilo>_<mode>): ...this.  Only define
19115         unsigned packs.
19116         (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
19117         define_expand, renaming the old pattern to...
19118         (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
19119         (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
19120         (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
19121         account when deciding which SVE instruction the optab should use.
19122         (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
19124 2018-03-13  Richard Sandiford  <richard.sandiford@linaro.org>
19126         * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
19127         (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
19128         (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
19129         (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
19130         (tlsdesc_small_<mode>): Turn a define_expand and use
19131         tlsdesc_small_sve_<mode> for SVE.  Rename original define_insn to...
19132         (tlsdesc_small_advsimd_<mode>): ...this.
19133         (tlsdesc_small_sve_<mode>): New pattern.
19135 2018-03-13  Richard Sandiford  <richard.sandiford@linaro.org>
19137         * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
19138         (UNSPEC_UMUL_HIGHPART): New constants.
19139         (MUL_HIGHPART): New int iteraor.
19140         (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
19141         * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
19142         define_expand.
19143         (*<su>mul<mode>3_highpart): New define_insn.
19145 2018-03-13  Eric Botcazou  <ebotcazou@adacore.com>
19147         PR lto/84805
19148         * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
19149         incomplete types.
19151 2018-03-13  Martin Liska  <mliska@suse.cz>
19153         PR ipa/84658.
19154         * (sem_item_optimizer::sem_item_optimizer): Initialize new
19155         vector.
19156         (sem_item_optimizer::~sem_item_optimizer): Release it.
19157         (sem_item_optimizer::merge_classes): Register variable aliases.
19158         (sem_item_optimizer::fixup_pt_set): New function.
19159         (sem_item_optimizer::fixup_points_to_sets): Likewise.
19160         * ipa-icf.h: Declare new variables and functions.
19162 2018-03-13  Jakub Jelinek  <jakub@redhat.com>
19164         PR middle-end/84834
19165         * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
19166         integer_pow2p@2 and test integer_pow2p in condition.
19167         (A < 0 ? C : 0): Similarly for @1.
19169         PR middle-end/84831
19170         * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
19171         characters starting at p contain '\0' character, don't look beyond
19172         that.
19174         PR target/84827
19175         * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
19176         pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
19178         PR target/84828
19179         * reg-stack.c (change_stack): Change update_end var from int to
19180         rtx_insn *, if non-NULL don't update just BB_END (current_block), but
19181         also call set_block_for_insn on the newly added insns and rescan.
19183         PR target/84786
19184         * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
19185         on the last operand.
19187         PR c++/84704
19188         * tree.c (stabilize_reference_1): Return save_expr (e) for
19189         STATEMENT_LIST even if it doesn't have side-effects.
19191 2018-03-12  Jonathan Wakely  <jwakely@redhat.com>
19193         * doc/invoke.texi (-mclflushopt): Fix spelling of option.
19195 2018-03-12  Renlin Li  <renlin.li@arm.com>
19197         * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
19198         aarch64_output_scalar_simd_mov_immediate.
19200 2018-03-12  Martin Sebor  <msebor@redhat.com>
19202         PR tree-optimization/83456
19203         * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
19204         for perfectly overlapping calls to memcpy.
19205         (gimple_fold_builtin_memory_chk): Same.
19206         (gimple_fold_builtin_strcpy): Handle no-warning.
19207         (gimple_fold_builtin_stxcpy_chk): Same.
19208         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
19210 2018-03-12  Segher Boessenkool  <segher@kernel.crashing.org>
19212         * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
19213         parameter.  Use it for SFmode.
19214         (rs6000_function_arg_advance_1): Adjust.
19215         (rs6000_function_arg): Adjust.
19216         (rs6000_gimplify_va_arg): Pass false for that new parameter.
19218 2018-03-12  Segher Boessenkool  <segher@kernel.crashing.org>
19220         PR rtl-optimization/84169
19221         PR rtl-optimization/84780
19222         * combine.c (can_combine_p): Check for a 2-insn combination whether
19223         the destination register is used between the two insns, too.
19225 2018-03-12  Richard Biener  <rguenther@suse.de>
19227         PR tree-optimization/84803
19228         * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
19229         for refs DR analysis didn't process.
19231 2018-03-12  Richard Biener  <rguenther@suse.de>
19233         PR tree-optimization/84777
19234         * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
19235         force-vectorize loops ignore whether we are optimizing for size.
19237 2018-03-12  Chung-Ju Wu  <jasonwucj@gmail.com>
19239         * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
19240         (TARGET_MD_ASM_ADJUST): Define.
19242 2018-03-12  Monk Chiang  <sh.chiang04@gmail.com>
19243             Kito Cheng  <kito.cheng@gmail.com>
19244             Chung-Ju Wu  <jasonwucj@gmail.com>
19246         * config/nds32/nds32.c (nds32_compute_stack_frame,
19247         nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
19248         nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
19249         nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
19250         nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
19251         * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
19252         NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
19253         * config/nds32/nds32.md (prologue, epilogue): Use macro
19254         NDS32_V3PUSH_AVAILABLE_P to do checking.
19256 2018-03-11  Jakub Jelinek  <jakub@redhat.com>
19258         PR debug/58150
19259         * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
19260         DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
19261         but on TYPE_SIZE.  Don't do anything for ENUM_IS_OPAQUE if not creating
19262         a new die.  Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE.  Guard
19263         addition of most attributes on !orig_type_die or the attribute not
19264         being present already.  Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
19266 2018-03-11  Kito Cheng  <kito.cheng@gmail.com>
19267             Chung-Ju Wu  <jasonwucj@gmail.com>
19269         * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
19270         __NDS32_VH__ macro.
19271         * config/nds32/nds32.opt (mvh): New option.
19273 2018-03-11  Kito Cheng  <kito.cheng@gmail.com>
19274             Chung-Ju Wu  <jasonwucj@gmail.com>
19276         * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
19277         function.
19278         * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
19279         * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
19280         definition.
19282 2018-03-11  Kito Cheng  <kito.cheng@gmail.com>
19283             Chung-Ju Wu  <jasonwucj@gmail.com>
19285         * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
19286         function.
19287         * config/nds32/nds32-multiple.md (strlensi): New pattern.
19288         * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
19290 2018-03-11  Monk Chiang  <sh.chiang04@gmail.com>
19291             Kito Cheng  <kito.cheng@gmail.com>
19292             Chung-Ju Wu  <jasonwucj@gmail.com>
19294         * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
19295         UNSPEC_FFMISM and UNSPEC_FLMISM.
19296         * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
19297         for ffb, ffmism and flmism.
19298         * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
19299         (unspec_ffmism): Ditto.
19300         (unspec_flmism): Ditto.
19301         (nds32_expand_builtin_impl): Check if string extension is available.
19302         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
19303         NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
19305 2018-03-10  Vladimir Makarov  <vmakarov@redhat.com>
19307         Reverting patch:
19308         2018-03-09  Vladimir Makarov  <vmakarov@redhat.com>
19310         PR target/83712
19311         * lra-assigns.c (assign_by_spills): Return a flag of reload
19312         assignment failure.  Do not process the reload assignment
19313         failures.  Do not spill other reload pseudos if they has the same
19314         reg class.
19315         (lra_assign): Add a return arg.  Set up from the result of
19316         assign_by_spills call.
19317         (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
19318         * lra-constraints.c (split_reg): Add a new arg.  Use it instead of
19319         usage_insns if it is not NULL.
19320         (spill_hard_reg_in_range): New function.
19321         (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
19322         * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
19323         function prototypes.
19324         (lra_assign): Change prototype.
19325         * lra.c (lra): Add code to deal with fails by splitting hard reg
19326         live ranges.
19328 2018-03-10  H.J. Lu  <hongjiu.lu@intel.com>
19330         PR target/84807
19331         * config/i386/i386.opt: Replace Enforcment with Enforcement.
19333 2018-03-10  Alexandre Oliva  <aoliva@redhat.com>
19335         PR debug/84620
19336         * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
19337         (dw_val_node): Add val_symbolic_view.
19338         * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
19339         (symview_upper_bound): New.
19340         (new_line_info_table): Initialize symviews_since_reset.
19341         (dwarf2out_source_line): Count symviews_since_reset and set
19342         symview_upper_bound.
19343         (dw_val_equal_p): Handle symview.
19344         (add_AT_symview): New.
19345         (print_dw_val): Handle symview.
19346         (attr_checksum, attr_checksum_ordered): Likewise.
19347         (same_dw_val_p, size_of_die): Likewise.
19348         (value_format, output_die): Likewise.
19349         (add_high_low_attributes): Use add_AT_symview for entry_view.
19350         (dwarf2out_finish): Reset symview_upper_bound, clear
19351         zero_view_p.
19353 2018-03-09  Peter Bergner  <bergner@vnet.ibm.com>
19355         PR target/83969
19356         * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
19357         Add strict argument and use it.
19358         (rs6000_split_multireg_move): Update for new strict argument.
19359         (mem_operand_gpr): Disallow all non-offsettable addresses.
19360         * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
19362 2018-03-09  Jakub Jelinek  <jakub@redhat.com>
19364         PR target/84772
19365         * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
19366         temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
19367         * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
19369         PR c++/84767
19370         * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
19371         decl, use remap_type if we want to use the type.
19373 2018-03-09  Martin Sebor  <msebor@redhat.com>
19375         PR tree-optimization/84526
19376         * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
19377         Remove dead code.
19378         (builtin_access::generic_overlap): Be prepared to handle non-array
19379         base objects.
19381 2018-03-09  Alexandre Oliva  <aoliva@redhat.com>
19383         PR rtl-optimization/84682
19384         * lra-constraints.c (process_address_1): Check is_address flag
19385         for address constraints.
19386         (process_alt_operands): Likewise.
19387         * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
19388         preprocess_constraints.
19389         * recog.h (preprocess_constraints): Add oploc parameter.
19390         Adjust callers.
19391         * recog.c (preprocess_constraints): Test address_operand for
19392         CT_ADDRESS constraints.
19394 2018-03-09  Vladimir Makarov  <vmakarov@redhat.com>
19396         PR target/83712
19397         * lra-assigns.c (assign_by_spills): Return a flag of reload
19398         assignment failure.  Do not process the reload assignment
19399         failures.  Do not spill other reload pseudos if they has the same
19400         reg class.
19401         (lra_assign): Add a return arg.  Set up from the result of
19402         assign_by_spills call.
19403         (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
19404         * lra-constraints.c (split_reg): Add a new arg.  Use it instead of
19405         usage_insns if it is not NULL.
19406         (spill_hard_reg_in_range): New function.
19407         (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
19408         * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
19409         function prototypes.
19410         (lra_assign): Change prototype.
19411         * lra.c (lra): Add code to deal with fails by splitting hard reg
19412         live ranges.
19414 2018-03-09  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
19416         PR target/83193
19417         * common/config/arm/arm-common.c (arm_parse_arch_option_name):
19418         Accept complain bool parameter.  Only emit errors if it is true.
19419         (arm_parse_cpu_option_name): Likewise.
19420         (arm_target_thumb_only): Adjust callers of the above.
19421         * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
19422         prototype to take a default true bool parameter.
19423         (arm_parse_arch_option_name): Likewise.
19425 2018-03-09  David Malcolm  <dmalcolm@redhat.com>
19426             Francois-Xavier Coudert  <fxcoudert@gcc.gnu.org>
19428         PR jit/64089
19429         PR jit/84288
19430         * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
19431         * configure: Regenerate.
19432         * configure.ac ("linker --version-script option"): New.
19433         ("linker soname option"): New.
19435 2018-03-09  Richard Biener  <rguenther@suse.de>
19437         PR tree-optimization/84775
19438         * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
19439         immediate uses of predicate stmts and mark them modified.
19441         Revert
19442         PR tree-optimization/84178
19443         * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
19444         to caller.
19445         (version_loop_for_if_conversion): Delay update_ssa call.
19446         (tree_if_conversion): Delay update_ssa until after predicate
19447         insertion.
19449 2018-03-09  Eric Botcazou  <ebotcazou@adacore.com>
19451         PR target/84763
19452         * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
19453         when the function accesses prior frames.
19455 2018-03-08  Jakub Jelinek  <jakub@redhat.com>
19457         PR debug/84456
19458         * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
19459         gen_llsym, otherwise call maybe_gen_llsym.
19461         PR inline-asm/84742
19462         * recog.c (asm_operand_ok): Return 0 if multi-character constraint
19463         has ',' character inside of it.
19465 2018-03-08  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
19467         PR target/84748
19468         * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
19469         as clobbering CC_REGNUM.
19471 2018-03-08  Richard Biener  <rguenther@suse.de>
19473         PR middle-end/84552
19474         * tree-scalar-evolution.c: Include tree-into-ssa.h.
19475         (follow_copies_to_constant): Do not follow SSA names registered
19476         for update.
19478 2018-03-08  Richard Biener  <rguenther@suse.de>
19480         PR tree-optimization/84178
19481         * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
19482         to caller.
19483         (version_loop_for_if_conversion): Delay update_ssa call.
19484         (tree_if_conversion): Delay update_ssa until after predicate
19485         insertion.
19487 2018-03-08  David Malcolm  <dmalcolm@redhat.com>
19489         PR tree-optimization/84178
19490         * tree-if-conv.c (release_bb_predicate): Remove the
19491         the assertion that the stmts have NULL use_ops.
19492         Discard the statements, asserting that they haven't
19493         yet been added to a BB.
19495 2018-03-08  Richard Biener  <rguenther@suse.de>
19497         PR tree-optimization/84746
19498         * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
19499         (phi_translate): Pass in destination ANTIC_OUT set.
19500         (phi_translate_1): Likewise.  For a simplified result lookup
19501         a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
19502         (phi_translate_set): Adjust.
19503         (do_pre_regular_insertion): Likewise.
19504         (do_pre_partial_partial_insertion): Likewise.
19506 2018-03-08  Martin Liska  <mliska@suse.cz>
19508         PR gcov-profile/84735
19509         * doc/gcov.texi: Document usage of profile files.
19510         * gcov-io.h: Document changes in the format.
19512 2018-03-08  Alexandre Oliva  <aoliva@redhat.com>
19514         PR debug/84404
19515         PR debug/84408
19516         * dwarf2out.c (struct dw_line_info_table): Update comments for
19517         view == -1.
19518         (FORCE_RESET_NEXT_VIEW): New.
19519         (FORCE_RESETTING_VIEW_P): New.
19520         (RESETTING_VIEW_P): Check for -1 too.
19521         (ZERO_VIEW_P): Likewise.
19522         (new_line_info_table): Force-reset next view.
19523         (dwarf2out_begin_function): Likewise.
19524         (dwarf2out_source_line): Simplify zero_view_p initialization.
19525         Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
19526         view directly.  Omit view when omitting .loc at line 0.
19528 2018-03-08  Jakub Jelinek  <jakub@redhat.com>
19530         PR tree-optimization/84740
19531         * tree-switch-conversion.c (process_switch): Call build_constructors
19532         only if info.phi_count is non-zero.
19534         PR tree-optimization/84739
19535         * tree-tailcall.c (find_tail_calls): Check call arguments against
19536         DECL_ARGUMENTS (current_function_decl) rather than
19537         DECL_ARGUMENTS (func) when checking for tail recursion.
19539 2018-03-07  Jakub Jelinek  <jakub@redhat.com>
19541         * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
19542         Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
19543         Volker Reichelt's entry and add entries for people that perform
19544         GCC fuzzy testing and report numerous bugs.
19546 2018-03-07  Segher Boessenkool  <segher@kernel.crashing.org>
19548         PR target/82411
19549         * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
19550         readonly data in sdata, if that is disabled.
19551         * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
19552         * doc/invoke.texi (RS/6000 and PowerPC Options): Document
19553         -mreadonly-in-sdata option.
19555 2018-03-07  Martin Sebor  <msebor@redhat.com>
19557         PR tree-optimization/84468
19558         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
19559         basic block when looking for nul assignment.
19561 2018-03-07  Eric Botcazou  <ebotcazou@adacore.com>
19563         PR target/84277
19564         * except.h (output_function_exception_table): Adjust prototype.
19565         * except.c (output_function_exception_table): Remove FNNAME parameter
19566         and add SECTION parameter.  Ouput one part of the table at a time.
19567         * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
19568         the first part of the exception table and emit unwind directives.
19569         * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
19570         (i386_pe_seh_cold_init): Likewise.
19571         * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
19572         (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
19573         * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
19574         (ix86_output_call_insn): Emit a nop in one more case for SEH.
19575         * config/i386/winnt.c: Include except.h.
19576         (struct seh_frame_state): Add reg_offset, after_prologue and
19577         in_cold_section fields.
19578         (i386_pe_seh_end_prologue): Set seh->after_prologue.
19579         (i386_pe_seh_cold_init): New function.
19580         (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
19581         to seh->in_cold_section.
19582         (seh_emit_push): Record the offset of the push.
19583         (seh_emit_save): Record the offet of the save.
19584         (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
19585         Test seh->after_prologue to disregard the epilogue.
19586         (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
19587         (i386_pe_end_cold_function): New function.
19589 2018-03-07  Jakub Jelinek  <jakub@redhat.com>
19591         PR fortran/84565
19592         * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
19593         aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
19595         PR c++/84704
19596         * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
19597         on tmp_var.
19598         * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
19599         don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
19601         PR middle-end/84723
19602         * multiple_target.c: Include tree-inline.h and intl.h.
19603         (expand_target_clones): Diagnose and fail if node->definition and
19604         !tree_versionable_function_p (node->decl).
19606 2018-03-06  John David Anglin  <danglin@gcc.gnu.org>
19608         * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
19609         sprint_ul.
19610         (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
19611         (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
19612         * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
19614 2018-03-06  Jakub Jelinek  <jakub@redhat.com>
19616         PR target/84710
19617         * combine.c (try_combine): Use reg_or_subregno instead of handling
19618         just paradoxical SUBREGs and REGs.
19620 2018-03-06  Claudiu Zissulescu  <claziss@synopsys.com>
19622         * config/arc/arc.c (arc_finalize_pic): Remove function.
19623         (arc_must_save_register): We use single base PIC register, remove
19624         checks to save/restore the PIC register.
19625         (arc_expand_prologue): Likewise.
19626         * config/arc/arc-protos.h (arc_set_default_type_attributes):
19627         Remove.
19628         (arc_verify_short): Likewise.
19629         (arc_attr_type): Likewise.
19630         * config/arc/arc.c (arc_set_default_type_attributes): Remove.
19631         (walk_stores): Likewise.
19632         (arc_address_cost): Make it static.
19633         (arc_verify_short): Likewise.
19634         (branch_dest): Likewise.
19635         (arc_attr_type): Likewise.
19636         * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
19637         (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
19638         (arc_final_prescan_insn): Remove inserting the nops due to
19639         hardware hazards.  It is done in reorg step.
19640         (insn_length_variant_t): Remove.
19641         (insn_length_parameters_t): Likewise.
19642         (arc_insn_length_parameters): Likewise.
19643         (arc_get_insn_variants): Likewise.
19644         * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
19646 2018-03-06  Jakub Jelinek  <jakub@redhat.com>
19648         PR inline-asm/84683
19649         * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
19650         assertion failure.
19652         PR tree-optimization/84687
19653         * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
19654         on new_node->decl.
19655         * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
19657 2018-03-05  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
19659         * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
19660         Rename to ppc_speculation_barrier.
19661         * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
19662         __builtin_ppc_speculation_barrier.
19664 2018-03-05  Jakub Jelinek  <jakub@redhat.com>
19666         PR target/84700
19667         * combine.c (combine_simplify_rtx): Don't try to simplify if
19668         if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
19669         are equal to x.
19671 2018-03-05  Segher Boessenkool  <segher@kernel.crashing.org>
19673         * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
19674         to 32 bytes when compiling for POWER9.
19676 2018-03-05  Jakub Jelinek  <jakub@redhat.com>
19678         PR target/84564
19679         * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
19680         regparm >= 3 with no arg reg available also for calls with
19681         flag_force_indirect_call.  Pass decl to ix86_function_regparm.
19683         PR target/84524
19684         * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
19685         orig,vex.
19686         (*<plusminus_insn><mode>3): Likewise.  Remove <mask_operand3> uses.
19688 2018-03-05  Peter Bergner  <bergner@vnet.ibm.com>
19690         PR target/84264
19691         * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
19693 2018-03-05  Richard Biener  <rguenther@suse.de>
19695         PR tree-optimization/84486
19696         * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
19697         When inserting a __builtin_assume_aligned call set the LHS
19698         SSA name alignment info accordingly.
19700 2018-03-05  Wilco Dijkstra  <wdijkstr@arm.com>
19702         PR tree-optimization/84114
19703         * config/aarch64/aarch64.c (aarch64_reassociation_width)
19704         Avoid reassociation of FLOAT_MODE addition.
19706 2018-03-05  Olga Makhotina  <olga.makhotina@intel.com>
19708         * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
19709         OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
19710         OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
19711         (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
19712         * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
19713         * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
19714         * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
19715         and -mwbnoinvd.
19716         * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
19717         __builtin_ia32_wbinvd): New builtins.
19718         (SPECIAL_ARGS2): New.
19719         * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
19720         (SPECIAL_ARGS2): New.
19721         * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
19722         (ix86_valid_target_attribute_inner_p): Ditto.
19723         (ix86_init_mmx_sse_builtins): Add special_args2.
19724         * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
19725         TARGET_WBNOINVD_P): New.
19726         * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
19727         (define_insn "wbinvd", define_insn "wbnoinvd"): New.
19728         * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
19729         * config/i386/immintrin.h (_wbinvd): New intrinsic.
19730         * config/i386/pconfigintrin.h: New file.
19731         * config/i386/wbnoinvdintrin.h: Ditto.
19732         * config/i386/x86intrin.h: Add headers pconfigintrin.h and
19733         wbnoinvdintrin.h.
19734         * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
19736 2018-03-05  Richard Biener  <rguenther@suse.de>
19738         PR tree-optimization/84670
19739         * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
19740         member.
19741         (BB_VISITED_WITH_VISITED_SUCCS): New define.
19742         (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
19743         (compute_antic_aux): Only assert the number of values in ANTIC_IN
19744         doesn't grow if all successors (recursively) were visited at least
19745         once.
19747 2018-03-05  Richard Biener  <rguenther@suse.de>
19749         PR tree-optimization/84650
19750         * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
19751         if executed in the loop pipeline.
19753 2018-03-05  Sandra Loosemore  <sandra@codesourcery.com>
19755         * doc/configfiles.texi (Configuration Files): Move info about
19756         conditionalizing $target-protos.h to...
19757         * doc/sourcebuild.texi (Back End): Here.  Explain how $target.h
19758         differs from $target-protos.h.
19760 2018-03-05  Kito Cheng  <kito.cheng@gmail.com>
19761             Chung-Ju Wu  <jasonwucj@gmail.com>
19763         * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
19764         * config/nds32/nds32-multiple.md (setmemsi): Define.
19765         * config/nds32/nds32-memory-manipulation.c
19766         (nds32_gen_dup_4_byte_to_word_value): New.
19767         (emit_setmem_word_loop): New.
19768         (emit_setmem_byte_loop): New.
19769         (nds32_expand_setmem_loop): New.
19770         (nds32_expand_setmem_loop_v3m): New.
19771         (nds32_expand_setmem_unroll): New.
19772         (nds32_expand_setmem): New.
19774 2018-03-04  Kito Cheng  <kito.cheng@gmail.com>
19775             Chung-Ju Wu  <jasonwucj@gmail.com>
19777         * config/nds32/nds32-memory-manipulation.c
19778         (nds32_emit_load_store): New.
19779         (nds32_emit_post_inc_load_store): New.
19780         (nds32_emit_mem_move): New.
19781         (nds32_emit_mem_move_block): New.
19782         (nds32_expand_movmemsi_loop_unknown_size): New.
19783         (nds32_expand_movmemsi_loop_known_size): New.
19784         (nds32_expand_movmemsi_loop): New.
19785         (nds32_expand_movmemsi_unroll): New.
19786         (nds32_expand_movmemqi): Rename ...
19787         (nds32_expand_movmemsi): ... to this.
19788         * config/nds32/nds32-multiple.md (movmemqi): Rename ...
19789         (movmemsi): ... to this.
19790         * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
19791         (nds32_expand_movmemsi): ... to this.
19793 2018-03-04  Kito Cheng  <kito.cheng@gmail.com>
19794             Monk Chiang  <sh.chiang04@gmail.com>
19795             Chung-Ju Wu  <jasonwucj@gmail.com>
19797         * config/nds32/nds32-protos.h
19798         (nds32_expand_load_multiple): New arguments.
19799         (nds32_expand_store_multiple): Ditto.
19800         (nds32_valid_multiple_load_store): Rename ...
19801         (nds32_valid_multiple_load_store_p): ... to this.
19802         * config/nds32/nds32-memory-manipulation.c
19803         (nds32_expand_load_multiple): Refine implementation.
19804         (nds32_expand_store_multiple): Ditto.
19805         * config/nds32/nds32-multiple.md
19806         (load_multiple): Update nds32_expand_load_multiple interface.
19807         (store_multiple): Update nds32_expand_store_multiple interface.
19808         * config/nds32/nds32-predicates.c
19809         (nds32_valid_multiple_load_store): Rename ...
19810         (nds32_valid_multiple_load_store_p): ... to this and refine
19811         implementation.
19812         * config/nds32/predicates.md
19813         (nds32_load_multiple_and_update_address_operation): New predicate.
19814         (nds32_store_multiple_and_update_address_operation): New predicate.
19816 2018-03-04  Kito Cheng  <kito.cheng@gmail.com>
19817             Chung-Ju Wu  <jasonwucj@gmail.com>
19819         * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
19820         (combo): New attribute.
19821         * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
19823 2018-03-03  Chung-Ju Wu  <jasonwucj@gmail.com>
19825         * config/nds32/nds32.opt: Change -mcmodel= default value.
19827 2018-03-03  Kito Cheng  <kito.cheng@gmail.com>
19828             Monk Chiang  <sh.chiang04@gmail.com>
19829             Chung-Ju Wu  <jasonwucj@gmail.com>
19831         * config/nds32/constants.md (unspec_element): New enum.
19832         * config/nds32/constraints.md (Umw): New constraint.
19833         * config/nds32/nds32-intrinsic.c: Add more builtin functions.
19834         * config/nds32/nds32-intrinsic.md: Likewise.
19835         * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
19836         (nds32_valid_smw_lwm_base_p): New.
19837         (nds32_output_smw_single_word): New.
19838         (nds32_output_lmw_single_word): New.
19839         (nds32_expand_unaligned_load): New.
19840         (nds32_expand_unaligned_store): New.
19841         * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
19842         (nds32_output_smw_single_word): Declare.
19843         (nds32_output_lmw_single_word): Declare.
19844         (nds32_expand_unaligned_load): Declare.
19845         (nds32_expand_unaligned_store): Declare.
19846         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
19847         NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
19848         NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
19849         NDS32_BUILTIN_UASTORE_DW.
19850         * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
19851         predicate.
19853 2018-03-03  Monk Chiang  <sh.chiang04@gmail.com>
19854             Kito Cheng  <kito.cheng@gmail.com>
19855             Chung-Ju Wu  <jasonwucj@gmail.com>
19857         * config/nds32/nds32-intrinsic.c
19858         (nds32_expand_builtin_null_ftype_reg): Delete.
19859         (nds32_expand_builtin_reg_ftype_imm): Ditto.
19860         (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
19861         (nds32_read_argument): New.
19862         (nds32_legitimize_target): Ditto.
19863         (nds32_legitimize_argument): Ditto.
19864         (nds32_check_constant_argument): Ditto.
19865         (nds32_expand_unop_builtin): Ditto.
19866         (nds32_expand_unopimm_builtin): Ditto.
19867         (nds32_expand_binop_builtin): Ditto.
19868         (nds32_builtin_decl_impl): Ditto.
19869         (builtin_description): Ditto.
19870         (nds32_expand_builtin_impl): Rewrite with new infrastructure.
19871         (nds32_init_builtins_impl): Ditto.
19872         * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
19873         (nds32_builtin_decl): New.
19874         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
19875         * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
19877 2018-03-02  Jeff Law  <law@redhat.com>
19879         * reorg.c (stop_search_p): Handle DEBUG_INSN.
19880         (redundant_insn, fill_simple_delay_slots): Likewise.
19881         (fill_slots_from_thread): Likewise.
19882         * resource.c (mark_referenced_resources): Likewise.
19883         (mark_set_resources, find_dead_or_set_registers): Likewise.
19885 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
19887         * substring-locations.h (format_warning_va): Formatting fix for
19888         ATTRIBUTE_GCC_DIAG.
19889         (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
19890         argument.
19891         (format_warning_n_va, format_warning_at_substring_n): New prototypes.
19892         * substring-locations.c: Include intl.h.
19893         (format_warning_va): Turned into small wrapper around
19894         format_warning_n_va, renamed to ...
19895         (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
19896         rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
19897         use ngettext.
19898         (format_warning_at_substring_n): New function.
19899         * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
19900         (fmtwarn): Add ATTRIBUTE_GCC_DIAG.  Turn into a copy of
19901         format_warning_at_substring with just a shorter name instead of
19902         const function pointer.
19903         (fmtwarn_n): New function.
19904         (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
19905         appropriate, get rid of all the fmtstr temporaries, move conditionals
19906         with G_() wrapped string literals directly into fmtwarn arguments,
19907         cast dir.len to (int), formatting fixes.
19909 2018-03-02  Thomas Schwinge  <thomas@codesourcery.com>
19911         * doc/invoke.texi: Remove "Cilk Plus" references.
19913 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
19914             Richard Biener  <rguenther@suse.de>
19916         PR ipa/84628
19917         * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
19918         for error or warning attributes if CALL_FROM_THUNK_P is set.
19919         Formatting fixes.
19921 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
19923         PR target/56540
19924         * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
19925         __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
19927         PR target/56540
19928         * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
19929         __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
19931         * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
19932         instead of -1U in last predictors element's probability member.
19934 2018-03-02  Eric Botcazou  <ebotcazou@adacore.com>
19936         PR ipa/83983
19937         * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
19938         arguments if they are comparable.
19940 2018-03-02  Richard Sandiford  <richard.sandiford@linaro.org>
19942         PR tree-optimization/84634
19943         * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
19944         masks and masked_loop_p with a single loop_masks, making sure it's
19945         null for bb vectorization.
19947 2018-03-02  Richard Sandiford  <richard.sandiford@linaro.org>
19949         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
19950         (vect_analyze_data_ref_access): Use loop->safe_len rather than
19951         loop->force_vectorize to check whether there is no alias.
19953 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
19955         PR target/84614
19956         * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
19957         prototypes.
19958         * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
19959         comments.
19960         (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
19961         * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
19962         instead of a loop around prev_real_insn.
19963         * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
19964         prev_real_insn.
19966         PR inline-asm/84625
19967         * config/i386/i386.c (ix86_print_operand): Use conditional
19968         output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
19969         zero vector.
19971 2018-03-02  Richard Biener  <rguenther@suse.de>
19973         PR tree-optimization/84427
19974         * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
19975         (bitmap_set_subtract_values): Rewrite to handle multiple
19976         exprs per value.
19977         (clean): Likewise.
19978         (prune_clobbered_mems): Likewise.
19979         (phi_translate): Take edge instead of pred/phiblock.
19980         (phi_translate_1): Likewise.
19981         (phi_translate_set): Likewise.  Insert all translated
19982         exprs for a value into the set, keeping possibly multiple
19983         expressions per value.
19984         (compute_antic_aux): Adjust for phi_translate changes.
19985         When intersecting union the expressions and prune those
19986         not in the final value set, keeping possibly multiple
19987         expressions per value.  Do not use value-insertion
19988         for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
19989         all expressions.  Add verification that the value-sets
19990         only shrink during iteration.
19991         (compute_partial_antic_aux): Adjust for the phi_translate changes.
19992         (do_pre_regular_insertion): Likewise.
19993         (do_pre_partial_partial_insertion): Likewise.
19995 2018-03-02  Richard Biener  <rguenther@suse.de>
19997         PR target/82005
19998         * config/darwin.c (saved_debug_info_level): New static global.
19999         (darwin_asm_lto_start): Disable debug info generation for LTO out.
20000         (darwin_asm_lto_end): Restore debug info generation settings.
20002 2018-03-01  Martin Liska  <mliska@suse.cz>
20004         PR sanitizer/82484
20005         * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
20006         volatile arguments.
20008 2018-03-01  Richard Biener  <rguenther@suse.de>
20010         PR debug/84645
20011         * dwarf2out.c (gen_variable_die): Properly handle late VLA
20012         type annotation with LTO when debug was disabled at compile-time.
20014 2018-03-01  Matthew Fortune  <mfortune@gmail.com>
20016         * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
20017         XINT with INTVAL.
20018         (mips_final_postscan_insn): Likewise.
20020 2018-03-01  Richard Sandiford  <richard.sandiford@linaro.org>
20022         PR rtl-optimization/84528
20023         * alias.c (init_alias_target): Add commentary.
20024         (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
20025         a unique base value if the frame pointer is not eliminated
20026         to the stack pointer.
20028 2018-03-01  Tom de Vries  <tom@codesourcery.com>
20030         PR rtl-optimization/83327
20031         * lra-int.h (hard_regs_spilled_into): Declare.
20032         * lra.c (hard_regs_spilled_into): Define.
20033         (init_reg_info): Init hard_regs_spilled_into.
20034         * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
20035         * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
20036         (process_bb_lives): Handle hard_regs_spilled_into.
20037         (lra_create_live_ranges_1): Before doing liveness propagation, clear
20038         regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
20040 2018-02-28  David Edelsohn  <dje.gcc@gmail.com>
20042         * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
20043         (powerpc-ibm-aix[789]*): Default to AIX 7.2.
20044         * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
20045         * config/rs6000/aix72.h: New file.
20047 2018-02-28  Jakub Jelinek  <jakub@redhat.com>
20049         * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
20050         instead of warning_at with conditional singular and plural messages
20051         where possible.
20053         PR target/52991
20054         * stor-layout.c (update_alignment_for_field): For
20055         targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
20056         && !DECL_PACKED (field), do the alignment update, just use
20057         only desired_align instead of MAX (type_align, desired_align)
20058         as the alignment.
20059         (place_field): Don't do known_align < desired_align handling
20060         early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
20061         is non-NULL, instead do it after rli->prev_field handling and
20062         only if not within a bitfield word.  For DECL_PACKED (field)
20063         use type_align of BITS_PER_UNIT.
20065 2018-02-28  Eric Botcazou  <ebotcazou@adacore.com>
20067         * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
20068         superfluous parentheses and trailing spaces.
20070 2018-02-28  Richard Biener  <rguenther@suse.de>
20072         PR tree-optimization/84584
20073         * graphite-scop-detection.c (scop_detection::add_scop): Discard
20074         SCoPs with fake exit edge.
20076 2018-02-28  Martin Liska  <mliska@suse.cz>
20078         PR testsuite/84597
20079         * timevar.c (timer::print): Fix format to properly print 100%
20080         values.
20082 2018-02-28  Richard Biener  <rguenther@suse.de>
20084         PR middle-end/84607
20085         * genmatch.c (capture_info::walk_match): Do not mark
20086         captured expressions without operands as expr_p given
20087         they act more like predicates and should be subject to
20088         "lost tail" side-effect preserving.
20090 2018-02-28  Alexandre Oliva  <aoliva@redhat.com>
20092         PR rtl-optimization/81611
20093         * auto-inc-dec.c (attempt_change): Move dead note from
20094         mem_insn if it's the next use of regno
20095         (find_address): Take address use of reg holding
20096         non-incremented value.  Add parm to limit search to the named
20097         reg only.
20098         (merge_in_block): Attempt to use a mem insn that is the next
20099         use of the original regno.
20101 2018-02-27  Martin Sebor  <msebor@redhat.com>
20103         PR c++/83871
20104         * doc/invoke.texi (-Wmissing-attributes): New option.
20105         * print-tree.c (print_node): Handle DECL_UNINLINABLE.
20107 2018-02-27  Martin Sebor  <msebor@redhat.com>
20109         PR translation/84207
20110         * diagnostic-core.h (warning_n, error_n, inform_n): Change
20111         n argument to unsigned HOST_WIDE_INT.
20112         * diagnostic.c (warning_n, error_n, inform_n): Ditto.
20113         (diagnostic_n_impl): Ditto.  Handle arguments in excess of LONG_MAX.
20114         * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
20115         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
20117 2018-02-27  Richard Biener  <rguenther@suse.de>
20119         PR tree-optimization/84512
20120         * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
20121         Do not use the estimate returned from record_stmt_cost for
20122         the scalar iteration cost but sum properly using add_stmt_cost.
20124 2018-02-27  Richard Biener  <rguenther@suse.de>
20126         PR tree-optimization/84466
20127         * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
20128         Adjust last change to less strictly validate use operands.
20130 2018-02-27  Martin Liska  <mliska@suse.cz>
20132         PR gcov-profile/84548
20133         * gcov.c (process_file): Allow partial overlap and consider it
20134         also as group functions.
20135         (output_lines): Properly calculate range of lines for a group.
20137 2018-02-27  Martin Liska  <mliska@suse.cz>
20139         * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
20140         'ggc' suffixes.  Change first column width.
20141         (timer::print): Fix formatting of the column.
20143 2018-02-27  Alexandre Oliva  <aoliva@redhat.com>
20145         * tree-ssa-live.c (remove_unused_scope_block_p): Do not
20146         preserve inline entry blocks for the sake of debug inline
20147         entry point markers alone.
20148         (remove_unused_locals): Suggest in comments a better place to
20149         force the preservation of inline entry blocks that are
20150         otherwise unused, but do not preserve them.
20152 2018-02-26  H.J. Lu  <hongjiu.lu@intel.com>
20154         * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
20156 2018-02-26  H.J. Lu  <hongjiu.lu@intel.com>
20158         PR target/84039
20159         * config/i386/constraints.md (Bs): Replace
20160         ix86_indirect_branch_register with
20161         TARGET_INDIRECT_BRANCH_REGISTER.
20162         (Bw): Likewise.
20163         * config/i386/i386.md (indirect_jump): Likewise.
20164         (tablejump): Likewise.
20165         (*sibcall_memory): Likewise.
20166         (*sibcall_value_memory): Likewise.
20167         Peepholes of indirect call and jump via memory: Likewise.
20168         (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
20169         (*sibcall_value_GOT_32): Likewise.
20170         * config/i386/predicates.md (indirect_branch_operand): Likewise.
20171         (GOT_memory_operand): Likewise.
20172         (call_insn_operand): Likewise.
20173         (sibcall_insn_operand): Likewise.
20174         (GOT32_symbol_operand): Likewise.
20175         * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
20177 2018-02-26  Eric Botcazou  <ebotcazou@adacore.com>
20179         PR rtl-optimization/83496
20180         * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
20181         booleans to RTXes.  Call fix_reg_dead_note on every non-null element.
20182         (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
20183         redundant insn, if any.
20184         (relax_delay_slots): Likewise.
20185         (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
20187 2018-02-26  Richard Sandiford  <richard.sandiford@linaro.org>
20189         PR tree-optimization/83965
20190         * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
20191         that grouped statements are part of a reduction chain.  Return
20192         true if the statement is not marked as a reduction itself but
20193         is part of a group.
20194         (vect_recog_dot_prod_pattern): Don't check whether the statement
20195         is part of a group here.
20196         (vect_recog_sad_pattern): Likewise.
20197         (vect_recog_widen_sum_pattern): Likewise.
20199 2018-02-26  Eric Botcazou  <ebotcazou@adacore.com>
20201         PR debug/84545
20202         * final.c (rest_of_clean_state): Also look for calls inside sequences.
20204 2018-02-26  H.J. Lu  <hongjiu.lu@intel.com>
20206         PR target/84530
20207         * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
20208         the bool argument.
20209         (ix86_output_indirect_function_return): New prototype.
20210         (ix86_split_simple_return_pop_internal): Likewise.
20211         * config/i386/i386.c (indirect_return_via_cx): New.
20212         (indirect_return_via_cx_bnd): Likewise.
20213         (indirect_thunk_name): Handle return va CX_REG.
20214         (output_indirect_thunk_function): Create alias for
20215         __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
20216         (ix86_output_indirect_jmp): Remove the bool argument.
20217         (ix86_output_indirect_function_return): New function.
20218         (ix86_split_simple_return_pop_internal): Likewise.
20219         * config/i386/i386.md (*indirect_jump): Don't pass false
20220         to ix86_output_indirect_jmp.
20221         (*tablejump_1): Likewise.
20222         (simple_return_pop_internal): Change it to define_insn_and_split.
20223         Call ix86_split_simple_return_pop_internal to split it for
20224         -mfunction-return=.
20225         (simple_return_indirect_internal): Call
20226         ix86_output_indirect_function_return instead of
20227         ix86_output_indirect_jmp.
20229 2018-02-26  Jakub Jelinek  <jakub@redhat.com>
20231         PR bootstrap/84405
20232         * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
20233         memset and value initialization afterwards.
20235 2018-02-26  Christophe Lyon  <christophe.lyon@linaro.org>
20237         * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
20239 2018-02-26  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
20241         PR target/84521
20242         * common/config/aarch64/aarch64-common.c
20243         (aarch_option_optimization_table[]): Switch
20244         off fomit-frame-pointer
20246 2018-02-26  Kito Cheng  <kito.cheng@gmail.com>
20247             Chung-Ju Wu  <jasonwucj@gmail.com>
20249         * config/nds32/nds32-multiple.md (load_multiple): Disallow
20250         volatile memory.
20251         (store_multiple): Ditto.
20253 2018-02-26  Kito Cheng  <kito.cheng@gmail.com>
20255         * config.gcc: Add --with-cpu support for nds32 target.
20256         * config/nds32/nds32-opts.h (nds32_cpu_type): New.
20257         * config/nds32/nds32.opt: Add -mcpu= option.
20259 2018-02-25  Segher Boessenkool  <segher@kernel.crashing.org>
20261         * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
20262         isel=yes): Warn for these deprecated options.
20264 2018-02-23  David Edelsohn  <dje.gcc@gmail.com>
20266         * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
20267         ISA_2_5_MASKS_EMBEDDED.
20269 2018-02-23  Jakub Jelinek  <jakub@redhat.com>
20271         * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
20272         p->max as pointers rather than using iterative_hash_expr.
20274 2018-02-23  Carl Love  <cel@us.ibm.com>
20276         * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
20277         macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
20278         BU_P8V_OVERLOAD_2.
20279         * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
20280         P8V_BUILTIN_VEC_VSIGNED2.  Change VSX_BUILTIN_VEC_VUNSIGNED2 to
20281         P8V_BUILTIN_VEC_VUNSIGNED2.
20283 2018-02-22  Vladimir Makarov  <vmakarov@redhat.com>
20285         PR target/81572
20286         * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
20287         * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
20288         LRA_UNKNOWN_ALT.
20289         * lra-constraints.c (curr_insn_transform): Set up
20290         LRA_NON_CLOBBERED_ALT for moves processed on the fast path.  Use
20291         LRA_UNKNOWN_ALT.
20292         (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
20293         * lra-eliminations.c (spill_pseudos): Ditto.
20294         (process_insn_for_elimination): Ditto.
20295         * lra-lives.c (reg_early_clobber_p): Use the new macros.
20296         * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
20297         LRA_NON_CLOBBERED_ALT.
20299 2018-02-22  Martin Sebor  <msebor@redhat.com>
20301         PR tree-optimization/84480
20302         * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
20303         to maybe_diag_stxncpy_trunc.  Call it.
20304         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
20305         from gimple_fold_builtin_strcpy.  Print inlining stack.
20306         (handle_builtin_stxncpy): Print inlining stack.
20307         * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
20309 2018-02-22  H.J. Lu  <hongjiu.lu@intel.com>
20311         PR target/84176
20312         * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
20313         error when -mindirect-branch=thunk-extern, -fcf-protection=branch
20314         and -fcheck-pointer-bounds are used together.
20315         (indirect_thunk_prefix): New enum.
20316         (indirect_thunk_need_prefix): New function.
20317         (indirect_thunk_name): Replace need_bnd_p with need_prefix.  Use
20318         "_nt" instead of "_bnd" for NOTRACK prefix.
20319         (output_indirect_thunk): Replace need_bnd_p with need_prefix.
20320         (output_indirect_thunk_function): Likewise.
20321         (): Likewise.
20322         (ix86_code_end): Update output_indirect_thunk_function calls.
20323         (ix86_output_indirect_branch_via_reg): Replace
20324         ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
20325         (ix86_output_indirect_branch_via_push): Likewise.
20326         (ix86_output_function_return): Likewise.
20327         * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
20328         incompatible with -fcf-protection=branch and
20329         -fcheck-pointer-bounds.
20331 2018-02-22  Steve Ellcey  <sellcey@cavium.com>
20333         PR target/83335
20334         * config/aarch64/aarch64.c (aarch64_print_address_internal):
20335         Change gcc_assert call to output_operand_lossage.
20337 2018-02-22  Steve Ellcey  <sellcey@cavium.com>
20339         * doc/extend.texi (__builtin_extend_pointer): Document builtin.
20341 2018-02-22  DJ Delorie  <dj@redhat.com>
20342             Sebastian Perta  <sebastian.perta@renesas.com>
20343             Oleg Endo  <olegendo@gcc.gnu.org>
20345         * config/rx/rx.c (rx_rtx_costs): New function.
20346         (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
20348 2018-02-22  Thomas Preud'homme  <thomas.preudhomme@arm.com>
20350         * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
20352 2018-02-22  Martin Liska  <mliska@suse.cz>
20354         PR driver/83193
20355         * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
20356         Add "native" as a possible value.
20358 2018-02-22  Martin Liska  <mliska@suse.cz>
20360         PR driver/83193
20361         * config/i386/i386.c (ix86_option_override_internal):
20362         Add "native" as a possible value for -march and -mtune.
20364 2018-02-22  Jakub Jelinek  <jakub@redhat.com>
20366         PR target/84502
20367         * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
20368         to all type variants.
20370         PR tree-optimization/84503
20371         * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
20372         width as info->bitpos + info->bitsize - start.
20373         (merged_store_group::merge_overlapping): Simplify width computation.
20374         (check_no_overlap): New function.
20375         (imm_store_chain_info::try_coalesce_bswap): Compute expected
20376         start + width and last_order of the group, fail if check_no_overlap
20377         fails.
20378         (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
20379         to group if check_no_overlap fails.
20381 2018-02-21  Segher Boessenkool  <segher@kernel.crashing.org>
20383         * config/rs6000/altivec.md: Delete contraint arguments to
20384         define_expand, define_split, and define_peephole2, and in
20385         define_insn_and_split if always unused.
20386         * config/rs6000/darwin.md: Ditto.
20387         * config/rs6000/dfp.md: Ditto.
20388         * config/rs6000/rs6000.md: Ditto.
20389         * config/rs6000/sync.md: Ditto.
20390         * config/rs6000/vector.md: Ditto.
20391         * config/rs6000/vsx.md: Ditto.
20393 2018-02-21  Segher Boessenkool  <segher@kernel.crashing.org>
20395         * config/rs6000/altivec.md: Write output control strings as braced
20396         blocks instead of double-quoted strings.
20397         * config/rs6000/darwin.md: Ditto.
20398         * config/rs6000/rs6000.md: Ditto.
20399         * config/rs6000/vector.md: Ditto.
20400         * config/rs6000/vsx.md: Ditto.
20402 2018-02-21  Jason Merrill  <jason@redhat.com>
20404         PR c++/84314 - ICE with templates and fastcall attribute.
20405         * attribs.c (build_type_attribute_qual_variant): Remove assert.
20407 2018-02-21  Jan Hubicka  <hubicka@ucw.cz>
20409         * ipa-cp.c (determine_versionability): Fix comment typos.
20411 2018-02-21  Jan Hubicka  <hubicka@ucw.cz>
20413         PR c/84229
20414         * ipa-cp.c (determine_versionability): Do not version functions caling
20415         va_arg_pack.
20417 2018-02-21  Martin Liska  <mliska@suse.cz>
20419         PR driver/83193
20420         * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
20421         Add "native" as a possible value.
20422         * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT):  Define
20423         the macro when native cpu detection is available.
20425 2018-02-21  Martin Liska  <mliska@suse.cz>
20427         PR driver/83193
20428         * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
20429         Add "native" as a possible value.
20430         * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
20431         when native cpu detection is available.
20433 2018-02-21  Jakub Jelinek  <jakub@redhat.com>
20434             Martin Sebor  <msebor@redhat.com>
20436         PR tree-optimization/84478
20437         * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
20438         false.
20439         * gimple-fold.c (get_range_strlen): Make minlen const and assume it
20440         can't be NULL.  Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
20441         support which is conservatively correct, for 2 only stay conservative
20442         for maxlen.  Formatting and comment capitalization fixes.  Add STRICT
20443         argument to the 2 argument get_range_strlen, adjust 6 arg
20444         get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
20445         false.
20446         (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
20447         (gimple_fold_builtin_strlen): Pass true as last argument to
20448         get_range_strlen.
20450 2018-02-20  Martin Sebor  <msebor@redhat.com>
20452         PR middle-end/84095
20453         * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
20454         (builtin_memref::set_base_and_offset): Same.  Handle inner references.
20455         (builtin_memref::builtin_memref): Factor out parts into
20456         set_base_and_offset and call it.
20458 2018-02-20  Richard Sandiford  <richard.sandiford@linaro.org>
20460         PR middle-end/84406
20461         * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
20462         is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
20463         greater precision.  If to_mode is a MODE_PARTIAL_INT, stop the
20464         search at the associated MODE_INT.
20466 2018-02-20  Jeff Law  <law@redhat.com>
20468         PR middle-end/82123
20469         PR tree-optimization/81592
20470         PR middle-end/79257
20471         * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
20472         for range data rather than using global data.
20473         * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
20474         range data rather than using global data.
20475         * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
20476         pass it to children as needed.
20477         (struct directive::fmtresult): Similarly.
20478         (struct directive::set_width): Similarly.
20479         (struct directive::set_precision): Similarly.
20480         (format_integer, format_directive, parse_directive): Similarly.
20481         (format_none): Accept unnamed vr_values parameter.
20482         (format_percent, format_floating, format_character): Similarly.
20483         (format_string, format_plain): Similarly.
20484         * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
20485         the EVRP range analyzer for range data rather than using global data.
20486         * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
20487         gimple-ssa-evrp-analyze.h
20488         (class sprintf_dom_walker): Add after_dom_children member function.
20489         Add evrp_range_analyzer member.
20490         (sprintf_dom_walker::before_dom_children): Call into the EVRP
20491         range analyzer as needed.
20492         (sprintf_dom_walker::after_dom_children): New member function.
20493         * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
20494         if not optimizing.
20495         (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
20496         (evrp_range_analyzer::pop_to_marker): Likewise.
20498 2018-02-20  Richard Sandiford  <richard.sandiford@linaro.org>
20500         PR tree-optimization/84419
20501         * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
20502         with the required type if its current type is compatible but
20503         different.
20505 2018-02-20  Jakub Jelinek  <jakub@redhat.com>
20507         PR middle-end/82004
20508         * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
20509         after vectorization.
20511 2018-02-20  Martin Liska  <mliska@suse.cz>
20513         PR driver/83193
20514         * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
20515         possible values if we don't have a hint.
20517 2018-02-20  Martin Liska  <mliska@suse.cz>
20519         PR c/84310
20520         PR target/79747
20521         * final.c (shorten_branches): Build align_tab array with one
20522         more element.
20523         * opts.c (finish_options): Add alignment option limit check.
20524         (MAX_CODE_ALIGN): Likewise.
20525         (MAX_CODE_ALIGN_VALUE): Likewise.
20526         * doc/invoke.texi: Document maximum allowed option value for
20527         all -falign-* options.
20529 2018-02-19  Jakub Jelinek  <jakub@redhat.com>
20531         PR target/84146
20532         * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
20533         * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
20534         * var-tracking.c (emit_note_insn_var_location): Remove all references
20535         to NOTE_INSN_CALL_ARG_LOCATION.
20536         (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
20537         the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
20538         Use copy_rtx_if_shared.
20539         * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
20540         NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
20541         (dwarf2out_var_location): Remove handling of
20542         NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
20543         on call_insn.
20544         * final.c (final_scan_insn): Remove all references to
20545         NOTE_INSN_CALL_ARG_LOCATION.
20546         (rest_of_clean_state): Likewise.  Remove REG_CALL_ARG_LOCATION notes
20547         before dumping final insns.
20548         * except.c (emit_note_eh_region_end): Remove all references to
20549         NOTE_INSN_CALL_ARG_LOCATION.
20550         * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
20551         * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
20552         * config/arc/arc.c (hwloop_optimize): Likewise.
20553         * config/arm/arm.c (create_fix_barrier): Likewise.
20554         * config/s390/s390.c (s390_chunkify_start): Likewise.
20555         * config/sh/sh.c (find_barrier): Likewise.
20556         * config/i386/i386.c (rest_of_insert_endbranch,
20557         ix86_seh_fixup_eh_fallthru): Likewise.
20558         * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
20559         * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
20560         * config/frv/frv.c (frv_function_prologue): Likewise.
20561         * emit-rtl.c (try_split): Likewise.  Copy over REG_CALL_ARG_LOCATION
20562         reg note.
20563         (note_outside_basic_block_p): Remove all references to
20564         NOTE_INSN_CALL_ARG_LOCATION.
20565         * gengtype.c (adjust_field_rtx_def): Likewise.
20566         * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
20567         Likewise.
20568         * jump.c (cleanup_barriers, delete_related_insns): Likewise.
20569         * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
20571         PR c++/84444
20572         * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
20573         is ADDR_EXPR.
20575         PR tree-optimization/84452
20576         * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
20577         expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
20578         is NULL.
20580 2018-02-19  Martin Liska  <mliska@suse.cz>
20582         PR sanitizer/82183
20583         * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
20585 2018-02-19  Martin Liska  <mliska@suse.cz>
20586             Richard Sandiford  <richard.sandiford@linaro.org>
20588         PR tree-optimization/82491
20589         * gimple-fold.c (get_base_constructor): Make earlier bail out
20590         to prevent ubsan.
20592 2018-02-19  Carl Love  <cel@us.ibm.com>
20594         * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
20595         BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
20596         BU_P8V_OVERLOAD_1.
20597         * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
20598         P8V_BUILTIN_VEC_NEG.
20600 2018-02-19  Sebastian Perta  <sebastian.perta@renesas.com>
20602         * config/rl78/rl78.md (movdf): New define expand.
20604 2018-02-19  Martin Liska  <mliska@suse.cz>
20606         PR other/80589
20607         * doc/invoke.texi: Fix typo.
20608         * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
20610 2018-02-18  Segher Boessenkool  <segher@kernel.crashing.org>
20612         * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
20613         handle rs6000_single_float and rs6000_double_float specially for
20614         e500 family CPUs.
20616 2018-02-16  Jeff Law  <law@redhat.com>
20618         * config/rx/rx.c (add_pop_cfi_notes): New function.;
20619         (pop_regs): Use it.
20621 2018-02-16  Jakub Jelinek  <jakub@redhat.com>
20623         PR ipa/84425
20624         * ipa-inline.c (inline_small_functions): Fix a typo.
20626 2018-02-16  Nathan Sidwell  <nathan@acm.org>
20628         * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
20630 2018-02-16  Carl Love  <cel@us.ibm.com>
20632         * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
20633         Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
20634         from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
20635         * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
20636         expansion to P8V_BUILTIN_VEC_FLOAT2.
20638 2018-02-16  Vladimir Makarov  <vmakarov@redhat.com>
20640         PR rtl-optimization/70023
20641         * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
20642         src_regno into account.
20644 2018-02-16  Carl Love  <cel@us.ibm.com>
20646         * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
20647         * config/rs6000/rs6000-builtin.def: Remove macro expansion for
20648         VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
20649         * config/rs6000/rs6000.c: Remove case statements for
20650         P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
20651         P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
20652         and P9V_BUILTIN_VEC_VINSERT4B.
20653         * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
20654         P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
20655         * config/rs6000/vsx.md:
20656         * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
20657         vec_insert4b.
20659 2018-02-16  Carl Love  <cel@us.ibm.com>
20661         * config/rs6000/altivec.h: Add builtin names vec_extract4b
20662         vec_insert4b.
20663         * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
20664         definitions.
20665         * config/rs6000/rs6000-c.c: Add the definitions for
20666         P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
20667         * config/rs6000/rs6000.c (altivec_expand_builtin): Add
20668         P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
20669         * config/rs6000/vsx.md: Add define_insn extract4b.  Add define_expand
20670         definition for insert4b and define insn *insert3b_internal.
20671         * doc/extend.texi: Add documentation for vec_extract4b.
20673 2018-02-16  Nathan Sidwell  <nathan@acm.org>
20675         * doc/extend.texi (Backwards Compatibility): Mention friend
20676         injection.  Note for-scope is deprecated.
20677         * doc/invoke.texi (-ffriend-injection): Deprecate.
20679 2018-02-16  Segher Boessenkool  <segher@kernel.crashing.org>
20681         * combine.c (try_combine): When adjusting LOG_LINKS for the destination
20682         that moved to I2, also allow destinations that are a paradoxical
20683         subreg (instead of a normal reg).
20685 2018-02-16  Oleg Endo  <olegendo@gcc.gnu.org>
20687         PR target/83831
20688         * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
20689         to QImode.
20691 2018-02-16  Richard Biener  <rguenther@suse.de>
20693         PR tree-optimization/84037
20694         PR tree-optimization/84016
20695         PR target/82862
20696         * config/i386/i386.c (ix86_builtin_vectorization_cost):
20697         Adjust vec_construct for the fact we need additional higher latency
20698         128bit inserts for AVX256 and AVX512 vector builds.
20699         (ix86_add_stmt_cost): Scale vector construction cost for
20700         elementwise loads.
20702 2018-02-16  Richard Biener  <rguenther@suse.de>
20704         PR tree-optimization/84417
20705         * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
20706         the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
20707         (non_rewritable_lvalue_p): Likewise, use poly-ints.
20709 2018-02-16  Martin Liska  <mliska@suse.cz>
20711         PR sanitizer/84307
20712         * internal-fn.def (ASAN_CHECK): Set proper flags.
20713         (ASAN_MARK): Likewise.
20715 2018-02-16  Julia Koval  <julia.koval@intel.com>
20717         * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
20718         from PTA_CANNONLAKE.
20720 2018-02-16  Jakub Jelinek  <jakub@redhat.com>
20722         PR target/84272
20723         * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
20724         Use ++iter rather than iter++ for std::list iterators.
20725         (func_fma_steering::dfs): Likewise.  Don't delete nodes right away,
20726         defer deleting them until all nodes in the forest are processed.  Do
20727         free even leaf nodes.  Change to_process into auto_vec.
20729         PR bootstrap/84405
20730         * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
20731         * vec.h (vec_default_construct): Use memset instead of placement new
20732         if BROKEN_VALUE_INITIALIZATION is defined.
20733         * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
20734         memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
20735         is defined.
20737         PR rtl-optimization/83723
20738         * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
20739         * lra.c (lra_substitute_pseudo): Likewise.  If true, use
20740         gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG.  Pass DEBUG_P to
20741         recursive calls.
20742         (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
20743         callers.
20744         * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
20746 2018-02-16  Eric Botcazou  <ebotcazou@adacore.com>
20748         PR rtl-optimization/81443
20749         * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
20750         from inner REGs to paradoxical SUBREGs.
20752 2018-02-16  Richard Biener  <rguenther@suse.de>
20754         PR tree-optimization/84399
20755         * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
20756         For operands we can analyze at their definition make sure we can
20757         analyze them at each use as well.
20759 2018-02-16  Richard Biener  <rguenther@suse.de>
20761         PR tree-optimization/84190
20762         * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
20763         volatile accesses if the decl isn't volatile.
20765 2018-02-15  Jason Merrill  <jason@redhat.com>
20767         PR c++/84314 - ICE with templates and fastcall attribute.
20768         * attribs.c (build_type_attribute_qual_variant): Don't clobber
20769         TYPE_CANONICAL on an existing type.
20771 2018-02-15  Jakub Jelinek  <jakub@redhat.com>
20773         PR tree-optimization/84383
20774         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
20775         dstoff nor call operand_equal_p if dstbase is NULL.
20777         PR tree-optimization/84334
20778         * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
20779         also a CONSTANT_CLASS_P, punt.
20781 2018-02-14  Jim Wilson  <jimw@sifive.com>
20783         * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
20784         first SMALL_OPERAND check.  New local min_second_step.  Move assert
20785         to where locals are set.  Add TARGET_RVC support.
20786         * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
20788 2018-02-14  Indu Bhagat  <indu.bhagat@oracle.com>
20790         * doc/invoke.texi: Correct -Wformat-overflow code sample.
20792 2018-02-14  Martin Sebor  <msebor@redhat.com>
20794         PR tree-optimization/83698
20795         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
20796         arrays constrain the offset range to their bounds.
20797         (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
20798         (builtin_access::overlap): Avoid setting the size of overlap if it's
20799         already been set.
20800         (maybe_diag_overlap): Also consider arrays when deciding what values
20801         of offsets to include in diagnostics.
20803 2018-02-14  Martin Sebor  <msebor@redhat.com>
20805         PR c/84108
20806         * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
20807         that correspond to the kind of a declaration.
20809 2018-02-14  John David Anglin  <danglin@gcc.gnu.org>
20811         PR target/83984
20812         * config/pa/pa.md: Load address of PIC label using the linkage table
20813         if the label is nonlocal.
20815 2018-02-14  Kelvin Nilsen  <kelvin@gcc.gnu.org>
20817         * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
20818         warning message if user requests -maltivec=be.
20819         * doc/invoke.texi: Document deprecation of -maltivec=be.
20821 2018-02-14  Will Schmidt  <will_schmidt@vnet.ibm.com>
20823         PR target/84220
20824         * config/rs6000/rs6000-c.c: Update definitions for
20825         ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
20826         VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
20828 2018-02-14  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
20830         PR target/84239
20831         * config/i386/cetintrin.h: Remove _rdssp[d|q] and
20832         add _get_ssp intrinsics. Remove argument from
20833         __builtin_ia32_rdssp[d|q].
20834         * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
20835         * config/i386/i386-builtin.def: Remove argument from
20836         __builtin_ia32_rdssp[d|q].
20837         * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
20838         ix86_expand_special_args_builtin for _rdssp[d|q].
20839         * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
20840         Clear register before usage.
20841         * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
20842         Add documentation for new _get_ssp and _inc_ssp intrinsics.
20844 2018-02-14  Richard Sandiford  <richard.sandiford@linaro.org>
20846         PR tree-optimization/84357
20847         * tree-data-ref.c (object_address_invariant_in_loop_p): Check
20848         operand 1 of an ARRAY_REF too.
20850 2018-02-14  Oleg Endo  <olegendo@gcc.gnu.org>
20852         PR target/83831
20853         * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
20854         rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
20855         declarations.
20856         (set_of_reg): New struct.
20857         (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
20858         * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
20859         rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
20860         functions.
20861         * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
20862         Split into bitclr, bitset, bitinvert patterns if appropriate.
20863         (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
20864         use rx_fuse_in_memory_bitop.
20865         (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
20866         to named insn, correct maximum insn length.
20868 2018-02-14  Jozef Lawrynowicz  <jozefl.gcc@gmail.com>
20870         PR target/79242
20871         * machmode.def: Define a complex mode for PARTIAL_INT.
20872         * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
20873         MODE_PARTIAL_INT.
20874         * doc/rtl.texi: Document CSPImode.
20875         * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
20876         handling.
20877         (msp430_hard_regno_nregs_with_padding): Likewise.
20879 2018-02-13  Peter Bergner  <bergner@vnet.ibm.com>
20881         PR target/84279
20882         * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
20884 2018-02-13  Segher Boessenkool  <segher@kernel.crashing.org>
20886         PR rtl-optimization/84169
20887         * combine.c (try_combine): New variable split_i2i3.  Set it to true if
20888         we generated a parallel as new i3 and we split that to new i2 and i3
20889         instructions.  Handle split_i2i3 similar to swap_i2i3: scan the
20890         LOG_LINKs of i3 to see which of those need to link to i2 now.  Link
20891         those to i2, not i1.  Partially rewrite this scan code.
20893 2018-02-13  Jakub Jelinek  <jakub@redhat.com>
20895         PR c/82210
20896         * stor-layout.c (place_field): For variable length fields, adjust
20897         offset_align afterwards not just based on the field's alignment,
20898         but also on the size.
20900         PR middle-end/84309
20901         * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
20902         of exps and logs in the use_exp2 case.
20904 2018-02-13  Jeff Law  <law@redhat.com>
20906         * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
20907         entry for "vector".
20909         * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
20910         ARGS as unused.
20912 2018-02-13  Alexandre Oliva  <aoliva@redhat.com>
20914         PR debug/84342
20915         PR debug/84319
20916         * common.opt (gas-loc-support, gas-locview-support): New.
20917         (ginline-points, ginternal-reset-location-views): New.
20918         * doc/invoke.texi: Document them.  Use @itemx where intended.
20919         (gvariable-location-views): Adjust.
20920         * target.def (reset_location_view): New.
20921         * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
20922         (TARGET_RESET_LOCATION_VIEW): New.
20923         * doc/tm.texi: Rebuilt.
20924         * dwarf2out.c (dwarf2out_default_as_loc_support): New.
20925         (dwarf2out_default_as_locview_support): New.
20926         (output_asm_line_debug_info): Use option variables.
20927         (dwarf2out_maybe_output_loclist_view_pair): Likewise.
20928         (output_loc_list): Likewise.
20929         (add_high_low_attributes): Check option variables.
20930         Don't output entry view attribute in strict mode.
20931         (gen_inlined_subroutine_die): Check option variables.
20932         (dwarf2out_inline_entry): Likewise.
20933         (init_sections_and_labels): Likewise.
20934         (dwarf2out_early_finish): Likewise.
20935         (maybe_reset_location_view): New, from...
20936         (dwarf2out_var_location): ... here.  Call it.
20937         * debug.h (dwarf2out_default_as_loc_support): Declare.
20938         (dwarf2out_default_as_locview_support): Declare.
20939         * hooks.c (hook_int_rtx_insn_0): New.
20940         * hooks.h (hook_int_rtx_insn_0): Declare.
20941         * toplev.c (process_options): Take -gas-loc-support and
20942         -gas-locview-support from dwarf2out.  Enable
20943         -gvariable-location-views by default only with locview
20944         assembler support.  Enable -ginternal-reset-location-views by
20945         default only if the target defines the corresponding hook.
20946         Enable -ginline-points by default if location views are
20947         enabled; force it disabled if statement frontiers are
20948         disabled.
20949         * tree-inline.c (expand_call_inline): Check option variables.
20950         * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
20952 2018-02-13  Richard Sandiford  <richard.sandiford@linaro.org>
20954         PR tree-optimization/84321
20955         * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
20956         handling.  Also check whether the anti-range contains any values
20957         that satisfy the mask; switch to a VR_RANGE if not.
20959 2018-02-13  Paolo Bonzini  <bonzini@gnu.org>
20961         PR sanitizer/84340
20962         * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
20964 2018-02-13  Martin Jambor  <mjambor@suse.cz>
20966         PR c++/83990
20967         * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
20968         of call statements, also set location of a load to a temporary.
20970 2018-02-13  Sebastian Perta  <sebastian.perta@renesas.com>
20972         * config/rl78/rl78.c (add_vector_labels): New function.
20973         * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
20974         * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
20975         * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
20976         which checks that no arguments are passed.
20977         * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
20978         * doc/extend.texi: Documentation for the new attribute.
20980 2018-02-13  Andreas Schwab  <schwab@suse.de>
20982         * config/riscv/linux.h (CPP_SPEC): Define.
20984 2018-02-13  Jakub Jelinek  <jakub@redhat.com>
20986         PR target/84335
20987         * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
20988         OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
20989         OPTION_MASK_ISA_AES as first argument to def_builtin_const
20990         for AES builtins.  Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
20991         instead of OPTION_MASK_ISA_PCLMUL as first argument to
20992         def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
20993         * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
20994         temporarily for AES and PCLMUL builtins.
20996         PR tree-optimization/84339
20997         * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
20998         ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
20999         Formatting fixes.
21001         PR middle-end/84309
21002         * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
21003         exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
21004         * generic-match-head.c (canonicalize_math_after_vectorization_p): New
21005         inline function.
21006         * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
21007         inline function.
21008         * omp-simd-clone.h: New file.
21009         * omp-simd-clone.c: Include omp-simd-clone.h.
21010         (expand_simd_clones): No longer static.
21011         * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
21012         cgraph.h and omp-simd-clone.h.
21013         (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
21014         (vect_recog_widen_shift_pattern): Formatting fix.
21015         (vect_pattern_recog_1): Don't check optab for calls.
21017         PR target/84336
21018         * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
21019         operands[2] into a REG before using gen_lowpart on it.
21021 2018-02-12  Jeff Law  <law@redhat.com>
21023         PR target/83760
21024         * config/sh/sh.c (find_barrier): Consider a sibling call
21025         a barrier as well.
21027         * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
21028         successfully back substituting a reg.
21030 2018-02-12  Richard Biener  <rguenther@suse.de>
21032         PR tree-optimization/84037
21033         * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
21034         parameter, move visited init to caller.
21035         (vect_slp_analyze_operations): Separate cost from validity
21036         check, initialize visited once for all instances.
21037         (vect_schedule_slp): Analyze map to CSE vectorized nodes once
21038         for all instances.
21039         * tree-vect-stmts.c (vect_model_simple_cost): Make early
21040         out an assert.
21041         (vect_model_promotion_demotion_cost): Likewise.
21042         (vectorizable_bswap): Guard cost modeling with !slp_node
21043         instead of !PURE_SLP_STMT to avoid double-counting on hybrid
21044         SLP stmts.
21045         (vectorizable_call): Likewise.
21046         (vectorizable_conversion): Likewise.
21047         (vectorizable_assignment): Likewise.
21048         (vectorizable_shift): Likewise.
21049         (vectorizable_operation): Likewise.
21050         (vectorizable_store): Likewise.
21051         (vectorizable_load): Likewise.
21052         (vectorizable_condition): Likewise.
21053         (vectorizable_comparison): Likewise.
21055 2018-02-12  Paolo Bonzini  <bonzini@gnu.org>
21057         PR sanitizer/84307
21058         * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
21059         (ASAN_MARK): Fix fnspec to account for return value, change pointer
21060         argument from 'R' to 'W' so that the pointed-to datum is clobbered.
21062 2018-02-08  Jan Hubicka  <hubicka@ucw.cz>
21064         PR middle-end/83665
21065         * params.def (inline-min-speedup): Increase from 8 to 15.
21066         (max-inline-insns-auto): Decrease from 40 to 30.
21067         * ipa-split.c (consider_split): Add some buffer for function to
21068         be considered inlining candidate.
21069         * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
21070         default values.
21072 2018-02-12  Richard Biener  <rguenther@suse.de>
21074         PR tree-optimization/84037
21075         * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
21076         matched stmts if we cannot swap the non-matched ones.
21078 2018-02-12  Olga Makhotina  <olga.makhotina@intel.com>
21080         * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
21081         _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
21082         _mm_maskz_scalef_round_ss): New intrinsics.
21083         (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
21084         * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
21085         __builtin_ia32_scalefss_round): Remove.
21086         (__builtin_ia32_scalefsd_mask_round,
21087         __builtin_ia32_scalefss_mask_round): New intrinsics.
21088         * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
21089         (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
21090         ((match_operand:VF_128 2 "<round_nimm_predicate>"
21091         "<round_constraint>")): Changed to ...
21092         ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
21093         "<round_scalar_constraint>")): ... this.
21094         ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
21095         %0, %1, %2<round_op3>}"): Changed to ...
21096         ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
21097         %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
21098         %2<round_scalar_mask_op3>}"): ... this.
21099         * config/i386/subst.md (round_scalar_nimm_predicate): New.
21101 2018-02-12  Olga Makhotina  <olga.makhotina@intel.com>
21103         * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
21104         (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
21105         (_mm_maskz_sqrt_round_ss): New intrinsics.
21106         (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
21107         (__builtin_ia32_sqrtsd_mask_round)
21108         (__builtin_ia32_sqrtss_mask_round): New builtins.
21109         * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
21110         (__builtin_ia32_sqrtss_round): Remove.
21111         (__builtin_ia32_sqrtsd_mask_round)
21112         (__builtin_ia32_sqrtss_mask_round): New builtins.
21113         * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
21114         (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
21115         ((match_operand:VF_128 1 "vector_operand"
21116         "xBm,<round_constraint>")): Changed to ...
21117         ((match_operand:VF_128 1 "vector_operand"
21118         "xBm,<round_scalar_constraint>")): ... this.
21119         (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
21120         %0, %2, %<iptr>1<round_op3>}): Changed to ...
21121         (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
21122         %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
21123         %<iptr>1<round_scalar_mask_op3>}): ... this.
21124         ((set_attr "prefix" "<round_prefix>")): Changed to ...
21125         ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
21127 2018-02-11  Steven Munroe  <munroesj@gcc.gnu.org>
21129         PR target/84266
21130         * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
21131         Cast vec_cmpeq result to correct type.
21132         * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
21133         Cast vec_cmpgt result to correct type.
21135 2018-02-11  Alexandre Oliva  <aoliva@redhat.com>
21137         * final.c (final_scan_insn_1): Renamed from...
21138         (final_scan_insn): ... this.  New wrapper, to recover
21139         seen from the outermost call in recursive ones.
21140         * config/sparc/sparc.c (output_return): Drop seen from call.
21141         (output_sibcall): Likewise.
21142         * config/visium/visium.c (output_branch): Likewise.
21144 2018-02-10  John David Anglin  <danglin@gcc.gnu.org>
21146         * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
21147         function label.
21149 2018-02-10  Alan Modra  <amodra@gmail.com>
21151         PR target/84300
21152         * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
21153         Specify LR as an input.
21155 2018-02-10  Jakub Jelinek  <jakub@redhat.com>
21157         PR sanitizer/83987
21158         * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
21159         remove_member_access_dummy_vars): New functions.
21160         (lower_omp_for, lower_omp_taskreg, lower_omp_target,
21161         lower_omp_1, execute_lower_omp): Use them.
21163         PR rtl-optimization/84308
21164         * shrink-wrap.c (spread_components): Release todo vector.
21166 2018-02-09  Vladimir Makarov  <vmakarov@redhat.com>
21168         PR rtl-optimization/57193
21169         * ira-color.c (struct allocno_color_data): Add member
21170         conflict_allocno_hard_prefs.
21171         (update_conflict_allocno_hard_prefs): New.
21172         (bucket_allocno_compare_func): Add a preference based on
21173         conflict_allocno_hard_prefs.
21174         (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
21175         (color_allocnos): Remove a dead code.  Initiate
21176         conflict_allocno_hard_prefs.  Call update_costs_from_prefs.
21178 2018-02-09  Jakub Jelinek  <jakub@redhat.com>
21180         PR target/84226
21181         * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
21182         constraint from =wa to wa.  Avoid a subreg on the output operand,
21183         instead use a pseudo and subreg it in a move.
21184         (p9_xxbrd_<mode>): Changed to ...
21185         (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
21186         (p9_xxbrd_v2df): New expander.
21187         (p9_xxbrw_<mode>): Changed to ...
21188         (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
21189         (p9_xxbrw_v4sf): New expander.
21191 2018-02-09  Sebastian Perta  <sebastian.perta@renesas.com>
21193         * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
21195 2018-02-09  Peter Bergner  <bergner@vnet.ibm.com>
21197         PR target/83926
21198         * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
21199         multiply in 32-bit mode.
21200         (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
21201         (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
21202         mode.
21204 2018-02-09  Sebastian Perta  <sebastian.perta@renesas.com>
21206         * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
21207         to allow or block "symbol_ref" depending on the value of TARGET_JSR.
21208         * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
21209         * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
21211 2018-02-09  Pierre-Marie de Rodat  <derodat@adacore.com>
21213         PR lto/84213
21214         * dwarf2out.c (is_trivial_indirect_ref): New function.
21215         (dwarf2out_late_global_decl): Do not generate a location
21216         attribute for variables that have a non-trivial DECL_VALUE_EXPR
21217         and that are not defined in the current unit.
21219 2018-02-09  Eric Botcazou  <ebotcazou@adacore.com>
21221         * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
21222         instead of a libcall for UNORDERED.
21224 2018-02-09  Tamar Christina  <tamar.christina@arm.com>
21226         PR target/82641
21227         * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
21228         __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
21230 2018-02-09  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
21232         PR target/PR84295
21233         * config/s390/s390.c (s390_set_current_function): Invoke
21234         s390_indirect_branch_settings also if fndecl didn't change.
21236 2018-02-09  Alexandre Oliva  <aoliva@redhat.com>
21238         * config/rs6000/rs6000.md (blockage): Set length to zero.
21240 2018-02-09  Eric Botcazou  <ebotcazou@adacore.com>
21242         * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
21244 2018-02-09  Jakub Jelinek  <jakub@redhat.com>
21246         PR sanitizer/84285
21247         * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
21248         STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
21249         -static-lib*san.
21251         PR debug/84252
21252         * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
21253         PARALLEL incoming that failed vt_get_decl_and_offset check.
21255         PR middle-end/84237
21256         * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
21257         * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
21258         TREE_READONLY bit.
21259         (get_variable_section): For decls in named .bss* sections pass true as
21260         second argument to bss_initializer_p.
21262 2018-02-09  Marek Polacek  <polacek@redhat.com>
21263             Jakub Jelinek  <jakub@redhat.com>
21265         PR c++/83659
21266         * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
21267         Formatting fixes.  Verify first that tree_fits_poly_int64_p (op01).
21268         Sync some changes from cxx_fold_indirect_ref.
21270 2018-02-09  Alexandre Oliva  <aoliva@redhat.com>
21272         * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
21273         markers.
21274         * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
21275         (BLOCK_INLINE_ENTRY_LABEL): New.
21276         (dwarf2out_var_location): Disregard inline entry markers.
21277         (inline_entry_data): New struct.
21278         (inline_entry_data_hasher): New hashtable type.
21279         (inline_entry_data_hasher::hash): New.
21280         (inline_entry_data_hasher::equal): New.
21281         (inline_entry_data_table): New variable.
21282         (add_high_low_attributes): Add DW_AT_entry_pc and
21283         DW_AT_GNU_entry_view attributes if a pending entry is found
21284         in inline_entry_data_table.  Add old entry_pc attribute only
21285         if debug nonbinding markers are disabled.
21286         (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
21287         markers are enabled.
21288         (block_within_block_p, dwarf2out_inline_entry): New.
21289         (dwarf2out_finish): Check that no entries remained in
21290         inline_entry_data_table.
21291         * final.c (reemit_insn_block_notes): Handle inline entry notes.
21292         (final_scan_insn, notice_source_line): Likewise.
21293         (rest_of_clean_state): Skip inline entry markers.
21294         * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
21295         markers.
21296         * gimple.c (gimple_build_debug_inline_entry): New.
21297         * gimple.h (enum gimple_debug_subcode): Add
21298         GIMPLE_DEBUG_INLINE_ENTRY.
21299         (gimple_build_debug_inline_entry): Declare.
21300         (gimple_debug_inline_entry_p): New.
21301         (gimple_debug_nonbind_marker_p): Adjust.
21302         * insn-notes.def (INLINE_ENTRY): New.
21303         * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
21304         inline entry marker notes.
21305         (print_insn): Likewise.
21306         * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
21307         (INSN_DEBUG_MARKER_KIND): Likewise.
21308         (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
21309         * tree-inline.c (expand_call_inline): Build and insert
21310         debug_inline_entry stmt.
21311         * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
21312         inline entry blocks early, if nonbind markers are enabled.
21313         (dump_scope_block): Dump fragment info.
21314         * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
21315         * doc/gimple.texi (gimple_debug_inline_entry_p): New.
21316         (gimple_build_debug_inline_entry): New.
21317         * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
21318         Enable/disable inline entry points too.
21319         * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
21320         (DEBUG_INSN): Describe inline entry markers.
21322         * common.opt (gvariable-location-views): New.
21323         (gvariable-location-views=incompat5): New.
21324         * config.in: Rebuilt.
21325         * configure: Rebuilt.
21326         * configure.ac: Test assembler for view support.
21327         * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
21328         * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
21329         * dwarf2out.c (var_loc_view): New typedef.
21330         (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
21331         (dwarf2out_locviews_in_attribute): New.
21332         (dwarf2out_locviews_in_loclist): New.
21333         (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
21334         (enum dw_line_info_opcode): Add LI_adv_address.
21335         (struct dw_line_info_table): Add view.
21336         (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
21337         (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
21338         (zero_view_p): New variable.
21339         (ZERO_VIEW_P): New macro.
21340         (output_asm_line_debug_info): New.
21341         (struct var_loc_node): Add view.
21342         (add_AT_view_list, AT_loc_list): New.
21343         (add_var_loc_to_decl): Add view param.  Test it against last.
21344         (new_loc_list): Add view params.  Record them.
21345         (AT_loc_list_ptr): Handle loc and view lists.
21346         (view_list_to_loc_list_val_node): New.
21347         (print_dw_val): Handle dw_val_class_view_list.
21348         (size_of_die): Likewise.
21349         (value_format): Likewise.
21350         (loc_list_has_views): New.
21351         (gen_llsym): Set vl_symbol too.
21352         (maybe_gen_llsym, skip_loc_list_entry): New.
21353         (dwarf2out_maybe_output_loclist_view_pair): New.
21354         (output_loc_list): Output view list or entries too.
21355         (output_view_list_offset): New.
21356         (output_die): Handle dw_val_class_view_list.
21357         (output_dwarf_version): New.
21358         (output_compilation_unit_header): Use it.
21359         (output_skeleton_debug_sections): Likewise.
21360         (output_rnglists, output_line_info): Likewise.
21361         (output_pubnames, output_aranges): Update version comments.
21362         (output_one_line_info_table): Output view numbers in asm comments.
21363         (dw_loc_list): Determine current endview, pass it to new_loc_list.
21364         Call maybe_gen_llsym.
21365         (loc_list_from_tree_1): Adjust.
21366         (add_AT_location_description): Create view list attribute if
21367         needed, check it's absent otherwise.
21368         (convert_cfa_to_fb_loc_list): Adjust.
21369         (maybe_emit_file): Call output_asm_line_debug_info for test.
21370         (dwarf2out_var_location): Reset views as needed.  Precompute
21371         add_var_loc_to_decl args.  Call get_attr_min_length only if we have the
21372         attribute.  Set view.
21373         (new_line_info_table): Reset next view.
21374         (set_cur_line_info_table): Call output_asm_line_debug_info for test.
21375         (dwarf2out_source_line): Likewise.  Output view resets and labels to
21376         the assembler, or select appropriate line info opcodes.
21377         (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
21378         (optimize_string_length): Catch it.  Adjust.
21379         (resolve_addr): Copy vl_symbol along with ll_symbol.  Handle
21380         dw_val_class_view_list, and remove it if no longer needed.
21381         (hash_loc_list): Hash view numbers.
21382         (loc_list_hasher::equal): Compare them.
21383         (optimize_location_lists): Check whether a view list symbol is
21384         needed, and whether the locview attribute is present, and
21385         whether they match.  Remove the locview attribute if no longer
21386         needed.
21387         (index_location_lists): Call skip_loc_list_entry for test.
21388         (dwarf2out_finish): Call output_asm_line_debug_info for test.
21389         Use output_dwarf_version.
21390         * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
21391         (struct dw_val_node): Add val_view_list.
21392         * final.c (SEEN_NEXT_VIEW): New.
21393         (set_next_view_needed): New.
21394         (clear_next_view_needed): New.
21395         (maybe_output_next_view): New.
21396         (final_start_function): Rename to...
21397         (final_start_function_1): ... this.  Take pointer to FIRST,
21398         add SEEN parameter.  Emit param bindings in the initial view.
21399         (final_start_function): Reintroduce SEEN-less interface.
21400         (final): Rename to...
21401         (final_1): ... this.  Take SEEN parameter.  Output final pending
21402         next view at the end.
21403         (final): Reintroduce seen-less interface.
21404         (final_scan_insn): Output pending next view before switching
21405         sections or ending a block.  Mark the next view as needed when
21406         outputting variable locations.  Notify debug backend of section
21407         changes, and of location view changes.
21408         (rest_of_handle_final): Adjust.
21409         * toplev.c (process_options): Autodetect value for debug variable
21410         location views option.  Warn on incompat5 without -gdwarf-5.
21411         * doc/invoke.texi (gvariable-location-views): New.
21412         (gvariable-location-views=incompat5): New.
21413         (gno-variable-location-views): New.
21415 2018-02-08  David Malcolm  <dmalcolm@redhat.com>
21417         PR tree-optimization/84136
21418         * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
21419         that the result of find_edge is non-NULL.
21421 2018-02-08  Sergey Shalnov  <sergey.shalnov@intel.com>
21423         PR target/83008
21424         * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
21425         storing integer register in SImode.  Fix cost of 256 and 512
21426         byte aligned SSE register store.
21428 2018-02-08  Sergey Shalnov  <sergey.shalnov@intel.com>
21430         * config/i386/i386.c (ix86_multiplication_cost): Fix
21431         multiplication cost for TARGET_AVX512DQ.
21433 2018-02-08  Marek Polacek  <polacek@redhat.com>
21435         PR tree-optimization/84238
21436         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
21437         get_range_strlen.
21439 2018-02-08  Richard Sandiford  <richard.sandiford@linaro.org>
21441         PR tree-optimization/84265
21442         * tree-vect-stmts.c (vectorizable_store): Don't treat
21443         VMAT_CONTIGUOUS accesses as grouped.
21444         (vectorizable_load): Likewise.
21446 2018-02-08  Richard Sandiford  <richard.sandiford@linaro.org>
21448         PR tree-optimization/81635
21449         * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
21450         * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
21451         (test_round_for_mask): New functions.
21452         (wide_int_cc_tests): Call test_round_for_mask.
21453         * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
21454         * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
21455         * tree-data-ref.c (split_constant_offset_1): Use it to refine the
21456         range returned by get_range_info.
21458 2018-02-08  Jan Hubicka  <hubicka@ucw.cz>
21460         PR ipa/81360
21461         * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
21462         * symtab.c: Include builtins.h
21463         (symtab_node::output_to_lto_symbol_table_p): Move here
21464         from lto-streamer-out.c:output_symbol_p.
21465         * lto-streamer-out.c (write_symbol): Turn early exit to assert.
21466         (output_symbol_p): Move all logic to symtab.c
21467         (produce_symtab): Update.
21469 2018-02-08  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
21471         * config/s390/s390-opts.h (enum indirect_branch): Define.
21472         * config/s390/s390-protos.h (s390_return_addr_from_memory)
21473         (s390_indirect_branch_via_thunk)
21474         (s390_indirect_branch_via_inline_thunk): Add function prototypes.
21475         (enum s390_indirect_branch_type): Define.
21476         * config/s390/s390.c (struct s390_frame_layout, struct
21477         machine_function): Remove.
21478         (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
21479         (indirect_branch_table_label_no, indirect_branch_table_name):
21480         Define variables.
21481         (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
21482         (enum s390_indirect_branch_option): Define.
21483         (s390_return_addr_from_memory): New function.
21484         (s390_handle_string_attribute): New function.
21485         (s390_attribute_table): Add new attribute handler.
21486         (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
21487         (s390_indirect_branch_via_thunk): New function.
21488         (s390_indirect_branch_via_inline_thunk): New function.
21489         (s390_function_ok_for_sibcall): When jumping via thunk disallow
21490         sibling call optimization for non z10 compiles.
21491         (s390_emit_call): Force indirect branch target to be a single
21492         register.  Add r1 clobber for non-z10 compiles.
21493         (s390_emit_epilogue): Emit return jump via return_use expander.
21494         (s390_reorg): Handle JUMP_INSNs as execute targets.
21495         (s390_option_override_internal): Perform validity checks for the
21496         new command line options.
21497         (s390_indirect_branch_attrvalue): New function.
21498         (s390_indirect_branch_settings): New function.
21499         (s390_set_current_function): Invoke s390_indirect_branch_settings.
21500         (s390_output_indirect_thunk_function):  New function.
21501         (s390_code_end): Implement target hook.
21502         (s390_case_values_threshold): Implement target hook.
21503         (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
21504         macros.
21505         * config/s390/s390.h (struct s390_frame_layout)
21506         (struct machine_function): Move here from s390.c.
21507         (TARGET_INDIRECT_BRANCH_NOBP_RET)
21508         (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
21509         (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
21510         (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
21511         (TARGET_INDIRECT_BRANCH_NOBP_CALL)
21512         (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
21513         (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
21514         (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
21515         (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
21516         * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
21517         (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
21518         (mnemonic attribute): Add values which aren't recognized
21519         automatically.
21520         ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
21521         pattern for branch conversion.  Fix mnemonic attribute.
21522         ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
21523         indirect branch via thunk if requested.
21524         ("indirect_jump", "<code>"): Expand patterns for branch conversion.
21525         ("*indirect_jump"): Disable for branch conversion using out of
21526         line thunks.
21527         ("indirect_jump_via_thunk<mode>_z10")
21528         ("indirect_jump_via_thunk<mode>")
21529         ("indirect_jump_via_inlinethunk<mode>_z10")
21530         ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
21531         ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
21532         ("casesi_jump_via_inlinethunk<mode>_z10")
21533         ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
21534         ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
21535         ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
21536         ("*indirect2_jump"): Disable for branch conversion.
21537         ("casesi_jump"): Turn into expander and expand patterns for branch
21538         conversion.
21539         ("return_use"): New expander.
21540         ("*return"): Emit return via thunk and rename it to ...
21541         ("*return<mode>"): ... this one.
21542         * config/s390/s390.opt: Add new options and and enum for the
21543         option values.
21545 2018-02-08  Richard Sandiford  <richard.sandiford@linaro.org>
21547         * lra-constraints.c (match_reload): Unconditionally use
21548         gen_lowpart_SUBREG, rather than selecting between that
21549         and equivalent gen_rtx_SUBREG code.
21551 2018-02-08  Richard Biener  <rguenther@suse.de>
21553         PR tree-optimization/84233
21554         * tree-ssa-phiprop.c (propagate_with_phi): Use separate
21555         changed flag instead of boguously re-using phi_inserted.
21557 2018-02-08  Martin Jambor  <mjambor@suse.cz>
21559         * hsa-gen.c (get_symbol_for_decl): Set program allocation for
21560         static local variables.
21562 2018-02-08  Richard Biener  <rguenther@suse.de>
21564         PR tree-optimization/84278
21565         * tree-vect-stmts.c (vectorizable_store): When looking for
21566         smaller vector types to perform grouped strided loads/stores
21567         make sure the mode is supported by the target.
21568         (vectorizable_load): Likewise.
21570 2018-02-08  Wilco Dijkstra  <wdijkstr@arm.com>
21572         * config/aarch64/aarch64.c (aarch64_components_for_bb):
21573         Increase LDP/STP opportunities by adding adjacent callee-saves.
21575 2018-02-08  Wilco Dijkstra  <wdijkstr@arm.com>
21577         PR rtl-optimization/84068
21578         PR rtl-optimization/83459
21579         * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
21581 2018-02-08  Aldy Hernandez  <aldyh@redhat.com>
21583         PR tree-optimization/84224
21584         * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
21585         * calls.c (gimple_alloca_call_p): Only return TRUE when we have
21586         non-zero arguments.
21588 2018-02-07  Iain Sandoe  <iain@codesourcery.com>
21590         PR target/84113
21591         * config/rs6000/altivec.md (*restore_world): Remove LR use.
21592         * config/rs6000/predicates.md (restore_world_operation): Adjust op
21593         count, remove one USE.
21595 2018-02-07  Michael Meissner  <meissner@linux.vnet.ibm.com>
21597         * doc/install.texi (Configuration): Document the
21598         --with-long-double-format={ibm,ieee} PowerPC configuration
21599         options.
21601         PR target/84154
21602         * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
21603         Convert from define_expand to be define_insn_and_split.  Rework
21604         float/double/_Float128 conversions to QI/HI/SImode to work with
21605         both ISA 2.07 (power8) or ISA 3.0 (power9).  Fix regression where
21606         conversions to QI/HImode types did a store and then a load to
21607         truncate the value.  For conversions to VSX registers, don't split
21608         the insn, instead emit the code directly.  Use the code iterator
21609         any_fix to combine signed and unsigned conversions.
21610         (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
21611         (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
21612         (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
21613         (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
21614         (fix_<mode>di2_hw): Likewise.
21615         (fixuns_<mode>di2_hw): Likewise.
21616         (fix_<mode>si2_hw): Likewise.
21617         (fixuns_<mode>si2_hw): Likewise.
21618         (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
21619         (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
21620         (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
21621         fix<uns>_trunc<SFDF:mode>si2_p8.
21622         (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer used.
21623         (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
21624         (fix<uns>_<mode>_mem): Likewise.
21625         (fctiw<u>z_<mode>_mem): Likewise.
21626         (fix<uns>_<mode>_mem): Likewise.
21627         (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
21628         the register allocator from doing a direct move to the GPRs to do
21629         a store, and instead use the ISA 3.0 store byte/half-word from
21630         vector register instruction.  For IEEE 128-bit floating point,
21631         also optimize stores of 32-bit ints.
21632         (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
21634 2018-02-07  Alan Hayward  <alan.hayward@arm.com>
21636         * genextract.c (push_pathstr_operand): New function to support
21637         [a-zA-Z].
21638         (walk_rtx): Call push_pathstr_operand.
21639         (print_path): Support [a-zA-Z].
21641 2018-02-07  Richard Biener  <rguenther@suse.de>
21643         PR tree-optimization/84037
21644         * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
21645         (cse_and_gimplify_to_preheader): Declare.
21646         (vect_get_place_in_interleaving_chain): Likewise.
21647         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
21648         ivexpr_map.
21649         (_loop_vec_info::~_loop_vec_info): Delete it.
21650         (cse_and_gimplify_to_preheader): New function.
21651         * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
21652         * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
21653         (vectorizable_load): Likewise.  For grouped stores always base
21654         the IV on the first element.
21655         * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
21656         condition before gimplifying.
21658 2018-02-07  Jakub Jelinek  <jakub@redhat.com>
21660         * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
21661         *DIV_EXPR and *MOD_EXPR.
21663 2018-02-07  H.J. Lu  <hongjiu.lu@intel.com>
21665         PR target/84248
21666         * config/i386/i386.c (ix86_option_override_internal): Mask out
21667         the CF_SET bit when checking -fcf-protection.
21669 2018-02-07  Tom de Vries  <tom@codesourcery.com>
21671         PR libgomp/84217
21672         * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
21673         enough.
21675 2018-02-07  Richard Biener  <rguenther@suse.de>
21677         PR tree-optimization/84204
21678         * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
21679         this place.
21681         PR tree-optimization/84205
21682         * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
21683         special-case isl_ast_op_zdiv_r.
21685         PR tree-optimization/84223
21686         * graphite-scop-detection.c (gather_bbs::before_dom_children):
21687         Only add conditions from within the region.
21688         (gather_bbs::after_dom_children): Adjust.
21690 2018-02-07  Georg-Johann Lay  <avr@gjlay.de>
21692         PR target/84209
21693         * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
21694         * config/avr/avr.md: Only post-reload split REG-REG moves if
21695         either register is GENERAL_REG_P.
21697 2018-02-07  Jakub Jelinek  <jakub@redhat.com>
21699         PR tree-optimization/84235
21700         * tree-ssa-scopedtables.c
21701         (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
21702         if the subtraction is performed in floating point type where NaNs are
21703         honored.  For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
21704         build 1.  Formatting fix.
21706 2018-02-06  Jakub Jelinek  <jakub@redhat.com>
21708         PR target/84146
21709         * config/i386/i386.c (rest_of_insert_endbranch): Only skip
21710         NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
21711         and skip it regardless of bb boundaries.  Use CALL_P macro,
21712         don't test INSN_P (insn) together with CALL_P or JUMP_P check
21713         unnecessarily, formatting fix.
21715 2018-02-06  Michael Collison  <michael.collison@arm.com>
21717         * config/arm/thumb2.md:
21718         (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
21719         (*thumb_mov_notscc): Ditto.
21721 2018-02-06  Michael Meissner  <meissner@linux.vnet.ibm.com>
21723         PR target/84154
21724         * config/rs6000/rs6000.md (su code attribute): Use "u" for
21725         unsigned_fix, not "s".
21727 2018-02-06  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
21729         * configure.ac (gcc_fn_eh_frame_ro): New function.
21730         (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
21731         correct .eh_frame permissions.
21732         * configure: Regenerate.
21734 2018-02-06  Andrew Jenner  <andrew@codeourcery.com>
21736         * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
21737         irrelevant options.
21739 2018-02-06  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
21741         * config/rs6000/rs6000.c (rs6000_option_override_internal):
21742         Display warning message for -mno-speculate-indirect-jumps.
21744 2018-02-06  Andrew Jenner  <andrew@codesourcery.com>
21746         * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
21747         Undocumented.
21748         * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
21750 2018-02-06  Aldy Hernandez  <aldyh@redhat.com>
21752         PR tree-optimization/84225
21753         * tree-eh.c (find_trapping_overflow): Only call
21754         operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
21756 2018-02-06  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
21758         PR target/84145
21759         * config/i386/i386.c: Reimplement the check of possible options
21760         -mibt/-mshstk conbination. Change error messages.
21761         * doc/invoke.texi: Fix a typo: remove extra '='.
21763 2018-02-06  Marek Polacek  <polacek@redhat.com>
21765         PR tree-optimization/84228
21766         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
21768 2018-02-06  Tamar Christina  <tamar.christina@arm.com>
21770         PR target/82641
21771         * config/arm/arm.c (arm_print_asm_arch_directives): Record already
21772         emitted arch directives.
21773         * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
21774         __ARM_FEATURE_COPROC before changing architectures.
21776 2018-02-06  Richard Biener  <rguenther@suse.de>
21778         * config/i386/i386.c (print_reg): Fix typo.
21779         (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
21781 2018-02-06  Eric Botcazou  <ebotcazou@adacore.com>
21783         * configure: Regenerate.
21785 2018-02-05  Martin Sebor  <msebor@redhat.com>
21787         PR tree-optimization/83369
21788         * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
21789         inlining context.
21791 2018-02-05  Martin Liska  <mliska@suse.cz>
21793         * doc/invoke.texi: Cherry-pick upstream r323995.
21795 2018-02-05  Richard Sandiford  <richard.sandiford@linaro.org>
21797         * ira.c (ira_init_register_move_cost): Adjust comment.
21799 2018-02-05  Martin Liska  <mliska@suse.cz>
21801         PR gcov-profile/84137
21802         * doc/gcov.texi: Fix typo in documentation.
21804 2018-02-05  Martin Liska  <mliska@suse.cz>
21806         PR gcov-profile/83879
21807         * doc/gcov.texi: Document necessity of --dynamic-list-data when
21808         using dlopen functionality.
21810 2018-02-05  Olga Makhotina  <olga.makhotina@intel.com>
21812         * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
21813         _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
21814         _mm_maskz_range_ss, _mm_mask_range_round_ss,
21815         _mm_maskz_range_round_ss): New intrinsics.
21816         (__builtin_ia32_rangesd128_round)
21817         (__builtin_ia32_rangess128_round): Remove.
21818         (__builtin_ia32_rangesd128_mask_round,
21819         __builtin_ia32_rangess128_mask_round): New builtins.
21820         * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
21821         __builtin_ia32_rangess128_round): Remove.
21822         (__builtin_ia32_rangesd128_mask_round,
21823         __builtin_ia32_rangess128_mask_round): New builtins.
21824         * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
21825         (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
21826         ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
21827         "<round_saeonly_constraint>")): Changed to ...
21828         ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
21829         "<round_saeonly_scalar_constraint>")): ... this.
21830         ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
21831         %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
21832         ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
21833         %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
21834         %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
21836 2018-02-02  Andrew Jenner  <andrew@codesourcery.com>
21838         * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
21839         options.
21840         * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
21841         Remove all values except native, 8540 and 8548.
21843 2018-02-02  H.J. Lu  <hongjiu.lu@intel.com>
21845         * config/i386/i386.c (ix86_output_function_return): Pass
21846         INVALID_REGNUM, instead of -1, as invalid register number to
21847         indirect_thunk_name and output_indirect_thunk.
21849 2018-02-02  Julia Koval  <julia.koval@intel.com>
21851         * config.gcc: Add -march=icelake.
21852         * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
21853         * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
21854         * config/i386/i386.c (processor_costs): Add m_ICELAKE.
21855         (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
21856         PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
21857         (processor_target_table): Add icelake.
21858         (ix86_option_override_internal): Handle new PTAs.
21859         (get_builtin_code_for_version): Handle icelake.
21860         (M_INTEL_COREI7_ICELAKE): New.
21861         (fold_builtin_cpu): Handle icelake.
21862         * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
21863         * doc/invoke.texi: Add -march=icelake.
21865 2018-02-02  Julia Koval  <julia.koval@intel.com>
21867         * config/i386/i386.c (ix86_option_override_internal): Change flags type
21868         to wide_int_bitmask.
21869         * wide-int-bitmask.h: New.
21871 2018-02-02  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
21873         PR target/84066
21874         * config/i386/i386.md: Replace Pmode with word_mode in
21875         builtin_setjmp_setup and builtin_longjmp to support x32.
21877 2018-02-01  Peter Bergner  <bergner@vnet.ibm.com>
21879         PR target/56010
21880         PR target/83743
21881         * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
21882         #include "opts.h".
21883         (rs6000_supported_cpu_names): New static variable.
21884         (linux_cpu_translation_table): Likewise.
21885         (elf_platform) <cpu>: Define new static variable and use it.
21886         Translate kernel AT_PLATFORM name to canonical name if needed.
21887         Error if platform name is unknown.
21889 2018-02-01  Aldy Hernandez  <aldyh@redhat.com>
21891         PR target/84089
21892         * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
21894 2018-02-01  Jeff Law  <law@redhat.com>
21896         PR target/84128
21897         * config/i386/i386.c (release_scratch_register_on_entry): Add new
21898         OFFSET and RELEASE_VIA_POP arguments.  Use SP+OFFSET to restore
21899         the scratch if RELEASE_VIA_POP is false.
21900         (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
21901         If we have to save a temporary register, decrement SIZE appropriately.
21902         Pass new arguments to release_scratch_register_on_entry.
21903         (ix86_adjust_stack_and_probe): Likewise.
21904         (ix86_emit_probe_stack_range): Pass new arguments to
21905         release_scratch_register_on_entry.
21907 2018-02-01  Uros Bizjak  <ubizjak@gmail.com>
21909         PR rtl-optimization/84157
21910         * combine.c (change_zero_ext): Use REG_P predicate in
21911         front of HARD_REGISTER_P predicate.
21913 2018-02-01  Georg-Johann Lay  <avr@gjlay.de>
21915         * config/avr/avr.c (avr_option_override): Move disabling of
21916         -fdelete-null-pointer-checks to...
21917         * common/config/avr/avr-common.c (avr_option_optimization_table):
21918         ...here.
21920 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
21922         PR tree-optimization/81635
21923         * tree-data-ref.c (split_constant_offset_1): For types that
21924         wrap on overflow, try to use range info to prove that wrapping
21925         cannot occur.
21927 2018-02-01  Renlin Li  <renlin.li@arm.com>
21929         PR target/83370
21930         * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
21931         TAILCALL_ADDR_REGS.
21932         (aarch64_register_move_cost): Likewise.
21933         * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
21934         TAILCALL_ADDR_REGS.
21935         (REG_CLASS_NAMES): Likewise.
21936         (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
21937         TAILCALL_ADDR_REGS. Remove IP registers.
21938         * config/aarch64/aarch64.md (Ucs): Update register constraint.
21940 2018-02-01  Richard Biener  <rguenther@suse.de>
21942         * domwalk.h (dom_walker::dom_walker): Add additional constructor
21943         for specifying RPO order and allow NULL for that.
21944         * domwalk.c (dom_walker::dom_walker): Likewise.
21945         (dom_walker::walk): Handle NULL RPO order.
21946         * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
21947         in RPO order.
21948         (rewrite_update_dom_walker): Likewise.
21949         (mark_def_dom_walker): Likewise.
21951 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
21953         * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
21954         (aarch64_maybe_expand_sve_subreg_move): Declare.
21955         * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
21956         * config/aarch64/predicates.md (aarch64_any_register_operand): New
21957         predicate.
21958         * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
21959         that are semantically a reverse operation.
21960         (*aarch64_sve_mov<mode>_subreg_be): New pattern.
21961         * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
21962         (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
21963         functions.
21964         (aarch64_can_change_mode_class): For big-endian, forbid changes
21965         between two SVE modes if they have different element sizes.
21967 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
21969         * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
21970         the TImode handling for big-endian targets.
21972 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
21974         * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
21975         (*sve_ld1rq<Vesize>): ... this new pattern.  Handle all element sizes,
21976         not just bytes.
21977         * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
21978         Remove BSWAP handing for big-endian targets and use the form of
21979         LD1RQ appropariate for the mode.
21981 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
21983         * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
21984         all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
21985         duplicated element.
21987 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
21989         PR tearget/83845
21990         * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
21991         check for operands that need to go through aarch64_sve_reload_be.
21993 2018-02-01  Jakub Jelinek  <jakub@redhat.com>
21995         PR tree-optimization/81661
21996         PR tree-optimization/84117
21997         * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
21998         * tree-eh.c: Include gimplify.h.
21999         (find_trapping_overflow, replace_trapping_overflow,
22000         rewrite_to_non_trapping_overflow): New functions.
22001         * tree-vect-loop.c: Include tree-eh.h.
22002         (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
22003         * tree-data-ref.c: Include tree-eh.h.
22004         (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
22006 2018-01-31  Uros Bizjak  <ubizjak@gmail.com>
22008         PR rtl-optimization/84123
22009         * combine.c (change_zero_ext): Check if hard register satisfies
22010         can_change_dest_mode before calling gen_lowpart_SUBREG.
22012 2018-01-31  Vladimir Makarov  <vmakarov@redhat.com>
22014         PR target/82444
22015         * ira.c (ira_init_register_move_cost): Remove assert.
22017 2018-01-31  Eric Botcazou  <ebotcazou@adacore.com>
22019         PR rtl-optimization/84071
22020         * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
22021         * doc/tm.texi: Regenerate.
22023 2018-01-31  Richard Biener  <rguenther@suse.de>
22025         PR tree-optimization/84132
22026         * tree-data-ref.c (analyze_miv_subscript): Properly
22027         check whether evolution_function_is_affine_multivariate_p
22028         before calling gcd_of_steps_may_divide_p.
22030 2018-01-31  Julia Koval  <julia.koval@intel.com>
22032         PR target/83618
22033         * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
22034         * config/i386/i386.md (rdpid_rex64) New.
22035         (rdpid): Make 32bit only.
22037 2018-01-29  Aldy Hernandez  <aldyh@redhat.com>
22039         PR lto/84105
22040         * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
22041         an IDENTIFIER_NODE for FUNCTION_TYPE's.
22043 2018-01-31  Eric Botcazou  <ebotcazou@adacore.com>
22045         Revert
22046         2018-01-12  Eric Botcazou  <ebotcazou@adacore.com>
22048         * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
22050 2018-01-31  Eric Botcazou  <ebotcazou@adacore.com>
22052         PR rtl-optimization/84071
22053         * combine.c (record_dead_and_set_regs_1): Record the source unmodified
22054         for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
22056 2018-01-31  Claudiu Zissulescu  <claziss@synopsys.com>
22058         * config/arc/arc.c (arc_handle_aux_attribute): New function.
22059         (arc_attribute_table): Add 'aux' attribute.
22060         (arc_in_small_data_p): Consider aux like variables.
22061         (arc_is_aux_reg_p): New function.
22062         (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
22063         (arc_get_aux_arg): New function.
22064         (prepare_move_operands): Handle aux-register access.
22065         (arc_handle_aux_attribute): New function.
22066         * doc/extend.texi (ARC Variable attributes): Add subsection.
22068 2018-01-31  Claudiu Zissulescu  <claziss@synopsys.com>
22070         * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
22071         * config/arc/arc.c (arc_handle_uncached_attribute): New function.
22072         (arc_attribute_table): Add 'uncached' attribute.
22073         (arc_print_operand): Print '.di' flag for uncached memory
22074         accesses.
22075         (arc_in_small_data_p): Do not consider for small data the uncached
22076         types.
22077         (arc_is_uncached_mem_p): New function.
22078         * config/arc/predicates.md (compact_store_memory_operand): Check
22079         for uncached memory accesses.
22080         (nonvol_nonimm_operand): Likewise.
22081         * doc/extend.texi (ARC Type Attribute): New subsection.
22083 2018-01-31  Jakub Jelinek  <jakub@redhat.com>
22085         PR c/84100
22086         * common.opt (falign-functions=, falign-jumps=, falign-labels=,
22087         falign-loops=): Add Optimization flag.
22089 2018-01-30  Jeff Law  <law@redhat.com>
22091         PR target/84064
22092         * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
22093         INT_REGISTERS_SAVED.  Check it prior to calling
22094         get_scratch_register_on_entry.
22095         (ix86_adjust_stack_and_probe): Similarly.
22096         (ix86_emit_probe_stack_range): Similarly.
22097         (ix86_expand_prologue): Corresponding changes.
22099 2018-01-30  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
22101         PR target/40411
22102         * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
22103         -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
22105 2018-01-30  Vladimir Makarov  <vmakarov@redhat.com>
22107         PR target/84112
22108         * lra-constraints.c (curr_insn_transform): Process AND in the
22109         address.
22111 2018-01-30  Jakub Jelinek  <jakub@redhat.com>
22113         PR rtl-optimization/83986
22114         * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
22115         dependence against last_pending_memory_flush in addition to
22116         pending_jump_insns.
22118 2018-01-30  Alexandre Oliva  <aoliva@redhat.com>
22120         PR tree-optimization/81611
22121         * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
22122         copies.
22124 2018-01-30  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
22126         PR target/83758
22127         * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
22128         a reg rtx.
22130 2018-01-30  Richard Biener  <rguenther@suse.de>
22131             Jakub Jelinek  <jakub@redhat.com>
22133         PR tree-optimization/84111
22134         * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
22135         inner loops added during recursion, as they don't have up-to-date
22136         SSA form.
22138 2018-01-30  Jan Hubicka  <hubicka@ucw.cz>
22140         PR ipa/81360
22141         * ipa-inline.c (can_inline_edge_p): Break out late tests to...
22142         (can_inline_edge_by_limits_p): ... here.
22143         (can_early_inline_edge_p, check_callers,
22144         update_caller_keys, update_callee_keys, recursive_inlining,
22145         add_new_edges_to_heap, speculation_useful_p,
22146         inline_small_functions,
22147         inline_small_functions, flatten_function,
22148         inline_to_all_callers_1): Update.
22150 2018-01-30  Jan Hubicka  <hubicka@ucw.cz>
22152         * profile-count.c (profile_count::combine_with_ipa_count): Handle
22153         zeros correctly.
22155 2018-01-30  Richard Biener  <rguenther@suse.de>
22157         PR tree-optimization/83008
22158         * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
22159         invariant and constant vector uses in stmts when they need
22160         more than one stmt.
22162 2018-01-30  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
22164         PR bootstrap/84017
22165         * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
22166         * configure: Regenerate.
22168 2018-01-30  Richard Sandiford  <richard.sandiford@linaro.org>
22170         * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
22171         pattern.
22172         (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
22173         Use gen_rtx_REG rather than gen_lowpart.
22175 2018-01-30  Richard Sandiford  <richard.sandiford@linaro.org>
22177         * lra-constraints.c (match_reload): Use subreg_lowpart_offset
22178         rather than 0 when creating partial subregs.
22180 2018-01-30  Richard Sandiford  <richard.sandiford@linaro.org>
22182         * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
22183         of usage.
22185 2018-01-29  Michael Meissner  <meissner@linux.vnet.ibm.com>
22187         PR target/81550
22188         * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
22189         and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
22190         -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
22191         flags.  This restores the settings used before the 2017-07-24.
22192         Turning off pre increment/decrement/modify allows IVOPTS to
22193         optimize DF/SF loops where the index is an int.
22195 2018-01-29  Richard Biener  <rguenther@suse.de>
22196             Kelvin Nilsen  <kelvin@gcc.gnu.org>
22198         PR bootstrap/80867
22199         * tree-vect-stmts.c (vectorizable_call): Don't call
22200         targetm.vectorize_builtin_md_vectorized_function if callee is
22201         NULL.
22203 2018-01-22  Carl Love  <cel@us.ibm.com>
22205         * doc/extend.tex: Fix typo in second arg in
22206         __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
22208 2018-01-29  Richard Biener  <rguenther@suse.de>
22210         PR tree-optimization/84086
22211         * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
22212         (flush_ssaname_freelist): When SSA names were released reset
22213         the SCEV hash table.
22215 2018-01-29  Richard Biener  <rguenther@suse.de>
22217         PR tree-optimization/84057
22218         * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
22219         removed paths when removing edges.
22221 2018-01-27  H.J. Lu  <hongjiu.lu@intel.com>
22223         * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
22224         -mfunction-return=@var{choice}.
22226 2018-01-27  Bernd Edlinger  <bernd.edlinger@hotmail.de>
22228         PR diagnostic/84034
22229         * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
22230         Handle CR like TAB.
22231         (layout::print_source_line): Likewise.
22232         (test_get_line_width_without_trailing_whitespace): Add test cases.
22234 2018-01-27  Jakub Jelinek  <jakub@redhat.com>
22236         PR middle-end/84040
22237         * sched-deps.c (sched_macro_fuse_insns): Return immediately for
22238         debug insns.
22240 2018-01-26  Jim Wilson  <jimw@sifive.com>
22242         * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
22244         * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
22245         specified.
22247 2018-01-26  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
22249         * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
22250         and CMP + SUB-immediate -> SUBS.
22252 2018-01-26  Martin Sebor  <msebor@redhat.com>
22254         PR tree-optimization/83896
22255         * tree-ssa-strlen.c (get_string_len): Rename...
22256         (get_string_cst_length): ...to this.  Return HOST_WIDE_INT.
22257         Avoid assuming length is constant.
22258         (handle_char_store): Use HOST_WIDE_INT for string length.
22260 2018-01-26  Uros Bizjak  <ubizjak@gmail.com>
22262         PR target/81763
22263         * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
22264         to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
22266 2018-01-26  Richard Biener  <rguenther@suse.de>
22268         PR rtl-optimization/84003
22269         * dse.c (record_store): Only record redundant stores when
22270         the earlier store aliases at least all accesses the later one does.
22272 2018-01-26  Jakub Jelinek  <jakub@redhat.com>
22274         PR rtl-optimization/83985
22275         * dce.c (deletable_insn_p): Return false for separate shrink wrapping
22276         REG_CFA_RESTORE insns.
22277         (delete_unmarked_insns): Don't ignore separate shrink wrapping
22278         REG_CFA_RESTORE insns here.
22280         PR c/83989
22281         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
22282         use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
22284 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
22286         * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
22287         * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
22288         (arc_init): Likewise.
22289         (arc_override_options): Likewise.
22290         (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
22291         value.
22292         (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
22293         support.
22294         * config/arc/arc.h (TARGET_DBNZ): Define.
22295         * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
22296         properly set the tune attribute.
22297         (dbnz): Use TARGET_DBNZ guard.
22298         * config/arc/arc.opt (mtune): Add core3 option.
22300 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
22302         * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
22303         recognize new pic like addresses.
22304         (arc_delegitimize_address): Clean up.
22306 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
22308         * config/arc/arc-arches.def: Option mrf16 valid for all
22309         architectures.
22310         * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
22311         * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
22312         * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
22313         * config/arc/arc-tables.opt: Regenerate.
22314         * config/arc/arc.c (arc_conditional_register_usage): Handle
22315         reduced register file case.
22316         (arc_file_start): Set must have build attributes.
22317         * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
22318         mrf16 option value.
22319         * config/arc/arc.opt (mrf16): Add new option.
22320         * config/arc/elf.h (ATTRIBUTE_PCS): Define.
22321         * config/arc/genmultilib.awk: Handle new mrf16 option.
22322         * config/arc/linux.h (ATTRIBUTE_PCS): Define.
22323         * config/arc/t-multilib: Regenerate.
22324         * doc/invoke.texi (ARC Options): Document mrf16 option.
22326 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
22328         * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
22329         * config/arc/arc.c (arc_handle_secure_attribute): New function.
22330         (arc_attribute_table): Add 'secure_call' attribute.
22331         (arc_print_operand): Print secure call operand.
22332         (arc_function_ok_for_sibcall): Don't optimize tail calls when
22333         secure.
22334         (arc_is_secure_call_p): New function.  * config/arc/arc.md
22335         (call_i): Add support for sjli instruction.
22336         (call_value_i): Likewise.
22337         * config/arc/constraints.md (Csc): New constraint.
22339 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
22340             John Eric Martin  <John.Martin@emmicro-us.com>
22342         * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
22343         * config/arc/arc.c (_arc_jli_section): New struct.
22344         (arc_jli_section): New type.
22345         (rc_jli_sections): New static variable.
22346         (arc_handle_jli_attribute): New function.
22347         (arc_attribute_table): Add jli_always and jli_fixed attribute.
22348         (arc_file_end): New function.
22349         (TARGET_ASM_FILE_END): Define.
22350         (arc_print_operand): Reuse 'S' letter for JLI output instruction.
22351         (arc_add_jli_section): New function.
22352         (jli_call_scan): Likewise.
22353         (arc_reorg): Call jli_call_scan.
22354         (arc_output_addsi): Remove 'S' from printing asm operand.
22355         (arc_is_jli_call_p): New function.
22356         * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
22357         operand.
22358         (movhi_insn): Likewise.
22359         (movsi_insn): Likewise.
22360         (movsi_set_cc_insn): Likewise.
22361         (loadqi_update): Likewise.
22362         (load_zeroextendqisi_update): Likewise.
22363         (load_signextendqisi_update): Likewise.
22364         (loadhi_update): Likewise.
22365         (load_zeroextendhisi_update): Likewise.
22366         (load_signextendhisi_update): Likewise.
22367         (loadsi_update): Likewise.
22368         (loadsf_update): Likewise.
22369         (movsicc_insn): Likewise.
22370         (bset_insn): Likewise.
22371         (bxor_insn): Likewise.
22372         (bclr_insn): Likewise.
22373         (bmsk_insn): Likewise.
22374         (bicsi3_insn): Likewise.
22375         (cmpsi_cc_c_insn): Likewise.
22376         (movsi_ne): Likewise.
22377         (movsi_cond_exec): Likewise.
22378         (clrsbsi2): Likewise.
22379         (norm_f): Likewise.
22380         (normw): Likewise.
22381         (swap): Likewise.
22382         (divaw): Likewise.
22383         (flag): Likewise.
22384         (sr): Likewise.
22385         (kflag): Likewise.
22386         (ffs): Likewise.
22387         (ffs_f): Likewise.
22388         (fls): Likewise.
22389         (call_i): Remove 'S' asm letter, add jli instruction.
22390         (call_value_i): Likewise.
22391         * config/arc/arc.op (mjli-always): New option.
22392         * config/arc/constraints.md (Cji): New constraint.
22393         * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
22394         operand.
22395         (subsf3_fpx): Likewise.
22396         (mulsf3_fpx): Likewise.
22397         * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
22398         asm operand.
22399         * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
22400         function attrbutes.
22401         * doc/invoke.texi (ARC): Document mjli-always option.
22403 2018-01-26  Sebastian Perta  <sebastian.perta@renesas.com>
22405         * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
22406         avoid addition with 0 and use incw and decw where possible.
22408 2018-01-26  Richard Biener  <rguenther@suse.de>
22410         PR tree-optimization/81082
22411         * fold-const.c (fold_plusminus_mult_expr): Do not perform the
22412         association if it requires casting to unsigned.
22413         * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
22414         from fold_plusminus_mult_expr to catch important cases late when
22415         range info is available.
22417 2018-01-26  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
22419         * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
22420         * configure.ac (hidden_linkonce): New test.
22421         * configure: Regenerate.
22422         * config.in: Regenerate.
22424 2018-01-26  Julia Koval  <julia.koval@intel.com>
22426         * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
22427         _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
22428         _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
22429         _mm_mask_bitshuffle_epi64_mask): Fix type.
22430         * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
22431         USI_FTYPE_V4DI_V4DI_USI): Remove.
22432         * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
22433         __builtin_ia32_vpshufbitqmb256_mask,
22434         __builtin_ia32_vpshufbitqmb128_mask): Fix types.
22435         * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
22436         * config/i386/sse.md (VI1_AVX512VLBW): Change types.
22438 2018-01-26  Alan Modra  <amodra@gmail.com>
22440         PR target/84033
22441         * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
22442         UNSPEC_VBPERMQ.  Sort other unspecs.
22444 2018-01-25  David Edelsohn  <dje.gcc@gmail.com>
22446         * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
22448 2018-01-25  Jan Hubicka  <hubicka@ucw.cz>
22450         PR middle-end/83055
22451         * predict.c (drop_profile): Do not push/pop cfun; update also
22452         node->count.
22453         (handle_missing_profiles): Fix logic looking for zero profiles.
22455 2018-01-25  Jakub Jelinek  <jakub@redhat.com>
22457         PR middle-end/83977
22458         * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
22459         on functions with #pragma omp declare simd or functions with simd
22460         attribute.
22461         * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
22462         * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
22463         Remove trailing \n from warning_at calls.
22465 2018-01-25  Tom de Vries  <tom@codesourcery.com>
22467         PR target/84028
22468         * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
22469         for neutered workers.
22471 2018-01-24  Joseph Myers  <joseph@codesourcery.com>
22473         PR target/68467
22474         * config/m68k/m68k.c (m68k_promote_function_mode): New function.
22475         (TARGET_PROMOTE_FUNCTION_MODE): New macro.
22477 2018-01-24  Jeff Law  <law@redhat.com>
22479         PR target/83994
22480         * i386.c (get_probe_interval): Move to earlier point.
22481         (ix86_compute_frame_layout): If -fstack-clash-protection and
22482         the frame is larger than the probe interval, then use pushes
22483         to save registers rather than reg->mem moves.
22484         (ix86_expand_prologue): Remove conditional for int_registers_saved
22485         assertion.
22487 2018-01-24  Vladimir Makarov  <vmakarov@redhat.com>
22489         PR target/84014
22490         * ira-build.c (setup_min_max_allocno_live_range_point): Set up
22491         min/max for never referenced object.
22493 2018-01-24  Jakub Jelinek  <jakub@redhat.com>
22495         PR middle-end/83977
22496         * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
22497         here.
22498         * omp-low.c (create_omp_child_function): Remove "omp declare simd"
22499         attributes from DECL_ATTRIBUTES (decl) without affecting
22500         DECL_ATTRIBUTES (current_function_decl).
22501         * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
22502         functions with non-NULL DECL_ABSTRACT_ORIGIN.
22504 2018-01-24  Richard Sandiford  <richard.sandiford@linaro.org>
22506         PR tree-optimization/83979
22507         * fold-const.c (fold_comparison): Use constant_boolean_node
22508         instead of boolean_{true,false}_node.
22510 2018-01-24  Jan Hubicka  <hubicka@ucw.cz>
22512         * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
22513         with zero counts.
22515 2018-01-24  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
22517         * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
22518         Simplify the clause that sets the length attribute.
22519         (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
22520         (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
22521         clause that sets the length attribute.
22522         (*sibcall_value_nonlocal_sysv<mode>): Likewise.
22524 2018-01-24  Tom de Vries  <tom@codesourcery.com>
22526         PR target/83589
22527         * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
22528         (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
22529         Add strict parameter.
22530         (prevent_branch_around_nothing): Insert dummy insn between branch to
22531         label and label with no ptx insn inbetween.
22532         * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
22534 2018-01-24  Tom de Vries  <tom@codesourcery.com>
22536         PR target/81352
22537         * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
22538         for neutered threads in warp.
22539         * config/nvptx/nvptx.md (define_insn "exit"): New insn.
22541 2018-01-24  Richard Biener  <rguenther@suse.de>
22543         PR tree-optimization/83176
22544         * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
22545         operands.
22547 2018-01-24  Richard Biener  <rguenther@suse.de>
22549         PR tree-optimization/82819
22550         * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
22551         code generating pluses that are no-ops in the target precision.
22553 2018-01-24  Richard Biener  <rguenther@suse.de>
22555         PR middle-end/84000
22556         * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
22558 2018-01-23  Jan Hubicka  <hubicka@ucw.cz>
22560         * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
22561         to merge probabilities.
22562         * predict.c (probably_never_executed): Also mark as cold functions
22563         with global 0 profile and guessed local profile.
22564         * profile-count.c (profile_probability::combine_with_count): New
22565         member function.
22566         * profile-count.h (profile_probability::operator*,
22567         profile_probability::operator*=, profile_probability::operator/,
22568         profile_probability::operator/=): Reduce precision to adjusted
22569         and set value to guessed on contradictory divisions.
22570         (profile_probability::combine_with_freq): Remove.
22571         (profile_probability::combine_wiht_count): Declare.
22572         (profile_count::force_nonzero):: Set to adjusted.
22573         (profile_count::probability_in):: Set quality to adjusted.
22574         * tree-ssa-tail-merge.c (replace_block_by): Use
22575         combine_with_count.
22577 2018-01-23  Andrew Waterman  <andrew@sifive.com>
22578             Jim Wilson  <jimw@sifive.com>
22580         * config/riscv/riscv.c (riscv_stack_boundary): New.
22581         (riscv_option_override): Set riscv_stack_boundary.  Handle
22582         riscv_preferred_stack_boundary_arg.
22583         * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
22584         (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
22585         (STACK_BOUNDARY): Set to riscv_stack_boundary.
22586         (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
22587         * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
22588         * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
22590 2018-01-23  H.J. Lu  <hongjiu.lu@intel.com>
22592         PR target/83905
22593         * config/i386/i386.c (ix86_expand_prologue): Use cost reference
22594         of struct ix86_frame.
22595         (ix86_expand_epilogue): Likewise.  Add a local variable for
22596         the reg_save_offset field in struct ix86_frame.
22598 2018-01-23  Bin Cheng  <bin.cheng@arm.com>
22600         PR tree-optimization/82604
22601         * tree-loop-distribution.c (enum partition_kind): New enum item
22602         PKIND_PARTIAL_MEMSET.
22603         (partition_builtin_p): Support above new enum item.
22604         (generate_code_for_partition): Ditto.
22605         (compute_access_range): Differentiate cases that equality can be
22606         proven at all loops, the innermost loops or no loops.
22607         (classify_builtin_st, classify_builtin_ldst): Adjust call to above
22608         function.  Set PKIND_PARTIAL_MEMSET for partition appropriately.
22609         (finalize_partitions, distribute_loop): Don't fuse partition of
22610         PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
22611         (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
22612         parloop is enabled.
22614 2018-01-23  Martin Liska  <mliska@suse.cz>
22616         * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
22617         order to ignore the predictor.
22618         (PRED_POLYMORPHIC_CALL): Likewise.
22619         (PRED_RECURSIVE_CALL): Likewise.
22621 2018-01-23  Martin Liska  <mliska@suse.cz>
22623         * tree-profile.c (tree_profiling): Print function header to
22624         aware reader which function we are working on.
22625         * value-prof.c (gimple_find_values_to_profile): Do not print
22626         not interesting value histograms.
22628 2018-01-23  Martin Liska  <mliska@suse.cz>
22630         * profile-count.h (enum profile_quality): Add
22631         profile_uninitialized as the first value. Do not number values
22632         as they are zero based.
22633         (profile_count::verify): Update sanity check.
22634         (profile_probability::verify): Likewise.
22636 2018-01-23  Nathan Sidwell  <nathan@acm.org>
22638         * doc/invoke.texi (ffor-scope): Deprecate.
22640 2018-01-23  David Malcolm  <dmalcolm@redhat.com>
22642         PR tree-optimization/83510
22643         * domwalk.c (set_all_edges_as_executable): New function.
22644         (dom_walker::dom_walker): Convert bool param
22645         "skip_unreachable_blocks" to enum reachability.  Move setup of
22646         edge flags to set_all_edges_as_executable and only do it when
22647         reachability is REACHABLE_BLOCKS.
22648         * domwalk.h (enum dom_walker::reachability): New enum.
22649         (dom_walker::dom_walker): Convert bool param
22650         "skip_unreachable_blocks" to enum reachability.
22651         (set_all_edges_as_executable): New decl.
22652         * graphite-scop-detection.c  (gather_bbs::gather_bbs): Convert
22653         from false for "skip_unreachable_blocks" to ALL_BLOCKS for
22654         "reachability".
22655         * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
22656         but converting true to REACHABLE_BLOCKS.
22657         * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
22658         * tree-vrp.c
22659         (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
22660         Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
22661         (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
22662         REACHABLE_BLOCKS.
22663         (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
22664         if check_all_array_refs will be called.
22666 2018-01-23  David Malcolm  <dmalcolm@redhat.com>
22668         * tree.c (selftest::test_location_wrappers): Add more test
22669         coverage.
22671 2018-01-23  David Malcolm  <dmalcolm@redhat.com>
22673         * sbitmap.c (selftest::test_set_range): Fix memory leaks.
22674         (selftest::test_bit_in_range): Likewise.
22676 2018-01-23  Richard Sandiford  <richard.sandiford@linaro.org>
22678         PR testsuite/83888
22679         * doc/sourcebuild.texi (vect_float): Say that the selector
22680         only describes the situation when -funsafe-math-optimizations is on.
22681         (vect_float_strict): Document.
22683 2018-01-23  Richard Sandiford  <richard.sandiford@linaro.org>
22685         PR tree-optimization/83965
22686         * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
22687         (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
22688         instead of checking only for a reduction.
22689         (vect_recog_widen_sum_pattern): Likewise.
22691 2018-01-23  Jan Hubicka  <hubicka@ucw.cz>
22693         * predict.c (probably_never_executed): Only use precise profile info.
22694         (compute_function_frequency): Skip after inlining hack since we now
22695         have quality checking.
22697 2018-01-23  Jan Hubicka  <hubicka@ucw.cz>
22699         * profile-count.h (profile_probability::very_unlikely,
22700         profile_probability::unlikely, profile_probability::even): Set
22701         precision to guessed.
22703 2018-01-23  Richard Biener  <rguenther@suse.de>
22705         PR tree-optimization/83963
22706         * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
22707         Properly terminate dominator walk when crossing the exit edge not
22708         when visiting its source block.
22710 2018-01-23  Jakub Jelinek  <jakub@redhat.com>
22712         PR c++/83918
22713         * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
22714         VIEW_CONVERT_EXPR to wrap CONST_DECLs.
22716 2018-01-22  Jakub Jelinek  <jakub@redhat.com>
22718         PR tree-optimization/83957
22719         * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs.  Remove
22720         semicolon after for body surrounded by braces.
22722         PR tree-optimization/83081
22723         * profile-count.h (profile_probability::split): New method.
22724         * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
22725         Use profile_probability::split.
22726         (do_compare_rtx_and_jump): Fix adjustment of probabilities
22727         when splitting a single conditional jump into 2.
22729 2018-01-22  David Malcolm  <dmalcolm@redhat.com>
22731         PR tree-optimization/69452
22732         * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
22733         decl.
22735 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
22737         * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
22738         * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
22739         * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
22741 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
22743         * config/rl78/rl78-protos.h (rl78_split_movdi): New function
22744         declaration.
22745         * config/rl78/rl78.md (movdi): New define_expand.
22746         * config/rl78/rl78.c (rl78_split_movdi): New function.
22748 2018-01-22  Michael Meissner  <meissner@linux.vnet.ibm.com>
22750         PR target/83862
22751         * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
22752         no longer used.
22753         * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
22754         * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
22755         128-bit to produce an UNSPEC move to get the double word with the
22756         signbit and then a shift directly to do signbit.
22757         (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
22758         implementation with a new version that just does either a direct
22759         move or a regular move.  Move memory interface to separate insns.
22760         Move insns so they are next to the expander.
22761         (signbit<mode>2_dm_mem_be): New combiner insns to combine load
22762         with signbit move.  Split big and little endian case.
22763         (signbit<mode>2_dm_mem_le): Likewise.
22764         (signbit<mode>2_dm_<su>ext): Delete, no longer used.
22765         (signbit<mode>2_dm2): Likewise.
22767 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
22769         * config/rl78/rl78.md (anddi3): New define_expand.
22771 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
22773         * config/rl78/rl78.md (umindi3): New define_expand.
22775 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
22777         * config/rl78/rl78.md (smindi3): New define_expand.
22779 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
22781         * config/rl78/rl78.md (smaxdi3): New define_expand.
22783 2018-01-22  Carl Love  <cel@us.ibm.com>
22785         * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
22786         LVX_V1TI): Add macro expansion.
22787         * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
22788         definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
22789         VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
22790         * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
22791         Change check to determine if the instruction is a byte reversing
22792         entry.  Fix typo in comment.
22793         * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
22794         for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
22795         Add def_builtin calls for new builtins.
22796         * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
22797         Add define_insn expansion.
22799 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
22801         * config/rl78/rl78.md (umaxdi3): New define_expand.
22803 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
22805         * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
22806         for non-QImode registers.
22808 2018-01-22  Richard Biener  <rguenther@suse.de>
22810         PR tree-optimization/83963
22811         * graphite-scop-detection.c (scop_detection::get_sese): Delay
22812         including the loop exit block.
22813         (scop_detection::merge_sese): Likewise.
22814         (scop_detection::add_scop): Do it here instead.
22816 2018-01-22  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
22818         * doc/sourcebuild.texi (arm_softfloat): Document.
22820 2018-01-21  John David Anglin  <danglin@gcc.gnu.org>
22822         PR gcc/77734
22823         * config/pa/pa.c (pa_function_ok_for_sibcall): Use
22824         targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
22825         Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
22827 2018-01-21  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
22828             David Edelsohn  <dje.gcc@gmail.com>
22830         PR target/83946
22831         * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
22832         Change "crset eq" to "crset 2".
22833         (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
22834         (*call_indirect_aix<mode>_nospec): Likewise.
22835         (*call_value_indirect_aix<mode>_nospec): Likewise.
22836         (*call_indirect_elfv2<mode>_nospec): Likewise.
22837         (*call_value_indirect_elfv2<mode>_nospec): Likewise.
22838         (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
22839         change assembly output from . to $.
22840         (*sibcall_value_nonlocal_sysv<mode>): Likewise.
22841         (indirect_jump<mode>_nospec): Change assembly output from . to $.
22842         (*tablejump<mode>_internal1_nospec): Likewise.
22844 2018-01-21  Oleg Endo  <olegendo@gcc.gnu.org>
22846         PR target/80870
22847         * config/sh/sh_optimize_sett_clrt.cc:
22848         Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
22850 2018-01-20  Richard Sandiford  <richard.sandiford@linaro.org>
22852         PR tree-optimization/83940
22853         * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
22854         offset_dt to vect_constant_def rather than vect_unknown_def_type.
22855         (vect_check_load_store_mask): Add a mask_dt_out parameter and
22856         use it to pass back the definition type.
22857         (vect_check_store_rhs): Likewise rhs_dt_out.
22858         (vect_build_gather_load_calls): Add a mask_dt argument and use
22859         it instead of a call to vect_is_simple_use.
22860         (vectorizable_store): Update calls to vect_check_load_store_mask
22861         and vect_check_store_rhs.  Use the dt returned by the latter instead
22862         of scatter_src_dt.  Use the cached mask_dt and gs_info.offset_dt
22863         instead of calls to vect_is_simple_use.  Pass the scalar rather
22864         than the vector operand to vect_is_simple_use when handling
22865         second and subsequent copies of an rhs value.
22866         (vectorizable_load): Update calls to vect_check_load_store_mask
22867         and vect_build_gather_load_calls.  Use the cached mask_dt and
22868         gs_info.offset_dt instead of calls to vect_is_simple_use.
22870 2018-01-20  Jakub Jelinek  <jakub@redhat.com>
22872         PR middle-end/83945
22873         * tree-emutls.c: Include gimplify.h.
22874         (lower_emutls_2): New function.
22875         (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
22876         with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
22877         it before further processing.
22879         PR target/83930
22880         * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
22881         UINTVAL (trueop1) instead of INTVAL (op1).
22883 2018-01-19  Jakub Jelinek  <jakub@redhat.com>
22885         PR debug/81570
22886         PR debug/83728
22887         * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
22888         INCOMING_FRAME_SP_OFFSET if not defined.
22889         (scan_trace): Add ENTRY argument.  If true and
22890         DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
22891         emit a note to adjust the CFA offset.
22892         (create_cfi_notes): Adjust scan_trace callers.
22893         (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
22894         INCOMING_FRAME_SP_OFFSET in the CIE.
22895         * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
22896         * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
22897         Likewise.
22898         * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
22899         * doc/tm.texi: Regenerated.
22901 2018-01-19  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
22903         PR rtl-optimization/83147
22904         * lra-constraints.c (remove_inheritance_pseudos): Use
22905         lra_substitute_pseudo_within_insn.
22907 2018-01-19  Tom de Vries  <tom@codesourcery.com>
22908             Cesar Philippidis  <cesar@codesourcery.com>
22910         PR target/83920
22911         * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
22913 2018-01-19  Cesar Philippidis  <cesar@codesourcery.com>
22915         PR target/83790
22916         * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
22917         spaces for function labels.
22919 2018-01-19  Martin Liska  <mliska@suse.cz>
22921         * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
22922         (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
22923         (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
22924         (PRED_OPCODE_POSITIVE): Change from 64 to 59.
22925         (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
22926         (PRED_CONST_RETURN): Change from 69 to 65.
22927         (PRED_NULL_RETURN): Change from 91 to 71.
22928         (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
22929         (PRED_LOOP_GUARD): Change from 66 to 73.
22931 2018-01-19  Martin Liska  <mliska@suse.cz>
22933         * predict.c (predict_insn_def): Add new assert.
22934         (struct branch_predictor): Change type to signed integer.
22935         (test_prediction_value_range): Amend test to cover
22936         PROB_UNINITIALIZED.
22937         * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
22938         (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
22939         (PRED_LOOP_ITERATIONS_MAX): Likewise.
22940         (PRED_LOOP_IV_COMPARE): Likewise.
22941         * predict.h (PROB_UNINITIALIZED): Define new constant.
22943 2018-01-19  Martin Liska  <mliska@suse.cz>
22945         * predict.c (dump_prediction): Add new format for
22946         analyze_brprob.py script which is enabled with -details
22947         suboption.
22948         * profile-count.h (precise_p): New function.
22950 2018-01-19  Richard Sandiford  <richard.sandiford@linaro.org>
22952         PR tree-optimization/83922
22953         * tree-vect-loop.c (vect_verify_full_masking): Return false if
22954         there are no statements that need masking.
22955         (vect_active_double_reduction_p): New function.
22956         (vect_analyze_loop_operations): Use it when handling phis that
22957         are not in the loop header.
22959 2018-01-19  Richard Sandiford  <richard.sandiford@linaro.org>
22961         PR tree-optimization/83914
22962         * tree-vect-loop.c (vectorizable_induction): Don't convert
22963         init_expr or apply the peeling adjustment for inductions
22964         that are nested within the vectorized loop.
22966 2018-01-19  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
22968         * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
22969         instead of NEG.
22971 2018-01-18  Jakub Jelinek  <jakub@redhat.com>
22973         PR sanitizer/81715
22974         PR testsuite/83882
22975         * function.h (gimplify_parameters): Add gimple_seq * argument.
22976         * function.c: Include gimple.h and options.h.
22977         (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
22978         for the added local temporaries if needed.
22979         * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
22980         if there are any parameter cleanups, wrap whole body into a
22981         try/finally with the cleanups.
22983 2018-01-18  Wilco Dijkstra  <wdijkstr@arm.com>
22985         PR target/82964
22986         * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
22987         Use GET_MODE_CLASS for scalar floating point.
22989 2018-01-18  Jan Hubicka  <hubicka@ucw.cz>
22991         PR ipa/82256
22992         patch by PaX Team
22993         * cgraphclones.c (cgraph_node::create_version_clone_with_body):
22994         Fix call of call_cgraph_insertion_hooks.
22996 2018-01-18  Martin Sebor  <msebor@redhat.com>
22998         * doc/invoke.texi (-Wclass-memaccess): Tweak text.
23000 2018-01-18  Jan Hubicka  <hubicka@ucw.cz>
23002         PR ipa/83619
23003         * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
23004         frequencies.
23006 2018-01-18  Boris Kolpackov  <boris@codesynthesis.com>
23008         PR other/70268
23009         * common.opt: (-ffile-prefix-map): New option.
23010         * opts.c (common_handle_option): Defer it.
23011         * opts-global.c (handle_common_deferred_options): Handle it.
23012         * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
23013         * file-prefix-map.h: New file.
23014         (remap_debug_filename, add_debug_prefix_map): ...here.
23015         (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
23016         * final.c (debug_prefix_map, add_debug_prefix_map
23017         remap_debug_filename): Move to...
23018         * file-prefix-map.c: New file.
23019         (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
23020         generalize, get rid of alloca(), use strrchr() instead of strchr().
23021         (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
23022         Implement in terms of add_prefix_map().
23023         (remap_macro_filename, remap_debug_filename): Implement in term of
23024         remap_filename().
23025         * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
23026         * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
23027         * dbxout.c: Include file-prefix-map.h.
23028         * varasm.c: Likewise.
23029         * vmsdbgout.c: Likewise.
23030         * xcoffout.c: Likewise.
23031         * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
23032         * doc/cppopts.texi (-fmacro-prefix-map): Document.
23033         * doc/invoke.texi (-ffile-prefix-map): Document.
23034         (-fdebug-prefix-map): Update description.
23036 2018-01-18  Martin Liska  <mliska@suse.cz>
23038         * config/i386/i386.c (indirect_thunk_name): Document that also
23039         lfence is emitted.
23040         (output_indirect_thunk): Document why both instructions
23041         (pause and lfence) are generated.
23043 2018-01-18  Richard Biener  <rguenther@suse.de>
23045         PR tree-optimization/83887
23046         * graphite-scop-detection.c
23047         (scop_detection::get_nearest_dom_with_single_entry): Remove.
23048         (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
23049         (scop_detection::merge_sese): Re-implement with a flood-fill
23050         algorithm that properly finds a SESE region if it exists.
23052 2018-01-18  Jakub Jelinek  <jakub@redhat.com>
23054         PR c/61240
23055         * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
23056         pointer_diff optimizations use view_convert instead of convert.
23058 2018-01-17  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
23060         * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
23061         Generate different code for -mno-speculate-indirect-jumps.
23062         (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
23063         (*call_indirect_aix<mode>): Disable for
23064         -mno-speculate-indirect-jumps.
23065         (*call_indirect_aix<mode>_nospec): New define_insn.
23066         (*call_value_indirect_aix<mode>): Disable for
23067         -mno-speculate-indirect-jumps.
23068         (*call_value_indirect_aix<mode>_nospec): New define_insn.
23069         (*sibcall_nonlocal_sysv<mode>): Generate different code for
23070         -mno-speculate-indirect-jumps.
23071         (*sibcall_value_nonlocal_sysv<mode>): Likewise.
23073 2018-01-17  Michael Meissner  <meissner@linux.vnet.ibm.com>
23075         * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
23076         long double type, set the flags for noting the default long double
23077         type, even if we don't pass or return a long double type.
23079 2018-01-17  Jan Hubicka  <hubicka@ucw.cz>
23081         PR ipa/83051
23082         * ipa-inline.c (flatten_function): Do not overwrite final inlining
23083         failure.
23085 2018-01-17  Will Schmidt  <will_schmidt@vnet.ibm.com>
23087         * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
23088         support for merge[hl].
23089         (fold_mergehl_helper): New helper function.
23090         (tree-vector-builder.h): New #include for tree_vector_builder usage.
23091         * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
23092         (altivec_vmrglw_direct): Add xxmrglw insn.
23094 2018-01-17  Andrew Waterman  <andrew@sifive.com>
23096         * config/riscv/riscv.c (riscv_conditional_register_usage): If
23097         UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
23099 2018-01-17  David Malcolm  <dmalcolm@redhat.com>
23101         PR lto/83121
23102         * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
23103         call the lto_location_cache before reading the
23104         DECL_SOURCE_LOCATION of the types.
23106 2018-01-17  Wilco Dijkstra  <wdijkstr@arm.com>
23107             Richard Sandiford  <richard.sandiford@linaro.org>
23109         * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
23110         * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
23111         (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
23112         SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
23113         * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
23114         Add declaration.
23115         * config/aarch64/constraints.md (aarch64_movti_operand):
23116         Limit immediates.
23117         * config/aarch64/predicates.md (Uti): Add new constraint.
23119 2018-01-17  Carl Love  <cel@us.ibm.com>
23121         * config/rs6000/vsx.md (define_expand xl_len_r,
23122         define_expand stxvl, define_expand *stxvl): Add match_dup argument.
23123         (define_insn): Add, match_dup 1 argument to define_insn stxvll and
23124         lxvll.
23125         (define_expand, define_insn): Move the shift left from  the
23126         define_insn to the define_expand for lxvl and stxvl instructions.
23127         * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
23128         and XL_LEN_R definitions to PURE.
23130 2018-01-17  Uros Bizjak  <ubizjak@gmail.com>
23132         * config/i386/i386.c (indirect_thunk_name): Declare regno
23133         as unsigned int.  Compare regno with INVALID_REGNUM.
23134         (output_indirect_thunk): Ditto.
23135         (output_indirect_thunk_function): Ditto.
23136         (ix86_code_end): Declare regno as unsigned int.  Use INVALID_REGNUM
23137         in the call to output_indirect_thunk_function.
23139 2018-01-17  Richard Sandiford  <richard.sandiford@linaro.org>
23141         PR middle-end/83884
23142         * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
23143         rather than the size of inner_type to determine the stack slot size
23144         when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
23146 2018-01-16  Sebastian Peryt  <sebastian.peryt@intel.com>
23148         PR target/83546
23149         * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
23150         to PTA_SILVERMONT.
23152 2018-01-16  Michael Meissner  <meissner@linux.vnet.ibm.com>
23154         * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
23155         endian Linux systems to optionally enable multilibs for selecting
23156         the long double type if the user configured an explicit type.
23157         * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
23158         have no long double multilibs if not defined.
23159         * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
23160         warn if the user used -mabi={ieee,ibm}longdouble and we built
23161         multilibs for long double.
23162         * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
23163         appropriate multilib option.
23164         (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
23165         multilib options.
23166         * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
23167         for building long double multilibs.
23168         * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
23170 2018-01-16  John David Anglin  <danglin@gcc.gnu.org>
23172         * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
23173         copies.
23175         * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
23176         64 bits.
23177         * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
23178         128 bits.
23180         * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
23181         variables.
23183         * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
23184         return value.
23186 2018-01-16  Eric Botcazou  <ebotcazou@adacore.com>
23188         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
23189         ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
23191 2018-01-16  Kelvin Nilsen  <kelvin@gcc.gnu.org>
23193         * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
23194         different rtl trees depending on TARGET_64BIT.
23195         (rs6000_gen_lvx): Likewise.
23197 2018-01-16  Eric Botcazou  <ebotcazou@adacore.com>
23199         * config/visium/visium.md (nop): Tweak comment.
23200         (hazard_nop): Likewise.
23202 2018-01-16  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
23204         * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
23205         -mspeculate-indirect-jumps.
23206         * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
23207         for -mno-speculate-indirect-jumps.
23208         (*call_indirect_elfv2<mode>_nospec): New define_insn.
23209         (*call_value_indirect_elfv2<mode>): Disable for
23210         -mno-speculate-indirect-jumps.
23211         (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
23212         (indirect_jump): Emit different RTL for
23213         -mno-speculate-indirect-jumps.
23214         (*indirect_jump<mode>): Disable for
23215         -mno-speculate-indirect-jumps.
23216         (*indirect_jump<mode>_nospec): New define_insn.
23217         (tablejump): Emit different RTL for
23218         -mno-speculate-indirect-jumps.
23219         (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
23220         (tablejumpsi_nospec): New define_expand.
23221         (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
23222         (tablejumpdi_nospec): New define_expand.
23223         (*tablejump<mode>_internal1): Disable for
23224         -mno-speculate-indirect-jumps.
23225         (*tablejump<mode>_internal1_nospec): New define_insn.
23226         * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
23227         option.
23229 2018-01-16  Artyom Skrobov tyomitch@gmail.com
23231         * caller-save.c (insert_save): Drop unnecessary parameter.  All
23232         callers updated.
23234 2018-01-16  Jakub Jelinek  <jakub@redhat.com>
23235             Richard Biener  <rguenth@suse.de>
23237         PR libgomp/83590
23238         * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
23239         return early, inline manually is_gimple_sizepos.  Make sure if we
23240         call gimplify_expr we don't end up with a gimple constant.
23241         * tree.c (variably_modified_type_p): Don't return true for
23242         is_gimple_constant (_t).  Inline manually is_gimple_sizepos.
23243         * gimplify.h (is_gimple_sizepos): Remove.
23245 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
23247         PR tree-optimization/83857
23248         * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
23249         vectorizable_live_operation for pure SLP statements.
23250         (vectorizable_live_operation): Handle PHIs.
23252 2018-01-16  Richard Biener  <rguenther@suse.de>
23254         PR tree-optimization/83867
23255         * tree-vect-stmts.c (vect_transform_stmt): Precompute
23256         nested_in_vect_loop_p since the scalar stmt may get invalidated.
23258 2018-01-16  Jakub Jelinek  <jakub@redhat.com>
23260         PR c/83844
23261         * stor-layout.c (handle_warn_if_not_align): Use byte_position and
23262         multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
23263         If off is not INTEGER_CST, issue a may not be aligned warning
23264         rather than isn't aligned.  Use isn%'t rather than isn't.
23265         * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
23266         into MULT_EXPR.
23267         <case MULT_EXPR>: Improve the case when bottom and one of the
23268         MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
23269         operand, in that case check if the other operand is multiple of
23270         bottom divided by the INTEGER_CST operand.
23272 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
23274         PR target/83858
23275         * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
23276         * config/pa/pa-protos.h (pa_function_arg_size): Declare.
23277         * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
23278         pa_function_arg_size instead of FUNCTION_ARG_SIZE.
23279         * config/pa/pa.c (pa_function_arg_advance): Likewise.
23280         (pa_function_arg, pa_arg_partial_bytes): Likewise.
23281         (pa_function_arg_size): New function.
23283 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
23285         * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
23286         in a separate statement.
23288 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
23290         PR tree-optimization/83847
23291         * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
23292         group gathers and scatters.
23294 2018-01-16  Jakub Jelinek  <jakub@redhat.com>
23296         PR rtl-optimization/86620
23297         * params.def (max-sched-ready-insns): Bump minimum value to 1.
23299         PR rtl-optimization/83213
23300         * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
23301         to last if both are JUMP_INSNs.
23303         PR tree-optimization/83843
23304         * gimple-ssa-store-merging.c
23305         (imm_store_chain_info::output_merged_store): Handle bit_not_p on
23306         store_immediate_info for bswap/nop orig_stores.
23308 2018-01-15  Andrew Waterman  <andrew@sifive.com>
23310         * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
23311         !TARGET_MUL.
23312         <UDIV>: Increase cost if !TARGET_DIV.
23314 2018-01-15  Segher Boessenkool  <segher@kernel.crashing.org>
23316         * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
23317         (define_attr "cr_logical_3op"): New.
23318         (cceq_ior_compare): Adjust.
23319         (cceq_ior_compare_complement): Adjust.
23320         (*cceq_rev_compare): Adjust.
23321         * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
23322         (is_cracked_insn): Adjust.
23323         (insn_must_be_first_in_group): Adjust.
23324         * config/rs6000/40x.md: Adjust.
23325         * config/rs6000/440.md: Adjust.
23326         * config/rs6000/476.md: Adjust.
23327         * config/rs6000/601.md: Adjust.
23328         * config/rs6000/603.md: Adjust.
23329         * config/rs6000/6xx.md: Adjust.
23330         * config/rs6000/7450.md: Adjust.
23331         * config/rs6000/7xx.md: Adjust.
23332         * config/rs6000/8540.md: Adjust.
23333         * config/rs6000/cell.md: Adjust.
23334         * config/rs6000/e300c2c3.md: Adjust.
23335         * config/rs6000/e500mc.md: Adjust.
23336         * config/rs6000/e500mc64.md: Adjust.
23337         * config/rs6000/e5500.md: Adjust.
23338         * config/rs6000/e6500.md: Adjust.
23339         * config/rs6000/mpc.md: Adjust.
23340         * config/rs6000/power4.md: Adjust.
23341         * config/rs6000/power5.md: Adjust.
23342         * config/rs6000/power6.md: Adjust.
23343         * config/rs6000/power7.md: Adjust.
23344         * config/rs6000/power8.md: Adjust.
23345         * config/rs6000/power9.md: Adjust.
23346         * config/rs6000/rs64.md: Adjust.
23347         * config/rs6000/titan.md: Adjust.
23349 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
23351         * config/i386/predicates.md (indirect_branch_operand): Rewrite
23352         ix86_indirect_branch_register logic.
23354 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
23356         * config/i386/constraints.md (Bs): Update
23357         ix86_indirect_branch_register check.  Don't check
23358         ix86_indirect_branch_register with GOT_memory_operand.
23359         (Bw): Likewise.
23360         * config/i386/predicates.md (GOT_memory_operand): Don't check
23361         ix86_indirect_branch_register here.
23362         (GOT32_symbol_operand): Likewise.
23364 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
23366         * config/i386/predicates.md (constant_call_address_operand):
23367         Rewrite ix86_indirect_branch_register logic.
23368         (sibcall_insn_operand): Likewise.
23370 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
23372         * config/i386/constraints.md (Bs): Replace
23373         ix86_indirect_branch_thunk_register with
23374         ix86_indirect_branch_register.
23375         (Bw): Likewise.
23376         * config/i386/i386.md (indirect_jump): Likewise.
23377         (tablejump): Likewise.
23378         (*sibcall_memory): Likewise.
23379         (*sibcall_value_memory): Likewise.
23380         Peepholes of indirect call and jump via memory: Likewise.
23381         * config/i386/i386.opt: Likewise.
23382         * config/i386/predicates.md (indirect_branch_operand): Likewise.
23383         (GOT_memory_operand): Likewise.
23384         (call_insn_operand): Likewise.
23385         (sibcall_insn_operand): Likewise.
23386         (GOT32_symbol_operand): Likewise.
23388 2018-01-15  Jakub Jelinek  <jakub@redhat.com>
23390         PR middle-end/83837
23391         * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
23392         type rather than type addr's type points to.
23393         (expand_omp_atomic_mutex): Likewise.
23394         (expand_omp_atomic): Likewise.
23396 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
23398         PR target/83839
23399         * config/i386/i386.c (output_indirect_thunk_function): Use
23400         ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
23401         for  __x86_return_thunk.
23403 2018-01-15  Richard Biener  <rguenther@suse.de>
23405         PR middle-end/83850
23406         * expmed.c (extract_bit_field_1): Fix typo.
23408 2018-01-15  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
23410         PR target/83687
23411         * config/arm/iterators.md (VF): New mode iterator.
23412         * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
23413         Remove integer-related logic from pattern.
23414         (neon_vabd<mode>_3): Likewise.
23416 2018-01-15  Jakub Jelinek  <jakub@redhat.com>
23418         PR middle-end/82694
23419         * common.opt (fstrict-overflow): No longer an alias.
23420         (fwrapv-pointer): New option.
23421         * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
23422         also for pointer types based on flag_wrapv_pointer.
23423         * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
23424         opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
23425         opts->x_flag_wrapv got set.
23426         * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
23427         changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
23428         POINTER_TYPE_OVERFLOW_UNDEFINED.
23429         * match.pd: Likewise in address comparison pattern.
23430         * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
23432 2018-01-15  Richard Biener  <rguenther@suse.de>
23434         PR lto/83804
23435         * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
23436         from TYPE_FIELDS.  Free TYPE_BINFO if not used by devirtualization.
23437         Reset type names to their identifier if their TYPE_DECL doesn't
23438         have linkage (and thus is used for ODR and devirt).
23439         (save_debug_info_for_decl): Remove.
23440         (save_debug_info_for_type): Likewise.
23441         (add_tree_to_fld_list): Adjust.
23442         * tree-pretty-print.c (dump_generic_node): Make dumping of
23443         type names more robust.
23445 2018-01-15  Richard Biener  <rguenther@suse.de>
23447         * BASE-VER: Bump to 8.0.1.
23449 2018-01-14  Martin Sebor  <msebor@redhat.com>
23451         PR other/83508
23452         * builtins.c (check_access): Avoid warning when the no-warning bit
23453         is set.
23455 2018-01-14  Cory Fields  <cory-nospam-@coryfields.com>
23457         * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
23458         * ira-color (allocno_hard_regs_compare): Likewise.
23460 2018-01-14  Nathan Rossi  <nathan@nathanrossi.com>
23462         PR target/83013
23463         * config/microblaze/microblaze.c (microblaze_asm_output_ident):
23464         Use .pushsection/.popsection.
23466 2018-01-14  Martin Sebor  <msebor@redhat.com>
23468         PR c++/81327
23469         * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
23471 2018-01-14  Jakub Jelinek  <jakub@redhat.com>
23473         * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
23474         entry from extra_headers.
23475         (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
23476         extra_headers, make the list bitwise identical to the i?86-*-* one.
23478 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
23480         * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
23481         -mcmodel=large with -mindirect-branch=thunk,
23482         -mindirect-branch=thunk-extern, -mfunction-return=thunk and
23483         -mfunction-return=thunk-extern.
23484         * doc/invoke.texi: Document -mcmodel=large is incompatible with
23485         -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
23486         -mfunction-return=thunk and -mfunction-return=thunk-extern.
23488 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
23490         * config/i386/i386.c (print_reg): Print the name of the full
23491         integer register without '%'.
23492         (ix86_print_operand): Handle 'V'.
23493         * doc/extend.texi: Document 'V' modifier.
23495 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
23497         * config/i386/constraints.md (Bs): Disallow memory operand for
23498         -mindirect-branch-register.
23499         (Bw): Likewise.
23500         * config/i386/predicates.md (indirect_branch_operand): Likewise.
23501         (GOT_memory_operand): Likewise.
23502         (call_insn_operand): Likewise.
23503         (sibcall_insn_operand): Likewise.
23504         (GOT32_symbol_operand): Likewise.
23505         * config/i386/i386.md (indirect_jump): Call convert_memory_address
23506         for -mindirect-branch-register.
23507         (tablejump): Likewise.
23508         (*sibcall_memory): Likewise.
23509         (*sibcall_value_memory): Likewise.
23510         Disallow peepholes of indirect call and jump via memory for
23511         -mindirect-branch-register.
23512         (*call_pop): Replace m with Bw.
23513         (*call_value_pop): Likewise.
23514         (*sibcall_pop_memory): Replace m with Bs.
23515         * config/i386/i386.opt (mindirect-branch-register): New option.
23516         * doc/invoke.texi: Document -mindirect-branch-register option.
23518 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
23520         * config/i386/i386-protos.h (ix86_output_function_return): New.
23521         * config/i386/i386.c (ix86_set_indirect_branch_type): Also
23522         set function_return_type.
23523         (indirect_thunk_name): Add ret_p to indicate thunk for function
23524         return.
23525         (output_indirect_thunk_function): Pass false to
23526         indirect_thunk_name.
23527         (ix86_output_indirect_branch_via_reg): Likewise.
23528         (ix86_output_indirect_branch_via_push): Likewise.
23529         (output_indirect_thunk_function): Create alias for function
23530         return thunk if regno < 0.
23531         (ix86_output_function_return): New function.
23532         (ix86_handle_fndecl_attribute): Handle function_return.
23533         (ix86_attribute_table): Add function_return.
23534         * config/i386/i386.h (machine_function): Add
23535         function_return_type.
23536         * config/i386/i386.md (simple_return_internal): Use
23537         ix86_output_function_return.
23538         (simple_return_internal_long): Likewise.
23539         * config/i386/i386.opt (mfunction-return=): New option.
23540         (indirect_branch): Mention -mfunction-return=.
23541         * doc/extend.texi: Document function_return function attribute.
23542         * doc/invoke.texi: Document -mfunction-return= option.
23544 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
23546         * config/i386/i386-opts.h (indirect_branch): New.
23547         * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
23548         * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
23549         with local indirect jump when converting indirect call and jump.
23550         (ix86_set_indirect_branch_type): New.
23551         (ix86_set_current_function): Call ix86_set_indirect_branch_type.
23552         (indirectlabelno): New.
23553         (indirect_thunk_needed): Likewise.
23554         (indirect_thunk_bnd_needed): Likewise.
23555         (indirect_thunks_used): Likewise.
23556         (indirect_thunks_bnd_used): Likewise.
23557         (INDIRECT_LABEL): Likewise.
23558         (indirect_thunk_name): Likewise.
23559         (output_indirect_thunk): Likewise.
23560         (output_indirect_thunk_function): Likewise.
23561         (ix86_output_indirect_branch_via_reg): Likewise.
23562         (ix86_output_indirect_branch_via_push): Likewise.
23563         (ix86_output_indirect_branch): Likewise.
23564         (ix86_output_indirect_jmp): Likewise.
23565         (ix86_code_end): Call output_indirect_thunk_function if needed.
23566         (ix86_output_call_insn): Call ix86_output_indirect_branch if
23567         needed.
23568         (ix86_handle_fndecl_attribute): Handle indirect_branch.
23569         (ix86_attribute_table): Add indirect_branch.
23570         * config/i386/i386.h (machine_function): Add indirect_branch_type
23571         and has_local_indirect_jump.
23572         * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
23573         to true.
23574         (tablejump): Likewise.
23575         (*indirect_jump): Use ix86_output_indirect_jmp.
23576         (*tablejump_1): Likewise.
23577         (simple_return_indirect_internal): Likewise.
23578         * config/i386/i386.opt (mindirect-branch=): New option.
23579         (indirect_branch): New.
23580         (keep): Likewise.
23581         (thunk): Likewise.
23582         (thunk-inline): Likewise.
23583         (thunk-extern): Likewise.
23584         * doc/extend.texi: Document indirect_branch function attribute.
23585         * doc/invoke.texi: Document -mindirect-branch= option.
23587 2018-01-14  Jan Hubicka  <hubicka@ucw.cz>
23589         PR ipa/83051
23590         * ipa-inline.c (edge_badness): Tolerate roundoff errors.
23592 2018-01-14  Richard Sandiford  <richard.sandiford@linaro.org>
23594         * ipa-inline.c (want_inline_small_function_p): Return false if
23595         inlining has already failed with CIF_FINAL_ERROR.
23596         (update_caller_keys): Call want_inline_small_function_p before
23597         can_inline_edge_p.
23598         (update_callee_keys): Likewise.
23600 2018-01-10  Kelvin Nilsen  <kelvin@gcc.gnu.org>
23602         * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
23603         New function.
23604         (rs6000_quadword_masked_address_p): Likewise.
23605         (quad_aligned_load_p): Likewise.
23606         (quad_aligned_store_p): Likewise.
23607         (const_load_sequence_p): Add comment to describe the outer-most loop.
23608         (mimic_memory_attributes_and_flags): New function.
23609         (rs6000_gen_stvx): Likewise.
23610         (replace_swapped_aligned_store): Likewise.
23611         (rs6000_gen_lvx): Likewise.
23612         (replace_swapped_aligned_load): Likewise.
23613         (replace_swapped_load_constant): Capitalize argument name in
23614         comment describing this function.
23615         (rs6000_analyze_swaps): Add a third pass to search for vector loads
23616         and stores that access quad-word aligned addresses and replace
23617         with stvx or lvx instructions when appropriate.
23618         * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
23619         New function prototype.
23620         (rs6000_quadword_masked_address_p): Likewise.
23621         (rs6000_gen_lvx): Likewise.
23622         (rs6000_gen_stvx): Likewise.
23623         * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
23624         VSX_D (V2DF, V2DI), modify this split to select lvx instruction
23625         when memory address is aligned.
23626         (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
23627         this split to select lvx instruction when memory address is aligned.
23628         (*vsx_le_perm_load_v8hi): Modify this split to select lvx
23629         instruction when memory address is aligned.
23630         (*vsx_le_perm_load_v16qi): Likewise.
23631         (four unnamed splitters): Modify to select the stvx instruction
23632         when memory is aligned.
23634 2018-01-13  Jan Hubicka  <hubicka@ucw.cz>
23636         * predict.c (determine_unlikely_bbs): Handle correctly BBs
23637         which appears in the queue multiple times.
23639 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
23640             Alan Hayward  <alan.hayward@arm.com>
23641             David Sherwood  <david.sherwood@arm.com>
23643         * tree-vectorizer.h (vec_lower_bound): New structure.
23644         (_loop_vec_info): Add check_nonzero and lower_bounds.
23645         (LOOP_VINFO_CHECK_NONZERO): New macro.
23646         (LOOP_VINFO_LOWER_BOUNDS): Likewise.
23647         (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
23648         * tree-data-ref.h (dr_with_seg_len): Add access_size and align
23649         fields.  Make seg_len the distance travelled, not including the
23650         access size.
23651         (dr_direction_indicator): Declare.
23652         (dr_zero_step_indicator): Likewise.
23653         (dr_known_forward_stride_p): Likewise.
23654         * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
23655         tree-ssanames.h.
23656         (runtime_alias_check_p): Allow runtime alias checks with
23657         variable strides.
23658         (operator ==): Compare access_size and align.
23659         (prune_runtime_alias_test_list): Rework for new distinction between
23660         the access_size and seg_len.
23661         (create_intersect_range_checks_index): Likewise.  Cope with polynomial
23662         segment lengths.
23663         (get_segment_min_max): New function.
23664         (create_intersect_range_checks): Use it.
23665         (dr_step_indicator): New function.
23666         (dr_direction_indicator): Likewise.
23667         (dr_zero_step_indicator): Likewise.
23668         (dr_known_forward_stride_p): Likewise.
23669         * tree-loop-distribution.c (data_ref_segment_size): Return
23670         DR_STEP * (niters - 1).
23671         (compute_alias_check_pairs): Update call to the dr_with_seg_len
23672         constructor.
23673         * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
23674         (vect_preserves_scalar_order_p): New function, split out from...
23675         (vect_analyze_data_ref_dependence): ...here.  Check for zero steps.
23676         (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
23677         (vect_vfa_access_size): New function.
23678         (vect_vfa_align): Likewise.
23679         (vect_compile_time_alias): Take access_size_a and access_b arguments.
23680         (dump_lower_bound): New function.
23681         (vect_check_lower_bound): Likewise.
23682         (vect_small_gap_p): Likewise.
23683         (vectorizable_with_step_bound_p): Likewise.
23684         (vect_prune_runtime_alias_test_list): Ignore cross-iteration
23685         depencies if the vectorization factor is 1.  Convert the checks
23686         for nonzero steps into checks on the bounds of DR_STEP.  Try using
23687         a bunds check for variable steps if the minimum required step is
23688         relatively small. Update calls to the dr_with_seg_len
23689         constructor and to vect_compile_time_alias.
23690         * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
23691         function.
23692         (vect_loop_versioning): Call it.
23693         * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
23694         when retrying.
23695         (vect_estimate_min_profitable_iters): Account for any bounds checks.
23697 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
23698             Alan Hayward  <alan.hayward@arm.com>
23699             David Sherwood  <david.sherwood@arm.com>
23701         * doc/sourcebuild.texi (vect_scatter_store): Document.
23702         * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
23703         optabs.
23704         * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
23705         Document.
23706         * genopinit.c (main): Add supports_vec_scatter_store and
23707         supports_vec_scatter_store_cached to target_optabs.
23708         * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
23709         IFN_MASK_SCATTER_STORE.
23710         * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
23711         functions.
23712         * internal-fn.h (internal_store_fn_p): Declare.
23713         (internal_fn_stored_value_index): Likewise.
23714         * internal-fn.c (scatter_store_direct): New macro.
23715         (expand_scatter_store_optab_fn): New function.
23716         (direct_scatter_store_optab_supported_p): New macro.
23717         (internal_store_fn_p): New function.
23718         (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
23719         IFN_MASK_SCATTER_STORE.
23720         (internal_fn_mask_index): Likewise.
23721         (internal_fn_stored_value_index): New function.
23722         (internal_gather_scatter_fn_supported_p): Adjust operand numbers
23723         for scatter stores.
23724         * optabs-query.h (supports_vec_scatter_store_p): Declare.
23725         * optabs-query.c (supports_vec_scatter_store_p): New function.
23726         * tree-vectorizer.h (vect_get_store_rhs): Declare.
23727         * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
23728         true for scatter stores.
23729         (vect_gather_scatter_fn_p): Handle scatter stores too.
23730         (vect_check_gather_scatter): Consider using scatter stores if
23731         supports_vec_scatter_store_p.
23732         * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
23733         scatter stores too.
23734         * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
23735         internal_fn_stored_value_index.
23736         (check_load_store_masking): Handle scatter stores too.
23737         (vect_get_store_rhs): Make public.
23738         (vectorizable_call): Use internal_store_fn_p.
23739         (vectorizable_store): Handle scatter store internal functions.
23740         (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
23741         when deciding whether the end of the group has been reached.
23742         * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
23743         * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
23744         (mask_scatter_store<mode>): New insns.
23746 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
23747             Alan Hayward  <alan.hayward@arm.com>
23748             David Sherwood  <david.sherwood@arm.com>
23750         * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
23751         * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
23752         * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
23753         function.
23754         (vect_use_strided_gather_scatters_p): Take a masked_p argument.
23755         Use vect_truncate_gather_scatter_offset if we can't treat the
23756         operation as a normal gather load or scatter store.
23757         (get_group_load_store_type): Take the gather_scatter_info
23758         as argument.  Try using a gather load or scatter store for
23759         single-element groups.
23760         (get_load_store_type): Update calls to get_group_load_store_type
23761         and vect_use_strided_gather_scatters_p.
23763 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
23764             Alan Hayward  <alan.hayward@arm.com>
23765             David Sherwood  <david.sherwood@arm.com>
23767         * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
23768         optional tree argument.
23769         * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
23770         null target hooks.
23771         (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
23772         but continue to use the current value as a fallback.
23773         (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
23774         to compare the updates.
23775         * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
23776         (get_load_store_type): Use it when handling a strided access.
23777         (vect_get_strided_load_store_ops): New function.
23778         (vect_get_data_ptr_increment): Likewise.
23779         (vectorizable_load): Handle strided gather loads.  Always pass
23780         a step to vect_create_data_ref_ptr and bump_vector_ptr.
23782 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
23783             Alan Hayward  <alan.hayward@arm.com>
23784             David Sherwood  <david.sherwood@arm.com>
23786         * doc/md.texi (gather_load@var{m}): Document.
23787         (mask_gather_load@var{m}): Likewise.
23788         * genopinit.c (main): Add supports_vec_gather_load and
23789         supports_vec_gather_load_cached to target_optabs.
23790         * optabs-tree.c (init_tree_optimization_optabs): Use
23791         ggc_cleared_alloc to allocate target_optabs.
23792         * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
23793         * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
23794         functions.
23795         * internal-fn.h (internal_load_fn_p): Declare.
23796         (internal_gather_scatter_fn_p): Likewise.
23797         (internal_fn_mask_index): Likewise.
23798         (internal_gather_scatter_fn_supported_p): Likewise.
23799         * internal-fn.c (gather_load_direct): New macro.
23800         (expand_gather_load_optab_fn): New function.
23801         (direct_gather_load_optab_supported_p): New macro.
23802         (direct_internal_fn_optab): New function.
23803         (internal_load_fn_p): Likewise.
23804         (internal_gather_scatter_fn_p): Likewise.
23805         (internal_fn_mask_index): Likewise.
23806         (internal_gather_scatter_fn_supported_p): Likewise.
23807         * optabs-query.c (supports_at_least_one_mode_p): New function.
23808         (supports_vec_gather_load_p): Likewise.
23809         * optabs-query.h (supports_vec_gather_load_p): Declare.
23810         * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
23811         and memory_type field.
23812         (NUM_PATTERNS): Bump to 15.
23813         * tree-vect-data-refs.c: Include internal-fn.h.
23814         (vect_gather_scatter_fn_p): New function.
23815         (vect_describe_gather_scatter_call): Likewise.
23816         (vect_check_gather_scatter): Try using internal functions for
23817         gather loads.  Recognize existing calls to a gather load function.
23818         (vect_analyze_data_refs): Consider using gather loads if
23819         supports_vec_gather_load_p.
23820         * tree-vect-patterns.c (vect_get_load_store_mask): New function.
23821         (vect_get_gather_scatter_offset_type): Likewise.
23822         (vect_convert_mask_for_vectype): Likewise.
23823         (vect_add_conversion_to_patterm): Likewise.
23824         (vect_try_gather_scatter_pattern): Likewise.
23825         (vect_recog_gather_scatter_pattern): New pattern recognizer.
23826         (vect_vect_recog_func_ptrs): Add it.
23827         * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
23828         internal_fn_mask_index and internal_gather_scatter_fn_p.
23829         (check_load_store_masking): Take the gather_scatter_info as an
23830         argument and handle gather loads.
23831         (vect_get_gather_scatter_ops): New function.
23832         (vectorizable_call): Check internal_load_fn_p.
23833         (vectorizable_load): Likewise.  Handle gather load internal
23834         functions.
23835         (vectorizable_store): Update call to check_load_store_masking.
23836         * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
23837         * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
23838         * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
23839         (aarch64_gather_scale_operand_d): New predicates.
23840         * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
23841         (mask_gather_load<mode>): New insns.
23843 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
23844             Alan Hayward  <alan.hayward@arm.com>
23845             David Sherwood  <david.sherwood@arm.com>
23847         * optabs.def (fold_left_plus_optab): New optab.
23848         * doc/md.texi (fold_left_plus_@var{m}): Document.
23849         * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
23850         * internal-fn.c (fold_left_direct): Define.
23851         (expand_fold_left_optab_fn): Likewise.
23852         (direct_fold_left_optab_supported_p): Likewise.
23853         * fold-const-call.c (fold_const_fold_left): New function.
23854         (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
23855         * tree-parloops.c (valid_reduction_p): New function.
23856         (gather_scalar_reductions): Use it.
23857         * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
23858         (vect_finish_replace_stmt): Declare.
23859         * tree-vect-loop.c (fold_left_reduction_fn): New function.
23860         (needs_fold_left_reduction_p): New function, split out from...
23861         (vect_is_simple_reduction): ...here.  Accept reductions that
23862         forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
23863         (vect_force_simple_reduction): Also store the reduction type in
23864         the assignment's STMT_VINFO_REDUC_TYPE.
23865         (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
23866         (merge_with_identity): New function.
23867         (vect_expand_fold_left): Likewise.
23868         (vectorize_fold_left_reduction): Likewise.
23869         (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION.  Leave the
23870         scalar phi in place for it.  Check for target support and reject
23871         cases that would reassociate the operation.  Defer the transform
23872         phase to vectorize_fold_left_reduction.
23873         * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
23874         * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
23875         (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
23877 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
23879         * tree-if-conv.c (predicate_mem_writes): Remove redundant
23880         call to ifc_temp_var.
23882 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
23883             Alan Hayward  <alan.hayward@arm.com>
23884             David Sherwood  <david.sherwood@arm.com>
23886         * target.def (legitimize_address_displacement): Take the original
23887         offset as a poly_int.
23888         * targhooks.h (default_legitimize_address_displacement): Update
23889         accordingly.
23890         * targhooks.c (default_legitimize_address_displacement): Likewise.
23891         * doc/tm.texi: Regenerate.
23892         * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
23893         as an argument, moving assert of ad->disp == ad->disp_term to...
23894         (process_address_1): ...here.  Update calls to base_plus_disp_to_reg.
23895         Try calling targetm.legitimize_address_displacement before expanding
23896         the address rather than afterwards, and adjust for the new interface.
23897         * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
23898         Match the new hook interface.  Handle SVE addresses.
23899         * config/sh/sh.c (sh_legitimize_address_displacement): Make the
23900         new hook interface.
23902 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
23904         * Makefile.in (OBJS): Add early-remat.o.
23905         * target.def (select_early_remat_modes): New hook.
23906         * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
23907         * doc/tm.texi: Regenerate.
23908         * targhooks.h (default_select_early_remat_modes): Declare.
23909         * targhooks.c (default_select_early_remat_modes): New function.
23910         * timevar.def (TV_EARLY_REMAT): New timevar.
23911         * passes.def (pass_early_remat): New pass.
23912         * tree-pass.h (make_pass_early_remat): Declare.
23913         * early-remat.c: New file.
23914         * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
23915         function.
23916         (TARGET_SELECT_EARLY_REMAT_MODES): Define.
23918 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
23919             Alan Hayward  <alan.hayward@arm.com>
23920             David Sherwood  <david.sherwood@arm.com>
23922         * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
23923         vfm1 with a bound_epilog parameter.
23924         (vect_do_peeling): Update calls accordingly, and move the prologue
23925         call earlier in the function.  Treat the base bound_epilog as 0 for
23926         fully-masked loops and retain vf - 1 for other loops.  Add 1 to
23927         this base when peeling for gaps.
23928         * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
23929         with fully-masked loops.
23930         (vect_estimate_min_profitable_iters): Handle the single peeled
23931         iteration in that case.
23933 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
23934             Alan Hayward  <alan.hayward@arm.com>
23935             David Sherwood  <david.sherwood@arm.com>
23937         * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
23938         single-element interleaving even if the size is not a power of 2.
23939         * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
23940         accesses for single-element interleaving if the group size is
23941         not a power of 2.
23943 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
23944             Alan Hayward  <alan.hayward@arm.com>
23945             David Sherwood  <david.sherwood@arm.com>
23947         * doc/md.texi (fold_extract_last_@var{m}): Document.
23948         * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
23949         * optabs.def (fold_extract_last_optab): New optab.
23950         * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
23951         * internal-fn.c (fold_extract_direct): New macro.
23952         (expand_fold_extract_optab_fn): Likewise.
23953         (direct_fold_extract_optab_supported_p): Likewise.
23954         * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
23955         * tree-vect-loop.c (vect_model_reduction_cost): Handle
23956         EXTRACT_LAST_REDUCTION.
23957         (get_initial_def_for_reduction): Do not create an initial vector
23958         for EXTRACT_LAST_REDUCTION reductions.
23959         (vectorizable_reduction): Leave the scalar phi in place for
23960         EXTRACT_LAST_REDUCTIONs.  Try using EXTRACT_LAST_REDUCTION
23961         ahead of INTEGER_INDUC_COND_REDUCTION.  Do not check for an
23962         epilogue code for EXTRACT_LAST_REDUCTION and defer the
23963         transform phase to vectorizable_condition.
23964         * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
23965         split out from...
23966         (vect_finish_stmt_generation): ...here.
23967         (vect_finish_replace_stmt): New function.
23968         (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
23969         * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
23970         pattern.
23971         * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
23973 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
23974             Alan Hayward  <alan.hayward@arm.com>
23975             David Sherwood  <david.sherwood@arm.com>
23977         * doc/md.texi (extract_last_@var{m}): Document.
23978         * optabs.def (extract_last_optab): New optab.
23979         * internal-fn.def (EXTRACT_LAST): New internal function.
23980         * internal-fn.c (cond_unary_direct): New macro.
23981         (expand_cond_unary_optab_fn): Likewise.
23982         (direct_cond_unary_optab_supported_p): Likewise.
23983         * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
23984         loops using EXTRACT_LAST.
23985         * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
23986         (extract_last_<mode>): ...this optab.
23987         (vec_extract<mode><Vel>): Update accordingly.
23989 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
23990             Alan Hayward  <alan.hayward@arm.com>
23991             David Sherwood  <david.sherwood@arm.com>
23993         * target.def (empty_mask_is_expensive): New hook.
23994         * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
23995         * doc/tm.texi: Regenerate.
23996         * targhooks.h (default_empty_mask_is_expensive): Declare.
23997         * targhooks.c (default_empty_mask_is_expensive): New function.
23998         * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
23999         if the target says that empty masks are expensive.
24000         * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
24001         New function.
24002         (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
24004 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
24005             Alan Hayward  <alan.hayward@arm.com>
24006             David Sherwood  <david.sherwood@arm.com>
24008         * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
24009         (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
24010         (vect_use_loop_mask_for_alignment_p): New function.
24011         (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
24012         * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
24013         niters_skip argument.  Make sure that the first niters_skip elements
24014         of the first iteration are inactive.
24015         (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
24016         Update call to vect_set_loop_masks_directly.
24017         (get_misalign_in_elems): New function, split out from...
24018         (vect_gen_prolog_loop_niters): ...here.
24019         (vect_update_init_of_dr): Take a code argument that specifies whether
24020         the adjustment should be added or subtracted.
24021         (vect_update_init_of_drs): Likewise.
24022         (vect_prepare_for_masked_peels): New function.
24023         (vect_do_peeling): Skip prologue peeling if we're using a mask
24024         instead.  Update call to vect_update_inits_of_drs.
24025         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
24026         mask_skip_niters.
24027         (vect_analyze_loop_2): Allow fully-masked loops with peeling for
24028         alignment.  Do not include the number of peeled iterations in
24029         the minimum threshold in that case.
24030         (vectorizable_induction): Adjust the start value down by
24031         LOOP_VINFO_MASK_SKIP_NITERS iterations.
24032         (vect_transform_loop): Call vect_prepare_for_masked_peels.
24033         Take the number of skipped iterations into account when calculating
24034         the loop bounds.
24035         * tree-vect-stmts.c (vect_gen_while_not): New function.
24037 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
24038             Alan Hayward  <alan.hayward@arm.com>
24039             David Sherwood  <david.sherwood@arm.com>
24041         * doc/sourcebuild.texi (vect_fully_masked): Document.
24042         * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
24043         default value to 0.
24044         * tree-vect-loop.c (vect_analyze_loop_costing): New function,
24045         split out from...
24046         (vect_analyze_loop_2): ...here. Don't check the vectorization
24047         factor against the number of loop iterations if the loop is
24048         fully-masked.
24050 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
24051             Alan Hayward  <alan.hayward@arm.com>
24052             David Sherwood  <david.sherwood@arm.com>
24054         * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
24055         (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
24056         (dump_groups): Update accordingly.
24057         (iv_use::mem_type): New member variable.
24058         (address_p): New function.
24059         (record_use): Add a mem_type argument and initialize the new
24060         mem_type field.
24061         (record_group_use): Add a mem_type argument.  Use address_p.
24062         Remove obsolete null checks of base_object.  Update call to record_use.
24063         (find_interesting_uses_op): Update call to record_group_use.
24064         (find_interesting_uses_cond): Likewise.
24065         (find_interesting_uses_address): Likewise.
24066         (get_mem_type_for_internal_fn): New function.
24067         (find_address_like_use): Likewise.
24068         (find_interesting_uses_stmt): Try find_address_like_use before
24069         calling find_interesting_uses_op.
24070         (addr_offset_valid_p): Use the iv mem_type field as the type
24071         of the addressed memory.
24072         (add_autoinc_candidates): Likewise.
24073         (get_address_cost): Likewise.
24074         (split_small_address_groups_p): Use address_p.
24075         (split_address_groups): Likewise.
24076         (add_iv_candidate_for_use): Likewise.
24077         (autoinc_possible_for_pair): Likewise.
24078         (rewrite_groups): Likewise.
24079         (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
24080         (determine_group_iv_cost): Update after split of USE_ADDRESS.
24081         (get_alias_ptr_type_for_ptr_address): New function.
24082         (rewrite_use_address): Rewrite address uses in calls that were
24083         identified by find_address_like_use.
24085 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
24086             Alan Hayward  <alan.hayward@arm.com>
24087             David Sherwood  <david.sherwood@arm.com>
24089         * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
24090         TARGET_MEM_REFs.
24091         * gimple-expr.h (is_gimple_addressable: Likewise.
24092         * gimple-expr.c (is_gimple_address): Likewise.
24093         * internal-fn.c (expand_call_mem_ref): New function.
24094         (expand_mask_load_optab_fn): Use it.
24095         (expand_mask_store_optab_fn): Likewise.
24097 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
24098             Alan Hayward  <alan.hayward@arm.com>
24099             David Sherwood  <david.sherwood@arm.com>
24101         * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
24102         (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
24103         (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
24104         (cond_umax@var{mode}): Document.
24105         * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
24106         (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
24107         (cond_umin_optab, cond_umax_optab): New optabs.
24108         * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
24109         (COND_IOR, COND_XOR): New internal functions.
24110         * internal-fn.h (get_conditional_internal_fn): Declare.
24111         * internal-fn.c (cond_binary_direct): New macro.
24112         (expand_cond_binary_optab_fn): Likewise.
24113         (direct_cond_binary_optab_supported_p): Likewise.
24114         (get_conditional_internal_fn): New function.
24115         * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
24116         Cope with reduction statements that are vectorized as calls rather
24117         than assignments.
24118         * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
24119         * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
24120         (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
24121         (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
24122         (UNSPEC_COND_EOR): New unspecs.
24123         (optab): Add mappings for them.
24124         (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
24125         (sve_int_op, sve_fp_op): New int attributes.
24127 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
24128             Alan Hayward  <alan.hayward@arm.com>
24129             David Sherwood  <david.sherwood@arm.com>
24131         * optabs.def (while_ult_optab): New optab.
24132         * doc/md.texi (while_ult@var{m}@var{n}): Document.
24133         * internal-fn.def (WHILE_ULT): New internal function.
24134         * internal-fn.h (direct_internal_fn_supported_p): New override
24135         that takes two types as argument.
24136         * internal-fn.c (while_direct): New macro.
24137         (expand_while_optab_fn): New function.
24138         (convert_optab_supported_p): Likewise.
24139         (direct_while_optab_supported_p): New macro.
24140         * wide-int.h (wi::udiv_ceil): New function.
24141         * tree-vectorizer.h (rgroup_masks): New structure.
24142         (vec_loop_masks): New typedef.
24143         (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
24144         and fully_masked_p.
24145         (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
24146         (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
24147         (vect_max_vf): New function.
24148         (slpeel_make_loop_iterate_ntimes): Delete.
24149         (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
24150         (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
24151         (vect_record_loop_mask, vect_get_loop_mask): Likewise.
24152         * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
24153         internal-fn.h, stor-layout.h and optabs-query.h.
24154         (vect_set_loop_mask): New function.
24155         (add_preheader_seq): Likewise.
24156         (add_header_seq): Likewise.
24157         (interleave_supported_p): Likewise.
24158         (vect_maybe_permute_loop_masks): Likewise.
24159         (vect_set_loop_masks_directly): Likewise.
24160         (vect_set_loop_condition_masked): Likewise.
24161         (vect_set_loop_condition_unmasked): New function, split out from
24162         slpeel_make_loop_iterate_ntimes.
24163         (slpeel_make_loop_iterate_ntimes): Rename to..
24164         (vect_set_loop_condition): ...this.  Use vect_set_loop_condition_masked
24165         for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
24166         (vect_do_peeling): Update call accordingly.
24167         (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
24168         loops.
24169         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
24170         mask_compare_type, can_fully_mask_p and fully_masked_p.
24171         (release_vec_loop_masks): New function.
24172         (_loop_vec_info): Use it to free the loop masks.
24173         (can_produce_all_loop_masks_p): New function.
24174         (vect_get_max_nscalars_per_iter): Likewise.
24175         (vect_verify_full_masking): Likewise.
24176         (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
24177         retries, and free the mask rgroups before retrying.  Check loop-wide
24178         reasons for disallowing fully-masked loops.  Make the final decision
24179         about whether use a fully-masked loop or not.
24180         (vect_estimate_min_profitable_iters): Do not assume that peeling
24181         for the number of iterations will be needed for fully-masked loops.
24182         (vectorizable_reduction): Disable fully-masked loops.
24183         (vectorizable_live_operation): Likewise.
24184         (vect_halve_mask_nunits): New function.
24185         (vect_double_mask_nunits): Likewise.
24186         (vect_record_loop_mask): Likewise.
24187         (vect_get_loop_mask): Likewise.
24188         (vect_transform_loop): Handle the case in which the final loop
24189         iteration might handle a partial vector.  Call vect_set_loop_condition
24190         instead of slpeel_make_loop_iterate_ntimes.
24191         * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
24192         (check_load_store_masking): New function.
24193         (prepare_load_store_mask): Likewise.
24194         (vectorizable_store): Handle fully-masked loops.
24195         (vectorizable_load): Likewise.
24196         (supportable_widening_operation): Use vect_halve_mask_nunits for
24197         booleans.
24198         (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
24199         (vect_gen_while): New function.
24200         * config/aarch64/aarch64.md (umax<mode>3): New expander.
24201         (aarch64_uqdec<mode>): New insn.
24203 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
24204             Alan Hayward  <alan.hayward@arm.com>
24205             David Sherwood  <david.sherwood@arm.com>
24207         * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
24208         (reduc_xor_scal_optab): New optabs.
24209         * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
24210         (reduc_xor_scal_@var{m}): Document.
24211         * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
24212         * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
24213         internal functions.
24214         * fold-const-call.c (fold_const_call): Handle them.
24215         * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
24216         internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
24217         * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
24218         (*reduc_<bit_reduc>_scal_<mode>): New patterns.
24219         * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
24220         (UNSPEC_XORV): New unspecs.
24221         (optab): Add entries for them.
24222         (BITWISEV): New int iterator.
24223         (bit_reduc_op): New int attributes.
24225 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
24226             Alan Hayward  <alan.hayward@arm.com>
24227             David Sherwood  <david.sherwood@arm.com>
24229         * doc/md.texi (vec_shl_insert_@var{m}): New optab.
24230         * internal-fn.def (VEC_SHL_INSERT): New internal function.
24231         * optabs.def (vec_shl_insert_optab): New optab.
24232         * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
24233         (duplicate_and_interleave): Likewise.
24234         * tree-vect-loop.c: Include internal-fn.h.
24235         (neutral_op_for_slp_reduction): New function, split out from
24236         get_initial_defs_for_reduction.
24237         (get_initial_def_for_reduction): Handle option 2 for variable-length
24238         vectors by loading the neutral value into a vector and then shifting
24239         the initial value into element 0.
24240         (get_initial_defs_for_reduction): Replace the code argument with
24241         the neutral value calculated by neutral_op_for_slp_reduction.
24242         Use gimple_build_vector for constant-length vectors.
24243         Use IFN_VEC_SHL_INSERT for variable-length vectors if all
24244         but the first group_size elements have a neutral value.
24245         Use duplicate_and_interleave otherwise.
24246         (vect_create_epilog_for_reduction): Take a neutral_op parameter.
24247         Update call to get_initial_defs_for_reduction.  Handle SLP
24248         reductions for variable-length vectors by creating one vector
24249         result for each scalar result, with the elements associated
24250         with other scalar results stubbed out with the neutral value.
24251         (vectorizable_reduction): Call neutral_op_for_slp_reduction.
24252         Require IFN_VEC_SHL_INSERT for double reductions on
24253         variable-length vectors, or SLP reductions that have
24254         a neutral value.  Require can_duplicate_and_interleave_p
24255         support for variable-length unchained SLP reductions if there
24256         is no neutral value, such as for MIN/MAX reductions.  Also require
24257         the number of vector elements to be a multiple of the number of
24258         SLP statements when doing variable-length unchained SLP reductions.
24259         Update call to vect_create_epilog_for_reduction.
24260         * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
24261         and remove initial values.
24262         (duplicate_and_interleave): Make public.
24263         * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
24264         * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
24266 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
24267             Alan Hayward  <alan.hayward@arm.com>
24268             David Sherwood  <david.sherwood@arm.com>
24270         * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
24271         (can_duplicate_and_interleave_p): New function.
24272         (vect_get_and_check_slp_defs): Take the vector of statements
24273         rather than just the current one.  Remove excess parentheses.
24274         Restriction rejectinon of vect_constant_def and vect_external_def
24275         for variable-length vectors to boolean types, or types for which
24276         can_duplicate_and_interleave_p is false.
24277         (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
24278         (duplicate_and_interleave): New function.
24279         (vect_get_constant_vectors): Use gimple_build_vector for
24280         constant-length vectors and suitable variable-length constant
24281         vectors.  Use duplicate_and_interleave for other variable-length
24282         vectors.  Don't defer the update when inserting new statements.
24284 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
24285             Alan Hayward  <alan.hayward@arm.com>
24286             David Sherwood  <david.sherwood@arm.com>
24288         * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
24289         min_profitable_iters doesn't go negative.
24291 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
24292             Alan Hayward  <alan.hayward@arm.com>
24293             David Sherwood  <david.sherwood@arm.com>
24295         * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
24296         (vec_mask_store_lanes@var{m}@var{n}): Likewise.
24297         * optabs.def (vec_mask_load_lanes_optab): New optab.
24298         (vec_mask_store_lanes_optab): Likewise.
24299         * internal-fn.def (MASK_LOAD_LANES): New internal function.
24300         (MASK_STORE_LANES): Likewise.
24301         * internal-fn.c (mask_load_lanes_direct): New macro.
24302         (mask_store_lanes_direct): Likewise.
24303         (expand_mask_load_optab_fn): Handle masked operations.
24304         (expand_mask_load_lanes_optab_fn): New macro.
24305         (expand_mask_store_optab_fn): Handle masked operations.
24306         (expand_mask_store_lanes_optab_fn): New macro.
24307         (direct_mask_load_lanes_optab_supported_p): Likewise.
24308         (direct_mask_store_lanes_optab_supported_p): Likewise.
24309         * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
24310         parameter.
24311         (vect_load_lanes_supported): Likewise.
24312         * tree-vect-data-refs.c (strip_conversion): New function.
24313         (can_group_stmts_p): Likewise.
24314         (vect_analyze_data_ref_accesses): Use it instead of checking
24315         for a pair of assignments.
24316         (vect_store_lanes_supported): Take a masked_p parameter.
24317         (vect_load_lanes_supported): Likewise.
24318         * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
24319         vect_store_lanes_supported and vect_load_lanes_supported.
24320         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
24321         * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
24322         parameter.  Don't allow gaps for masked accesses.
24323         Use vect_get_store_rhs.  Update calls to vect_store_lanes_supported
24324         and vect_load_lanes_supported.
24325         (get_load_store_type): Take a masked_p parameter and update
24326         call to get_group_load_store_type.
24327         (vectorizable_store): Update call to get_load_store_type.
24328         Handle IFN_MASK_STORE_LANES.
24329         (vectorizable_load): Update call to get_load_store_type.
24330         Handle IFN_MASK_LOAD_LANES.
24332 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
24333             Alan Hayward  <alan.hayward@arm.com>
24334             David Sherwood  <david.sherwood@arm.com>
24336         * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
24337         modes for SVE.
24338         * config/aarch64/aarch64-protos.h
24339         (aarch64_sve_struct_memory_operand_p): Declare.
24340         * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
24341         (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
24342         (VPRED, vpred): Handle SVE structure modes.
24343         * config/aarch64/constraints.md (Utx): New constraint.
24344         * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
24345         (aarch64_sve_struct_nonimmediate_operand): New predicates.
24346         * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
24347         * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
24348         (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
24349         structure modes.  Split into pieces after RA.
24350         (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
24351         (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
24352         New patterns.
24353         * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
24354         SVE structure modes.
24355         (aarch64_classify_address): Likewise.
24356         (sizetochar): Move earlier in file.
24357         (aarch64_print_operand): Handle SVE register lists.
24358         (aarch64_array_mode): New function.
24359         (aarch64_sve_struct_memory_operand_p): Likewise.
24360         (TARGET_ARRAY_MODE): Redefine.
24362 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
24363             Alan Hayward  <alan.hayward@arm.com>
24364             David Sherwood  <david.sherwood@arm.com>
24366         * target.def (array_mode): New target hook.
24367         * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
24368         * doc/tm.texi: Regenerate.
24369         * hooks.h (hook_optmode_mode_uhwi_none): Declare.
24370         * hooks.c (hook_optmode_mode_uhwi_none): New function.
24371         * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
24372         targetm.array_mode.
24373         * stor-layout.c (mode_for_array): Likewise.  Support polynomial
24374         type sizes.
24376 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
24377             Alan Hayward  <alan.hayward@arm.com>
24378             David Sherwood  <david.sherwood@arm.com>
24380         * fold-const.c (fold_binary_loc): Check the argument types
24381         rather than the result type when testing for a vector operation.
24383 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
24385         * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
24386         * doc/tm.texi: Regenerate.
24388 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
24389             Alan Hayward  <alan.hayward@arm.com>
24390             David Sherwood  <david.sherwood@arm.com>
24392         * doc/invoke.texi (-msve-vector-bits=): Document new option.
24393         (sve): Document new AArch64 extension.
24394         * doc/md.texi (w): Extend the description of the AArch64
24395         constraint to include SVE vectors.
24396         (Upl, Upa): Document new AArch64 predicate constraints.
24397         * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
24398         enum.
24399         * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
24400         (msve-vector-bits=): New option.
24401         * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
24402         SVE when these are disabled.
24403         (sve): New extension.
24404         * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
24405         modes.  Adjust their number of units based on aarch64_sve_vg.
24406         (MAX_BITSIZE_MODE_ANY_MODE): Define.
24407         * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
24408         aarch64_addr_query_type.
24409         (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
24410         (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
24411         (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
24412         (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
24413         (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
24414         (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
24415         (aarch64_simd_imm_zero_p): Delete.
24416         (aarch64_check_zero_based_sve_index_immediate): Declare.
24417         (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
24418         (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
24419         (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
24420         (aarch64_sve_float_mul_immediate_p): Likewise.
24421         (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
24422         rather than an rtx.
24423         (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
24424         (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
24425         (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
24426         (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
24427         (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
24428         (aarch64_regmode_natural_size): Likewise.
24429         * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
24430         (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
24431         left one place.
24432         (AARCH64_ISA_SVE, TARGET_SVE): New macros.
24433         (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
24434         for VG and the SVE predicate registers.
24435         (V_ALIASES): Add a "z"-prefixed alias.
24436         (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
24437         (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
24438         (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
24439         (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
24440         (REG_CLASS_NAMES): Add entries for them.
24441         (REG_CLASS_CONTENTS): Likewise.  Update ALL_REGS to include VG
24442         and the predicate registers.
24443         (aarch64_sve_vg): Declare.
24444         (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
24445         (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
24446         (REGMODE_NATURAL_SIZE): Define.
24447         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
24448         SVE macros.
24449         * config/aarch64/aarch64.c: Include cfgrtl.h.
24450         (simd_immediate_info): Add a constructor for series vectors,
24451         and an associated step field.
24452         (aarch64_sve_vg): New variable.
24453         (aarch64_dbx_register_number): Handle VG and the predicate registers.
24454         (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
24455         (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
24456         (VEC_ANY_DATA, VEC_STRUCT): New constants.
24457         (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
24458         (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
24459         (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
24460         (aarch64_get_mask_mode): New functions.
24461         (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
24462         and FP_LO_REGS.  Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
24463         (aarch64_hard_regno_mode_ok): Handle VG.  Also handle the SVE
24464         predicate modes and predicate registers.  Explicitly restrict
24465         GPRs to modes of 16 bytes or smaller.  Only allow FP registers
24466         to store a vector mode if it is recognized by
24467         aarch64_classify_vector_mode.
24468         (aarch64_regmode_natural_size): New function.
24469         (aarch64_hard_regno_caller_save_mode): Return the original mode
24470         for predicates.
24471         (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
24472         (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
24473         (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
24474         (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
24475         functions.
24476         (aarch64_add_offset): Add a temp2 parameter.  Assert that temp1
24477         does not overlap dest if the function is frame-related.  Handle
24478         SVE constants.
24479         (aarch64_split_add_offset): New function.
24480         (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
24481         them aarch64_add_offset.
24482         (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
24483         and update call to aarch64_sub_sp.
24484         (aarch64_add_cfa_expression): New function.
24485         (aarch64_expand_prologue): Pass extra temporary registers to the
24486         functions above.  Handle the case in which we need to emit new
24487         DW_CFA_expressions for registers that were originally saved
24488         relative to the stack pointer, but now have to be expressed
24489         relative to the frame pointer.
24490         (aarch64_output_mi_thunk): Pass extra temporary registers to the
24491         functions above.
24492         (aarch64_expand_epilogue): Likewise.  Prevent inheritance of
24493         IP0 and IP1 values for SVE frames.
24494         (aarch64_expand_vec_series): New function.
24495         (aarch64_expand_sve_widened_duplicate): Likewise.
24496         (aarch64_expand_sve_const_vector): Likewise.
24497         (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
24498         Handle SVE constants.  Use emit_move_insn to move a force_const_mem
24499         into the register, rather than emitting a SET directly.
24500         (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
24501         (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
24502         (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
24503         (offset_9bit_signed_scaled_p): New functions.
24504         (aarch64_replicate_bitmask_imm): New function.
24505         (aarch64_bitmask_imm): Use it.
24506         (aarch64_cannot_force_const_mem): Reject expressions involving
24507         a CONST_POLY_INT.  Update call to aarch64_classify_symbol.
24508         (aarch64_classify_index): Handle SVE indices, by requiring
24509         a plain register index with a scale that matches the element size.
24510         (aarch64_classify_address): Handle SVE addresses.  Assert that
24511         the mode of the address is VOIDmode or an integer mode.
24512         Update call to aarch64_classify_symbol.
24513         (aarch64_classify_symbolic_expression): Update call to
24514         aarch64_classify_symbol.
24515         (aarch64_const_vec_all_in_range_p): New function.
24516         (aarch64_print_vector_float_operand): Likewise.
24517         (aarch64_print_operand): Handle 'N' and 'C'.  Use "zN" rather than
24518         "vN" for FP registers with SVE modes.  Handle (const ...) vectors
24519         and the FP immediates 1.0 and 0.5.
24520         (aarch64_print_address_internal): Handle SVE addresses.
24521         (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
24522         (aarch64_regno_regclass): Handle predicate registers.
24523         (aarch64_secondary_reload): Handle big-endian reloads of SVE
24524         data modes.
24525         (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
24526         (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
24527         (aarch64_convert_sve_vector_bits): New function.
24528         (aarch64_override_options): Use it to handle -msve-vector-bits=.
24529         (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
24530         rather than an rtx.
24531         (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
24532         Handle SVE vector and predicate modes.  Accept VL-based constants
24533         that need only one temporary register, and VL offsets that require
24534         no temporary registers.
24535         (aarch64_conditional_register_usage): Mark the predicate registers
24536         as fixed if SVE isn't available.
24537         (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
24538         Return true for SVE vector and predicate modes.
24539         (aarch64_simd_container_mode): Take the number of bits as a poly_int64
24540         rather than an unsigned int.  Handle SVE modes.
24541         (aarch64_preferred_simd_mode): Update call accordingly.  Handle
24542         SVE modes.
24543         (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
24544         if SVE is enabled.
24545         (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
24546         (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
24547         (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
24548         (aarch64_sve_float_mul_immediate_p): New functions.
24549         (aarch64_sve_valid_immediate): New function.
24550         (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
24551         Explicitly reject structure modes.  Check for INDEX constants.
24552         Handle PTRUE and PFALSE constants.
24553         (aarch64_check_zero_based_sve_index_immediate): New function.
24554         (aarch64_simd_imm_zero_p): Delete.
24555         (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
24556         vector modes.  Accept constants in the range of CNT[BHWD].
24557         (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
24558         ask for an Advanced SIMD mode.
24559         (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
24560         (aarch64_simd_vector_alignment): Handle SVE predicates.
24561         (aarch64_vectorize_preferred_vector_alignment): New function.
24562         (aarch64_simd_vector_alignment_reachable): Use it instead of
24563         the vector size.
24564         (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
24565         (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
24566         functions.
24567         (MAX_VECT_LEN): Delete.
24568         (expand_vec_perm_d): Add a vec_flags field.
24569         (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
24570         (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
24571         (aarch64_evpc_ext): Don't apply a big-endian lane correction
24572         for SVE modes.
24573         (aarch64_evpc_rev): Rename to...
24574         (aarch64_evpc_rev_local): ...this.  Use a predicated operation for SVE.
24575         (aarch64_evpc_rev_global): New function.
24576         (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
24577         (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
24578         MAX_VECT_LEN.
24579         (aarch64_evpc_sve_tbl): New function.
24580         (aarch64_expand_vec_perm_const_1): Update after rename of
24581         aarch64_evpc_rev.  Handle SVE permutes too, trying
24582         aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
24583         than aarch64_evpc_tbl.
24584         (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
24585         (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
24586         (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
24587         (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
24588         (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
24589         (aarch64_expand_sve_vcond): New functions.
24590         (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
24591         of aarch64_vector_mode_p.
24592         (aarch64_dwarf_poly_indeterminate_value): New function.
24593         (aarch64_compute_pressure_classes): Likewise.
24594         (aarch64_can_change_mode_class): Likewise.
24595         (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
24596         (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
24597         (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
24598         (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
24599         (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
24600         (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
24601         * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
24602         (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
24603         constraints.
24604         (Dn, Dl, Dr): Accept const as well as const_vector.
24605         (Dz): Likewise.  Compare against CONST0_RTX.
24606         * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
24607         of "vector" where appropriate.
24608         (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
24609         (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
24610         (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
24611         (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
24612         (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
24613         (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
24614         (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
24615         (v_int_equiv): Extend to SVE modes.
24616         (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
24617         mode attributes.
24618         (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
24619         (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
24620         (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
24621         (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
24622         (SVE_COND_FP_CMP): New int iterators.
24623         (perm_hilo): Handle the new unpack unspecs.
24624         (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
24625         attributes.
24626         * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
24627         (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
24628         (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
24629         (aarch64_equality_operator, aarch64_constant_vector_operand)
24630         (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
24631         (aarch64_sve_nonimmediate_operand): Likewise.
24632         (aarch64_sve_general_operand): Likewise.
24633         (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
24634         (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
24635         (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
24636         (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
24637         (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
24638         (aarch64_sve_float_arith_immediate): Likewise.
24639         (aarch64_sve_float_arith_with_sub_immediate): Likewise.
24640         (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
24641         (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
24642         (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
24643         (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
24644         (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
24645         (aarch64_sve_float_arith_operand): Likewise.
24646         (aarch64_sve_float_arith_with_sub_operand): Likewise.
24647         (aarch64_sve_float_mul_operand): Likewise.
24648         (aarch64_sve_vec_perm_operand): Likewise.
24649         (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
24650         (aarch64_mov_operand): Accept const_poly_int and const_vector.
24651         (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
24652         as well as const_vector.
24653         (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
24654         in file.  Use CONST0_RTX and CONSTM1_RTX.
24655         (aarch64_simd_or_scalar_imm_zero): Likewise.  Add match_codes.
24656         (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
24657         Use aarch64_simd_imm_zero.
24658         * config/aarch64/aarch64-sve.md: New file.
24659         * config/aarch64/aarch64.md: Include it.
24660         (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
24661         (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
24662         (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
24663         (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
24664         (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
24665         (sve): New attribute.
24666         (enabled): Disable instructions with the sve attribute unless
24667         TARGET_SVE.
24668         (movqi, movhi): Pass CONST_POLY_INT operaneds through
24669         aarch64_expand_mov_immediate.
24670         (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
24671         CNT[BHSD] immediates.
24672         (movti): Split CONST_POLY_INT moves into two halves.
24673         (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
24674         Split additions that need a temporary here if the destination
24675         is the stack pointer.
24676         (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
24677         (*add<mode>3_poly_1): New instruction.
24678         (set_clobber_cc): New expander.
24680 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
24682         * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
24683         parameter and use it instead of GET_MODE_SIZE (innermode).  Use
24684         inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
24685         Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
24686         GET_MODE_NUNITS (innermode).  Also add a first_elem parameter.
24687         Change innermode from fixed_mode_size to machine_mode.
24688         (simplify_subreg): Update call accordingly.  Handle a constant-sized
24689         subreg of a variable-length CONST_VECTOR.
24691 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
24692             Alan Hayward  <alan.hayward@arm.com>
24693             David Sherwood  <david.sherwood@arm.com>
24695         * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
24696         (add_offset_to_base): New function, split out from...
24697         (create_mem_ref): ...here.  When handling a scale other than 1,
24698         check first whether the address is valid without the offset.
24699         Add it into the base if so, leaving the index and scale as-is.
24701 2018-01-12  Jakub Jelinek  <jakub@redhat.com>
24703         PR c++/83778
24704         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
24705         fold_for_warn before checking if arg2 is INTEGER_CST.
24707 2018-01-12  Segher Boessenkool  <segher@kernel.crashing.org>
24709         * config/rs6000/predicates.md (load_multiple_operation): Delete.
24710         (store_multiple_operation): Delete.
24711         * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
24712         * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
24713         * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
24714         guarded by TARGET_STRING.
24715         (rs6000_output_load_multiple): Delete.
24716         * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
24717         OPTION_MASK_STRING / TARGET_STRING handling.
24718         (print_operand) <'N', 'O'>: Add comment that these are unused now.
24719         (const rs6000_opt_masks) <"string">: Change mask to 0.
24720         * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
24721         (MASK_STRING): Delete.
24722         * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
24723         parts.  Simplify.
24724         (load_multiple): Delete.
24725         (*ldmsi8): Delete.
24726         (*ldmsi7): Delete.
24727         (*ldmsi6): Delete.
24728         (*ldmsi5): Delete.
24729         (*ldmsi4): Delete.
24730         (*ldmsi3): Delete.
24731         (store_multiple): Delete.
24732         (*stmsi8): Delete.
24733         (*stmsi7): Delete.
24734         (*stmsi6): Delete.
24735         (*stmsi5): Delete.
24736         (*stmsi4): Delete.
24737         (*stmsi3): Delete.
24738         (movmemsi_8reg): Delete.
24739         (corresponding unnamed define_insn): Delete.
24740         (movmemsi_6reg): Delete.
24741         (corresponding unnamed define_insn): Delete.
24742         (movmemsi_4reg): Delete.
24743         (corresponding unnamed define_insn): Delete.
24744         (movmemsi_2reg): Delete.
24745         (corresponding unnamed define_insn): Delete.
24746         (movmemsi_1reg): Delete.
24747         (corresponding unnamed define_insn): Delete.
24748         * config/rs6000/rs6000.opt (mno-string): New.
24749         (mstring): Replace by deprecation warning stub.
24750         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
24752 2018-01-12  Jakub Jelinek  <jakub@redhat.com>
24754         * regrename.c (regrename_do_replace): If replacing the same
24755         reg multiple times, try to reuse last created gen_raw_REG.
24757         PR debug/81155
24758         * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
24759         main to workaround a bug in GDB.
24761 2018-01-12  Tom de Vries  <tom@codesourcery.com>
24763         PR target/83737
24764         * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
24766 2018-01-12  Vladimir Makarov  <vmakarov@redhat.com>
24768         PR rtl-optimization/80481
24769         * ira-color.c (get_cap_member): New function.
24770         (allocnos_conflict_by_live_ranges_p): Use it.
24771         (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
24772         (setup_slot_coalesced_allocno_live_ranges): Ditto.
24774 2018-01-12  Uros Bizjak  <ubizjak@gmail.com>
24776         PR target/83628
24777         * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
24778         (*saddl_se_1): Ditto.
24779         (*ssubsi_1): Ditto.
24780         (*ssubl_se_1): Ditto.
24782 2018-01-12  Richard Sandiford  <richard.sandiford@linaro.org>
24784         * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
24785         rather than wi::to_widest for DR_INITs.
24786         * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
24787         wi::to_poly_offset rather than wi::to_offset for DR_INIT.
24788         (vect_analyze_data_ref_accesses): Require both DR_INITs to be
24789         INTEGER_CSTs.
24790         (vect_analyze_group_access_1): Note that here.
24792 2018-01-12  Richard Sandiford  <richard.sandiford@linaro.org>
24794         * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
24795         polynomial type sizes.
24797 2018-01-12  Richard Sandiford  <richard.sandiford@linaro.org>
24799         * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
24800         poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
24801         (gimple_add_tmp_var): Likewise.
24803 2018-01-12  Martin Liska  <mliska@suse.cz>
24805         * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
24806         (gimple_alloc_sizes): Likewise.
24807         (dump_gimple_statistics): Use PRIu64 in printf format.
24808         * gimple.h: Change uint64_t to int.
24810 2018-01-12  Martin Liska  <mliska@suse.cz>
24812         * tree-core.h: Use uint64_t instead of int.
24813         * tree.c (tree_node_counts): Likewise.
24814         (tree_node_sizes): Likewise.
24815         (dump_tree_statistics): Use PRIu64 in printf format.
24817 2018-01-12  Martin Liska  <mliska@suse.cz>
24819         * Makefile.in: As qsort_chk is implemented in vec.c, add
24820         vec.o to linkage of gencfn-macros.
24821         * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
24822         passing the info to record_node_allocation_statistics.
24823         (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
24824         and pass the info.
24825         * ggc-common.c (struct ggc_usage): Add operator== and use
24826         it in operator< and compare function.
24827         * mem-stats.h (struct mem_usage): Likewise.
24828         * vec.c (struct vec_usage): Remove operator< and compare
24829         function. Can be simply inherited.
24831 2018-01-12  Martin Jambor  <mjambor@suse.cz>
24833         PR target/81616
24834         * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
24835         * tree-ssa-math-opts.c: Include domwalk.h.
24836         (convert_mult_to_fma_1): New function.
24837         (fma_transformation_info): New type.
24838         (fma_deferring_state): Likewise.
24839         (cancel_fma_deferring): New function.
24840         (result_of_phi): Likewise.
24841         (last_fma_candidate_feeds_initial_phi): Likewise.
24842         (convert_mult_to_fma): Added deferring logic, split actual
24843         transformation to convert_mult_to_fma_1.
24844         (math_opts_dom_walker): New type.
24845         (math_opts_dom_walker::after_dom_children): New method, body moved
24846         here from pass_optimize_widening_mul::execute, added deferring logic
24847         bits.
24848         (pass_optimize_widening_mul::execute): Moved most of code to
24849         math_opts_dom_walker::after_dom_children.
24850         * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
24851         * config/i386/i386.c (ix86_option_override_internal): Added
24852         maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
24854 2018-01-12  Richard Biener  <rguenther@suse.de>
24856         PR debug/83157
24857         * dwarf2out.c (gen_variable_die): Do not reset old_die for
24858         inline instance vars.
24860 2018-01-12  Oleg Endo  <olegendo@gcc.gnu.org>
24862         PR target/81819
24863         * config/rx/rx.c (rx_is_restricted_memory_address):
24864         Handle SUBREG case.
24866 2018-01-12  Richard Biener  <rguenther@suse.de>
24868         PR tree-optimization/80846
24869         * target.def (split_reduction): New target hook.
24870         * targhooks.c (default_split_reduction): New function.
24871         * targhooks.h (default_split_reduction): Declare.
24872         * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
24873         target requests first reduce vectors by combining low and high
24874         parts.
24875         * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
24876         (get_vectype_for_scalar_type_and_size): Export.
24877         * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
24878         * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
24879         * doc/tm.texi: Regenerate.
24880         * config/i386/i386.c (ix86_split_reduction): Implement
24881         TARGET_VECTORIZE_SPLIT_REDUCTION.
24883 2018-01-12  Eric Botcazou  <ebotcazou@adacore.com>
24885         PR target/83368
24886         * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
24887         in PIC mode except for TARGET_VXWORKS_RTP.
24888         * config/sparc/sparc.c: Include cfgrtl.h.
24889         (TARGET_INIT_PIC_REG): Define.
24890         (TARGET_USE_PSEUDO_PIC_REG): Likewise.
24891         (sparc_pic_register_p): New predicate.
24892         (sparc_legitimate_address_p): Use it.
24893         (sparc_legitimize_pic_address): Likewise.
24894         (sparc_delegitimize_address): Likewise.
24895         (sparc_mode_dependent_address_p): Likewise.
24896         (gen_load_pcrel_sym): Remove 4th parameter.
24897         (load_got_register): Adjust call to above.  Remove obsolete stuff.
24898         (sparc_expand_prologue): Do not call load_got_register here.
24899         (sparc_flat_expand_prologue): Likewise.
24900         (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
24901         (sparc_use_pseudo_pic_reg): New function.
24902         (sparc_init_pic_reg): Likewise.
24903         * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
24904         (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
24906 2018-01-12  Christophe Lyon  <christophe.lyon@linaro.org>
24908         * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
24909         Add item for branch_cost.
24911 2018-01-12  Eric Botcazou  <ebotcazou@adacore.com>
24913         PR rtl-optimization/83565
24914         * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
24915         not extend the result to a larger mode for rotate operations.
24916         (num_sign_bit_copies1): Likewise.
24918 2018-01-12  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
24920         PR target/40411
24921         * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
24922         -symbolic.
24923         Use values-Xc.o for -pedantic.
24924         Link with values-xpg4.o for C90, values-xpg6.o otherwise.
24926 2018-01-12  Martin Liska  <mliska@suse.cz>
24928         PR ipa/83054
24929         * ipa-devirt.c (final_warning_record::grow_type_warnings):
24930         New function.
24931         (possible_polymorphic_call_targets): Use it.
24932         (ipa_devirt): Likewise.
24934 2018-01-12  Martin Liska  <mliska@suse.cz>
24936         * profile-count.h (enum profile_quality): Use 0 as invalid
24937         enum value of profile_quality.
24939 2018-01-12  Chung-Ju Wu  <jasonwucj@gmail.com>
24941         * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
24942         -mext-string options.
24944 2018-01-12  Richard Biener  <rguenther@suse.de>
24946         * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
24947         DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
24948         * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
24949         Likewise.
24950         * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
24952 2018-01-11  Michael Meissner  <meissner@linux.vnet.ibm.com>
24954         * configure.ac (--with-long-double-format): Add support for the
24955         configuration option to change the default long double format on
24956         PowerPC systems.
24957         * config.gcc (powerpc*-linux*-*): Likewise.
24958         * configure: Regenerate.
24959         * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
24960         double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
24961         used without modification.
24963 2018-01-11  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
24965         * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
24966         (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
24967         * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
24968         MISC_BUILTIN_SPEC_BARRIER.
24969         (rs6000_init_builtins): Likewise.
24970         * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
24971         enum value.
24972         (speculation_barrier): New define_insn.
24973         * doc/extend.texi: Document __builtin_speculation_barrier.
24975 2018-01-11  Jakub Jelinek  <jakub@redhat.com>
24977         PR target/83203
24978         * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
24979         is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
24980         * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
24981         iterators.
24982         (ssescalarmodesuffix): Add 512-bit vectors.  Use "d" or "q" for
24983         integral modes instead of "ss" and "sd".
24984         (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
24985         vectors with 32-bit and 64-bit elements.
24986         (vecdupssescalarmodesuffix): New mode attribute.
24987         (vec_dup<mode>): Use it.
24989 2018-01-11  H.J. Lu  <hongjiu.lu@intel.com>
24991         PR target/83330
24992         * config/i386/i386.c (ix86_compute_frame_layout): Align stack
24993         frame if argument is passed on stack.
24995 2018-01-11  Jakub Jelinek  <jakub@redhat.com>
24997         PR target/82682
24998         * ree.c (combine_reaching_defs): Optimize also
24999         reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
25000         reg2=any_extend(exp); reg1=reg2;, formatting fix.
25002 2018-01-11  Jan Hubicka  <hubicka@ucw.cz>
25004         PR middle-end/83189
25005         * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
25007 2018-01-11  Jan Hubicka  <hubicka@ucw.cz>
25009         PR middle-end/83718
25010         * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
25011         after they are computed.
25013 2018-01-11  Bin Cheng  <bin.cheng@arm.com>
25015         PR tree-optimization/83695
25016         * gimple-loop-linterchange.cc
25017         (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
25018         reset cached scev information after interchange.
25019         (pass_linterchange::execute): Remove call to scev_reset_htab.
25021 2018-01-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
25023         * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
25024         vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
25025         vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
25026         vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
25027         vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
25028         vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
25029         * config/arm/arm_neon_builtins.def (vfmal_lane_low,
25030         vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
25031         vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
25032         vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
25033         vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
25034         * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
25035         (V_lane_reg): Likewise.
25036         * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
25037         New define_expand.
25038         (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
25039         (vfmal_lane_low<mode>_intrinsic,
25040         vfmal_lane_low<vfmlsel2><mode>_intrinsic,
25041         vfmal_lane_high<vfmlsel2><mode>_intrinsic,
25042         vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
25043         vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
25044         vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
25045         vfmsl_lane_high<mode>_intrinsic): New define_insns.
25047 2018-01-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
25049         * config/arm/arm-cpus.in (fp16fml): New feature.
25050         (ALL_SIMD): Add fp16fml.
25051         (armv8.2-a): Add fp16fml as an option.
25052         (armv8.3-a): Likewise.
25053         (armv8.4-a): Add fp16fml as part of fp16.
25054         * config/arm/arm.h (TARGET_FP16FML): Define.
25055         * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
25056         when appropriate.
25057         * config/arm/arm-modes.def (V2HF): Define.
25058         * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
25059         vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
25060         vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
25061         * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
25062         vfmsl_low, vfmsl_high): New set of builtins.
25063         * config/arm/iterators.md (PLUSMINUS): New code iterator.
25064         (vfml_op): New code attribute.
25065         (VFMLHALVES): New int iterator.
25066         (VFML, VFMLSEL): New mode attributes.
25067         (V_reg): Define mapping for V2HF.
25068         (V_hi, V_lo): New mode attributes.
25069         (VF_constraint): Likewise.
25070         (vfml_half, vfml_half_selector): New int attributes.
25071         * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
25072         define_expand.
25073         (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
25074         vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
25075         New define_insn.
25076         * config/arm/t-arm-elf (v8_fps): Add fp16fml.
25077         * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
25078         * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
25079         * doc/invoke.texi (ARM Options): Document fp16fml.  Update armv8.4-a
25080         documentation.
25081         * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
25082         Document new effective target and option set.
25084 2018-01-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
25086         * config/arm/arm-cpus.in (armv8_4): New feature.
25087         (ARMv8_4a): New fgroup.
25088         (armv8.4-a): New arch.
25089         * config/arm/arm-tables.opt: Regenerate.
25090         * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
25091         * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
25092         * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
25093         Add matching rules for -march=armv8.4-a and extensions.
25094         * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
25096 2018-01-11  Oleg Endo  <olegendo@gcc.gnu.org>
25098         PR target/81821
25099         * config/rx/rx.md (BW): New mode attribute.
25100         (sync_lock_test_and_setsi): Add mode suffix to insn output.
25102 2018-01-11  Richard Biener  <rguenther@suse.de>
25104         PR tree-optimization/83435
25105         * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
25106         * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
25107         * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
25109 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
25110             Alan Hayward  <alan.hayward@arm.com>
25111             David Sherwood  <david.sherwood@arm.com>
25113         * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
25114         field.
25115         (aarch64_classify_address): Initialize it.  Track polynomial offsets.
25116         (aarch64_print_address_internal): Use it to check for a zero offset.
25118 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
25119             Alan Hayward  <alan.hayward@arm.com>
25120             David Sherwood  <david.sherwood@arm.com>
25122         * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
25123         * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
25124         Return a poly_int64 rather than a HOST_WIDE_INT.
25125         (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
25126         rather than a HOST_WIDE_INT.
25127         * config/aarch64/aarch64.h (aarch64_frame): Protect with
25128         HAVE_POLY_INT_H rather than HOST_WIDE_INT.  Change locals_offset,
25129         hard_fp_offset, frame_size, initial_adjust, callee_offset and
25130         final_offset from HOST_WIDE_INT to poly_int64.
25131         * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
25132         to_constant when getting the number of units in an Advanced SIMD
25133         mode.
25134         (aarch64_builtin_vectorized_function): Check for a constant number
25135         of units.
25136         * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
25137         GET_MODE_SIZE.
25138         (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
25139         attribute instead of GET_MODE_NUNITS.
25140         * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
25141         (aarch64_class_max_nregs): Use the constant_lowest_bound of the
25142         GET_MODE_SIZE for fixed-size registers.
25143         (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
25144         (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
25145         (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
25146         (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
25147         (aarch64_print_operand, aarch64_print_address_internal)
25148         (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
25149         (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
25150         (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
25151         Handle polynomial GET_MODE_SIZE.
25152         (aarch64_hard_regno_caller_save_mode): Likewise.  Return modes
25153         wider than SImode without modification.
25154         (tls_symbolic_operand_type): Use strip_offset instead of split_const.
25155         (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
25156         (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
25157         passing and returning SVE modes.
25158         (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
25159         rather than GEN_INT.
25160         (aarch64_emit_probe_stack_range): Take the size as a poly_int64
25161         rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
25162         (aarch64_allocate_and_probe_stack_space): Likewise.
25163         (aarch64_layout_frame): Cope with polynomial offsets.
25164         (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
25165         start_offset as a poly_int64 rather than a HOST_WIDE_INT.  Track
25166         polynomial offsets.
25167         (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
25168         (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
25169         poly_int64 rather than a HOST_WIDE_INT.
25170         (aarch64_get_separate_components, aarch64_process_components)
25171         (aarch64_expand_prologue, aarch64_expand_epilogue)
25172         (aarch64_use_return_insn_p): Handle polynomial frame offsets.
25173         (aarch64_anchor_offset): New function, split out from...
25174         (aarch64_legitimize_address): ...here.
25175         (aarch64_builtin_vectorization_cost): Handle polynomial
25176         TYPE_VECTOR_SUBPARTS.
25177         (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
25178         GET_MODE_NUNITS.
25179         (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
25180         number of elements from the PARALLEL rather than the mode.
25181         (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
25182         rather than GET_MODE_BITSIZE.
25183         (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
25184         (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
25185         (aarch64_expand_vec_perm_const_1): Handle polynomial
25186         d->perm.length () and d->perm elements.
25187         (aarch64_evpc_tbl): Likewise.  Use nelt rather than GET_MODE_NUNITS.
25188         Apply to_constant to d->perm elements.
25189         (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
25190         polynomial CONST_VECTOR_NUNITS.
25191         (aarch64_move_pointer): Take amount as a poly_int64 rather
25192         than an int.
25193         (aarch64_progress_pointer): Avoid temporary variable.
25194         * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
25195         the mode attribute instead of GET_MODE.
25197 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
25198             Alan Hayward  <alan.hayward@arm.com>
25199             David Sherwood  <david.sherwood@arm.com>
25201         * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
25202         x exists before using it.
25203         (aarch64_add_constant_internal): Rename to...
25204         (aarch64_add_offset_1): ...this.  Replace regnum with separate
25205         src and dest rtxes.  Handle the case in which they're different,
25206         including when the offset is zero.  Replace scratchreg with an rtx.
25207         Use 2 additions if there is no spare register into which we can
25208         move a 16-bit constant.
25209         (aarch64_add_constant): Delete.
25210         (aarch64_add_offset): Replace reg with separate src and dest
25211         rtxes.  Take a poly_int64 offset instead of a HOST_WIDE_INT.
25212         Use aarch64_add_offset_1.
25213         (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
25214         an rtx rather than an int.  Take the delta as a poly_int64
25215         rather than a HOST_WIDE_INT.  Use aarch64_add_offset.
25216         (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
25217         (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
25218         aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
25219         (aarch64_expand_epilogue): Update calls to aarch64_add_offset
25220         and aarch64_add_sp.
25221         (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
25222         aarch64_add_constant.
25224 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
25226         * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
25227         Use scalar_float_mode.
25229 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
25231         * config/aarch64/aarch64-simd.md
25232         (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
25233         (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
25234         (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
25235         (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
25236         (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
25237         (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
25238         (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
25239         (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
25240         (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
25241         (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
25243 2018-01-11  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
25245         PR target/83514
25246         * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
25247         targ_options->x_arm_arch_string is non NULL.
25249 2018-01-11  Tamar Christina  <tamar.christina@arm.com>
25251         * config/aarch64/aarch64.h
25252         (AARCH64_FL_FOR_ARCH8_4): Add  AARCH64_FL_DOTPROD.
25254 2018-01-11  Sudakshina Das  <sudi.das@arm.com>
25256         PR target/82096
25257         * expmed.c (emit_store_flag_force): Swap if const op0
25258         and change VOIDmode to mode of op0.
25260 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
25262         PR rtl-optimization/83761
25263         * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
25264         than bytes to mode_for_size.
25266 2018-01-10  Jan Hubicka  <hubicka@ucw.cz>
25268         PR middle-end/83189
25269         * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
25270         * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
25271         profile.
25273 2018-01-10  Jan Hubicka  <hubicka@ucw.cz>
25275         PR middle-end/83575
25276         * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
25277         when in layout mode.
25278         (cfg_layout_finalize): Do not verify cfg before we are out of layout.
25279         * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
25280         partition fixup.
25282 2018-01-10  Michael Collison  <michael.collison@arm.com>
25284         * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
25285         * config/aarch64/aarch64-option-extension.def: Add
25286         AARCH64_OPT_EXTENSION of 'fp16fml'.
25287         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
25288         (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
25289         * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
25290         * config/aarch64/constraints.md (Ui7): New constraint.
25291         * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
25292         (VFMLA_SEL_W): Ditto.
25293         (f16quad): Ditto.
25294         (f16mac1): Ditto.
25295         (VFMLA16_LOW): New int iterator.
25296         (VFMLA16_HIGH): Ditto.
25297         (UNSPEC_FMLAL): New unspec.
25298         (UNSPEC_FMLSL): Ditto.
25299         (UNSPEC_FMLAL2): Ditto.
25300         (UNSPEC_FMLSL2): Ditto.
25301         (f16mac): New code attribute.
25302         * config/aarch64/aarch64-simd-builtins.def
25303         (aarch64_fmlal_lowv2sf): Ditto.
25304         (aarch64_fmlsl_lowv2sf): Ditto.
25305         (aarch64_fmlalq_lowv4sf): Ditto.
25306         (aarch64_fmlslq_lowv4sf): Ditto.
25307         (aarch64_fmlal_highv2sf): Ditto.
25308         (aarch64_fmlsl_highv2sf): Ditto.
25309         (aarch64_fmlalq_highv4sf): Ditto.
25310         (aarch64_fmlslq_highv4sf): Ditto.
25311         (aarch64_fmlal_lane_lowv2sf): Ditto.
25312         (aarch64_fmlsl_lane_lowv2sf): Ditto.
25313         (aarch64_fmlal_laneq_lowv2sf): Ditto.
25314         (aarch64_fmlsl_laneq_lowv2sf): Ditto.
25315         (aarch64_fmlalq_lane_lowv4sf): Ditto.
25316         (aarch64_fmlsl_lane_lowv4sf): Ditto.
25317         (aarch64_fmlalq_laneq_lowv4sf): Ditto.
25318         (aarch64_fmlsl_laneq_lowv4sf): Ditto.
25319         (aarch64_fmlal_lane_highv2sf): Ditto.
25320         (aarch64_fmlsl_lane_highv2sf): Ditto.
25321         (aarch64_fmlal_laneq_highv2sf): Ditto.
25322         (aarch64_fmlsl_laneq_highv2sf): Ditto.
25323         (aarch64_fmlalq_lane_highv4sf): Ditto.
25324         (aarch64_fmlsl_lane_highv4sf): Ditto.
25325         (aarch64_fmlalq_laneq_highv4sf): Ditto.
25326         (aarch64_fmlsl_laneq_highv4sf): Ditto.
25327         * config/aarch64/aarch64-simd.md:
25328         (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
25329         (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
25330         (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
25331         (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
25332         (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
25333         (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
25334         (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
25335         (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
25336         (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
25337         (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
25338         (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
25339         (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
25340         (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
25341         (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
25342         (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
25343         (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
25344         (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
25345         (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
25346         (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
25347         (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
25348         * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
25349         (vfmlsl_low_u32): Ditto.
25350         (vfmlalq_low_u32): Ditto.
25351         (vfmlslq_low_u32): Ditto.
25352         (vfmlal_high_u32): Ditto.
25353         (vfmlsl_high_u32): Ditto.
25354         (vfmlalq_high_u32): Ditto.
25355         (vfmlslq_high_u32): Ditto.
25356         (vfmlal_lane_low_u32): Ditto.
25357         (vfmlsl_lane_low_u32): Ditto.
25358         (vfmlal_laneq_low_u32): Ditto.
25359         (vfmlsl_laneq_low_u32): Ditto.
25360         (vfmlalq_lane_low_u32): Ditto.
25361         (vfmlslq_lane_low_u32): Ditto.
25362         (vfmlalq_laneq_low_u32): Ditto.
25363         (vfmlslq_laneq_low_u32): Ditto.
25364         (vfmlal_lane_high_u32): Ditto.
25365         (vfmlsl_lane_high_u32): Ditto.
25366         (vfmlal_laneq_high_u32): Ditto.
25367         (vfmlsl_laneq_high_u32): Ditto.
25368         (vfmlalq_lane_high_u32): Ditto.
25369         (vfmlslq_lane_high_u32): Ditto.
25370         (vfmlalq_laneq_high_u32): Ditto.
25371         (vfmlslq_laneq_high_u32): Ditto.
25372         * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
25373         (AARCH64_FL_FOR_ARCH8_4): New.
25374         (AARCH64_ISA_F16FML): New ISA flag.
25375         (TARGET_F16FML): New feature flag for fp16fml.
25376         (doc/invoke.texi): Document new fp16fml option.
25378 2018-01-10  Michael Collison  <michael.collison@arm.com>
25380         * config/aarch64/aarch64-builtins.c:
25381         (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
25382         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
25383         (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
25384         * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
25385         (AARCH64_ISA_SHA3): New ISA flag.
25386         (TARGET_SHA3): New feature flag for sha3.
25387         * config/aarch64/iterators.md (sha512_op): New int attribute.
25388         (CRYPTO_SHA512): New int iterator.
25389         (UNSPEC_SHA512H): New unspec.
25390         (UNSPEC_SHA512H2): Ditto.
25391         (UNSPEC_SHA512SU0): Ditto.
25392         (UNSPEC_SHA512SU1): Ditto.
25393         * config/aarch64/aarch64-simd-builtins.def
25394         (aarch64_crypto_sha512hqv2di): New builtin.
25395         (aarch64_crypto_sha512h2qv2di): Ditto.
25396         (aarch64_crypto_sha512su0qv2di): Ditto.
25397         (aarch64_crypto_sha512su1qv2di): Ditto.
25398         (aarch64_eor3qv8hi): Ditto.
25399         (aarch64_rax1qv2di): Ditto.
25400         (aarch64_xarqv2di): Ditto.
25401         (aarch64_bcaxqv8hi): Ditto.
25402         * config/aarch64/aarch64-simd.md:
25403         (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
25404         (aarch64_crypto_sha512su0qv2di): Ditto.
25405         (aarch64_crypto_sha512su1qv2di): Ditto.
25406         (aarch64_eor3qv8hi): Ditto.
25407         (aarch64_rax1qv2di): Ditto.
25408         (aarch64_xarqv2di): Ditto.
25409         (aarch64_bcaxqv8hi): Ditto.
25410         * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
25411         (vsha512h2q_u64): Ditto.
25412         (vsha512su0q_u64): Ditto.
25413         (vsha512su1q_u64): Ditto.
25414         (veor3q_u16): Ditto.
25415         (vrax1q_u64): Ditto.
25416         (vxarq_u64): Ditto.
25417         (vbcaxq_u16): Ditto.
25418         * config/arm/types.md (crypto_sha512): New type attribute.
25419         (crypto_sha3): Ditto.
25420         (doc/invoke.texi): Document new sha3 option.
25422 2018-01-10  Michael Collison  <michael.collison@arm.com>
25424         * config/aarch64/aarch64-builtins.c:
25425         (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
25426         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
25427         (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
25428         (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
25429         * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
25430         (AARCH64_ISA_SM4): New ISA flag.
25431         (TARGET_SM4): New feature flag for sm4.
25432         * config/aarch64/aarch64-simd-builtins.def
25433         (aarch64_sm3ss1qv4si): Ditto.
25434         (aarch64_sm3tt1aq4si): Ditto.
25435         (aarch64_sm3tt1bq4si): Ditto.
25436         (aarch64_sm3tt2aq4si): Ditto.
25437         (aarch64_sm3tt2bq4si): Ditto.
25438         (aarch64_sm3partw1qv4si): Ditto.
25439         (aarch64_sm3partw2qv4si): Ditto.
25440         (aarch64_sm4eqv4si): Ditto.
25441         (aarch64_sm4ekeyqv4si): Ditto.
25442         * config/aarch64/aarch64-simd.md:
25443         (aarch64_sm3ss1qv4si): Ditto.
25444         (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
25445         (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
25446         (aarch64_sm4eqv4si): Ditto.
25447         (aarch64_sm4ekeyqv4si): Ditto.
25448         * config/aarch64/iterators.md (sm3tt_op): New int iterator.
25449         (sm3part_op): Ditto.
25450         (CRYPTO_SM3TT): Ditto.
25451         (CRYPTO_SM3PART): Ditto.
25452         (UNSPEC_SM3SS1): New unspec.
25453         (UNSPEC_SM3TT1A): Ditto.
25454         (UNSPEC_SM3TT1B): Ditto.
25455         (UNSPEC_SM3TT2A): Ditto.
25456         (UNSPEC_SM3TT2B): Ditto.
25457         (UNSPEC_SM3PARTW1): Ditto.
25458         (UNSPEC_SM3PARTW2): Ditto.
25459         (UNSPEC_SM4E): Ditto.
25460         (UNSPEC_SM4EKEY): Ditto.
25461         * config/aarch64/constraints.md (Ui2): New constraint.
25462         * config/aarch64/predicates.md (aarch64_imm2): New predicate.
25463         * config/arm/types.md (crypto_sm3): New type attribute.
25464         (crypto_sm4): Ditto.
25465         * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
25466         (vsm3tt1aq_u32): Ditto.
25467         (vsm3tt1bq_u32): Ditto.
25468         (vsm3tt2aq_u32): Ditto.
25469         (vsm3tt2bq_u32): Ditto.
25470         (vsm3partw1q_u32): Ditto.
25471         (vsm3partw2q_u32): Ditto.
25472         (vsm4eq_u32): Ditto.
25473         (vsm4ekeyq_u32): Ditto.
25474         (doc/invoke.texi): Document new sm4 option.
25476 2018-01-10  Michael Collison  <michael.collison@arm.com>
25478         * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
25479         * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
25480         (AARCH64_FL_FOR_ARCH8_4): New.
25481         (AARCH64_FL_V8_4): New flag.
25482         (doc/invoke.texi): Document new armv8.4-a option.
25484 2018-01-10  Michael Collison  <michael.collison@arm.com>
25486         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
25487         (__ARM_FEATURE_AES): Define if TARGET_AES is true.
25488         (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
25489         * config/aarch64/aarch64-option-extension.def: Add
25490         AARCH64_OPT_EXTENSION of 'sha2'.
25491         (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
25492         (crypto): Disable sha2 and aes if crypto disabled.
25493         (crypto): Enable aes and sha2 if enabled.
25494         (simd): Disable sha2 and aes if simd disabled.
25495         * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
25496         New flags.
25497         (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
25498         (TARGET_SHA2): New feature flag for sha2.
25499         (TARGET_AES): New feature flag for aes.
25500         * config/aarch64/aarch64-simd.md:
25501         (aarch64_crypto_aes<aes_op>v16qi): Make pattern
25502         conditional on TARGET_AES.
25503         (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
25504         (aarch64_crypto_sha1hsi): Make pattern conditional
25505         on TARGET_SHA2.
25506         (aarch64_crypto_sha1hv4si): Ditto.
25507         (aarch64_be_crypto_sha1hv4si): Ditto.
25508         (aarch64_crypto_sha1su1v4si): Ditto.
25509         (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
25510         (aarch64_crypto_sha1su0v4si): Ditto.
25511         (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
25512         (aarch64_crypto_sha256su0v4si): Ditto.
25513         (aarch64_crypto_sha256su1v4si): Ditto.
25514         (doc/invoke.texi): Document new aes and sha2 options.
25516 2018-01-10  Martin Sebor  <msebor@redhat.com>
25518         PR tree-optimization/83781
25519         * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
25520         as string arrays.
25522 2018-01-11  Martin Sebor  <msebor@gmail.com>
25523             Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
25525         PR tree-optimization/83501
25526         PR tree-optimization/81703
25528         * tree-ssa-strlen.c (get_string_cst): Rename...
25529         (get_string_len): ...to this.  Handle global constants.
25530         (handle_char_store): Adjust.
25532 2018-01-10  Kito Cheng  <kito.cheng@gmail.com>
25533             Jim Wilson  <jimw@sifive.com>
25535         * config/riscv/riscv-protos.h (riscv_output_return): New.
25536         * config/riscv/riscv.c (struct machine_function): New naked_p field.
25537         (riscv_attribute_table, riscv_output_return),
25538         (riscv_handle_fndecl_attribute, riscv_naked_function_p),
25539         (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
25540         (riscv_compute_frame_info): Only compute frame->mask if not a naked
25541         function.
25542         (riscv_expand_prologue): Add early return for naked function.
25543         (riscv_expand_epilogue): Likewise.
25544         (riscv_function_ok_for_sibcall): Return false for naked function.
25545         (riscv_set_current_function): New.
25546         (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
25547         (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
25548         * config/riscv/riscv.md (simple_return): Call riscv_output_return.
25549         * doc/extend.texi (RISC-V Function Attributes): New.
25551 2018-01-10  Michael Meissner  <meissner@linux.vnet.ibm.com>
25553         * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
25554         check for 128-bit long double before checking TCmode.
25555         * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
25556         128-bit long doubles before checking TFmode or TCmode.
25557         (FLOAT128_IBM_P): Likewise.
25559 2018-01-10  Martin Sebor  <msebor@redhat.com>
25561         PR tree-optimization/83671
25562         * builtins.c (c_strlen): Unconditionally return zero for the empty
25563         string.
25564         Use -Warray-bounds for warnings.
25565         * gimple-fold.c (get_range_strlen): Handle non-constant lengths
25566         for non-constant array indices with COMPONENT_REF, arrays of
25567         arrays, and pointers to arrays.
25568         (gimple_fold_builtin_strlen): Determine and set length range for
25569         non-constant character arrays.
25571 2018-01-10  Aldy Hernandez  <aldyh@redhat.com>
25573         PR middle-end/81897
25574         * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
25575         empty blocks.
25577 2018-01-10  Eric Botcazou  <ebotcazou@adacore.com>
25579         * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
25581 2018-01-10  Peter Bergner  <bergner@vnet.ibm.com>
25583         PR target/83399
25584         * config/rs6000/rs6000.c (print_operand) <'y'>: Use
25585         VECTOR_MEM_ALTIVEC_OR_VSX_P.
25586         * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
25587         indexed_or_indirect_operand predicate.
25588         (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
25589         (*vsx_le_perm_load_v8hi): Likewise.
25590         (*vsx_le_perm_load_v16qi): Likewise.
25591         (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
25592         (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
25593         (*vsx_le_perm_store_v8hi): Likewise.
25594         (*vsx_le_perm_store_v16qi): Likewise.
25595         (eight unnamed splitters): Likewise.
25597 2018-01-10  Peter Bergner  <bergner@vnet.ibm.com>
25599         * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
25600         * config/rs6000/emmintrin.h: Likewise.
25601         * config/rs6000/mmintrin.h: Likewise.
25602         * config/rs6000/xmmintrin.h: Likewise.
25604 2018-01-10  David Malcolm  <dmalcolm@redhat.com>
25606         PR c++/43486
25607         * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
25608         "public_flag".
25609         * tree.c (tree_nop_conversion): Return true for location wrapper
25610         nodes.
25611         (maybe_wrap_with_location): New function.
25612         (selftest::check_strip_nops): New function.
25613         (selftest::test_location_wrappers): New function.
25614         (selftest::tree_c_tests): Call it.
25615         * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
25616         (maybe_wrap_with_location): New decl.
25617         (EXPR_LOCATION_WRAPPER_P): New macro.
25618         (location_wrapper_p): New inline function.
25619         (tree_strip_any_location_wrapper): New inline function.
25621 2018-01-10  H.J. Lu  <hongjiu.lu@intel.com>
25623         PR target/83735
25624         * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
25625         stack_realign_offset for the largest alignment of stack slot
25626         actually used.
25627         (ix86_find_max_used_stack_alignment): New function.
25628         (ix86_finalize_stack_frame_flags): Use it.  Set
25629         max_used_stack_alignment if we don't realign stack.
25630         * config/i386/i386.h (machine_function): Add
25631         max_used_stack_alignment.
25633 2018-01-10  Christophe Lyon  <christophe.lyon@linaro.org>
25635         * config/arm/arm.opt (-mbranch-cost): New option.
25636         * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
25637         account.
25639 2018-01-10  Segher Boessenkool  <segher@kernel.crashing.org>
25641         PR target/83629
25642         * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
25643         load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
25645 2018-01-10  Richard Biener  <rguenther@suse.de>
25647         PR debug/83765
25648         * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
25649         early out so it also covers the case where we have a non-NULL
25650         origin.
25652 2018-01-10  Richard Sandiford  <richard.sandiford@linaro.org>
25654         PR tree-optimization/83753
25655         * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
25656         for non-strided grouped accesses if the number of elements is 1.
25658 2018-01-10  Jan Hubicka  <hubicka@ucw.cz>
25660         PR target/81616
25661         * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
25662         * i386.h (TARGET_USE_GATHER): Define.
25663         * x86-tune.def (X86_TUNE_USE_GATHER): New.
25665 2018-01-10  Martin Liska  <mliska@suse.cz>
25667         PR bootstrap/82831
25668         * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
25669         * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
25670         partitioning.
25671         * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
25672         CLEANUP_NO_PARTITIONING is not set.
25674 2018-01-10  Richard Sandiford  <richard.sandiford@linaro.org>
25676         * doc/rtl.texi: Remove documentation of (const ...) wrappers
25677         for vectors, as a partial revert of r254296.
25678         * rtl.h (const_vec_p): Delete.
25679         (const_vec_duplicate_p): Don't test for vector CONSTs.
25680         (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
25681         * expmed.c (make_tree): Likewise.
25683         Revert:
25684         * common.md (E, F): Use CONSTANT_P instead of checking for
25685         CONST_VECTOR.
25686         * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
25687         checking for CONST_VECTOR.
25689 2018-01-09  Jan Hubicka  <hubicka@ucw.cz>
25691         PR middle-end/83575
25692         * predict.c (force_edge_cold): Handle in more sane way edges
25693         with no prediction.
25695 2018-01-09  Carl Love  <cel@us.ibm.com>
25697         * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
25698         V4SI, V4SF types.
25699         (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
25700         * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
25701         VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
25702         VMRGOW_V2DI, VMRGOW_V2DF.  Remove definition for VMRGOW.
25703         * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
25704         P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW):  Add definitions.
25705         * config/rs6000/rs6000-protos.h: Add extern defition for
25706         rs6000_generate_float2_double_code.
25707         * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
25708         function.
25709         * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
25710         (float2_v2df): Add define_expand.
25712 2018-01-09  Uros Bizjak  <ubizjak@gmail.com>
25714         PR target/83628
25715         * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
25716         op_mode in the force_to_mode call.
25718 2018-01-09  Richard Sandiford  <richard.sandiford@linaro.org>
25720         * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
25721         instead of checking each element individually.
25722         (aarch64_evpc_uzp): Likewise.
25723         (aarch64_evpc_zip): Likewise.
25724         (aarch64_evpc_ext): Likewise.
25725         (aarch64_evpc_rev): Likewise.
25726         (aarch64_evpc_dup): Test the encoding for a single duplicated element,
25727         instead of checking each element individually.  Return true without
25728         generating rtl if
25729         (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
25730         whether all selected elements come from the same input, instead of
25731         checking each element individually.  Remove calls to gen_rtx_REG,
25732         start_sequence and end_sequence and instead assert that no rtl is
25733         generated.
25735 2018-01-09  Richard Sandiford  <richard.sandiford@linaro.org>
25737         * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
25738         order of HIGH and CONST checks.
25740 2018-01-09  Richard Sandiford  <richard.sandiford@linaro.org>
25742         * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
25743         if the destination isn't an SSA_NAME.
25745 2018-01-09  Richard Biener  <rguenther@suse.de>
25747         PR tree-optimization/83668
25748         * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
25749         move prologue...
25750         (canonicalize_loop_form): ... here, renamed from ...
25751         (canonicalize_loop_closed_ssa_form): ... this and amended to
25752         swap successor edges for loop exit blocks to make us use
25753         the RPO order we need for initial schedule generation.
25755 2018-01-09  Joseph Myers  <joseph@codesourcery.com>
25757         PR tree-optimization/64811
25758         * match.pd: When optimizing comparisons with Inf, avoid
25759         introducing or losing exceptions from comparisons with NaN.
25761 2018-01-09  Martin Liska  <mliska@suse.cz>
25763         PR sanitizer/82517
25764         * asan.c (shadow_mem_size): Add gcc_assert.
25766 2018-01-09  Georg-Johann Lay  <avr@gjlay.de>
25768         Don't save registers in main().
25770         PR target/83738
25771         * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
25772         * config/avr/avr.opt (-mmain-is-OS_task): New target option.
25773         * config/avr/avr.c (avr_set_current_function): Don't error if
25774         naked, OS_task or OS_main are specified at the same time.
25775         (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
25776         OS_main.
25777         (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
25778         attribute.
25779         * common/config/avr/avr-common.c (avr_option_optimization_table):
25780         Switch on -mmain-is-OS_task for optimizing compilations.
25782 2018-01-09  Richard Biener  <rguenther@suse.de>
25784         PR tree-optimization/83572
25785         * graphite.c: Include cfganal.h.
25786         (graphite_transform_loops): Connect infinite loops to exit
25787         and remove fake edges at the end.
25789 2018-01-09  Jan Hubicka  <hubicka@ucw.cz>
25791         * ipa-inline.c (edge_badness): Revert accidental checkin.
25793 2018-01-09  Jan Hubicka  <hubicka@ucw.cz>
25795         PR ipa/80763
25796         * ipa-comdats.c (set_comdat_group): Only set comdat group of real
25797         symbols; not inline clones.
25799 2018-01-09  Jakub Jelinek  <jakub@redhat.com>
25801         PR target/83507
25802         * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
25803         hard registers.  Formatting fixes.
25805         PR preprocessor/83722
25806         * gcc.c (try_generate_repro): Pass
25807         &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
25808         &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
25809         do_report_bug.
25811 2018-01-08  Monk Chiang  <sh.chiang04@gmail.com>
25812             Kito Cheng  <kito.cheng@gmail.com>
25814         * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
25815         (riscv_leaf_function_p): Delete.
25816         (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
25818 2018-01-08  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
25820         * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
25821         function.
25822         (do_ifelse): New function.
25823         (do_isel): New function.
25824         (do_sub3): New function.
25825         (do_add3): New function.
25826         (do_load_mask_compare): New function.
25827         (do_overlap_load_compare): New function.
25828         (expand_compare_loop): New function.
25829         (expand_block_compare): Call expand_compare_loop() when appropriate.
25830         * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
25831         option description.
25832         (-mblock-compare-inline-loop-limit): New option.
25834 2018-01-08  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
25836         PR target/83677
25837         * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
25838         Reverse order of second and third operands in first alternative.
25839         * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
25840         of first and second elements in UNSPEC_VPERMR vector.
25841         (altivec_expand_vec_perm_le): Likewise.
25843 2018-01-08  Jeff Law  <law@redhat.com>
25845         PR rtl-optimizatin/81308
25846         * tree-switch-conversion.c (cfg_altered): New file scoped static.
25847         (process_switch): If group_case_labels makes a change, then set
25848         cfg_altered.
25849         (pass_convert_switch::execute): If a switch is converted, then
25850         set cfg_altered.  Return TODO_cfg_cleanup if cfg_altered is true.
25852         PR rtl-optimization/81308
25853         * recog.c (split_all_insns): Conditionally cleanup the CFG after
25854         splitting insns.
25856 2018-01-08  Vidya Praveen  <vidyapraveen@arm.com>
25858         PR target/83663 - Revert r255946
25859         * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
25860         generation for cases where splatting a value is not useful.
25861         * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
25862         across a vec_duplicate and a paradoxical subreg forming a vector
25863         mode to a vec_concat.
25865 2018-01-08  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
25867         * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
25868         -march=armv8.3-a variants.
25869         * config/arm/t-multilib: Likewise.
25870         * config/arm/t-arm-elf: Likewise.  Handle dotprod extension.
25872 2018-01-08  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
25874         * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
25875         to generate rtl.
25876         (cceq_ior_compare_complement): Give it a name so I can use it, and
25877         change boolean_or_operator predicate to boolean_operator so it can
25878         be used to generate a crand.
25879         (eqne): New code iterator.
25880         (bd/bd_neg): New code_attrs.
25881         (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
25882         a single define_insn.
25883         (<bd>tf_<mode>): A new insn pattern for the conditional form branch
25884         decrement (bdnzt/bdnzf/bdzt/bdzf).
25885         * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
25886         with the new names of the branch decrement patterns, and added the
25887         names of the branch decrement conditional patterns.
25889 2018-01-08  Richard Biener  <rguenther@suse.de>
25891         PR tree-optimization/83563
25892         * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
25893         cache.
25895 2018-01-08  Richard Biener  <rguenther@suse.de>
25897         PR middle-end/83713
25898         * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
25900 2018-01-08  Richard Biener  <rguenther@suse.de>
25902         PR tree-optimization/83685
25903         * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
25904         references to abnormals.
25906 2018-01-08  Richard Biener  <rguenther@suse.de>
25908         PR lto/83719
25909         * dwarf2out.c (output_indirect_strings): Handle empty
25910         skeleton_debug_str_hash.
25911         (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
25913 2018-01-08  Claudiu Zissulescu  <claziss@synopsys.com>
25915         * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
25916         (emit_store_direct): Likewise.
25917         (arc_trampoline_adjust_address): Likewise.
25918         (arc_asm_trampoline_template): New function.
25919         (arc_initialize_trampoline): Use asm_trampoline_template.
25920         (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
25921         * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
25922         * config/arc/arc.md (flush_icache): Delete pattern.
25924 2018-01-08  Claudiu Zissulescu  <claziss@synopsys.com>
25926         * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
25927         * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
25928         munaligned-access.
25930 2018-01-08  Sebastian Huber  <sebastian.huber@embedded-brains.de>
25932         PR target/83681
25933         * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
25934         by not USED_FOR_TARGET.
25935         (make_pass_resolve_sw_modes): Likewise.
25937 2018-01-08  Sebastian Huber  <sebastian.huber@embedded-brains.de>
25939         * config/nios2/nios2.h (nios2_section_threshold): Guard by not
25940         USED_FOR_TARGET.
25942 2018-01-08  Richard Biener  <rguenther@suse.de>
25944         PR middle-end/83580
25945         * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
25947 2018-01-08  Richard Biener  <rguenther@suse.de>
25949         PR middle-end/83517
25950         * match.pd ((t * 2) / 2) -> t): Add missing :c.
25952 2018-01-06  Aldy Hernandez  <aldyh@redhat.com>
25954         PR middle-end/81897
25955         * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
25956         basic blocks with a small number of successors.
25957         (convert_control_dep_chain_into_preds): Improve handling of
25958         forwarder blocks.
25959         (dump_predicates): Split apart into...
25960         (dump_pred_chain): ...here...
25961         (dump_pred_info): ...and here.
25962         (can_one_predicate_be_invalidated_p): Add debugging printfs.
25963         (can_chain_union_be_invalidated_p): Improve check for invalidation
25964         of paths.
25965         (uninit_uses_cannot_happen): Avoid unnecessary if
25966         convert_control_dep_chain_into_preds yielded nothing.
25968 2018-01-06  Martin Sebor  <msebor@redhat.com>
25970         PR tree-optimization/83640
25971         * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
25972         subtracting negative offset from size.
25973         (builtin_access::overlap): Adjust offset bounds of the access to fall
25974         within the size of the object if possible.
25976 2018-01-06  Richard Sandiford  <richard.sandiford@linaro.org>
25978         PR rtl-optimization/83699
25979         * expmed.c (extract_bit_field_1): Restrict the vector usage of
25980         extract_bit_field_as_subreg to cases in which the extracted
25981         value is also a vector.
25983         * lra-constraints.c (process_alt_operands): Test for the equivalence
25984         substitutions when detecting a possible reload cycle.
25986 2018-01-06  Jakub Jelinek  <jakub@redhat.com>
25988         PR debug/83480
25989         * toplev.c (process_options): Don't enable debug_nonbind_markers_p
25990         by default if flag_selective_schedling{,2}.  Formatting fixes.
25992         PR rtl-optimization/83682
25993         * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
25994         if it has non-VECTOR_MODE element mode.
25995         (vec_duplicate_p): Likewise.
25997         PR middle-end/83694
25998         * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
25999         and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
26001 2018-01-05  Jakub Jelinek  <jakub@redhat.com>
26003         PR target/83604
26004         * config/i386/i386-builtin.def
26005         (__builtin_ia32_vgf2p8affineinvqb_v64qi,
26006         __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
26007         Require also OPTION_MASK_ISA_AVX512F in addition to
26008         OPTION_MASK_ISA_GFNI.
26009         (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
26010         __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
26011         OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
26012         to OPTION_MASK_ISA_GFNI.
26013         (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
26014         OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
26015         OPTION_MASK_ISA_AVX512BW.
26016         (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
26017         OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
26018         addition to OPTION_MASK_ISA_GFNI.
26019         (__builtin_ia32_vgf2p8affineinvqb_v16qi,
26020         __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
26021         Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
26022         to OPTION_MASK_ISA_GFNI.
26023         * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
26024         a requirement for all ISAs rather than any of them with a few
26025         exceptions.
26026         (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
26027         processing.
26028         (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
26029         bitmasks to be enabled with 3 exceptions, instead of requiring any
26030         enabled ISA with lots of exceptions.
26031         * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
26032         vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
26033         Change avx512bw in isa attribute to avx512f.
26034         * config/i386/sgxintrin.h: Add license boilerplate.
26035         * config/i386/vaesintrin.h: Likewise.  Fix macro spelling __AVX512F
26036         to __AVX512F__ and __AVX512VL to __AVX512VL__.
26037         (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
26038         _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
26039         defined.
26040         * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
26041         _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
26042         temporarily sse2 rather than sse if not enabled already.
26044         PR target/83604
26045         * config/i386/sse.md (VI248_VLBW): Rename to ...
26046         (VI248_AVX512VL): ... this.  Don't guard V32HI with TARGET_AVX512BW.
26047         (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
26048         vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
26049         vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
26050         vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
26051         mode iterator instead of VI248_VLBW.
26053 2018-01-05  Jan Hubicka  <hubicka@ucw.cz>
26055         * ipa-fnsummary.c (record_modified_bb_info): Add OP.
26056         (record_modified): Skip clobbers; add debug output.
26057         (param_change_prob): Use sreal frequencies.
26059 2018-01-05  Richard Sandiford  <richard.sandiford@linaro.org>
26061         * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
26062         punt for user-aligned variables.
26064 2018-01-05  Richard Sandiford  <richard.sandiford@linaro.org>
26066         * tree-chrec.c (chrec_contains_symbols): Return true for
26067         POLY_INT_CST.
26069 2018-01-05  Sudakshina Das  <sudi.das@arm.com>
26071         PR target/82439
26072         * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
26073         of (x|y) == x for BICS pattern.
26075 2018-01-05  Jakub Jelinek  <jakub@redhat.com>
26077         PR tree-optimization/83605
26078         * gimple-ssa-strength-reduction.c: Include tree-eh.h.
26079         (find_candidates_dom_walker::before_dom_children): Ignore stmts that
26080         can throw.
26082 2018-01-05  Sebastian Huber  <sebastian.huber@embedded-brains.de>
26084         * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
26085         * config/epiphany/rtems.h: New file.
26087 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
26088             Uros Bizjak  <ubizjak@gmail.com>
26090         PR target/83554
26091         * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
26092         QIreg_operand instead of register_operand predicate.
26093         * config/i386/i386.c (ix86_rop_should_change_byte_p,
26094         set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
26095         comments instead of -fmitigate[-_]rop.
26097 2018-01-04  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
26099         PR bootstrap/81926
26100         * cgraphunit.c (symbol_table::compile): Switch to text_section
26101         before calling assembly_start debug hook.
26102         * run-rtl-passes.c (run_rtl_passes): Likewise.
26103         Include output.h.
26105 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
26107         * tree-vrp.c (extract_range_from_binary_expr_1): Check
26108         range_int_cst_p rather than !symbolic_range_p before calling
26109         extract_range_from_multiplicative_op_1.
26111 2018-01-04  Jeff Law  <law@redhat.com>
26113         * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
26114         redundant test in assertion.
26116 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
26118         * doc/rtl.texi: Document machine_mode wrapper classes.
26120 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
26122         * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
26123         using tree_to_uhwi.
26125 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
26127         * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
26128         the VEC_PERM_EXPR fold to fail.
26130 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
26132         PR debug/83585
26133         * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
26134         to switched_sections.
26136 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
26138         PR target/83680
26139         * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
26140         test for d.testing.
26142 2018-01-04  Peter Bergner  <bergner@vnet.ibm.com>
26144         PR target/83387
26145         * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
26146         allow arguments in FP registers if TARGET_HARD_FLOAT is false.
26148 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
26150         PR debug/83666
26151         * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
26152         is BLKmode and bitpos not zero or mode change is needed.
26154 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
26156         PR target/83675
26157         * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
26158         TARGET_VIS2.
26160 2018-01-04  Uros Bizjak  <ubizjak@gmail.com>
26162         PR target/83628
26163         * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
26164         instead of MULT rtx.  Update all corresponding splitters.
26165         (*saddl_se): Ditto.
26166         (*ssub<modesuffix>): Ditto.
26167         (*ssubl_se): Ditto.
26168         (*cmp_sadd_di): Update split patterns.
26169         (*cmp_sadd_si): Ditto.
26170         (*cmp_sadd_sidi): Ditto.
26171         (*cmp_ssub_di): Ditto.
26172         (*cmp_ssub_si): Ditto.
26173         (*cmp_ssub_sidi): Ditto.
26174         * config/alpha/predicates.md (const23_operand): New predicate.
26175         * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
26176         Look for ASHIFT, not MULT inner operand.
26177         (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
26179 2018-01-04  Martin Liska  <mliska@suse.cz>
26181         PR gcov-profile/83669
26182         * gcov.c (output_intermediate_file): Add version to intermediate
26183         gcov file.
26184         * doc/gcov.texi: Document new field 'version' in intermediate
26185         file format. Fix location of '-k' option of gcov command.
26187 2018-01-04  Martin Liska  <mliska@suse.cz>
26189         PR ipa/82352
26190         * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
26192 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
26194         * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
26196 2018-01-03  Martin Sebor  <msebor@redhat.com>
26198         PR tree-optimization/83655
26199         * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
26200         checking calls with invalid arguments.
26202 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
26204         * tree-vect-stmts.c (vect_get_store_rhs): New function.
26205         (vectorizable_mask_load_store): Delete.
26206         (vectorizable_call): Return false for masked loads and stores.
26207         (vectorizable_store): Handle IFN_MASK_STORE.  Use vect_get_store_rhs
26208         instead of gimple_assign_rhs1.
26209         (vectorizable_load): Handle IFN_MASK_LOAD.
26210         (vect_transform_stmt): Don't set is_store for call_vec_info_type.
26212 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
26214         * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
26215         split out from..,
26216         (vectorizable_mask_load_store): ...here.
26217         (vectorizable_load): ...and here.
26219 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
26221         * tree-vect-stmts.c (vect_build_all_ones_mask)
26222         (vect_build_zero_merge_argument): New functions, split out from...
26223         (vectorizable_load): ...here.
26225 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
26227         * tree-vect-stmts.c (vect_check_store_rhs): New function,
26228         split out from...
26229         (vectorizable_mask_load_store): ...here.
26230         (vectorizable_store): ...and here.
26232 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
26234         * tree-vect-stmts.c (vect_check_load_store_mask): New function,
26235         split out from...
26236         (vectorizable_mask_load_store): ...here.
26238 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
26240         * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
26241         (vect_model_store_cost): Take a vec_load_store_type instead of a
26242         vect_def_type.
26243         * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
26244         (vect_model_store_cost): Take a vec_load_store_type instead of a
26245         vect_def_type.
26246         (vectorizable_mask_load_store): Update accordingly.
26247         (vectorizable_store): Likewise.
26248         * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
26250 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
26252         * tree-vect-loop.c (vect_transform_loop): Stub out scalar
26253         IFN_MASK_LOAD calls here rather than...
26254         * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
26256 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
26257             Alan Hayward  <alan.hayward@arm.com>
26258             David Sherwood  <david.sherwood@arm.com>
26260         * expmed.c (extract_bit_field_1): For vector extracts,
26261         fall back to extract_bit_field_as_subreg if vec_extract
26262         isn't available.
26264 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
26265             Alan Hayward  <alan.hayward@arm.com>
26266             David Sherwood  <david.sherwood@arm.com>
26268         * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
26269         they are variable or constant sized.
26270         (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
26271         slots for constant-sized data.
26273 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
26274             Alan Hayward  <alan.hayward@arm.com>
26275             David Sherwood  <david.sherwood@arm.com>
26277         * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
26278         handling COND_EXPRs with boolean comparisons, try to find a better
26279         basis for the mask type than the boolean itself.
26281 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
26283         * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
26284         is calculated and how it can be overridden.
26285         * genmodes.c (max_bitsize_mode_any_mode): New variable.
26286         (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
26287         if defined.
26288         (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
26289         if nonzero.
26291 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
26292             Alan Hayward  <alan.hayward@arm.com>
26293             David Sherwood  <david.sherwood@arm.com>
26295         * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
26296         Remove the mode argument.
26297         (aarch64_simd_valid_immediate): Remove the mode and inverse
26298         arguments.
26299         * config/aarch64/iterators.md (bitsize): New iterator.
26300         * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
26301         (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
26302         * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
26303         aarch64_simd_valid_immediate.
26304         * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
26305         (aarch64_reg_or_bic_imm): Likewise.
26306         * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
26307         with an insn_type enum and msl with a modifier_type enum.
26308         Replace element_width with a scalar_mode.  Change the shift
26309         to unsigned int.  Add constructors for scalar_float_mode and
26310         scalar_int_mode elements.
26311         (aarch64_vect_float_const_representable_p): Delete.
26312         (aarch64_can_const_movi_rtx_p)
26313         (aarch64_simd_scalar_immediate_valid_for_move)
26314         (aarch64_simd_make_constant): Update call to
26315         aarch64_simd_valid_immediate.
26316         (aarch64_advsimd_valid_immediate_hs): New function.
26317         (aarch64_advsimd_valid_immediate): Likewise.
26318         (aarch64_simd_valid_immediate): Remove mode and inverse
26319         arguments.  Rewrite to use the above.  Use const_vec_duplicate_p
26320         to detect duplicated constants and use aarch64_float_const_zero_rtx_p
26321         and aarch64_float_const_representable_p on the result.
26322         (aarch64_output_simd_mov_immediate): Remove mode argument.
26323         Update call to aarch64_simd_valid_immediate and use of
26324         simd_immediate_info.
26325         (aarch64_output_scalar_simd_mov_immediate): Update call
26326         accordingly.
26328 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
26329             Alan Hayward  <alan.hayward@arm.com>
26330             David Sherwood  <david.sherwood@arm.com>
26332         * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
26333         (mode_nunits): Likewise CONST_MODE_NUNITS.
26334         * machmode.def (ADJUST_NUNITS): Document.
26335         * genmodes.c (mode_data::need_nunits_adj): New field.
26336         (blank_mode): Update accordingly.
26337         (adj_nunits): New variable.
26338         (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
26339         parameter.
26340         (emit_mode_size_inline): Set need_bytesize_adj for all modes
26341         listed in adj_nunits.
26342         (emit_mode_nunits_inline): Set need_nunits_adj for all modes
26343         listed in adj_nunits.  Don't emit case statements for such modes.
26344         (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
26345         and CONST_MODE_PRECISION.  Make CONST_MODE_SIZE expand to
26346         nothing if adj_nunits is nonnull.
26347         (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
26348         (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
26349         (emit_mode_fbit): Update use of print_maybe_const_decl.
26350         (emit_move_size): Likewise.  Treat the array as non-const
26351         if adj_nunits.
26352         (emit_mode_adjustments): Handle adj_nunits.
26354 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
26356         * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
26357         * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
26358         (VECTOR_MODES): Use it.
26359         (make_vector_modes): Take the prefix as an argument.
26361 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
26362             Alan Hayward  <alan.hayward@arm.com>
26363             David Sherwood  <david.sherwood@arm.com>
26365         * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
26366         * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
26367         for MODE_VECTOR_BOOL.
26368         * machmode.def (VECTOR_BOOL_MODE): Document.
26369         * genmodes.c (VECTOR_BOOL_MODE): New macro.
26370         (make_vector_bool_mode): New function.
26371         (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
26372         MODE_VECTOR_BOOL.
26373         * lto-streamer-in.c (lto_input_mode_table): Likewise.
26374         * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
26375         Likewise.
26376         * stor-layout.c (int_mode_for_mode): Likewise.
26377         * tree.c (build_vector_type_for_mode): Likewise.
26378         * varasm.c (output_constant_pool_2): Likewise.
26379         * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
26380         CONSTM1_RTX (BImode) are the same thing.  Initialize const_tiny_rtx
26381         for MODE_VECTOR_BOOL.
26382         * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
26383         of mode class checks.
26384         * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
26385         instead of a list of mode class checks.
26386         (expand_vector_scalar_condition): Likewise.
26387         (type_for_widest_vector_mode): Handle BImode as an inner mode.
26389 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
26390             Alan Hayward  <alan.hayward@arm.com>
26391             David Sherwood  <david.sherwood@arm.com>
26393         * machmode.h (mode_size): Change from unsigned short to
26394         poly_uint16_pod.
26395         (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
26396         (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
26397         or if measurement_type is not polynomial.
26398         (fixed_size_mode::includes_p): Check for constant-sized modes.
26399         * genmodes.c (emit_mode_size_inline): Make mode_size_inline
26400         return a poly_uint16 rather than an unsigned short.
26401         (emit_mode_size): Change the type of mode_size from unsigned short
26402         to poly_uint16_pod.  Use ZERO_COEFFS for the initializer.
26403         (emit_mode_adjustments): Cope with polynomial vector sizes.
26404         * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
26405         for GET_MODE_SIZE.
26406         * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
26407         for GET_MODE_SIZE.
26408         * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
26409         * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
26410         * caller-save.c (setup_save_areas): Likewise.
26411         (replace_reg_with_saved_mem): Likewise.
26412         * calls.c (emit_library_call_value_1): Likewise.
26413         * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
26414         * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
26415         (gen_lowpart_for_combine): Likewise.
26416         * convert.c (convert_to_integer_1): Likewise.
26417         * cse.c (equiv_constant, cse_insn): Likewise.
26418         * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
26419         (cselib_subst_to_values): Likewise.
26420         * dce.c (word_dce_process_block): Likewise.
26421         * df-problems.c (df_word_lr_mark_ref): Likewise.
26422         * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
26423         * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
26424         (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
26425         (rtl_for_decl_location): Likewise.
26426         * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
26427         * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
26428         * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
26429         (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
26430         (expand_expr_real_1): Likewise.
26431         * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
26432         (pad_below): Likewise.
26433         * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
26434         * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
26435         * ira.c (get_subreg_tracking_sizes): Likewise.
26436         * ira-build.c (ira_create_allocno_objects): Likewise.
26437         * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
26438         (ira_sort_regnos_for_alter_reg): Likewise.
26439         * ira-costs.c (record_operand_costs): Likewise.
26440         * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
26441         (resolve_simple_move): Likewise.
26442         * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
26443         (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
26444         (lra_constraints): Likewise.
26445         (CONST_POOL_OK_P): Reject variable-sized modes.
26446         * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
26447         (add_pseudo_to_slot, lra_spill): Likewise.
26448         * omp-low.c (omp_clause_aligned_alignment): Likewise.
26449         * optabs-query.c (get_best_extraction_insn): Likewise.
26450         * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
26451         * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
26452         (expand_mult_highpart, valid_multiword_target_p): Likewise.
26453         * recog.c (offsettable_address_addr_space_p): Likewise.
26454         * regcprop.c (maybe_mode_change): Likewise.
26455         * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
26456         * regrename.c (build_def_use): Likewise.
26457         * regstat.c (dump_reg_info): Likewise.
26458         * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
26459         (find_reloads, find_reloads_subreg_address): Likewise.
26460         * reload1.c (eliminate_regs_1): Likewise.
26461         * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
26462         * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
26463         (simplify_binary_operation_1, simplify_subreg): Likewise.
26464         * targhooks.c (default_function_arg_padding): Likewise.
26465         (default_hard_regno_nregs, default_class_max_nregs): Likewise.
26466         * tree-cfg.c (verify_gimple_assign_binary): Likewise.
26467         (verify_gimple_assign_ternary): Likewise.
26468         * tree-inline.c (estimate_move_cost): Likewise.
26469         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
26470         * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
26471         (get_address_cost_ainc): Likewise.
26472         * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
26473         (vect_supportable_dr_alignment): Likewise.
26474         * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
26475         (vectorizable_reduction): Likewise.
26476         * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
26477         (vectorizable_operation, vectorizable_load): Likewise.
26478         * tree.c (build_same_sized_truth_vector_type): Likewise.
26479         * valtrack.c (cleanup_auto_inc_dec): Likewise.
26480         * var-tracking.c (emit_note_insn_var_location): Likewise.
26481         * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
26482         (ADDR_VEC_ALIGN): Likewise.
26484 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
26485             Alan Hayward  <alan.hayward@arm.com>
26486             David Sherwood  <david.sherwood@arm.com>
26488         * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
26489         unsigned short.
26490         (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
26491         or if measurement_type is polynomial.
26492         * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
26493         * combine.c (make_extraction): Likewise.
26494         * dse.c (find_shift_sequence): Likewise.
26495         * dwarf2out.c (mem_loc_descriptor): Likewise.
26496         * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
26497         (extract_bit_field, extract_low_bits): Likewise.
26498         * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
26499         (optimize_bitfield_assignment_op, expand_assignment): Likewise.
26500         (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
26501         * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
26502         * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
26503         * reload.c (find_reloads): Likewise.
26504         * reload1.c (alter_reg): Likewise.
26505         * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
26506         * targhooks.c (default_secondary_memory_needed_mode): Likewise.
26507         * tree-if-conv.c (predicate_mem_writes): Likewise.
26508         * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
26509         * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
26510         * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
26511         * valtrack.c (dead_debug_insert_temp): Likewise.
26512         * varasm.c (mergeable_constant_section): Likewise.
26513         * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
26515 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
26516             Alan Hayward  <alan.hayward@arm.com>
26517             David Sherwood  <david.sherwood@arm.com>
26519         * expr.c (expand_assignment): Cope with polynomial mode sizes
26520         when assigning to a CONCAT.
26522 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
26523             Alan Hayward  <alan.hayward@arm.com>
26524             David Sherwood  <david.sherwood@arm.com>
26526         * machmode.h (mode_precision): Change from unsigned short to
26527         poly_uint16_pod.
26528         (mode_to_precision): Return a poly_uint16 rather than an unsigned
26529         short.
26530         (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
26531         or if measurement_type is not polynomial.
26532         (HWI_COMPUTABLE_MODE_P): Turn into a function.  Optimize the case
26533         in which the mode is already known to be a scalar_int_mode.
26534         * genmodes.c (emit_mode_precision): Change the type of mode_precision
26535         from unsigned short to poly_uint16_pod.  Use ZERO_COEFFS for the
26536         initializer.
26537         * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
26538         for GET_MODE_PRECISION.
26539         * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
26540         for GET_MODE_PRECISION.
26541         * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
26542         as polynomial.
26543         (try_combine, find_split_point, combine_simplify_rtx): Likewise.
26544         (expand_field_assignment, make_extraction): Likewise.
26545         (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
26546         (get_last_value): Likewise.
26547         * convert.c (convert_to_integer_1): Likewise.
26548         * cse.c (cse_insn): Likewise.
26549         * expr.c (expand_expr_real_1): Likewise.
26550         * lra-constraints.c (simplify_operand_subreg): Likewise.
26551         * optabs-query.c (can_atomic_load_p): Likewise.
26552         * optabs.c (expand_atomic_load): Likewise.
26553         (expand_atomic_store): Likewise.
26554         * ree.c (combine_reaching_defs): Likewise.
26555         * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
26556         * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
26557         * tree.h (type_has_mode_precision_p): Likewise.
26558         * ubsan.c (instrument_si_overflow): Likewise.
26560 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
26561             Alan Hayward  <alan.hayward@arm.com>
26562             David Sherwood  <david.sherwood@arm.com>
26564         * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
26565         polynomial numbers of units.
26566         (SET_TYPE_VECTOR_SUBPARTS): Likewise.
26567         (valid_vector_subparts_p): New function.
26568         (build_vector_type): Remove temporary shim and take the number
26569         of units as a poly_uint64 rather than an int.
26570         (build_opaque_vector_type): Take the number of units as a
26571         poly_uint64 rather than an int.
26572         * tree.c (build_vector_from_ctor): Handle polynomial
26573         TYPE_VECTOR_SUBPARTS.
26574         (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
26575         (uniform_vector_p, vector_type_mode, build_vector): Likewise.
26576         (build_vector_from_val): If the number of units is variable,
26577         use build_vec_duplicate_cst for constant operands and
26578         VEC_DUPLICATE_EXPR otherwise.
26579         (make_vector_type): Remove temporary is_constant ().
26580         (build_vector_type, build_opaque_vector_type): Take the number of
26581         units as a poly_uint64 rather than an int.
26582         (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
26583         VECTOR_CST_NELTS.
26584         * cfgexpand.c (expand_debug_expr): Likewise.
26585         * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
26586         (store_constructor, expand_expr_real_1): Likewise.
26587         (const_scalar_mask_from_tree): Likewise.
26588         * fold-const-call.c (fold_const_reduction): Likewise.
26589         * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
26590         (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
26591         (native_encode_vector, vec_cst_ctor_to_array): Likewise.
26592         (fold_relational_const): Likewise.
26593         (native_interpret_vector): Likewise.  Change the size from an
26594         int to an unsigned int.
26595         * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
26596         TYPE_VECTOR_SUBPARTS.
26597         (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
26598         (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
26599         duplicating a non-constant operand into a variable-length vector.
26600         * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
26601         TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
26602         * ipa-icf.c (sem_variable::equals): Likewise.
26603         * match.pd: Likewise.
26604         * omp-simd-clone.c (simd_clone_subparts): Likewise.
26605         * print-tree.c (print_node): Likewise.
26606         * stor-layout.c (layout_type): Likewise.
26607         * targhooks.c (default_builtin_vectorization_cost): Likewise.
26608         * tree-cfg.c (verify_gimple_comparison): Likewise.
26609         (verify_gimple_assign_binary): Likewise.
26610         (verify_gimple_assign_ternary): Likewise.
26611         (verify_gimple_assign_single): Likewise.
26612         * tree-pretty-print.c (dump_generic_node): Likewise.
26613         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
26614         (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
26615         * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
26616         (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
26617         (vect_shift_permute_load_chain): Likewise.
26618         * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
26619         (expand_vector_condition, optimize_vector_constructor): Likewise.
26620         (lower_vec_perm, get_compute_type): Likewise.
26621         * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
26622         (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
26623         * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
26624         (vect_recog_mask_conversion_pattern): Likewise.
26625         * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
26626         (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
26627         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
26628         (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
26629         (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
26630         (vectorizable_shift, vectorizable_operation, vectorizable_store)
26631         (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
26632         (supportable_widening_operation): Likewise.
26633         (supportable_narrowing_operation): Likewise.
26634         * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
26635         Likewise.
26636         * varasm.c (output_constant): Likewise.
26638 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
26639             Alan Hayward  <alan.hayward@arm.com>
26640             David Sherwood  <david.sherwood@arm.com>
26642         * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
26643         so that both the length == 3 and length != 3 cases set up their
26644         own permute vectors.  Add comments explaining why we know the
26645         number of elements is constant.
26646         (vect_permute_load_chain): Likewise.
26648 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
26649             Alan Hayward  <alan.hayward@arm.com>
26650             David Sherwood  <david.sherwood@arm.com>
26652         * machmode.h (mode_nunits): Change from unsigned char to
26653         poly_uint16_pod.
26654         (ONLY_FIXED_SIZE_MODES): New macro.
26655         (pod_mode::measurement_type, scalar_int_mode::measurement_type)
26656         (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
26657         (complex_mode::measurement_type, fixed_size_mode::measurement_type):
26658         New typedefs.
26659         (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
26660         (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
26661         or if measurement_type is not polynomial.
26662         * genmodes.c (ZERO_COEFFS): New macro.
26663         (emit_mode_nunits_inline): Make mode_nunits_inline return a
26664         poly_uint16.
26665         (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
26666         Use ZERO_COEFFS when emitting initializers.
26667         * data-streamer.h (bp_pack_poly_value): New function.
26668         (bp_unpack_poly_value): Likewise.
26669         * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
26670         for GET_MODE_NUNITS.
26671         * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
26672         for GET_MODE_NUNITS.
26673         * tree.c (make_vector_type): Remove temporary shim and make
26674         the real function take the number of units as a poly_uint64
26675         rather than an int.
26676         (build_vector_type_for_mode): Handle polynomial nunits.
26677         * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
26678         * emit-rtl.c (const_vec_series_p_1): Likewise.
26679         (gen_rtx_CONST_VECTOR): Likewise.
26680         * fold-const.c (test_vec_duplicate_folding): Likewise.
26681         * genrecog.c (validate_pattern): Likewise.
26682         * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
26683         * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
26684         * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
26685         (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
26686         (expand_vec_cond_expr, expand_mult_highpart): Likewise.
26687         * rtlanal.c (subreg_get_info): Likewise.
26688         * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
26689         (vect_grouped_load_supported): Likewise.
26690         * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
26691         * tree-vect-loop.c (have_whole_vector_shift): Likewise.
26692         * simplify-rtx.c (simplify_unary_operation_1): Likewise.
26693         (simplify_const_unary_operation, simplify_binary_operation_1)
26694         (simplify_const_binary_operation, simplify_ternary_operation)
26695         (test_vector_ops_duplicate, test_vector_ops): Likewise.
26696         (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
26697         instead of CONST_VECTOR_NUNITS.
26698         * varasm.c (output_constant_pool_2): Likewise.
26699         * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
26700         explicit-encoded elements in the XVEC for variable-length vectors.
26702 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
26704         * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
26706 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
26707             Alan Hayward  <alan.hayward@arm.com>
26708             David Sherwood  <david.sherwood@arm.com>
26710         * coretypes.h (fixed_size_mode): Declare.
26711         (fixed_size_mode_pod): New typedef.
26712         * builtins.h (target_builtins::x_apply_args_mode)
26713         (target_builtins::x_apply_result_mode): Change type to
26714         fixed_size_mode_pod.
26715         * builtins.c (apply_args_size, apply_result_size, result_vector)
26716         (expand_builtin_apply_args_1, expand_builtin_apply)
26717         (expand_builtin_return): Update accordingly.
26719 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
26721         * cse.c (hash_rtx_cb): Hash only the encoded elements.
26722         * cselib.c (cselib_hash_rtx): Likewise.
26723         * expmed.c (make_tree): Build VECTOR_CSTs directly from the
26724         CONST_VECTOR encoding.
26726 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
26727             Jeff Law  <law@redhat.com>
26729         PR target/83641
26730         * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
26731         noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
26732         only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
26733         and add REG_CFA_ADJUST_CFA notes in that case to both insns.
26735         PR target/83641
26736         * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
26737         explicitly probe *sp in a noreturn function if there were any callee
26738         register saves or frame pointer is needed.
26740 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
26742         PR debug/83621
26743         * cfgexpand.c (expand_debug_expr): Return NULL if mode is
26744         BLKmode for ternary, binary or unary expressions.
26746         PR debug/83645
26747         * var-tracking.c (delete_vta_debug_insn): New inline function.
26748         (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
26749         insns from get_insns () to NULL instead of each bb separately.
26750         Use delete_vta_debug_insn.  No longer static.
26751         (vt_debug_insns_local, variable_tracking_main_1): Adjust
26752         delete_vta_debug_insns callers.
26753         * rtl.h (delete_vta_debug_insns): Declare.
26754         * final.c (rest_of_handle_final): Call delete_vta_debug_insns
26755         instead of variable_tracking_main.
26757 2018-01-03  Martin Sebor  <msebor@redhat.com>
26759         PR tree-optimization/83603
26760         * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
26761         arguments past the endof the argument list in functions declared
26762         without a prototype.
26763         * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
26764         Avoid checking when arguments are null.
26766 2018-01-03  Martin Sebor  <msebor@redhat.com>
26768         PR c/83559
26769         * doc/extend.texi (attribute const): Fix a typo.
26770         * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
26771         issuing -Wsuggest-attribute for void functions.
26773 2018-01-03  Martin Sebor  <msebor@redhat.com>
26775         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
26776         offset_int::from instead of wide_int::to_shwi.
26777         (maybe_diag_overlap): Remove assertion.
26778         Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
26779         * gimple-ssa-sprintf.c (format_directive): Same.
26780         (parse_directive): Same.
26781         (sprintf_dom_walker::compute_format_length): Same.
26782         (try_substitute_return_value): Same.
26784 2018-01-03  Jeff Law  <law@redhat.com>
26786         PR middle-end/83654
26787         * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
26788         non-constant residual for zero at runtime and avoid probing in
26789         that case.  Reorganize code for trailing problem to mirror handling
26790         of the residual.
26792 2018-01-03  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
26794         PR tree-optimization/83501
26795         * tree-ssa-strlen.c (get_string_cst): New.
26796         (handle_char_store): Call get_string_cst.
26798 2018-01-03  Martin Liska  <mliska@suse.cz>
26800         PR tree-optimization/83593
26801         * tree-ssa-strlen.c: Include tree-cfg.h.
26802         (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
26803         (strlen_dom_walker): Add new member variable m_cleanup_cfg.
26804         (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
26805         to false.
26806         (strlen_dom_walker::before_dom_children): Call
26807         gimple_purge_dead_eh_edges. Dump tranformation with details
26808         dump flags.
26809         (strlen_dom_walker::before_dom_children): Update call by adding
26810         new argument cleanup_eh.
26811         (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
26813 2018-01-03  Martin Liska  <mliska@suse.cz>
26815         PR ipa/83549
26816         * cif-code.def (VARIADIC_THUNK): New enum value.
26817         * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
26818         thunks.
26820 2018-01-03  Jan Beulich  <jbeulich@suse.com>
26822         * sse.md (mov<mode>_internal): Tighten condition for when to use
26823         vmovdqu<ssescalarsize> for TI and OI modes.
26825 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
26827         Update copyright years.
26829 2018-01-03  Martin Liska  <mliska@suse.cz>
26831         PR ipa/83594
26832         * ipa-visibility.c (function_and_variable_visibility): Skip
26833         functions with noipa attribure.
26835 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
26837         * gcc.c (process_command): Update copyright notice dates.
26838         * gcov-dump.c (print_version): Ditto.
26839         * gcov.c (print_version): Ditto.
26840         * gcov-tool.c (print_version): Ditto.
26841         * gengtype.c (create_file): Ditto.
26842         * doc/cpp.texi: Bump @copying's copyright year.
26843         * doc/cppinternals.texi: Ditto.
26844         * doc/gcc.texi: Ditto.
26845         * doc/gccint.texi: Ditto.
26846         * doc/gcov.texi: Ditto.
26847         * doc/install.texi: Ditto.
26848         * doc/invoke.texi: Ditto.
26850 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
26852         * vector-builder.h (vector_builder::m_full_nelts): Change from
26853         unsigned int to poly_uint64.
26854         (vector_builder::full_nelts): Update prototype accordingly.
26855         (vector_builder::new_vector): Likewise.
26856         (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
26857         (vector_builder::operator ==): Likewise.
26858         (vector_builder::finalize): Likewise.
26859         * int-vector-builder.h (int_vector_builder::int_vector_builder):
26860         Take the number of elements as a poly_uint64 rather than an
26861         unsigned int.
26862         * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
26863         from unsigned int to poly_uint64.
26864         (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
26865         (vec_perm_indices::new_vector): Likewise.
26866         (vec_perm_indices::length): Likewise.
26867         (vec_perm_indices::nelts_per_input): Likewise.
26868         (vec_perm_indices::input_nelts): Likewise.
26869         * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
26870         number of elements per input as a poly_uint64 rather than an
26871         unsigned int.  Use the original encoding for variable-length
26872         vectors, rather than clamping each individual element.
26873         For the second and subsequent elements in each pattern,
26874         clamp the step and base before clamping their sum.
26875         (vec_perm_indices::series_p): Handle polynomial element counts.
26876         (vec_perm_indices::all_in_range_p): Likewise.
26877         (vec_perm_indices_to_tree): Likewise.
26878         (vec_perm_indices_to_rtx): Likewise.
26879         * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
26880         * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
26881         (tree_vector_builder::new_binary_operation): Handle polynomial
26882         element counts.  Return false if we need to know the number
26883         of elements at compile time.
26884         * fold-const.c (fold_vec_perm): Punt if the number of elements
26885         isn't known at compile time.
26887 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
26889         * vec-perm-indices.h (vec_perm_builder): Change element type
26890         from HOST_WIDE_INT to poly_int64.
26891         (vec_perm_indices::element_type): Update accordingly.
26892         (vec_perm_indices::clamp): Handle polynomial element_types.
26893         * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
26894         (vec_perm_indices::all_in_range_p): Likewise.
26895         (tree_to_vec_perm_builder): Check for poly_int64 trees rather
26896         than shwi trees.
26897         * vector-builder.h (vector_builder::stepped_sequence_p): Handle
26898         polynomial vec_perm_indices element types.
26899         * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
26900         * fold-const.c (fold_vec_perm): Likewise.
26901         * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
26902         * tree-vect-generic.c (lower_vec_perm): Likewise.
26903         * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
26904         * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
26905         element type to HOST_WIDE_INT.
26907 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
26908             Alan Hayward  <alan.hayward@arm.com>
26909             David Sherwood  <david.sherwood@arm.com>
26911         * alias.c (addr_side_effect_eval): Take the size as a poly_int64
26912         rather than an int.  Use plus_constant.
26913         (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
26914         Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
26916 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
26917             Alan Hayward  <alan.hayward@arm.com>
26918             David Sherwood  <david.sherwood@arm.com>
26920         * calls.c (emit_call_1, expand_call): Change struct_value_size from
26921         a HOST_WIDE_INT to a poly_int64.
26923 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
26924             Alan Hayward  <alan.hayward@arm.com>
26925             David Sherwood  <david.sherwood@arm.com>
26927         * calls.c (load_register_parameters): Cope with polynomial
26928         mode sizes.  Require a constant size for BLKmode parameters
26929         that aren't described by a PARALLEL.  If BLOCK_REG_PADDING
26930         forces a parameter to be padded at the lsb end in order to
26931         fill a complete number of words, require the parameter size
26932         to be ordered wrt UNITS_PER_WORD.
26934 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
26935             Alan Hayward  <alan.hayward@arm.com>
26936             David Sherwood  <david.sherwood@arm.com>
26938         * reload1.c (spill_stack_slot_width): Change element type
26939         from unsigned int to poly_uint64_pod.
26940         (alter_reg): Treat mode sizes as polynomial.
26942 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
26943             Alan Hayward  <alan.hayward@arm.com>
26944             David Sherwood  <david.sherwood@arm.com>
26946         * reload.c (complex_word_subreg_p): New function.
26947         (reload_inner_reg_of_subreg, push_reload): Use it.
26949 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
26950             Alan Hayward  <alan.hayward@arm.com>
26951             David Sherwood  <david.sherwood@arm.com>
26953         * lra-constraints.c (process_alt_operands): Reject matched
26954         operands whose sizes aren't ordered.
26955         (match_reload): Refer to this check here.
26957 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
26958             Alan Hayward  <alan.hayward@arm.com>
26959             David Sherwood  <david.sherwood@arm.com>
26961         * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
26962         that the mode size is in the set {1, 2, 4, 8, 16}.
26964 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
26965             Alan Hayward  <alan.hayward@arm.com>
26966             David Sherwood  <david.sherwood@arm.com>
26968         * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
26969         Use plus_constant instead of gen_rtx_PLUS.
26971 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
26972             Alan Hayward  <alan.hayward@arm.com>
26973             David Sherwood  <david.sherwood@arm.com>
26975         * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
26976         * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
26977         * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
26978         * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
26979         * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
26980         * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
26981         * config/i386/i386-protos.h (ix86_push_rounding): Declare.
26982         * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
26983         * config/i386/i386.c (ix86_push_rounding): ...this new function.
26984         * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
26985         a poly_int64.
26986         * config/m32c/m32c.c (m32c_push_rounding): Likewise.
26987         * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
26988         * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
26989         * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
26990         * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
26991         * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
26992         * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
26993         * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
26994         * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
26995         * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
26996         function.
26997         * expr.c (emit_move_resolve_push): Treat the input and result
26998         of PUSH_ROUNDING as a poly_int64.
26999         (emit_move_complex_push, emit_single_push_insn_1): Likewise.
27000         (emit_push_insn): Likewise.
27001         * lra-eliminations.c (mark_not_eliminable): Likewise.
27002         * recog.c (push_operand): Likewise.
27003         * reload1.c (elimination_effects): Likewise.
27004         * rtlanal.c (nonzero_bits1): Likewise.
27005         * calls.c (store_one_arg): Likewise.  Require the padding to be
27006         known at compile time.
27008 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27009             Alan Hayward  <alan.hayward@arm.com>
27010             David Sherwood  <david.sherwood@arm.com>
27012         * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
27013         Use plus_constant instead of gen_rtx_PLUS.
27015 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27016             Alan Hayward  <alan.hayward@arm.com>
27017             David Sherwood  <david.sherwood@arm.com>
27019         * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
27020         rather than an int.
27022 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27023             Alan Hayward  <alan.hayward@arm.com>
27024             David Sherwood  <david.sherwood@arm.com>
27026         * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
27027         instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
27028         via stack temporaries.  Treat the mode size as polynomial too.
27030 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27031             Alan Hayward  <alan.hayward@arm.com>
27032             David Sherwood  <david.sherwood@arm.com>
27034         * expr.c (expand_expr_real_2): When handling conversions involving
27035         unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
27036         multiplying int_size_in_bytes by BITS_PER_UNIT.  Treat GET_MODE_BISIZE
27037         as a poly_uint64 too.
27039 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27040             Alan Hayward  <alan.hayward@arm.com>
27041             David Sherwood  <david.sherwood@arm.com>
27043         * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
27045 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27046             Alan Hayward  <alan.hayward@arm.com>
27047             David Sherwood  <david.sherwood@arm.com>
27049         * combine.c (can_change_dest_mode): Handle polynomial
27050         REGMODE_NATURAL_SIZE.
27051         * expmed.c (store_bit_field_1): Likewise.
27052         * expr.c (store_constructor): Likewise.
27053         * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
27054         and polynomial REGMODE_NATURAL_SIZE.
27055         (gen_lowpart_common): Likewise.
27056         * reginfo.c (record_subregs_of_mode): Likewise.
27057         * rtlanal.c (read_modify_subreg_p): Likewise.
27059 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27060             Alan Hayward  <alan.hayward@arm.com>
27061             David Sherwood  <david.sherwood@arm.com>
27063         * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
27064         numbers of elements.
27066 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27067             Alan Hayward  <alan.hayward@arm.com>
27068             David Sherwood  <david.sherwood@arm.com>
27070         * match.pd: Cope with polynomial numbers of vector elements.
27072 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27073             Alan Hayward  <alan.hayward@arm.com>
27074             David Sherwood  <david.sherwood@arm.com>
27076         * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
27077         in a POINTER_PLUS_EXPR.
27079 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27080             Alan Hayward  <alan.hayward@arm.com>
27081             David Sherwood  <david.sherwood@arm.com>
27083         * omp-simd-clone.c (simd_clone_subparts): New function.
27084         (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
27085         (ipa_simd_modify_function_body): Likewise.
27087 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27088             Alan Hayward  <alan.hayward@arm.com>
27089             David Sherwood  <david.sherwood@arm.com>
27091         * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
27092         (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
27093         (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
27094         (expand_vector_condition, vector_element): Likewise.
27095         (subparts_gt): New function.
27096         (get_compute_type): Use subparts_gt.
27097         (count_type_subparts): Delete.
27098         (expand_vector_operations_1): Use subparts_gt instead of
27099         count_type_subparts.
27101 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27102             Alan Hayward  <alan.hayward@arm.com>
27103             David Sherwood  <david.sherwood@arm.com>
27105         * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
27106         (vect_compile_time_alias): ...this new function.  Do the calculation
27107         on poly_ints rather than trees.
27108         (vect_prune_runtime_alias_test_list): Update call accordingly.
27110 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27111             Alan Hayward  <alan.hayward@arm.com>
27112             David Sherwood  <david.sherwood@arm.com>
27114         * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
27115         numbers of units.
27116         (vect_schedule_slp_instance): Likewise.
27118 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27119             Alan Hayward  <alan.hayward@arm.com>
27120             David Sherwood  <david.sherwood@arm.com>
27122         * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
27123         constant and extern definitions for variable-length vectors.
27124         (vect_get_constant_vectors): Note that the number of units
27125         is known to be constant.
27127 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27128             Alan Hayward  <alan.hayward@arm.com>
27129             David Sherwood  <david.sherwood@arm.com>
27131         * tree-vect-stmts.c (vectorizable_conversion): Treat the number
27132         of units as polynomial.  Choose between WIDE and NARROW based
27133         on multiple_p.
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         * tree-vect-stmts.c (simd_clone_subparts): New function.
27140         (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
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         * tree-vect-stmts.c (vectorizable_call): Treat the number of
27147         vectors as polynomial.  Use build_index_vector for
27148         IFN_GOMP_SIMD_LANE.
27150 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27151             Alan Hayward  <alan.hayward@arm.com>
27152             David Sherwood  <david.sherwood@arm.com>
27154         * tree-vect-stmts.c (get_load_store_type): Treat the number of
27155         units as polynomial.  Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
27156         for variable-length vectors.
27157         (vectorizable_mask_load_store): Treat the number of units as
27158         polynomial, asserting that it is constant if the condition has
27159         already been enforced.
27160         (vectorizable_store, vectorizable_load): Likewise.
27162 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27163             Alan Hayward  <alan.hayward@arm.com>
27164             David Sherwood  <david.sherwood@arm.com>
27166         * tree-vect-loop.c (vectorizable_live_operation): Treat the number
27167         of units as polynomial.  Punt if we can't tell at compile time
27168         which vector contains the final result.
27170 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27171             Alan Hayward  <alan.hayward@arm.com>
27172             David Sherwood  <david.sherwood@arm.com>
27174         * tree-vect-loop.c (vectorizable_induction): Treat the number
27175         of units as polynomial.  Punt on SLP inductions.  Use an integer
27176         VEC_SERIES_EXPR for variable-length integer reductions.  Use a
27177         cast of such a series for variable-length floating-point
27178         reductions.
27180 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27181             Alan Hayward  <alan.hayward@arm.com>
27182             David Sherwood  <david.sherwood@arm.com>
27184         * tree.h (build_index_vector): Declare.
27185         * tree.c (build_index_vector): New function.
27186         * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
27187         of units as polynomial, forcibly converting it to a constant if
27188         vectorizable_reduction has already enforced the condition.
27189         (vect_create_epilog_for_reduction): Likewise.  Use build_index_vector
27190         to create a {1,2,3,...} vector.
27191         (vectorizable_reduction): Treat the number of units as polynomial.
27192         Choose vectype_in based on the largest scalar element size rather
27193         than the smallest number of units.  Enforce the restrictions
27194         relied on above.
27196 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27197             Alan Hayward  <alan.hayward@arm.com>
27198             David Sherwood  <david.sherwood@arm.com>
27200         * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
27201         number of units as polynomial.
27203 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27204             Alan Hayward  <alan.hayward@arm.com>
27205             David Sherwood  <david.sherwood@arm.com>
27207         * target.h (vector_sizes, auto_vector_sizes): New typedefs.
27208         * target.def (autovectorize_vector_sizes): Return the vector sizes
27209         by pointer, using vector_sizes rather than a bitmask.
27210         * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
27211         * targhooks.c (default_autovectorize_vector_sizes): Likewise.
27212         * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
27213         Likewise.
27214         * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
27215         * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
27216         * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
27217         * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
27218         * omp-general.c (omp_max_vf): Likewise.
27219         * omp-low.c (omp_clause_aligned_alignment): Likewise.
27220         * optabs-query.c (can_vec_mask_load_store_p): Likewise.
27221         * tree-vect-loop.c (vect_analyze_loop): Likewise.
27222         * tree-vect-slp.c (vect_slp_bb): Likewise.
27223         * doc/tm.texi: Regenerate.
27224         * tree-vectorizer.h (current_vector_size): Change from an unsigned int
27225         to a poly_uint64.
27226         * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
27227         the vector size as a poly_uint64 rather than an unsigned int.
27228         (current_vector_size): Change from an unsigned int to a poly_uint64.
27229         (get_vectype_for_scalar_type): Update accordingly.
27230         * tree.h (build_truth_vector_type): Take the size and number of
27231         units as a poly_uint64 rather than an unsigned int.
27232         (build_vector_type): Add a temporary overload that takes
27233         the number of units as a poly_uint64 rather than an unsigned int.
27234         * tree.c (make_vector_type): Likewise.
27235         (build_truth_vector_type): Take the number of units as a poly_uint64
27236         rather than an unsigned int.
27238 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27239             Alan Hayward  <alan.hayward@arm.com>
27240             David Sherwood  <david.sherwood@arm.com>
27242         * target.def (get_mask_mode): Take the number of units and length
27243         as poly_uint64s rather than unsigned ints.
27244         * targhooks.h (default_get_mask_mode): Update accordingly.
27245         * targhooks.c (default_get_mask_mode): Likewise.
27246         * config/i386/i386.c (ix86_get_mask_mode): Likewise.
27247         * doc/tm.texi: Regenerate.
27249 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27250             Alan Hayward  <alan.hayward@arm.com>
27251             David Sherwood  <david.sherwood@arm.com>
27253         * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
27254         * omp-general.c (omp_max_vf): Likewise.
27255         * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
27256         (expand_omp_simd): Handle polynomial safelen.
27257         * omp-low.c (omplow_simd_context): Add a default constructor.
27258         (omplow_simd_context::max_vf): Change from int to poly_uint64.
27259         (lower_rec_simd_input_clauses): Update accordingly.
27260         (lower_rec_input_clauses): Likewise.
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-vectorizer.h (vect_nunits_for_cost): New function.
27267         * tree-vect-loop.c (vect_model_reduction_cost): Use it.
27268         * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
27269         (vect_analyze_slp_cost): Likewise.
27270         * tree-vect-stmts.c (vect_model_store_cost): Likewise.
27271         (vect_model_load_cost): Likewise.
27273 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27274             Alan Hayward  <alan.hayward@arm.com>
27275             David Sherwood  <david.sherwood@arm.com>
27277         * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
27278         (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
27279         from an unsigned int * to a poly_uint64_pod *.
27280         (calculate_unrolling_factor): New function.
27281         (vect_analyze_slp_instance): Use it.  Track polynomial max_nunits.
27283 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27284             Alan Hayward  <alan.hayward@arm.com>
27285             David Sherwood  <david.sherwood@arm.com>
27287         * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
27288         from an unsigned int to a poly_uint64.
27289         (_loop_vec_info::slp_unrolling_factor): Likewise.
27290         (_loop_vec_info::vectorization_factor): Change from an int
27291         to a poly_uint64.
27292         (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
27293         (vect_get_num_vectors): New function.
27294         (vect_update_max_nunits, vect_vf_for_cost): Likewise.
27295         (vect_get_num_copies): Use vect_get_num_vectors.
27296         (vect_analyze_data_ref_dependences): Change max_vf from an int *
27297         to an unsigned int *.
27298         (vect_analyze_data_refs): Change min_vf from an int * to a
27299         poly_uint64 *.
27300         (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
27301         than an unsigned HOST_WIDE_INT.
27302         * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
27303         (vect_analyze_data_ref_dependence): Change max_vf from an int *
27304         to an unsigned int *.
27305         (vect_analyze_data_ref_dependences): Likewise.
27306         (vect_compute_data_ref_alignment): Handle polynomial vf.
27307         (vect_enhance_data_refs_alignment): Likewise.
27308         (vect_prune_runtime_alias_test_list): Likewise.
27309         (vect_shift_permute_load_chain): Likewise.
27310         (vect_supportable_dr_alignment): Likewise.
27311         (dependence_distance_ge_vf): Take the vectorization factor as a
27312         poly_uint64 rather than an unsigned HOST_WIDE_INT.
27313         (vect_analyze_data_refs): Change min_vf from an int * to a
27314         poly_uint64 *.
27315         * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
27316         vfm1 as a poly_uint64 rather than an int.  Make the same change
27317         for the returned bound_scalar.
27318         (vect_gen_vector_loop_niters): Handle polynomial vf.
27319         (vect_do_peeling): Likewise.  Update call to
27320         vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
27321         (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
27322         be constant.
27323         * tree-vect-loop.c (vect_determine_vectorization_factor)
27324         (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
27325         (vect_get_known_peeling_cost): Likewise.
27326         (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
27327         (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
27328         (vect_transform_loop): Likewise.  Use the lowest possible VF when
27329         updating the upper bounds of the loop.
27330         (vect_min_worthwhile_factor): Make static.  Return an unsigned int
27331         rather than an int.
27332         * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
27333         polynomial unroll factors.
27334         (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
27335         (vect_make_slp_decision): Likewise.
27336         (vect_supported_load_permutation_p): Likewise, and polynomial
27337         vf too.
27338         (vect_analyze_slp_cost): Handle polynomial vf.
27339         (vect_slp_analyze_node_operations): Likewise.
27340         (vect_slp_analyze_bb_1): Likewise.
27341         (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
27342         than an unsigned HOST_WIDE_INT.
27343         * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
27344         (vectorizable_load): Handle polynomial vf.
27345         * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
27346         a poly_uint64.
27347         (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
27349 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27350             Alan Hayward  <alan.hayward@arm.com>
27351             David Sherwood  <david.sherwood@arm.com>
27353         * match.pd: Handle bit operations involving three constants
27354         and try to fold one pair.
27356 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27358         * tree-vect-loop-manip.c: Include gimple-fold.h.
27359         (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
27360         niters_maybe_zero parameters.  Handle other cases besides a step of 1.
27361         (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
27362         Add a path that uses a step of VF instead of 1, but disable it
27363         for now.
27364         (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
27365         and niters_no_overflow parameters.  Update calls to
27366         slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
27367         Create a new SSA name if the latter choses to use a ste other
27368         than zero, and return it via niters_vector_mult_vf_var.
27369         * tree-vect-loop.c (vect_transform_loop): Update calls to
27370         vect_do_peeling, vect_gen_vector_loop_niters and
27371         slpeel_make_loop_iterate_ntimes.
27372         * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
27373         (vect_gen_vector_loop_niters): Update declarations after above changes.
27375 2018-01-02  Michael Meissner  <meissner@linux.vnet.ibm.com>
27377         * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
27378         128-bit round to integer instructions.
27379         (ceil<mode>2): Likewise.
27380         (btrunc<mode>2): Likewise.
27381         (round<mode>2): Likewise.
27383 2018-01-02  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
27385         * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
27386         unaligned VSX load/store on P8/P9.
27387         (expand_block_clear): Allow the use of unaligned VSX
27388         load/store on P8/P9.
27390 2018-01-02  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
27392         * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
27393         New function.
27394         (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
27395         swap associated with both a load and a store.
27397 2018-01-02  Andrew Waterman  <andrew@sifive.com>
27399         * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
27400         * config/riscv/riscv.md (clear_cache): Use it.
27402 2018-01-02  Artyom Skrobov  <tyomitch@gmail.com>
27404         * web.c: Remove out-of-date comment.
27406 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
27408         * expr.c (fixup_args_size_notes): Check that any existing
27409         REG_ARGS_SIZE notes are correct, and don't try to re-add them.
27410         (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
27411         (emit_single_push_insn): ...here.
27413 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
27415         * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
27416         (const_vector_encoded_nelts): New function.
27417         (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
27418         (const_vector_int_elt, const_vector_elt): Declare.
27419         * emit-rtl.c (const_vector_int_elt_1): New function.
27420         (const_vector_elt): Likewise.
27421         * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
27422         of CONST_VECTOR_ELT.
27424 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
27426         * expr.c: Include rtx-vector-builder.h.
27427         (const_vector_mask_from_tree): Use rtx_vector_builder and operate
27428         directly on the tree encoding.
27429         (const_vector_from_tree): Likewise.
27430         * optabs.c: Include rtx-vector-builder.h.
27431         (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
27432         sequence of "u" values.
27433         * vec-perm-indices.c: Include rtx-vector-builder.h.
27434         (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
27435         directly on the vec_perm_indices encoding.
27437 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
27439         * doc/rtl.texi (const_vector): Describe new encoding scheme.
27440         * Makefile.in (OBJS): Add rtx-vector-builder.o.
27441         * rtx-vector-builder.h: New file.
27442         * rtx-vector-builder.c: Likewise.
27443         * rtl.h (rtx_def::u2): Add a const_vector field.
27444         (CONST_VECTOR_NPATTERNS): New macro.
27445         (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
27446         (CONST_VECTOR_DUPLICATE_P): Likewise.
27447         (CONST_VECTOR_STEPPED_P): Likewise.
27448         (CONST_VECTOR_ENCODED_ELT): Likewise.
27449         (const_vec_duplicate_p): Check for a duplicated vector encoding.
27450         (unwrap_const_vec_duplicate): Likewise.
27451         (const_vec_series_p): Check for a non-duplicated vector encoding.
27452         Say that the function only returns true for integer vectors.
27453         * emit-rtl.c: Include rtx-vector-builder.h.
27454         (gen_const_vec_duplicate_1): Delete.
27455         (gen_const_vector): Call gen_const_vec_duplicate instead of
27456         gen_const_vec_duplicate_1.
27457         (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
27458         (gen_const_vec_duplicate): Use rtx_vector_builder.
27459         (gen_const_vec_series): Likewise.
27460         (gen_rtx_CONST_VECTOR): Likewise.
27461         * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
27462         (swap_const_vector_halves): Take an rtx pointer rather than rtx.
27463         Build a new vector rather than modifying a CONST_VECTOR in-place.
27464         (handle_special_swappables): Update call accordingly.
27465         * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
27466         (swap_const_vector_halves): Take an rtx pointer rather than rtx.
27467         Build a new vector rather than modifying a CONST_VECTOR in-place.
27468         (handle_special_swappables): Update call accordingly.
27470 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
27472         * simplify-rtx.c (simplify_const_binary_operation): Use
27473         CONST_VECTOR_ELT instead of XVECEXP.
27475 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
27477         * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
27478         the selector elements to be different from the data elements
27479         if the selector is a VECTOR_CST.
27480         * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
27481         ssizetype for the selector.
27483 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
27485         * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
27486         before testing each element individually.
27487         * tree-vect-generic.c (lower_vec_perm): Likewise.
27489 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
27491         * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
27492         * selftest-run-tests.c (selftest::run_tests): Call it.
27493         * vector-builder.h (vector_builder::operator ==): New function.
27494         (vector_builder::operator !=): Likewise.
27495         * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
27496         (vec_perm_indices::all_from_input_p): New function.
27497         * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
27498         (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
27499         * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
27500         instead of reading the VECTOR_CST directly.  Detect whether both
27501         vector inputs are the same before constructing the vec_perm_indices,
27502         and update the number of inputs argument accordingly.  Use the
27503         utility functions added above.  Only construct sel2 if we need to.
27505 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
27507         * optabs.c (expand_vec_perm_var): Use an explicit encoding for
27508         the broadcast of the low byte.
27509         (expand_mult_highpart): Use an explicit encoding for the permutes.
27510         * optabs-query.c (can_mult_highpart_p): Likewise.
27511         * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
27512         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
27513         (vectorizable_bswap): Likewise.
27514         * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
27515         explicit encoding for the power-of-2 permutes.
27516         (vect_permute_store_chain): Likewise.
27517         (vect_grouped_load_supported): Likewise.
27518         (vect_permute_load_chain): Likewise.
27520 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
27522         * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
27523         * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
27524         * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
27525         * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
27526         * tree-vect-stmts.c (vectorizable_bswap): Likewise.
27527         (vect_gen_perm_mask_any): Likewise.
27529 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
27531         * int-vector-builder.h: New file.
27532         * vec-perm-indices.h: Include int-vector-builder.h.
27533         (vec_perm_indices): Redefine as an int_vector_builder.
27534         (auto_vec_perm_indices): Delete.
27535         (vec_perm_builder): Redefine as a stand-alone class.
27536         (vec_perm_indices::vec_perm_indices): New function.
27537         (vec_perm_indices::clamp): Likewise.
27538         * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
27539         (vec_perm_indices::new_vector): New function.
27540         (vec_perm_indices::new_expanded_vector): Update for new
27541         vec_perm_indices class.
27542         (vec_perm_indices::rotate_inputs): New function.
27543         (vec_perm_indices::all_in_range_p): Operate directly on the
27544         encoded form, without computing elided elements.
27545         (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
27546         encoding.  Update for new vec_perm_indices class.
27547         * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
27548         the given vec_perm_builder.
27549         (expand_vec_perm_var): Update vec_perm_builder constructor.
27550         (expand_mult_highpart): Use vec_perm_builder instead of
27551         auto_vec_perm_indices.
27552         * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
27553         vec_perm_indices instead of auto_vec_perm_indices.  Use a single
27554         or double series encoding as appropriate.
27555         * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
27556         vec_perm_indices instead of auto_vec_perm_indices.
27557         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
27558         * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
27559         (vect_permute_store_chain): Likewise.
27560         (vect_grouped_load_supported): Likewise.
27561         (vect_permute_load_chain): Likewise.
27562         (vect_shift_permute_load_chain): Likewise.
27563         * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
27564         (vect_transform_slp_perm_load): Likewise.
27565         (vect_schedule_slp_instance): Likewise.
27566         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
27567         (vectorizable_mask_load_store): Likewise.
27568         (vectorizable_bswap): Likewise.
27569         (vectorizable_store): Likewise.
27570         (vectorizable_load): Likewise.
27571         * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
27572         vec_perm_indices instead of auto_vec_perm_indices.  Use
27573         tree_to_vec_perm_builder to read the vector from a tree.
27574         * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
27575         vec_perm_builder instead of a vec_perm_indices.
27576         (have_whole_vector_shift): Use vec_perm_builder and
27577         vec_perm_indices instead of auto_vec_perm_indices.  Leave the
27578         truncation to calc_vec_perm_mask_for_shift.
27579         (vect_create_epilog_for_reduction): Likewise.
27580         * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
27581         from auto_vec_perm_indices to vec_perm_indices.
27582         (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
27583         instead of changing individual elements.
27584         (aarch64_vectorize_vec_perm_const): Use new_vector to install
27585         the vector in d.perm.
27586         * config/arm/arm.c (expand_vec_perm_d::perm): Change
27587         from auto_vec_perm_indices to vec_perm_indices.
27588         (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
27589         instead of changing individual elements.
27590         (arm_vectorize_vec_perm_const): Use new_vector to install
27591         the vector in d.perm.
27592         * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
27593         Update vec_perm_builder constructor.
27594         (rs6000_expand_interleave): Likewise.
27595         * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
27596         (rs6000_expand_interleave): Likewise.
27598 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
27600         * optabs-query.c (can_vec_perm_var_p): Check whether lowering
27601         to qimode could truncate the indices.
27602         * optabs.c (expand_vec_perm_var): Likewise.
27604 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
27606         * Makefile.in (OBJS): Add vec-perm-indices.o.
27607         * vec-perm-indices.h: New file.
27608         * vec-perm-indices.c: Likewise.
27609         * target.h (vec_perm_indices): Replace with a forward class
27610         declaration.
27611         (auto_vec_perm_indices): Move to vec-perm-indices.h.
27612         * optabs.h: Include vec-perm-indices.h.
27613         (expand_vec_perm): Delete.
27614         (selector_fits_mode_p, expand_vec_perm_var): Declare.
27615         (expand_vec_perm_const): Declare.
27616         * target.def (vec_perm_const_ok): Replace with...
27617         (vec_perm_const): ...this new hook.
27618         * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
27619         (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
27620         * doc/tm.texi: Regenerate.
27621         * optabs.def (vec_perm_const): Delete.
27622         * doc/md.texi (vec_perm_const): Likewise.
27623         (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
27624         * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
27625         expand_vec_perm for constant permutation vectors.  Assert that
27626         the mode of variable permutation vectors is the integer equivalent
27627         of the mode that is being permuted.
27628         * optabs-query.h (selector_fits_mode_p): Declare.
27629         * optabs-query.c: Include vec-perm-indices.h.
27630         (selector_fits_mode_p): New function.
27631         (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
27632         is defined, instead of checking whether the vec_perm_const_optab
27633         exists.  Use targetm.vectorize.vec_perm_const instead of
27634         targetm.vectorize.vec_perm_const_ok.  Check whether the indices
27635         fit in the vector mode before using a variable permute.
27636         * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
27637         vec_perm_indices instead of an rtx.
27638         (expand_vec_perm): Replace with...
27639         (expand_vec_perm_const): ...this new function.  Take the selector
27640         as a vec_perm_indices rather than an rtx.  Also take the mode of
27641         the selector.  Update call to shift_amt_for_vec_perm_mask.
27642         Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
27643         Use vec_perm_indices::new_expanded_vector to expand the original
27644         selector into bytes.  Check whether the indices fit in the vector
27645         mode before using a variable permute.
27646         (expand_vec_perm_var): Make global.
27647         (expand_mult_highpart): Use expand_vec_perm_const.
27648         * fold-const.c: Includes vec-perm-indices.h.
27649         * tree-ssa-forwprop.c: Likewise.
27650         * tree-vect-data-refs.c: Likewise.
27651         * tree-vect-generic.c: Likewise.
27652         * tree-vect-loop.c: Likewise.
27653         * tree-vect-slp.c: Likewise.
27654         * tree-vect-stmts.c: Likewise.
27655         * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
27656         Delete.
27657         * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
27658         * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
27659         (aarch64_vectorize_vec_perm_const_ok): Fuse into...
27660         (aarch64_vectorize_vec_perm_const): ...this new function.
27661         (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
27662         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
27663         * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
27664         * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
27665         * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
27666         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
27667         (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
27668         into...
27669         (arm_vectorize_vec_perm_const): ...this new function.  Explicitly
27670         check for NEON modes.
27671         * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
27672         * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
27673         * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
27674         (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
27675         into...
27676         (ix86_vectorize_vec_perm_const): ...this new function.  Incorporate
27677         the old VEC_PERM_CONST conditions.
27678         * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
27679         * config/ia64/vect.md (vec_perm_const<mode>): Delete.
27680         * config/ia64/ia64.c (ia64_expand_vec_perm_const)
27681         (ia64_vectorize_vec_perm_const_ok): Merge into...
27682         (ia64_vectorize_vec_perm_const): ...this new function.
27683         * config/mips/loongson.md (vec_perm_const<mode>): Delete.
27684         * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
27685         * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
27686         * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
27687         * config/mips/mips.c (mips_expand_vec_perm_const)
27688         (mips_vectorize_vec_perm_const_ok): Merge into...
27689         (mips_vectorize_vec_perm_const): ...this new function.
27690         * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
27691         * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
27692         * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
27693         * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
27694         * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
27695         (rs6000_expand_vec_perm_const): Delete.
27696         * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
27697         Delete.
27698         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
27699         (altivec_expand_vec_perm_const_le): Take each operand individually.
27700         Operate on constant selectors rather than rtxes.
27701         (altivec_expand_vec_perm_const): Likewise.  Update call to
27702         altivec_expand_vec_perm_const_le.
27703         (rs6000_expand_vec_perm_const): Delete.
27704         (rs6000_vectorize_vec_perm_const_ok): Delete.
27705         (rs6000_vectorize_vec_perm_const): New function.
27706         (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
27707         an element count and rtx array.
27708         (rs6000_expand_extract_even): Update call accordingly.
27709         (rs6000_expand_interleave): Likewise.
27710         * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
27711         * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
27712         * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
27713         * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
27714         (rs6000_expand_vec_perm_const): Delete.
27715         * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
27716         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
27717         (altivec_expand_vec_perm_const_le): Take each operand individually.
27718         Operate on constant selectors rather than rtxes.
27719         (altivec_expand_vec_perm_const): Likewise.  Update call to
27720         altivec_expand_vec_perm_const_le.
27721         (rs6000_expand_vec_perm_const): Delete.
27722         (rs6000_vectorize_vec_perm_const_ok): Delete.
27723         (rs6000_vectorize_vec_perm_const): New function.  Remove stray
27724         reference to the SPE evmerge intructions.
27725         (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
27726         an element count and rtx array.
27727         (rs6000_expand_extract_even): Update call accordingly.
27728         (rs6000_expand_interleave): Likewise.
27729         * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
27730         * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
27731         new function.
27732         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
27734 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
27736         * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
27737         vector mode and that that mode matches the mode of the data
27738         being permuted.
27739         (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
27740         out into expand_vec_perm_var.  Do all CONST_VECTOR handling here,
27741         directly using expand_vec_perm_1 when forcing selectors into
27742         registers.
27743         (expand_vec_perm_var): New function, split out from expand_vec_perm.
27745 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
27747         * optabs-query.h (can_vec_perm_p): Delete.
27748         (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
27749         * optabs-query.c (can_vec_perm_p): Split into...
27750         (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
27751         (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
27752         particular selector is valid.
27753         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
27754         * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
27755         (vect_grouped_load_supported): Likewise.
27756         (vect_shift_permute_load_chain): Likewise.
27757         * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
27758         (vect_transform_slp_perm_load): Likewise.
27759         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
27760         (vectorizable_bswap): Likewise.
27761         (vect_gen_perm_mask_checked): Likewise.
27762         * fold-const.c (fold_ternary_loc): Likewise.  Don't take
27763         implementations of variable permutation vectors into account
27764         when deciding which selector to use.
27765         * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
27766         vec_perm_const_optab is supported; instead use can_vec_perm_const_p
27767         with a false third argument.
27768         * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
27769         to test whether the constant selector is valid and can_vec_perm_var_p
27770         to test whether a variable selector is valid.
27772 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
27774         * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
27775         * optabs-query.c (can_vec_perm_p): Likewise.
27776         * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
27777         instead of vec_perm_indices.
27778         * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
27779         (vect_gen_perm_mask_checked): Likewise,
27780         * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
27781         (vect_gen_perm_mask_checked): Likewise,
27783 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
27785         * optabs-query.h (qimode_for_vec_perm): Declare.
27786         * optabs-query.c (can_vec_perm_p): Split out qimode search to...
27787         (qimode_for_vec_perm): ...this new function.
27788         * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
27790 2018-01-02  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
27792         * rtlanal.c (canonicalize_condition): Return 0 if final rtx
27793         does not have a conditional at the top.
27795 2018-01-02  Richard Biener  <rguenther@suse.de>
27797         * ipa-inline.c (big_speedup_p): Fix expression.
27799 2018-01-02  Jan Hubicka  <hubicka@ucw.cz>
27801         PR target/81616
27802         * config/i386/x86-tune-costs.h: Increase cost of integer load costs
27803         for generic 4->6.
27805 2018-01-02  Jan Hubicka  <hubicka@ucw.cz>
27807         PR target/81616
27808         Generic tuning.
27809         * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
27810         cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
27811         and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
27812         cond_taken_branch_cost 3->4.
27814 2018-01-01  Jakub Jelinek  <jakub@redhat.com>
27816         PR tree-optimization/83581
27817         * tree-loop-distribution.c (pass_loop_distribution::execute): Return
27818         TODO_cleanup_cfg if any changes have been made.
27820         PR middle-end/83608
27821         * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
27822         convert_modes if target mode has the right side, but different mode
27823         class.
27825         PR middle-end/83609
27826         * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
27827         last argument when extracting from CONCAT.  If either from_real or
27828         from_imag is NULL, use expansion through memory.  If result is not
27829         a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
27830         the parts directly to inner mode, if even that fails, use expansion
27831         through memory.
27833         PR middle-end/83623
27834         * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
27835         check for bswap in mode rather than HImode and use that in expand_unop
27836         too.
27838 Copyright (C) 2018 Free Software Foundation, Inc.
27840 Copying and distribution of this file, with or without modification,
27841 are permitted in any medium without royalty provided the copyright
27842 notice and this notice are preserved.